Microgrids: Advances in Operation, Control, and Protection (Power Systems) 3030597490, 9783030597498
146 50 29MB
English Pages [642]
Preface
Contents
Part I Operation of Microgrids
1 An Introduction to Microgrids, Concepts, Definition, and Classifications
1.1 Introduction
1.2 Microgrid Components
1.3 Classification
1.3.1 Type
1.3.2 Size
1.3.3 Application
1.3.4 Operation Mode
1.3.5 Configuration
1.3.6 Characteristics/Properties of the Feeder
1.4 Control
1.4.1 Hierarchical Control
1.4.1.1 Primary Control
1.4.1.2 Secondary Control
1.4.1.3 Tertiary Control
1.4.2 Control Coordination
1.4.2.1 Centralized Control
1.4.2.2 Distributed Control
1.4.2.3 Hybrid Control
1.5 Stability
1.5.1 Grid-Connected MG Stability
1.5.2 Islanded MG Stability
1.6 Protection
1.7 Microgrid's Advantages and Challenges
1.7.1 Advantages
1.7.2 Challenges
1.7.2.1 Technical
1.7.2.2 Regulation
1.7.2.3 Economical
1.7.2.4 Marketing
References
2 Operation Management of Microgrid Clusters
2.1 Introduction
2.2 Benefits of NMCs
2.2.1 Best Utilization of DERs
2.2.2 Reduction of Overall Cost
2.2.3 Ancillary Services Improvement
2.2.4 Resiliency Improvement
2.2.5 Reliability Improvement
2.2.6 Bilateral and out of Market Transactions
2.3 Challenges of NMCs
2.3.1 Stability of the System
2.3.2 Protection Coordination
2.3.3 Privacy of MGs
2.3.4 Threat of Cyberattack
2.3.5 Disallowed Transactions
2.4 Main Objectives and Constraints of NMCs
2.5 Typical Architecture of NMCs
2.5.1 Serial MGs on a Single Distribution Feeder
2.5.2 Parallel MGs on a Single Distribution Feeder
2.5.3 Interconnected MGs on Multiple Distribution Feeders
2.6 Control Strategy of NMCs
2.6.1 Peer-to-Peer Control Strategy in NMCs
2.6.2 Master-Slave Control Strategy in NMCs
2.6.3 Hierarchical Control Strategy in NMCs
2.6.4 Distributed Control Strategy in NMCs
2.7 Energy Management and Operation of NMCs
2.7.1 Energy Management Strategies of NMCs
2.7.2 Compare EMS Structures
2.7.3 Overview of Energy Management Modeling and Solution Methods in NMCs
2.8 Objective Functions Formulation
2.8.1 Cost Operation Modeling
2.8.2 Pollution Emission Modeling
2.8.3 Problem Constraints
2.8.3.1 Power Balance Constraint
2.8.3.2 Generation Capacity Limit
2.8.3.3 Charge and Discharge Rate Limit
2.8.3.4 Self-Sufficiency Constraint
2.8.3.5 Reserve Constraint
2.8.4 Demand Response Pogromming (DRP)
2.8.5 Modeling of Generation Units in the MGs
2.8.5.1 Fuel Cell Model
2.8.5.2 Micro-Turbines Model
2.8.5.3 Combined Heat and Power Model
2.9 Numerical Result
2.10 Conclusion
References
3 Energy Management Systems for Microgrids
3.1 Introduction
3.2 An Overview of the EMS of the MGs
3.3 Monitoring System
3.4 Control System
3.5 Decision-Making System
3.5.1 Assessment Tools
3.5.1.1 Network and Component Modeler
3.5.1.2 Security Assessment Tool
3.5.1.3 Remedial Action Scheduling Tool
3.5.1.4 State Estimation
3.5.1.5 Load Forecast/Estimation
3.5.1.6 Load Flow
3.5.1.7 Short Circuit Calculation
3.5.1.8 Uncertainty Assessment
3.5.2 Optimization Tools
3.5.2.1 Unit Commitment
3.5.2.2 Economic Dispatch
3.5.2.3 Optimal Power Flow
3.5.2.4 Network Reconfiguration Tool
3.5.2.5 Decision Making under Uncertainty
3.5.3 Restoration Tools
3.6 Interaction with Other Systems
3.6.1 Distribution Management System (DMS)
3.6.2 Advanced Metering Infrastructure (AMI)
3.6.3 Outage Management System (OMS)
3.6.4 Maintenance Scheduling System
3.6.5 Weather Forecasting System
3.6.6 Electricity Market
3.6.7 Bid/Offer Interface
3.7 Centralized and Decentralized Energy Management
3.8 The Necessity of EMS in the Scheduling of MG
3.9 EMS Functions in the MG Scheduling
3.9.1 Microgrid's Hierarchical Scheduling
3.9.2 System Operation Strategies
3.9.2.1 Economic Aspects
3.9.2.2 Technical Aspects
3.10 Mathematical Modeling of Different MG's Components
3.10.1 Loads
3.10.2 Dispatchable Generation Resources
3.10.3 Renewable-Based Units with MPPT
3.10.4 Energy Storages
3.10.5 Reactive Power Resources
3.10.6 Combined Heat and Power (CHP) and Boiler
3.10.7 Electrical Network
3.10.7.1 Second-Order Cone Programming (SOCP)
3.10.7.2 Linear DistFlow
3.10.8 Energy Exchange with the Main Grid
3.11 Mathematical Modeling of System Security
3.11.1 Security Modeling in HAS
3.11.2 Security Modeling in RTS
3.12 Conclusion
References
4 Optimal Dispatch and Unit Commitment in Microgrids
4.1 Introduction
4.2 Microgrid
4.3 Demand Response Program
4.4 Distributed Generation Units
4.4.1 Nonrenewable Units
4.4.1.1 Micro Turbine
4.4.1.2 Fuel Cell
4.4.2 Renewable Units
4.4.2.1 Photovoltaic Panel
4.4.2.2 Wind Turbine
4.5 Energy Storage System
4.6 Objective Functions and Constraints
4.6.1 Objective Functions
4.6.1.1 The Profit of MGDC
4.6.1.2 Pollution Emission of MG
4.6.2 Constraints
4.6.2.1 Power Balance Constraint
4.6.2.2 Distributed Generation Constraint
4.6.2.3 Energy Storage System Constraint
4.6.3 Optimization Algorithm
4.6.3.1 Multi-Objective Grey Wolf Optimization Algorithm
4.6.3.2 Fuzzy Method
4.7 Numerical Results
4.7.1 Without DR Program
4.7.2 With DR Program
4.8 Conclusion
References
5 The Role of Energy Storage Systems in Microgrids Operation
5.1 Introduction
5.1.1 Background
5.1.2 Land-based Microgrids
5.1.2.1 Residential Microgrid
5.1.2.2 Industrial Microgrids
5.1.3 Mobile Microgrids
5.1.4 Comparisons between Different Types of Microgrids
5.1.4.1 Operation Modes
5.1.4.2 Load Demand Types
5.2 Energy Storage Technologies
5.2.1 Classification of Energy Storage Technologies
5.2.2 Single Energy Storage Technologies
5.2.2.1 Pumped Hydro Storage (PHS)
5.2.2.2 Compressed Air Energy Storage (CAES)
5.2.2.3 Battery Energy Storage (BES)
5.2.2.4 Flow Battery Energy Storage (FBES)
5.2.2.5 Flywheel Energy Storage (FES)
5.2.2.6 Superconducting Magnetics Energy Storage (SMES)
5.2.2.7 Supercapacitor Energy Storage (SES)
5.2.2.8 Fuel Cell
5.2.3 Hybrid Energy Storage Technologies
5.2.3.1 Battery Supercapacitor
5.2.3.2 Battery–Fuel Cell
5.2.3.3 Battery–Flywheel
5.3 Energy Storage Applications in Microgrids
5.3.1 Load Leveling
5.3.2 Power Quality
5.4 Conclusions
References
6 Microgrids and Local Markets
6.1 Introduction
6.2 Local Markets
6.2.1 Definition
6.2.2 Benefits
6.2.3 Objectives
6.2.4 Services
6.2.5 Value Streams for Microgrids
6.3 Key Elements in Local Markets Framework
6.4 Local Market Models
6.5 Trading Approaches in Local Markets
6.5.1 Pool-Based Trading
6.5.2 P2P Trading
6.5.3 Hybrid Trading
6.6 Market Settlement Approaches in Local Markets
6.6.1 Auction-Based Approach
6.6.2 Optimization-Based Approach
6.6.2.1 Distributed Clearing
6.6.2.2 Decentralized Clearing
6.7 Case Example of Local Markets for Microgrids: The Monash Microgrid
6.8 Case Examples of Market Settlement in Local Markets
6.9 Conclusions
References
7 An Economic Demand Management Strategy for Passive Consumers Considering Demand-Side Management Schemes and Microgrid Operation
7.1 Introduction
7.2 Types of DR Programs
7.2.1 PDR Strategies
7.2.2 IDR Strategies
7.2.3 DR Programs and Microgrid Operation
7.3 Classification of Microgrid Applications
7.3.1 Industrial Sector
7.3.2 Military Sector
7.3.3 Campus/Institutional Sector
7.3.4 Commercial Sector
7.3.5 Healthcare Sector
7.3.6 Residential Sector
7.3.7 Remote or Rural Microgrids
7.3.8 Other Microgrids
7.4 The Decision Procedure for Operating of a Microgrid Integrated with Demand Response
7.4.1 Microgrid Cost Modeling
7.4.1.1 Microgrid Installation Cost
7.4.1.2 Microgrid Maintenance Cost
7.4.1.3 Microgrid Operation Cost
7.4.1.4 Microgrid Start-Up Cost
7.4.2 DR Cost Modeling
7.4.2.1 PDR-Based Cost
7.4.2.2 IDR-Based Cost
7.4.3 The Decision Algorithm
7.5 Numerical Studies
7.5.1 Case I: Industrial Load
7.5.2 Case II: Commercial Load
7.5.3 Case III: Hospital Load
7.5.4 Final Deduction
7.6 Conclusions
References
8 Real-Time Perspective in Distributed Robust Operation of Networked Microgrids
8.1 Introduction
8.2 Methodology Description
8.3 Problem Formulation
8.3.1 Microgrids' Real-Time Energy Management
8.3.2 Distributed Energy Management
8.3.3 Numerical Studies and Result Analysis
8.4 Conclusion
Nomenclature
Sets
Indexes
Parameters
Variables
References
9 Application of Heuristic Techniques and Evolutionary Algorithms in Microgrids Optimization Problems
9.1 Introduction
9.2 Brief Introduction to Evolutionary Algorithms
9.2.1 Genetic Algorithm (GA)
9.2.2 Particle Swarm Optimization Algorithm (PSO)
9.2.3 Ant Colony Optimization (ACO)
9.2.4 Biogeography Based Optimization (BBO)
9.2.5 Harmony Search Algorithm (HSA)
9.2.6 Cuckoo Search Algorithm (CSA)
9.2.7 Artificial Bee Colonies (ABC)
9.2.8 Grey Wolf Optimization (GWO)
9.2.9 Firefly Algorithm (FA)
9.2.10 Grasshopper Optimization Algorithm (GOA)
9.2.11 Whale Optimization Algorithm (WOA)
9.3 Illustrative Examples on Application of EAs in Microgrids
9.3.1 Energy Management and Operation Scheduling
9.3.2 Optimal Placement and Sizing of Energy-Related Devices
9.3.3 Microgrid Optimal Voltage and Frequency Control
9.4 Conclusion
References
Part II Control of Microgrids
10 Conventional Droop Methods for Microgrids
10.1 Introduction
10.2 Conventional Droop Method for AC Microgrid
10.2.1 Mathematical Analysis
10.2.2 Synchronous Generator
10.2.3 Renewable Energy Sources
10.2.3.1 IBR
10.2.3.2 Non-IBR
10.2.4 Energy Storage System
10.2.5 Frequency and Voltage Responses
10.3 Conventional Droop Control Method in DC Microgrid
10.3.1 Mathematical Analysis
10.3.2 Converter-Based Generator
10.4 Conclusion
References
11 Distributed Control Approaches for Microgrids
11.1 Introduction
11.2 Hierarchical Control Structure of AC and DC MGs
11.2.1 DC MGs
11.2.2 AC MGs
11.3 Distributed Control of DC MGs
11.3.1 DG Model in a DC MG
11.3.2 Distributed Secondary Control of DC MGs
11.4 Distributed Control of AC MGs
11.4.1 Frequency Control
11.4.2 Voltage Control
11.5 Conclusion and Future Trend
References
12 On Control of Energy Storage Systems in Microgrids
12.1 Introduction
12.2 Overview of Energy Storage Systems
12.2.1 Characteristics of ESSs
12.2.2 Power Electronic Interface
12.2.3 Battery Management System
12.3 ESS Control Strategies in Islanded Microgrids
12.3.1 Coordinated Control of Multiple ESSs
12.3.2 Coordinated Control of RESs and ESSs
12.4 ESS Control Strategies in Grid-Connected Microgrids
12.4.1 Voltage Regulation
12.4.2 Frequency Regulation
12.5 Research Trends and Opportunity
12.6 Conclusion
References
13 Microgrid Stability Definition, Analysis, and Examples
13.1 Introduction
13.2 Microgrid Modeling
13.2.1 The Fixed-Speed Wind Turbine Model
13.2.2 CHP Model
13.2.3 Synchronous Reference Frame
13.2.4 Reduced Network Model
13.3 AFPID Controller
13.4 Global Design of the Optimization
13.4.1 Cost Function
13.4.2 System Optimization Variables
13.4.2.1 Stage I
13.4.2.2 Stage II
13.4.2.3 Stage III
13.4.3 NSIDE Algorithm
13.4.3.1 Improved Differential Evolution Algorithms
13.4.3.2 Nondominated Sorting-Based Multi-Objective Algorithm
13.5 Simulation Results
13.5.1 Scenario 1: Symmetrical Three-Phase Fault
13.5.2 Scenario 2: Load Shedding in Bus 8
13.5.3 Scenario 3: 30% Load Decrease
13.5.4 Scenario 4: 30% Load Increase
13.6 Conclusion
References
14 Voltage Unbalance Compensation in AC Microgrids
14.1 Introduction
14.2 Inverter-Interfaced DG
14.2.1 VSC Control
14.3 Harmonic Compensation
14.3.1 Identification of Reference Currents
14.3.1.1 Instantaneous Active and Reactive Power Method
14.3.1.2 Synchronous Reference Frame (SRF) Method
14.3.2 Performance of Pq and SRF Methods in Ideal and Nonideal Conditions
14.3.2.1 Case of the pq Method
14.3.2.2 Case of the SRF Method
14.4 Imbalance Compensation
14.4.1 Direct Extraction Methods
14.4.1.1 In Abc Frame
14.4.1.2 In αβ Frame
14.4.1.3 In dq Frame
14.4.2 Indirect Extraction Methods
14.4.2.1 Double Synchronous Reference Frame (DSRF)
14.4.2.2 DSOGI-FLL
14.4.2.3 DSOGI-PLL
14.4.3 Instantaneous Power under Unbalanced Conditions
14.5 Case Study
References
15 WAM-Based Hierarchical Control of Islanded AC Microgrids
15.1 Introduction
15.2 Design of WAMS-Based Hierarchical Controller Considering Signal Transmission Time Delays
15.3 Results and Discussions
15.3.1 Eigenvalue Analysis
15.3.2 Time-Domain Simulations
15.4 Conclusion
Appendix
References
Part III Protection of Microgrids
16 Fault Ride Through and Fault Current Management forMicrogrids
16.1 Introduction
16.2 MG FRT in Grid Codes
16.3 FRT Control in Microgrids
16.3.1 AC Microgrids
16.3.2 DC Microgrids
16.3.3 Traditional Reference Current Control Methods
16.3.3.1 Balanced Positive Sequence Control Method
16.3.3.2 Decoupled Double Synchronous Reference Control Method
16.3.4 Natural Phase-Coordinates Approach
16.3.4.1 Symmetric NPC Method
16.3.4.2 Generalized NPC Method
16.4 Fault Current Management for Microgrids
16.4.1 Single MG Fault Current Management
16.4.2 Multiple Microgrids Fault Current Management
16.5 Hardware-in-the-Loop Platform for MG FRT
16.6 Analysis and Results
16.7 Summary
Appendix
16.7.1 FRT Reactive Power Injection
16.7.2 Synchronous Reference Frame PLL
16.7.3 Deadbeat Current Control
References
17 Microgrid Protection
17.1 Introduction
17.2 Microgrids
17.2.1 MG Protection
17.2.2 Problems and Functional Solutions for Relays
17.2.3 Discussion
17.3 Multi-Agent System Proposals
17.3.1 Definition of a Smart Agent
17.3.2 Multi-Agents in the Electricity Sector
17.3.3 Gang of Agents
17.3.3.1 Agents in JADE
17.4 Implementation in JADE Architecture
17.4.1 Communication Language Used by Agents
17.4.2 Structure of the JADE Work Environment
17.4.3 Testing and Measurement Tool
17.4.4 Communication Between Matlab/Simulink and JADE
17.5 Test Scenarios
17.5.1 Detection of Weak Infeed Conditions
17.5.1.1 Small Conventional Generators
17.5.1.2 DG Generators
17.5.2 Detection of Islanding Condition
17.5.2.1 Detection Schemes
17.5.2.2 Implementation
17.5.3 Coordination of DOCRs
17.5.3.1 Optimization Problem
17.5.3.2 Optimization Algorithms
17.5.3.3 On-Line Implementation
17.5.3.4 Test Systems
17.5.3.5 Results and Discussion
17.6 Conclusions
References
18 A New Second Central Moment-Based Algorithm for Differential Protection in Micro-Grids
18.1 Introduction
18.2 Differential Protection Based on SCM
18.2.1 Differential Protection Principle
18.2.1.1 Percentage Differential Protection
18.2.1.2 Negative-Sequence Differential Principle
18.2.1.3 Current Transformer Saturation
18.2.1.4 Methods Summary
18.3 Second Central Moment Applied to Differentiate Inrush Currents from Internal Faults
18.3.1 Computational Complexity
18.4 Results
18.4.1 Power Transformer Protection
18.4.2 Generator Protection
18.4.3 Reactor Differential Protection
18.4.4 Busbar Differential Protection
18.4.5 Power Transformer Energization with Harmonic Content
18.4.6 Power Reactor Energization with High Harmonic Distortion
References
19 Microgrid Protection with Conventional and Adaptive Protection Schemes
19.1 Introduction
19.2 Microgrid Protection Issues
19.2.1 DER Unit Fault Behavior and Effect on Microgrid Protection
19.2.2 Example—Microgrid Transition to Islanded Operation
19.3 Protection Requirements
19.3.1 Sensitivity
19.3.2 Selectivity
19.3.3 Reliability
19.3.4 Adaptivity
19.3.5 Re-Synchronization
19.3.5.1 LV Microgrid Synchronized Reconnection
19.3.6 Circuit Breaker Technology
19.4 Conventional Protections for AC Microgrids
19.4.1 Overcurrent Protection (Definite-Time Versus Inverse Time)
19.4.1.1 Protection Coordination in a Grid-Connected Mode without DERs
19.4.1.2 Protection Coordination in Grid-Connected Mode with DERs
19.4.1.3 Protection Coordination in Islanded Mode with DERs and BESS
19.5 Adaptive Protection for AC Microgrids
19.5.1 Communication-Based Adaptive Protection
19.5.1.1 Centralized Adaptive Protection
19.5.1.2 Decentralized Adaptive Protection
19.5.2 Review of Communication-Based Protection Schemes for AC Microgrids
19.5.2.1 Directional Overcurrent Protection
19.5.2.2 Current Symmetrical Components Based Protection
19.5.2.3 Distance Protection
19.5.2.4 Voltage Based Protection Schemes
19.5.2.5 Current Differential
19.5.2.6 Protection Based on Voltage and Directional Overcurrent
19.5.3 Protection of LV AC Microgrids
19.5.3.1 Proposed LV Microgrid Protection Scheme 1
19.5.3.2 Proposed LV Microgrid Protection Scheme 2
19.5.4 Protection of MV AC Microgrids
19.5.4.1 Proposed MV Microgrid Protection Scheme 1
19.5.4.2 Enhanced MV Microgrid Protection Scheme 1—HIF Detection Included
19.5.5 Communication-Less Adaptive Protection
19.5.6 Islanding Detection during Island Operation of Nested Microgrid
19.5.7 Need for Microgrid Grid Codes
19.6 Protection of DC Microgrids
References
20 Fault Identification, Protection Schemes, and Restoration Requirements of Microgrids
20.1 Introduction
20.2 Fault Locating and Restoration Requirements in MG
20.2.1 Fault Indication and Restoration Devices
20.2.2 FI Placement
20.2.3 Simultaneous Placement of FIs and RCSs
20.3 The Need for Selective and Fast Protection Schemes in MGs
20.3.1 Importance of Selective Protection
20.3.1.1 Selective Dead Time in Reclosing Scheme of MGs
20.3.1.2 Nuisance Tripping of Protective Devices in MG
20.3.1.3 Blinding of Protective Devices along with Inverter-Based DERs
20.3.2 Essence of Fast-Response Protection in Restoration
20.3.3 Selective and Fast Solutions
20.3.3.1 Multi Agent-Based Protection
20.3.3.2 Overcurrent Protection Multi-Inverse Characteristic
20.3.3.3 Overcurrent Protection with Multi-Function Characteristic
20.3.3.4 Time-Current-Voltage Characteristics
20.3.3.5 Dual Time–Current–Voltage Characteristics
20.3.3.6 Piece-Wise Linear Characteristic
20.4 Reconfigurable Topology of MG and Enhancing the Restoration Capability
20.5 Conclusion
References
21 Real-Time Testing of Microgrids
21.1 Introduction
21.2 Real-Time Testing Methods
21.3 Digital Real-Time Testing: Concept
21.4 Hardware in the Loop Testing
21.5 Real-Time Emulation
21.6 Novel/Hybrid Testing Approaches
21.7 Conclusions
References
Index
Recommend Papers
![Control and Dynamics in Power Systems and Microgrids (Solutions, Instructor Solution Manual) [1 ed.]
1138034991, 9781138034990](https://ebin.pub/img/200x200/control-and-dynamics-in-power-systems-and-microgrids-solutions-instructor-solution-manual-1nbsped-1138034991-9781138034990.jpg)
- Author / Uploaded
- Amjad Anvari-Moghaddam
- Hamdi Abdi
- Behnam Mohammadi-Ivatloo
- Nikos Hatziargyriou
File loading please wait...
Citation preview
Power Systems
Amjad Anvari-Moghaddam Hamdi Abdi Behnam Mohammadi-Ivatloo Nikos Hatziargyriou Editors
Microgrids Advances in Operation, Control, and Protection
Power Systems
Electrical power has been the technological foundation of industrial societies for many years. Although the systems designed to provide and apply electrical energy have reached a high degree of maturity, unforeseen problems are constantly encountered, necessitating the design of more efficient and reliable systems based on novel technologies. The book series Power Systems is aimed at providing detailed, accurate and sound technical information about these new developments in electrical power engineering. It includes topics on power generation, storage and transmission as well as electrical machines. The monographs and advanced textbooks in this series address researchers, lecturers, industrial engineers and senior students in electrical engineering. **Power Systems is indexed in Scopus**
More information about this series at http://www.springer.com/series/4622
Amjad Anvari-Moghaddam • Hamdi Abdi Behnam Mohammadi-Ivatloo • Nikos Hatziargyriou Editors
Microgrids Advances in Operation, Control, and Protection
Editors Amjad Anvari-Moghaddam Department of Energy Technology Aalborg University Aalborg, Denmark
Hamdi Abdi Electrical Engineering Department Razi University Kermanshah, Iran
Behnam Mohammadi-Ivatloo Faculty of Electrical and Computer Engineering University of Tabriz Tabriz, Iran
Nikos Hatziargyriou Electrical and Computer Engineering Department National Technical University of Athens Zografou, Attika, Greece
ISSN 1612-1287 ISSN 1860-4676 (electronic) Power Systems ISBN 978-3-030-59749-8 ISBN 978-3-030-59750-4 (eBook) https://doi.org/10.1007/978-3-030-59750-4 © Springer Nature Switzerland AG 2021 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Switzerland AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
Increased penetration of distributed generations (DGs) in the power system has crucially changed the control and operation relevant concepts. In this way, the microgrids are being expanded and developed as a fundamental and essential building block for future smart power systems. The microgrid concept-introduced by the Consortium for Electric Reliability Technology Solutions (CERTS) to improve the reliability, sustainability, and efficiency of the modern power systemis an aggregation of DG units, distributed energy storage (DES), sensitive and nonsensitive loads, and centralized/decentralized control system, operating as a controllable subsystem which can operate in grid-connected as well as in an islanded mode. These new systems requested a revisited definition of the most well-known issues in the power system control aspects. AC, DC, and AC/DC networks, or hybrid microgrids are the most dominant nature in these new systems. Therefore, key issues for the operation, control, and protection of these systems include integration technologies, hierarchical control techniques, and optimization methods that should be carefully updated, focusing on primary, secondary, and tertiary control layers in both islanded and grid-connected modes. The importance of revising the relevant operation studies in the presence of microgrids is very crucial and vital. The impacts of different uncertainties arising from the increased penetration of renewables sources, power market pricing policies, electric vehicles, storage system devices, and demand-side management are the most important features. This book covers a comprehensive study on the control, operation, and protection of microgrids with related strategies to analyze and understand the salient features of modern control and optimization techniques applied to these systems. It also discusses emerging concepts, key drivers and new players in microgrids, and local energy markets while addressing various aspects from day-ahead scheduling to realtime testing of microgrids.
v
vi
Preface
This book is divided into three parts, including operation (Chaps. 1–9), control (Chaps. 10–15), and protection (Chaps. 16–21) of microgrids. Part I covers the studies related to microgrid operation. The first chapter provides an overview of the definitions and clustering of microgrids from different perspectives. Chapter 2 studies the advantages, challenges, objectives, architecture, control strategies, and operation of networked microgrid and those in clusters. This chapter also proposes a model for daily energy management and scheduling of networked microgrids considering different generation resources and loads. Chapter 3 discusses energy management systems (EMSs) for microgrids in normal and contingency conditions and appropriate objectives are defined for each condition. This chapter also provides an overview of the different aspects that should be considered in EMS design and implementation for microgrids like technical and security issues, economic objectives, power flow management, reconfiguration, etc. Optimal dispatch and unit commitment in microgrids are studied in Chap. 4, considering economic and environmental aspects at the same time, which results in a multi-objective optimization problem. The fuzzy decision-making method is utilized to find the best compromise operational schedule of energy units. Chapter 5 focuses on energy storage systems in different microgrids, such as land-based microgrids and mobile microgrids. The application of energy storage for load leveling and power quality improvement in microgrids are also studied in this chapter. The authors of Chap. 6 study the formation of local markets in microgrids and provide an overview of definitions in this area, potential benefits, and objectives. A summary of key enabler elements for local market implementation is given, and different trading approaches, as well as, market settlement approaches for local markets are presented with detailed case examples to help readers with outlining attributes of different market models. Chapter 7 covers a summary of demand response (DR) programs applicable to microgrids in different sectors such as residential and commercial and explores the impact of customers’ participation level in both price-based and incentive-based DR programs from the economic point of view. Chapter 8 elaborates on a framework for operation management of networked microgrids that have different owners. A combination of the alternating direction method of multipliers and robust optimization methods is introduced and implemented to solve effectively such a problem at the operating layer. Chapter 9 focuses on recent progress in the application of computational intelligence and heuristic techniques in microgrids and provides an overview of the application of evolutionary algorithms to the energy management problem of microgrids. Part II covers the topics related to the control of microgrids ranging from conventional droop methods at local levels to wide-area measurement system (WAMS)-based hierarchical control techniques. Chapter 10 focuses on droop control concept in microgrids and discusses the application of droop-based control methods in both AC and DC microgrids, considering different characteristics and features of inverter-based renewable energy sources, dispersed generation units, and energy storage systems. Chapter 11 addresses the hierarchical control structure of microgrids, where the primary, secondary, and tertiary control levels are discussed in detail. The chapter also gives a focus on distributed control
Preface
vii
of both AC and DC microgrids and covers the distributed control techniques utilized for voltage/frequency control as well as active/reactive power-sharing. Chapter 12 introduces the control and application of energy storage systems (ESSs) in microgrid systems. The characteristics of energy storage techniques, power electronic interfaces, and battery management systems are also discussed. Finally, a comprehensive review of ESSs in both islanded and grid-connected microgrids is conducted and future research roadmaps of multifunctional ESSs together with cyber-security issues of ESSs are outlined. Chapter 13 elaborates on microgrid’s stability definitions and analysis, followed by examples on voltage and frequency stability improvement in islanded microgrids. The chapter also introduces a novel multi-machine structure-based simulation model for the study of the dynamic behavior of microgrids. Chapter 14 studies the control methods used to compensate harmonics and voltage unbalance disturbances in microgrids as the main problems of power quality in the steady state. Finally, Chap. 15 in Part II of the book presents a WAMS-based hierarchical control for islanded microgrids where a stable operation is an important concern due to the low-inertial nature of power electronics interfaced units. Part III covers recent advances in the protection of microgrids. This part of the book elaborates on conventional and emerging schemes for fault identification, protection, and restoration in microgrids, followed by various real-time testing methods and the categories of simulation suitable for control and protection of microgrids. In this regard, Chap. 16 is devoted to Fault Ride Through (FRT), as the ability of distributed energy resources to stay connected during the faulty conditions, and Fault Current Management (FCM) in microgrids. This chapter also provides the microgrid FRT in different grid codes and proposes new methods for FRT control and FCM. The proposed model is validated using real-time simulation based on hardware in the loop tests. Chapter 17 provides an overview of the challenges in microgrid protection and employed relays. The application of multiagent systems (MAS) to microgrid protection is investigated in this chapter, where different studies related to relay coordination and islanding detection are carried out. Chapter 18 provides an overview of requirements for the protection system from the point of view of differential protection used in different elements of microgrids and distribution systems such as bus bars, transformers, generators, and reactors. As a complementary example to this chapter, the second central moment (SCM) scheme is presented as a fast and low computation method for fault detection in different zones. Chapter 19 discusses the conventional and adaptive protection schemes for microgrids during grid-connected and islanded operation conditions. The chapter also addresses the issues related to protection schemes in systems with a high number of distributed energy sources, gives an overview of the existing and new requirements of protection schemes, and analyzes their potentials in microgrids. Chapter 20 mostly focuses on fault identification and restoration of microgrids after fault clearance. The challenges related to direction detection, blinding issues, and possible mal-operations are studied as well in this chapter. Finally, Chap. 21 studies the real-time testing of microgrids. This chapter explores the basics and makes a comparison of the various testing methods ranging from
viii
Preface
off-line to real-time simulations, rapid controller prototyping, hardware in the loop (HIL): both the signal level (CHIL) and power level (PHIL), real-time power level emulation, test-bed platforms, hybrid approaches/combinations of these techniques and novel solutions such as digital twin, blockchain and internet of things (IoT) based approaches. Aalborg, Denmark Kermanshah, Iran Tabriz, Iran Zografou, Attika, Greece
Amjad Anvari-Moghaddam Hamdi Abdi Behnam Mohammadi-Ivatloo Nikos Hatziargyriou
Contents
Part I Operation of Microgrids 1
An Introduction to Microgrids, Concepts, Definition, and Classifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Maryam Shahbazitabar, Hamdi Abdi, Hossein Nourianfar, Amjad Anvari-Moghaddam, Behnam Mohammadi-Ivatloo, and Nikos Hatziargyriou
3
2
Operation Management of Microgrid Clusters . . . . . . . . . . . . . . . . . . . . . . . . . Meisam Moradi and Asghar Akbari Foroud
17
3
Energy Management Systems for Microgrids . . . . . . . . . . . . . . . . . . . . . . . . . . . Seyed Mohsen Hashemi and Vahid Vahidinasab
61
4
Optimal Dispatch and Unit Commitment in Microgrids . . . . . . . . . . . . . . . Hossein Shayeghi and Masoud Alilou
97
5
The Role of Energy Storage Systems in Microgrids Operation . . . . . . . 127 Sidun Fang and Yu Wang
6
Microgrids and Local Markets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151 Mohsen Khorasany and Reza Razzaghi
7
An Economic Demand Management Strategy for Passive Consumers Considering Demand-Side Management Schemes and Microgrid Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179 Mohammad Esmaeil Honarmand, Vahid Hosseinnezhad, Barry Hayes, Behnam Mohammadi-Ivatloo, and Pierluigi Siano
8
Real-Time Perspective in Distributed Robust Operation of Networked Microgrids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205 Mehdi Jalali, Manijeh Alipour, and Kazem Zare
ix
x
Contents
9
Application of Heuristic Techniques and Evolutionary Algorithms in Microgrids Optimization Problems . . . . . . . . . . . . . . . . . . . . . 219 Amir Aminzadeh Ghavifekr
Part II Control of Microgrids 10
Conventional Droop Methods for Microgrids . . . . . . . . . . . . . . . . . . . . . . . . . . . 255 Kwang Woo Joung and Jung-Wook Park
11
Distributed Control Approaches for Microgrids . . . . . . . . . . . . . . . . . . . . . . . . 275 Tohid Khalili and Ali Bidram
12
On Control of Energy Storage Systems in Microgrids . . . . . . . . . . . . . . . . . 289 Yu Wang, Sidun Fang, and Yan Xu
13
Microgrid Stability Definition, Analysis, and Examples . . . . . . . . . . . . . . . 305 Hossein Shayeghi, Hamzeh Aryanpour, Masoud Alilou, and Aref Jalili
14
Voltage Unbalance Compensation in AC Microgrids . . . . . . . . . . . . . . . . . . . 337 Shahram Karimi, Mehdi Norianfar, and Josep M. Guerrero
15
WAM-Based Hierarchical Control of Islanded AC Microgrids . . . . . . . 375 E. S. N. Raju P and Trapti Jain
Part III Protection of Microgrids 16
Fault Ride Through and Fault Current Management for Microgrids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403 Wei Kou and Sung-Yeul Park
17
Microgrid Protection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 437 Arturo Conde Enríquez, Yendry González Cardoso, and José Treviño Martínez
18
A New Second Central Moment-Based Algorithm for Differential Protection in Micro-Grids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 489 Ernesto Vázquez, Héctor Esponda, and Manuel A. Andrade
19
Microgrid Protection with Conventional and Adaptive Protection Schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 523 Aushiq Ali Memon, Hannu Laaksonen, and Kimmo Kauhaniemi
20
Fault Identification, Protection Schemes, and Restoration Requirements of Microgrids. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 581 Amin Yazdaninejadi, Amir Hamidi, Sajjad Golshannavaz, and Daryoush Nazarpour
Contents
21
xi
Real-Time Testing of Microgrids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 615 A. S. Vijay and Suryanarayana Doolla
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 631
Part I
Operation of Microgrids
Chapter 1
An Introduction to Microgrids, Concepts, Definition, and Classifications Maryam Shahbazitabar, Hamdi Abdi, Hossein Nourianfar, Amjad Anvari-Moghaddam , Behnam Mohammadi-Ivatloo, and Nikos Hatziargyriou
1.1 Introduction Traditional electric power systems are rapidly transforming by increased renewable energy sources (RESs) penetration resulting in more efficient and clean energy production while requiring advanced control and management functions. Microgrids (MGs) are significant parts of this transformation at the distribution level. As a fact, since the year 2004, in which the MG was defined as “a better way to realize the emerging potential of distributed generation in a systematic approach which views generation and associated loads as a subsystem” [1], significant improvements and innovations have been made. The MG concept was firstly introduced by the USA’s Consortium for Electric Reliability Technology Solutions (CERTS) to reduce the cost, and increase the power quality, effectively all around the world [2]. Among various definitions, the U.S. Department of Energy (DOE) Microgrid Exchange Group (MEG) has used the following [3]: “A microgrid is a group of interconnected loads and distributed energy resources within clearly defined electrical boundaries that acts as a single controllable entity
M. Shahbazitabar () · H. Abdi · H. Nourianfar Department of Electrical Engineering, Faculty of Engineering, Razi University, Kermanshah, Iran e-mail: [email protected] A. Anvari-Moghaddam Department of Energy Technology, Aalborg University, Aalborg, Denmark B. Mohammadi-Ivatloo Faculty of Electrical and Computer Engineering, University of Tabriz, Tabriz, Iran N. Hatziargyriou Electrical and Computer Engineering Department, National Technical University of Athens Zografou, Attika, Greece © Springer Nature Switzerland AG 2021 A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems, https://doi.org/10.1007/978-3-030-59750-4_1
3
4
M. Shahbazitabar et al.
with respect to the grid. A microgrid can connect and disconnect from the grid to enable both grid-connected and island-modes of operation.” In a widely accepted definition “Microgrids are electricity distribution systems containing loads and distributed energy resources, (such as distributed generators, storage devices, or controllable loads) that can be operated in a controlled, coordinated way, either while connected to the main power network and/or while islanded” [4]. The MG is a flexible and dispatchable system that is capable of operating in both modes of grid-connected or stand-alone. It can potentially reduce the dependency of its consumers on traditional generation systems by providing different types of energy, such as electrical and thermal energy, and provide ancillary services trading activity between the MG and the main grid. The MG configuration can be AC, DC, or hybrid. This chapter, as an introduction to the MG concept, tries to present some practical and useful information for MG integration. Definitions, classifications, components, control methods, and protection schemes of MGs are also addressed briefly along with their merits or demerits.
1.2 Microgrid Components Global warming and growing energy demand are the most significant drivers spurring renewable energy sources (RESs) to reduce greenhouse gas (GHG) emissions by fossil fuel-based electricity generation. Distributed energy resources (DERs) such as solar photovoltaic (PV) modules, wind turbines (WTs), combined heat and power (CHP) units, and controllable loads such as electric vehicles (EVs) are expected to play a considerable role in future electricity supply because of their significant benefits such as carbon emissions reduction, energy efficiency enhancement, power quality and reliability (PQR) improvement, and line losses reduction and deferral of grid expansion plans [5]. The intermittent nature of renewable-based DERs is the main challenge for their integration into traditional power systems. The presence of MGs helps the increase of DGs penetration, more specifically at low voltages (LV) distribution networks. DERs integration, along with energy storage systems (ESSs) and controllable loads near power consumers within MGs, provides economic and environmental benefits [6]. MGs as parts of the distribution systems are connected to the upstream network at a single point of common coupling (PCC) often by power electronic-based switchgear. The operation of MGs in islanded mode is particularly challenging. In such an operation, the ESSs activity is particularly essential to improve power quality, stability, and reliability of supply, at least for critical loads. The capability of MGs to switch into the islanded mode in case of faults in the upstream network increases the reliability of customer supply and the resilience of
1 An Introduction to Microgrids, Concepts, Definition, and Classifications Wind turbine Solar PV
5
Loads
MG control center
Battery ESS
Micro-Grid Main Grid
Conventional DG
Controllable Loads
Fig. 1.1 A simple MG structure
the local distribution networks. This is particularly important, in case of natural or human-made disasters, when MGs can isolate and continue to provide electricity to critical loads. In fact, resilience is nowadays one of the key drivers for the development of MGs in developed countries, such as the United States, Japan, etc. Perhaps, the most common application of MGs is found in rural electrification. In developing countries, MGs can be used for the electricity supply of remote communities or to support some facilities, such as healthcare, water use, food preservation, waste treatment, telecommunication support, etc. Figure 1.1 shows a simple MG comprising DGs, ESS, and flexible loads connected via power converters. Extensive research is currently underway in MG development and demonstration to solve several technical and economic challenges such as accurate and integrated management of all energy sources, increasing the penetration of hybrid AC/DC power networks in the various areas of planning, operation, protection, and control [7].
1.3 Classification According to Navigant Research, MG can be classified into different groups based on various aspects [8, 9]:
1.3.1 Type MGs could be categorized in different types scuh as campus, military, residential, commercial, and industrial. The campus MG includes onsite generation, while it is managed by a single owner. Military MGs are used for improved efficiency and resilience.
6
M. Shahbazitabar et al.
1.3.2 Size MGs can exist in different sizes: small and simple (about hundreds of kW) at low or medium voltages [4] supplying just a few customers to large and complex ones (few MWs) [10].
1.3.3 Application MGs can be used to provide premium power or resilience-oriented services. The premium power is the power that provides a stable level of voltage noise free to its end consumers [11]. Also, the resilience-oriented MG is referred to as an MG with the ability to withstand and recover from “high impact–low-frequency” events [12]. The accidents, such as deliberate attacks, or naturally occurring incidents, are considered in this regard, while the negative impacts during both long-term and short-term horizons should be minimized [13].
1.3.4 Operation Mode MGs can operate in two modes: grid-connected and islanded. In grid-connected mode, the MG can exchange power with the upstream grid, depending on the electricity generated and its load demand [14]. The MG can be disconnected from the utility grid due to faults or in planned maintenance and operate autonomously [15]. Unlike grid-connected mode, an islanded MG may face challenges in regulating voltage and frequency or maintain the required quality of the power.
1.3.5 Configuration The topologies in which components of an MG, namely loads, micro-sources, and storage devices, are integrated lead to different configurations: AC network MGs, DC network MGs, and hybrid AC–DC MGs. Emerging DC sources and loads have given rise to the application of DC–MGs in recent years. Distribution in AC–MG can be one of the following three types: single phase, three phase with neutral, and three phase without neutral, while in DC–MG, it can be monopolar, bipolar, and homopolar [16]. In AC-one, DC-based DERs, ESSs, and loads are connected to a universal AC bus via DC-to-AC inverters. On the other hand, DC–MGs could offer various merits compared to AC–MGs: more efficient supply of DC loads, loss reduction via decreasing the multiple converters used for DC loads, facilitate various DC-DERs integration such as fuel cells (FC) and photovoltaic systems (PV) to the
1 An Introduction to Microgrids, Concepts, Definition, and Classifications
7
Urban
Rural
Off-grid
Feeders are loaded in a congested industrial area with low imbalance degree becouse of short distance between the main body and the laterals. The voltage and frequency are controled by the macrogrid so the voltage drop is low.
Feeders are placed in a rather populated area, the distance between the main body and the laterals is fairly long becouse of scattered loads; the voltage profile is not flat. DERs have effects on voltage fluctuation and must be controlled to facilitate in the feeder voltage regulation.
By definition, an off-grid microgrid always operates in islanded mode, remote area assighned with no possibility for macrogrid, where large-size DER integration is occurring faster.
Fig. 1.2 Feeder-based classification
common node with simplified interfaces, and decreasing the need for synchronizing generators and the buses versus several challenges in control, and operation [8, 17, 18].
1.3.6 Characteristics/Properties of the Feeder MGs can be categorized into three groups via feeder properties, as shown in Fig. 1.2.
1.4 Control Control of MGs is the one significant feature that distinguishes them from simple distribution lines with DER. This is further discussed in this section.
1.4.1 Hierarchical Control Hierarchical, multilevel control is adopted for the effective control of MGs, including the following three levels [19, 20].
1.4.1.1
Primary Control
Its task is the control of voltage, current, and local power. The set points of inverters in this level are changed based on their droops.
8
M. Shahbazitabar et al.
1.4.1.2
Secondary Control
It deals with optimal load sharing, frequency restoration, voltage regulation at pilot points, etc., and this is the level where the MGCC determines the set points needed to be followed by local controllers at the primary level. Some important subjects, such as forecasting functions and economic dispatch, could be also implemented at this level.
1.4.1.3
Tertiary Control
This level deals with upstream networks, like MG synchronization, and electricity market trading. The time scales of MG control functions can be divided into different levels. For example, primary level control actions such as voltage and current control should be executed in a couple of seconds to meet the system’s security constraints. Secondary and tertiary control functions require a couple of minutes, and hours, respectively. Three-level control functions for the MG are illustrated in Fig. 1.3, with time scaling definition.
Terti ary
Electricity Market Trading MG Synchronization
Demand Response
Secondary
Economic Load Sharing Frequency Restoration
Forecasting Functions
Economic Dispatch
Prim ary
Power quality Changing the Set points of Inverters Frequency and Voltage Control
s
ms
Fig. 1.3 Different control level for the MG
s
min
h
Day
1 An Introduction to Microgrids, Concepts, Definition, and Classifications Central Controller Communication Device
9
Central Controller
Central Controller
Communication Device
Communication Device
Comm. Device
Comm. Device
Comm. Device
Comm. Device
Comm. Device
Comm. Device
Comm. Device
Comm. Device
Comm. Device
Comm. Device
Comm. Device
Comm. Device
LC
LC
LC
LC
LC
LC
LC
LC
LC
LC
LC
LC
DG
DG
DG
DG
DG
DG
DG
DG
DG
DG
DG
DG
Point of common coupling
Point of common coupling
a) Centralized MG
b) Distributed MG
Point of common coupling
c) Hybrid MG
Fig. 1.4 Different types of control coordination structures from the communication perspective [23]
1.4.2 Control Coordination Control coordination of MG can be divided into a centralized, distributed, or a combination of both, named as a hybrid [21]. The type of coordinated control structure designed for the MG depends on DER ownership, MG size, available technologies, and communication infrastructures. The following is a classification of coordinated control, as shown in Fig. 1.4.
1.4.2.1
Centralized Control
Building a central controller that can communicate with all controlled units requires extensive communication infrastructures and significant computer resources. The main advantage of the centralized control structure is that it can apply optimal solutions. When the MG switches from grid-connected to islanded mode, one microsource can act as a master controller, providing voltage and frequency reference to others [21]. It allows simple algorithms to be used in the MG energy management unit. One of the major drawbacks of centralized control is that it suffers from a single point of failure. A centralized control structure is typically recommended for small environments such as educational centers and hospitals.
1.4.2.2
Distributed Control
In distributed control, each local controller operates on its own without instructions from a central controller. Any appropriate control actions are specified locally based on local evaluation and the information shared among neighboring local controllers of the MG through peer-to-peer communication. Since limited information about the entire MG status is communicated among neighboring nodes, optimal global
10
M. Shahbazitabar et al.
performance is not generally achieved in this topology as compared to the centralized control scheme. However, a “plug and play” feature is satisfied, meaning that when a DER is connected or removed from the system, the MG will continuously operate without reconfiguration. Multi-agent system (MAS) control is an example of such topology that allows every component to exchange information with its neighbors as an autonomous entity that can be decided based on its own status with no external command [22].
1.4.2.3
Hybrid Control
A hybrid control scheme, as the name suggests, is a combination of central and distributed controllers. Hybrid hierarchical control consists of several central controllers with a distributed topology that is coordinated. Each central controller contains several local controllers that can operate independently.
1.5 Stability RES integration with stochastic, uncontrollable, and intermittent nature is one of the attractive points of MGs, which in turn necessitates proper mechanisms in the system to assure reliable operation in transient events such as a sudden drop in wind speed or a cloud passing over a solar array [24]. DGs may be inverter-based or directly connected to MGs. Since MGs typically use renewable energy sources, most DGs are inverter-based. The existence of a variety of DGs can create diverse features in MG stability problems. Another issue affecting MG stability is its small inertia. An MG is stable if all the state variables are recovered to steady-state values after being subjected to a disturbance so that all constraints are satisfied [10]. It should be mentioned that, in MGs which generally are equipped with the inverter-based DER units, the inertia is zero or very low and the reference signal is used to set their output frequency, internally [25, 26]. As a matter of fact, the frequency will not inherently vary as the active power changes [27]. In traditional power systems, there is the chance to store energy by using the rotating masses of synchronous generators to deal with load changes. However, in MGs, some form of energy storage is needed to deal with transients happening during the islanded operation. As an effective manner, some load shedding strategies and storage techniques can be applied in MGs [27]. MG stability is divided into grid-connected MG stability issues and islanded MG stability subjects. DER integration with intermittent nature, which leads to stability issues in MGs, has been addressed extensively in research works considering both large and small-signal stabilities. Linear analysis tools such as Nyquist or Routh– Hurwitz for synchronous generators, inverters, rectifiers, and motors are in common. On the other hand, large-signal nonlinear stability studies such as the Lyapunovbased technique are employed for intrinsically nonlinear converters for integrating
1 An Introduction to Microgrids, Concepts, Definition, and Classifications
11
Fig. 1.5 Stability definition
RES and energy storage devices. Instability in conventional power systems can be classified into three general categories [28]: rotor angle stability, voltage stability, and frequency stability. All these three mentioned categories are highly dependent on the dynamic behavior of the synchronous generators [29].
1.5.1 Grid-Connected MG Stability When the MG is connected, its voltage and frequency are maintained by the utility grid. Therefore, stability studies of DGs with small capacities are not required. Since the capacity of the MG is much smaller than the utility grid, the disturbances in the MG will have little effect on the network frequency regulation. Therefore, rotor angle and frequency stability are not relevant in the grid-connected mode.
1.5.2 Islanded MG Stability When the MG is disconnected (islanded mode), it must support its voltage and frequency. In other words, it must preserve the balance between power generation and load. Therefore, it is crucial to investigate the voltage and frequency stability in an islanded MG. Long transmission lines are one of the main causes of voltage instability in conventional power systems, which limits the transmission of power between loads and generation. However, in MGs, the feeders are relatively short, resulting in relatively small voltage drops. Indeed, in an islanded MG, frequency stability is more significant than voltage one, due to its small inertia. Figure 1.4 shows the classification of MG stability types. The study of the stability types is beyond the scope of this chapter. A comprehensive review of MG stability can be found in [30]. A brief stability definition is clarified in Fig. 1.5.
12
M. Shahbazitabar et al.
1.6 Protection Electronically coupled unit control has significant effects on the MG transient behavior, particularly subsequent to faults. According to its fault specification, the fault types can be pole to pole or pole to the ground, bus or feeder. According to the fault location, the fault would be more severe near the energy sources; so, the bus fault is crucial in overall the system. Pole-to-pole short circuit may be occurring inside the capacitors and batteries that generally cannot be quickly diagnosed. In these cases, using fuses and circuit breakers along with device replacement could be profitable. Another challenge that has to be gradually overcome is an arc, which is created by the current interruption and extinguished hardly in a DC system without the current crossing through zero [17]. Also, low fault currents due to the power electronics interfaces, and adaptive protection because of the variety of generation sources, are two main subjects which should be clearly discussed and addressed in MGs. We should note that unlike conventional distribution systems in which the power flow is unidirectional and protection schemes are simple, in the MGs, the power flow is bidirectional, so their protection schemes are more complex. The main challenges of protection in DCMGs are related to following issues [31]: lack of phasor, and frequency data making it difficult to detect and accurate location of faults; absence of natural zero crossings to extinguish the arc occurring in circuit breaker opening; rising the fast current imposing strict time limits needed for fault interruption; protection coordination issues because of intermittent nature of DERs; and need for suitable protection standards.
1.7 Microgrid’s Advantages and Challenges In this section, the main advantages, and challenges of MGs are briefly addressed.
1.7.1 Advantages The main advantages of MGs can be categorized as follows: • Decreasing CO2 emission and fuel cost by using renewable energy supply instead of conventional fossil-fueled energy sources. • Increasing consumer reliability and power system flexibility. • Increasing the power transmission capacity in DC–MGs. • Loss reduction with local power delivery. • Having much smaller financial commitments.
1 An Introduction to Microgrids, Concepts, Definition, and Classifications
13
Operation Components and compability
MG challenges
Technical Integration of DERs Regulation Protection Economical marketing
Fig. 1.6 MG challenges
• Requiring fewer technical skills for operation and relying more on a remote control, and automation. • Isolating from any grid disturbance or outage.
1.7.2 Challenges Despite some clear advantages of MG, there are several challenges must be overcome. The main types of MG challenges are illustrated in Fig. 1.6.and summarized as follow:
1.7.2.1
Technical
• The electricity generation of some RESs, such as wind and solar, is highly dependent on weather conditions; hence their generation is unpredictable. Because their capacity is small, they are sensitive to unpredictable changes. This causes problems with operational capability. • Due to the smaller number of loads and interrelated changes in available energy sources, the uncertainty of islanded MGs is much greater than that of the utility grid. Even though the reliability of MG could be higher than the utility grid. • One of the challenges of islanded MGs is their low inertia characteristic in comparison with the bulk power systems due to numerous power electronicbased units and lack of conventional synchronous generators. This low inertia in the system can lead to intense frequency deviations. • One of the major challenges faced by MGs in islanded mode is maintaining the balance between generation and load continuously. Large disturbances can easily lead to MG instability. • In order to ensure the safety and reliability of the system, islanding conditions must be quickly detected.
14
1.7.2.2
M. Shahbazitabar et al.
Regulation
The current regulatory framework was not designed to incorporate DERs or MGs. Therefore, in some countries, changes have been made to the regulatory framework to influence the benefits of MGs to the entire community. Some laws about renewable DGs and energy storage systems have been incorporated into the new regulatory framework [32]. As an example, some of the common grid services provided by ESSs are categorized as power quality, transient stability, and regulation services; spinning reserve, voltage control, arbitrage (energy), load balancing, congestion relief, firm capacity, and upgrade deferral [33].
1.7.2.3
Economical
Despite advances in technology, the investment cost remains high in MGs. The cost of energy storage systems, some of DGs such as photovoltaic (PV) and fuel cells, is still high and not affordable. However, today in most countries, there are various types of financial support to facilitate conditions for investment in this field. Another economic challenge of MGs is its efficient energy management.
1.7.2.4
Marketing
MGs, in addition to supplying local loads, can sell their additional generation power to the utility grid or purchase some power from the utility grid. Thus, MGs can participate in the market by selling their products and services. MGs also play an important role in developing free/local energy markets by encouraging energy users to install DERs, offering new services, and supporting self-consumption [32]. While regulatory gaps can be primarily followed to the origin of DSO (distribution system operator) concerns, market gaps lying before the road of commercializing MG are mainly related to the direct economic interests of end users and DG owners. Based on these conditions, denial of local energy trading among the DG and demand, and MG market positioning difficulty can be determined as two major market challenges in this context [34].
References 1. Lasseter, R.H. and P. Paigi. Microgrid: A conceptual solution. In 2004 IEEE 35th Annual Power Electronics Specialists Conference (IEEE Cat. No. 04CH37551). 2004. IEEE. 2. Planas, E., et al. (2015). AC and DC technology in microgrids: A review. Renewable and Sustainable Energy Reviews, 43, 726–749. 3. Energy, U., DOE microgrid workshop report. 2018. 4. Hatziargyriou, N. (2014). Microgrids: Architectures and control. John Wiley & Sons.
1 An Introduction to Microgrids, Concepts, Definition, and Classifications
15
5. Asano, H., et al. (2007). Microgrids: An overview of ongoing research, development, and demonstration projects. IEEE Power Energy Magazine, 78–94. 6. Shayeghi, H., et al. (2019). A survey on microgrid energy management considering flexible energy sources. Energies, 12(11), 2156. 7. Abdi, H., & Shahbazitabar, M. (2020). Smart city: A review on concepts, definitions, standards, experiments, and challenges. Journal of Energy Management and Technology, 4(3), 1–6. 8. Lotfi, H., & Khodaei, A. (2015). AC versus DC microgrid planning. IEEE Transactions on Smart Grid, 8(1), 296–304. 9. Hooshyar, A., & Iravani, R. (2017). Microgrid protection. Proceedings of the IEEE, 105(7), 1332–1353. 10. Farrokhabadi, M., et al. (2019). Microgrid stability definitions, analysis, and examples. IEEE Transactions on Power Systems, 35(1), 13–29. 11. Azar, K. Power consumption and generation in the electronics industry. A perspective. In Sixteenth Annual IEEE Semiconductor Thermal Measurement and Management Symposium (Cat. No. 00CH37068). 2000. IEEE. 12. Chi, Y. and Y. Xu. Resilience-oriented microgrids: A comprehensive literature review. In 2017 IEEE Innovative Smart Grid Technologies-Asia (ISGT-Asia). 2017. IEEE. 13. Watson, J.-P., et al., Conceptual framework for developing resilience metrics for the electricity oil and gas sectors in the United States. Sandia national laboratories, albuquerque, nm (united states), tech. rep, 2014. 14. Pedrasa, M.A. and T. Spooner. A survey of techniques used to control microgrid generation and storage during island operation. In Proceedings of the 2006 Australasian Universities Power Engineering Conference (AUPEC’06). 2006. 15. Lopes, J.P., et al. Control strategies for microgrids emergency operation. In 2005 International Conference on Future Power Systems. 2005. IEEE. 16. Justo, J. J., et al. (2013). AC-microgrids versus DC-microgrids with distributed energy resources: A review. Renewable and Sustainable Energy Reviews, 24, 387–405. 17. Zhang, L., et al. (2018). A review on protection of DC microgrids. Journal of Modern Power Systems and Clean Energy, 6(6), 1113–1127. 18. Rodriguez-Diaz, E., et al. Multi-level energy management and optimal control of a residential DC microgrid. In 2017 IEEE International Conference on Consumer Electronics (ICCE). 2017. IEEE. 19. Meng, L., et al. (2017). Review on control of DC microgrids and multiple microgrid clusters. IEEE Journal of Emerging and Selected Topics in Power Electronics, 5(3), 928–948. 20. Shotorbani, A. M., et al. (2018). Distributed secondary control of battery energy storage systems in a stand-alone microgrid. IET Generation, Transmission & Distribution, 12(17), 3944–3953. 21. Yamashita, D. Y., Vechiu, I., & Gaubert, J.-P. (2020). A review of hierarchical control for building microgrids. Renewable and Sustainable Energy Reviews, 118, 109523. 22. Zhou, Y. and C.N.-M. Ho. A review on microgrid architectures and control methods. In 2016 IEEE 8th International Power Electronics and Motion Control Conference (IPEMC-ECCE Asia). 2016. IEEE. 23. Meng, L., Hierarchical control for optimal and distributed operation of microgrid systems. 2015, Ph. D. dissertation, 10 2015. 24. Kabalan, M., Singh, P., & Niebur, D. (2016). Large signal Lyapunov-based stability studies in microgrids: A review. IEEE Transactions on Smart Grid, 8(5), 2287–2295. 25. De Brabandere, K., et al. (2007). A voltage and frequency droop control method for parallel inverters. IEEE Transactions on Power Electronics, 22(4), 1107–1115. 26. Blaabjerg, F., et al. (2006). Overview of control and grid synchronization for distributed power generation systems. IEEE Transactions on Industrial Electronics, 53(5), 1398–1409. 27. Dag, O. and B. Mirafzal. On stability of islanded low-inertia microgrids. In 2016 Clemson University Power Systems Conference (PSC). 2016. IEEE.
16
M. Shahbazitabar et al.
28. Kundur, P., et al. (2004). Definition and classification of power system stability IEEE/CIGRE joint task force on stability terms and definitions. IEEE Transactions on Power Systems, 19(3), 1387–1401. 29. Maniatopoulos, M., et al. (2017). Combined control and power hardware in-the-loop simulation for testing smart grid control algorithms. IET Generation, Transmission & Distribution, 11(12), 3009–3018. 30. Shuai, Z., et al. (2016). Microgrid stability: Classification and a review. Renewable and Sustainable Energy Reviews, 58, 167–179. 31. Jayamaha, D., Lidula, N., & Rajapakse, A. (2020). Protection and grounding methods in DC microgrids: Comprehensive review and analysis. Renewable and Sustainable Energy Reviews, 120, 109631. 32. Bellido, M. H., et al. (2018). Barriers, challenges and opportunities for microgrid implementation: The case of Federal University of Rio de Janeiro. Journal of Cleaner Production, 188, 203–216. 33. Castillo, A., & Gayme, D. F. (2014). Grid-scale energy storage applications in renewable energy integration: A survey. Energy Conversion and Management, 87, 885–894. 34. Tao, L., et al. From laboratory Microgrid to real markets—Challenges and opportunities. In 8th International Conference on Power Electronics-ECCE Asia. 2011. IEEE.
Chapter 2
Operation Management of Microgrid Clusters Meisam Moradi and Asghar Akbari Foroud
2.1 Introduction This chapter deals with the operation management of networked microgrid clusters (NMCs) or networked microgrids (NMGs). The system that contains a connection of two or more microgrids with the ability to exchange energy with each other and with a distribution system (DS) is called NMCs. The NMCs differ from the DS includes multi-microgrids which exchange energy only with DS. In NMCs, power flow from one MG to another MG or DS is possible bidirectional and the topology of the network in the NMCs can change continuously. NMCs in the normal mode is similar to the DS with several MGs and have the benefits of these networks. The main difference of this structure is adding the networked mode between MGs. Designing and operation of a set of multiple MGs with DS as NMCs, will lead to increasing the resiliency of the network significantly. Networked mode operation in NMCs reduces system loss. The loss parameter is defined as the loss of power network persistence and failure to supply demand after an extreme event. Since the resiliency of the system is inversely related to this parameter, so, the resiliency of NMG is higher than the resiliency of the distribution network containing several MGs and traditional networks. Also, the improvement of resiliency increases system reliability. Studies show that the average system interruption duration index (SAIDI) and the average system interruption frequency index (SAIFI) are decreased by adding MGs to the network [1]. As the resiliency and reliability of NMCs are more than the DS with several MGs, the value of these mentioned two indices also improves more in the NMCs.
M. Moradi · A. Akbari Foroud () Semnan University, Semnan, Iran e-mail: [email protected]; [email protected] © Springer Nature Switzerland AG 2021 A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems, https://doi.org/10.1007/978-3-030-59750-4_2
17
18
M. Moradi and A. Akbari Foroud
To indicate the difference between NMCs and DS containing several MGs, Fig. 2.1 and Fig. 2.2 are presented. Figure 2.1 demonstrates the DS with NMCs. There are five MGs with one radial 33-bus IEEE DS. Each MG is capable of handling its loads. Also, in NMCs, each MG has one or more PCCs which makes the possibility of exchanging power with other MGs or DS. As can be seen in this figure, MG1 and MG2 can operate as the grid-connected mode by closing switch SW1 via PCC at bus 25 and switches SW2 and SW3 via PCCs at bus 33 or 14, respectively. These two MGs can operate as islanded or grid-connected. In addition to the island and grid-connected mode, the other three MGs can exchange power with each other in networked mode. By closing switches SW7, SW8, or both, the networked mode of MGs operation can be enabled. It is noteworthy that MG3 and MG5 with closing SW8 and SW7 through PCC at bus 18 and bus 22 can exchange power with MG4, respectively. Figure 2.2 shows the DS with two MGs. In this structure, the MGs are connected to the network via PCC on buses 12 and 33. Also, in this network, MGs are capable of supplying their loads and in order to have higher reliability, they are connected to the DS. Clearly, the first network is more resilient to severe contingencies than the second. In sever contingencies it is possible not to provide critical loads of the entire network or MGs and it is impossible to return to the initial conditions quickly. So, the resiliency of the system decreases against an extreme event. As shown in Fig. 2.1, NMCs had a higher level of resiliency due to its ability to change configuration against extreme events. Due to the flexible NMCs configuration, any interruption factor outside of the MGs is simply cut off. In this condition, the NMCs can exchange power through their physical network. By MGs isolation, the repair process in power distribution systems becomes faster. After removing the interruption factor, it is possible to start supplying the power of critical loads from the MGs. Accordingly, system performance can be recovered much faster with a simultaneous bidirectional supply method. Therefore, it can be
MG1 MG2 SW1
Main Grid
1
30
29
25
28
24
27
23
26
32
31
33 SW2
SW3
2
18
12 3
4
19
5
6
7
8
9
10
11
13
14
15
SW5
20
16
17 SW4
21 SW6
SW8
SW7
22
MG5
Fig. 2.1 Structure of NMCs
MG4
MG3
2 Operation Management of Microgrid Clusters
19 15
1
2
3
4
5
6
17
16
14
9
8
10
12
13
7 11 35
22
18
19
20
21
MG1 31 32
27 23
24
33
34
29
26 28
30
25
MG2
Fig. 2.2 Structure of DS with several MGs
concluded that the NMCs have more resiliency and less power return time than several MGs or traditional DSs. On the other hand, some researches are being done to develop the NMCs, such as Bronzeville Community Microgrid (BCM)1 and Illinois Institute of Technology (IIT)2 [1]. These researches demonstrate that NMCs can reduce contamination significantly and improve ancillary services, such as sustainability, security, efficiency, reliability, and cost reduction of providing customers demand. Also, according to research done by the Los Alamos National Laboratory, the NMCs can reduce the operating costs of MGs by at least 10% [1]. This is a significant reduction in operating costs. Given these cases, further studies are necessary on the NMCs. The major aspects of NMCs can be divided into four categories including architecture, control strategy, communication, and operation. In this chapter, in addition to the advantages and challenges of NMCs, the main objectives, planning, and operation of NMCs and the motivations that lead to the evolution of this power system, will be described. Also, the different types of energy management systems (EMS) and problem-solving methods in the NMCs will be analyzed. Microgrids can be “networked” in physical layers by distribution systems with closing or opening one or more keys or controlling layers by independent local controllers or both of them. It is important to note that the priority of any microgrid is the balancing of power and economic operation in the management of its internal
1 https://bronzevillecommunityofthefuture.com/ 2 http://www.iitmicrogrid.net/microgrid.aspx
20
M. Moradi and A. Akbari Foroud
resources, and then, if the microgrid is not capable of feeding its loads, it can buy the amount of its energy deficit from the other microgrids. In emergency conditions, the microgrid is separated from the upstream network and it uses the output power within the microgrid or if needed, it uses the output power from the other microgrids. In such conditions, it is observed that the EMS protects critical loads from being outage. These structures are highly flexible in critical situations and also resistant to unnecessary outages. However, distribution systems with multi-microgrids are not always in networked mode and switch to networked mode according to predefined conditions and planning. Figure 2.3 shows the operation of physical and control layers. It is to mention that the range of opportunities and potential architectures of NMCs like the individual MGs are very diverse, however, there are very few examples are deployed in NMCs. Also, there are very few tools to simulate or analyze the behavior of NMCs. Nevertheless, several national laboratories are developing design and analysis software tools that partially address some components of the quantitative evaluation of the economic, reliability, and resilience benefits of networked microgrids.
Network Controller Fig. 2.3 Performance of physical and control layers in NMCs
Controller N
Loads
Gen 2
Gen 1
Controller 2
Loads
Gen 2
Gen 1
Controller 1
Loads
Gen 2
Gen 1
Other loads and Generations
2 Operation Management of Microgrid Clusters
21
2.2 Benefits of NMCs Considering NMCs advantages, the number of MGs will increase and NMCs in the DS will be created. In developed countries, NMCS was considered as the transition state of smart grids [2].
2.2.1 Best Utilization of DERs In general, using DERs locally is more accessible in the microgrids network. These resources can be connected to the nearest distribution network in larger networks. In fact, the EMS in NMCs can network multiple microgrids and distributed generation resources can also feed the intended consumer in this condition. Therefore, EMS can reduce the total cost of the operating system in NMCs by integrating local DERs. In other words, in distribution systems with microgrids, if the use of DERs is done with a proper control or coordinator system, the uncertainty of wind and solar renewable energy sources in different microgrids can also be reduced by information aggregation and correlation. Besides, microgrids that have energy storage systems can play an important role in smoothing the power generation of DGs. For example, a storage system of one MG can be entered at peak load times of other MGs and reduce the startup cost and total operating cost of the system by preventing the installation of thermal units. This work reduces the peak load and facilitates demand response. Also, at the transmission level, the integration of DERs using the microgrids network can increase the efficiency of the transmission system and postpone the investment of transmission substruction and new substations.
2.2.2 Reduction of Overall Cost The integration of DERs using the microgrids network can decrease the total cost of the required generation power. For example, a study done by the Los Alamos National Laboratory shows that operation in NMG mode reduces at least 10% of costs compared to normal distribution network mode. This amount of reduction in operating costs is significant. Also, it has been shown in [3] that the interaction between interconnected and independent microgrids in the form of NMCs, with planning strategy and energy trading development, can reduce the total operating cost by 13.2% and reduce the cost by 29.4% for one MG.
22
M. Moradi and A. Akbari Foroud
2.2.3 Ancillary Services Improvement Optimal use of DERs can improve ancillary services in NMCs. In fact, ancillary services keep flowing and direction of power by maintaining a balance between supply and demand and it can help the system returning to normal mode when faults occur. Also, since EVs and ESSs have been used in NMCs and EVs are capable of transferring stored power, proper planning of this equipment can further enhance the ancillary services market. Therefore, at the planning stage of NMCs, the addition of power storage equipment can add benefits to the ancillary market. Studies show that using NMCs improves 10% of service costs and the ancillary services market [3].
2.2.4 Resiliency Improvement Improving system resilience is one of the most important goals of power systems. The use of NMCs allows the generation resources to be distributed between microgrids and loads near to DSs, which can lead to additional resiliency and lower investment costs. On the other hand, networked mode operation in NMCs reduces system loss. The loss parameter is defined as the loss of power network persistence and failure to supply demand after an extreme event. Since the resiliency of the system is inversely related to this parameter, so, the resiliency of NMG is higher than that of the distribution network containing several MGs and traditional networks.
2.2.5 Reliability Improvement One of the benefits of NMCs is the increment of system reliability. This advantage can be achieved by reducing the amount of load which is not provided or reducing the utility outages. This reduction of blackouts in NMCs may increase costs by balancing between generation and consumption, but there must be a balance between the costs of removing remaining blackouts against the entire cost of the network. This means that removing the load may be more economical for the network. However, networking the MGs and sharing their resources reduce the total investment costs of reservation generation and increment of reliability. It is usually suggested that critical peak load be considered to account for approximately 80% of the backup generation capacity [3]. In a simple example, suppose if two MGs have a peak load-to-average load ratio of 1 MW to 0.5 MW and the critical peak load time of them are not the same, instead of buying two generators with a nominal capacity (1 MW/0.80 = 1.25 MW), in NMG mode, the total required capacity of 1.5 MW/0.80 = 1.88 MW can be used. This capacity reduction will reduce the cost of $186,000 by considering the cost of 0.30 $/W for a diesel generator. However, if the capacity of the bought generators
2 Operation Management of Microgrid Clusters
23
is not reduced, a higher capacity of the generators can increase the reliability of the system. Also, the improvement of resiliency increases system reliability. Studies show that the average of system interruption duration index (SAIDI) and the average of system interruption frequency index (SAIFI) are decreased by adding MGs to the network. As the resiliency and reliability of NMCs are more than the DS with several MGs, the value of these mentioned two indices also improves more in the NMCs.
2.2.6 Bilateral and out of Market Transactions Since in NMCs, power exchanging between MGs exists, there is also the possibility of power exchanging outside the wholesale market without controlling. Suppose in the series structure of NMCs, upper MGs buy the required energy from the distribution network with wholesale price. Since the only way to meet the shortage of power required by the second MG is to purchase it from the first MG distribution feeder. Therefore, the second MG has to buy a lack of power at the retail price from the first MG, this bilateral transaction is not acceptable in NMCs. In other words, if the pure measured energy from the upstream system is purchased with wholesale rate while being sold with retail rate to the other consumers or MGs, bilateral transaction rules in NMCs are allowed to prevent exporting energy. Because in the retail transaction, it is possible to increase the price anomalistic price. In many areas, this price increment can reach 50$/MWh. For example, if multiple microgrids on a 1 MW grid can prevent 1 MW from exporting over 4 hours per day, the total profit would be 73,000$/year [4].
2.3 Challenges of NMCs 2.3.1 Stability of the System Disconnecting from the upstream network and switching to the network mode can cause system stability problems. By isolating MG, the penetration level of resources and uncertainties of them are increased compared to the previous one and can cause challenges in the stability of the system. In Ref [5], the stability problems associated with NMCs have been discussed exactly. By increasing the penetration level of resources with uncertainty, there is the possibility of transition from stability margin. Therefore, for any change in the structure of NMCs, stability limits must be examined. As the topology of NMCs is changing constantly, stability assessment is essential for this structure. Therefore, due to widespread and rapid changes in NMCs, control systems should be able to analyze network stability with high speed. In other words, any change in NMCs, especially when operating in the island state, should be done to ensure sufficient stability margins of them. Once the margin of proper stability is assured, economic dispatch can be done, and only
24
M. Moradi and A. Akbari Foroud
when networked MGs have sufficient stability margins, they can serve as resiliency sources to actively and coordinately provide ancillary services that stabilize, restore, or black start the main grid.
2.3.2 Protection Coordination Proper protection coordination is one of the most essential requirements for using high-reliability NMCs. As mentioned in the previous section, to achieve the best performance, the network structure in NMCs has been modified frequently and this can cause a challenge in coordinating protection equipment. It should be noted that by changing the network topology, the amplitude and direction of fault currents are constantly changing and can be quite different from the previous one in each case. Also, the coordination of protective equipment in the grid-connected mode, isolate mode, and networked mode is different from each other. Therefore, a large number of possible topologies of NMCs changes must be preplanned. On the other hand, the equipment fault current rate of each MG is usually considered above the nominal limit, but it is not valid for upstream equipment. The expansion of grid-connected MGs increases the level of system fault current and increases the nominal value of equipment such as transformers, breakers, and protection equipment in the upstream network (DS). Replacing equipment with higher fault current endurance is not an economical solution. One of the proper methods is to use an adaptive protection scheme, but it will increase the complexity of the protection system.
2.3.3 Privacy of MGs In terms of ownership, MG is divided into three categories, utility MG, community MG, and private MG. Privacy is a social concern and since all MG owners may not want to share all the information except the details of power exchange, this can be a challenge for NMCs. It should be noted that excess or deficit of power in NMCs is usually compensated within the internal network of MGs and this will reduce the privacy challenge partly.
2.3.4 Threat of Cyberattack Information and Communication Technology (ICT) is one of the main elements of NMCs. Therefore, there is always the possibility of a cyberattack that can prevent the proper entire function of the system. Therefore, cybersecurity systems must be strong enough to prevent any probable cyberattacks. Any disturbance of the processing of information or sending it to unidirectional or bidirectional communication
2 Operation Management of Microgrid Clusters
25
protocols can endanger the entire security of the system. Both in the centralized and distributed control scheme of NMCs, there is the possibility of cyberattacks and serious damage to the system. Of course, centralized control systems are more resistant to cyberattacks [6]. The vulnerability is relatively higher in the distributed control scheme, because of requiring partial data of local controllers. For example, a destructive factor can attack to one node or linkage between nodes and exchange inaccurate data to local or central controllers and causes instability of the entire system. So, NMCs that require extensive communications for optimal control are very susceptible to future cyberattack.
2.3.5 Disallowed Transactions As mentioned in Sect. 2.6, in systems with NMCs, it is possible to exchange power out of the framework. This will significantly increase the price of energy for the consumer. Therefore, the exchange of power between MGs in the grid mode and disconnected from the main grid requires an appropriate regulatory framework to minimize illegal transactions or nodal prices increment.
2.4 Main Objectives and Constraints of NMCs The main objectives of NMCs are divided into two levels of the distribution network and the MG level. The objective at the distribution network level is to achieve economic dispatch and maintain power quality throughout the studied system. To achieve this objective, using distributed generation resources should be optimized and the EMS should be implemented effectively in both the distribution system and all MGs. The main operation of EMS in distribution systems includes optimal economic performance and energy quality throughout the system. The main objective at the level of MGs is to control voltage frequency, power supply, and generation balance, and to manage energy storage and reservation at each MG. In NMCs, EMS for each MG supplies the power of each MG. The extra power generation of MG is stored in the storage system or delivered directly to the DS or adjacent MG through EMS coordination at the distribution level. Similarly, the lack of power of MG is supplied either through the distribution network or through a grid connection between the MGs, directly from the MGs of itself. Therefore, with EMS, the power supply throughout the system continues uninterrupted and economically. This economic dispatch program in NMCs is done by EMS at predetermined time intervals. It should be noted that the studied system can be grid-connected, disconnected from the grid, and a combination of both of them (Only NMCs should be disconnected from the grid). NMCs are capable of sharing power between different MGs of one network or between MG and DS. In an emergency condition and
26
M. Moradi and A. Akbari Foroud
isolation state from the distribution network, all MGs or some of them with more significant loads operate as a grid by EMS. In this condition, only critical loads are initially provided in MG. Extra power can be sent to the main grid after providing significant loads in the DS depending on the communication between the DSO and the transmission system operator. However, if power generation is insufficient to supply all the loads on the DS, only significant system loads can be supplied by activating energy storage equipment in each MG. Therefore, NMCs can feed critical loads for longer periods by DERs and ESSs in critical conditions. In Table 2.1, general considerations, the main purposes, and constraints used in NMCs are discussed.
2.5 Typical Architecture of NMCs In general, the connection of MGs to each other or connection of them to a distribution network is different. Different connections can have advantages and disadvantages for NMCs. Since the structure of NMCs has many variations for optimal control, it is necessary to know the types of NMCs architecture. In different architectures, the role of each MG can vary depending on the laws of interaction between them. There are generally two methods of interaction. For example, in one structure, one MG may be considered as a generating and controlling member but in another structure, it can be a controllable member [1, 3, 31, and]. In general, the architecture of NMCs can be divided into three overall categories: serial MGs on a single feeder, parallel MGs on a single feeder, and interconnected MGs with multiple feeders. Depending on the architecture considered for NMCs, a set of unique communication, control, protection, and economic requirements are considered. Figure 2.4 shows the different types of architectures used for NMCs.
2.5.1 Serial MGs on a Single Distribution Feeder If there is no external grid, the interconnected system must be able to control the voltage and frequency. To achieve this goal, coordination between MGs is essential to balance power. Unlike parallel architecture, in the event of an occurrence or disconnection, the system can be divided into subclusters with similar architecture. According to this topology, MGs within the subcluster do not lose completely external support. So, this topology performs better than the off-grid operation. The serial architecture needs further evaluation and study due to its potential performance and benefits. According to Fig. 2.4a, if two or more MGs connect together by a single interconnection and install on one feeder, they will form the structure of the serial MGs on a single feeder. In this configuration, there is only one way of communication through MG A to MG B. As MGs are interconnected and have
2 Operation Management of Microgrid Clusters
27
Table 2.1 General considerations and the main objective in NMCs General considerations Internal communications of MGs Minimize load shading Stability issues and high penetration of DERs MG syncs issues and pay attention to cyberattack
Main objective Methodology Develop Software PSCAD hierarchical control Stability analysis FARs and Gersgorin theorem with LP Pay attention to Observer based cyberattack
Ref [19]
Offline studies without ESS
[5]
Limited to [6] inverter-based MGs Convex programming The role of the [7] with ADMM network operator has been ignored OBS and ODD Uncertainty not [8] seen
Incentive-based model with the presence of PV, WTs, and ESSs Priority indexing of sources and loads Optimum energy scheduling Coordinated energy management In the presence of uncertainty and scenario reduction methods Two-layered stochastic energy modeling with loads and DG uncertainties Distributed energy management Uncertainty in DERs and loads Energy management Online monitoring Controller for a wide range of functions and communication failure is modeled. Resiliency improvement N–k contingency modeling Self-healing and autonomous operation Energy storage and MTs Dynamic electricity pricing based on incentive model, PV, WTs, MTs, FCs, and ESSs Service and maintenance With DG and ESSs
Ancillary service analysis
Voltage control and power management
Voltage stability Simulink modeling
Economic scheduling
Highlights Offline study
Energy management
Stochastic programming KKT and big-M method
ESS has not been [9] seen.
Minimize operating costs
Stochastic optimization
Only active power flow
[10]
Energy management
Robust optimization ADMM
High-volume computing
[11]
Energy management Develop hierarchical control
ADMM online
High-volume computing PV and wind uncertainties are not seen.
[12]
Software simulation
NMCs planning Robust optimization CPLEX Energy Consensus algorithm management Economic scheduling
Fuzzy adaptive MOPSO
Planning service MILP stochastic and maintenance
[13]
ESS has not been [14] seen. Dynamic [16] stability is not guaranteed. Uncertainty not [17] seen
High[18] performance PC is needed. Delay in [20] communication causes instability. (continuted)
28
M. Moradi and A. Akbari Foroud
Table 2.1 (continued) General considerations Outage detection Reconfiguration and load curtailment
Main objective Economic scheduling
Methodology MIQP
Highlights An advanced metering infrastructure system is required. The study is offline and requires large ESS. It has an uncertain solution. Privacy issues are not considered
Ref [21]
V/F stability
Develop hierarchical control
Software PSCAD
P/F/V control With linear approximation
Economic scheduling
MILP
Suitable only for MGs with nondispatchable DGs. It is a day-ahead scheduling Reliable communication is required. Uncertainty is not considered.
[25]
NMCs planning ICA
ESS not seen.
[29]
Demand response and energy management programs
PSO
It is a day-ahead scheduling
[30]
Energy management Exchange of active and reactive power
Self-healing control
Load and generation uncertainties are not considered.
[15]
High-volume computing
[15]
Combined AC/DC MGs with various DGs
Develop hierarchical control Economic dispatch problem Economic Non-dispatchable DG and scheduling ESS
Simulink
Hybrid energy management MGs privacy Auction-based production control short-term energy management Energy modeling with MPC framework Different types of DGs and ESSs Minimize operating costs Uncertainties in DERs and loads Maximize the profits of customers and energy producers Reduce production costs/pollution/uncertainty and maximize load supply Minimize the operating costs of each MG and automatic power supply in NMCs and charge control of energy storage before and after fault NMCs resiliency DC/AC MGs and uncertainties
CPLEX
Energy management Automatic generation control Energy management
MOIA
OPC toolbox in MATLAB MILP
NMCs planning Robust optimization consensus algorithm
[22]
[23]
[24]
[26] [27]
[28]
2 Operation Management of Microgrid Clusters
29
1
2
3
Main Grid SW1 Communication line
SW2
Power line
MG B
MG A
DMS (a) Serial MGs on a single feeder of DS
1
2
3
Main Grid SW1 Communication line
SW3 SW2
Power line
MG B
MG A
DMS (b) Parallel MGs on a single feeder of DS
1
2 SW1
MG A
Main Grid
SW2
DMS
MG B SW3
3
4
(c) Interconnected MGs on multiple feeders of DS
Fig. 2.4 Types of architectures used in NMCs
only one way of communication, a communication protocol with a master-slave architecture or hierarchical communication system can be used to control this structure. In this structure, the distribution management system (DMS) provides the communication requirements for the main controller of MG A, and the MG A controller (MCA) also provides the communication requirements for the MCB similarly. Since there is only one connector between the equipment, each DMS needs to exchange information with the MCB.
30
M. Moradi and A. Akbari Foroud
The following situations may occur for this architecture: If both SW1 and SW2 switches are closed. The MGs are connected to the distribution feeder. In this case, each MG controller is optimized and controlled according to its information and orders and the other MGs. In other words, for optimization of the MGA, lack or extra of MGB power should also be considered. If both SW1 and SW2 switches are opened. MGs are isolated from the distribution feeder and operates as an island mode. In this case, each MG controller is optimized and controlled according to its information and orders. If the SW1 switch is closed and SW2 is opened, MG A is connected to the grid and MG B operates as an island mode. MCA controls MG A based on information obtained from DMS, while MCB controls only MG B through the information of internal resources. If the SW2 switch is closed and SW1 is opened, MG A is disconnected from the grid and MG B is connected to MG A through internal communication. In the hierarchical control architecture, MG A can consider MG B as part of its system and optimize it individually or optimize MG B with terms of total cost objective function and voltage and power limit of point common coupling (PCC). For these conditions, MCA should have plug-and-play capability. But, if a distributed control system is used, MCA and MCB will interact to reach an agreement about the voltage and power limitation of the AB at the PCC.
2.5.2 Parallel MGs on a Single Distribution Feeder According to Fig. 2.4b, if two or more MGs are installed individually like tie interconnection on one feeder and connected together, they will form a parallel MG structure on a single feeder. In this structure, all microgrids are connected to the same external network and each MG needs a PCC for connecting to the network. Therefore, any electrical path between MGs is performed through an external grid. Regarding architecture and the grid-connected mode, the MGs can provide ancillary services to the external grid. Also, the operator of an external grid can send orders to the MG with the DMS controller. When one MG operates in island mode, it must be self-sufficient because it has no other external electrical connection to support it. In this structure, the master-slave or hierarchical control architecture can also be used and different modes of this connection to the distribution system can be considered: If SW1 and SW3 are closed and SW2 is opened, both MGs A and B are connected separately to the distribution network. In this case, the controller of each MG is optimized and controlled by DMS based on their information or orders. If all three switches are opened, both MGs A and B are disconnected from the grid and each MG is controlled and optimized based on its information and internal resources.
2 Operation Management of Microgrid Clusters
31
If SW1 is closed and SW2 and SW3 are opened, MG A is connected to the grid, but MG B is island mode. In this case, the controller of MG A is optimized and controlled based on the information or order obtained from the DMS, while MG B is controlled based on its information and internal resources. If SW1or SW3 is opened and SW2 is closed, MG A or MG B is connected to the grid and the second MG is connected to the other MG by internal connection and they form an internal grid. In this case, the grid structure becomes a serial architecture connected to the single feeder.
2.5.3 Interconnected MGs on Multiple Distribution Feeders Interconnected MGs in multi-feeders’ structure, that two or more MGs are connected separately to different feeders, are interconnected MGs on multiple distribution feeders. In this architecture, MGs can be connected directly to the external grid or can form clusters of series interconnected MGs. Each of these clusters has, at least, one interconnection with the external grid. It allows MGs to get support from other grids when they are disconnected or when the external network is overloaded. Besides, in grid-connected mode, they can provide the ancillary services required by the main grid operator. Assuming a master-slave architecture for this structure, the following modes associated with the distribution network may be formed: 1. Both MGs A and B are connected separately to the grid with different feeders. Each MG controller is optimized and controlled by the DMS based on its information and orders. 2. Both MGs A and B are disconnected from the grid and are optimized and controlled based on their information and internal resources. 3. If MG A is connected to the grid but MG B is island mode or vice versa. In this case, control and optimization are similar to the previous sections. 4. If MG A or MG B are connected to the grid, the other is connected via PCC AB and communication switch. By coordinating the separator switches in the feeders, different paths can be provided to transfer power between the feeders or other MGs. This is a unique feature. A comparison between different architectures of NMCs is not easy as two compared systems must be similar in terms of amount and rating. For this purpose, a comparison is made between two parallel and serial structures in Ref [31]. To compare both architectures, similar features must be considered. So, according to Fig. 2.5, both examples contain 4 MGs in the form of a ring. To reduce this effect, the size of the transmission cables should be large enough. However, in both plans, cable size increment should be considered. But, in serial structure, interface cables size between MGs should be increased. It can be noted that power can flow both through the cables and the interface device. But in the parallel structure, the interface elements are designed solely on their input and output power and they have fewer elements than the series structure.
32
M. Moradi and A. Akbari Foroud PCC 1
MG1
Link 1-2
Link 1-3
MG3
MG2
PCC 2
MG1
MG2
MG3
MG4
Link 2-4
Link 3-4
MG4
PCC 3
PCC 4
Fig. 2.5 Series architectures (left) and parallel (right) with 4 MGs
For example, in Fig. 2.5, the serial structure requires eight interface devices, but the parallel structure requires four interface devices. Therefore, according to the required higher rating equipment and the cost of interface devices, the serial structure has higher costs than the parallel structure. The cost of the hybrid structure is also expected to be between parallel and serial structures.
2.6 Control Strategy of NMCs Coordinated control strategies for MGs are divided into four general categories: (a) peer-to-peer control strategy; (b) master-slave control strategy; (c) hierarchical control strategy; and (d) distributed control strategy.
2.6.1 Peer-to-Peer Control Strategy in NMCs Peer-to-peer control systems are recommended for the plugging and playing of the MG controller. The peer-to-peer control strategy in NMCs has been developed to remove the problem of centralized control. This method avoids the increment of communication links and it can develop well and decrease the cost. This type of control strategy has not a central controller and it is inspired by P2P computer networking. As shown in Fig. 2.6, all agents or local DERs can communicate with other agents and have the same importance. In peer-to-peer, MGs are autonomous due to the absence of a central controller, and its communication is used for dissemination of the grid states to all required agents in the MG. The grid-supporting agents can then act according to the received information and in cooperation with each other. In this way, they should be able to reach an optimal operation of the considered microgrid. When a single agent fails, the others can still stably manage the grid. Also, when a single communication
2 Operation Management of Microgrid Clusters
33
Fig. 2.6 P2P control strategy
channel fails, the required information can still reach all necessary participants, through other agents. Because of these properties, this strategy is a robust way to control an MG. Also, the elimination of possible privacy concerns is reached due to keeping all information locally. So, all agents need a considerable amount of local intelligence, as they need to be able to execute the necessary optimizations.
2.6.2 Master-Slave Control Strategy in NMCs In master-slaves strategy, one section is known as Master and the other is known as Slave. The information is transmitted between the master controller and the slave controllers. The technical difficulties and risks of this control system are low for NMCs, but if they fail, the main controller of the MGCs and the entire control system of the NMCs will not be able to perform well. Therefore, if the system relies too much on the main controller, MGC reliability will get into trouble. It is noteworthy that the master-slave control is mostly used in the island mode of NMCs and the peer-to-peer control is mostly used in grid-connected mode.
2.6.3 Hierarchical Control Strategy in NMCs A hierarchical control strategy (HCS) is used to solve the stability problem when changing the control mode. Hierarchical control is the most common method in MGs and MGCs and is also suitable for more complex systems including NMCs. A multilayer HCS is commonly used to control NMCs. Figure 2.7 represents the
34
M. Moradi and A. Akbari Foroud
• • • •
Primary Control
Secondary Control
Tertiary Control
(Distributed)
(Quasi-Centralized)
(Centralized)
Optimized Energy Management Economic dispatch Unit commitment Optimal power flow Reactive power control (Distributed/networked level)
• • • • •
Regular Error Correction Real-time load management Load-frequency control Voltage control Synchronization Automatic generation control (Networked/ MG level)
Fast Load Sharing • Droop control • Local protection (MG level only)
Measurement Signal Control Signal Milliseconds
Minutes
Fig. 2.7 Hierarchical control strategy for networked MGs
HCS for controlling the performance of NMCs. This control structure has primary, secondary, and tertiary layers. The main objectives for each control layer are shown in Fig. 2.7. The main objective function of the primary (first) control layer is the exclusive and local control of the performance of each MG equipment such as V/f, P/Q, and P/V. Power quality, power flow, and frequency synchronization control are performed in the second layer. Optimal energy management such as economic load dispatch, load forecasting, and resource optimization at the distribution network level is performed in the third layer of the hierarchical control structure. Also, DMS or EMS, in general, has direct control over all separator switches and controls the PCC through MG controllers. It is noteworthy that a reliable communication channel is an important section of this type of control strategy [32]. In the grid-connected mode of NMCs, DSO communicates with all grid MGs to control and optimize the economic dispatch and internal resources. In fact, in this control system, DSO monitors the power flow of all PCCs in the grid. This monitoring is done based on price signals/incentives and general objectives/optimization requirements and is based on solving the economic dispatch problem. Also, when island mode occurs, DSO sends an islanding mode signal to MG or the MG may also send the DSO disconnection signal after the disturbance occurs. In such cases, the DSO may also request from neighboring MGs to switch the island mode. The synchronization of MGs to the distribution network or switching to the networked mode is performed by the DSO and the request of the incoming MG controller. However, the reconnection of the two islanded MGs is done only when the agreement is reached between the two MGs. First, the V/f control responsibility is assigned to one MG and the PCC connection point voltage is kept constant, then the PCC is closed and the networking signal is sent to the two MG sources. HCS
2 Operation Management of Microgrid Clusters
35
requires data collection from all the major components of MGs, and links need to be established between the DSO and each local controller.
2.6.4 Distributed Control Strategy in NMCs In distributed control strategy (DCS), each microgrid controls all performances locally and independently and shares all important information with the other MGs and DSO controllers. Figure 2.8 represents a typical DCS for controlling NMCs. Performances such as V/f control, P/Q and P/V control, power distribution, frequency synchronization, and optimal EM at the MG level are fully distributed. To achieve optimal performance across the control system, information is shared with all local controllers in each MG as well as with the DSO. Local controllers continue their performance by sharing information to optimize performance at the MG level. This control method can be used in NMCs because of its high resiliency and “plugand-play” capability. Because NMCs have high changes in performance level, this control strategy can maintain optimal system performance in consecutive changes. Figure 2.9, shows the adopted DCS to control networked MGs. In this strategy, the primary (distributed) control is performed in the shortest possible time to maintain always a balance between load and generation, and the secondary control that is quasi-centralized is implemented in the next step. Primary control is similar in both distributed and hierarchical control strategies. By changing the system loads, the PCC voltage and MG frequency will be changed. The voltage and frequency
Primary Control
Secondary Control
(Distributed)
(Quasi-Centralized)
Information Monitoring • Information collection from agents of the other MGs •
Load Sharing and Error Correction and Energy Management • Droop control • Local-frequency control • Synchronization • Real-time load management • Optimal power flow (MG level only)
Information sharing with the agents of the other MGs (Distribution/ networked level)
Local MG information Other MG information Milliseconds
Seconds
Fig. 2.8 Distributed control strategy for NMCs
36
M. Moradi and A. Akbari Foroud
DG 1
Tertiary Control
Primary Control
Droop Control
Reference Generator
V&I Control
Local Microgrid Power Network
Local Consensus
PWM + UPS Inverter
f&V Restoration Virtual Impedance
Secondary Control
Local Consensus
DG 2 Primary Control
Droop Control
Reference Generator
V&I Control
Networked Microgrids Power Network
Secondary Control
Local Communication Network
Networked Microgrids Communication Network
Global Consensus
PWM + UPS Inverter
F&V Restoration Virtual Impedance
PCC 1
Fig. 2.9 Proper operation of NMCs based on an overall adopted control strategy Table 2.2 Comparison of control strategies Features Reliability Plug-and-play Flexibility/expandability Communication bandwidth Time/space complexity Design complexity Economic operation Hardware platform
Hierarchical control Moderate Low Low Low High Complex Optimal Powerful computer
Distributed control High High High High Low Simple Suboptimal Embedded controller
correction in secondary control is performed in both control strategies to return them to their previous states. A comparison of some features of both control strategies is presented in Table 2.2 [33]. In the DCS, the master control gives the command to the power generation units with the aim of setting and supplying the required demand. However, in this strategy, the economic objectives are arriving through the consensus of all local controllers per MG. In primary control, the amount of power of the generation units and in secondary control, the amount of voltage and frequency in each MG are kept within the allowable range. For this purpose, power electronic converters (PECs) are responsible for restoring v/f. The adopted distributed control strategy of NMCs with utilizing PECs is shown in Fig. 2.9.
2 Operation Management of Microgrid Clusters
37
2.7 Energy Management and Operation of NMCs This section deals with the topic of Energy Management System in NMCs. In systems with one MG, the MG Central Controller (MGCC) is responsible for energy management and it has the highest priority in hierarchical control. But if two or more MGs form the structure of NMCs, a NMG central controller (NMGCC) can be considered to take the task of energy management and the highest priority of hierarchical control and coordinate with all existing MGCCs. As stated (mentioned) in Sect. 6.3, in control systems, control is performed in millisecond up to a few minutes. The EMS manages three levels of control during this period. At the first and second levels, voltage and frequency control, power balance, and load management are performed. But at the third level, EMS has the task of feeding critical loads under any circumstances by optimally managing power transmission with other DGs, distribution networks, feeders, or neighboring MGs. The time framework of the EMS is shown in Fig. 2.10a. If the EMS is also used in a distributed control system, according to Sect. 6.4, functions such as V/f control, P/Q, and P/V control, power distribution, frequency synchronization, and optimal energy management at MG level are done fully distributed. Then, in the networked mode of the MGs, to achieve optimal performance across the system, the information is shared with all local controllers in each MG as well as with the DSO. This is the task of the NMGCC. In fact, in systems with NMCs, another control layer called NMGCC is added to communicate with MGCC, DMS, and tie-line control and optimize the network in this case. This framework is shown in Fig. 2.10b.
2.7.1 Energy Management Strategies of NMCs Achieving optimal economic dispatch and making the best use of DERs are the main objectives of NMCs. To achieve these objectives, the EMS must operate properly at distribution and MG levels. The major objective of EMS at the DS level is improving overall economic performance and power quality of the entire system. Also, the main purpose of EMS at the MGs level is monitoring and improving power balance performance, optimizing the use of ESSs, and controlling effectively V and F across all MGs. Figure 2.11 shows the overall performance of EMS in NMCs [1]. In NMCs, the EMS maintains the power supply for each MG. An excess power generated in the MGs is stored in a storage or delivered directly to the DS or adjacent MGs via EMS coordination at the distribution level. Similarly, if one MG has a lack of power, the EMS system supplies the required power to the MG through the distribution network or adjacent MGs. It is noteworthy that the MG controller, due to economic considerations, can buy its requested power deficit from the distribution network or other MGs, or it can regard buying this lack of power by using the DR program and eliminating the load. Therefore, according to economic purposes, EMS
38
M. Moradi and A. Akbari Foroud
Several minutes - Day CO2 Emission limit
Several Seconds- Minute
Millisecond - Second
Load Management
Fast-Dynamic Storage Availability
Economic Dispatch
Power Balancing
Electricity Market
Unit Commitment
Lond Forecasting
RES Production prediction
Energy Storage Availability
Long Term
Dispatchable DG(s), Tie- Line Converter(s), Switch(es)
a
Power Capability of RE-DG
Short Term
b
DMS Tie – Line Control
Network MG Central Controller MGCC
MGCC
Local Control
Local Control
MGCC
MG1
Local Control
Local Control
MG3 DG Control
Hybrid PV – Bat Control
Power Transfer Information
MG2
Fig. 2.10 (a) Time framework of the energy management system, (b) NMGCC location in the control system
2 Operation Management of Microgrid Clusters
39
Main Grid
EMS Loads
MG1
P1 Q1
EMS1
PN
(c) Energy-storage management
Storage
MGN
(a) Voltage-frequency control (b) Load generation monitoring (c) Energy-storage management
(b) Load generation monitoring
Generators
QN
EMSN
MG1 to MGN
(a) Voltage-frequency control
Loads
DGs
(a) Economic power dispatch (b) Power quality monitoring
Loads
Generators
Storage
Fig. 2.11 Overall performance of EMS in NMCs
can maintain the power supply of the entire system. This optimal economic dispatch of the entire system is done by the EMS at regular intervals and the EMS can change MGs to the island mode, grid-connected, or NMCs depending on the calculated optimal solution. In an emergency, if the connection is separated from the main grid, all MGs will be connected to the network throughout PCC by predetermined operation done by the EMS. In this condition, only the critical loads in each MG are supplied and the extra power is sent to the DS. If power generation is not enough to supply all the existing loads in the DS, only the critical loads of the distribution system are supplied by the ESSs of MGs. This will allow the EMS to handle critical loads of MG for more hours. Of course, different scenarios can be programmed to provide power depending on predetermined conditions and rules. Different types of EMS strategies in NMCs are discussed in the following.
2.7.2 Compare EMS Structures In recent years, various types of EMS have been studied and used in various MGs researches. EMS can be divided into three general categories. Centralized, decentralized (distributed), and hybrid EMS. In a centralized structure, all MGs are controlled through a single management system. This system reduces the total operating cost by preventing the removal of critical loads under any possible conditions. The system has a simple structure and in island mode, it has acceptable reliability. Because of requiring high communication infrastructure, this structure
40
M. Moradi and A. Akbari Foroud
can impose heavy costs on the system. It also has less resiliency in faults and is more useful in slow-changing situations. In a decentralized EMS, each MG has its local control center that can operate independently [2]. Each MG balances its generation and loads through power sharing with the distribution network or other adjacent MGs. Compared to centralized EMS, this type of EMS can work properly in NMCs mode. But this structure, in grid-connected structure mode, is highly dependent on the main network. This increases uneconomical operating costs. Therefore, the decentralized or distributed EMS structure in the island mode of MGs or gridconnected mode is not very useful and resilient. According to the advantages and disadvantages of both energy management systems, a hybrid management system can solve the problems of centralized and decentralized structures. Various hybrid systems have been proposed in different references. The hybrid EMS structure is usually such that local resources in each MG are optimized and the central EMS modifies the extra or lack of power in each MG. This hybrid system significantly reduces operating costs compared to the decentralized EMS strategy. However, these hybrid systems also have problems in the island mode. In these EMSs, because of the parallel performance of the hybrid system in the MGs (optimize the MGs in parallel) and since the MGs are unaware of the local data of the other MGs, it may not achieve the economic benefit of the entire network. So, especially when MGs are used as NMCs, the hybrid EMS structure must receive all local data of MGS and manage the system according to their analysis. In this regard, the EMS structure based on NMCs provides an opportunity for all MGs to become aware of data from other MGs such as power generation level, required load for customers, and the required amount of buying or selling of extra or lack of power. Thus, in the hybrid EMSs used in NMCs, each MG is responsible not only for balancing its power but also for providing optimal economic performance in the interaction of energy with other MGs. In this case, slight changes in the performance of each MG can make a big difference in the power generation of other MGs. Accordingly, different hybrid EMS strategies are observed in different references. Various EMS strategies are shown in Fig. 2.12. A summary of studies done on EMS strategies is shown in Table 2.3. Also, the merits and demerits of EMS types are shown in Table 2.4 [26].
2.7.3 Overview of Energy Management Modeling and Solution Methods in NMCs As shown in Fig. 2.13a, several cases can be included in the EMS problem formulation in NMCs. Usually, in articles, the type of EMS strategy and selected control and items shown in Fig. 2.13a, are added to the problem formulation based on the studied case. Then, based on the type of created problem, and according to the objective functions and constraints are shown in Fig. 2.13b and c, the appropriate solution is chosen. The methods and solutions used for problems with MGs are shown in Fig. 2.13d.
2 Operation Management of Microgrid Clusters
Centralized
Central EMS
41
External EMS
Decentralized
Main Grid
Central EMS
External EMS
Hybrid EMS
Fig. 2.12 Various EMS strategies
2.8 Objective Functions Formulation In the following, one example of a simple formulation of objectivefunctions and constraints in the EMS problem in NMCs is shown. In this formulation, a model is presented for the operation of NMCs that are connected to each other and fed loads. In this model, it is possible to communicate several MGs continuously, and one MG can buy a lack of its power from other MGs or sell its extra power to other MGs. This will improve the reliability of the MGs and their better stability. On the other hand, the addition of renewable units, which have uncertain and imbalance nature, cause problems such as load mismatch and voltage instability in using of MGs. Therefore, the simultaneous use of heat and renewable units together contributes greatly to reducing these problems. Therefore, units such as energy storage systems and batteries can overcome this uncertainty in scheduling. But due to the high level of uncertainty caused by wind and solar units, the reservation issue is also considered in situations where the forecasting power generation of these units does not match
Decentralized
Centralized with sequential operation
EMS type Centralized
Multi-agent system with contract net protocol
1. Competitive environment among suppliers 2. Islanded mode only
Cost minimization
Minimization of load shedding
Sequential quadratic programming
Cost minimization
Mixed integer linear programming (MILP)
System reliability maximization
1. Distribution of computational load 2. Grid-connected mode only
Imperialistic competitive algorithm
Cost minimization and reliability maximization
Cost minimization
Optimization algorithm Particle swarm optimization
Optimization objective Cost minimization
Table 2.3 Summary of studies on EMS strategies
1. Load-generation uncertainties 2. Islanded and grid-connected mode 1. Unbalanced systems with detailed modeling 2. Islanded mode only 1. Failure to preserve customer privacy of microgrids 2. Requirement of extensive communication setup 1. Three-level control and demand bidding 2. Grid-connected mode only 1. Autonomous operation and privacy of customer 2. Grid-connected mode only
Major considerations 1. Load and renewable power uncertainties 2. Only grid-connected mode
(continued)
1. Unawareness of system-level resources 2. Due to individual objectives, equilibrium may exist and further optimization may not be possible
Limitations 1. Increase in the computational burden of central EMS 2. Failure to preserve customer privacy of microgrids 3. Requirement of extensive communication setup
42 M. Moradi and A. Akbari Foroud
Nested hybrid
Hybrid
EMS type
Cost minimization Privacy maximization
Cost minimization
Cost minimization and reliability maximization
Cost minimization
MILP
Average consensus algorithm
Supply adequacy maximization
Cost minimization and reliability maximization
Optimization algorithm Stochastic optimization
Optimization objective Cost minimization and reliability maximization
Table 2.3 (continued)
1. Control of voltage, frequency, and power 2. Islanded and grid-connected modes 1. Hybrid microgrids and diversity gain 2. Islanded and grid-connected modes 1. Load smoothening and cooperative operation 2. Islanded mode only 1. Hybrid MGs 2. Island, grid-connected and NMCs 3. Trade-off between various contradictory objectives such as operation cost, privacy of the customer, and network resiliency
Major considerations 1. Two-layered stochastic model and uncertainties 2. Islanded and grid-connected modes 1. Autonomous operation and self-healing . Islanded and grid-connected modes 1. Three-level control and dynamic conditions 2. Islanded and grid-connected modes
1. Upper operation cost vs. centralized EMSs 2. Lesser flexibility
1. Parallel operation of microgrids 2. Single-level privacy of customers, easy to reveal . Reduction in the resiliency of disconnected microgrids 4. Central EMS failure results in autonomous operation
Limitations
2 Operation Management of Microgrid Clusters 43
44
M. Moradi and A. Akbari Foroud
Table 2.4 Merits and demerits of EMS types EMS type Centralized
Merits 1. Minimized operation cost with global optimization. 2. Efficient components of each MG is used. 3. Standardize and implement are easy. 4. External trading is reduced. 5. Higher reliability in islanded mode.
Decentralized 1. Preserve the privacy of the customer. 2. Ensures plug-and-play flexibility. 3. Distribution of computational load. 4. Influence of forecasting errors on the MG level only. Hybrid 1. Preserve single level the privacy of the customer. 2. Higher flexibility vs. centralized EMSs. 3. Ensure plug-and-play flexibility. 4. Distribution of computational load. 5. Lesser operation cost vs. decentralized EMSs.
Demerits 1. Extensive communication infrastructure with powerful central EMS is needed. 2. Weak plug-and-play capability. 3. Without ensuring preserve the privacy of customer 4. After a small change in the system onerous testing is needed. 5. Lower flexibility, adaptability, and propagation of forecasting errors 1. Increased operation cost 2. The resiliency in islanded mode is decreased; 3. Excessive power trading with utility grid in grid-connected mode 1. Parallel operation of MGs, unaware of other MGs. 2. Easy to disclose the consumers’ privacy due to single-level privacy. 3. Resilient performance of disconnected MGs is decreased. 4. If central EMS is compromised, all MGs will operate in an autonomously decentralized mode.
with the actual value. The scheduling problem can be considered with the set of objectives shown in Fig. 2.13b and optimized by one of the methods shown in Fig. 2.13d. For example, objective functions can be considered to minimize operating costs connected to grid and island mode, minimize the costs due to environmental pollutant emissions, and optimally exchange of power between MGs and with the distribution network. It is noteworthy that the formulation and solution method vary depending on the selected objective function and different constraints. In this formulation, we can consider the uncertainties of wind and solar power generation units and loads [30]. MinOF =
Cost t,m + Emissiont,m
∀t ∈ T , ∀m ∈ {MG1 , . . . , MGm }
t
(2.1)
Privacy of MGs
Maximized
ESS Modeling
Reliability
(b)
ESS Cost
Main objectives in EMS
(a)
Demand Side Management Modeling
Emission Cost
Energy Market Modeling
Load Shedding
Operation Cost
Modeling of power exchange between MGs
Minimized
Generation Units Modeling
Reserve Cost
Modeling of power exchange betwwen MGs & DS
Fig. 2.13 Overview of EMS modeling and solution methods in NMCs. (a) EMS problem formulation in NMCs, (b) EMS objective functions in NMCs, (c) EMS constraints in NMCs, and (d) EMS solution method in NMCs
Resiliency
Load Modeling
Formulation for EMS in NMGs
2 Operation Management of Microgrid Clusters 45
Game Theory
Markov Decision
Two-level Optimization
Chance Constraint
ADMM
Benders
Fig. 2.13 (continued)
Artificial Intelligence
Demand
Stochastic & Robust Programming
Power Balance
Decomposition Methods
Reserve
(c)
Storage
(d)
Model Predictive Control
Multi-Agent System Programming
Carbon Emission
NMGs EMS Solution Method
Meta-Heuristic Methods
Operation
Constraints in NMGs EMS
Heuristic Methods
Max & Min generation unit
ScenarioBased
Priority-based
Dynamic Programming
Price
MIQP
MINLP
MILP
Linear and NonLinear Programming
Self efficiency
46 M. Moradi and A. Akbari Foroud
2 Operation Management of Microgrid Clusters
47
2.8.1 Cost Operation Modeling The operation cost of the MGs includes the fuel cost of the generating units, the cost of energy storage units, and the cost of power exchanged between the MGs and the main grid. This function is formulated in Eq. (2.2): T T ⎧Ng,m ⎨
F Pi,m = PGi,m (t)BGi,m (t) Cost t,m = ⎩ m t=1
t=1
i=1
⎫ ⎬
PSj,m (t)BSj,m (t) +PGrid,m (t)BGrid,m (t) + ⎭ Ns,m j =1
∀m ∈ {MG1 , . . . , MGm } (2.2) where F(Pi ) is the total operating cost (in term of $), Trepresents the total studied hours,Ng ,NS represent the number of generating and storing units, respectively. PGi (t) andPSj (t)are output power of ith generation unit and ith storage unit in the time, respectively.BGi (t)and BSj (t)are the proposed energy price of ith generation unit and ith storage unit in the time, respectively. PGrid (t) and BGrid (t) are the amount of exchanging power with the main grid and its proposed price at the time t and the index m represents the desired MG.
2.8.2 Pollution Emission Modeling Pollution emission level is proportional to the output power of units. The formulation of this function is shown in Eq. (2.3): ⎧ g,m T T ⎨N
C Pi,m = Emissiont,m = PGi,m (t)EGi,m (t) ⎩ t=1
t=1
i=1
⎫ ⎬
PSj,m (t)ESj,m (t) + PGrid,m (t)EGrid,m (t) + ⎭ Ns,m j =1
∀m ∈ {MG1 , . . . , MGm } (2.3)
48
M. Moradi and A. Akbari Foroud
In Eq. (2.3), EGi (t) ESj (t) and EGrid (t) indicate the amount of pollution caused by the ith generation unit, jth storage, and the global network in time t, respectively, Kg . and they are in terms of MW h
2.8.3 Problem Constraints 2.8.3.1
Power Balance Constraint
In the studied MGs, the sum of the power of generation units, storages, and the main distribution network should be equal to the required load of the network. m
⎛ ⎝
Ng,m
i=1
Ns,m
PGi,m (t) +
⎞ PSj,m (t)⎠ + PGrid (t) = PL (t)
(2.4)
i=1
where p is the load demanded by the system at time t. (In the case of load response programs or shedding of the load, this limitation is violated. However, the penalty coefficient can be considered a penalty of this limitation).
2.8.3.2
Generation Capacity Limit
All generation units in the MG, storages, and distribution networks have their own generation power limit. These limitations are described in Eq. (2.5). PGi,min (t) ≤ PGi (t) ≤ PGi,max (t) PSj,min (t) ≤ PSj (t) ≤ PSj,max (t) PGrid,min (t) ≤ PGrid (t) ≤ PGrid,max (t)
2.8.3.3
(2.5)
Charge and Discharge Rate Limit
Batteries as a storage system have limitations such as charging and discharging rate and stored energy that is given in Eq. (2.6). SoCsj (t) = SoC sj (t − 1) + Pchg/Dchg (t) 0 ≤ Pchg/Dchgj (t) ≤ PCDSj ,max
(2.6)
where SoCsj (t) and SoCsj (t − 1) denote the amount of charge storage at the present time and the hour ago. Pchg/Dchg (t)displays the amount of charge during tth hour and PCDSj ,max denotes the maximum charge rate.
2 Operation Management of Microgrid Clusters
2.8.3.4
49
Self-Sufficiency Constraint
NMCs, under certain conditions, may be disconnected from the distribution network and operated as an island. Under these conditions, operator decisions and planning must be such that the distribution generation resources can feed the loads on the MG completely without interruption. In the proposed model, self-sufficiency is added to the optimization operation as an index and it has been discussed. In the following equation, the influence of MG self-sufficiency index on problem formulation is expressed mathematically in Eq. (2.7). max PM,t = (1 − SSI)
Dd,t
(2.7)
d max is the capacity of MG lines for In Eq. (2.7), SSI is self-sufficiency index, PM,t power transmission and Dd, t is the amount of power demand associated with the load d at time t. It is found in this equation that if the SI index is greater, the capacity of the lines will be lower and MG can receive less power from the distribution network. As a result, it is more self-sufficient.
2.8.3.5
Reserve Constraint
In MGs, both loads and distributed generation resources have uncertainty. In a comprehensive (general) model for 24-hour scheduling of MG, some power must be considered as a reservation to overcome these faults. A probable index called PSS is considered for reservation modeling. This index indicates the probable capability of MG to feed existing loads. To introduce the PSS index, it is assumed that the prediction fault of the load consumption power and the generation power of the scattered generation resources variables. On the are modeled asnormal distributed 2 andΔ ∼ N μ , σ 2 . Where, μ other hand, Δd ∼ N μed , σed w ew ew ew and μed are the average generation power of wind units and the consumption power of loads, 2 and σ 2 are the standard deviation of the generation power of wind respectively. σew ed units and the consumption power of loads, respectively. N represents the normal distribution of a probable variable. However, the required reservation of MG to overcome the prediction error can be added to problem formulation as one of the optimization constraints. The mathematical form of these explanations is shown in the Eq. (2.8). i
Pi,t ≥
f Di,t −pwt
−1 2 2 − (μew − μed ) + 2 σew +σed × erf (1−2P SS)
d
(2.8)
50
M. Moradi and A. Akbari Foroud
In Eq. (2.8), Pi, t is the output power of ith unit at time t, Di, t is the predicted f consumption power in MG, pwt is the predicted generation power at time t, erf is the error function and PSS is a probable index used in reservation modeling.
2.8.4 Demand Response Pogromming (DRP) Economic models that are presented for the DR are trying to measure demand from parameters that are important from a marketing point of view [34]. This is based on microeconomic theories. Economic models are derived from data gained from actual experience and are used to evaluate DRPs. The elasticity of the demand is defined as the demand sensitivity to the price: E=
ρ0 ∂d new × do ∂ρ new
(2.9)
where, ρ 0 is an initial energy price regardless of DRP and do is the initial demand for energy before considering and applying DRP. Also, dnew is the amount of demand after applying DRP and ρ new is the energy price after applying the charge. According to Eq. (2.9), this demand elasticity proportion to price can be divided into two parts: 1. The sensitivity of a consumer’s demand at a given time proportion to consumer energy price changes at the same time elasticity. 2. The sensitivity of a consumer’s demand at a given time proportion to energy price change at the other times (mutual elasticity). The self-elasticity and mutual elasticity can be written mathematically in the form of Eq. (2.10) and Eq. (2.11), respectively. Self and mutual elasticity are also considered to be −0.2 and 0.01, respectively. E (i, i) =
E (i, j ) =
ρ0 (i) ∂d new (i) × do (i) ∂ρ new (i)
(2.10)
ρ0 (j ) ∂d new (i) × do (i) ∂ρ new (j )
(2.11)
It can be noted that self-elasticity and mutual elasticity have negative and positive values, respectively.
E (i, j ) ≤ 0 E (i, j ) ≥ 0
i = j, ∀i, j ∈ T i = j, ∀i, j ∈ T
(2.12)
2 Operation Management of Microgrid Clusters
51
According to the demand response relative to the price change, two models can be presented: • One-time model. • Multi-time model. In a one-time model, the loads are only sensitive to the price changes at the same time and so, do not have the ability to transfer to another time. In this model, whenever the price of energy changes, the loads react to it. Therefore, only selfelasticity is involved in this model. The economic one-time model will be as follows: ρnew (i) − ρ0 (i) dnew (i) = d0 (i) × 1 + E (i, i) ∀i ∈ T ρ0 (i)
(2.13)
In the multi-time model, the mutual elasticity which is mentioned previously is used. According to this model, the demand changes at a special hour depends not only on the price change at that time but also on price change at the other hours. If we want to see the effect of encouragement and penalty on determining the level of demand we have: d(i) = d0 (i) +
24
i=1 i = j
E (i, j )
d0 (i) [ρ(j ) − ρnew (i)] ρ0 (i)
(2.14)
By combining the model of one time and multi times, a comprehensive economic model of the load is obtained [34]. ⎧ ⎫ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ 24 ⎨ ⎬ ρnew (i)−ρ0 (i) 1 d (i)= d0 (i) 1+E (i, i) + E (i, j ) [ρ(j )−ρnew (i)] ⎪ ⎪ ρ0 (i) ρ0 (i) ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ i=1 ⎪ ⎪ ⎪ ⎪ ⎩ ⎭ i = j (2.15) The Eq. (2.15) shows that a consumer needs to adjust his consumption at any time of the day to get the most profit.
2.8.5 Modeling of Generation Units in the MGs In this section, all units in the MGs are modeled mathematically. These units include fuel cell units, micro-turbines, and CHP units.
52
2.8.5.1
M. Moradi and A. Akbari Foroud
Fuel Cell Model
The cost of fuel for each fuel cell is calculated as a function of its efficiency. The fuel cell efficiency is related to its operating point and is referenced in terms of output power ratio to the input energy capacity in natural gas, which must be in the same unit (W). Fuel cost for each fuel cells is defined as follows: CF C = Cnl
PJ J
ηJ
(2.16)
where Cnl is the price of natural gas for feeding fuel cell, PJ is the active power generated at the time interval J, ηJ is the efficiency of the cell at the time interval J and CFC is the cost of fuel consumption of fuel cells.
2.8.5.2
Micro-Turbines Model
Unlike fuel cells, the efficiency of the micro-turbine increases with increasing operating power. The overall and electrical efficiency of the micro-turbine is expressed as: ηtot =
PJ + Qdh Qf uel
(2.17)
PJ Qf uel
(2.18)
ηJ =
where, PJ is pure generated electric power, Qdh is the generated heat of each units and Qfuel is input fuel energy. According to Eq. (2.19), the cost of MTs fuel is calculated as follows: CMT = Cnl
PJ ηlJ
(2.19)
J
where PJ is pure generated electric power at the time interval J, Cnl is the cost of natural gas for feeding the MT and CMT is the cost of fuel consumption of MT.
2.8.5.3
Combined Heat and Power Model
The large-scale old generators have an efficiency of about 35%. While combined heat and power (CHP) units are able to increase the efficiency of micro-turbines to 80–85%. Without considering CHP units, MTs will have less efficiency than
2 Operation Management of Microgrid Clusters
53
traditional power generation, and the operation of MGs is not economically feasible. The fuel cost of CHPs is as follows: t t rec ηCH P − ηe CCH P = CMT × 1 − (2.20) ηb where, rec is the heat recovery factor in CHP and is assumed 0.95. Also, ηb and ηet are assumed 0.8 and 0.35, respectively.
2.9 Numerical Result In this section, three scenarios are assumed. To calculate the costs and emission, all of the required parameters are shown in Table 2.5 and Table 2.6. Also, the amount of loads and the TOU method pertinent prices are given in Table 2.7. As mentioned earlier, NMCs should be able to perform well in different operational conditions, such as island mode. For this purpose, three scenarios with different SSI indicators have been considered. The transfer power limit between MGs and the upstream network is assumed 300 kW. In general, as the SSI increases, the amount of transmission power purchased between the MGs and the upstream network decreases. The reason for the increase in the cost of operation is that considering the SSI, the capacity of the lines between the MGs is limited and as a result, the volume of energy exchanges between the MGs is more limited. Thus, an MG has to use more expensive units to meet its power deficit, and therefore increase operating costs. Also, the DGs of each MG must generate extra power to prevent shedding of the load. Figures 2.14, 2.15 and 2.16 (includes a, b, and c) show the effect of increasing the SSI on unit commitment and battery charge/discharge status per MG (includes 1, 2, and 3). The comparison of energy scheduling costs in different scenarios is shown in Table 2.8. As shown in Table 2.8, with the increase of SSI and the decrease in the transfer power purchased from the upstream network, the cost of reservation has also increased. As the output power of DGs increases, the uncertainty of the system, and the number of required reservations to overcome the prediction error will be increased. To demonstrate the effect of the SSI on unit scheduling, the generation of WT and PV units are obtained in the first scenario and are not changed for the rest of the scenarios. As the price of power changes throughout the day, the amount of demand reacts to price changes per hour. Also, by discharging the batteries during peak hours, the purchase of additional energy is prevented. In fact, batteries are very effective to reduce the dependence of MGs on the upstream network and improve the reliability and resilience of the network in the critical event.
54
M. Moradi and A. Akbari Foroud
Table 2.5 Required parameters to calculate the operation costs of MGs MG No. Unit type Operation and maintenance cost (λOM ) Efficiency (η) Min power generation (kw) Max power generation (kw) Sold power costs ($/kwh) Purchased power costs ($/kwh)
MG 1 MG 2 MG 3 CHP PV WT FC MT WT CHP PV FC 0.088 0.1095 0.1095 0.016 0.088 0.1095 0.088 0.1095 0.016
0.85 50
– 0
– 0
0.6 30
0.8 50
– 0
0.85 50
– 0
0.6 30
500
150
250
250
500
250
500
150
250
d1 = 0.14
d2 = 0.13
d3 = 0.12
C1 = 0.18
C2 = 0.17
C3 = 0.17
Table 2.6 Required parameters to calculate emissions costs kg kg kg $ Gas Type γ kg ρMT kwh ρCH P kwh ρF C kwh NOx SO2 CO2
10.0714 2.3747 0.0336
0.0001 0.000007 0.00137
Natural gas low-hot value: L = 9.7
0.00003 0.000006 0.001078 KW h m3
0.00044 0.000008 0.001596
ρP V
0 0 0
, natural gas price: Cnl = 0.76
kg kwh
$ m3
ρW T
kg kwh
0 0 0
Table 2.7 Mean value of MGs load and consumption prices based on TOU Hours 1 2 3 4 5 6 7 8 9 10 11 12
Load (Kw) MG1 MG2 100.36 77.87 87.13 64.17 83.55 53.42 105.25 74.16 73.5 55.15 72.16 57.67 137.35 94.58 171.8 133.41 231.08 172.55 266.28 218.85 148.19 128.19 164.7 125.97
MG3 111.08 108.31 90.09 134.2 100.48 93.78 148.98 189.42 319.1 342.73 215.65 214.86
Price ($/KWh) 0.023 0.023 0.023 0.023 0.023 0.023 0.023 0.034 0.04 0.04 0.034 0.034
Hours 13 14 15 16 17 18 19 20 21 22 23 24
Load (Kw) MG1 MG2 143.89 144.77 174.59 117.35 146.8 113.31 172.82 159.38 258.05 252.04 318.7 264.33 367.4 268.67 372.66 222.92 294.51 251.99 238.72 223.91 214.65 159.97 135.92 137.86
MG3 246 204.46 190.61 207.45 402.02 375.03 509.36 373.83 381.99 340.36 276.6 223.7
Price ($/KWh) 0.034 0.034 0.034 0.034 0.04 0.04 0.04 0.04 0.04 0.04 0.04 0.023
2.10 Conclusion The NMCs are cost-effective and beneficial for both owners of resource and energy consumers. In this chapter, the 24-hour scheduling of the NMCs, consisting of several DGs and loads has been discussed. This programming is an optimization
2 Operation Management of Microgrid Clusters
55 MG1
400
200
350
150
300
Power (KW)
Power(kw) in MG1
250
100 50
Demand Generated Power
250 200 150
0
100 CHP
Wind
Battery
PV
50
-50 0
5
10
Hour
15
20
25
0
5
10
Hour
15
20
25
(a) 70
MG2
300
Demand Generated
50
250
40
Power (KW)
Power(kw) in MG2
60
30 20 10 0
150 100
-10 -20 -30
FC
Wind
0
Battery
MT
5
10
Hour
50 15
20
Power (KW)
200 150 100 50 0 FC
CHP
0
Battery
PV
5
10
Hour
5
10
(b)
250
-50
0
25
300
Power(kw) in MG3
200
15
20
25
550 500 450 400 350 300 250 200 150 100 50
Hour
15
20
25
MG3 Demand Generated
0
5
10
15
20
25
Hour
(c)
Fig. 2.14 Units generation scheduling with considering of DRPs and SSI = 0 in (a) MG1 (b) MG2 (c) MG3
350
150
300
100 50
200 150 100
CHP
Wind
PV
Battery
50 0
5
10
Hour
15
20
0
25
10
Hour
15
20
Demand Generated
50 250
40
Power (KW)
30 20 10 0 -10
200
150
100
-20 -30
25
MG2
300
60
Power(kw) in MG2
5
(a)
70
FC
Wind
0
MT
Battery
5
10
Hour
15
20
25
50
(b)
0
300
550
250
500
5
10
Hour
15
20
25
MG3 Demand Generated
450 200
Power (KW)
Power(kw) in MG3
Demand Generated
250
0 -50
MG1
400
200
Power (KW)
Power(kw) in MG1
250
150 100 50
350 300 250 200 150
0 FC
-50
400
0
CHP
PV
5
100
Battery
10
Hour
15
20
25
50
(c)
0
5
10
Hour
15
20
25
Fig. 2.15 Units generation scheduling in with considering of DRPs and SSI = 0.5 in (a) MG1 (b) MG2 (c) MG3
M. Moradi and A. Akbari Foroud 250
400
200
350
MG1 Demand Generated
300
150 Power (KW)
Power(kw) in MG1
56
100 50
250 200 150
0
100 CHP
-50
Wind
0
PV
battery
5
10
Hour
15
20
25
50
(a)
0
5
10
Hour
15
20
25
MG2
80
300 250
40
Power (KW)
Power(kw) in MG2
Demand Generated
60
20 0
200 150 100
-20 FC
-40
Wind
0
MT
battery
5
10
Hour
15
20
25
50
(b)
300
5
10
Hour
15
20
25
MG3 Demand Generated
500
250
450 200
Power (KW)
Power(kw) in MG3
0
550
150 100
400 350 300 250 200
50
150 0
100 FC
CHP
PV
battery
50
-50 0
5
10
15 Hour
20
25
(c)
0
5
10
15
20
25
Hour
Fig. 2.16 Units generation scheduling with considering of DRPs and SSI = 0.7 in (a) MG1 (b) MG2 (c) MG3
problem that contains two objective functions, operating costs, and the emission cost. Also, SSI, reservation constraint, and active and reactivepower exchange have been considered in the optimization problem. In this programming, a structure for the operation of NMCs is described, that structure provides the possibility of exchanging power between each MG, and MGs with the upstream distribution network. The uncertainty of the wind generation and solar units is investigated using the probabilistic distribution function of the Monte Carlo method. To overcome the difference caused by uncertainty, the required reserve is generated and the cost of it is added to the objective function. Also, the DRP based on time of use (TOU) is employed due to that customers tend to reduce some of their power consumption during peak hours and transfer some of their power consumption to off-peak hours. On the other hand, the SSI has been added to the scheduling constraint to increase the reliability of the MGs, especially in islanding mode. Also, the limitations of the operating systems are considered in all operating conditions based on the uncertainty of units and DGs and load variation (in terms of quantity, power factor, etc.). In addition, the ability to exchange power between any MG with the others is possible. In the networked mode, a local EMS located on each of the MGs optimizes the DERs in the MGs according to the information of other MGs and it can generate excessive power requirements. So, it can compensate for the power shortages of the other MGs. This is done in coordination with the distribution network operator and the external EMS. It is necessary to note that the priority of any MGs is to balance the power and economic operation to manage its resources. In
Scenario No. 1 2 3
SSI 0 0.5 0.7
Reserve cost ($) 379.681 390.037 394.073
Operation cost ($) 1155.794 1174.379 1225.997
Emission cost ($) 33.772 32.338 30.184
Table 2.8 Comparison of energy scheduling costsin different scenarios Purchased Power cost ($) MG1 MG2 MG3 −27.36 −45.99 −108.43 −32.64 −45.60 −97.60 −35.91 −44.22 −128.08
Sold power cost ($) MG1 MG2 318.970 70.094 267.245 69.648 217.616 70.255
MG3 229.791 198.245 163.304
Total cost ($) 1132.181 1237.468 1407.305
2 Operation Management of Microgrid Clusters 57
58
M. Moradi and A. Akbari Foroud
the next step, if the MG cannot be able to feed its demands, it should buy its energy shortage from the other MGs. In order to examine the uncertainty of loads, normal distribution function with an average value obtained from DRP, and the standard deviation value equals to 0.3 are considered. The simulation results are validated by MATLAB software and with the PSO algorithm. The simulation results show that NMCs can prevent load shedding by reducing the power dependence by exchanging power between MGs. This will increase resilience and reliability in NMCs.
References 1. Alam, M. N., Chakrabarti, S., & Ghosh, A. (2019). Networked microgrids: State-of-the-art and future perspectives. IEEE Transactions on Industrial Informatics, 15(3), 1238–1250. 2. Tang, C., Liu, M., Dai, Y., Wang, Z., & Xie, M. (2019). Decentralized saddle-point dynamics solution for optimal power flow of distribution systems with multi-microgrids. Applied Energy, 252. 3. Liu, G., Starke, M. R., Ollis, B., & Xue, Y. (2016). Networked Microgrids Scoping Study. Oak Ridge National Laboratory (ORNL/TM-2016/294). 4. Backhaus, S., Dobriansky, L., et al. (2016). Networked microgrids scoping study. Los Alamos National Laboratory. 5. Hussain, A., Bui, V., & Kim, H. (2019). Resilience-oriented optimal operation of networked hybrid microgrids. IEEE Transactions on Smart Grid, 10(1), 204–215. 6. Zamora, R., & Srivastava, A. K. (2018). Multi-layer architecture for voltage and frequency control in networked microgrids. IEEE Transactions on Smart Grid, 9(3), 2076–2085. 7. Wang, H., & Huang, J. (July). Incentivizing energy trading for interconnected microgrids. IEEE Transactions on Smart Grid, 9(4), 2647–2657. 8. Jadhav, A. M., & Patne, N. R. (2017). Priority-based energy scheduling in a smart distributed network with multiple microgrids. IEEE Transactions on Industrial Informatics, 13(6), 3134– 3143. 9. Wang, Z., Chen, B., Wang, J., Begovic, M. M., & Chen, C. (2015). Coordinated energy Management of Networked Microgrids in distribution systems. IEEE Transactions on Smart Grid, 6(1), 45–53. 10. Wang, Z., Chen, B., Wang, J., & kim, J. (2016). Decentralized energy management system for networked microgrids in grid-connected and islanded modes. IEEE Transactions on Smart Grid, 7(2), 1097–1105. 11. Gao, H., Liu, J., Wang, L., & Wei, Z. (2018). Decentralized energy Management for Networked Microgrids in future distribution systems. IEEE Transactions on Power Systems, 33(4), 3599– 3610. 12. Ma, W., Wang, J., Gupta, V., & Chen, C. (2018). Distributed energy Management for Networked Microgrids Using Online ADMM with regret. IEEE Transactions on Smart Grid, 9(2), 847–856. 13. Yuan, W., Wang, J., Qiu, F., Chen, C., Kang, C., & Zeng, B. (2016). Robust optimization-based resilient distribution network planning against natural disasters. IEEE Transactions on Smart Grid, 7(6), 2817–2826. 14. Abhinav, S., Modares, H., Lewis, F. L., Ferrese, F., & Davoudi, A. (2018). Synchrony in networked microgrids under attacks. IEEE Transactions on Smart Grid, 9(6), 6731–6741. 15. Wang, Z., Chen, B., Wang, J., & Chen, C. (2016). Networked microgrids for self-healing power systems. IEEE Transactions on Smart Grid, 7(1), 310–319. 16. Li, Y., Zhang, P., & Luh, P. B. (2018). Formal analysis of networked microgrids dynamics. IEEE Transactions on Power Systems, 33(3), 3418–3427.
2 Operation Management of Microgrid Clusters
59
17. Hu, X., & Liu, T. (2017). Co-optimisation for distribution networks with multi-microgrids based on a two-stage optimisation model with dynamic electricity pricing. IET generation, Transmission & Distribution, 11(9), 2251–2259. 18. Arif, A., & Wang, Z. (2017). Networked microgrids for service restoration in resilient distribution systems. IET generation, Transmission & Distribution, 11(14), 3612–3619. 19. Shahnia, F., Chandrasena, R. P. S., Rajakaruna, S., & Ghosh, A. (2014). Primary control level of parallel distributed energy resources converters in system of multiple interconnected autonomous microgrids within self-healing networks. IET Generation, Transmission & Distribution, 8(2), 203–222. 20. Golsorkhi, M. S., Hill, D. J., & Karshenas, H. R. (2018). Distributed voltage control and power Management of Networked Microgrids. IEEE Journal of Emerging and Selected Topics in Power Electronics, 6(4), 1892–1902. 21. Wang, Z., & Wang, J. (2017). Service restoration based on AMI and networked MGs under extreme weather events. IET generation, Transmission & Distribution, 11(2), 401–408. 22. Zhang, F., Zhao, H., & Hong, M. (2015). Operation of networked microgrids in a distribution system. CSEE Journal of Power and Energy Systems, 1(4), 12–21. 23. Tian, P., Xiao, X., Wang, K., & Ding, R. (2016). A hierarchical energy management system based on hierarchical optimization for microgrid community economic operation. IEEE Transactions on Smart Grid, 7(5), 2230–2241. 24. Che, L., Shahidehpour, M., Alabdulwahab, A., & Al-Turki, Y. (2015). Hierarchical coordination of a community microgrid with AC and DC microgrids. IEEE Transactions on Smart Grid, 6(6), 3042–3051. 25. Chiu, W., Sun, H., & Poor, H. V. (2015). A multiobjective approach to Multimicrogrid system design. IEEE Transactions on Smart Grid, 6(5), 2263–2272. 26. Hussain, A., Bui, V., & Kim, H. (2018). A resilient and privacy-preserving energy management strategy for networked microgrids. IEEE Transactions on Smart Grid, 9(3), 2127–2139. 27. Cintuglu, M. H., & Mohammed, O. A. (2017). Behavior modeling and auction architecture of networked microgrids for frequency support. IEEE Transactions on Industrial Informatics, 13(4), 1772–1782. 28. Parisio, A., Wiezorek, C., Kyntaja, T., Elo, J., Strunz, K., & Johansson, K. H. (2017). Cooperative MPC-based energy management for networked microgrids. IEEE Transactions on Smart Grid, 8(6), 3066–3074. 29. Mojtahedzadeh, S., Ravadanegh, S. N., & Haghifam, M. (2017). Optimal multiple microgrids based forming of greenfield distribution network under uncertainty. IET Renewable Power Generation, 11(7), 1059–1068. 30. Nikmehr, N., Najafi-Ravadanegh, S., & Khodaei, A. (2017). Probabilistic optimal scheduling of networked microgrids considering time-based demand response programs under uncertainty. Applied Energy, 198, 267–269.15. 31. Bullich-Massague, E., Díaz-González, F., et al. (2018). Microgrid clustering architectures. Applied Energy, 212, 340–361. 32. Hu, M., Wang, Y.-W., Xiao, J.-W., & Lin, X. (2019). Multi-energy management with hierarchical distributed multi-scale strategy for pelagic islanded microgrid clusters. Energy, 185, 910–92115. 33. Antoniadou-Plytaria, K. E., Kouveliotis-Lysikatos, I. N., Georgilakis, P. S., & Hatziargyriou, N. D. (2017). Distributed and decentralized voltage control of smart distribution networks: Models, methods, and future research. IEEE Transactions on Smart Grid, 8(6), 2999–3008. 34. Ajoulabadi, A., Ravadanegh, S. N., & Ivatloo, B. M. (2020). Flexible scheduling of reconfigurable microgrid-based distribution networks considering demand response program. Energy, 196, 1.
Chapter 3
Energy Management Systems for Microgrids Seyed Mohsen Hashemi and Vahid Vahidinasab
3.1 Introduction Increasing use of the small-scale generation resources and energy storage in the distribution networks and close to the end users has led to several technical and economic advantages. Microgrids (MGs) provide a systematic approach for operating an energy system with these features. A strong Energy management system (EMS) enables the MG to monitor and control the resources in the time steps near the real operation time. The EMS of an MG can be introduced as an integrated software and hardware system that brings a decision support system for the MGs’ operator in which all of the tasks of monitoring, decision-making, and controls could be done using it. Furthermore, it determines how to exchange energy with the main grid considering the operational constraints and the system conditions such as load consumption, electricity market price, generation capability of the units, and the stored energy of the energy storage systems. In addition to the normal operation state, EMS should be attentive to the critical conditions of the contingencies. For this purpose, security constraints are considered in the mathematical model of the MG’s operation using the frequency-related constraints.
S. M. Hashemi · V. Vahidinasab () Department of Electrical Engineering, Abbaspour School of Engineering, Shahid Beheshti University, Tehran, Iran e-mail: [email protected]; [email protected] © Springer Nature Switzerland AG 2021 A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems, https://doi.org/10.1007/978-3-030-59750-4_3
61
62
S. M. Hashemi and V. Vahidinasab
3.2 An Overview of the EMS of the MGs The EMS of an MG enables different functionalities such as: • • • • •
Monitoring of the MG in different conditions. Analyzing the system’s condition in different operational states. Ability to deal with various threats. Making quick decisions in critical situations. Performing control actions.
Different input signals are provided for an EMS and are analyzed through its computational core. The output results or decisions can be used either for the assessment of system conditions through the graphical reports or be automatically sent to the different agents as the dispatching orders. EMS has a critical function in real-time control, where the input data has high variability and, in such conditions, it would be impossible for manual control to handle the system. As indicated in Fig. 3.1, an MG has multiple interactions with different agents that should be handled by the EMS. As previously mentioned, EMS has three tasks of monitoring, decision making, and control that are indicated in Fig. 3.2. They will be analyzed in the following subsections.
Fig. 3.1 Structure of a typical MG [1]
3 Energy Management Systems for Microgrids
63
EMS Decision making system Security assessment
Remedial action system
State estimation
Load forecast/ estimation
Load flow
Short circuit calculation
Uncertainty assessment
Optimization tools Unit commitment
Economic dispatch
Optimal power flow
Decision making under uncertainty
Network reconfiguration
Control system
Monitoring system
Assessment tools Network/component modeler
Restoration tools FLISR
MG Fig. 3.2 General structure of EMS in MG
3.3 Monitoring System The monitoring system of EMS provides a detailed and comprehensive picture of the microgrid at any time. This information includes the status and performance of switches, lines, and transformers. As well as the output power of DGs, the output power of storages and their state of charge (SOC), available fuel of the Microturbines (MT), and the information of the reactors and capacitors of the network. This information is gathered by the meters and sensors and sent to EMS by the remote terminal units (RTU) through the communication infrastructure.
3.4 Control System EMS controls the MG in different manners. They are determined considering the complexity and importance of the MG’s procedures. In some cases, the automatic control is applied in which EMS directly controls the components. For example, in the MGs equipped with the self-healing strategies, locating and isolating the fault and restoring the system are automatically performed. Another control scheme is
64
S. M. Hashemi and V. Vahidinasab
the supervisory control in which the system operator can directly open or close the switches by the EMS if needed. In cases where there is no remote access to the components, the manual control is used in which the system operator calls the plant operator to close or open the switches.
3.5 Decision-Making System Decision-making can be considered as the most valuable feature of the EMS. It enables the EMS with the capabilities of the assessment, optimization, and restoration of the MG. They will be analyzed in the following subsections.
3.5.1 Assessment Tools Assessment tools enable the operator to find out about the MGs condition in the current operation regime or any possible operation scenario. These tools are presented in the following.
3.5.1.1
Network and Component Modeler
An important tool including the mathematical model of the system components’ behaviors. With an extension of the MG by the addition of the new components, it provides the capability of mathematical modeling.
3.5.1.2
Security Assessment Tool
This tool empowers the system operator to assess the system condition in different contingencies, in order to identify the most critical contingencies. According to the low inertia of MGs, this tool considers the dynamic aspects of the contingencies such as the frequency deviation. Sudden islanding is usually considered as the contingency of the MGs. Also, in the isolated MGs, EMS considers events such as the resource outage that affect the dynamic variables of the system.
3.5.1.3
Remedial Action Scheduling Tool
Using this tool, the system operator can deal with the system’s contingencies. It provides solutions such as system reconfiguration, generation unit startup, and so on. Some of the demand response programs such as the emergency demand response
3 Energy Management Systems for Microgrids
65
program (EDRP) are valuable resources to provide remedial actions. The response rate of the mentioned resources is an important factor in MGs.
3.5.1.4
State Estimation
A useful tool to estimate the state variables of the system by the limited number of metered data.
3.5.1.5
Load Forecast/Estimation
This tool uses the current load consumption, the historical data, and the forecasted weather condition to determine the load consumption in the coming hours or minutes. Also, it approximately calculates the load of the downstream buses of the metered transformers. This capability is based on the load pattern of the different load groups.
3.5.1.6
Load Flow
Using this tool, the system operator will be able to calculate the line flows and bus voltage of the network in different configurations or load levels.
3.5.1.7
Short Circuit Calculation
This tool calculates the short circuit current of the circuit breakers and switches and compares them with their allowed values.
3.5.1.8
Uncertainty Assessment
This tool studies uncertain parameters such as renewable power generation and proposes an uncertainty model for them. For example, in the form of some of the scenarios and their probabilities. Moreover, it provides scenario generation and scenario reduction tools.
3.5.2 Optimization Tools Decision-making in MGs is a complex process that requires a systematic approach as there are different resources and strategies to supply the load. Depending on the operation strategy and considering different economic and technical aspects, the
66
S. M. Hashemi and V. Vahidinasab
MG’s operator requires an optimization tool that provides different applications as will be expressed in the following subsections.
3.5.2.1
Unit Commitment
An optimization model to determine the online generation units in the operation horizon. Usually, it considers the economic objectives and includes the technical properties of different generation units and storages.
3.5.2.2
Economic Dispatch
This optimization module determines the optimal output power of the online resources using the linear models that can be solved very fast.
3.5.2.3
Optimal Power Flow
An optimization tool with different technical or economic objectives including the network constraints. Loss minimization, cost minimization, and profit maximization can be different objectives of the OPF module.
3.5.2.4
Network Reconfiguration Tool
This tool determines the best supplying path of the loads in different viewpoints. Although the network is designed in a mesh structure this module opens or closes the switches to achieve a radial structure with low power loss or high resiliency. The other objectives can be considered for reconfiguration such as sharing the load between different lines.
3.5.2.5
Decision Making under Uncertainty
This tool is able to provide different decision-making methods considering uncertainty. The system operator applies each of these methods based on the uncertainty model of the parameters. This module includes different methods of stochastic programming, robust optimization, information gap decision theory, and so on.
3 Energy Management Systems for Microgrids
67
3.5.3 Restoration Tools This tool empowers the system operator to deal with system faults. Locating and isolation of the network faults are two important tasks that can be performed in the automatic or manual manners. As well as, the interrupted loads should be restored fast. In the MGs equipped with the smart meters and switches, the process of finding the fault location, isolation, and system restoration (FLISR) is automatically performed.
3.6 Interaction with Other Systems EMS interacts with the other systems in order to perform its mentioned functionalities. These systems are introduced in the following (see Fig. 3.3).
3.6.1 Distribution Management System (DMS) This system similarly manages the main distribution grid to the EMS. Information such as the scheduled outage of the microgrid’s connecting feeder can be received from DMS. In this condition, EMS should prepare the MG for the islanded mode operation.
Distribution management system
Outage management system
EMS Advanced metering infrastructure
Monitoring
Bid/offer interface
Decision making
Electricity market
Fig. 3.3 Interaction of EMS with other systems
Control
Weather forecasting system
Maintenance scheduling system
68
S. M. Hashemi and V. Vahidinasab
3.6.2 Advanced Metering Infrastructure (AMI) AMI interacts with EMS by providing different facilities such as smart meters, two-way communication infrastructure, and other data gathering and transmission facilities to interact with EMS in real-time operation.
3.6.3 Outage Management System (OMS) In OMS, data of the components’ outage and load interruption are collected and analyzed based on different data resources such as the customer information system (CIS). The outage’s information such as fault location, duration of the load interruption, and the number of the affected customers are determined in OMS.
3.6.4 Maintenance Scheduling System This system schedules the preventive maintenance of the lines, transformers, generation resources, and the storages, leading to their planned outages. EMS considers these schedules in its scheduling procedure.
3.6.5 Weather Forecasting System This system forecasts the data such as environmental temperature and wind speed that are used in EMS to determine the load consumption and the power generation of renewable resources in the coming hours or minutes.
3.6.6 Electricity Market EMS can participate in the electricity market as a buyer or seller to determine its energy exchange with the main grid. Additionally, EMS can be a player of the ancillary services market by providing the reserve capacity.
3 Energy Management Systems for Microgrids
69
3.6.7 Bid/Offer Interface Using this system, generation resources and consumers can declare the supply and demand curves indicating their cost function or price sensitivity. Although different aspects of the EMS were previously analyzed, the next sections will focus on EMS functionality in short-term scheduling of MG, in which EMS dispatches/commits the existing resources for the next minutes/hours to supply the load in the most economic manner. Hereinafter, we mean EMS a decisionmaking module that considers different technical and economic aspects to schedule the resources.
3.7 Centralized and Decentralized Energy Management Generally, an EMS can be implemented in centralized and decentralized ways. Different local controllers are used in both of these methods. In a centralized manner, a computational core sends the dispatching commands to the local controllers to operate different components. In return, in a decentralized structure, the local controllers are intelligent agents analyzing the conditions to dispatch their associated components. The centralized manner applies optimization procedures using different gathered information to economically operate the MG. Unit commitment (UC) and economic dispatch (ED) are two popular functions of the centralized energy management system which are applied in different time scales. In order to reduce the effect of uncertainties, scheduling should be continuously renewed by updating the input data. This procedure can be applied to EMS as a rolling horizon (RH) strategy [2]. There are different decision-making methods for the implementation of UC and ED. For instance, they can be considered as the optimization problems containing continuous or discrete variables [3–10]. Also, the machine learning systems [11, 12], and rule-based systems [13] can be used as decision-making strategies. Centralized EMS is appropriate for the real-time application [14] because different parts of MG are monitored and analyzed in a central agent. In this structure, EMS should be equipped with a powerful computing unit to process a high amount of input data. Decentralized energy management has a lower computational burden in which different agents schedules a part of MG. This method is appropriate for large MGs, where the resources and loads are dispersed and centralized data gathering is impossible or costly. A multi-agent system (MAS) is one of the important techniques in decentralized decision making for power management in different time scales [15, 16]. In MAS, each component is managed by an agent that makes decisions to achieve its desired goals. The agent uses its local information and the received data from the other agents. In this method, the final operation scheme is achieved by passing an iterative process. For applying MAS to real-time scheduling, the convergence speed is a vital factor [17].
70
S. M. Hashemi and V. Vahidinasab
3.8 The Necessity of EMS in the Scheduling of MG With the advancement of technology and the rise of the smart grid concept, MGs have become more and more popular. MG is a small power system that supplies loads using different distributed energy resources (DER). Besides, there are advanced monitoring and control facilities. Distinguishing features of the MGs are as follows: • These power systems contain different distributed generation (DG) technologies. • The number of end users and energy resources of the MGs are very lower than that of the bulk power systems. Therefore, each consumer has a higher impact on the system’s load, and the system condition is so variable. • MG’s inertia constant is very low. The bulk power systems containing many resources and loads have high inertia constant. So, sudden outage of the lines or generators has no significant impact on the system frequency. However, the MGs having low inertia are very vulnerable to disturbances. • Having the radial structure, MGs are so vulnerable against the line outages. Bulk power systems have a circular structure and supply the load points from different paths. On the other hand, MGs’ access to the resources is reduced in the line outage conditions. • Considering the appropriate access to the end users, MGs are capable to exercise different control actions based on the load’s properties. According to the mentioned properties, only a powerful EMS can operate the MG efficiently and reliably.
3.9 EMS Functions in the MG Scheduling EMS is the main decision module of the MG scheduling the resources to serve the loads and exchange the energy with the main grid based on several input parameters indicating the MG’s condition. Figure 3.4 presents the inputs and outputs of the EMS. Some of these input data such as network configuration and technical properties of the resources are about constant during the scheduling horizon. On the other hand, load, electricity market price, and the renewables’ power generation are rapidly changing during the operation horizon. Example 3.1 A simple MG containing two MTs (with the capacity of 10 MW) and a photovoltaic (PV) unit (with an installed capacity of 5 MW) is depicted in Fig. 3.5. As indicated in Fig. 3.6a, EMS schedules the resources to serve the load during a 24-hours horizon aiming the minimum operation cost. The operation cost of MTs is 5 (cent/kWh) and PV’s generation cost is about 0. Also, the hourly price forecast of the electricity market and the forecasted power generation of the PV unit are respectively indicated in Fig. 3.6b–c. As can be observed, the PV unit is operated in its maximum available capacity during the hours 8–21. During the hours 1:00 to
3 Energy Management Systems for Microgrids
71
Input data Technical properties of resources
Electricity market price
Optimization module of EMS Load forecast Power generation forecast of the renewable resources Power generation cost of the resources
Operation strategies: Cost minimization, Profit maximization Constraints: Technical constraints of components, Reliability constraints
Scheduling decisions Output power of generation resources Charge/discharge power of energy storages Consumption scheduling of the controllable loads
Internal network topology System operation strategy
System operation horizon
Fig. 3.4 The structure of MG scheduling by EMS Fig. 3.5 A simple MG
Main grid
Microgrid
PV
MT2
MT1
11:00, all of the consumption demand is supplied by the main grid and both of the MTs are off. By increasing the electricity market price beyond the 5 cents/kWh after the 12:00, MTs are started to supply the load. It was a very simple example showing the function of EMS, and it may not be so complicated to schedule the MG without the EMS, in this example. However, by increasing the number of resources (generation and storage units) and consumers, considering the network constraints and need to fast decisions near the real operation time, the presence of a strong EMS will be essential. In the following, EMS functions will be analyzed in the aspects of the planning hierarchy and operation strategies.
72
S. M. Hashemi and V. Vahidinasab
a
b
c
Fig. 3.6 The inputs and outputs of EMS in Example 3.1: (a) Power generation scheduling. (b) Generation cost of MTs and the electricity market price. (c) Forecasted power generation of the PV units
3.9.1 Microgrid’s Hierarchical Scheduling EMS schedules the MG using forecasted data of the loads and power generation of the renewable generation units. According to the variable condition of the system during the operation horizon, results of the hour-ahead scheduling (HAS) such as 24-hour scheduling of Example 3.1, are not valid for the real-time system operation. Therefore, EMS should rerun an optimum scheduling problem near the real operation time, based on the results of the HAS and using the updated input data. Indeed, a hierarchical scheduling framework is needed that firstly applies HAS over the hours of the operation horizon and then uses real-time scheduling (RTS) module for energy management within the small time steps near the real operation time. Although RTS as the latest operational planning of the system has a significant role in the system operation procedure it still needs the HAS for the following reasons: • Some of the decision variables should be inherently considered in the HAS problem and they cannot be determined in real-time scheduling. For example, scheduling of shiftable loads should be performed considering the whole opera-
3 Energy Management Systems for Microgrids
73
HAS1
U p d a tin g th e fo r ec a s te d input data
h2
h1
...
RTSk,m
h3
...
RTSk,2
...
RTSk,1
RTS2,2
RTS2,1
RTS1,m RTS1,2
...
. . .
HASk
RTS2,m
. . .
RTS1,1
M G s c h ed u li n g by EM S
HAS2
... hk+1
hk
...
h24
. . .
Fig. 3.7 Hierarchical scheduling of MG in EMS
tion horizon [18]. Also, the state of charge (SOC) of the energy storage should be determined considering the whole operation horizon [19]. • Repeated start and shut down of microturbines in the small time steps is not desired and amortizes them. Therefore, the status of the microturbine is determined in the HAS and is not changed in RTS. • RTS should be fast because it is run repeatedly near the real operation time. HAS can provide an initial solution for the optimization model of the RTS module that reduces the time of solving the problem. Also, to further reduce the problem’s solving time, some of the decision variables can be fixed into the values calculated by HAS. Generally, the module of HAS can be hourly run for the next hours up to the end of the operation horizon, applying the updated data, as indicated in Fig. 3.7. Then, the results are applied to the RTS which is repeatedly run in 5-minute time steps. Although the RTS schedules the MG only for the next time steps to the end of the current hour it indirectly considers the system’s forecasted condition of the next hours through applying the results of the latest HAS problem. Example 3.2 Near the real operation time, the electricity market price is determined by the real-time market. It deviates from the forecasted values. This deviation can be modeled by the Normal probability distribution. Figure 3.8 indicates the real-time market prices in the 15-minute time steps. Without the HAS, the on/off status of the MTs should be determined in RTS. In this condition, severe price fluctuations can result in the start or shut down of the MTs, as indicated in Fig. 3.8. As a result, EMS should benefit from both HAS and RTS modules to appropriately schedule the MG.
74
S. M. Hashemi and V. Vahidinasab
Fig. 3.8 The inefficiency of RTS to determine the on/off status of MTs
3.9.2 System Operation Strategies Different operation strategies can be considered in the EMS, such as [1] operational and maintenance cost of resources, energy transaction cost, battery degradation cost [20], outages and interruption cost, demand response incentives, losses cost, load shedding penalty cost, emission cost and levelized cost of renewable energy resources. Additionally, the EMS may include technical constraints such as electrical network loading, energy balancing, output limits of the renewable energy resources, demand response, reactive power support, reliability, and physical limits of resources. In summary, the operation strategies can be categorized into the economic strategies and technical strategies that will be presented in the following subsections.
3.9.2.1
Economic Aspects
Minimizing the total operation cost or maximizing the profit are the most popular economic objectives of EMS. When the MG is connected to the main grid, it can export its surplus power to the main grid to make a profit. In return, it can import power from the main grid to supply internal consumers. There is an economic competition between the internal resources of the MG and the upstream network to meet the consumers. The electricity market price will be a vital input parameter of the decision-making, and a decision with a higher profit is the best choice in this condition. In a market-oriented decision-making, EMS generates the price bids to
3 Energy Management Systems for Microgrids
75
Exported power to the main grid
Fig. 3.9 Power generation scheduling of a profit maximizer EMS
participate in the power market [13]. MG’s profitability is not constrained to the energy exchange with the main grid. According to the fast response resources and the controllable loads, EMS can make a profit by providing reserve capacity for the main grid [21]. For the isolated MGs, cost minimization is the main economic strategy. Example 3.3 In Example 3.2, When EMS operates the MG for the profit maximization, the scheduling outputs will be different from the cost minimization strategy, as indicated in Fig. 3.9. As can be observed, MG exports energy to the main grid in some hours. When the. market price is higher than the generation cost of MTs and there is surplus generation capacity.
3.9.2.2
Technical Aspects
Aside from the economic aspects, there are some technical criteria for the system operation problem. Maximizing the reliability indices of the system and minimizing the network loss or fuel consumption [22] are important technical decision strategies. The two later strategies may be in line with the cost minimization or profit maximization. While the reliability maximization may be against the cost/profitoriented decisions. Suppose a sudden contingency leads to the load shedding. Considering the main task of the power system is supplying the load, the system operation strategy is to minimize the load shedding or to maximize the amount of the restored loads. In fact, in the contingency condition, minimum cost or maximum profit is not considered by the MG’s operator. Although the MGs are operated in the radial structure, they are designed in the mesh structure, using the
76
S. M. Hashemi and V. Vahidinasab
System operation strategies
Cost minimization
Profit maximization
Technical aspects
Operation cost Maintenance cost Startup/shut down cost
Exporting energy to main grid Providing ancillary services for main grid Strategic bidding to the electricity markets
Minimizing the energy loss Minimizing the fuel consumption
Reliability and Resiliency
Minimum load shedding. Maximum load restoration
Fig. 3.10 System operation strategies of EMS
normally open and normally closed switches. In some scheduling problems, system reconfiguration variables are determined to minimize the system loss [23]. As well as, reconfiguration can be applied for the reliability or resiliency goals. For example, in some of the isolated MGs such as the electric ships, EMS should select the best supplying path of the loads during the critical condition aiming the minimum load curtailment [24]. As well, EMS may be designed to deal with the physical attacks or cyberattacks using the system reconfiguration [25]. Figure 3.10 summarizes the operation strategies of the EMS.
3.10 Mathematical Modeling of Different MG’s Components To design and implementation of the EMS, the mathematical models of different elements of the MG are presented in the following.
3.10.1 Loads There are different electrical and thermal loads in the MG. Generally, the electrical load model is expressed by the ZIP model indicated in (3.1) and (3.2) [26]. It means that each load can be considered as the combination of three load models of constant impedance, constant current, and constant power.
3 Energy Management Systems for Microgrids
P = P0 Zp Q = Q0 Zq
V V0
2
+ Ip
V V0
2
V V0
+ Iq
77
+ Pp ,
V V0
Zp + Ip + Pp = 1
(3.1)
Zq + Iq + Pq = 1
(3.2)
+ Pq ,
The hourly thermal load is generally modeled as a constant energy consumption within an hour. In terms of controllability, loads are categorized into controllable and uncontrollable loads. In the controllable loads, the power consumption (P) can be controlled in the range of (3.3), in which PL is the forecasted power consumption. Eq. (3.4) guarantees that the total power consumption of the load is constant during the operation horizon. 0 ≤ P ≤ PL
Tn h=1
Ph =
Tn
P Lh
(3.3)
(3.4)
h=1
3.10.2 Dispatchable Generation Resources Dispatchable generation (DG) units have the main function of energy generation in MG. Diesel generators and microturbines (MT) that consume natural gas and produce electrical energy are categorized as the DG units. The generation cost of DGs is a quadratic function as indicated in (3.5). To linearize the optimization model, this cost function can be replaced by some of the estimating linear functions [27] as indicated in Fig. 3.11. Cg = AP 2g + BP g + C
(3.5)
The startup cost of the generation units is another cost term that is indicated in (3.6). SC g,t ≥ 0, SC g,t ≥ ug,t -ug,t -1 scg
(3.6)
The active and reactive power of the online DGs should be within the allowable range of (3.7) and (3.8). In these inequalities, u is a binary variable indicating the status of the units.
78
S. M. Hashemi and V. Vahidinasab
Cg ln l3
l2
l1 Pg Fig. 3.11 Cost function of the dispatchable gas-burning generation units
Pgmin ug,t ≤ Pg,t ≤ Pgmax ug,t
(3.7)
max Qmin g ug,t ≤ Qg,t ≤ Qg ug,t
(3.8)
There are ramping limits between the consecutive time intervals as indicated in (3.9)–(3.14). URg and DRg are respectively the up and down ramping capabilities. z and y are binary variables respectively showing the start and stop of the units. Also, P0 indicates the initial power generation of the units at the beginning of the scheduling. The inequalities of (3.9) and (3.10) are the up-ramping constraints. After starting a unit, its generation should be fixed to Pmin . As well as, (3.11) and (3.12) are the down ramping constraints. Before stopping a generation unit its power should be fixed to Pmin . Pg,t − Pg,t−1 ≤ 1 − zg,t U R g + zg,t Pgmin Pg,t − Pg0 ≤ 1 − zg,t U R g + zg,t Pgmin Pg,t−1 − Pg,t ≤ 1 − yg,t DR g + yg,t Pgmin Pg0 − Pg,t ≤ 1 − yg,t DR g + yg,t Pgmin
∀ t >1
∀t =1
∀ t >1 ∀t =1
(3.9)
(3.10)
(3.11) (3.12)
3 Energy Management Systems for Microgrids
79
zg,t + yg,t ≤ 1 ∀ t
(3.13)
zg,t − yg,t = ug,t − ug,t−1 ∀ g, t
(3.14)
Repeated starting and stopping of DGs in short time intervals amortizes their mechanical elements. Hence, the minimum up/down time is defined. According to (3.15) and (3.16), when a unit is started/stopped, its status should not change up to the MUT/MDT time interval. T
ug,t+k−1 ≥ zg,t T ,
! T = min MU T , T h − t + 1
(3.15)
k=1 T 1 − ug,t+k−1 ≥ yg,t T ,
! T = min MDT , T h − t + 1
(3.16)
k=1
3.10.3 Renewable-Based Units with MPPT Renewable resources are popular generation units of MG. Their generation cost is very low and they are categorized as clean energy technologies. Considering their primary energy resources, they have lower controllability than conventional DGs. For example, the power generation ability of the PV units is directly related to the solar irradiance and ambient temperature. In fact, they cannot generate power as much as their installed capacity at any time. Generally, the wind turbines and PV panels are equipped with the maximum power point tracking (MPPT) algorithm [28]. MPPT trackers find the best working point of the mentioned generation units in which the maximum power is produced based on different environmental conditions. In fact, MPPT trackers improve the controllability of the renewable resources. Considering P* as the maximum forecasted power of renewable resources, their actual power generation should be within the specified range of (3.21). ∗ 0 ≤ Pg,h ≤ Pg,h , g ∈ ΩW
(3.17)
3.10.4 Energy Storages Energy storages have a vital function in the MGs. They can be operated as generators or consumers. They are charged in some hours and generate energy in the next hours. By the storage units, it will be possible to benefit from the energy price difference
80
S. M. Hashemi and V. Vahidinasab
in different hours. As well as, in the light load condition, they can be charged by the renewable resources and supply the load at peak hours. SOC of the storages is calculated by (3.18) which considers the initial stored energy (E0) and charge or discharge power (Psc/d ) at different time intervals. Since a part of the energy is wasted during the power charge or discharge, energy efficiency (ηc/d ) is included in (3.18). Energy storage capacity and charge and discharge rate should be considered as indicated in (3.19)–(3.22). Est,t = E0st +
t
c ηst · P s cst,k −
k=1
t d P sst,k k=1
d ηst
(3.18)
min max Est ≤ Est,h ≤ Est
(3.19)
0 ≤ P s cst,h ≤ P s max st
(3.20)
0 ≤ P s dst,h ≤ P s max st
(3.21)
P s st,h = P s dst,h − P s cst,h
(3.22)
The operation of energy storage has no significant operational cost.
3.10.5 Reactive Power Resources Generally, capacitors and reactors of the electrical network participate in the voltage regulation as the reactive power resources. Their reactive power generation should be constrained by their installed capacities as indicated in (3.23). Qcmin ≤ Qcc,t ≤ Qcmax c c
(3.23)
3.10.6 Combined Heat and Power (CHP) and Boiler CHPs having high energy efficiency are usually installed in the MGs containing the thermal loads. CHP simultaneously generates electrical and thermal power using natural gas. The allowed operational zone of the CHPs has been indicated in Fig.
3 Energy Management Systems for Microgrids
81
PCHP
Maximum heat generation
HCHP
Fig. 3.12 Allowable thermal-electrical operating zone of CHP
3.12 and (3.28). Operational constraints presented for DGs can be applied for CHPs. The constraints such as ramping limitations. chp
th CH P th CH P th αCH P Pch,t + βCH P Hch,t ≤ γCH P ug,t th ∈ {1, 2, 3}
(3.24)
In general, the operation cost of the CHP unit is a nonlinear function as indicated in (3.25) [29]. p
p
2 h 2 h + βch Pch,t + αch Hch,t + βch Hch,t + λch Pch,t Hch,t + σch Cch,t = αch Pch,t (3.25)
In [30] a linearized version has been proposed for operation cost of CHP units in which piecewise linearization has been applied for two variables: output electrical and thermal power. Besides, some researches [31] use a linear cost function for CHP CH P ) and gas price (π gas ) as indicated using an average energy efficiency factor (ηch MWh/m3 (3.30)–(3.31). In these constraints, σ is a conversion factor that includes the thermal value of the natural gas. P Cch,t = π gas gas CH ch,t
P MW h/m3 gas CH ch,t = σ
CH P + H CH P Pch,t ch,t CH P ηch
(3.26)
(3.27)
To increase the flexibility and to serve the thermal load, auxiliary boilers are installed beside the CHP units. Their thermal power generation should be within the allowed range as indicated in (3.32).
82
S. M. Hashemi and V. Vahidinasab boi,min boi,max boi uboi ≤ Hb,t ≤ uboi b,t Hb,t b,t Hb,t
(3.28)
Similar to the operation cost of CHP units in (3.30)–(3.31), the operation cost of the boilers is modeled by (3.29) and (3.30) in which uboi b,t is a binary variable indicating the on/off status of the boiler. Cb,t = π gas gas boi b,t
MW h/m3 gas boi b,t = σ
boi Hb,t
ηbboi
(3.29)
(3.30)
3.10.7 Electrical Network Operational variables of the electrical network such as line flows and bus voltages should be modeled in EMS. They can be computed using the injected power of the buses and the properties of the lines through the power flow equations. Although methods such as forward–backward sweep [32] calculate the line flows and bus voltages of the radial networks in an iterative procedure they are not appropriate for the centralized optimization models of EMS. Direct current (DC) power flow is a popular method for large power systems. Although it is a linear and simple method, it is not appropriate for the MGs and distribution networks in which the R/X ratio is high. On the other hand, using the conventional AC power flow complicates the optimization problem because it contains the non-convex constraints. In the following subsections, two mathematical models are presented for the power flow.
3.10.7.1
Second-Order Cone Programming (SOCP)
According to [33], an exact second-order cone relaxation method can be defined for the convexification of the AC load flow of the radial systems. The convexified AC load flow equations are indicated in (3.27)–(3.37). The Eqs. (3.31) and (3.32) are respectively the active and reactive power balance in each bus and (3.33)-(3.34) are the injected active and reactive power of the buses. According to (3.35), any bus voltage is determined based on the input power from the upstream bus and its voltage. Eq. (3.36) relates the active and reactive flows to the bus voltage and line currents. In the conventional formulation, (3.36) is a non-convex equality constraint. In return, the indicated inequality constraint is convex. This relaxation will be exact, if the mentioned constraint is active, after solving the optimization problem. In p q these equations, Fn,t and Fn,t are active or reactive input powers of bus n from its upstream line. r and x are respectively the resistance and reactance of the lines. Pg, t , Psst, t and PLn, t are respectively the output active power of generators and
3 Energy Management Systems for Microgrids
83
storage units and the energy consumption of the load. As well as Qg, t , Qcc, t and QLn, t are respectively the reactive power generation of the generators and reactive power resources, and the reactive loads. Vn, t and In, t are the bus voltages and line currents that are indicated in Fig. 3.13. p
inj
Fn,t + Pn,t =
p Fm,t + rm m,t ,
p
(3.31)
q
(3.32)
P s st,t − P Ln,t
(3.33)
Qcc,t − QLn,t
(3.34)
Fn=0,t = 0
m∈Ωn
q
inj
Fn,t + Qn,t =
q Fm,t + xm m,t ,
Fn=0,t = 0
m∈Ωn
inj
Pn,t =
g∈Ωgn
inj
Qn,t =
Pg,t +
g∈Ωgn
st∈Ωst n
Qg,t +
c∈Ωcn
p q 2 2 m,t , m ∈ Ωnn υn,t = υm,t + 2 rm Fm,t + xm Fm,t + rm + xm p 2 q 2 Fn,t + Fn,t ≤ υn,t n,t , n = 0
2 2 υn,t = Vn,t , n,t = In,t
V 02 (1 − 0 )2 ≤ υn,t ≤ V 02 (1 + 0 )2
3.10.7.2
(3.35)
(3.36)
(3.37)
Linear DistFlow
The presented SOCP method is an exact convex model for power flow. However, it is a MIQCP model. There is another power flow method that is a linear model. The mathematical formulation of the Linear DistFlow [34] method is indicated in Fig. 3.13 Line flow and bus voltage
84
S. M. Hashemi and V. Vahidinasab
(3.34)–(3.44). As can be observed, its accuracy is higher than the DC load flow method. It determines the bus voltages, while it does not model the power loss of the lines. It should be noted that Eq. (3.44) is not included in the optimization model. It is used to calculate the bus voltages after solving the optimization problem. p
inj
Fn,t + Pn,t =
Fm,t ,
p
Fn=0,t = 0
p
(3.38)
q
Fn=0,t = 0
q
(3.39)
P sst,t − P Ln,t
(3.40)
Qcc,t − QLn,t
(3.41)
m∈Ωn
q
inj
Fn,t + Qn,t =
Fm,t ,
m∈Ωn
inj
Pn,t =
Pg,t +
g∈Ωgn
inj
Qn,t =
st∈Ωst n
Qg,t +
g∈Ωgn
c∈Ωcn
p q υn,t = υm,t + 2 rm Fm,t + xm Fm,t , m ∈ Ωnn
(3.42)
V 02 0 2 − 20 ≤ υn,t ≤ V 02 0 2 + 20
(3.43)
2 − V 02 υn,t = Vn,t
(3.44)
3.10.8 Energy Exchange with the Main Grid MG can export/import electrical energy to/from the main grid. The cost of the maingrid energy exchange with the main grid is calculated by (3.45) in which Pt is the electrical imported power from the main grid and prct is the electricity market price. maingrid The negative value of Pt means the power export to the main grid resulting in the negative cost that means the profit. grid
Ct
maingrid
= Pt
prct
(3.45)
3 Energy Management Systems for Microgrids
85
3.11 Mathematical Modeling of System Security Power system reliability is a concept introducing different indices showing the system capability to deal with the planned and unplanned outages of the equipment. It is divided into two popular terms of adequacy and security. The reliability of the system in a short-term scheduling problem is analyzed through the concept of system security. n − 1 is a well-known criterion in security modeling. It means that the normal state working point of the system should be robust against the outage of any single element of line or power plant so that the operating limits are satisfied and all the loads are supplied. Usually, MG’s security is analyzed when it is suddenly isolated from the main grid [35]. In comparison to the large power systems, MGs have lower inertia. So, the power disturbance severely affects the MGs frequency. In fact, in addition to the reserve constraints, some frequency-related constraints should be considered in the scheduling procedure as dynamic security. Here, two formulations are presented for the dynamic security which is compatible with HAS and RTS.
3.11.1 Security Modeling in HAS It is not possible to include the exact frequency control models in the HAS because the on/off status of the resources is unknown and the frequency response model is undefined. HAS should provide enough online generators to increase the system’s inertia and its capability to respond to the power disturbance. Here, an energyrelated frequency response model is presented [36, 37]. The kinetic energy of the MG is indicated in (3.46) in which JT and f are respectively the total inertia moment of MG and the frequency of MG. E=
1 T J (2πf )2 2
(3.46)
JT is the sum of inertia moments of the online generators as indicated in (3.47). In this equation, Hk is the inertia constant of the kth generator.
JT =
k∈online generators
Jk =
k
uk Jk =
k
uk
2P max Hk 4π 2 f 2
(3.47)
A power mismatch of D > 0 (kW) for t seconds reduces the system’s energy as "t much as ΔE1 = Ddτ = Dt. Generation units change their production to deal with 0
"t the disturbance. This changes the systems’ energy as much as ΔE2 = rτ dτ = 0
rt 2 2 ,
in the opposite direction of E1 , in which r is the equal ramp rate of the generators.
86
S. M. Hashemi and V. Vahidinasab
Considering E = E1 − E2 as the net energy deviation of the system, the time of minimum kinetic energy (tnadir ) can be calculated by (3.48). D d (ΔE) = 0 → t nadir = dt r
(3.48)
According to the minimum allowed value of the system frequency (fmin ), a lower bound is determined for the systems’ kinetic energy in t = tnadir as indicated in (3.49). E − ΔE ≥ E min →
2 1 1 T D2 J (2πf )2 − ≥ J T 2πf min 2 2r 2
(3.49)
Replacing (3.47) and (3.48) into (3.49) results in (3.50). t
nadir
2 2 f − f min ≤ uk Pkmax Hk f 2D
(3.50)
k
On the other hand, it takes tr minutes to deliver Ri MW of reserve capacity by the generation unit of i. It is determined considering its ramp rate (ai ) as indicated in (3.51). tr =
Ri ai
(3.51)
When tr of the reserve units is lower than tnadir , the system frequency won’t be lower than fmin . Considering (3.50) and (3.51), reserve allocation should be constrained as indicated in (3.52). 2 2 f − f min Ri ≤ uk Pkmax Hk ai f 2D
(3.52)
k
Also, the sum of the reserve capacity of the generators should be equal to the power mismatch (3.53).
Ri = D
(3.53)
i
In the connected mode operation of MG, the power mismatch of D is equal to the amount of power exchange with the main grid. In the isolated MGs, any load connection or the generation outage can be considered as the power disturbance. Example 3.4 An isolated MG is analyzed in this example. If in Example 3.1 generation capacity of MTs is increased to 40 kW, MG will be able to serve the loads
3 Energy Management Systems for Microgrids
87
Table 3.1 Technical and economic properties of the generation units of MG MT1 MT2 PV
H (sec) 1.5 1.5 0
Pmax (kW) 40 40 Variable
Pmin (kW) 4 4 0
Cost (cent/kW) 4 4.4 0
Ramp rate (kW/sec) 10 12 50
in the islanded mode. The technical and economic specifications of the generation units are indicated in Table 3.1. As indicated in Fig. 3.14a, if the frequency deviation is not considered in the system scheduling, PV and MT1 will supply the load during the whole operation horizon. Given the generation cost of resources, that is the right decision. By adding the frequency-related constraints, different factors affect the power generation schedule. Factors such as inertia constants of the online generators, minimum allowed frequency, ramp rate of the resources, and the power mismatch. Assuming fmin = 49 Hz and D = 5% load, generation scheduling will be as Fig. 3.14b. It is worth noting that the PV power plant can participate in the primary frequency regulation procedure [38–41]. As can be observed, MT2 that is an expensive generation unit is online in many hours, and it generates as much as its minimum power. In fact, on the one hand, it increases the system’s inertia and on the other hand, it provides a fast reserve resource. Figure 3.14c indicates the reserve capacity of different resources. PV unit having low operation cost is often operated on its maximum generation ability and does not provide reserve capacity. Figure 3.14d indicates tr of the units and tnadir of the system. Reserve capacities are deployed before tnadir . Example 3.5 Increasing the generation cost of MT2 to 6 (cent/kW) changes the scheduling results. In this condition, as indicated in Fig. 3.15a MT2 is turned off in many hours that results in the low inertia constant of the MG. To deal with this problem, the PV unit having a very high ramp rate reduces the power generation (Fig. 3.14d) and provides a large share of the reserve capacity (Fig. 3.14b). Despite the reduction of the system’s inertia during the hours 9–19, the reserve capacity is deployed very fast as indicated in Fig. 3.14c.
3.11.2 Security Modeling in RTS In RTS, online generators are determined by the latest HAS. Hence, the frequency response model can be constructed. It can be used to limit the disturbance power of the MG within a specified margin avoiding frequency deviation beyond the allowed values. For instance, in the connected operation mode of MG, the sudden islanding severely affects the MG’s frequency. Especially when the power exchange with the main grid is high, the RTS should limit the power exchange. Frequency response
Fig. 3.14 Results of HAS for islanded MG of Example 3.4: (a) Power generation of different resources in the insecure operation. (b) Power generation of different resources in the secure operation. (c) Reserve capacity of different resources. (d) The reaction time of the reserve resources versus the frequency drop time
88 S. M. Hashemi and V. Vahidinasab
Fig. 3.15 Results of HAS for islanded MG of Example 3.5: (a) Power generation of different resources. (b) Reserve capacity of the MG’s resources. (c) The reaction time of the reserve resources versus the frequency drop time. (d) PV unit’s generated power versus its available power
3 Energy Management Systems for Microgrids 89
90
−k s
S. M. Hashemi and V. Vahidinasab
1 R
Secondary control loop
Primary control loop (MT) Micro turine
_ +
1
1
+
1+ T S
1+ T S
Governer
Turbine
v
t
_ +
1 2Hs + D
f
kss ΔP=P main grid
1 + TsS Virtual inertia control system
Fig. 3.16 Frequency response diagram of MG
Frequency
0
Frequency response for power disturbance of pα
-f α
-Hzmax
0
Time (sec)
Fig. 3.17 The impact of power disturbance on the MG’s frequency deviation
in the downside of the network can be calculated by the linear frequency response model of [42] that is indicated in Fig. 3.16. As indicated, the input power of MG is considered as the disturbance power of the frequency response model. According to Fig. 3.17, Assuming fα as the maximum frequency deviation for power disturbance of pα , power exchange should be within the specified margin of (3.54) avoiding frequency to go beyond HZmax [43]. maingrid
−
H zmax P ≤ t α ftα p
≤
H zmax ftα
(3.54)
In order to calculate the dynamic parameters of the frequency response model of the MG’s equal machine, equations of (3.55)–(3.61) should be considered [44].
3 Energy Management Systems for Microgrids
91
# H =
Pkmax Hk
k∈online generator
#
Pkmax
k∈online generator
# 1 = R
Pkmax R1k
k∈online generator
#
Pkmax
k∈online generator
# Tv =
#
# #
k∈online generator
k=
i∈online generator
i∈online generator
ks =
i∈online storages
i∈online storages
Ts =
k∈online storages
Pkmax
(3.58)
Pimax
(3.59)
Pimax kis
#
#
(3.57)
Pimax ki
#
#
Pkmax
Pkmax Tkt
k∈online generator
#
(3.56)
Pkmax Tkv
k∈online generator
k∈online generator
Tt =
(3.55)
Pimax
(3.60)
Pkmax Tks
#
k∈online storages
Pkmax
(3.61)
In the islanded operation mode in which the outage of generation units leads to power disturbance, the frequency limit of (3.54) can be considered in several single contingencies of the online generators. Example 3.6 Suppose that in Example 3.4, MG is connected to the main grid. Dynamic parameters of MTs are indicated in Table 3.2. Also, ks and Ts are respectively 0.5 and 10 seconds.
92
S. M. Hashemi and V. Vahidinasab
Table 3.2 Dynamic parameters of the MTs
MT1, MT2
a
d
H (p.u. MW/sec) 1.5
Tv (sec) Tt (sec) R (Hz /p.u. MW) 0.1 0.4 2.4
D K (p.u. (MW/Hz) MW/Hz) 0.015 0.05
ks (sec)
T3 (sec)
0.5
10
b
c
Fig. 3.18 The impact of frequency constraints on the RTS: (a) Given the status of MTs from HAS. (b) generation cost of MTS and the electricity market price in RTS. (c) Power generation scheduling in an insecure operation. (d) Power generation scheduling in the secure operation
We want to schedule the MG in real time during 04:00–08:45. The results of the latest HAS are provided and determine the online generators in different hours as indicated in Fig. 3.18a. The electricity market price and generation cost of MTs are indicated in Fig. 3.18b and RTS is run in 15 minutes intervals. When there is no attention to the frequency response of the MG in the sudden islanding, MG imports power from the main grid and reduces the power generation of MTs in most of the operation horizon as indicated in Fig. 3.18c. According to the market price variations, it is the best decision from the economic viewpoint. Adding security constraints changes the MG’s scheduling. When the maximum frequency deviation is limited to 0.1 Hz, input power from the main grid will be reduced as indicated in Fig. 3.18d. This scheduling guarantees the secure operation of MG in the case of sudden islanding. As can be observed, during the hours 7 and 8 in which both of the MTs are online, power scheduling has no significant sensitivity to the frequency constraints because in this condition high inertia constant of MG avoids the severe frequency deviation. In return, the low inertia constant of MG during the hours 4 to 6, activates the frequency constraint and changes the generation scheduling decisions.
3 Energy Management Systems for Microgrids
93
3.12 Conclusion The EMS is a vital element in an MG that monitors and controls all the interactions and operates different resources based on the determined strategies. To analyze the MG operation, EMS provides different analytical tools, the most widely used of which is the optimal operation tool. This chapter focused on this aspect to present the scheduling models of MG. Cost minimization and profit maximization are two popular operation strategies of the MG. According to the intrinsic properties of some of the decision variables and the uncertainties of the input parameters, a two-stage scheduling structure was used for EMS which includes the HAS and RTS. HAS includes a UC model and is hourly run to schedule the MG from the running time to the end of the operation horizon. Hence, all of the forecasted parameters of the next hours are considered in HAS. According to the variability of the input parameters such as load, renewable power generation, and electricity market price, the results of HAS are not appropriate for the system operation and MG should be rescheduled near the real operation time by RTS. By receiving the on/off status of the resources from the results of the latest HAS, the RTS only focuses on the current hour using an ED model. To deal with contingencies, security was considered in EMS using the frequency response model. MG’s inertia is very low and sudden contingencies severely deviate the frequency. An energy-based frequency response model was applied to HAS to the commitment of fast response generators in accordance with the MG’s inertia. In RTS, the online generators are given by the latest HAS. So, the exact frequency deviation was calculated by the frequency response diagram, and an allowed range was determined for it. The mathematical formulations and technical details of this chapter were included both the islanded and connected operation mode of MG.
References 1. Zia, M. F., Elbouchikhi, E., & Benbouzid, M. (2018). Microgrids energy management systems: A critical review on methods, solutions, and prospects. Applied Energy, 222, 1033–1055. 2. Palma-Behnke, R., et al. (2013). A microgrid energy management system based on the rolling horizon strategy. IEEE Transactions on Smart Grid, 4(2), 996–1006. 3. G. Graditi, M. G. Ippolito, E. R. Sanseverino, and G. Zizzo, Optimal set points regulation of distributed generation units in micro-grids under islanded operation, In 2010 IEEE International Symposium on Industrial Electronics, 2010, pp. 2253–2260. 4. Chen, C., Duan, S., Cai, T., Liu, B., & Hu, G. (2011). Smart energy management system for optimal microgrid economic operation. IET Renewable Power Generation, 5(3), 258–267. 5. M. L. Di Silvestre, G. Graditi, M. G. Ippolito, E. R. Sanseverino, and G. Zizzo, “Robust multi-objective optimal dispatch of distributed energy resources in micro-grids,” In 2011 IEEE Trondheim Power Tech, 2011, pp. 1–5. 6. Sanseverino, E. R., Di Silvestre, M. L., Ippolito, M. G., De Paola, A., & Re, G. L. (2011). An execution, monitoring and replanning approach for optimal energy management in microgrids. Energy, 36(5), 3429–3436.
94
S. M. Hashemi and V. Vahidinasab
7. R. Palma-Behnke, C. Benavides, E. Aranda, J. Llanos, and D. Saez, “Energy management system for a renewable based microgrid with a demand side management mechanism,” In 2011 IEEE symposium on computational intelligence applications in smart grid (CIASG), 2011, pp. 1–8. 8. Bustos, C., Watts, D., & Ren, H. (2012). Microgrid operation and design optimization with synthetic wins and solar resources. IEEE Latin America Transactions, 10(2), 1550–1562. 9. X. Zhang, R. Sharma, and Y. He, “Optimal energy management of a rural microgrid system using multi-objective optimization,” In 2012 IEEE PES innovative smart grid technologies (ISGT), 2012, pp. 1–8. 10. Mohamed, F. A., & Koivo, H. N. (2012). Multiobjective optimization using mesh adaptive direct search for power dispatch problem of microgrid. International Journal of Electrical Power & Energy Systems, 42(1), 728–735. 11. G. Celli, F. Pilo, G. Pisano, and G. G. Soma, “Optimal participation of a microgrid to the energy market with an intelligent EMS,” In 2005 International Power Engineering Conference, 2005, pp. 663–668. 12. Siano, P., Cecati, C., Yu, H., & Kolbusz, J. (2012). Real time operation of smart grids via FCN networks and optimal power flow. IEEE Trans Ind Inform, 8(4), 944–952. 13. A. G. Tsikalakis and N. D. Hatziargyriou, “Centralized control for optimizing microgrids operation,” In 2011 IEEE power and energy society general meeting, 2011, pp. 1–8. 14. Meng, L., Sanseverino, E. R., Luna, A., Dragicevic, T., Vasquez, J. C., & Guerrero, J. M. (2016). Microgrid supervisory controllers and energy management systems: A literature review. Renewable and Sustainable Energy Reviews, 60, 1263–1273. 15. Logenthiran, T., Srinivasan, D., Khambadkone, A. M., & Aung, H. N. (2012). Multiagent system for real-time operation of a microgrid in real-time digital simulator. IEEE Transactions on Smart Grid, 3(2), 925–933. 16. Colson, C. M., & Nehrir, M. H. (2013). Comprehensive real-time microgrid power management and control with distributed agents. IEEE Transactions on Smart Grid, 4(1), 617–627. 17. Liang, H., Choi, B. J., Abdrabou, A., Zhuang, W., & Shen, X. S. (2012). Decentralized economic dispatch in microgrids via heterogeneous wireless networks. IEEE Journal on Selected Areas in Communications, 30(6), 1061–1074. 18. Kanchev, H., Lu, D., Colas, F., Lazarov, V., & Francois, B. (2011). Energy management and operational planning of a microgrid with a PV-based active generator for smart grid applications. IEEE Transactions on Industrial Electronics, 58(10), 4583–4592. 19. Olivares, D. E., Lara, J. D., Cañizares, C. A., & Kazerani, M. (2015). Stochastic-predictive energy management system for isolated microgrids. IEEE Transactions on Smart Grid, 6(6), 2681–2693. 20. L. N. An and T. Quoc-Tuan, “Optimal energy management for grid connected microgrid by using dynamic programming method,” In 2015 IEEE Power & Energy Society General Meeting, 2015, pp. 1–5. 21. Shen, J., Jiang, C., Liu, Y., & Qian, J. (2016). A microgrid energy management system with demand response for providing grid peak shaving. Electric Power Components & Systems, 44(8), 843–852. 22. Anglani, N., Oriti, G., & Colombini, M. (2017). Optimized energy management system to reduce fuel consumption in remote military microgrids. IEEE Transactions on Industry Applications, 53(6), 5777–5785. 23. Zidan, A., & El-Saadany, E. F. (2013). Distribution system reconfiguration for energy loss reduction considering the variability of load and local renewable generation. Energy, 59, 698– 707. 24. S. B. Ganesh, K. Schoder, H.-J. Lai, A. Al-Hinai, and A. Feliachi, “Energy management system with automatic reconfiguration for electric shipboard power systems,” In Proc. of ASNE Reconfiguration and Survivability Symposium, RSS 2005, 2005. 25. Paridari, K., O’Mahony, N., Mady, A. E.-D., Chabukswar, R., Boubekeur, M., & Sandberg, H. (2017). A framework for attack-resilient industrial control systems: Attack detection and controller reconfiguration. Proceedings of the IEEE, 106(1), 113–128.
3 Energy Management Systems for Microgrids
95
26. M. Sadeghi, “Determination of ZIP parameters with least squares optimization method,” In 2009 IEEE Electrical Power & Energy Conference (EPEC), 2009, pp. 1–6. 27. H. Ahmadi, J. R. Martí, and A. Moshref, “Piecewise linear approximation of generators cost functions using max-affine functions,” In 2013 IEEE Power & Energy Society General Meeting, 2013, pp. 1–5. 28. M. A. G. De Brito, L. Galotto, L. P. Sampaio, G. de A. e Melo, and C. A. Canesin, “Evaluation of the main MPPT techniques for photovoltaic applications,” IEEE Transactions on Industrial Electronics, vol. 60, no. 3, pp. 1156–1167, 2012. 29. Sashirekha, A., Pasupuleti, J., Moin, N. H., & Tan, C. S. (2013). Combined heat and power (CHP) economic dispatch solved using Lagrangian relaxation with surrogate subgradient multiplier updates. International Journal of Electrical Power & Energy Systems, 44(1), 421– 430. 30. Moradi, S., Ghaffarpour, R., Ranjbar, A. M., & Mozaffari, B. (2017). Optimal integrated sizing and planning of hubs with midsize/large CHP units considering reliability of supply. Energy Conversion and Management, 148, 974–992. 31. Zhang, Y., Wang, R., Zhang, T., Liu, Y., & Guo, B. (2016). Model predictive control-based operation management for a residential microgrid with considering forecast uncertainties and demand response strategies. IET Generation Transmission and Distribution, 10(10), 2367– 2378. 32. Panwar, L. K., Konda, S. R., Verma, A., Panigrahi, B. K., & Kumar, R. (2017). Operation window constrained strategic energy management of microgrid with electric vehicle and distributed resources. IET Generation Transmission and Distribution, 11(3), 615–626. 33. Gan, L., Li, N., Topcu, U., & Low, S. H. (2015). Exact convex relaxation of optimal power flow in radial networks. IEEE Transactions on Automatic Control, 60(1), 72–87. 34. Šulc, P., Backhaus, S., & Chertkov, M. (2014). Optimal distributed control of reactive power via the alternating direction method of multipliers. IEEE Transactions on Energy Conversion, 29(4), 968–977. 35. Vergara, P. P., López, J. C., da Silva, L. C., & Rider, M. J. (2017). Security-constrained optimal energy management system for three-phase residential microgrids. Electric Power Systems Research, 146, 371–382. 36. S. M. Hashemi, V. Vahidinasab, M. S. Ghazizadeh, and J. Aghaei, “Reliability-oriented DG allocation in radial Microgrids equipped with smart consumer switching capability,” In 2019 International Conference on Smart Energy Systems and Technologies (SEST), 2019, pp. 1–6. 37. Lee, Y.-Y., & Baldick, R. (2013). A frequency-constrained stochastic economic dispatch model. IEEE Transactions on Power Apparatus and Systems, 28(3), 2301–2312. 38. A. Hoke and D. Maksimovi´c, “Active power control of photovoltaic power systems,” In 2013 1st IEEE Conference on Technologies for Sustainability (SusTech), 2013, pp. 70–77. 39. A. S. Dobakhshari, S. Azizi, and A. M. Ranjbar, “Control of microgrids: Aspects and prospects,” In 2011 International Conference on Networking, Sensing and Control, 2011, pp. 38–43. 40. Rahmann, C., & Castillo, A. (2014). Fast frequency response capability of photovoltaic power plants: The necessity of new grid requirements and definitions. Energies, 7(10), 6306–6322. 41. P. Pourbeik, S. Soni, A. Gaikwad, and V. Chadliev, “Providing primary frequency response from photovoltaic power plants,” in Proc. CIGRE Symp., 2017, pp. 1–10. 42. Kerdphol, T., Rahman, F. S., Mitani, Y., Watanabe, M., & Küfeoˇglu, S. (2018). Robust virtual inertia control of an islanded microgrid considering high penetration of renewable energy. IEEE Access, 6, 625–636. 43. Hashemi, S. M., Vahidinasab, V., Ghazizadeh, M. S., & Aghaei, J. (2018). Valuing consumer participation in security enhancement of microgrids. IET Gener Transm Distrib. 44. Shekari, T., Aminifar, F., & Sanaye-Pasand, M. (2016). An analytical adaptive load shedding scheme against severe combinational disturbances. IEEE Transactions on Power Apparatus and Systems, 31(5), 4135–4143.
Chapter 4
Optimal Dispatch and Unit Commitment in Microgrids Hossein Shayeghi and Masoud Alilou
4.1 Introduction In the last years, the concept of microgrid (MG) has been introduced for better managing an extensive and complex power network. In other words, the total power system operates efficiently if each of the microgrids is managed properly. The microgrid is defined as a group of loads, distributed generation (DG) units, and electrical energy storage systems (ESS) [1]. The microgrid is connected to the upstream network for buying energy from the grid when the produced power of DG units and ESS is lower than the demand of the MG. Moreover, MG can sell energy to the power network when there is extra power of energy sources. Therefore, the microgrid has the ability for increasing the local reliability and flexibility of the electric power network [2]. Optimal management of the energy sources of the microgrid has a high effect on the performance of both microgrid and power network. The operator of the microgrid has to schedule the operational time of renewable and nonrenewable DG units and energy storage system of the MG in order to improve the technical, economic, and environmental indices of the MG. The variations of market price, load, and weather affect the operational time of energy sources [3]. In microgrids, demand-side management (DSM) is a practical method to increase the efficiency of the microgrid. DSM is the planning, implementation, and monitoring of utility activities that are designed to control the consumption of customers. DSM causes to change the daily pattern and magnitude of various loads [4]. One of the useful technologies of DSM is the demand response (DR). Demand response
H. Shayeghi () · M. Alilou Energy Management Research Center, University of Mohaghegh Ardabili, Ardabil, Iran © Springer Nature Switzerland AG 2021 A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems, https://doi.org/10.1007/978-3-030-59750-4_4
97
98
H. Shayeghi and M. Alilou
programs can be utilized due to the two-way communication between the suppliers and consumers. These programs are practical to better match the consumption pattern with the production curve of the microgrid [5]. So it can be said that the performance of the microgrid improves significantly when a demand response program and an optimal unit commitment are simultaneously implemented on the microgrid. In the last years, some researchers have studied the unit commitment and demand response programs in microgrids. For instance, optimizing the operational schedule of multi-DG and grid has been studied in [6] in the presence of a timeof-use demand response program. In this study, the considered objective functions were optimized by the combination of a non-dominated sorting firefly algorithm and a fuzzy method. In another study, the authors have proposed an intelligent energy management framework for optimizing the operational schedule of units [7]. The minimization of electrical peak load and reduction of electricity costs are considered as objective functions in this research. Wu et al. have studied the optimal unit commitment in a grid-connected photovoltaic-battery hybrid system in the [8]. The main purpose of this research is to sufficiently explore solar energy and benefit customers at microgrid. Kotur and Durisic in [9] have studied on managing the microgrid through the spatial and temporal demand-side management. The considered objective function of this research consists of daily energy losses and daily operating costs of units. In another research, unit commitment has been investigated in the residential microgrid in the presence of rooftop photovoltaic units [10]. An autonomous energy consumption scheduling algorithm has been proposed to decrease the peak load and reverse power flow. In some studies, the optimization of the capacity of local energy sources has been investigated. For instance, in [11], the simultaneous capacity optimization of distributed generation and energy storage system has been solved by a sequential quadratic programming method. Both the grid-connected and stand-alone microgrid has been considered as the main network of study. The minimization of the total annual energy losses and the reduction of the cost of the energy have been investigated during the optimization. In another study, a bi-level programming model has been proposed to optimize the location and size of the battery energy storage system by a numerical optimization algorithm [12]. The microgrid has been considered in the presence of a wind farm. The result of numerical simulations indicates that the located battery at the load center has higher performance than the located battery at the wind farm. In [13], an offshore wind farm has been considered. In this work, the unit commitment has been studied in the presence of high wind penetration scenarios in order to increase hourly spinning reserve capacity for covering the uncertainty of wind power and load. Garlik and Krivan formulated a unit commitment optimization problem for renewable energy units in a microgrid in [14]. The proposed unit commitment has been done using a simulated annealing heuristic optimization technique. In another study, the modified particle swarm optimization algorithm has been used for optimal unit commitment of renewable energy sources including wind turbines, photovoltaic panels, and combined heat and power plants in a microgrid [15]. Minimizing the cost of the microgrid is the main objective function of the
4 Optimal Dispatch and Unit Commitment in Microgrids
99
proposed unit commitment in this study. Experimental studies of microgrid and its devices have been presented in some papers. For instance, in [16], multi-objective optimization has been presented for simultaneously minimizing the total cost of the distribution system and maximizing the stability of the microgrid. The particle swarm optimization has been utilized to find the best combination of battery and hydrogen storage system based on the total cost, the occupied area, legislation, and local pollution. According to previous studies, it can be said that optimizing as a multi-objective function and considering both supplier and consumer sides of the microgrid are the topics that have been less studied. The innovative contributions of this study are highlighted as follows: • Simultaneously optimizing both local energy sources and loads of the microgrid. • Multi-objective optimization of the proposed method by considering an economic–environmental objective function. • Considering renewable and nonrenewable energy sources. • Utilizing a new intelligent algorithm based on the grey wolf optimization algorithm and the fuzzy method for multi-objective optimizing the unit commitment. So in this chapter, a multi-objective optimization algorithm is proposed to optimal manage the operational schedule of the energy sources of the microgrid in order to improve the performance of the MG. Renewable distributed generation such as wind turbine and photovoltaic and nonrenewable units such as micro turbine and fuel cell and also energy storage system are energy sources of the microgrid. For better managing the MG, a price-based demand response program is also considered in the proposed method. The considered objective function of the optimization consists of economic and environmental indices. A new method based on the grey wolf optimization algorithm and fuzzy method is introduced for multi-objective optimizing the hourly performance of energy sources of the microgrid in the presence of a demand response program in order to improve the efficiency of the MG. Finally, the proposed unit commitment is applied to a sample microgrid. Ultimately, the efficiency of the proposed method is pondered based on the simulation results.
4.2 Microgrid In the last decade, microgrid has been introduced for better controlling the total power network. Figure 4.1 shows a sample power system. This grid is divided into the number of microgrids in order to increase the performance of the system. Each microgrid has some energy sources and loads that should be managed optimally based on technical, economic, and environmental parameters. Thus, microgrids are utilized in the power network for improving the local reliability and flexibility of electric power systems so that the total grid is operated efficiently if each of the microgrids is managed and operated optimally. Although
100
H. Shayeghi and M. Alilou
Microgrid
Microgrid
Upper network
Microgrid
Microgrid
Fig. 4.1 A sample power system with the number of microgrids
microgrids can operate in islanded mode, microgrids in grid-connected mode have a useful ability for buying power from the upstream network when the demand of the microgrid is more than the produced power of local energy sources. Moreover, the microgrid can sell its extra energy to the upstream network. In this chapter, the grid-connected microgrid is studied.
4.3 Demand Response Program Demand response is a method for increasing the participation of consumers in managing the microgrid. The proper applying the DR programs to the microgrid causes to shave the peak demand, manage risk and reliability, and reduce carbon emission and energy cost. Totally, DR programs can be divided into Time/Price-based programs and Incentive-based programs. The various methods for applying the DR program are demonstrated in Fig. 4.2 [17]. The time or price-based DR programs use the variation of electricity cost during the 24-hour as a control signal to affect the electricity consumption of consumers while Incentive-based DR programs utilize the incentive and penalty to encourage customers for managing the load profile. In this chapter, the real-time pricing (RTP)
4 Optimal Dispatch and Unit Commitment in Microgrids
101
Time/Price DR programs Emergency DR
Time of use
Capacity market
Real time pricing
Ancillary service
Demand response programs
Critical peak pricing
Demand bidding Interruptible service Direct load control Incentive DR programs
Fig. 4.2 The different methods of demand response programs
method, which is a price based DR program, is utilized to participate the consumers in the demand side management. Electricity tariff varies continuously in the RTP program based on the climate zone, seasonal weather, and time. The dynamic price is available to the consumers one day ahead or an hour. This DR program causes that the distribution company of MG better distribute the electricity tariff reflecting the demand-supply elastics. Thus, the RTP program encourages the consumers to optimally manage the load profile and decrease their energy consumption during peak times or move the period of energy use to off-peak hours. Therefore, the RTP program is used as a demand response program in this chapter because it is one of the common and useful DR programs in microgrids. Equation (4.1) presents the considered model of demand response program [17]. ⎡
⎤
P (j ) − P0 (j ) ⎥ P (i) − P0 (i) 24 ⎢ d(i) = d0 (i) × ⎣1 + E (i, i) + E (i, j ) ⎦ j = 1 P0 (i) P0 (j ) j = i (4.1) In Eq. (4.1), E(i, i) and E(i, j) are the self-elasticity and cross-elasticity of demand response, respectively. d0 (i) and d(i) are the demand of system (Consumers) before and after applying for the RTP program, respectively. Moreover, the initial price of each hour is represented by using P0 (i) while P(i) demonstrates the price of an hour i after applying the RTP program.
102
H. Shayeghi and M. Alilou
4.4 Distributed Generation Units Distributed generation units are one of the important technologies of microgrids because the local production of electricity is the main proviso for calling a system as a microgrid. The produced power of DG sources increases the reliability and independence of the microgrid. The distribution company of MG utilizes the energy of these technologies for supplying the demand of the MG and selling energy to the upstream network. Of course, the distribution company has to buy energy from the upstream grid when the local sources cannot provide the MG’s demand. Totally, distributed generation units can be divided into nonrenewable and renewable units. Both types of DG units are described in the following [18, 19].
4.4.1 Nonrenewable Units The produced energy of nonrenewable units is stable and dispatchable because they can generate the electricity until their initial energy is supplied. So, this type of DG unit has a high effect on the reliability and stability of microgrids. In this chapter, two technologies of nonrenewable units including micro turbine (MT) and fuel cell (FC) are considered in the microgrids. The description and modeling of MT and FC are presented as following [18, 19]:
4.4.1.1
Micro Turbine
Micro turbine is one of the useful technologies of nonrenewable units. An overview of micro turbine is shown in Fig. 4.3. Totally, the micro turbine is a small turbine like a jet engine that has the capability to operate on a variety of gaseous and liquid fuels. This small turbine is connected to an electric generator. The combination of a small turbine, electric generator, power electronic devices, and control equipment is called micro turbine. The not nature dependent behavior of micro turbine helps the operator of the microgrid for providing the demand of the grid in urgently times using local energy sources instead of buying energy from the upstream network. Moreover, the MT has the unique ability to simultaneously produce electricity and heat. Micro turbine can run on a variety of fuels, including natural gas, propane, and fuel oil. Also, MTs have the ability to inject both active and reactive powers into the network.
4.4.1.2
Fuel Cell
A fuel cell is a nonrenewable DG technology that converts the chemical energy from a fuel into electricity. Figure 4.4 demonstrates an overview of fuel cells.
4 Optimal Dispatch and Unit Commitment in Microgrids
103
Exhaust Heat
Heat Exchanger
Gas compressor Fuel input Produced electricity
Combustor Air
Generator
Turbine
Compressor
Fig. 4.3 An overview of micro turbine Produced Electricity e–
e–
Hydrogen In
e–
Oxygen In
e– e–
e–
e– e– e– e–
e–
e– Water and Heat Out Anode
Cathode Electrolyte
Fig. 4.4 An overview of fuel cells
The fuel cell can provide highly reliable electricity so that it produces almost no pollutants. The efficiency of fuel cells is more than 85%. Fuel cells produce only water and heat in their energy-generating process. Thus, the fuel cell can be said as an environmentally friendly technology.
104
H. Shayeghi and M. Alilou
It is worth mentioning that the technology of a fuel cell is different from the technology of a battery because the FC can produce electricity continuously for as long as fuel is supplied while in the battery, the electrical energy comes from chemicals already present. The fuel cell produces just active power.
4.4.2 Renewable Units Renewable distributed generation units are called to technologies that their initial energy such as solar irradiance will never be exhausted. These units are also called Green because they have an eco-friendly technology. Although renewable units improve environmental indices of the microgrid, one of the main disadvantages of these technologies is their unstable behavior due to their initial energy. In the following, description and modeling of two common types of renewable units, photovoltaic panel (PV) and wind turbine (WT), are presented [18, 19].
4.4.2.1
Photovoltaic Panel
Photovoltaic is one of the popular technologies of renewable DGs, especially in rural and urban areas. PV is a solar power unit that uses solar cells or solar photovoltaic arrays to convert the light of the sun (solar irradiance) directly into electrical power. An overview of PV panels is shown in Fig. 4.5. PV panels are the practical choice for providing the electricity demand of remote areas due to the availability of solar energy approximately all points of the world. The output power of each PV panel relates to the amount of solar irradiance, the area, and the efficiency of the solar panel. Mathematically, the active power of the photovoltaic panel can be calculated by Eq. (4.2). PP V = Apv βμ
(4.2)
In this equation, β is solar irradiance. The parameters Apv and μ show the area and efficiency of the solar panel, respectively. PV panels can produce only active power.
4.4.2.2
Wind Turbine
Renewable wind energy is approximately available in most parts of the world during the year. For this reason, it is one of the proper sources for producing electricity. The technology which is used to convert wind energy into electrical energy is called the wind turbine. Figure 4.6 demonstrates an overview of wind turbines. The output power of WT has a direct relation to wind speed and swept area of the turbine. Moreover, air density and power coefficient affect the power of the
4 Optimal Dispatch and Unit Commitment in Microgrids
105
Produced electricity
Solar irradiance
Photons Electron flow N-type silicon Junction P-type silicon
Fig. 4.5 An overview of photovoltaic panels
Propeller leg Brake
Electricity Mechanical regulation system transfer Housing
Rotor management system
Generator Turning system
Tower
Foundation
Fig. 4.6 An overview of wind turbine
Connection to the grid
106
H. Shayeghi and M. Alilou
wind turbine. Therefore, Eq. (4.3) can be utilized to calculate the active power of WT units. PW T =
1 ρAwt Vw3 Cp 2
(4.3)
Here, ρ and VW are air density and wind speed, respectively. The parameters Awt and Cp demonstrate swept area and power coefficient of the wind turbine, respectively. It is worth mentioning that the wind turbine consumes reactive power to produce active power due to its induction generator.
4.5 Energy Storage System Energy storage system (ESS) is a device that has the ability to store electrical energy in the form of chemical energy and convert chemical energy into electricity when required. The optimal charging and discharging schedule of energy storage systems based on the condition of market price, the demand of the grid, and state of charge can significantly improve the technical, economic, and environmental indices of the microgrid. An overview of the energy storage system is shown in Fig. 4.7. As mentioned earlier, one of the main advantages of microgrids is the ability of local electricity production. The combination of distributed generation units and energy storage system increases the level of self-consumed electricity in the microgrid. With an ESS, the extra electrical energy of the MG is stored and used later. Totally, the microgrid equipped with both DG and ESS units can reduce the energy drawn from the upstream network. Therefore, the self-sufficiency of the microgrid is increased. Moreover, using the ESS beside the DG units decrease the dependence of the microgrids on the upper grid as well as reducing carbon emissions [20].
DC current Temperature controller
DC switch
AC breaker
AC transformer AC current
Storage
Monitor and control
Power conversion system
Fig. 4.7 An overview of the energy storage system
4 Optimal Dispatch and Unit Commitment in Microgrids
107
4.6 Objective Functions and Constraints In this chapter, the operational schedule of energy sources is optimized as a multiobjective problem with considering some constraints and predetermined parameters. In the following, the objective functions and constraints are explained completely.
4.6.1 Objective Functions An economic–environmental objective function is considered for finding the best unit commitment in the microgrid. Mathematically, the main objective function is presented in Eq. (4.4). Objectivefunction = max {PMGDC } , min {EMG }
(4.4)
Here, PMGDC is the profit of the microgrid distribution company and EMG is the pollution emission of the microgrid. They are formulated as follows.
4.6.1.1
The Profit of MGDC
The profit of the microgrid distribution company (MGDC) is considered as the technical index of the unit commitment. The MGDC has to provide the demand of the microgrid including a load of customers and the energy storage system. It is presumed that the MGDC owns the renewable and nonrenewable distributed generation units and the electrical storage system. So, the MGDC can use the power of DGs and ESS to provide the part of the total demand of the grid; another part of the demand is purchased from the upstream network. According to the abovementioned data, the profit of the MGDC is composed of the following terms: 1. 2. 3. 4. 5.
The income from selling the energy to the customers. The income from selling the energy to the upstream network. The cost of purchased energy from the upstream network. The cost of the produced energy of DG units. The cost of saving energy to the ESS in charging mode.
Therefore, the profit of the MGDC is formulated based on the incomes and costs in Eq. (4.5). PMGDC = ID + ISB − CW M − CDG − CESS
(4.5)
ID is the income from selling the energy to the customers. This income is presented in Eq. (4.6). In this equation, nc and PijD are the number of customers and their
108
H. Shayeghi and M. Alilou
consumption at each interval, respectively. PrM j is the electricity price (market price) at interval j. H is the number of time intervals in each day. It can be noted that the time interval in this chapter is 1 h; so H is equal to 24. ID =
nc H
PijD × PrM j
(4.6)
i=1 j =1
ISB is the income from selling the extra produced energy to the upstream network. So, when the sum of the produced power of DG units and the discharged energy of ESS is more than a load of the microgrid, the MGDC could sell the extra power to the upstream network. This income can be calculated by Eq. (4.7). In this equation, PjSB and PrSB j are the amount of sold power to the upstream network and the price of power at interval j, respectively. ISB =
H
PjSB × PrSB j
(4.7)
j =1
CWM is the cost of purchased energy from the upstream network. It is worth noting that the MGDC, firstly, provides the possible amount of energy of the microgrid using DGs and ESS; then it purchases extra energy from the upstream grid. This M cost is formulated in Eq. (4.8). In this equation, PjW M and PrW are the purchased j power from the upstream network and the price of purchased electricity at time interval j, respectively. CW M =
H
M PjW M × PrW j
(4.8)
j =1
CDG is the cost of the produced power of distributed generation units. This cost can be calculated by Eq. (4.9). In this equation, nt is the number of different types of DG units (Renewable and nonrenewable). nDG and PsDG are the number of DGs from i,j the considered type and their capacity at time interval j, respectively. PrDG is the s price of the production active power of the different types of DGs. CDG =
nt n H DG
PsDG × PrDG s i,j
(4.9)
s=1 i=1 j =1
CESS is the cost of the saved energy to the electrical storage system. It is worth noting here that the charging of the ESS unit has only cost for the MGDC. In other words, the discharging of ESS doesn’t have an extra cost for the MGDC because the MGDC is the owner of the ESS unit. Mathematically, the cost of charging the ESS Ch are the number of electrical is given by Eq. (4.10). In this equation, nESS and PESS ij storage systems and their charging energy at time interval j, respectively.
4 Optimal Dispatch and Unit Commitment in Microgrids
CESS =
n H ESS i=1 j =1
4.6.1.2
Ch M PESS × PrW j ij
109
(4.10)
Pollution Emission of MG
Nowadays, most countries have special attention to the environmental aspects of their decisions because the environmental condition gets worse every year. In this chapter, pollution emission of the microgrid is considered as the environmental issue of the optimization. Moreover, the pollution emission of central power plants is combined with the environmental index of the MG when the MGDC has to buy energy from the upstream network. The following five pollutant gasses are considered for calculating the pollutant emission of the MG: 1. 2. 3. 4. 5.
Carbon Monoxide (CO). Carbon Dioxide (CO2 ). Sulfur Dioxide (SO2 ). Nitrogen Oxides (NOx ). Particulate Matter (PM10 ).
Mathematically, the pollution emission of the microgrid considering the above pollutant gasses is calculated by Eq. (4.11). EMG =
H nunit nP G h=1
i=1
j =1
Puniti (h) × P Gij
(4.11)
Here, nunit and nPG are the number of energy sources (DG, ESS, and upstream network) and the number of pollutant gasses, respectively. Punit _ i (h) shows the power of unit i at hour h while PGij is the rate of pollutant gas j of unit i.
4.6.2 Constraints There are the following constraints during the implementation of the proposed unit commitment in the microgrid.
4.6.2.1
Power Balance Constraint
This constraint presents the sum of the total power of various types of distributed generation units (PDG ), the power energy storage system in discharge mode (PESS, Dis ) and the purchased power from the upstream grid (PG2MG ) should be equal to the sum of the total demand of the microgrid (Pdemand ), the power of the energy storage system in charge mode (PESS, Ch ) and the sold power to the upstream
110
H. Shayeghi and M. Alilou
network (PMG2G ). Mathematically, the power balance constraint is calculated by Eq. (4.12). nDG i=1
PDGi + PG2MG + PESS,Dis =
n j =1
Pdemandj + PMG2G + PESS,Ch (4.12)
4.6.2.2
Distributed Generation Constraint
The produced power of each type of DG unit should be in allowable size as the following range: PDG min ≤ PDGi ≤ PDG max
(4.13)
Here, PDG min and PDG max are the minimum and maximum power of each DG unit for producing the electricity, respectively.
4.6.2.3
Energy Storage System Constraint
There are limitations of charging and discharging in the energy storage system during each hour. The Eqs. (4.14–4.18) are considered as constraints of the ESS. XtESS,Ch + XtESS,Dis ≤ 1
(4.14)
ESS SOC min ≤ SOC max ESS ≤ SOC t ESS
(4.15)
Ch 0 ≤ PtESS,Ch ≤ RESS
(4.16)
Dis 0 ≤ PtESS,Dis ≤ RESS
(4.17)
ESS,Ch Ch Dis SOC ESS − PtESS,Dis × ηESS = SOC ESS × ηESS t t−1 + Pt
(4.18)
where, Eq. (4.14) shows that the charge and discharge of ESS are not simultaneous. In this equation, XtESS,Ch and XtESS,Dis indicate a binary variable that shows the charge and discharge state of ESS, respectively (0 or 1). The limitation of the total state of charge (SOC) of ESS is shown in Eq. (4.15). In this equation, SOC ESS t max demonstrates SOC of ESS at time interval t while SOC min ESS and SOC ESS indicate
4 Optimal Dispatch and Unit Commitment in Microgrids
111
the minimum and maximum rate of SOC of ESS, respectively. The maximum amount of charged power of the ESS from the grid and the maximum amount of discharged power of ESS to the grid are satisfied by Eq. (4.16) and Eq. (4.17), Ch and R Dis show the maximum charge and discharge rate of respectively. RESS ESS the ESS, respectively. Eq. (4.18) indicates that SOC of ESS at each time interval (SOC ESS ) consists of the remaining SOC of ESS from the previous interval t ESS,Ch and (SOC ESS t−1 ), the amount of power exchanged with the grid and ESS (Pt ESS,Dis Ch Dis Pt ) and the charge and discharge efficiency (ηESS and ηESS ).
4.6.3 Optimization Algorithm In this chapter, the combination of multi-objective grey wolf optimization (MOGWO) and fuzzy method is utilized to multi-objective optimize the operational schedule of energy sources of the microgrid. For getting the best result, firstly, the MOGWO is used to optimize the multi-objective functions including economic and environmental indices, and create the optimal Pareto-front. After applying the MOGWO algorithm, the fuzzy method is used to find the optimal particle from the non-dominated particles.
4.6.3.1
Multi-Objective Grey Wolf Optimization Algorithm
Intelligent algorithms are usually inspired by existing natural behaviors of nature. The optimization method of MOGWO is inspired by the grey wolves. Grey wolves that belong to the Canidae family are considered apex predators. This means that they are at the top of the food chain. Grey wolves mostly prefer to live in a pack with group size. The MOGWO is based on the social behavior of Wolves. In this meta-heuristic algorithm, the best solution is considered as α wolf. β and δ wolves are the second and third solutions, respectively. The rest of the particles are assumed as ω wolves. The optimization method of the MOGWO is managed by alpha, beta, and delta so that the omega wolves follow α, β, and δ wolves for reaching the best result. The position updating of search agents (ω wolves) based on the position of alpha, beta, and delta wolves is demonstrated in Fig. 4.8. So in this algorithm, the position of each particle is updated using Eq. (4.19) [21]. X (t + 1) = XP (t) − A.D
(4.19)
D = |C.XP (t) − X(t)|
(4.20)
Here,
112
H. Shayeghi and M. Alilou
C2
a1 a C2 R
a2 b
prey Dalpha Dbeta
Move
d
Ddelta
w
a3 C3
Fig. 4.8 Position updating of search agents (ω wolves) based on α, β, and δ in MOGWO algorithm
A = 2a.r1 − a
(4.21)
C = 2r2
(4.22)
Here, XP and X show the position vector of the prey and the position vector of a grey wolf, respectively. Vectors of A and C are the coefficients. Moreover, a linearly decrease from 2 to 0 over the iterations while vectors of r1 and r2 are random in [0 and 1]. The extra detail of the multi-objective grey wolf optimization algorithm is available in reference [21].
4.6.3.2
Fuzzy Method
After optimizing the economic–environmental objective functions of unit commitment in the microgrid, the fuzzy decision-making method run to select the optimal compromise solution that represents the optimal amount of economic and
4 Optimal Dispatch and Unit Commitment in Microgrids
113
environmental indices equal to the best operational schedule of various type of distributed generation units and energy storage system in the microgrid. In the fuzzy decision-making method, the best result is selected in three steps [22]: • Firstly, the membership values of each particle based on various objective functions are calculated by Eq. (4.23).
μi k =
⎧ ⎪ ⎨ ⎪ ⎩
Fi k ≤ Fi min Fi min < Fi k < Fi max
1 Fi max −Fi k Fi max −Fi min
0
Fi
max
≤ Fi
(4.23)
k
Here, Fi min and Fi max demonstrate the lower and upper bound of index i, respectively. Fi k shows the amount of particle k based on the objective function i. • Secondly, Eq. (4.24) is utilized to calculate the total membership value of each non-dominated particle which is in the Pareto-front. #N O k μi μk = #N K i=1 #N O k=1
i=1
μi k
(4.24)
• Thirdly, each particle that has the highest amount of total membership value is selected as the best compromise solution. Consequently, the complete method for optimizing the hourly energy schedule of various energy sources of the microgrid is demonstrated in Fig. 4.9. This figure presents that the best unit commitment in the microgrid is selected in four sections including (1) input details of the problem; (2) initial steps; (3) MOGWO steps; (4) fuzzy decision steps.
4.7 Numerical Results In this section, the proposed unit commitment method is tested on a sample microgrid. The hourly electricity consumption of the microgrid is demonstrated in Fig. 4.10. The distribution company of microgrid has a micro turbine, fuel cell, wind turbine, photovoltaic panel, and energy storage system. Thus, the distribution company of microgrid can utilize from the produced power of energy sources of the microgrid and the bought energy from the upstream grid for providing the electricity demand of the microgrid. The micro turbine is a CAPSTONE C1000S with a capacity of 1000 KW. The HD85 type of BALLARD Company is considered as fuel cell units of the microgrid.
114
H. Shayeghi and M. Alilou Start
Input details of the problem including: • Hourly loads of the microgrid • Details of DG units and ESS • Hourly variation of solar irradiance and wind speed • Economic and environmental parameters • Parameters of considered constraints • Details of demand response program Initial steps including: • Create random particles • Evaluate the constraints • Calculate the amounts of objective functions MOGWO steps including: • Select the alpha, beta and delta particles • Update position of particles based on MOGWO method • Evaluate the constraints • Calculate the amount of objective functions • Update the iteration counter • Repeat above steps until reaching the maximum iteration Fuzzy decision steps including: • Calculate the membership values of each non-dominated particle • Calculate the total membership value of each particle • Select the best particle with the highest amount of total membership value • Determine the best unit commitment in the microgrid • Calculate the amount of economic and environmental objective functions End
Fig. 4.9 Flowchart of the unit commitment in the microgrid
This type of fuel cell unit can inject 85 KW to the grid. The used wind turbine in this chapter is AW1500 type of NORDEX Company with 1500 KW capacity. Moreover, the cut-in, normal, and cutout speeds are 4, 13, and 25 m/s, respectively [23, 24]. The 310 OS5 types of IBC-SOLAR Company are utilized for producing electrical energy from solar power. The capacity of the PV panel is 310 KW. Moreover, the energy storage system of ECCINC Company with a capacity of 400 KWh is operated as an ESS unit of the microgrid. The hourly charging and discharging limit of ESS battery is 100 KW. According to variations of wind speed and solar irradiance [6], the hourly abilities of wind turbines and photovoltaic panels for producing electrical energy are shown in Figs. 4.11 and 4.12, respectively. Figure 4.13 shows the hourly market price of the microgrid. This price is based on the RTP method. Moreover, The costs of producing 1 Kwh using a micro turbine, fuel cell, wind turbine, and photovoltaic are 0.043, 0.039, 0.037, and 0.038 $,
4 Optimal Dispatch and Unit Commitment in Microgrids
115
4000 3500 3000
Kw
2500 2000 1500 1000 500 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour Fig. 4.10 Hourly demand for considered microgrid 1600 1400 1200
Kw
1000 800
600 400 200
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour Fig. 4.11 Hourly ability of WT for producing electricity
respectively. Microgrid buys energy from the upstream network with a price of 0.12 $/Kwh. Moreover, microgrid sells energy to the upstream grid with the mean price of the daily market price. The amounts of emitted pollutant gases (carbon monoxide, carbon dioxide, sulfur dioxide, nitrogen oxides, and particulate matter) of renewable and nonrenewable DGs, ESS and central power plant are presented in Table 4.1. For better evaluating the results, the proposed unit commitment method is evaluated in the sample microgrid without and with considering the demand response program in the sections 4.8.1 and 4.8.2, respectively.
H. Shayeghi and M. Alilou
Kw
116 350 325 300 275 250 225 200 175 150 125 100 75 50 25 0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour
$/Kwh
Fig. 4.12 Hourly ability of PV for producing electricity 0.6 0.55 0.5 0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour Fig. 4.13 Hourly market price of electricity
4.7.1 Without DR Program In this section, it is considered that the demand response program is not applied to the microgrid. So, the hourly load variation of the microgrid is equal to Fig. 4.10. As mentioned earlier, the combination of the MOGWO and fuzzy method is utilized to optimize the economic–environmental objective function and find the best unit commitment in the considered microgrid. The optimal hourly power of various energy sources of the microgrid without real-time pricing DR program is presented in Table 4.2.
4 Optimal Dispatch and Unit Commitment in Microgrids
117
Table 4.1 The environmental details of various energy sources Type of energy source Micro turbine Fuel cell Photovoltaic panel Wind turbine Energy storage system Central power plant (received energy from upstream network) Table 4.2 The optimal unit commitment in the microgrid without RTP program
Pollution gases rate (Kg/Kwh) CO2 SO2 NOx 0.72 0.002 0.091 0.46 0.012 0.006 0 0 0 0 0 0 0.02 0 0.00001 0.85 2.14 9.723
Hour 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
CO 0.247 0.002 0 0 0.0003 6.043
Power of energy resources (KW) MT FC PV WT 0.0000 0.00 0.000 1411.80 21.000 10.0 0.000 1429.35 221.00 71.0 0.000 1376.40 156.00 21.0 0.000 1411.80 69.000 74.0 0.000 1235.25 115.00 85.0 15.50 904.270 0.0000 0.00 46.50 1052.12 615.80 2.00 104.0 1235.25 1000.0 41.0 201.5 1147.05 1000.0 12.0 279.0 882.300 1000.0 85.0 303.8 705.900 1000.0 85.0 310.0 794.100 1000.0 85.0 279.0 882.300 1000.0 85.0 217.0 1023.60 1000.0 85.0 170.5 1058.85 1000.0 85.0 93.00 1235.25 1000.0 85.0 62.00 1500.00 1000.0 85.0 15.50 1358.85 1000.0 85.0 0.000 1235.25 1000.0 85.0 0.000 1411.80 1000.0 85.0 0.000 1376.40 1000.0 85.0 0.000 1235.25 1000.0 85.0 0.000 1129.35 1000.0 59.0 0.000 1200.00
ESS 100.0 0.000 −100.0 −100.0 0.000 −100.0 66.78 47.00 −37.00 29.42 0.000 0.000 0.000 100.0 100.0 100.0 −100.0 0.000 0.000 100.0 100.0 100.0 −100.0 −14.00
PM10 0.018 0 0 0 0.001 0.87
Grid 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 419.700 1256.23 1307.46 1177.44 1216.59 1300.65 1056.12 803.930 745.580 1030.68 681.316 716.716 493.424 435.255 0.00000
According to Table 4.2, the renewable DG units have an important role in power generation; the produced power of wind turbines and photovoltaic panels is about 48.97% of total provided energy. Of course, the wind turbine unit produces more electricity than a photovoltaic panel. According to the result, the produced power of wind turbines is 93% of total daily renewable energy while the produced power of
118
H. Shayeghi and M. Alilou 48.97 50 45 40 35
30.15
30
20.41
% 25 20
15 10
0.47
5 0
Nonrenewable DGs
Renewable DGs
ESS
grid
Fig. 4.14 The percentage participation of various types of sources in unit commitment
the photovoltaic panel is only 7% of total renewable electrical power. The 24.36% of the load is produced by nonrenewable DG units. The energy storage system provides only 0.47% of total electricity. Although the upstream grid provided the total electricity demand of the microgrid before utilizing DG units and the optimal unit commitment, the provided power by the grid is about 20% of total demand after applying the proposed method to the microgrid. In other words, energy sources of the microgrid can provide the load of system at most times; therefore, it can be said that the upstream grid is the backup source after doing the proposed unit commitment. Moreover, it is worth mentioning that the microgrid sells electrical power to the upstream network at some hours. According to statistics, the microgrid has the ability to inject about 2718 KWh electrical energy into the upstream grid during the day. The percentage participation of various energy sources in the proposed unit commitment in the microgrid without considering the demand response program is demonstrated in Fig. 4.14. Table 4.3 demonstrates hourly amounts of economic and environmental indices of the microgrid with and without considering the proposed unit commitment method. Results demonstrate that the economic and environmental indices of the microgrid are improved after applying the proposed unit commitment method. The hourly profits of the distribution company of the microgrid increase between 14% and 677% after utilizing the energy sources using the optimal unit commitment. The daily statistic data of the economic index is presented in Table 4.4. As can be shown in this table, the daily profit of the microgrid is about 4500$ more than the initial case. Moreover, the minimum and maximum amounts of hourly profit of the microgrid are increased so that the daily average of the considered economic index is improved by about 24.95% after utilizing the proposed unit commitment. Thus, the distribution company of the microgrid earns more profit after optimal utilizing the nonrenewable and renewable DG sources and energy storage system.
4 Optimal Dispatch and Unit Commitment in Microgrids
119
Table 4.3 Hourly amounts of economic and environmental indices of the microgrid without RTP program Hour 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Without DGs and ESS Profit ($) Pollution emission (Kg) 346.69866 15633.63315 209.77119 14714.28358 87.406520 13794.93402 30.593025 12875.58446 52.082443 11956.23489 158.06396 12875.58446 238.90607 14714.28358 515.74230 22073.77065 655.54890 27589.86804 812.94974 33110.65598 1072.2976 42308.84218 1153.8641 44147.54131 1095.5981 43228.19174 1147.2886 45986.24044 1133.0750 46905.59000 1088.6585 45066.89087 1139.3162 42308.84218 1217.8736 40465.45249 1440.8999 42308.84218 1163.7312 41389.49262 1081.7783 41389.49262 961.51258 36788.05424 777.62937 32191.30642 530.29396 23912.46978
With unit commitment Profit ($) Pollution emission (Kg) 550.374473 2.1310000 411.147892 27.438000 302.747966 270.18700 237.969899 176.11700 283.353635 109.90200 198.899998 162.63900 338.058257 1.4229752 749.641502 665.74007 874.221344 1096.8915 990.560855 6383.5191 1239.68390 16979.959 1329.07944 17626.765 1275.59198 15985.208 1345.92639 16481.546 1330.82555 17542.937 1294.69521 14455.527 1301.55603 11267.089 1415.81414 10532.499 1627.31065 14132.165 1376.79558 9723.2268 1291.90448 10170.187 1159.92334 7350.9087 904.021256 6612.1923 800.100575 1106.0217
Table 4.4 Daily statistics parameters of the economic index of the microgrid
Sum of hourly profits Minimum of hourly profits Maximum of hourly profits Average of hourly profits
Profit ($) Without DGs and ESS 18111.58 30.59000 1440.900 754.6500
With unit commitment 22630.21 198.9100 1627.310 942.9300
The environmental index of the microgrid is improved considerably after applying the proposed unit commitment method. The impact of renewable DG sources in this improvement is significant. According to the results, the emitted pollutant gasses of the microgrid are reduced by about 76% after utilizing the DGs and ESS. Minimum hourly amounts of the environmental index without and with unit commitment are 11.96 Mg and 1.42 Kg, respectively. On the other hand, the
120
H. Shayeghi and M. Alilou
maximum hourly amounts of this index without and with unit commitment are 46.91 Mg and 17.62 Mg, respectively. Totally, it can be said that the proposed unit commitment method considerably improves the economic and environmental indices of the microgrid without considering the demand response program using the nonrenewable and renewable DG units and energy storage system.
4.7.2 With DR Program In this section, the unit commitment is applied to the microgrid in the presence of a demand response program. So, in this situation, the consumers of the microgrid are encouraged to improve their consumption curve based on the real-time pricing method. After applying for the demand response program, the hourly load variation of the microgrid becomes more linear than without considering the demand response program. The consumption pattern of the microgrid in the presence of the RTP program is demonstrated in Fig. 4.15. As can be shown in this figure, the difference between the minimum and maximum electricity demand of microgrid is considerably decreased by participating the customers in a demand response program. This significant reduction causes to improve the efficiency of the electricity and increase the performance of the microgrid. As mentioned earlier, the unit commitment of the microgrid in the presence of the RTP program is optimized using the combination method of multi-objective grey wolf optimization and fuzzy method. The hourly produced power of various energy sources is presented in Table 4.5. According to this table, the impact of
3500 3000
Kw
2500 2000 1500 1000 500 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour Fig. 4.15 Improved demand of the microgrid after applying for the DR program
4 Optimal Dispatch and Unit Commitment in Microgrids Table 4.5 The optimal unit commitment in the microgrid with RTP program
Hour 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Power of energy resources (KW) MT FC PV WT 651.76 85 0.000 1399.19 656.00 18 0.000 1429.35 1000.0 85 0.000 1376.40 569.92 85 0.000 1411.80 1000.0 85 0.000 1235.25 1000.0 85 15.50 1252.95 1000.0 85 46.50 1252.95 254.75 82 124.0 1235.25 914.45 81 201.5 1147.05 1000.0 85 279.0 882.300 1000.0 85 303.8 705.900 1000.0 85 310.0 794.100 1000.0 85 279.0 882.300 1000.0 85 217.0 1023.60 1000.0 85 170.5 1058.85 1000.0 85 93.00 1235.25 1000.0 85 62.00 1500.00 1000.0 85 15.50 1358.85 1000.0 85 0.000 1235.25 1000.0 85 0.000 1411.80 984.39 16 0.000 1376.40 1000.0 85 0.000 1235.25 1000.0 85 0.000 1129.35 1000.0 85 0.000 1200.00
121
ESS 73.19 0.000 44.00 −100.0 −100.0 −100.0 47.00 89.00 34.00 100.0 70.00 0.000 100.0 100.0 100.0 0.000 −100.0 100.0 18.12 −100.0 100.0 56.00 0.000 100.0
Grid 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 348.932 312.091 578.946 858.631 415.260 644.951 573.238 148.233 427.138 793.750 0.00000 0.00000 173.355 335.254 310.232
renewable energy sources is increased than the microgrid without the demand response program. Totally, the produced power of wind turbines and photovoltaic panels is about 50.34% of all provided energy. The participation of renewable units is increased by about 3% than the situation that the RTP program is not considered. Moreover, the produced power of wind turbines is 93.14% of total daily renewable energy while the produced power of photovoltaic panels is only 6.86% of total renewable electrical power. The difference between the produced power of the wind turbine and the photovoltaic panel is related to their capacity and availability of their initial energy including wind speed and solar irradiance. The participation of nonrenewable DG units is also increased when the demand response program is utilized in the microgrid. The 38.99% of total electrical energy is produced by nonrenewable DG units. Therefore, the produced power of this type of DG technologies is increased by about 30% than the microgrid without the demand response program. The injected power of the energy storage system is also raised more than 100% so that it provides 1.03% of total electrical energy. Although the upstream grid provided about 20% of the total demand of the microgrid without applying for the demand response program, the participation of it
122
H. Shayeghi and M. Alilou
50 45 40 35 30 % 25 20 15 10 5 0
46.88 35.91
9.64 3.09 MT
FC
3.45 PV
1.03 WT
ESS
Grid
Fig. 4.16 The percentage participation of energy sources in unit commitment in the presence of DR program
in providing the daily demand is only 9% after utilizing both unit commitment and demand response program. Thus the impact of the upstream grid in providing the demand of the microgrid is reduced 53% using the demand response program. The percentage participation of nonrenewable and renewable distributed generation units, energy storage systems and grid in the proposed unit commitment in the microgrid with considering the real-time pricing program is demonstrated in Fig. 4.16. The wind turbine has the highest impact on providing the daily demand of the microgrid. The hourly amounts of economic and environmental indices of the microgrid in the presence of a demand response program with and without considering the proposed unit commitment method are presented in Table 4.6. According to the results, the profit of the distribution company and the pollution gasses of the microgrid are improved after utilizing the proposed unit commitment method and demand response program. According to the results of Table 4.6, the hourly profits of the distribution company of the microgrid increase between 15% and 250% after utilizing the energy sources using the optimal unit commitment. The daily statistic data of the economic index is presented in Table 4.7. As can be shown in Table 4.7, the daily profit of the microgrid is about 4914.8$ more than the initial case. Moreover, the minimum and maximum amounts of hourly profit of the microgrid are increased so that the daily average of the considered economic index is improved by about 29.11% after utilizing the proposed unit commitment. Thus, the distribution company of the microgrid earns more profit after optimal utilizing the nonrenewable and renewable DG sources and energy storage system using the proposed unit commitment method. Moreover, the demand response program increases the effect of the energy sources of the microgrid on the energy management of the microgrid. The environmental index of the microgrid is also improved considerably after applying for both the unit commitment method and demand response program.
4 Optimal Dispatch and Unit Commitment in Microgrids
123
Table 4.6 Hourly amount of economic and environmental indices of the microgrid with RTP program Hour 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Without DGs and ESS Profit ($) Pollution emission (Kg) 611.84564 27589.8681 377.61489 26487.5866 151.51256 23912.4698 59.001630 24831.8193 100.16011 22993.1202 277.77798 22627.2566 367.38378 22627.2566 515.74231 22073.7706 655.54890 27589.8681 835.52208 34030.0055 792.57296 31271.9569 913.45534 34949.3551 1025.5777 40465.4525 894.87106 35868.7047 933.08726 38626.7534 910.87899 37707.4038 916.37905 34030.0056 1134.8656 37707.4038 1346.8101 39546.1029 879.26063 31271.9569 817.34086 31271.9569 841.36949 32191.3064 777.62937 32191.3064 754.66510 34030.0056
With unit commitment Profit ($) Pollution emission (Kg) 800.36088 744.95089 549.71568 715.80800 521.42321 1119.7376 175.72078 653.04383 344.02882 1116.6690 541.31707 1116.6690 738.80195 1119.8015 675.37263 315.87709 900.83861 1025.3816 1027.5160 5526.5527 968.35926 5060.7463 1088.6706 8428.5785 1217.5716 11961.999 1093.5089 6364.0100 1130.8378 9264.1012 1104.9157 8356.5093 1078.6189 2988.2525 1344.8062 6513.9817 1535.3948 11141.073 1007.3151 1116.6690 1020.6761 1070.9839 1034.5002 3308.7672 955.25042 5351.7233 950.15010 5037.9265
Table 4.7 Daily statistics parameters of the economic index of the microgrid in the presence of the RTP program
Sum of hourly profits Minimum of hourly profits Maximum of hourly profits Average of hourly profits
Profit ($) Without DGs and ESS 16890.87 59.00000 1346.810 703.7800
With unit commitment 21805.67 175.7200 1535.390 908.5600
The impact of renewable DG sources in this improvement is significant because their pollution is zero. According to the results, the emitted pollutant gasses of the microgrid are reduced by about 86.67% after utilizing the DGs and ESS. Moreover, the environmental index is also decreased by about 13.91% than the situation that unit commitment is utilized in the microgrid without considering the demand response program. Minimum hourly amounts of the environmental index without
124
H. Shayeghi and M. Alilou
and with unit commitment are 22.07 Mg and 0.31 Mg, respectively. On the other hand, the maximum hourly amounts of this index without and with unit commitment are 40.46 Mg and 11.96 Mg, respectively. Therefore, the combination of the proposed unit commitment method and demand response program significantly increases the economic index and decreases the environmental index of the microgrid.
4.8 Conclusion Optimal dispatch and unit commitment of various energy sources of the microgrid has a high effect on the efficiency of both microgrid and upstream grid. Moreover, increasing the interest of consumers for optimal managing the consumption using demand response programs increases the performance of the microgrid. For these reasons, the hourly operational schedule of energy sources of the microgrid was optimized with and without the demand response program in this chapter. Micro turbine, fuel cell, wind turbine, photovoltaic panel, and energy storage system were the energy sources of the microgrid which provide the demand of the microgrid. Also, the microgrid could buy/sell energy from/to the upstream grid when the produced power of energy sources of the microgrid is lower/more than the demand of the microgrid. The combination of multi-objective grey wolf optimization algorithm and the fuzzy set method was used to optimize the economic–environmental objective function and select the best operational schedule of energy units. The numerical results demonstrate that the proposed method can properly optimize a schedule for unit commitment in the microgrid so that considered indices of the microgrid are improved considerably after applying the proposed method. The economic index of the microgrid is increased by about 25–29% in various cases. On the other hand, the considered environmental index is decreased by about 76–87% in different tests. Based on the participation of energy sources in unit commitment, it can be said that renewable units have a high effect on providing the demand of the microgrid. The wind turbine produces the highest amount of electricity during the day. The demand response program has also a positive effect on the performance of the microgrid. Totally, the results prove that the proposed unit commitment method has a suitable performance in increasing the efficiency of the microgrid.
References 1. Shayeghi, H., & Alilou, M. (2015). Application of multi objective HFAPSO algorithm for simultaneous placement of DG, capacitor and protective device in radial distribution network. Journal of Operation and Automation in Power Engineering, 3, 131–146.
4 Optimal Dispatch and Unit Commitment in Microgrids
125
2. Shayeghi, H., & Shahryari, E. (2017). Integration and management technique of renewable energy resources in microgrid. Energy Harvesting and Energy Efficiency, 37, 393–421. 3. Gelazanskas, L., & Gamage, A. (2014). Demand side Management in Smart Grid: A review and proposals for future direction. Sustainable Cities and Society, 11, 22–30. 4. Behrangrad, M. (2015). A review of demand side management business models in the electricity market. Renewable and Sustainable Energy Reviews, 47, 270–283. 5. Heydarian, E., & Aalami, H. A. (2016). Multi objective scheduling of utility-scale energy storages and demand response programs portfolio for grid integration of wind power. Journal of Operation and Automation in Power Engineering, 4, 104–116. 6. Alilou, M., Nazarpour, D., & Shayeghi, H. (2018). Multi-objective optimization of demand side management and multi DG in the distribution system with demand response. Journal of Operation and Automation in Power Engineering, 6, 230–242. 7. Shakouri, H., & Kazemi, A. (2017). Multi-objective cost-load optimization for demand side Management of a Residential Area in smart grids. Sustainable Cities and Society, 32, 171– 180. 8. Wu, Z., Tazvinga, H., & Xia, X. (2015). Demand side Management of Photovoltaic-Battery Hybrid System. Applied Energy, 148, 294–304. 9. Yao, E., Samadi, P., Wong, V., & Schober, R. (2016). Residential demand side management under high penetration of rooftop photovoltaic units. IEEE Transactions on Smart Grid, 7, 1597–1608. 10. Kotur, D., & Durisic, Z. (2017). Optimal spatial and temporal demand side Management in a Power System Comprising Renewable Energy Sources. Renewable Energy, 108, 533–547. 11. Sfikas, E., Katsigiannis, Y., & Georgilakis, P. (2015). Simultaneous capacity optimization of distributed generation and storage in medium voltage microgrids. Electrical Power and Energy Systems, 67, 101–113. 12. Xu, G., Cheng, H., Fang, S., Ma, Z., Zeng, P., & Yao, L. (2018). Optimal size and location of battery energy storage Systems for Reducing the wind power curtailments. Electric Power Components and Systems, 46, 342–352. 13. Y. Wu, Sh. Chang, L. Chang, D. Viet, “Unit commitment in a high wind-power penetration system”, Energy Procedia, Vo. 156, pp.18–22, 2019. 14. Garlik, B., & Krivan, M. (2013). Renewable energy unit commitment, with different acceptance of balanced power, solved by simulated annealing. Energy and Buildings, 67, 392–402. 15. Gholami, K., & Dehnavi, E. (2019). A modified particle swarm optimization algorithm for scheduling renewable generation in a micro-grid under load uncertainty. Applied Soft Computing, 78, 496–514. 16. Avril, S., Arnaud, G., Florentin, A., & Vinard, M. (2010). Multi-objective optimization of batteries and hydrogen storage Technologies for Remote Photovoltaic Systems. Energy, 35, 5300–5308. 17. Aalami, H., Moghaddam, M., & Yousefi, G. (2010). Modeling and prioritizing demand response programs in power markets. Electric Power Systems Research, 80, 426–435. 18. Viral, R., & Khatod, D. (2012). Optimal planning of distributed generation systems in distribution system: A review. Renewable and Sustainable Energy Reviews, 16, 5146–5165. 19. Maczulak, A. (2010). Renewable energy sources and methods. Green Technology - Facts on File, Inc.. 20. Akbari, H., Browne, M. C., Ortega, A., Huang, M. J., Hewitt, N. J., Norton, B., & McCormack, S. J. (2019). Efficient energy storage Technologies for Photovoltaic Systems. Solar Energy, 192, 144–168. 21. Mirjalili, S., Saremi, S., Mirjalili, M., & Coelho, L. (2016). Multi-objective grey wolf optimizer: A novel algorithm for multi-criterion optimization. Expert Systems with Applications, 47, 106–119. 22. Alilou, M., Talavat, V., & Shayeghi, H. (2018). Simultaneous placement of renewable DGs and protective devices for improving the loss, reliability and economic indices of distribution system with nonlinear load model. International Journal of Ambient Energy, In Press.
126
H. Shayeghi and M. Alilou
23. Modi, A., Bühler, F., Andreasen, J., & Haglind, F. (2017). A review of solar energy based heat and power generation systems. Renewable and Sustainable Energy Reviews, 67, 1047–1064. 24. Stathopoulos, T., Alrawashdeh, H., Al-Quraan, A., Blocken, B., Dilimulati, A., Paraschivoiu, M., & Pilay, P. (2018). Urban wind energy: Some views on potential and challenges. Journal of Wind Engineering and Industrial Aerodynamics, 179, 146–157.
Chapter 5
The Role of Energy Storage Systems in Microgrids Operation Sidun Fang and Yu Wang
5.1 Introduction 5.1.1 Background Generally, a microgrid can be defined as a local energy district that incorporates electricity, heat/cooling power, and other energy forms, and can work in connection with the traditional wide area synchronous grid (macrogrid) or “isolated mode” [1]. The flexible operation pattern makes the microgrid become an effective and efficient interface to integrate multiple energy sources, such as distributed generators, energy storage, and so on [2]. Additionally, with the development of transportation electrification, electrified vehicles, ships, or even aircraft become available, which introduces another type of special microgrids, that is, mobile microgrids [3]. This type of microgrids mostly work in autonomous conditions and sometimes they will connect to the main grid, that is, when electrified vehicles are charging, or electrified ships berth into a seaport. No matter which type of microgrid is, the grid-connected and islanded modes are two typical operation patterns, and to accomplish different tasks and needs, microgrids will supply power to different equipment, such as the cooling/heat equipment in a residential area, or the air-conditioning power for a data center, or the power consumed by port cranes in a seaport, or charging the parked electrical vehicles. In this way, the energy storage system (ESS) is an important component in a microgrid to act as an energy/power buffer between the generation side and demand side. Lots of literature focus on this topic and fundamentally prove the great effects of ESS in microgrid operation, that is, to facilitate the mitigation of
S. Fang () · Y. Wang School of Electrical and Electronics Engineering, Nanyang Technological University, Singapore, Singapore © Springer Nature Switzerland AG 2021 A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems, https://doi.org/10.1007/978-3-030-59750-4_5
127
128
S. Fang and Y. Wang
Fig. 5.1 Classification of microgrids
renewable energy source, or to benefit the autonomous operation, or to provide uninterruptible power to critical components, and so on [4]. In this chapter, the role of ESS in different types of microgrids will be illustrated in detail, that is, in both conventional land-based microgrids and mobile microgrids, and the microgrids discussed in this chapter are classified as the following Fig. 5.1.
5.1.2 Land-based Microgrids 5.1.2.1
Residential Microgrid
Residential microgrid is the most conventional type of microgrid since the concept of microgrid is raised. It is designed to supply power to a residential area, which may consist of several buildings and blocks. The following figure gives the topology of a typical AC/DC multi-energy microgrid (MEMG) [5]. From Fig. 5.2, the energy of MEMG is supplied from three sources, that is, photovoltaics(PVs), electrical substation, and gas pressure house. The energy from PVs is collected by the DC bus and the substation injects electricity to the AC bus. Additionally, to improve the system flexibility, a battery ESS, two thermal storages, and gas storage are incorporated. In the MEMG, multiple energy forms, that is, electricity, heat/cooling power, and gas are coupled together. Excess electricity can convert to gas by the power to gas (P2G) equipment and heating power by the power to thermal (P2T) equipment. The gas can convert to electricity by a combined heat plant (CHP) and the by-product is thermal power, or directly convert to thermal power by the gas boiler. In the MEMG, the generation side and demand side may not be always matched to each other, which motivates the investment of ESS. Additionally, with the development of transportation electrification, electric vehicles will become more popular in the future, and the charging power will become an important service
5 The Role of Energy Storage Systems in Microgrids Operation
129
Fig. 5.2 A typical AC/DC multi-energy residential microgrid [5] (Permission for usage from the author)
load in the MEMG. However, the arrivals/departures of electric vehicles are usually uncertain and cannot be accurately forecasted, which further inquires the optimal operation of ESS.
5.1.2.2
Industrial Microgrids
Industrial microgrids provide power to different commercial consumers, such as the data center, industrial park, and seaport. The following figure gives a typical structure of seaport microgrid [3]. Compared with MEMG, seaport microgrid is a newly proposed concept for seaport management after a high level of electrification [6]. It generally defines as a harbor territory that has the own energy plan and involves multiple renewable energy penetrations [7]. From Fig. 5.3, the seaport microgrid is connected with the main grid by a transformer. The harbor wind farm and PV are integrated to enhance the energy efficiency of the seaport. Electricity is the only secondary energy form to drive various logistic equipment, such as cold-ironing, and port cranes. Two practical seaport microgrid projects in Hamburg (German) and Genoa (Italy) [8] have proved its validity. Similar to the seaport microgrid, other industrial microgrids also have similar topologies. The main difference lies in the service load demand type. For example, seaport microgrid serves the logistic load demand, and the data center microgrid supplies power to the data center and the cooling power. Generally, for industrial applications, reliability is highly correlated with the economy, which motivates the ESS installment in industrial microgrids.
130
S. Fang and Y. Wang
Fig. 5.3 A typical industrial microgrid (seaport) [3] (Permission for usage from the author)
5.1.3 Mobile Microgrids The mobile microgrid is a new type of microgrids in the trend of transportation electrification, including various electric vehicles, ships, and aircrafts [3, 9]. Mobile microgrids mostly work in isolated mode and also can connect to the main grid in some operating conditions, such as charging of electrical vehicles, and berthed in of ships. Since the mobile microgrids mostly work in an isolated mode, the integration of ESS is essential for the operation and control of mobile microgrids. In the following, an all-electric ship (AES) is illustrated as a representative case of a mobile microgrid. AES is a new type of ship recently which replaces the mechanically driven propulsion system by the electrically driven propulsion system. AES has proved to have high energy efficiency than conventional ships and becomes the future trend of shipping design. As illustrated in Fig. 5.4, AES dispatches energy via an integrated power system onboard (shipboard microgrid), which consists of an energy network (blue lines and arrows) and a communication network (green lines and arrows). The power sources of an energy network include generators and battery, and some ships may have renewable energy integrations (PV modules in Fig. 5.4). The load demands of energy networks include the propulsion load and service load, that is, the onboard radar, navigation system, air conditioning, and so on. The communication network is used to send dispatch signals to each component to control the energy network.
5 The Role of Energy Storage Systems in Microgrids Operation
131
Fig. 5.4 Typical structure of all-electric ship [3] (Permission for usage from the author)
5.1.4 Comparisons between Different Types of Microgrids From above, microgrids are defined as a local energy district to integrate various energy sources to supply multiple energy demands, but there are still some differences in the following two aspects.
5.1.4.1
Operation Modes
Generally, microgrids can work in both grid-connected mode and isolated mode. However, different types of microgrids have different durations of operation modes, which will influence the planning, operation, and control of ESS. For example, the residential microgrids mostly work in grid-connected mode and the ESS is therefore acting as auxiliary equipment to adjust the power demand. For the mobile microgrids, such as AESs, they mostly work in isolated mode and the ESS is an important energy source, which needs to sustain the system operations independently for a certain period. The difference in operation modes among various types of microgrids also has great impacts on the ESS operation, which is further declared in the following context.
132
5.1.4.2
S. Fang and Y. Wang
Load Demand Types
The load demand types are diversified in different microgrids. For example, electric vehicle charging power is an important load demand type in residential microgrids, and the propulsion load is an important load demand type in AES. Those different load demand types certainly have different characteristics, such as the timedependent charging power profile and propulsion fluctuation, which gives different requirements on the ESS. From above, ESS gradually becomes an indispensable component in different types of microgrids. To comprehensively illustrate the role of ESS, this chapter is organized into three parts: at first, this chapter reviews current energy storage technologies and compares their differences. Secondly, two typical application scenarios are selected to show the roles of energy storage in microgrids, that is, load leveling and the power quality issues. At last, the conclusions are summarized.
5.2 Energy Storage Technologies In this chapter, the nomenclature of various energy storage technologies is shown in Table 5.1. Table 5.1 Nomenclature of different Energy Storage technologies
BES: Battery energy storage CAES: Compressed air energy storage FBES: Flow battery energy storage FESS: Flywheel energy storage Li ion: Lithium ion SMES: Superconducting magnetic energy storage SCES: Supercapacitor energy storage VRB: Vanadium redox battery ZBB: Zinc–bromine flow battery NaS: Sodium–sulfur Ni–Cd: Nickel–cadmium Ni–Mn: Nickel–manganese PHS: Pumped hydro storage PEM: Proton exchange membrane MCFC: Molten carbonate fuel cell SOFC: Solid oxide fuel cell
5 The Role of Energy Storage Systems in Microgrids Operation
133
5.2.1 Classification of Energy Storage Technologies Nowadays, there already exist many energy storage technologies, which are suitable for microgrid usage or not. In this section, several energy storage technologies available now are reviewed for clarifying their applications. Generally, electricity can be converted to many different forms for storage, which are shown as following Fig. 5.5, and Table 5.2 shows the parameters of different ESSs collected from several publications and manufacturers.
5.2.2 Single Energy Storage Technologies 5.2.2.1
Pumped Hydro Storage (PHS)
Currently, PHS is the most mature one among various energy storage technologies. In 2018, the global PHS capacity is already more than 1675GW. The typical structure of PHS is shown as the following Fig. 5.6. From Fig. 5.6, a PHS usually consists of two reservoirs, that is, upper reservoir and lower reservoir. The energy stored is proportional to the water volume in the upper reservoir and the height of the waterfall. The water is pumped upwards to the upper reservoir or flowed to the lower reservoir when low/high electricity price. An illustrative PHS example is in operation by the State Grid of China, named as the Tianhuangping PHS [29], which is the biggest one in Asia and the third biggest around the world. The installation was commissioned in 1992 and completed in 2000. The rated power of this PHS is 1800 MW. It locates in Zhejiang Province and is near the load central of eastern China, that is, Shanghai. Every summer,
Fig. 5.5 Classification of Energy Storage
Technologies PHS CAES VRB ZBB PSB NaS Lead-acid Ni–cd Li ion SMES FES SCES FC
Investment (US$/kWh) 10–15 [10] 2–4 [10] 600 [14] 500 [18] 450 [14] 170–200 [10] 50–100 [13] 400–2400 [13] 900–1300 [13] 200–300 [16] 400–800 [13] 100–300 [28] 10,000+ [28]
Energy rating (MWh) 500–8000 [11] 580, 2860 [12] 1.2–60 [15] 0.1–4 [19] 0.005–120 [16] 0.4–244.8 [21] 0.001–40 [23] 6.75 [25] 0.001–50 [27] 0.015 [13] 0.025–5 [16] 0.01 [25] –
Table 5.2 Characteristics of different ESS Power rating (MW) 10–1000 [11] 50–300 [12] 0.2–10 [15] 0.1–1 [19] 0.1–15 [16] 0.05–34 [21] 0.05–10 [23] 45 [25] 0.01–50 [27] 1–100 0.1–20[12] 0.05–0.2 [16] 0.5–5 [12]
Specific energy (kWh/kg) – 3.2–5.5 [13] 25–35 [16] 70–90 [13] – 100 [14] 30–50 [24] 30–80 [26] 80–200 [13] 10–70 [28] 5–100 [14] 5–15 [12] 800–10,000 [16]
Specific power (kW/kg) – – 166 [17] 45 [20] – 90–230 [22] 180–200 [13] 100–150 [26] 200–2000 [13] 400–2000 10,000+ [25] 10,000+ [28] 500+ [14]
134 S. Fang and Y. Wang
5 The Role of Energy Storage Systems in Microgrids Operation
135
Fig. 5.6 Illustration of pumped hydro storage (PHS)
Fig. 5.7 Illustration of compressed air energy storage (CAES)
Tianhuangping PHS can shift more than 500 MW peak load per day. However, the construction of PHS highly depends on the geographic conditions, and may not suitable for microgrid usage.
5.2.2.2
Compressed Air Energy Storage (CAES)
CAES stores energy in the form of compressed air and the air is usually stored in an underground cavern or man-made storage tank. A typical structure of CAES is shown as Fig. 5.7.
136
S. Fang and Y. Wang
When CAES outputs electricity, the compressed air is released from the air tank, then the released air is heated and expanded to drive the high/low pressure turbines. After that, the air is combusted with the natural gas to drive the generator. The by-produced waste heat is used to heat the compressed air in the recuperator. The overall lifetime of CAES is approximately 40 years, and the energy efficiency is about 71% [13]. Currently, CAES systems are not widely used. As far as the authors’ knowledge, until now there are only two practical projects of CAES globally so far, the one is invested in Germany (290 MW) and the other one is constructed in the USA (110 MW) [30]. However, the size of CAES is generally too large for microgrid usage and the installment of CAES also needs a very large space. Since the capacities of conventional microgrids are no more than100MW, CAES can be used only in some special microgrids which have large energy storage requirement, such as wind farm microgrids and tidal energy microgrids [28].
5.2.2.3
Battery Energy Storage (BES)
Due to high energy density and flexibility for installment, the battery is the most commonly used ESS on the market. The energy is stored as the electrochemical energy in each battery cell, and multiple battery cells are accumulated in series/in parallel to make up the desired voltage and capacity. A typical battery structure is shown in Fig. 5.8. From Fig. 5.8, a battery pack consists of several battery cells and each battery cell consists of two electrodes immersed in the electrolyte. All the battery packs are sealed in a container and then integrated into the external grid.
Fig. 5.8 Illustration of battery energy storage packs
5 The Role of Energy Storage Systems in Microgrids Operation
137
In the last decade, the technologies of the battery have become much more mature. The main representatives are the lead–acid battery, nickel–cadmium battery, and lithium–ion battery. The lead–acid batteries, which have been researched for more than 140 years, is the most mature battery technology now. Currently, tremendous efforts have been devoted to the technologies like nickel–cadmium and lithium–ion batteries to achieve more cost-effective options in power grid applications.
5.2.2.4
Flow Battery Energy Storage (FBES)
FBES is an emerging technology only recently. The principle of FBES is based on the reversible electrochemical reactions similar to conventional BES. The main difference between the flow FBES and conventional BES is two different aqueous electrolytic of the flow battery are in the liquid solutions and stored in separate tanks. These aqueous solutions are pumped to the cell when operation [31]. There are three types of commercially available flow batteries currently, that is, Vanadium Redox Battery (VRB), Zinc Bromine Battery (ZBB), and Polysulphide Bromide Battery (PSB). Their illustrations are presented in Fig. 5.9. The main advantage of the FBES is the scalable energy capacity. The volume of the stored electrolyte can be designed to meet the desired energy requirement [31],
Fig. 5.9 Illustration of flow battery energy storage
138
S. Fang and Y. Wang
Fig. 5.10 Illustration of flywheel energy storage
and the power capacity of the FBES depends on the cell number and the size of the electrodes. FBES also have significant advantages over conventional batteries, (1) FBES can be fully discharged without any damage and (2) FBES have very low selfdischarge since the electrolytes are stored in separate sealed tanks. In this manner, FBES has a quite long life and low maintenance cost than conventional batteries and is also able to store energy over a long period with slight energy loss.
5.2.2.5
Flywheel Energy Storage (FES)
FES stores electricity in kinetic energy of rotating mass or rotor. The stored energy is proportional to the rotor mass, location of the mass, and the rotor’s rotational speed. When FES charges, it absorbs the energy from outside and accelerates the rotational speed of the mass. On the other side, when FES discharges, the rotating mass drives a generator to produce electricity, and the rotational speed slows down. An illustration of FES is shown in Fig. 5.10. Generally, FES can quickly respond to the power demand, and therefore widely used in the fast-acting scenario, such as for the power quality improvement, fast load demand shaving, or uninterrupted power supply (UPS) [32], frequency response [33], power smoothing [34], and port crane locomotivations [35]. The advantage of FES is the intermediate characteristics between the batteries and supercapacitors, that is, the FES has much higher power density than batteries and much higher energy density than supercapacitors. In addition, FES also has many advantages compared with prior other energy storage technologies, such as less sensitivity to temperature, chemical hazardless, higher life cycle, reduced space and weight, which is suitable for many applications. But the FES also has its shortcomings, that is, the complex maintenance process for rotating mass and the high self-discharging rate.
5 The Role of Energy Storage Systems in Microgrids Operation
139
Fig. 5.11 Illustration of superconducting magnetics energy storage
5.2.2.6
Superconducting Magnetics Energy Storage (SMES)
SMES uses a magnetic field created by the superconducting coil to store energy [36]. The structure of SMES is shown in Fig. 5.11. However, in order to keep the coil in superconductive states, the working temperature of SMES should below the critical temperature by an external cooling system. Since zero resistance, SMES storage devices have very high energy efficiency, that is, usually more than 95%. The main energy loss is due to the power electronic interfaces, which accounts for about 2–3% loss in both charging/discharging. The advantage of SMES is the high lifecycles and the disadvantages are (1) high rate of self-discharge, (2) very large installment space because of the external cooling system, and (3) mechanical stability issues led by the cryogenic liquid.
5.2.2.7
Supercapacitor Energy Storage (SES)
Capacitors store energy in the electric field, and therefore, have very short responding time. Currently, the supercapacitor is the most commonly used capacitor, and the corresponding SES is shown in the following Fig. 5.12. The main advantages of SES include high power density, shorter charging/discharging time, quite long lifecycles compared with other ESSs. The disadvantages are the low voltage of each cell, and much higher investment cost per watt-hour, that is, more than 10 times compared with a lithium battery. Other drawbacks of SES include relatively low energy density compared with battery and very high self-discharge.
140
S. Fang and Y. Wang
Fig. 5.12 Illustration of super-capacitor energy storage Fig. 5.13 Illustration of fuel cell Fuel
Oxygen
Electrolyte
Waste Anode
5.2.2.8
Cathode
Fuel Cell
Different from the above energy storage technologies, a fuel cell is more like a generator that directly transforms chemical energy into electrical energy. But since there is no combustion process and no rotating equipment, a fuel cell has similar operating characteristics with conventional ESS and superior efficiency than generators. A typical structure of the fuel cell is shown in Fig. 5.13. Since the reduced space, scalable capacity, fuel cells are viewed as promising energy source for mobile microgrids. In this field, polymer exchange membrane (PEM) fuel cell is the most mature technology and has been already used in submarine applications and other propulsion usages, ranging from 30–40 kW [37]. The main advantage of the fuel cell in a maritime application is the great reductions
5 The Role of Energy Storage Systems in Microgrids Operation
141
on gas emissions [38], but the storage technologies for hydrogen fuel is the main challenging features since it is hard to store the hydrogen fuel as conventional hydrocarbon fuels, that is, liquid natural gas (LNG) or Marine Diesel Oil (MDO) [39]. Currently, fuel cell technologies using conventional hydrocarbon fuels have gained high concerns, such as Molten Carbonate Fuel Cell (MCFC) and Solid Oxide Fuel Cell (SOFC).
5.2.3 Hybrid Energy Storage Technologies In current microgrid usage, the battery is the most commonly used energy storage technology to act as an energy buffer. However, the battery usually has high energy density but the power density is low. Therefore, hybrid ESSs are used to combine the advantages of different energy storage technology. Three combinations are frequently used in microgrid operation.
5.2.3.1
Battery Supercapacitor
Among all the hybridization technologies, the battery-supercapacitor combination has been studied quite extensively. In battery-supercapacitor hybrid ESS, the batteries serve as an energy buffer, which generally has high energy density but relatively low power density, and the life cycle of the battery is also limited, only 1500 to 4500 full cycles, meanwhile the supercapacitor has 100,000+ lifecycles. On the other side, the supercapacitor energy storage device has low energy density but very high power density. Therefore, the supercapacitor serves as a power buffer to undertake the high power demand, and the hybridization with battery can combine the advantages of high energy density and high power density. Currently, there are two types of hybridization for the battery-supercapacitor hybrid energy storage, that is, internal hybridization and external hybridization, which are shown as Fig. 5.14 a, b, respectively.
5.2.3.2
Battery–Fuel Cell
Since the reduced scale and convenient maintenance, battery–fuel cell hybrid ESS has shown great advantages in the maritime applications. The main advantages include the reduction of fuel consumption, reduced emissions, lower noise, lower maintenance requirements, and minimal vibration. For example, fuel cells are first integrated into a tourist vessel in 2008, named as Alster-Touristik, which has two 50 kW fuel cells in combination with lead–acid batteries [40]. The vessel can hold 100 passengers at maximum and the cruising speed is 8 knots. Compared with the traditional diesel-powered vessel, the practical operation of this fuel–cell driven vessel demonstrates that the integration of battery–fuel cell hybrid ESS benefits the
142
S. Fang and Y. Wang
Fig. 5.14 Illustration of battery-supercapacitor hybridization
Fig. 5.15 Typical topologies of fuel cell integrated shipboard microgrid
gas emission control. A typical topology of fuel cell integrated shipboard microgrid is shown in Fig. 5.15. Other applications include “Nemo H2”, a zero-emission ferry launched in 2009, of which the propulsion system comprises of 60–70 kW PEM-based fuel cells and 30–50 kW batteries [41]. In the future, since those great advantages, the hybridization of battery and fuel cell will certainly have wider applications in microgrid operation.
5 The Role of Energy Storage Systems in Microgrids Operation
5.2.3.3
143
Battery–Flywheel
From above, the battery has a relatively low power density and is suitable to undertake the long timescale power demand, and the flywheel can quickly respond to the power demand fluctuations. In this sense, the combination of battery and flywheel can gain both high power density and high energy density. In [33], the author uses hybrid battery–flywheel energy storage to mitigate the long-term and short-term load fluctuations. The proposed method can be applied in wind power integration, renewable uncertainty mitigation, as well as the shocking load mitigation [34, 35]. Other applications also proposed different control methods to balance the power demand on battery and flywheel for optimal microgrid control, battery lifetime extension, or facilitating PV integration.
5.3 Energy Storage Applications in Microgrids According to timescales, the applications of ESS in microgrids can be classified as long-term timescale and short-term timescale. The long-term timescale applications are shifting or shaving the load hourly or even longer, and the short-term timescale applications are the quick load response to mitigate the fluctuations, which are illustrated as follow:
5.3.1 Load Leveling In microgrid operation, ESS acts as an energy/power buffer to keep the power balance, and proper management of ESS can shave the peak load or level the load demand, which is illustrated in Fig. 5.16. From the above Fig. 5.16, when the load demand is high, the ESS discharges power to share the demand of load, and when the load demand is low, the ESS can absorb power. With the iteration of charging and discharging, the load demand is leveled or shaved. The load leveling and peak shaving can bring lots of benefits to microgrid operations. For example, the ESS can charge when the electricity price is low and discharge when the electricity price is high, to gain extra benefits. In land-based microgrids, there are many researches addressing the load leveling and peak shaving. In Ref. [42], a practical energy storage project is analyzed to show the advantages of an integrated energy storage system, which shows the integrated battery ESS can gain extra economic benefits. In Ref. [43], the authors proposed an effective energy storage sizing method and an optimal peak shaving strategy to reduce the peak load of a residential area. In Ref. [44], a novel algorithm to control the charging/discharging of a battery is presented, for the peak load shaving, power curve smoothing, and voltage regulation of a distribution transformer. In Ref. [45], the battery energy storage system is used to shave the peak load of a
144
S. Fang and Y. Wang
Fig. 5.16 Illustration of peak load shaving
residential building. In Ref. [46], battery planning is used to enhance the generation hosting capacity of a distribution network. In Ref. [47], the vanadium redox flow battery is used to coordinately control the frequency fluctuations and shave the peak load. In [48], the centralized and distributed energy storage systems are both used to facilitate the expansion of an active distribution network for maximizing the overall economic benefits. In Ref. [49], a cost–benefit analysis of Na–S battery is conducted for the peak load shaving. In [50], the battery energy storage anticipates the electricity market to conduct a price-aware optimal dispatch. In [51], the authors propose a model predictive control method to make battery tracking and shaving the power demand. In mobile microgrids, Ref. [52, 53] proposes energy storage management methods to reduce the fuel consumption and gas emission of ships, respectively. Ref. [54] uses battery energy management in a multi-objective framework, that is, both the fuel consumption and gas emission. Ref. [55] uses hybrid energy storage to mitigate the load fluctuations. Ref. [56] uses the hybrid energy storage for reducing the fuel consumption and gas emission and extends the lifetime of the battery. Ref. [57] proposes a battery planning method in a ferry and demonstrates the integration of battery is necessary for the optimal operation of mobile microgrids. Ref. [58, 59] use a battery energy storage system to facilitate the integration of shipboard photovoltaic modules. In summary, the integration of energy storage into microgrids greatly facilitates the optimal operation. The peak shaving and load leveling can make the generation system of microgrids works in a more economic and environmental way. The energy storage with high energy density usually serves in these scenarios, that is, the battery or flow battery.
5 The Role of Energy Storage Systems in Microgrids Operation
145
5.3.2 Power Quality There exist many load fluctuations in microgrid operation, such as renewable energy uncertainties and pulsed loads, which might result in voltage and frequency fluctuations, that is, voltage and frequency fluctuations and harmonic contamination, and so on. Since the ability to keep power balance and the convenient deployment requirement, energy storage systems are viewed as promising routes to mitigate power quality issues. Various high power density energy storage system, such as supercapacitor, flywheel, are used to handle the power quality issues. In land-based microgrids, Ref. [60] comprehensively analyzes the control architecture for microgrids which addresses the power quality by energy storage. Ref. [61] uses expanded energy storage planning to address the power quality issues introduced by renewable power generation. Ref. [62] coordinately uses DSTATCOM and energy storage to mitigate the power quality issues. Ref. [63] uses the battery to reduce the fluctuations of photovoltaic integration. Ref. [64] proposes a control strategy to integrate battery to support the frequency in microgrids. In mobile microgrids, Ref. [65] proposes a hybrid battery–flywheel to address the power quality issue in a shipboard microgrid, which can store energy up to 80 MJ. Ref. [66] uses FESS to mitigate the voltage sags for improving the ship’s survivability. Ref. [67] proposed a multi-modular DC–DC configuration for the HESS integration into the shipboard microgrid.
5.4 Conclusions This chapter introduces the role of energy storage systems in microgrids operation. The main types of microgrids, and the requirements on the ESS, and the operation characteristics of ESS are comprehensively illustrated in this chapter. The main conclusions of this chapter can be summarized as follow: 1. There exist many different types of microgrids and their operating characteristics will greatly influence the ESS, that is, requirements on the power/energy densities, operation conditions, and so on. 2. Different ESS technologies have their applications and the hybrid ESS technology is promising for the future microgrid applications since it can combine the high power density and high energy density. 3. Since its high efficiency, small scales, and less noise, the combination of fuel cell and other ESS will provide a promising energy source configuration for future microgrids.
146
S. Fang and Y. Wang
References 1. Vandoorn, T., Vasquez, J., et al. (2013). Microgrids: Hierarchical control and an overview of the control and reserve management strategies. IEEE Industrial Electronics Magazine, 7(4), 42–55. 2. Hafez, O., & Bhattacharya, K. (2012). Optimal planning and design of a renewable energy based supply system for microgrids. Renewable Energy, 45(9), 7–15. 3. Fang, S., Gou, B., et al. (2020). Towards future green maritime transportation: Overview of seaport microgrids and all-electric ships. IEEE Transactions on Vehicular Technology, 69(1), 207–219. 4. Tan, X., Li, Q., et al. (2013). Advances and trends of energy storage technology in microgrid. International Journal of Electrical Power & Energy Systems, 44(1), 179–191. 5. Fang, S., Zhao, T., et al. (2020). Coordinated Chance Constrained Programming of a Multienergy Microgrid System for Balancing Operational Efficiency and Quality-of-service. Modern Power System and Clean Energy, In press. 6. Parise, G., Parise, L., et al. (2016). Wise-port business energy management: Port facilities, electrical power distribution. IEEE Transactions on Industry Application, 52(1), 18–24. 7. Lamberti, T., Sorce, A., et al. (2015). Smart port: Exploiting renewable energy and storage potential of moored boats. Genova, Italy: Oceans-2015. 8. Acciaro, M., Ghiara, H., et al. (2014). Energy management in seaports: A new role for port authorities. Energy Policy, 71, 4–12. 9. Fang, S., & Xu, Y. (2019). Multi-objective coordinated scheduling of energy and flight for hybrid electric unmanned aircrafts. IEEE Transactions on Industrial Electronics, 66(7), 5686– 5695. 10. Kaldellis, J. K., & Zafirakis, D. (2007). Optimum energy storage techniques for the improvement of renewable energy sources-based electricity generation economic efficiency. Energy, 32, 2295–2305. 11. Kazempour, J., Moghaddam, P., et al. (2009). Electric energy storage systems in a market-based economy: Comparison of emerging and traditional technologies. Renewable Energy, 34(12), 2630–2639. 12. Ridge Energy Storage & Grid Services L.P. The economic impact of CAES on wind in TX, OK, and NM, http://www.ridgeenergystorage.com/re wind projects-compressed2005.pdf. 13. Hadjipaschalis, I., Poullikkas, A., et al. (2009). Overview of current and future energy storage technologies for electric power applications. Renewable and Sustainable Energy Reviews, 13(6), 1513–1522. 14. Beck, F., & Ruëtschi, P. (2000). Rechargeable batteries with aqueous electrolytes. Electrochimica-Acta, 45, 2467–2482. 15. Prudent Energy Corp. website, http://www.pdenergy.com/. 16. Ponce-de-León, C., Frías-Ferrer, A., et al. (2006). Redox flow cells for energy conversion. Journal of Power Sources, 160, 716–732. 17. L. Barote, R. Weissbach, et al. Stand-alone wind system with vanadium redox battery energy storage. In 11th international conference on optimization of electrical and electronic equipment, 407–412, 2008. 18. D. Ton, C. Hanley, et al. Solar energy grid integration systems-energy storage (SEGIS-ES) http://prod.sandia.gov/techlib/accesscontrol. cgi/2008/084247.pdf. 19. G. Corey. An assessment of the state of the zinc–bromine battery development effort, www.redflow.com. 20. Emerging electric technologies whitepaper. Parcon website, http://www.parcon.uci.edu/paper/ EmergingElectricalTechnologies/eeenergy.htm. 21. N. Kawakami, Y. Iijima, et al. Development and field experiences of stabilization system using 34 MW NaS batteries for a 51 MW wind farm IEEE International Symposium on Industrial Electronics (ISIE), 2010.
5 The Role of Energy Storage Systems in Microgrids Operation
147
22. EA Technology Review of electrical energy storage technologies and systems and of their potential for the UK [R], http://www.wearemichigan.com/JobsAndEnergy/documents/ file15185.pdf, 2004. 23. McDowall, J. (2006). Integrating energy storage with wind power in weak electricity grids. Journal of Power Sources, 162, 959–964. 24. Baker, J. (2008). New technology and possible advances in energy storage. Energy Policy, 36, 4368–4373. 25. Alaska Energy Authority Energy Storage Review, http://www.aidea.org/aea/. 26. Harding Energy Inc. website, http://www.hardingenergy.com/. 27. Saft Batteries website, http://www.saftbatteries.com/. 28. Francisco, D., Andreas, S., et al. (2012). A review of energy storage technologies for wind power applications. Renewable and Sustainable Energy Reviews, 16(4), 2154–2171. 29. Tianhuangping pumped storage power station., https://en.wikipedia.org/wiki/ Tianhuangping_Pumped_Storage_Power_Station. 30. Ibrahim, H., Ilinca, A., et al. (2008). Energy storage systems—Characteristics and comparisons. Renewable and Sustainable Energy Reviews, 12(5), 1221–1250. 31. Skyllas-Kazacos, M., Chakrabarti, M., et al. (2011). Progress in flow battery research and development. Journal of the Electrochemical Society, 158(8), 55–79. 32. Park, J., Kalev, C., et al. (2008). Control of high-speed solid-rotor synchronous reluctance motor/generator for flywheel-based uninterruptible power supplies. IEEE Transactions on Industrial Electronics, 55(8), 3038–3046. 33. Cheng, M., Sami, S., et al. (2017). Benefits of using virtual energy storage system for power system frequency response. Applied Energy, 194, 376–385. 34. Díaz-González, F., Sumper, A., et al. (2013). Energy management of flywheel-based energy storage device for wind power smoothing. Applied Energy, 110, 207–219. 35. Spiryagin, M., Wolfs, P., et al. (2015). Application of flywheel energy storage for heavy haul locomotives. Applied Energy, 157, 607–618. 36. Hsony, W., & Dodds, S. (1993). Applied superconductivity developments in Japan. Power Engineering Journal, 7(4), 170–176. 37. German HDW Submarine. Accessed https://www.naval-technology.com/projects/type_212/ on 27 August 2018. 38. Fuel Cell Ship Alsterwasser. Accessed https://www.drewsmarine.com/en/references/ passenger-ferries/fcs-alsterwasser on 27 August 2018. 39. Symington, W., Belle, A., et al. (2016). Emerging technologies in marine electric propulsion. Journal of Engineering for the Maritime Environment, 230(1), 187–198. 40. A. Anvari-Moghaddam, T. Dragicevic, et al. Optimal planning and operation management of a ship electrical power system with energy storage system. In Proceedings of the IECON 42nd annual conference of the IEEE industrial electronics society, Florence, Italy, 24–27 October 2016, pp. 2095–2099. 41. Fuel Cell Boat (Nemo H2). Accessed http://www.opr-advies.nl/fuelcellboat/efcbboot.html on 27 August 2018. 42. Koller, M., Borsche, T., et al. (2015). Review of grid applications with the Zurich 1 MW battery energy storage system. Electric Power Systems Research, 120, 128–135. 43. Chua, K., Lim, Y., et al. (2016). Energy storage system for peak shaving. International Journal of Energy Sector Management, 10(1), 3–18. 44. Reihani, E., Sepasi, S., et al. (2016). Energy management at the distribution grid using a battery energy storage system (BESS). International Journal of Electrical Power & Energy Systems, 77, 337–344. 45. Leadbetter, J., & Swan, L. (2012). Battery storage system for residential electricity peak demand shaving. Energy and Buildings, 55, 685–692. 46. Jayasekara, N., Masoum, M., et al. (2015). Optimal operation of distributed energy storage systems to improve distribution network load and generation hosting capability. IEEE Transactions on Sustainable Energy, 7(1), 250–261.
148
S. Fang and Y. Wang
47. Lucas, A., & Chondrogiannis, S. (2016). Smart grid energy storage controller for frequency regulation and peak shaving, using a vanadium redox flow battery. International Journal of Electrical Power & Energy Systems, 80, 26–36. 48. Shen, X., Shahidehpour, M., et al. (2016). Expansion planning of active distribution networks with centralized and distributed energy storage systems. IEEE Transactions on Sustainable Energy, 8(1), 126–134. 49. Liao, Q., Sun, B., et al. (2016). A techno-economic analysis on NaS battery energy storage system supporting peak shaving. International Journal of Energy Research, 40(2), 241–247. 50. Awad, A., Fuller, J., et al. (2014). Impact of energy storage systems on electricity market equilibrium. IEEE Transactions on Sustainable Energy, 5(3), 875–885. 51. Di-Giorgio, A., Liberati, F., et al. (2016). Model predictive control of energy storage systems for power tracking and shaving in distribution grids. IEEE Transactions on Sustainable Energy, 8(2), 496–504. 52. Kanellos, F. D. (2014). Optimal power management with GHG emissions limitation in allelectric ship power systems comprising energy storage systems. IEEE Transactions on Power System, 29(1), 330–339. 53. Kanellos, F. D., Tsekouras, G. J., & Hatziargyriou, N. D. (2014). Optimal demand-side management and power generation scheduling in an all-electric ship. IEEE Transactions on Sustainable Energy, 5(4), 1166–1175. 54. Shang, C., Srinivasan, D., & Reindl, T. (2016). Economic and environmental generation and voyage scheduling of all-electric ships. IEEE Transactions on Power Systems, 31(5), 4087– 4096. 55. Hou, J., Sun, J., et al. (2018). Mitigating power fluctuations in electric ship propulsion with hybrid energy storage system: Design and analysis. IEEE Journal of Oceanic Engineering, 43(1), 93–107. 56. Fang, S., Xu, Y., Li, Z., et al. (2019). Two-step multi-objective management of hybrid energy storage system in all-electric ship microgrids. IEEE Transactions on Vehicular Technology, 68(4), 3361–3372. 57. Boveri, A., Silvestro, F., Molinas, M., et al. (2018). Optimal sizing of energy storage systems for shipboard applications. IEEE Transactions on Energy Conversion, 34(2), 801–811. 58. Lan, H., Wen, S., et al. (2015). Optimal sizing of hybrid PV/diesel/battery in ship power system. Applied Energy, 158, 26–34. 59. Wen, S., Lan, H., et al. (2016). Allocation of ESS by interval optimization method considering impact of ship swinging on hybrid PV/diesel ship power system. Applied Energy, 175, 158– 167. 60. Guerrero, J., Loh, P., et al. (2012). Advanced control architectures for intelligent microgridspart II: Power quality, energy storage, and AC/DC microgrids. IEEE Transactions on Industrial Electronics, 60(4), 1263–1270. 61. Fu, Q., Montoya, L., et al. (2012). Microgrid generation capacity design with renewables and energy storage addressing power quality and surety. IEEE Transactions on Smart Grid, 3(4), 2019–2027. 62. Mahela, O., & Shaik, A. (2016). Power quality improvement in distribution network using DSTATCOM with battery energy storage system. International Journal of Electrical Power & Energy Systems, 83, 229–240. 63. Li, X., Hui, D., et al. (2013). Battery energy storage station (BESS)-based smoothing control of photovoltaic (PV) and wind power generation fluctuations. IEEE Transactions on Sustainable Energy, 4(2), 464–473. 64. Serban, I., & Marinescu, C. (2013). Control strategy of three-phase battery energy storage systems for frequency support in microgrids and with uninterrupted supply of local loads. IEEE Transactions on Power Electronics, 29(9), 5010–5020. 65. C. Xie, C. Zhang. Research on the ship electric propulsion system network power quality with flywheel energy storage. In Proceedings of the Power and Energy Engineering Conference (APPEEC) Asia-Pacific, Chengdu, China, 28–31 March 2010; pp. 1–3.
5 The Role of Energy Storage Systems in Microgrids Operation
149
66. Samineni, S., Johnson, B., et al. (2006). Modeling and analysis of a flywheel energy storage system for voltage sag correction. IEEE Transations Industry Application, 42(1), 42–52. 67. Mo, R., & Li, H. (2017). Hybrid energy storage system with active filter function for shipboard MVDC system applications based on isolated modular multilevel DC/DC converter. IEEE Journal of Emerging and Selected Topics in Power Electronics, 5(1), 79–87.
Chapter 6
Microgrids and Local Markets Mohsen Khorasany and Reza Razzaghi
6.1 Introduction Progressive integration of Distributed Energy Resources (DERs) aligned with recent advances in smart technologies are accelerating the pace of the global energy transition to a decentralized, decarbonized, and digitalized network. These trends in the energy sector are challenging well-established practices for supplying and selling electricity by introducing new models for DER integration such as virtual power plants, smart grids, and microgrids. The efficient integration of DERs requires appropriate operational and control layers, which coordinate these resources. This coordination provides substantial benefits for the stability of power networks and increases the value of the local management of these resources [1]. Through participating in new markets, DER owners can benefit from this coordination by maximizing the return on their investment. From the networks perspective, coordinated and controlled use of DERs, in combination with load management technologies and services such as demand-side response can provide significant benefits for power systems stability [2]. The flexibility of DERs in generating and absorbing energy is a key value, which prevents or relieves localized network performance issues such as voltage fluctuation, where the network is at the risk of breaching its performance limits. Nonetheless, the complexity of the energy system avoids DER owners to access values from their assets and investments. Hence, there is a need to develop new systems such as active distribution networks that are capable of providing value streams to customers to incentivize them to share their flexibilities [3]. In a network with high penetration of DERs, new market models can emerge that favor a local
M. Khorasany () · R. Razzaghi Department of Electrical and Computer Systems Engineering, Monash University, Clayton, VIC, Australia e-mail: [email protected] © Springer Nature Switzerland AG 2021 A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems, https://doi.org/10.1007/978-3-030-59750-4_6
151
152
M. Khorasany and R. Razzaghi Distribution
Transmission
On-site Storage DNSP
Prosumers
Consumers
TNSP
DSO
Microgrid Operator
Market Operator
MICROGRID
Retailer
Generation
On-site Generation Local Market
Retail Market
Wholesale Market
Fig. 6.1 Interaction of local market with existing systems
usage of the generated electricity [4]. Different services can be provided through local markets, such as aggregation of local resources, Peer-to-Peer (P2P) trading, and demand-side flexibility, which can help to overcome system balancing issues at the local level. The provided flexibility by local market participants can be utilized by network operators to manage capacity and voltage constraints in networks. Through implementing local markets, the expensive grid reinforcement costs can be avoided by employing the flexibility of local resources. Also, it can add value to the whole energy system by enabling more intelligent management of local energy resources. In this context, microgrids can be used as an architecture capable of making a local market to utilize resources at the local level and to maximize the local consumption of electricity generated in a distributed manner [5]. Microgrids provide the opportunity for sharing and monetizing values from the increasing volume of DERs and the emergence of prosumers (proactive consumers) in the electricity system. Through a monitoring and control platform, the microgrid operator can coordinate the supply and demand of customers connected to the microgrid. Through coordination, a microgrid can maximize the value of the connected DERs for microgrid participants, the network, and the broader market. Microgrids, as a decentralized business model, help customers to participate in the market and access values from services they provide to the grid. A microgrid as a benchmark for local markets and its interaction with different stakeholders is shown in Fig. 6.1. Local market in microgrids fulfills customers’ demand from local energy resources and reduces the need for expensive and inefficient energy transportation with substantial losses [6]. It increases social welfare in the community and encourages reinvestments in an additional renewable generation [7]. This chapter is focused on local markets for microgrid provision. An overview of local markets’ definition, their potential benefits, and objectives are provided.
6 Microgrids and Local Markets
153
A summary of key enabler elements for local market implementation is given. Different trading approaches for local markets are presented with a detailed literature review on each model. Also, an overview of different market settlement approaches for local markets is given. Case examples of local market business models and market settlements in local markets are provided to help the reader with outlining the attributes of different market models.
6.2 Local Markets 6.2.1 Definition Different references have used diverse definitions for local markets. The authors of [8] define the local market as a marketplace, which allows prosumers and consumers to trade electricity directly within their community at variable prices. In [9], the local market is defined as a group of interconnected loads and DERs that provide a market platform for locally generated energy to their participants. According to the definition in [10], a local community including different types of prosumers, consumers, and producers, as well as storage facilities constitute a local market, where community members are engaged in an array of commercial activities that serve to create a better and more sustainable energy experience for all involved parties. In [11], the local market is defined as a highly flexible market platform to coordinate self-interested energy agents. A local market is rooted in a residential area or similar and is based on a micro-market [12] concept that includes prosumers and consumers, as well as storage facilities within such a community [13]. Munné-Collado et al. [14] define a local market as a trading arena located within a local energy community, operated in a public grid to provide two different services: energy and flexibility. A local market is a platform on which individual consumers and prosumers trade energy supporting regional scopes such as a neighborhood environment [15]. The authors of [16] state that local markets can promote renewable energy development more efficiently and define the local market as a marketplace, where local cogeneration plants, wind power plants, and consumers are incentivized to trade internally and to cope locally with the wind power fluctuations. All of these definitions have two common aspects: customer engagement and the locality of the market. Customer engagement means that customers in the local market are active players who participate in market activities in different ways. The locality of activities underlines the location of customers as a key point in forming local markets. Local markets can counteract current issues in energy systems and solve grid imbalances on a local level with a market-based approach [17]. The local market can be interpreted as an approach for energy management based on market rules, where price signals can be developed to manage players’ behavior out with a market or contractual obligation. Transactive energy management is an example
154
M. Khorasany and R. Razzaghi
of market-based energy management, which uses the value as a key operational parameter to balance demand and supply in the network [18]. Transactive energy provides economic signals aligned with operational goals to encourage demand-side activities based on economic incentives [19].
6.2.2 Benefits Developing local markets can provide a range of different benefits for different stakeholders. Customers in local markets are able to consume self-generated electricity and trade their energy surplus within their neighborhood area. Consuming locally generated electricity reduces transport losses and the risk of backfeeding at medium voltage or low voltage transformers [20]. Local markets strengthen customers’ position by enabling them to play a more active role in the market instead of being submissive ratepayers. Through aggregating and coordinating small-scale resources, local markets can provide different services, such as the sale of electricity into wholesale and retail markets, and the provision of voltage and frequency control services to the local network. Improvement in system efficiency due to a reduction in losses and improved network agility due to reduced dependency on the main grid are other benefits of local markets. Furthermore, local markets support the development of a smart grid [21]. Energy consumers may have different preferences on the type of energy source based on factors like environmental preferences, such as being carbon-free, pollution-free, exclusively renewable, and locally generated [22]. Consumers’ preferences can be incorporated in local markets by providing those customers with an opportunity to choose the source of their energy. Also, local markets support the local economy, which provides new opportunities for local industries and regional businesses. Sharing energy in a community of prosumers and consumers increases social cohesion and improves the sense of community [23].
6.2.3 Objectives Local markets can be established to achieve different objectives. Designing an appropriate model for local markets is a challenging task, as it needs to incorporate several inconsistent objectives [24]. Several principles need to be considered in the market design, including (1) maximizing market surplus for involved parties, (2) incentivizing market players to follow their commitments, (3) clarifying the market price details for all players, and (4) providing an opportunity for all suppliers to recover their costs. According to these principles and the specified list of objectives in [14], the following objectives can be considered for a local market: • Maximizing social welfare of market players.
6 Microgrids and Local Markets
155
• Minimizing energy costs by providing access to cheaper locally generated energy. • Increasing utilization of renewable energy resources by creating an attractive and competitive marketplace that incentives to buy energy from local and renewable resource. • Supporting power systems by providing services in energy and flexibility markets.
6.2.4 Services Different services can be provided in local markets, which can be categorized as energy and flexibility services exchange [14]. Energy services aim to manage load and generation resources to minimize total energy costs. Local markets can provide energy related services by facilitating local energy trading and managing local resources for participation in external energy markets. On the other side, flexibility services aim to adjust power in a given moment for a given duration in a specific location within a network. This service can be provided in local markets by modifying the generation and consumption patterns of local resources in reaction to market signals. In the literature, there are several examples of local market design for both energy and flexibility services. In [25], the local market is defined as a flexibility market, which employs regional flexibility to resolve grid violations issues. Authors of [26] presents an aggregator-based flexibility market for the participation of small scale DERs in flexibility services trading. A framework for flexibility trading by prosumers is presented in [27], where a local market operator plans the flexibility to avoid congestion problems in the distribution grids. Authors of [28] propose a market design, which allows aggregators and prosumers to provide flexibility in response to distribution system operator requests. Different market mechanisms for coordinating the retailer, transmission system operator, and distribution system operator as buyers of flexibility are studied in [29]. The proposed local flexibility market in [30] is designed as a market-based management mechanism for aggregators, in which multiple participants compete for selling or buying flexibility. Market design for local energy trading is addressed in several studies. A blockchain-based platform for the microgrid energy market is presented in [31], where a classification of required components for an efficient market design is presented. In [32], the energy trading of microgrids in a transactive system is modeled as an auction-based electricity market, which manages the local demand and supply. Authors of [33] proposes a double-sided auction for energy trading in an islanded microgrid, where a distributed algorithm is implemented by an aggregator to clear the market. The presented framework in [34] is an energy exchange model for several microgrids, which aims to lower the exchanged energy between the microgrids and the utility. A contribution-based mechanism is used in [35] for energy trading among microgrids, in which a noncooperative energy
156
M. Khorasany and R. Razzaghi
competition game among the consumers is modeled that allows them to maximize their satisfaction. In [36], the proposed trading framework enables microgrids to trade energy in an independent market to maximize their average revenue, in which a trading scheme is developed employing learning automation and noncooperative repeated game. In [37], the energy trading is formulated using the Stackelberg game model and a reward concept is proposed to promote energy trading for consumers. Energy trading and scheduling of several interconnected microgrids modeled by Nash bargaining theory are presented in [38]. In [39], a game-theoretic approach is employed to design pricing mechanisms for energy trading in a prosumer-based community microgrid. Dynamic price-based demand response for energy sharing model among prosumers in a microgrid is presented in [40], where through a distributed approach, prosumers can co-decide on internal prices.
6.2.5 Value Streams for Microgrids Developing local markets in microgrids unlocks several value streams for microgrids such as cost saving on electricity bills, increasing energy efficiency, and reducing grid dependency. Through the establishment of a local market, microgrids can coordinate local resources for providing different services. For instance, microgrids can participate in Frequency Control Ancillary Services (FCAS) to provide services requested by network operators. Peak demand is a significant challenge in power networks, increasing infrastructure costs, and making balancing of supply and demand difficult. Peak demand charges can be reduced in microgrids through local markets. Sale of electricity into wholesale and retail markets, sale of voltage control services to the local network, and avoiding network augmentation are other value steams available to microgrids. Table 6.1 summarizes the types of value streams and the ways in which economic and commercial value are delivered.
6.3 Key Elements in Local Markets Framework In setting the framework for a local market, several key elements should be considered. Key elements in the local markets framework can be grouped into five layers as discussed in [41]. – Component layer: Active participation of end users is essential in local market development. Key enablers for active participation are smart meters, energy management systems, and communication devices. Smart meters provide data about demand and generation at different times. Energy management system enables market participants to determine the amount of energy for trade at different times of the day based on the energy requirements and preferences, weather conditions, and market prices. A communication device assists market players with connecting to the market platform through the communication layer.
Reduce peak demand, peak power flow in the network
Lowering ancillary service costs by providing distributed provision of frequency control services
It can change the wholesale price, depending on scale, time of day Substituting centralized voltage control with distributed control defers the investment for reactors and capacitor banks
Peak demand management
Frequency control services to the power system
Trading energy in wholesale market and with retailer Providing voltage control services to the local network
NSP: network service provider
Market impact Control demand side to maintain system reliability
Type of value stream Demand side response
Table 6.1 Local Markets value streams for Microgrids
Deferred investment in centralized voltage control equipment Improved quality of service in weak networks
Increased market efficiency
Reduced costs and increased revenue for generators by providing frequency control
NSP: Cost saving from avoided network investment and associated maintenance costs
Source of economic value NSP: Cost saving from avoided network investment and associated maintenance costs
Source of commercial value Reduces energy costs for customers. Potential increase in NSP’s profit by deferring capital investment Reduced network charges for customers by lowering peak demand, and importing energy during peak periods Increased profit for prosumers who provide frequency control which is more than offsets the lost value from reduced power production from solar power or battery Customers benefit from pooling their energy for these markets Customers benefit from providing network ancillary services for voltage control
6 Microgrids and Local Markets 157
158
M. Khorasany and R. Razzaghi
– Information layer: This layer includes a secure information system that enables all market participants to exchange required information during the negotiation and energy trading. In implementing local markets, a secure information system is an essential element, which enables market players to be integrated with the market mechanism. – Communication layer: This layer embraces Information and Communication Technology (ICT) devices that are essential for the communication of all market participants. – Function layer: The key elements of this layer are market settlement and pricing mechanism, trading approach, and market participants trading strategies. The market settlement indicates the allocation of energy, pricing mechanism, and payment rule. Trading approach guides the market operation by specifying the way market participants interact in the local market. The strategy of participants in reacting to market signals is another element of the function layer. – Business layer: The role of stakeholders and their interaction are defined in the business layer. The main stakeholders involved in the local market are consumers, producers, prosumers, local market operators, distribution system operators, and retailers. In the case of microgrid local market, the role of the market operator can be fulfilled by a microgrid operator. In order to leverage economic benefits from local markets, appropriate business models need to be designed [42]. A business model describes benefits that an enterprise will deliver to customers, how it will do so, and how it will capture a portion of the value it delivers [43]. Local markets change the traditional unidirectional energy value chain to a multidirectional by allowing the participation of new stakeholders, and hence, the business models for local markets are known to be disruptive [44]. Different business models for local markets can be considered, and for each model, a unique market model needs to be designed.
6.4 Local Market Models The market model for microgrids depends on the trading approaches, and the way market participants interact during market settlement. Different trading floors can be formed in microgrids, in which market participants can trade at the microgrid level, trade with other microgrids at a higher level, or trade with upstream markets. The formation of different trading floors enables market participants to participate in virtual bidding for trading at different levels. During the bidding process, market players interact with each other and can negotiate to agree on the price and amount of energy to be traded. The virtual bidding is enabled by online services based on the communication and information layers, which provide the required infrastructure for communication and information exchange between market players. The trading approaches in local markets can be classified based on the interaction of market participants as pool-based, P2P, and hybrid trading, which are explained in the rest of this section. The approaches implemented for market settlement can
6 Microgrids and Local Markets
Trading approach
159
Pool-based
P2P
Hybrid
Local Market Model for Microgrids
Market settlement approach
Centralized
Distributed
Decentralized
Fig. 6.2 Formation of different local market models based on the trading and market settlement approaches
also change the local market model, as the interaction of market players in the market depends on the market settlement mechanism. The combination of different trading approaches and market settlement mechanisms yields different models for local markets, as illustrated in Fig. 6.2.
6.5 Trading Approaches in Local Markets 6.5.1 Pool-Based Trading In pool-based trading (Fig. 6.3a), which is also known as indirect customer-tocustomer trading [45], there is a community manager who coordinates trading among sellers and buyers and clears the market. Pool-based trading allows market players to pool their resources to achieve their objectives such as reducing their costs, increasing their revenues, and utilizing their assets more efficiently [46]. The community manager can be a nonprofit operator, responsible for providing the platform for market participants and facilitating trading activities [47]. Also, it can be a third party that aggregates and orchestrates players in the local market and interacts with external markets to maximize its own profit as well as market players’ profits [28]. Microgrids can be used as a benchmark for pool-based trading. Each microgrid can form a local community, where community members can collaborate with each other to share their investments and reduce their costs. For instance, [46] proposes a framework for a community microgrid to share revenues and costs among members, where a marginal pricing scheme is implemented as a pricing mechanism in the local market. A community market to support residents in a community microgrid is presented in [48], which allows agents to effectively dispatch their resources by estimating the operation of the market.
160
M. Khorasany and R. Razzaghi
Fig. 6.3 Different trading approaches in local markets
6.5.2 P2P Trading The other trading approach is P2P trading illustrated in Fig. 6.3b, which is also known as “Energy AirBnB” [49]. P2P trading allows market players to directly interact with each other without the intervention of an intermediary entity. It provides a democratized environment, in which players can freely communicate with each other. Microgrids provide infrastructures and technologies required for monitoring, communication, and control that are important enablers for P2P markets [50]. P2P trading among prosumers and consumers in a microgrid increases cost saving in the microgrid. DER owners can achieve effective cost saving compared with directly trading with the utility grid under the feed-in-tariff scheme. Microgrids as the benchmark for P2P trading have recently received increasing momentum as an innovative solution for allowing prosumers to engage directly in energy trading. For example, a P2P market for a community microgrid is proposed in [39], where the game-theoretic approach was employed to model the interaction of sellers and buyers. Key drivers for prosumers’ participation in P2P microgrids are identified in [51], where motivating initial user participation and ensuring the long-term sustainability of microgrid development are introduced as fundamental goals of P2P microgrid development. Authors of [52] propose a motivational gametheoretic approach for P2P trading in which different motivational models that a prosumer needs to satisfy before being convinced to participate in energy trading are considered. Different market paradigms for P2P trading in a community microgrid is investigated in [46]. The main difference between community-based and full P2P markets is in the pricing, where in a P2P structure each transaction can have a singular price, whereas in a community-based market all transactions usually have a uniform price. Nonetheless, there are several P2P market models, which consider a uniform price for all transactions [53].
6 Microgrids and Local Markets
161
6.5.3 Hybrid Trading Besides pool-based and P2P market models, there are other models, which allow both pool-based trading and bilateral trading with neighborhood markets, named as hybrid models (Fig. 6.3c). In the hybrid market structure, players can trade within a community (e.g., a microgrid) in a pool-based market at the lower level, and communities (e.g., microgrids) can trade with each other in a P2P based market at the upper level. For instance, [54] presents a three-level energy trading platform for P2P trading within a microgrid, trading between multi microgrids, and trading between cells (multi microgrids). A feeder-based local market is proposed in [55], where prosumers can trade energy locally and with neighborhood areas. The hierarchical trading scheme in [56] enables energy trading among nanogrids, community microgrids, and multi-microgrid systems. A hybrid trading scheme is presented in [57], which enables market players to trade energy in transactive energy markets at different levels, including community, trading with neighborhood areas, and traditional trading with the upstream grid.
6.6 Market Settlement Approaches in Local Markets In local markets, the market settlement and supply and demand balancing can be carried out in three ways: centralized, distributed, and decentralized. In the centralized method, a central operator gathers information and data from all market players to settle the market. The central operator is the decision maker for all players and sends control signals to them to manage their actions in the market. Hence, prosumers and consumers cannot actively engage in the market. This approach is easy to implement and does not need to install new infrastructure in market players’ premises. However, centralized methods can endanger the privacy of players, as they need to reveal their private information with the central operator. Also, there are concerns about the scalability of these approaches, when employing for a largescale market. These issues can be resolved by using distributed and decentralized methods. In these methods, all market players are individual decision makers who share information and reach an agreement once they agree on the value of the shared information. These approaches can be implemented as an iterative price negotiation mechanism, where a signal such as voltage, estimated power mismatches, and the market price will be shared among market players to coordinate them. By applying distributed and decentralized methods, the computation can be distributed among players, which reduces computation complexity. These approaches protect market players’ privacy and decrease the number of communication links. The main difference between distributed and decentralized methods is that decentralized methods remove the need for a coordinator and there is no need for communicating with a third party. Hence, decentralized methods are appropriate for market structure
162
M. Khorasany and R. Razzaghi
with P2P trading, in which all peers directly negotiate to reach agreement on their activities in the market. In the following sections, three different algorithms for market settlement in local markets are introduced, namely, auction-based, distributed, and decentralized approach.
6.6.1 Auction-Based Approach Auctions are widely used for market settlements in different markets. An auction is a negotiation mechanism, where market players can express their interest in energy trading by submitting their offers or bids. Depending on the market model, different types of auctions can be implemented for local markets. One-sided auction is appropriate for flexibility markets with one seller and several buyers. Twosided auctions can be applied for energy markets with several sellers and buyers to enhance the social welfare of the market players through participation in a competitive market [58]. Double auction has received significant attention in the literature as a market settlement method in local markets and microgrids [59– 62]. In the auction-based approach, a market settlement has three steps, namely Determination where the number of sellers and buyers who win the auction is identified, Allocation where energy transfer from sellers to buyers is resolved. The third step is Payment, in which the auction prices for buyers and sellers for traded energy are settled [63]. Both uniform price and differential pricing can be applied for the payment step. Figure 6.4 illustrates an auction-based settlement for a double auction. Determination: In the double auction, sellers and buyers submit their offers/bids to the market operator. Let denote seller’s offer with Oi = (pi , λi ), where pi is the amount of offered energy by seller i, and λi is the reservation price for this offer. The buyer’s bid is denoted by Oj = (pj , λj ), where pj is demanded energy by buyer j, and λj is the corresponding price for this bid. After collecting all offers and bids, sellers’
Sellers
Buyers Offers
Payment/Allocation
Fig. 6.4 Auction-based settlement
Auctioneer
Bids
Payment/Allocation
6 Microgrids and Local Markets
163
offers are arranged in ascending order and buyers’ bids are arranged in descending order to generate aggregated demand and supply curves. The intersection of these curves indicates players who can participate in the market, such that the first K buyers with λj higher than the price in intersection point and the first L sellers with λi lower than the price at the intersection point can participate in the market. This step determines the allocated energy from sellers to buyers Allocation: pi∗ , pj∗ . Energy allocation from sellers to buyers is in a greedy manner, which means that energy from the seller with the lowest reservation price should be allocated to the buyer with the highest bid. Depending on market players energy requirements, they can participate in the market in two ways: (i) Fractional participation: the allocated energy to sellers/buyers can be a fraction of their offers/bids, (ii) Non-fractional participation: Players can only win for their total offers/bids energy or lose. Energy allocation can be considered as a knapsack problem and can be done from sellers’ or buyers’ perspectives. Payment: After allocating energy from sellers to buyers, the next step is to indicate the price for transactions in the market (λ∗ ). Different auction mechanisms use diverse rules for payment and these rules can be divided into two groups; (1) Fixed price: in which the price for all players in the market is the same and (2) Variable price: in which based on the allocation rule, different buyers should pay different prices and sellers will be paid at different prices. Different scenarios for payment rules are tabulated in Table 6.2.
Table 6.2 Different payment rules Fixed price
Payment mechanism Uniform price
VCG mechanism
Average mechanism
Vickery Trade reduction mechanism
Variable price
Pay as bid Generalized second price
Description Auction price for all buyers and sellers in the same and is equal to lowest winning bid or highest losing bid Each buyer pays the lowest equilibrium price max (λL , λK + 1 ), and each seller receives the highest equilibrium price min (λK , λL + 1 ) Auction price for all buyers and sellers in the same and is equal to average of all bids and offers of sellers and buyers All buyers pay the price equal to the second highest winning offer The first L − 1 sellers sell energy with price λL and the first K − 1 buyers receive the item and pay λK to the auctioneer Buyers pay their bid and sellers receive what buyers pay The highest bidder pays the price bid by the second highest bidder, the second-highest pays the price bid by the third-highest, and so on. Sellers get what the buyers pay
164
M. Khorasany and R. Razzaghi
6.6.2 Optimization-Based Approach 6.6.2.1
Distributed Clearing
In the distributed market clearing, market players optimize their own objectives and participate in the market to maximize their profits. Market players iteratively communicate through two-way communication links with a coordinator [64] to reach an agreement on their actions in the market. The coordinator is responsible for clearing the market and determines the clearing price to balance demand and supply. The coordinator needs only the demanded/supplied power by each player to clear the market such that utility and cost function parameters would remain private. To explain the distributed method for market settlement, we consider a local market consists of NS sellers indexed by i ∈ NS {1, . . . , NS } , and NB buyers indexed by j ∈ NB {1, . . . , NB }. We consider a general model for market players equipped with different assets and capability to manage these assets in response to market signals. Market players are defined by the range of flexibility that they can provide in the local market, which allows us to model different DERs as market players. The distributed market settlement is visualized in Fig. 6.5. The market objective is to maximize the social welfare of all players in the local market. Social welfare is the sum of all sellers and buyers surplus, which contributes to more user comfort with lower utility company costs [65]. The total welfare of players in the local market can be maximized through social welfare maximization, considering that each individual player’s profit is maximized too [66]. Social welfare can be formulated as the sum of the welfare of all players as in (6.1), and welfare of sellers and buyers can be modeled with (6.2) and (6.3), respectively: SW tot =
NS
SW i +
i=1
Fig. 6.5 Optimization-based settlement: distributed
NB j =1
SW j
(6.1)
6 Microgrids and Local Markets
165
SW i = λpi − Ci (pi )
Ci (pi ) =
Uj yj =
1 αi pi2 + βi pi + γi 2
(6.2a)
(6.2b)
p i ≤ pi ≤ p i
(6.2c)
SW j = Uj pj − λpj
(6.3a)
*
βj pj − βj2 /2αj ,
αj pj2 2 ,
0 ≤ pj ≤ βj /αj pj ≥ βj /αj
pj ≤ pj ≤ p j
(6.3b)
(6.3c)
where, pi and pj are traded energy by seller i and buyer j, respectively and λ is the market clearing price. We employ a quadratic cost function to approximate the cost of providing energy by the seller to the market as in (6.2b), with α i , β i , and γ i as predetermined positive constants, denoting the seller’s willingness to sell energy at different prices. These parameters depend on the type of generation and are private information of the seller, which should not be revealed during the market clearing process. The satisfaction level of the buyers from consuming the power is modeled by a quadratic utility function as in (6.3b), with α j and β j as positive constants determining the willingness of the buyer to pay for different levels of energy [68]. Substituting (6.2a) and (6.3a) in (6.1), yields (6.4) which is a convex optimization problem. Market clearing should manage demand and supply such that the total demanded energy by buyers is equal to the total supplied energy by sellers at the end of the market clearing as in (6.5), which is the constraint of the objective function in (6.4). max
pi ,pj
NB
NS Uj p j − Ci (pi )
j =1
(6.4)
i=1
Subject to. (6.2c), (6.3c), and NS i=1
pi =
NB j =1
pj .
(6.5)
166
M. Khorasany and R. Razzaghi
The optimization problem is a convex problem and can be relaxed using Lagrangian multipliers. Then, dual decomposition can be employed to develop a distributed iterative approach to maximize (6.4) subject to constraints without any need to have individual parameters of all market players. The updating rules for primal variables (supply and demand), and dual variable (price), are developed using the primal-dual gradient descent method [67] as presented in (6.6), (6.7), and (6.8), respectively pik+1 = argmin Ci pik − λk pik
(6.6)
λk pjk − Uj pjk
(6.7)
pi ≤pi ≤pi
pjk+1 = argmin
pj ≤pj ≤pj
⎡
λk+1
⎛ ⎞ ⎤+ NS NB = ⎣λk − ρ ⎝ pik = pjk ⎠⎦ i=1
k+1 − λk < λ
(6.8)
j =1
(6.9)
where λ represents the Lagrangian multiplier and is the same as the market clearing price, k is iteration index, ρ denotes step size and the notation [.]+ denotes max (0, .). In each iteration, sellers and buyers update their supply and demand using (6.6), (6.7) respectively and send their updated supply/demand to the market operator. Then, the market operator updates the market price using (6.8) and sends the updated price to all market players. This algorithm repeats to meet the stopping criterion in (6.9).
6.6.2.2
Decentralized Clearing
Another approach for market settlement in local markets is decentralized clearing, where market players can directly negotiate without the interaction with any central entity, as represented in Fig. 6.6. This method is different from the distributed method as there is no need for a supervisory node. In this section, a decentralized market clearing method using a primal-dual gradient algorithm is explained. This algorithm is a resource allocation problem to dispatch generation and demand of market players in the local market. In the decentralized clearing, each transaction can have a different price depending on players’ agreements in each bilateral trade. Hence, (6.2a) and (6.3a) can be rewritten as (6.10) and (6.11) respectively:
6 Microgrids and Local Markets
167
Fig. 6.6 Optimization-based settlement: decentralized
SW i =
NB λij pij − Ci (pi )
(6.10a)
j =1
pi =
NB
pij
(6.10b)
j =1 NS SW j = Uj pj − λij pij
(6.11a)
i=1
pj =
NS
pj i
(6.11b)
i=1
where, pij is the traded energy between seller i and buyer j, which should be the same as pji , and λij is the price for this transaction. The market objective is the same as (6.1) to maximize the social welfare of all players subject to constraints. In each bilateral trade, the offered energy by the seller should be equal to the demanded energy by the buyer, at the end of the market settlement as in (6.12): pij = pj i ,
∀i ∈ NS , ∀j ∈ NB
(6.12)
The optimization problem can be solved centrally or in a distributed manner. Central optimization needs a central controller who is aware of all market players’ characteristics and endangers the privacy of the players. Distributed optimization resolves this issue by decomposing the problem into several locally solvable subproblems. However, it needs a coordinator to coordinate local optimizations.
168
M. Khorasany and R. Razzaghi
The decentralized optimization eliminates the need for a coordinator, as all data and related computations are handled locally by agents. To design a decentralized solution for the optimization problem, constraints in (6.2c), (6.3c), and (6.12c) should be relaxed using Lagrangian multipliers as in: NS NB L pij , pj i , λij , μi , μi , μj , μj = Uj p j − Ci (pi ) j =1
+
NS NB
λij pij − pj i −
i=1 j =1
−
NB
NS
i=1
μi (pi − pi ) −
i=1
μi pi − pi
i=1
NB μj pj − pj − μj pi − pj
j =1
NS
(6.13)
j =1
where, λij , μi , μi , μj , μj are Lagrangian multipliers associated with constraints. Updating rules for both primal and dual variables can be developed using the proposed method in [69]. In this method, sellers update their prices for different buyers and send these prices to them, and buyers need to respond to these prices by updating their demanded energy from different sellers. Sellers and buyers use (6.14) and (6.15) respectively to update their parameters in the negotiation process. ,+ + k λk+1 = λkij − ρλk pij − pjki ij
(6.14a)
,+ + k k k p μk+1 = μ + ρ − p i μ i i i
(6.14b)
+ k k k p μk+1 = μ + ρ − p i μ i i i
(6.14c)
,+ + k+1 k+1 k pij = pij + ζijk p˜ ij − pik
(6.14d)
k+1 p˜ ij
=
k+1 + μk+1 − βi λk+1 ij − μi i
αi
k + vk pij ζijk = # NB k + vk p ij j =1
,+ + k k k p μk+1 = μ + ρ − p j μ j j j
(6.14e)
(6.14f)
(6.15a)
6 Microgrids and Local Markets
169
μk+1 j
=
μkj
+ ρμk
+
pj − pjk
,+ + k+1 k k k pjk+1 i = pj i + ζj i p˜ j i − pj
p˜ jk+1 i =
k+1 + μk+1 βj − λk+1 ij − μj j
αj
pjki + v k ζijk = # NS k + vk p ji i=1
(6.15b)
(6.15c)
(6.15d)
(6.15e)
k+1 k+1 , p˜ ij are power set points of sellers and buyers in each transaction, where, p˜ ij which are obtained by taking the first-order derivative of the relaxed problem in (6.13) with respect to pij , and pji , and ζijk , ζijk are asymptotically proportional factors. In each iteration, sellers update their prices for different transactions using (6.14a) and their offered energy using (6.14b–f). Then, all buyers in parallel update their demanded energy from different sellers using (6.15a–e). The stopping criteria for the negotiation process are as
k+1 λij − λkij <
(6.16a)
k+1 μi − μki <
(6.16b)
k+1 μi − μki <
(6.16c)
k+1 μj − μkj <
(6.16d)
k+1 μj − μkj < .
(6.16e)
6.7 Case Example of Local Markets for Microgrids: The Monash Microgrid The Monash Microgrid is a case example of real-world implementation of local markets for microgrids [70]. The microgrid system is a fully functioning local electricity network and trading market with dynamic optimization of resources
170
M. Khorasany and R. Razzaghi
interacting with external energy markets. The Monash Microgrid is grounded on the idea of orchestrating and coordinating DERs through a transactive energy market (TEM), where each DER will act as an independent customer that will, where it chooses to, offer and commit to providing their flexibility as a commercial service to the TEM. The TEM will then be able to use this internal market functionality to aggregate the microgrid’s available flexibility. The microgrid provides the complete hardware and software requirements to implement a marketplace at the local level to trade energy and flexibility in both internal and external markets. A distributed and ICT-oriented real-time integration platform is implemented to perform the monitoring and control of the microgrid and all connected components to support a real market implementation. The Monash Microgrid includes a range of customers and DER assets such as flexible buildings, a battery storage system, solar generation units, and EV chargers that participate in the market independently to provide the requested services. The local market enables microgrid’s participants to trade electricity and access revenues from the market and ancillary services they provide to the broader electricity network (such as demand response and frequency and voltage control). The establishing of a competitive internal market provides the Monash Microgrid with the opportunity to reduce pressure on the network during high and peak demand, sell renewable generation in the wholesale market, provide frequency and voltage control services to the grid through the ancillary services market, and help the grid respond to emergencies.
6.8 Case Examples of Market Settlement in Local Markets This section provides illustrative examples of different market settlement approaches for a local market. A microgrid with four sellers and five buyers is considered, where market players participate in a forward market with a duration of one hour to trade energy locally. Market players’ parameters are given in Table 6.3. First, a double auction with a uniform price is implemented to clear the market. Sellers and buyers submit their offers/bids to participate in the auction. Each seller/buyer participates in the market by offering/bidding its maximum generation/demand and uses (6.2a)/(6.3a) to calculate its marginal price to maximize its welfare. After receiving all offers and bids, an auctioneer generates the aggregated demand-supply curves and allocates energy from sellers to buyers at the clearing Table 6.3 Market players parameters
i
pi
pi
αi
βi
j
pj
pj
αj
βj
1 2 3 4
0 0 0 0
6 5 5 3
0.168 0.15 0.102 0.158
4.21 6.35 8.98 10.25
1 2 3 4 5
0 0 0 0 0
6 4 4 3 2
0.144 0.168 0.102 0.158 0.122
8.21 11.21 12.98 13.24 15.21
6 Microgrids and Local Markets
171
Fig. 6.7 Simulation results: auction-based settlement
price. The market clearing price for all players is λ∗ = 9.49 ¢/kWh. Market settlement results are illustrated in Fig. 6.7. In this case study, the first three sellers and first four buyers win the auction, as their offers/bids are lower/higher than market clearing price. The total welfare of market players is 83.72 ¢, where among sellers, the first seller has the highest welfare in the market due to having the lowest marginal cost as λ1 =5.21 ¢/kWh. The second implemented method for market clearing is a distributed method, where market players use (6.6) and (6.7) to update their supply and demand and the coordinator updates the price at each iteration to clear the market. This method can be interpreted as an iterative auction process, where market players actively negotiate on their actions in the market. The results for this approach are shown in Fig. 6.8. The total welfare of all players is 83.77 ¢. This approach allows market players to optimize their actions at any iteration, while in the auction-based approach, players cannot update their actions once the market is settled. Although auction-based and distributed methods are easy to implement, the need for a central coordinator makes it challenging to develop these methods for the market with a large number of players. Hence, decentralized market settlements can be employed to clear the market without any interaction of a third party. Similar to the distributed approach, the decentralized method allows market players to actively negotiate their actions in the market. This method is implemented for the presented market and results are illustrated in Fig. 6.9. Figure 6.9a shows results from the seller 1 perspective, where allocated energy from seller 1 to all buyers is indicated. As in this approach, sellers are price makers, they use the same price for all buyers, and buyers demand different amounts of energy from sellers based on their preferences. Market settlement from buyer 3 perspective is presented in Fig. 6.9b. Due to the fact that buyers can negotiate with different sellers simultaneously, the offered prices by sellers 1 to 3 are converged to the same value, while seller 4 is not selling any energy to the market due to its high offered
172
M. Khorasany and R. Razzaghi
Fig. 6.8 Simulation results: distributed settlement
price. Results verify that the decentralized method can reach approximately the same results as the distributed method. However, it needs a two-way communication network between all market players. Also, a higher number of iterations is needed for market settlement in a decentralized method. As a comparison, the required number of iterations for players to reach an agreement in a distributed method is 9, while this value for a decentralized method is 293.
6.9 Conclusions Deployment of local markets in microgrids facilitates the integration of DERs in the power systems. Local markets enable DERs to provide services to networks and to be rewarded for these services. Microgrids can be integrated with communication networks, smart devices, and monitoring systems to provide the required platform for local markets. Through establishing a local market, microgrids can play the role of a flexible power supply or a flexible load by aggregating flexibilities of different resources. This chapter has reviewed local market attributes including, definition, benefits, objective, and services which can be provided through local markets. Different trading approaches for local markets are presented. Three different methods for market settlement are discussed, namely auction-based, distributed, and decentralized clearing. For each method, a detailed formulation is presented. Also, case examples are provided to demonstrate market settlement methods for local markets. Research in the field of local markets for microgrids is a new trend, emerged from interest to employ more DERs and decentralization of energy markets. Future
6 Microgrids and Local Markets
173
Fig. 6.9 Simulation results: decentralized settlement, 1) Seller 1 perspective, 2) Buyer 3 perspective
studies should analyze regulatory regimes to determine how microgrid local markets fit into current energy policies. Compared to traditional markets, local markets are smaller in size. Thus, they would have a different dynamic, which might require different regulations. An important aspect of market regulation is to allocate costs such as network utilization costs to market players at a reasonable rate to motivate them for active participation in the market. Communication networks are the key enablers for local market implementation. Without having a two-way communication network, the active participation of mar-
174
M. Khorasany and R. Razzaghi
ket players in the local market is not feasible. An appropriate network architecture needs to be designed for microgrids to serve as a benchmark for local markets. Information flow among microgrid components raises privacy protection concerns due to cyberattack issues, which raises the need for countermeasure techniques to keep private data secure with minimal loss and latency. Acknowledgments Icons in Figs. 6.1 and 6.3 to 6.6 are made by Freepik, Eucalyp, Pixel Perfect, and Smashicons from www.flaticon.com.
References 1. Pudjianto, D., Ramsay, C., & Strbac, G. (2007). Virtual power plant and system integration of distributed energy resources. IET Renewable Power Generation, 1(1), 10–16. 2. Eftekharnejad, S., et al. (2013). Impact of increased penetration of photovoltaic generation on power systems. IEEE Transactions on Power Systems, 28(2), 893–901. 3. Energy Networks Australia, AEMO, Interim Report: Required Capabilities and Recommended Actions. Technical Report, 2019. 4. Morales, J. M., et al. (2013). Integrating renewables in electricity markets: Operational problems (Vol. 205, pp. 1–13). Springer Science & Business Media. 5. Hirsch, A., Parag, Y., & Guerrero, J. (2018). Microgrids: A review of technologies, key drivers, and outstanding issues. Renewable and Sustainable Energy Reviews, 90, 402–411. 6. Jimeno, J., et al. (2011). Architecture of a microgrid energy management system. European Transactions on Electrical Power, 21(2), 1142–1158. 7. Koirala, B. P., et al. (2016). Local alternative for energy supply: Performance assessment of integrated community energy systems. Energies, 9(12), 981. 8. Mengelkamp, E., et al. Trading on local energy markets: A comparison of market designs and bidding strategies. in 14th International Conference on the European Energy Market (EEM). IEEE, June 2017. Dresden, Germany. 9. Ton, D. T., & Smith, M. A. (2012). The US department of energy’s microgrid initiative. The Electricity Journal, 25(8), 84–94. 10. Ilieva, I., et al. Design characteristics of a smart grid dominated local market. in CIRED Workshop. IEEE, June 2016, Helsinki, Finland. 11. Lamparter, S., S. Becher, and J.-G. Fischer. An agent-based market platform for smart grids. in Proceedings of the 9th international conference on autonomous agents and multiagent systems: industry track. May 2010, Toronto, Canada. 12. Cox, W. and T. Considine, Energy, micromarkets, and microgrids. Grid-Interop 2011, December 2011, Phoenix, AZ, USA: p. 1-8. 13. Bremdal, B.A., P. Olivella, and J. Rajasekharan. EMPOWER: A network market approach for local energy trade. in 2017 IEEE Manchester PowerTech. IEEE, June 2017, Manchester, United Kingdom. 14. Munné-Collado, ´i., et al. (2019). Local and micro power markets. In Micro and local power markets (pp. 37–96). Wiley. 15. Da Silva, P. G., Ili´c, D., & Karnouskos, S. (2013). The impact of smart grid prosumer grouping on forecasting accuracy and its benefits for local electricity market trading. IEEE Transactions on Smart Grid, 5(1), 402–410. 16. Hvelplund, F. (2006). Renewable energy and the need for local energy markets. Energy, 31(13), 2293–2302. 17. Olivella-Rosell, P., et al. (2018). Optimization problem for meeting distribution system operator requests in local flexibility markets with distributed energy resources. Applied Energy, 210, 881–895.
6 Microgrids and Local Markets
175
18. Melton, R. (2015). Pacific northwest smart grid demonstration project technology performance report volume 1: Technology performance. Richland, WA: Pacific Northwest National Lab. (PNNL). 19. Holmberg, D. G., et al. (2016). Transactive energy application landscape scenarios. Technical Report. 20. Kok, J.K., C.J. Warmer, and I. Kamphuis. PowerMatcher: multiagent control in the electricity infrastructure. in Proceedings of the fourth international joint conference on Autonomous agents and multiagent systems. July 2005. Ultrecht, The Netherlands. 21. Rahimiyan, M., & Baringo, L. (2015). Strategic bidding for a virtual power plant in the dayahead and real-time markets: A price-taker robust optimization approach. IEEE Transactions on Power Systems, 31(4), 2676–2687. 22. Lane, C. W., et al. (2013). Costing consumer preferences for a micro energy market. Center for nation reconstruction and capacity development, Technical Report. 23. Jogunola, O., et al. (2017). State-of-the-art and prospects for peer-to-peer transaction-based energy system. Energies, 10(12), 2106. 24. Ela, E., et al. (2017). Electricity markets and renewables: A survey of potential design changes and their consequences. IEEE Power and Energy Magazine, 15(6), 70–82. 25. Pavlovic, M., et al., SGAM business layer for a local flexibility market. in CIRED Workshop. June 2016. Helsinki, Finland. 26. Zhang, C., et al. A flex-market design for flexibility services through DERs. in IEEE PES ISGT Europe. IEEE, October, 2013. Copenhagen, Denmark. 27. Torbaghan, S.S., et al. Local market framework for exploiting flexibility from the end users. in 13th International Conference on the European Energy Market (EEM). IEEE, June 2016, Porto, Portugal. 28. Morstyn, T., Teytelboym, A., & McCulloch, M. D. (2018). Designing decentralized markets for distribution system flexibility. IEEE Transactions on Power Systems, 34(3), 2128–2139. 29. Vicente-Pastor, A., et al. (2018). Evaluation of flexibility Markets for Retailer–DSO–TSO coordination. IEEE Transactions on Power Systems, 34(3), 2003–2012. 30. Olivella-Rosell, P., et al. (2018). Local flexibility market design for aggregators providing multiple flexibility services at distribution network level. Energies, 11(4), 822. 31. Mengelkamp, E., et al. (2018). Designing microgrid energy markets: A case study: The Brooklyn microgrid. Applied Energy, 210, 870–880. 32. Nunna, H. K., & Srinivasan, D. (2017). Multiagent-based transactive energy framework for distribution systems with smart microgrids. IEEE Transactions on Industrial Informatics, 13(5), 2241–2250. 33. Faqiry, M. N., & Das, S. (2016). Double-sided energy auction in microgrid: Equilibrium under price anticipation. IEEE Access, 4, 3794–3805. 34. Li, P., Zan, P., & Xu, H. (2016). Integrated energy exchange scheduling for multi-microgrid system with electric vehicles. IEEE Transactions on Smart Grid, 7(4), 1762–1774. 35. Park, S., et al. (2016). Contribution-based energy-trading mechanism in microgrids for future smart grid: A game theoretic approach. IEEE Transactions on Industrial Electronics, 63(7), 4255–4265. 36. Wang, H., et al. (2016). Reinforcement learning in energy trading game among smart microgrids. IEEE Transactions on Industrial Electronics, 63(8), 5109–5119. 37. Park, S., et al. (2017). Event-driven energy trading system in microgrids: Aperiodic market model analysis with a game theoretic approach. IEEE Access, 5, 26291–26302. 38. Wang, H., & Huang, J. (2016). Incentivizing energy trading for interconnected microgrids. IEEE Transactions on Smart Grid, 9(4), 2647–2657. 39. Paudel, A., et al. (2018). Peer-to-peer energy trading in a prosumer-based community microgrid: A game-theoretic model. IEEE Transactions on Industrial Electronics, 66(8), 6087– 6097. 40. Liu, N., et al. (2017). Energy-sharing model with price-based demand response for microgrids of peer-to-peer prosumers. IEEE Transactions on Power Systems, 32(5), 3569–3583.
176
M. Khorasany and R. Razzaghi
41. Kirpes, B., et al. (2019). Design of a microgrid local energy market on a blockchain-based information system. Information Technology, 61(2-3), 87–99. 42. Mengelkamp, E., Schlund, D., & Weinhardt, C. (2019). Development and real-world application of a taxonomy for business models in local energy markets. Applied Energy, 256, 113913. 43. Chesbrough, H., & Rosenbloom, R. S. (2002). The role of the business model in capturing value from innovation: Evidence from Xerox Corporation’s technology spin-off companies. Industrial and Corporate Change, 11(3), 529–555. 44. Richter, M. (2012). Utilities’ business models for renewable energy: A review. Renewable and Sustainable Energy Reviews, 16(5), 2483–2493. 45. Chen, T., Pourbabak, H., & Su, W. (2019). Electricity market reform. In The energy internet (pp. 97–121). Elsevier. 46. Cornélusse, B., et al. (2019). A community microgrid architecture with an internal local market. Applied Energy, 242, 547–560. 47. Khorasany, M., Y. Mishra, and G. Ledwich. Distributed market clearing approach for local energy trading in transactive market. in IEEE Power & Energy Society General Meeting (PESGM). IEEE, August 2018. Portland, OR, USA. 48. Shamsi, P., et al. (2015). Economic dispatch for an agent-based community microgrid. IEEE Transactions on Smart Grid, 7(5), 2317–2324. 49. Luo, J., W. Su, and A.Q. Huang. Bit-energy: An innovative bitcoin-style distributed transactional model for a competitive electricity market. in IEEE Power & Energy Society General Meeting (PESGM). IEEE, July 2017. Chicago, IL, USA. 50. Sousa, T., et al. (2019). Peer-to-peer and community-based markets: A comprehensive review. Renewable and Sustainable Energy Reviews, 104, 367–378. 51. Kirchhoff, H., & Strunz, K. (2019). Key drivers for successful development of peer-to-peer microgrids for swarm electrification. Applied Energy, 244, 46–62. 52. Tushar, W., et al. (2019). A motivational game-theoretic approach for peer-to-peer energy trading in the smart grid. Applied Energy, 243, 10–20. 53. Tushar, W., et al. (2018). Peer-to-peer energy trading with sustainable user participation: A game theoretic approach. IEEE Access, 6, 62932–62943. 54. Long, C., et al. (2017). Feasibility of peer-to-peer energy trading in low voltage electrical distribution networks. Energy Procedia, 105, 2227–2232. 55. Khorasany, M., Mishra, Y., & Ledwich, G. (2019). Two-step market clearing for local energy trading in feeder-based markets. The Journal of Engineering, 2019(18), 4775–4779. 56. Paudel, A. and G.H. Beng. A hierarchical peer-to-peer energy trading in community microgrid distribution systems. in IEEE Power & Energy Society General Meeting (PESGM). IEEE, August 2018. Portland, OR, USA. 57. Khorasany, M., Mishra, Y., & Ledwich, G. (2019). Hybrid trading scheme for peer-to-peer energy trading in transactive energy markets. IET Generation, Transmission & Distribution, 14(2), 245–253. 58. Munné-Collado, ´i., P. Olivella-Rosell, and A. Sumper, Power Market Fundamentals. Micro and Local Power Markets. Wiley, 2019: p. 1.35. 59. Bhattacharjee, S. and S.B. Raha. Double auction bidding based economics for a microgrid model. in Innovations in Power and Advanced Computing Technologies (i-PACT). IEEE, April 2017. Vellore, India. 60. Tushar, W., et al. (2016). Energy storage sharing in smart grid: A modified auction-based approach. IEEE Transactions on Smart Grid, 7(3), 1462–1475. 61. Khorasany, M., Y. Mishra, and G. Ledwich. Auction based energy trading in transactive energy market with active participation of prosumers and consumers. in Australasian Universities Power Engineering Conference (AUPEC). IEEE, November 2017. Melbourne, Australia. 62. El-Baz, W., Tzscheutschler, P., & Wagner, U. (2019). Evaluation of energy market platforms potential in microgrids: Scenario analysis based on a double-sided auction. Frontiers in Energy Research, 7, 41.
6 Microgrids and Local Markets
177
63. Khorasany, M., Mishra, Y., & Ledwich, G. (2019). Design of auction-based approach for market clearing in peer-to-peer market platform. The Journal of Engineering, 2019(18), 4813– 4818. 64. Pourbabak, H., et al. (2017). A novel consensus-based distributed algorithm for economic dispatch based on local estimation of power mismatch. IEEE Transactions on Smart Grid, 9(6), 5930–5942. 65. Zou, X. (2009). Double-sided auction mechanism design in electricity based on maximizing social welfare. Energy Policy, 37(11), 4231–4239. 66. Stiglitz, J.E., Whither socialism?. MIT press. 1996. (1). 67. Boyd, S., et al. (2011). Distributed optimization and statistical learning via the alternating direction method of multipliers. Foundations and Trends® in Machine learning, 3(1), 1–122. 68. Samadi, P., et al. Optimal Real-Time Pricing Algorithm Based on Utility Maximization for Smart Grid. in 1st IEEE International Conference on Smart Grid Communications. IEEE, October 2010. Gaithersburg, MD, USA. 69. Khorasany, M., Mishra, Y., & Ledwich, G. (2020). A decentralised bilateral energy trading system for peer-to-peer electricity markets. IEEE Transactions on Industrial Electronics, 67(6), 5930–5942. 70. Khorasany, M., et al. (2020). Transactive energy market for energy Management in Microgrids: The Monash microgrid case study. Energies, 13(8), 1–23.
Chapter 7
An Economic Demand Management Strategy for Passive Consumers Considering Demand-Side Management Schemes and Microgrid Operation Mohammad Esmaeil Honarmand, Vahid Hosseinnezhad, Barry Hayes, Behnam Mohammadi-Ivatloo, and Pierluigi Siano
7.1 Introduction In recent years, there are strong incentives to utilize electricity end users for reducing greenhouse gases, competitive energy policies, and participate in consumption management. The increased penetration of microgrids and the extended strategies of demand-side management have created an excellent opportunity for consumers to acquire these benefits in smart systems. Consequently, in this circumstance, a passive consumer, in addition to purchasing energy from the electricity market directly, can manage to supply its demand economically by choosing the right strategy. To this end, it can utilize the new concept of microgrids and local production to meet the need. Consumers can also participate in load management programs independently or integrated with the microgrid concept. Nowadays, with the development of smart infrastructures, microgrids can bridge the gaps between electricity market prices and consumer behavior. A microgrid can be considered as a single electrical load from a utility’s viewpoint, and from behind the consumer meter, a microgrid can function as a distributed energy resource [1]. Due to seasonal peak loads, increased network reliability, and energy shortages, this
M. E. Honarmand Gilan Electric Power Distribution Co., Rasht, Iran V. Hosseinnezhad () · B. Hayes School of Engineering, University College Cork (UCC), Cork, Ireland e-mail: [email protected] B. Mohammadi-Ivatloo Faculty of Electrical and Computer Engineering, University of Tabriz, Tabriz, Iran P. Siano Department of Management and Innovation Systems, University of Salerno, Fisciano, Italy © Springer Nature Switzerland AG 2021 A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems, https://doi.org/10.1007/978-3-030-59750-4_7
179
180
M. E. Honarmand et al.
controllable load can be an excellent platform for the implementation of demandside management schemes. Demand response (DR) programs as a part of demand-side management are powerful tools that facilitate the process of transforming conventional microgrids into green systems by consumption management and efficient utilization of renewable sources. These programs involve the modification of customer’s demand for energy by various means, such as financial incentives and behavioral change through educational approaches. Accordingly, microgrid and customer reliability will be improved by implementing the DR programs in microgrids. Besides, this benefit can be achieved through the reduction of demand during critical times. Recently, due to the propulsion of the smart grid paradigm, numerous efforts have been made to integrate the schemes of DR and microgrid operation. In [2], the impact of the customer participation level in emergency DR for microgrid operation is analyzed in the presence of different uncertainties. In addition, this work examines the effect of different incentives on total costs of operation by a model that is presented based on price elasticity and customer benefit. The scheduling problem based on various incentive rewards and constraints related to microgrid operation is presented in [3]. The authors of [4] propose an optimization function of microgrid operation with the purpose of minimizing operational costs and emissions by considering the DR programs. Due to the economic dispatch of a renewable microgrid, the operational costs are investigated in [5] by considering the participation of the consumers in DR schemes. To solve the economic dispatch to minimize the operation cost of microgrid, in [6], the authors proposed the model of different DR programs to prioritize running the plans in the presence of microgrid. In [7], the optimal operation of a microgrid is assessed by the combination of various DR programs. A mathematical model has been developed for the microgrid system considering the impact of different DR schemes to minimize the objective function. In order to maximize net income in [8], a cost–benefit analysis is presented to plan the operation of a grid-connected microgrid in combination with DR programs. The authors in [9] introduced an incentive payment oriented DR scheme for microgrid operational planning. A stochastic optimization approach is proposed to consider the presence of different types of customers. In [10], a comprehensive DR framework in a microgrid environment is proposed to mitigate peak demand and energy saving. Considering DR schemes for a microgrid retailer, an optimal strategy for energy dispatch and pricing is presented in [11]. An economic optimization model is introduced in [12] to manage microgrid and allocate the shiftable loads in the residential sector. The authors in [13] proposed a dynamic optimization model based on DR schemes to minimize the operation cost and maintain the supply–demand balance in microgrids. Despite the reviewed literature, the question that authors are interested in here is quite different. This work explores which strategy would be best suited for a typical passive load to manage its demand economically; purchasing electricity from the grid and participating in load management plans, or transferring to a microgrid and integrating with DR programs. To this end, developing an integrated procedure that covers both microgrid establishment and DR strategy effects from the economic
7 An Economic Demand Management Strategy for Passive Consumers. . .
181
viewpoint, can be vital. The proposed methodology here is concentrated to find a tradeoff point between the costs of microgrid and the DR. The scope of this work is to propose a procedure that can address this economic problem as much as possible. Accordingly, the intended method is aimed to consider microgrid costs and DR scheme costs regardless of the market mechanisms to manage the network load economically. On this base, first, the corresponding cost model for the installation and operation of a microgrid is formulated. For simplicity, in the formation of the microgrid, renewable energy-based units are not considered. Then, the output is evaluated alongside DR cost with the viewpoint of reaching the compromise point. Finally, the proper DR program related to the best performance is selected as the final strategy. Two DR programs including pricebased DR (PDR) and incentive-based DR (IDR) are considered in the studies. The introduced procedure is implemented on several real loads and is investigated under different case studies. The detailed results are presented in the analysis of significant contributions and benefits of the proposed method. This chapter covers a summary of DR programs in Sect. 2. Classification of microgrid applications is provided in Sect. 3. The cost model for installation and operation of the microgrid, the cost function related to the run of the various DR programs, and the proposed decision-making method are presented in Sect. 4. Finally, to investigate and analyze the proposed algorithm, three case studies are presented in Sect. 5.
7.2 Types of DR Programs DR can be defined as changes in electricity usage by end use customers from their usual consumption patterns in response to changes in prices. DR programs may be classified either according to how the enrolled consumers respond or by their type (motivation procedure and trigger criteria) regarding the characterization of their load. The U.S. Department of Energy (DOE) categorizes these programs into PDR and IDR [14]. PDR offers collaboration in time-varying rates that reflect the value and cost of electricity for different periods. However, in IDR consumers voluntarily provide load reductions by responding to economic signals. Indeed, the PDR includes the actual cost for the electricity, while the IDR provides customers with peak shaving incentives [15]. This section includes detailed discussions on some of the DR strategies for both categories. In the end, the overall impact of these strategies on microgrid operation is investigated.
7.2.1 PDR Strategies In different articles, various strategies have been studied for PDR. However, here, only three schemes are considered. These include time of use (TOU), critical peak
182
M. E. Honarmand et al.
pricing (CPP), and real-time pricing (RTP) [16]. These methods are used for cost modeling and economic evaluation in the proposed model. • Time of Use (TOU) pricing. Electricity consumers that are charged with flat prices are not aware of the varying cost of electricity. One way in which consumers will be incentivized to change their consumption patterns is through price signals delivered via TOU tariffs, in which the price of electricity varies depending on factors such as electricity network constraints and the wholesale price of electricity [17]. These tariffs are designed to more closely reflect the investment and the production cost structure, so key issues such as the duration of individual periods and related price levels are involved in the design of TOU rates. Many countries divide a year into peak periods and valley periods according to summer or non-summer months to charge differently. This pricing is easy to be implemented and has a great effect on load shifting, although it cannot be strictly called a dynamic pricing strategy for its high consistency [18]. Static TOU tariffs are the stepped rate structure, which varies into several periods (usually less than five periods) during the day fixedly and regularly. The pricing scheme of TOU tariff can be introduced as follows: ⎧ F1 , i ∈ (h1 , h2 ) ⎪ ⎪ ⎪ ⎨ F2 , i ∈ (h2 , h3 ) PT OU −i = (7.1) .. ⎪ ⎪ . ⎪ ⎩ Fk , i ∈ (hk−1 , hk ) In (7.1), a day is separated into k periods (h stands for hour) and a certain price level is provided for each period known as Fi . • Critical Peak Pricing (CPP). The long-term electricity supply costs associated with using electricity during a specific period of the day are reflected by TOU tariffs. In order to capture the short-term costs of considered critical periods for the power system, CPP tariffs may be employed [19]. This strategy, also called peak-load pricing, has both characteristics of TOU tariffs and emergency load control, therefore, it can be a supplement to TOU pricing which has some mandatory restrictions for electricity demand in critical peak periods. A CPP period is announced ahead of time, typically day-ahead, and customers on the CPP rate can reduce their bill by shifting or reducing their loads during these peak times. In this way, the higher rates on emergency or critical peak periods (e.g., unavailability of reserves, extreme outages, etc.) are charged by CPP tariffs, while the prices in other times remain the same [20]. The power utility may sign contracts with consumers to specify the maximum number of days per year that may be considered critical and the number of periods for which the CPP rate is applied otherwise; they will get some punishments. Equation (7.2) provides the pricing scheme of CPP tariff:
7 An Economic Demand Management Strategy for Passive Consumers. . .
PCP P −i =
⎧ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎨
RN 1 , RN k ,
i ∈ (h1 , h2 ) .. . i ∈ (hk−1 , hk )
183
If not Critical Periods
⎪ ⎪ ⎪ RC1 , i ∈ (hC−1 , hC−2 ) ⎪ ⎪ ⎪ .. ⎪ ⎪ ⎪ . ⎪ ⎪ ⎩ RCm , i ∈ hC−(m−1) , hC−m
(7.2) Otherwise
where RNk and RCm are the separated price levels for normal and critical tariffs, respectively. A day is split into k periods and also m part is introduced for critical times. • Real-time pricing (RTP). This strategy separates a day into several short time slots similar to the TOU tariff. Generally, RTP is an electricity pricing strategy that directly according to the real-time supply and demand situation, reflects the marginal value of electricity [21]. Therefore, prices vary in real time (e.g., an hour or a half-hour) depending on the current wholesale cost of electricity so it can be said that the scheme is theoretically most reasonable. Furthermore, RTP and realtime power load have a positive correlation for the general operating state. A typical relationship of the pricing scheme is obtained as follows [22]: PRT P −i = α × ALoad i
(7.3)
In (7.3), α is the rate between load and price, and A_Loadi is the overall consumed load of a certain end user. This correlation depends on different factors such as real-time status of the operation, line losses, wholesale price, and so on. Besides, electricity regulations and the policy of governments and other organizations, for example, ISOs, limit the setting of electricity prices in practice.
7.2.2 IDR Strategies In dynamic pricing, the load control scheme is defined without a third-party operator that manages load-shedding. However, IDR programs are employed by the power utility to control the load of the consumer directly or based on the response of the incentive measures by consumers. Here, three schemes of IDR include direct load control (DLC), emergency DR (EDR), and interruptible load program (ILP), are discussed. • Direct Load Control (DLC). Ordinarily, DLC programs involve a utility or system operator that allows them to switch on and off specific customers’ appliances for
184
M. E. Honarmand et al.
a short time during peak periods and critical events. In return for participating, consumers are usually rewarded by way of a financial incentive such as a one–off sign-up payment, recurring annual payment, and ongoing electricity bill discounts [23]. This type of program is mostly applied to residential or small commercial customers. PDR schemes could be supplemented by DLC to reinforce gains of these schemes. This service of contract-based reliability enhancing can be planned to control loads in 10–14 hours. • Interruptible load program (ILP). This program considers curtailment options to a predefined level. In order to turn off specific loads by participants in these services, customers receive a discount or bill credit in exchange to reduce load during system contingencies. Besides, participants may face penalties in case they fail to respond to a DR event [24]. These are offered for typical customer size from 200 kW up to 3 MW so that customers on these tariffs must curtail within 30–60 minutes when being notified by the utility. Also, the total amount of period that a utility can call interruption often is not more than 200 hours per year [25]. • Emergency DR (EDR) program. The incentive payments are considered to consumers due to the reducing power consumption during reliability triggered events. These programs can be also known as a combination of DLC and ILP programs. In contrast to ILP, since there is no contractual obligation, this scheme does not impose any penalties if consumers cannot participate [26]. However, these programs have a narrow application, and they are called a very limited number of times per year (less than 5).
7.2.3 DR Programs and Microgrid Operation As mentioned earlier, DR programs are divided into two categories. PDR programs depend on the behavior and response of electricity consumers to the suggested prices. Therefore, from the viewpoint of participants, the utilization of microgrids by the consumers is related to the associated cost–benefit. Due to the type of PDR programs, the operation of microgrid will be relatively long term (at least monthly). On this way, if the microgrid costs involved are higher than the cost of PDR implementation, the participant’s preference will be to use the DR scheme into the desired period. Assuming a variety of PDR schemes, the participation priority can be a criterion to choose the economic microgrid operation. Since the period and times of participation in IDR programs are predefined, it is possible to combine them with PDR programs. Generally, if the microgrid costs involved are lower than the cost of PDR implementation, then these costs can equilibrate with IDR program cost. With each IDR scheme implemented, the microgrid can be run for the customer’s electricity availability. Accordingly, when the IDR contract is signed, the microgrid is engaged if needed. In this case, the microgrid is run at the specific periods then turns off.
7 An Economic Demand Management Strategy for Passive Consumers. . .
185
7.3 Classification of Microgrid Applications In the literature, conventional micro-generators, renewable energy resources, and energy storage systems are often described as distributed energy resources (DERs) that are on-site generation sources in the distribution system [27]. Generally, a microgrid is well-known for the integration of DERs into the power grid, as well as its ability to operate in an islanded mode during certain cases. Therefore, the microgrid is described as a low-voltage distribution network of interconnected DERs, controllable loads, and critical loads that can operate in either grid-connected or islanded mode [28]. Besides, the microgrid may be used as a remote/off-grid case that is not considered in this article. In general, the microgrid may comprise several distributed generation systems, renewable (such as wind power, photovoltaic, hydro, and fuel-cell devices) or conventional generation (such as micro-turbines, diesel generators, and internal combustion engines) and a cluster of loads [29]. In this way, the customers can utilize from on-site generators, with the intent of adding additional resources over time, such as energy storage or other renewable sources [30]. However, due to the specific conditions of use of renewable technologies (availability of wind or solar radiation in the time of load demand), in this study, it is assumed that the conventional generation is used only as a microgrid. Therefore, customers can deploy this solution to manage their electric load consumption. Whereas DR is primarily focused on loads of consumer side, features analysis of various loads in the presence of a microgrid can facilitate to operate microgrid integrated with DR. The consumption pattern of loads may be indicated as urban, semi-urban, and rural or island. However, the practical application of microgrid for various loads can be mostly classified into eight sectors: industrial, military, campus/institutional, commercial, healthcare, residential, remote or rural, and others (such as data centers and cell phone towers). These applications are explained from the viewpoint of DR as follows.
7.3.1 Industrial Sector Industrial facilities, which are increasingly being established in remote locations, may not have continuous access to the main electricity grid. Therefore, this sector may be dependent on fuel and tend to use microgrid. Suppose an industrial consumer can manage electrical demand by utilizing distributed generation, energy storage, and load shifting [31]. Furthermore, the motivation of microgrid operation is the increased security and reliability needs in a grid-connected industrial site. On the other hand, the industrial sector is suitable for developing DR programs; however, adopting DR programs may be challenging for industrial firms. For instance, temporarily interrupting one or more processes may result in significant
186
M. E. Honarmand et al.
load reductions [32]. On this base, certain industrial sectors are only suitable for load management due to their technical restricts. By deploying microgrids in place, the industrial customer may solve this problem to integrate microgrid and DR strategies. Therefore, to provide flexible load solutions, a smart microgrid scheme can establish continuity of the industrial performance by implementing DR strategies.
7.3.2 Military Sector Military sector bases require reliable and resilient power to accommodate a variety of missions. Microgrids are easy to communicate on a community level but have more specific benefits when installed in military applications. Indeed, the military sector can enhance the security of critical electrical loads against the threat of grid outages by microgrids and this can be useful for DR schemes implementation. In order to develop a better DR management system, smart technologies can be used to communicate critical loads performance in real time. This can help make decisions and participate in DR programs, using microgrid without military sector interruption. Besides, cost-effective energy security is a driver to use the other military microgrids such as renewable resources [33]. It should be noted if these technologies are implemented in a secure procedure and well protected from cyber threats, it can be an opportunity for the military sector [34].
7.3.3 Campus/Institutional Sector So far, deploying on-site generation on a campus with multiple loads has been a successful procedure. Typically, the operation of microgrids with capacities ranging from 4 to over 40 MW has been common in this sector [35]. Furthermore, various abilities and numerous advantages to the management of net load shape during grid needs driven development of smart microgrids in the institutional sector. Generally, campus systems may include university and government campuses, and corporate parks. These are geographically large systems covering many buildings (residential, commercial, and/or industrial) but within a single ownership boundary that does not cross public rights of way [36]. Therefore, the potential for automated demand response is one of the key benefits of this sector along with the use of microgrids.
7.3.4 Commercial Sector Nowadays, microgrids have gained popularity to provide for economic requirements and demand management of commercial building installations, thereby supplementing the conventional grid. Practical applications of microgrids in smart commercial
7 An Economic Demand Management Strategy for Passive Consumers. . .
187
buildings are to increase renewable generation contribution and provide a high level of reliability and resiliency in response to grid outages [37]. One of the most direct ways a commercial microgrid can be used to cut costs is as a means to hedge power prices so that the system controls can be programmed to optimize for price. The usage peaks in commercial buildings typically align with an electric utility’s overall demand peaks. This means that utilities are particularly incentivized to participate with commercial customers to reduce or shift load through DR. Accordingly, a microgrid could use utility power until prices rise and then switch to its own, lower cost power by participating in DR Schemes [38].
7.3.5 Healthcare Sector A healthcare facility or medical center requires reliable electricity, heating, and cooling for running high-tech equipment and keeping patients healthy and comfortable. The microgrids system provides significant economic and environmental benefits to the most advanced healthcare sectors, ensuring the medical center’s sustainability and reducing its carbon footprint [39]. These sectors must care for patients 24/7, which creates greater demand for lights, heat, and cooling so that their consumption is much more than a commercial building of the same size. Due to the special energy requirements of this sector, huge opportunities can be offered in healthcare facilities by new technologies and procedures. These may be constituted to adopt new enablers and install advanced systems such as smart microgrids, which can empower them to participate in DR programs.
7.3.6 Residential Sector The main challenges in this sector are in making decisions for integrating individual home residential customers into large microgrids, and the deployment of microgrid technology at the level of individual homes. In the first case, it is possible to serve anywhere from a few up to thousands of customers and to support the penetration of local energy sources (electricity, heating, and cooling). In this situation, some houses may have some renewable sources that can supply their demand as well as that of their neighbors within the same community. In addition, this microgrid may have centralized or several distributed energy storage [40]. However, a decentralized building-integrated microgrid approach has the advantages of control over energy resources by customers. Besides, by adding microgrid capabilities, any changes performed behind the utility meter will likely not introduce significant legal or regulatory complications beyond because individual homes are already connected to the electrical distribution network [41]. Accordingly, using a variety of DR strategies, an appropriate framework for the management of energy can be developed for a residential microgrid [42].
188
M. E. Honarmand et al.
7.3.7 Remote or Rural Microgrids These microgrids never connect to the utility grid and instead operate in an island mode at all times because of economic issues or geographical position. To incorporate renewable energy sources as an add-on to diesel generator-based systems that so-called hybrid microgrids, provide great potential to diversify generation and lower microgrid operating costs in rural areas [43]. The careful resource assessment and understanding of demand profiles based on local conditions should be employed for the selection of remote microgrids. On this basis, the effective and economic operation of a microgrid is vital for sustained development; therefore, some DR strategies may be used as an appropriate method to operate rural microgrids [44].
7.3.8 Other Microgrids In addition to the above sectors, microgrids can also be used in other cases. For example, today’s data centers are trying to make their operations more resilient and efficient, and this is creating the perfect environment for new technologies like advanced microgrids to flourish. Because of data center investments in backup power equipment such as battery storage, a participant between a data center and its local microgrid may improve the situation for all parties. In this way, the DR schemes can be employed by interacting with data center facilities and microgrids [45]. Furthermore, the electric vehicle (EV) is to be viewed as a distributed resource and is becoming an enabling technology for microgrids. The integration EVbased microgrid, and operation planning strategies can be created under different vehicle behaviors with the minimum total cost goal [46]. Accordingly, the EV-based microgrid, as well as renewable resources, can present new opportunities for DR strategies so that can be employed to store energy when electricity consumption is low and discharge it in times of peak demands [47]. Microgrids connected to cell phone towers could help nearby communities gain access to electricity. In this way, energy service utilities can provide cell phone tower owners and operators with electricity at a competitive price while also providing electricity to nearby communities, with everyone being connected via the microgrid. These towers typically rely on expensive diesel generators, but now, renewable microgrids can offer the less expensive electricity prices, reliable and clean alternative [48].
7 An Economic Demand Management Strategy for Passive Consumers. . .
189
7.4 The Decision Procedure for Operating of a Microgrid Integrated with Demand Response In this section, details of the proposed decision procedure are provided to select DR schemes. For this purpose, the main concepts involved in this procedure are initially discussed. Therefore, this section begins with the cost model of investment and operation corresponding to the typical microgrid. Then, the impact of DR programs on microgrid costs is investigated by modeling the cost of these programs. Finally, the detailed descriptions about the decision procedure are presented.
7.4.1 Microgrid Cost Modeling This section presents the costs related to installation, maintenance, operation, and start-up of a microgrid. In order to compare these to the annual DR costs, microgrid costs are expressed annually.
7.4.1.1
Microgrid Installation Cost
The cost of microgrid installation is included in the purchased cost of distributed generation (DG) with the specified capacity and distribution infrastructure costs. The first cost element can be formulated as the following equation [49]: CDG−I = CIDG × P DG
(7.4)
where CDG-I is the total installation cost of the microgrid. Also CIDG and PDG are installation cost of DG ($/kW) and capacity of DG unit (kW), respectively. Moreover, the distribution infrastructure costs are comprised of the network costs and likely transformer cost. These can be calculated as follows: CN T −I = CIN T × LN T
(7.5)
CT R−I = CIT R × P T R
(7.6)
where CNT-I , CIN T , and LNT indicate the overall installation cost of a private network, the installation cost of a private network ($/m), and length of the network unit (m), respectively. Furthermore, CTR-I ,CIT R , and PTR show the total installation cost of the private transformer, the installation cost of a private transformer ($/kVA), and capacity of the transformer unit (kVA), respectively.
190
M. E. Honarmand et al.
These costs should be converted to the annualized cost (CIAnn ) for a payback period of n years and interest rate r, using the following equation: CIAnn =
7.4.1.2
r(r + 1)n × (CDG−I + CN T −I + CT R−I ) (r + 1)n − 1
(7.7)
Microgrid Maintenance Cost
This cost includes the annual mechanical and electrical reformation costs. Generally, this term is presented as a percentage of installation cost that can be calculated as [50]: CM = CI × ρ
(7.8)
where ρ is a constant value in terms of percentage and CM is the maintenance cost per year.
7.4.1.3
Microgrid Operation Cost
The cost of microgrid operation can consist of fuel cost, workforce, and so on. This cost depends on the duration of microgrid operation; therefore, this equation can be written as [49]: DG DG × CO × T DG /24 CO = PAv
(7.9)
DG is average generated power by DG. C DG and TDG are the operation cost where PAv O of DG source and duration of operating hours in a year, respectively.
7.4.1.4
Microgrid Start-Up Cost
Generally, this cost is considered only for fuel-consuming DG units. By definition, the start-up cost can be shown as a function of two parts, that is, the hot and cold start-up cost. This function is expressed as follows [51]: + , CS = CSH + CSC 1 − e(Toff /TC ) × N
(7.10)
where CSH and CSC are hot and cold start-up costs ($), respectively. In addition, Toff is the shutdown time of the DG unit, TC is the time constant of DG cooling and N indicates the number of start-ups.
7 An Economic Demand Management Strategy for Passive Consumers. . .
191
Fig. 7.1 The typical curves of load and TOU price
Finally, the overall cost of microgrid deployment can be calculated as: i = CIAnn + CM + CO + CS CMG
(7.11)
7.4.2 DR Cost Modeling The costs of participating in DR schemes are considered as DR costs. In practice, the customers will participate in these programs according to the comparison of the cost of microgrid utilization and DR strategies. As mentioned previously, these programs are categorized into PDR and IDR. In order to make proper decisions about the use of microgrid or the participant in DR schemes, the cost model should be developed for each scheme.
7.4.2.1
PDR-Based Cost
In PDR programs, electricity tariffs are defined based on different hours of consumption in various sectors. The typical curves of power load and TOU prices in each period are shown in Fig. 7.1. The load curve may be divided into three parts: peak, mid-peak, and off-peak. On this base, TOU prices are selected as a fixed tariff for each part. By predicting the consumption of each part, the participant can calculate the related cost using different tariffs as follows: CTi OU =
3
pr T OU −m × WT OU −m
(7.12)
m=1
where WTOU − 1 , WTOU − 2 and WTOU − 3 are the forecasted consumptions in total periods of the peak, mid-peak, and off-peak, respectively. Also, prTOU − 1 , prTOU − 2 , and prTOU − 3 are the tariffs corresponding in the TOU scheme.
192
M. E. Honarmand et al.
Fig. 7.2 The typical curves of load and CPP price
CPP pricing is similar to the TOU strategy. In this scheme, the price level of different periods is slightly lower than the TOU scheme. However, if an event happens, the price is suddenly raised. Therefore, in this case, two values of electricity consumption, that is, the normal and the event period, should be estimated to calculate the total cost. A typical CPP scheme is illustrated in Fig. 7.2. As mentioned earlier, if participants in this scheme fail to abide by their mandate, they have to pay the penalty. Assuming all obligations are met, the estimated cost of participating in the scheme is calculated as follows: i CCP P =
3
pr CP P −m × WCP P −m +
m=1
3
’ ’ pr CP P −m × WCP P −m
(7.13)
m=1
where prCPP − m and pr ’CP P −m are related to the normal and event tariffs in each ’ period, respectively. Also WCPP − m and WCP P −m indicate the estimated consumptions in the normal and the event periods, respectively. Since RTP pricing expresses better flexibility than TOU and CPP schemes, this tariff is indicated usually based on the consumption of the electric load. In this study, the common RTP is described in relation to the load of microgrid, given as [13]: 2 pr RT P −k = ak WRT P −k + bk WRT P −k + ck
(7.14)
where prRTP − k and WRTP − k indicate the RTP price and the general consumption of microgrid at time step k, respectively. Besides, the different values for ak , bk , and ck can be selected based on the actual demand at various time steps. Therefore, the cost of this scheme can be calculated to estimate the consumption at K periods as follows: i CRT P =
K k=1
pr RT P −k × WRT P −k
(7.15)
7 An Economic Demand Management Strategy for Passive Consumers. . .
7.4.2.2
193
IDR-Based Cost
Despite the popularity of these programs, the determination of the incentive amount is often arbitrary. Generally, these schemes are bilateral contracts in which the participants can attend at certain periods and specific tariffs. Depending on the type of scheme, there may be occasional penalties if the program does not run. In fact, if the IDR contract seems profitable, the customers will be eager to participate in the scheme. As a result, the costs of these programs are seen from the viewpoint of electricity utility. In order to increase the motivation to participate in the scheme, the related incentives are often constituted from two terms: readiness charge and participation incentive. The readiness charge is related to the fee of capability for the predefined demand interrupt or curtailment that is specified in the contract. But after participating in the scheme, the incentive quantity is calculated based on the content of the reduction in demand. Considering these points, by assuming the participation in the total contract, the cost function is expressed as follows: CI DRk =
N
PI DRk −i × TI DRk −i pr I DRk + PI’DRk × pr I’ DRk
(7.16)
i=1
where CI DRk can be related to kth scheme, that is, DLC, ILP, and EDR. Also PI DRk −i and TI DRk −i are ith demand amount and ith period participating in kth scheme. Besides, pr I DRk indicates the price of participating in kth scheme per $/kWh. The second element is introduced for readiness charge. Therefore, P ’ is I DRk
associated with the agreed interrupted/curtailment power and pr I’ DRk is related to power price per $/kW, in kth scheme.
7.4.3 The Decision Algorithm In order to provide flexible load side services, there are various options, such as the use of microgrids and DR schemes, namely PDR and IDR. Because of the economic savings resulting from the use of these options, customers have to consider one of these choices. This is particularly visible and palpable to large consumers. Generally, demand side plans are implemented according to the overall policies of each distribution company. The PDR program, usually defined as a one- or several month (almost long-term) period, seeks to modify or reduce end user consumption at different times. By saving and managing consumption at different time slots, consumers can control their costs at PDR prices. However, in practice, an electricity consumer may not be willing to cooperate in reducing or disrupting electricity due to the sensitivity of production lines or activities, specific requirements for a high level of service, or the effort required on the consumer’s part to participate. Therefore, the
194
M. E. Honarmand et al.
consumer should either pay the costs in full by participating in the PDR program or cover its entire load through the microgrid. The IDR program, which is often offered to customers in critical condition of the network, results in short interruptions (1 to 4 hours). Contracts for these programs usually offer attractive tariffs to entice more consumer participation. Despite these attractions, in practice, these plans may also cause problems for some customers to reduce or disconnect, which may prevent them from fully cooperating. In these cases, too, the microgrid can be used as a backup to prevent customers from being interrupted at a given time. Therefore, comparing the rewards of this program with the cost of microgrids will be crucial for selecting the appropriate IDR program and the related interval. Thus, a combination of IDR and microgrid can be an economic option for consumers. Therefore, the integration of microgrid technology and demand-side management plans can increase customer satisfaction. Here, to plan the economic demanding strategy of passive load considering DR programs and operating in the form of microgrid, a procedure based on different cost evaluation is proposed. In fact, in this approach, the cost of DR plans is examined along with the costs of microgrid technology for a typical load. The flow chart of this procedure is illustrated in Fig. 7.3. As observed, the investigation of two DR schemes is carried out separately in the presence of microgrids. To this end, the PDR scheme is first evaluated alongside the microgrid. Given the cost of the types of PDR programs, the consumer can decide on whether or not to cooperate, in other words, to use microgrid or not. However, by participating in IDR programs, the consumer can utilize the microgrid to eliminate the problems of cutting or reducing the load due to participate in these programs. Regarding the proposed tariffs, this algorithm assumes that the cost of a DR scheme is comparable to the microgrid operation costs. As it is observed, this flowchart is divided into two steps, and each step consists of three levels (L-I, LII, and L-III), which are described as follows: Step I: Microgrid operation and PDR programs. At this step, the cost of PDR schemes is compared to the costs of microgrids. For this purpose, first, one of the plans is selected, and the cost of one-day collaboration in that PDR scheme is calculated. Then, the cost of one-day performance in the form of a microgrid is obtained (L-I). These two costs are compared and if the cost of cooperating in the PDR were less than the cost of the microgrid, the day’s counter would increase. This increase will continue to a threshold value, M1, defined by the operating company unless the cost of cooperating in the PDR exceeds the cost of the imaginary microgrid for a day. From that day on, consumer performance in the form of a microgrid would be economical. This is the point of decision and discrimination between the choice of transferring into the form of a microgrid or participating merely in DR programs (L-2). These calculations are performed for other PDR schemes in the same manner and compared with the cost of microgrids, and a discrimination point is specified for all schemes. Becoming a microgrid for the number of days beyond this point will be the economical choice of the passive load (L-3). It should be mentioned that there may be no discrimination point for any
7 An Economic Demand Management Strategy for Passive Consumers. . .
195
Start Select C1 , C2 = 0
Step I
Step II
Select a PDR scheme ii =1
Select a IDR scheme kk =1
Number of day i =1 Calculate
kk = kk +1
Calculate
i = i +1
i , i C MG –T CPDR ii
L-I
Number of day k =1
ii = ii +1
k = k +1
k , k C MG –T CIDR kk
Yes i i C MG –T < C PDR ii
No
Yes i i C MG –T < C PDR ii
i+ 1 ≤ M 1
No
k+ 1 ≤ M 2
No
No
Yes
ii + 1 ≤ 3 No
No
i C MG –T < C 1 ≠ 0
Yes
L-II
L-III
No Planning of MG operation for i days integrated with PDRii scheme
kk + 1 ≤ 3 No
k C MG –T < C 2 ≠ 0
Yes
No
Yes
i Select C 1 = C MG –T ; day = i ; PDR = ii
ii + 1 ≤ 3
Yes
Yes
Yes
k Select C 2 =C MG –T ; day = k ;IDR = kk
kk + 1 ≤ 3
Yes
No Planning of MG operation for k days integrated with IDRkk scheme
No use of MG
END
Fig. 7.3 The flowchart of the proposed framework to decide about economic demanding strategy for a passive consumer
of the PDR schemes, which means that becoming a microgrid versus PDR plans is not economical. Step II: Microgrid operation and IDR programs. At this stage, similar to the previous step, computation is executed at the first level, and at the second level, comparisons are completed, and finally, at the third level, the selection is made. The only difference is the nature of the IDR scheme. Thus, for each time of execution of a particular IDR program in a defined interval (one hour or more per interval), the value of the reward to the plan is calculated. Accordingly, the rewards for IDR should be greater than the cost of transferring
196
M. E. Honarmand et al.
the consumer into the microgrid at each stage to benefit customers. Therefore, the counter increases to the threshold value, M2, specified by the company, and the discrimination points are specified for all the schemes. It is ultimately up to the user to decide on the best strategy. Ultimately, the consumer makes the best choice by comparing costs.
7.5 Numerical Studies To assess the efficiency of the proposed procedure, three case studies of industrial, commercial, and hospital customers are evaluated while they can be operated as a microgrid and numerical results for DR strategies are analyzed thoroughly. Customer data from the Guilan electrical distribution company are used in this study. This company is responsible for providing distribution services for an approximate of 1.5 million customers to the south of the Caspian Sea. The load curves of selected customers are illustrated in Fig. 7.4. Also, the cost parameters of a typical microgrid are presented in Table 7.1. Furthermore, in all cases, it is assumed that seven events in peak, two events in mid-peak, and one event in off-peak have been considered in the CPP scheme. In addition, the penalties in the schemes are ignored. For each load, the microgrid is connected to the external grid with 5800, 450, and 1600 kVA capacity, respectively, which can be operated at 90% of its capacity due
Fig. 7.4 The typical load curves (a) industrial, (b) commercial, and (c) hospital
7 An Economic Demand Management Strategy for Passive Consumers. . . Table 7.1 Typical information of microgrid
Parameter Installation cost Operation cost Maintenance cost Startup cost Interest rate Planning period
197 Unit $/MW $/MWh $/MWh $ % Year
Value 320 29 7 0.15 12.5 15
Table 7.2 Tariffs of TOU and CPP schemes in case I Period Peak (8 h) Mid-peak (6 h) Off-peak (10 h)
Hours 12–17 20–24 9–12 17–20 00–9
TOU price ($/kWh) 0.14
CPP price ($/kWh) Nonevent day Event day 0.096 0.46
0.09
0.065
0.15
0.055
0.039
0.039
Table 7.3 Tariffs of IDR schemes in case I Scheme DLC ILP EDR
Hour 4h 2h 1h
For load reduction ($/kW) 1.52 – 4.56
For readiness ($/kW) – 3.04 –
For consumption reduction ($/kWh) – 0.09 –
to operational constraints. In this study, considering centralized-style consumption of case studies, the cost terms related to the private distribution network and transformer are ignored. First, the effect of running PDR schemes on microgrid cost is investigated. Then an economic assessment of IDR program implementation is provided. All the schemes compared with the state that the load can act as a microgrid. In the end, the best option of the DR program is selected to operate the microgrid.
7.5.1 Case I: Industrial Load The peak load curve of a typical industrial firm is illustrated in Fig. 7.4a. This large electrical load has various industrial production lines and a connected microgrid to medium voltage (MV) with 5.8 MVA capacity. The different tariffs of TOU and CPP schemes are shown in Table 7.2, whereas the proposed prices for IDR programs are introduced in Table 7.3. The RTP rates are calculated using (7.14) by constant coefficients 5 × 10−9 , 2 × 10−5 , and 0.01 for ak , bk , and ck , respectively.
198
M. E. Honarmand et al.
Fig. 7.5 The various costs for (a) PDR and (b) IDR, compared with microgrid cost in case I
It is assumed that the participation occurs during the summer season, and the overall electric load is charged by microgrid when the customer participates in the DR scheme. Accordingly, the costs of different DR schemes should be calculated and compared with the deploying microgrid cost. In order to estimate the industrial load consumption, historical data are used. In this way, firstly, the cost of various PDR programs for different days are calculated and depicted as Fig. 7.5a in comparison with microgrid utilization cost. As it is observed, if the TOU program is planned for less than 55 days, the participation in this program is economical compared to microgrid running; whereas, the use of microgrid can be more economical for more days. Besides, this number of days for the CPP program is equal to 75 days, while in RTP program is roughly the same as the TOU scheme. In this way, the similar results of the IDR program running are shown in Fig. 7.5b. Generally, if the DLC contract is conducted 36 times a year, this contract may be economical for this industrial consumer. Obviously, the contract is more economical with more times a year. Furthermore, this value is for the ILP scheme equal to 17 times yearly, whereas, EDR contract can be economical annually for more 11 times.
7.5.2 Case II: Commercial Load In this section, the impact of DR programs in the presence of microgrid for a hypermarket with the peak load curve shown in Fig. 7.4b is investigated. The microgrid is also connected to the MV grid at 450 kVA capacity. The tariffs of different PDR and IDR schemes are shown in Table 7.4 and Table 7.5, respectively. Besides, the constant coefficients 8 × 10−7 , 4 × 10−4 , and 0.05 for ak , bk , and ck , respectively, are used to calculate RTP rates. Cost calculations are considered by assuming the participation of total electric load in the summer and the use of microgrid. Similar to case I, the costs of DR program implementation are compared to microgrid cost. The results are shown in Fig. 7.6. As can be seen in Fig. 7.6a,
7 An Economic Demand Management Strategy for Passive Consumers. . .
199
Table 7.4 Tariffs of TOU and CPP schemes in case II Period Peak (7 h) Mid-peak (7 h) Off-peak (10 h)
Hours 17–24 9–17 00–9
TOU price ($/kWh) 0.3 0.22 0.13
CPP price ($/kWh) Nonevent day 0.2 0.15 0.09
Event day 0.94 0.7 0.09
Table 7.5 Tariffs of IDR schemes in case II Scheme DLC ILP EDR
Hour 4h 2h 1h
For load reduction ($/kW) 1.73 – 3.46
For readiness ($/kW) – 3.00 –
For consumption reduction ($/kWh) – 0.057 –
Fig. 7.6 The various costs for (a) PDR and (b) IDR, compared with microgrid cost in case II
the TOU, CPP, and RTP schemes are economical for 14, 21, and 11 days in a year, respectively, compared with microgrid cost. Indeed, the microgrid can be used instead of participating in the above schemes with longer intervals. Accordingly, the results of IDR programs running are shown in Fig. 7.6b. In this way, the contract of DLC, ILP, and EDR schemes are economical for more 29, 16, and 14 times per year, respectively. Therefore, for a commercial consumer is economical to use a microgrid if periods of IDR contract are greater than the mentioned values.
7.5.3 Case III: Hospital Load In this case, the deployment cost of the microgrid in a hospital integrated with DR programs that has the peak load curve in Fig. 7.4c, is assessed. The microgrid is connected to the MV grid at 1.6 MVA capacity. The TOU and CPP tariffs in
200
M. E. Honarmand et al.
Table 7.6 Tariffs of TOU and CPP schemes in case III Period Peak (8 h) Mid-peak (13 h) Off-peak (3 h)
Hours 12–17 20–23 7–12 17–20 4–7
TOU price ($/kWh) 0.14 0.09 23–4 0.055
CPP price ($/kWh) Nonevent day Event day 0.096 0.46 0.065
0.15
0.039
0.039
Table 7.7 Tariffs of IDR schemes in case III Scheme DLC ILP EDR
Hour 4h 2h 1h
For load reduction ($/kW) 2.6 – 5.2
For readiness ($/kW) – 3.04 –
For consumption reduction ($/kWh) – 0.0176 –
Fig. 7.7 The various costs for (a) PDR and (b) IDR, compared with microgrid cost in case III
Table 7.6 and the IDR tariffs in Table 7.7 are shown. The RTP rates are calculated by the constant coefficients 8 × 10−7 , 5 × 10−5 , and 0.022 for ak , bk , and ck , respectively. In the presence of a microgrid, if the overall electric load in the summer can contribute to DR schemes, different costs are calculated. The calculation results for PDR and IDR schemes are depicted in Fig. 7.7. The CPP program for less than 58 days is economical compared to microgrid utilization, whereas the TOU scheme is more suitable for less than 36 days. Similarly, for less than 32 days, the RTP program is an economic scheme. These results are shown in Fig. 7.7a. Obviously, the use of microgrid may be cost-effective if there are more days in the proposed schemes. Figure 7.7b illustrates the cost of IDR schemes compared to microgrid cost. Accordingly, the contract of DLC, ILP, and EDR schemes is economical for more 19, 16, and 9 times per year, respectively. As a result, in order to be economical for
7 An Economic Demand Management Strategy for Passive Consumers. . . Table 7.8 A summary of results to use different schemes integrated with microgrid
Load Industrial Commercial Hospital
PDR scheme (days) TOU CPP RTP 55 75 52 14 21 11 36 58 32
201 IDR scheme (hours) DLC ILP EDR 36 17 11 29 16 14 19 16 9
microgrid utilization, this consumer should conclude a contract with more than the mentioned values for each scheme
7.5.4 Final Deduction The results presented in the previous sections are analyzed here. Table 7.8 summarizes these results. By comparing the cost of microgrid and PDR schemes, it is economical to use these schemes for the maximum number of days specified in Table 7.8. Since PDR schemes should be deployed for the summer season (90 days), the operation of a microgrid for the commercial load is entirely economical instead of PDR tariffs. For hospital load, microgrid operation is more suitable compared to TOU and RTP schemes; however, with these tariffs, the least economic benefit of microgrid utilization is for the industrial load. An IDR scheme alongside with microgrid operation is economical when the number of times running the program in contract exceeds the mentioned values in Table 7.8. Although the microgrid is operated when a load is interrupted, an important limitation of these schemes is the execution number; the load may be disconnected or connected time after time which can cause problems for some customers. On the whole, the participation of all consumers in ILP scheme can be economical for values greater than those presented in the Table 7.8.
7.6 Conclusions In this chapter, an economic demand management strategy for a passive consumer considering demand-side management schemes and microgrid operation was proposed. The main categories of DR programs, including PDR and IDR schemes were examined. Microgrid utilization for different types of loads was investigated in detail. To this end, the cost models for microgrid and DR schemes were extended. Based on these models, a decision criterion for determining the best choice for supplying the consumer demand was developed. In this regard, the cost of microgrid utilization alongside DR schemes was analyzed and compared. Several practical examples were provided and the results of the proposed method in the real case studies were presented.
202
M. E. Honarmand et al.
References 1. Zia, M. F., Elbouchikhi, E., & Benbouzid, M. (2018). Microgrids energy management systems: A critical review on methods, solutions, and prospects. Applied Energy, 222, 1033–1055. 2. Hosseini Imani, M., Niknejad, P., & Barzegaran, M. R. (2018). The impact of customers’ participation level and various incentive values on implementing emergency demand response program in microgrid operation. Electrical Power and Energy Systems, 96, 114–125. 3. Shen, J., Jiang, C., Liu, Y., & Qian, J. (2016). A microgrid energy management system with demand response for providing grid peak shaving. Electric Power Components and Systems, 44(8), 843–852. 4. Aghajani, G. R., Shayanfar, H. A., & Shayeghi, H. (2017). Demand-side management in a smart micro-grid in the presence of renewable generation and demand response. Energy, 126, 622–637. 5. Kumar, D., Verma, Y. P., & Khanna, R. (2019). Demand response-based dynamic dispatch of microgrid system in hybrid electricity market. International Journal of Energy Sector Management, 13(2), 318–340. 6. Hosseini Imani, M., Jabbari Ghadi, M., Ghavidel, S., & Li, L. (2018). Demand response modeling in microgrid operation: A review and application for incentive-based and time-based programs. Renewable and Sustainable Energy Reviews, 94, 486–499. 7. Nguyen, A. D., Bui, V. H., Hussain, A., Nguyen, D. H., & Kim, H. M. (2018). Impact of demand response programs on optimal operation of multi-microgrid system. Energies, 11(6) 1452, 1–18. 8. Han, X., Zhang, H., Yu, X., & Wang, L. (2016). Economic evaluation of grid-connected microgrid system with photovoltaic and energy storage under different investment and financing models. Applied Energy, 184, 103–118. 9. Zakariazadeh, A., Jadid, S., & Siano, P. (2014). Smart microgrid energy and reserve scheduling with demand response using stochastic optimization. Electrical Power and Energy Systems, 63, 523–533. 10. Zunnurain, I., Maruf, M. N. I., Rahman, M. M., & Shafiullah, G. M. (2018). Implementation of advanced demand-side management for microgrid incorporating demand response and home energy management system. Infrustructures, 3(4), 1–25. 11. Jin, M., Feng, W., Marrnay, C., & Spanos, C. (2018). Microgrid to enable optimal distributed energy retail and end-user demand response. Applied Energy, 210, 1321–1335. 12. Ferruzzi, G., Graditi, G., Rossi, F., & Russo, A. (2015). Optimal operation of a residential microgrid: The role of demand-side management. Intelligent Industrial Systems, 1, 61–82. 13. Yang, X., Zhang, Y., He, H., & Weng, G. (2019). Real-time demand-side management for a microgrid considering uncertainties. IEEE Trans. on Smart Grid, 10(3), 3401–3414. 14. Department of Energy, Benefits of demand response in electricity markets and recommendations for achieving them, a report to the United States congress pursuant to section 1252 of the energy policy act of 2005, 2006. 15. Hussain, M., & Gao, Y. (2018). A review of demand response in an efficient smart grid environment. The Electricity Journal, 31(5), 55–63. 16. Siano, P. (2014). Demand response and smart grids- a survey. Renewable and Sustainable Energy Reviews, 30, 461–478. 17. Nicolson, M. L., Fell, J., & Huebner, M. (2018). Consumer demand for time of use electricity tariffs: A systematized review of the empirical evidence. Renewable and Sustainable Energy Reviews, 97, 276–289. 18. Hung, M. F., & Huang, T. H. (2015). Dynamic demand for residential electricity in Taiwan under seasonality and increasing-block pricing. Energy Economics, 48, 168–177. 19. Paterakis, N. G., Erdinc, O., & Catalao, J. P. S. (2017). An overview of demand response: Key-elements and international experience. Renewable and Sustainable Energy Reviews, 69, 871–891.
7 An Economic Demand Management Strategy for Passive Consumers. . .
203
20. Yang, C., Meng, C., & Zhaou, K. (2018). Residential electricity pricing in China: The context of price-based demand response. Renewable and Sustainable Energy Reviews, 81, part 2, 2870–2878. 21. Xu, F. Y., Zhang, T., Lai, L. L., & Zhou, H. (2015). Shifting boundary for price-based residential demand response and applications. Applied Energy, 146, 353–370. 22. Davito, B., Tai, H., & Uhlaner, R. (2010). The smart grid and the promise of demand-side management. McKinsey on Smart Grid Summer, 2010, 38–44. 23. Shen, B., Ghatikar, G., Lei, Z., Li, J., Wikler, G., & Martin, P. (2014). The role of regulatory reforms, market changes, and technology development to make demand response a viable resource in meeting energy challenges. Applied Energy, 130, 814–823. 24. Vardakas, J. S., Zorba, N., & Verikoukis, C. V. (2015). A survey on demand response programs in smart grids: Pricing methods and optimization algorithms. IEEE Communications Surveys and Tutorials, 17(1), 152–178. 25. Aghaei, J., Alizadeh, M.-I., Siano, P., & Heidari, A. (2016). Contribution of emergency demand response programs in power system reliability. Energy, 103, 688–696. 26. Parhizi, S., Lotfi, H., Khodaei, A., & Bahramirad, S. (2015). State of the art in research on microgrids: A review. IEEE Access, 3, 890–925. 27. Hatziargyrion, N., Asano, H., Iravani, R., & Marnay, C. (2007). Microgrids: An overview of ongoing research, development, and demonstration projects. IEEE Power Energy Magazine, 5(4), 78–94. 28. Lasseter, R. H., & Paigi, P. (2004). Microgrid: A conceptual solution. IEEE 35th Annual Power Electronics Specialists Conference, 4258–4290. 29. Microgrid Knowledge, Microgrids and distributed energy: How customers can get full value, 2019. Available at: https://microgridknowledge.com/asset-backed-demand-responsemicrogrids/. 30. Shoreh, M. H., Siano, P., Shafie-khah, M., Loia, V., & Catalao, J. P. S. (2016). A survey of industrial applications of demand response. Electric Power Systems Research, 141, 31–49. 31. Mohagheghi, S., & Raji, S. (2014). Managing industrial energy intelligently: Demand response scheme. IEEE Industrial Applications Magazine, 20(2), 53–62. 32. Smith, D. M., St. Leger, A., & Severson, B. (2015). Automated demand response of thermal load with a photovoltaic source for military microgrids. IEEE Clemson University Power Systems Conference, 1–8. 33. Van Broekhoven, S., Judson, N., Galvin, J., & Marqusee, J. (2013). Leading the charge: Microgrids for domestic military installations. IEEE Power and Energy Magazine, 11(4), 40– 45. 34. Hirsch, A., Parag, Y., & Guerrero, J. (2018). Microgrids: A review of technologies, key drivers, and outstanding issues. Renewable and Sustainable Energy Reviews, 90, 402–411. 35. Hanna, R., Ghonima, M., Kleissl, J., Tynan, G., & Victor, D. G. (2017). Evaluating business models for microgrids: Interactions of technology and policy. Energy Policy, 103, 47–61. 36. Meegahapola, L. G., Robinson, D., Agalgaonkar, A. P., Perera, S., & Ciufo, P. (2014). Microgrids of commercial buildings: Strategies to manage mode transfer from grid connected to islanded mode. IEEE Trans. on Sustainable Energy, 5(4), 1337–1347. 37. Wang, Y., Wang, B., Chu, C. C., Pota, H., & Gadh, R. (2016). Energy management for a commercial building microgrid with stationary and mobile battery storage. Energy and Buildings, 116, 141–150. 38. Cotto, S. V., & Lee, W. J. (2016). Challenges and opportunities: Microgrid modular design for tribal healthcare facilities. IEEE North American Power Symposium, 1–6. 39. Dincer, I., & Abu-Rayash, A. (2019). Energy sustainability. Elsevier Science & Technology. 40. Sechilariu, M., Wang, B., & Locment, F. (2013). Building-integrated microgrid: Advanced local energy management for forthcoming smart power grid communication. Energy and Buildings, 59, 236–243. 41. Jamshidi Monfared, H., Ghasemi, A., Loni, A., & Marzband, M. (2019). A hybrid price-based demand response program for the residential micro-grid. Energy, 185, 274–285.
204
M. E. Honarmand et al.
42. Kuang, Y., Zhang, Y., Zhou, B., Li, C., Cao, Y., Li, Y., & Zeng, L. (2016). A review of renewable energy utilization in islands. Renewable and Sustainable Energy Reviews, 59, 504– 513. 43. Anand, H., & Ramasubbu, R. (2018). A real time pricing strategy for remote micro-grid with economic emission dispatch and stochastic renewable energy sources. Renewable Energy, 127, 779–789. 44. Thompson, C. C., Oikonomou, P. E. K., Etemadi, A. H., & Sorger, V. J. (2015). Optimization of data center battery storage investments for microgrid cost saving, emissions reduction, and reliability enhancement. IEEE Industry Applications Society Annual Meeting, 1–7. 45. Aluisio, B., Dicorato, M., Ferrini, I., Forte, G., Sbrizzai, R., & Trovato, M. (2019). Optimal sizing procedure for electric vehicle supply infrastructure based on DC microgrid with station commitment. Energies, 12(10), 1–19. 46. Tushar, M. H. K., Zeineddine, A. W., & Assi, C. (2018). Demand-side management by regulating charging and discharging of the EV, ESS, and utilizing renewable energy. IEEE Trans. on Industrial Informatics, 14(1), 117–126. 47. Aliasghari, P., Mohammadi-Ivatloo, B., Alipour, M., Abapour, M., & Zare, K. (2018). Optimal scheduling of plug-in electric vehicles and renewable micro-grid in energy and reserve markets considering demand response program. Journal of Cleaner Production, 186, 293–303. 48. Kansal, S., Tyagi, B., & Kumar, V. (2017). Cost-benefit analysis for optimal DG placement in distribution system. International Journal of Ambient Energy, 38(1), 45–54. 49. Chen, C., & Duan, S. (2014). Optimal allocation of distributed generation and energy storage system in microgrids. IET Renewable Power Generation, 8(6), 581–589. 50. An, L. N., Quoc-Tuan, T., Seddik, B., & Van-Linh, N. (2015). Optimal sizing of a gridconnected microgrid, IEEE International Conference on Industrial Technology (ICIT), 1–6.
Chapter 8
Real-Time Perspective in Distributed Robust Operation of Networked Microgrids Mehdi Jalali, Manijeh Alipour, and Kazem Zare
8.1 Introduction By extending distributed energy resources, the structure of traditional distribution networks is moving toward distribution systems. Furthermore, some consumers are equipping to the local generation units and controlling elements to obtain more reliable electricity and more economic benefits by constructing microgrids (MGs). According to CIGRE WG C6.21, microgrids are defined as “small controllable energy systems including loads and energy resources (such as micro-turbines, storage devices, or controllable loads) that can be operated in a coordinated way either while connected to the main network or while islanded” [1]. In the future distribution systems, there will be multiple distinct microgrids that each of them is responsible to provide its own consumption and can operate in grid connected and islanded modes based on the owner’s strategies. At the same time, microgrids will consider participating in day-ahead (DA) and real-time (RT) markets besides managing their own distributed energy resources (DERs). Therefore, designing a trading mechanism among networked microgrids is an urgent task in which energy exchange between MGs will be done by the agreement of both the seller and buyer. In this respect, real-time scheduling based on real-time forecasts tackle balance between supply and demand at each entity. Some operation management models for distribution systems with individual microgrids are presented in literature which can be distinguished in the terms of dependency in the objective function and ownerships and market clearing entity. In [2], based on the transactive energy framework, an energy management model for DERs in the form of a virtual power plant is presented. In [3], the
M. Jalali () · M. Alipour · K. Zare Faculty of Electrical and Computer Engineering, University of Tabriz, Tabriz, Iran e-mail: [email protected]; [email protected]; [email protected] © Springer Nature Switzerland AG 2021 A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems, https://doi.org/10.1007/978-3-030-59750-4_8
205
206
M. Jalali et al.
distribution system operator (DSO) determines distributed locational marginal price in the distribution network, then calculates transactive energy incentive signals and shares with participants. A distributed approach for energy sharing between several residential microgrids is presented in [4]. The optimal energy sharing problem in [6] to tackle the less computational burden and sharing private information is decomposed into one master problem and two subproblems. The decision variables of subproblems are sent to the master problem which controls the energy interaction between microgrids. In [5], a multi-follower bi-level programming approach is presented which by using KKT conditions from the DSO’s point of view could not be implemented for the case of independent microgrids [5]. In [6], the alternating direction method of multiplier (ADMM) is implemented to relax the complicating power balance between the distribution system operator and microgrids. In fact, the result of this approach is corresponding to the optimal sum of DSO and MGs’ costs. Obviously, the results of [5] would not lead to the Nash equilibrium point. The main idea lied in this chapter is presenting a framework for networked MGs that have different owners. Compared with the discussed methods, we designed a new third-part economic entity (TPEE) that it is not on the domination of any operator (DSO or MGs) and does not allow the utility operators to access the other players’ information and their economic values. Microgrids’ bids and offers, pairs of price, and corresponding quantity that are named as economic values, are submitted to the TPEE. Then, TPEE clears the bids and offers by determining feasible transactions and sellers and buyers. TPEE determines the feasible transactions’ social welfare values, provide integrated buying and selling bid and offers. At the next stage, the physical power between sellers and buyers corresponding to the approved transactions and satisfaction would be exchanged by satisfying the equality of technical parameters. To have optimal self-scheduling within the MGs, ADMM as a distributed optimization technique is implemented. In other words, distributed optimization provides a framework for the satisfaction of voltage magnitude and voltage angle equality at the connection nodes. The presented mechanism feeds by updated forecasts and prepares real-time energy management schemes through rolling horizons.
8.2 Methodology Description In this section, the functions of the proposed TPEE and energy management within the MGs are described in detail. To satisfy the secure information sharing and equality in terms of voltage magnitudes and angles, the trading mechanism and physical interactions have been distinguished. TPEE is considered as an intermediary in exchanging economic values (selling offers and buying bids) between individual microgrid owners. This mechanism does not allow any operator to have knowledge of the behavior of other ones and cannot impose its volition on other participants. TPEE receives economic values from the
8 Real-Time Perspective in Distributed Robust Operation of Networked Microgrids Fig. 8.1 Information sharing in the presented framework
207
TPEE
MG1
MGn
MG2
Power delivering signal
MGi
Economic value signals
MGs and clears the offers and bids. The economic values can be determined by using different methods. We supposed that MGs provide economic values (i.e., bid and offer curves) respect to their neighborhood entities and submit to the TPEE. TPEE determines sets of sellers and buyers based on submitted economic values. It should be noted that TPEE would not consider the MGs’ technical constraints. The summary of proposed energy trading is presented in Algorithm 8.1.
Algorithm 8.1 The proposed transactive energy trading 1. Each entity submits economic (selling offers and buying bids) to TPEE. 2. TPEE selects the transaction corresponds which are profitable for both of the participants. 3. Each entity sends its own final satisfactory by a signal to TPEE. 4. TPEE finalizes all deliverable and approved transactions. 5. MGs schedule their own resources considering approved transactions via ADMM.
The proposed energy trading microgrids and utility is illustrated in Fig. 8.1. According to Fig. 8.1, participants share their economic values with TPEE and after local market clearing by TPEE, participants share power delivering signal (dual variable of complicated constraints) with each other. The power exchange and self-scheduling of resources within the microgrids and distribution network are tackled using distributed optimization and is described in the next section.
208
M. Jalali et al.
8.3 Problem Formulation In this section, the physical energy exchange between entities and self-scheduling of resources with the individual entities have been addressed.
8.3.1 Microgrids’ Real-Time Energy Management According to Fig. 8.1, the interconnected MGs can participate in the main electricity DA and RT markets. Based on the DA market’s gate at a specific time prior to the trading day will be closed and after a short delay, the DA market’s price will be published. The RT market opens at each time slot of the trading day and closes specific minutes before the start of the trading hour and results will be published prior to the trading start time. Since MGs do not have any information about prices in the DA and RT markets before the markets’ clearing times, it needs to have appropriate forecasts as far as possible. After the DA market’s clearing time, MGs provides an optimal operating scheme considering updated forecasts. On the other hand, MGs by participating in the proposed energy framework can sell or buy power from each other. To have proper operation scheduling, the entities need to have an accurate prediction of their electricity consumption and price of DA and RT market. Based on rolling horizon load and electricity markets’ prices, the objective function from the MG i’s point of views can be formulated as: ⎡ ⎢ DA,i DA,i RT ,i RT ,i MT MT MT Min + ai,g Pi,g,t|k + bi,g ⎣ λˆ t|k Pt|k + λˆ t|k Pt|k
ΨiMG k=t:T
g∈ΩiMT MT +SU C MT i,g,t|k + SDC i,g,t|k
, (8.1)
The first term in (8.1) denotes the cost of purchased power from day-ahead ,i and real-time markets. Price uncertain parameters (λˆ DA,i andλˆ RT t|k t|k ) are forecasted day-ahead and real-time markets’ prices at time slot t for the k-hour ahead. The second term of (8.1) is used to model operation, start-up, and shut-down costs of micro-turbines within the MG i. Since TPEE clears trades before real-time, the transactions’ cost of the next time horizon for the player are predetermined. Therefore, these terms are not included in the objective function. The operating constraint and load follow equations in the microgrid can be expressed as: MT MT MT δi,g,t|k SU C MT − δi,g,t|k−1 i,g,t|k ≥ SU i,g
(8.2)
8 Real-Time Perspective in Distributed Robust Operation of Networked Microgrids
I nj
Pi,n,t|k =
209
MT MT MT δ SDC MT ≥ SD − δ i,g,t|k i,g i,g,t|k i,g,t|k−1
(8.3)
on MT MT δ Ti,g,t|k ≥ U T MT − δ i,g i,g,t|k i,g,t|k−1
(8.4)
off MT MT δ Ti,g,t|k ≥ DT MT − δ i,g i,g,t|k i,g,t|k−1
(8.5)
MT Min MT MT Max δi,g,t|k × Pi,g ≤ Pi,g,t|k ≤ δi,g,t|k × Pi,g
(8.6)
MT MT MT Max × QMin δi,g,t|k i,g ≤ Qi,g,t|k ≤ δi,g,t|k × Qi,g
(8.7)
DA,i RT ,i Pt|k + Pt|k +
s∈ΩnU,i
MT d Pi,g,t|k − Pˆi,n,t|k −
g∈ΩnMT ,i
P Linm,t|k
m∈ΩnBus,i
(8.8)
I nj
Qi,n,t|k =
QG t|k +
s∈ΩnU,i
I nj n −1 + Pi,n,t|k = Ginn 2Vi,t|k
g∈ΩnMT ,i
ˆd QMT i,g,t|k − Qi,n,t|k −
QLinm,t|k
(8.9)
m∈ΩnBus,i
i n − θm i n m Bnm θi,t|k i,t|k + Gnm Vi,t|k + Vi,t|k − 1
(8.10)
n − θm i n m Ginm θi,t|k i,t|k − Bnm Vi,t|k + Vi,t|k − 1
(8.11)
m∈ΩnBus,i
I nj i n Qi,n,t|k = Bnn 1 − 2Vi,t|k +
m∈ΩnBus,i
n m i n m + Bnm θi,t|k P Linm,t|k = Ginm Vi,t|k − Vi,t|k − θi,t|k
(8.12)
i n m n m Vi,t|k + Ginm θi,t|k QLinm,t|k = −Bnm − Vi,t|k − θi,t|k
(8.13)
210
M. Jalali et al. i SLinm,t|k ∼ .QLinm,t|k = P Linm,t|k + ξnm,t|k
(8.14)
≤ SLinm,t|k ≤ SLi,Max −SLi,Max nm nm
(8.15)
n ViMin ≤ Vi,t|k ≤ ViMax
(8.16)
The set of MG i’s independent decision-making variables is defined as DA,i RT ,i MT n n MT ΨiMG ={Pt|k ,Pt|k ,Pi,g,t|k ,QMT i,g,t|k ,δi,g,t|k ,Vi,t|k ,θi,t|k }. The start-up and shoutdown costs are formulated as (8.2) and (8.3), respectively. The micro-turbines’ minimum uptime and downtime constraints are subjected to (8.4) and (8.5), respectively. Constraints (8.6) and (8.7) satisfy the minimum and maximum active and reactive power generation of micro-turbines. The details of the implemented power flow method can be followed in [6]. According to the power flow method presented in [6], the active and reactive power balances can be considered in (8.8) and (8.9). The injected active and reactive powers into each node are equal to (8.10) and (8.11), respectively. The linearized form of active and reactive powers which flow in-line between node n and m are modeled according to (8.12) and (8.13), respectively. By implementing Taylor series, the apparent power flow through the line between node n and node m is approximated introducing an auxiliary i parameter (ξnm,t|k ), in (8.14), and is enforced to be limited according to the line capacity in (8.15). The maximum and minimum limits on voltage magnitudes are shown in (8.16). As addressed in microgrid planning related works, microgrids within the distribution systems can be operated in islanded and grid-connected modes. It should be noted that considered power flow eqs. (8.2)–(8.16) would be duplicated for each microgrid. The connection point of two entities corresponding to each transaction is imagined as a zero impedance line between MG i and MG j. By exchanging approved power between seller and buyer at nodes i and j, the constraints corresponding to the voltage magnitudes and voltage angles can be written as follows: CPj
CPi = Vj,t
(8.17)
CPj
CPi = θj,t
(8.18)
Vi,t
θi,t
The Eqs. (8.17) and (8.18) satisfy the voltage magnitude and voltage angle equality at connection points (CP) of microgrid i and j. It should be noted that the value of power exchange between entities is predetermined by TPEE.
8 Real-Time Perspective in Distributed Robust Operation of Networked Microgrids
211
8.3.2 Distributed Energy Management As mentioned in stage 5 of Algorithm 8.1, ADMM is implemented to distribute the operation scheduling and physical exchanging of the approved transactions. The general form of implemented ADMM for the relaxing complicated constraints (see. (8.17) and (8.18)) between individual MGs is presented [7]. Using ADMM, the modified objective functions of each MG by relaxing complicated constraints (8.17) and (8.18) at iteration κ + 1can be written as: ⎡ ⎤ # DA,i DA,i RT ,i RT ,i MT MT MT MT ⎣ λˆ Min + λˆ t|k Pt|k ai,g Pi,g,t|k + bi,g + +SDC i,g,t|k ⎦ + t|k Pt|k ΨiMG k=t:T g∈ΩiMT -2 # CPj CPj CPi κ CPi κ ρV V ,κ ˆ ˆ + + 2 -Vi,t − Vj,t λij,t Vi,t − Vj,t (8.19) 2 j ∈ΩiMG κ κ -2 # CPj ρθ -θ CPj − θˆ CPi ˆ CPi θ + − θ λθ,κ + i,t j,t i,t j,t ij,t 2 #
2
j ∈ΩiMG
In comparison with (8.1), the first and second terms added in (8.19) are augmented lagrangian terms based on ADMM. The objective function (8.19) subject to constraints (8.3)–(8.16) would be minimized by MG i at each time t, before real-time market gate closing time and after publishing TPEE. Parameter ρ V and ρ θ are positive values. The constraints (8.17) and (8.18) duality variables as power delivering signals between MG i and j will be updated at iteration κ + 1 at time slot t as: CP κ CPi κ V ,κ+1 (8.20) λij,t = λVij,t,κ + ρV Vˆi,t j − Vˆj,t
θ,κ+1 θ λij,t = λθ,κ ij,t + ρ
CP κ CPi κ θˆi,t j − θˆj,t
(8.21)
To have a robustness vision about the uncertain parameters, uncertain parameters are fixed at their worst scenarios.
8.3.3 Numerical Studies and Result Analysis A case study is conducted based on a test system with four microgrids which are shown in Fig. 8.2. Each MG consists of several micro-turbines with the location and operation characteristics including feasible maximum, minimum active, and reactive powers as well as coefficient of cost functions are presented in Table 8.1.
212
M. Jalali et al.
Fig. 8.2 Characteristics test networked MGs Table 8.1 Characteristics of the micro-turbines
Entity
Bus
Pmax [MW]
Qmax [MW]
MT [$/MW] ai,g
MG1
2 3 5 6 6 9 10 3 5 5 6
3 3 3 2 2 2 1.5 3 2.5 3 2.5
1.2 1.2 1.2 0.8 0.8 0.8 0.6 1.2 1 1.2 1
0.025 0.026 0.027 0.028 0.029 0.027 0.027 0.024 0.024 0.021 0.021
MG2
MG3 MG 4
The magnitudes of base apparent power and base voltage are considered to be 10 MVAr and 20 kV, respectively. The lines resistance and reactance are 0.05 p.u and 0.1 p.u, respectively. The implemented forecasting framework for the electricity demand, price is inspired by the rolling horizon technique [8]. To provide a simple and applicable forecasting method, auto-regressive with exogenous input variables (ARX) is used to train forecasting models. The most informative input features are selected based on the technique which is presented in [9]. The candidate inputs for the load forecasting consist of two sets. The first set includes the lagged value corresponding to 1, 24, 48, 72, 96, 120, 144, and 168 hours ago of the electricity load and the second set is corresponding to the exogenous variables which are minimum and maximum loads of the previous day and the same day in the previous week, and the hours corresponding to occurrences of peak load and ramp-up hours of the previous day
8 Real-Time Perspective in Distributed Robust Operation of Networked Microgrids Table 8.2 1-Hour ahead load forecasts in the distribution network and individual microgrids
Time 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
MG1 1158.50 1134.64 1184.24 1253.37 1348.92 1409.19 1495.77 1497.29 1436.91 1408.13 1335.63 1241.29 1198.63 1155.17 1144.44 1144.29 1167.25 1197.32 1219.82 1244.99 1231.34 1220.24 1179.76 1147.55
MG2 737.20 704.53 691.70 730.86 805.24 873.10 908.14 907.17 872.26 851.57 809.79 741.67 711.07 686.46 680.17 734.05 813.55 879.85 928.34 939.39 922.57 871.46 793.23 741.95
MG3 462.53 455.79 447.97 443.11 444.25 450.55 462.90 469.36 478.03 478.74 491.14 490.46 497.04 493.18 491.44 503.98 518.13 563.51 616.12 641.90 648.50 639.96 628.21 643.28
213 MG4 2640.74 2582.51 2564.61 2755.95 3034.38 3070.22 3153.46 3131.52 3073.04 2941.85 2814.18 2729.38 2664.55 2622.46 2574.77 2578.03 2624.61 2503.48 2619.79 2647.13 2625.74 2627.73 2525.90 2448.08
and the same day in the previous week. The 1-hour ahead electricity load forecasts of microgrids are reported in Table 8.2. According to the feature selection technique which is presented in [10], the considered features for the electricity price of dayahead market consist eight lagged prices (1, 2, 12, 23, 24, 48, 72, and 168 hour ago prices) plus four lagged loads (1, 12, 24, and 73 hour ago loads) as exogenous variables. The historic data for price forecasting is obtained from [11]. As described in Algorithm 1, the entities submit their selling and buying bids and offers to the IEE. IEE collects the offers and determines the transactions. IEE determines the seller and buyers as well as the price of each transaction. TPEE sorts buyers’ bids and sellers’ offers in the highest and lowest bids and offers, respectively. This mechanism has been illustrated in Fig. 8.3. As shown in Fig. 8.3a, at time slot t = 1, the bids of buyers are 0.0295, 0.029, 0.027, and 0.021 respectively belong to MG2, MG3, MG1, and MG4. The offered price for the selling are 0.0232 and 0.0335 belonging to MG4 and MG3. Therefore, TPEE based on the value of feasible transactions selects participants in an ordered pair form (SELLER, BUYER). The selected transactions for time slot t = 1 are: (MG4, MG1), (MG4, MG2), (MG4, MG3). For time slot t = 23, based on the suggested price for the
214
M. Jalali et al.
Fig. 8.3 Transaction clearing by TPEE at time slot (a) t = 1 and (b) t = 23
(a) - Price and participating in DA and RT markets
(b) - Price and energy exchange MG4 with other MGs
Fig. 8.4 MG4’s participating in electricity markets and interactions with other MGs. (a) Price and participating in DA and RT markets. (b) Price and energy exchange MG4 with other MGs
buying and selling as well as the quantities, TPEE accepts the transaction between TPEE and MG1. Since the proposed energy management is established for real-time applications, TPEE should clear transactions between MGs before the final scheduling of entities. Entities by receiving their own approved transactions for the next time slot, solve the operation problem. Due to handling the approved transactions, the alternating direction method of multipliers according to Algorithm 8.1 has been implemented in an iterative manner until achieving convergence criteria. MGs in the connection with upper grids can participate in day-ahead and real-time markets. According to results, MG4 provides the economic values corresponding to the energy trading with other MGs based on its own forecasts of day-ahead and real-time markets’ prices as well as the operation cost of its own micro-turbines. It should be mentioned that the real-time and day-ahead electricity markets’ prices are published after their closing times. Therefore, microgrids’ strategies are dependent on the price and load forecasts’ accuracy. In this case, study decisions are made based on rolling forecasting of load and prices. In Fig. 8.4a, the last updates forecasts of day-ahead (DA) and real-time (RT) markets’ prices and purchased powers by MG4 are illustrated.
8 Real-Time Perspective in Distributed Robust Operation of Networked Microgrids
215
As shown in Fig. 8.4a, at time slots between t = 1 until t = 9, the prices of RT or DA are less than the generation cost of micro-turbines (see. Table 8.1), MG4 purchase power from the upper grid by participating in DA and RT markets. At time slots t = 3 until t = 5, the price of DA was less than the predicted RT market’s price, therefore, MG4 purchases power from the DA. In real-time decision making fashion, microgrids similar to MG4 submit selling offers and buying bids to the TPEE. It should be noted that all the accepted transactions are double-checked by the corresponding sellers and buyers. The double-checking stage validates deliverable transactions and prevents technical challenges such as congestion, overvoltage, and stability problems in microgrids. For our test system, all the feasible transactions which are approved by TPEE and corresponding seller and buyer are shown in Fig.8.4b. At each time slot, the number of transactions and buyers’ bids are depicted in Fig. 8.4b. As can be concluded from Fig. 8.4b, buyers with maximum bids are selected by TPEE as approved buyers. Since MG4’s offered selling price is lower than other MGs, MG4 is selected as the seller over the operating horizon. At time slots between t = 1 and t = 9, the price of electricity markets is low. Therefore, MG4 benefits from arbitrage of buying power from the DA or RT and selling to the microgrids. MG4 sells to all the microgrids at time slots t 2 [1; 9]. Because MG4 by buying from markets at a low price, it obtains more capacity respect to time slots t = 2 [10; 24]. By comparing exchanged power at time slot t = 23 and t = 24, it can figure out that buyers are selected according to their bids. At time slot t = 23, MG4 sells to MG1 but at time slot t = 24, MG4 sells to MG3. The real-time scheduling of energy resources within the entities are presented in Fig. 8.5. Operation schedules of micro-turbines within the MG1, MG2, MG3, and MG4 are illustrated in Figs. 8.5a–d, respectively. Since the generation costs of MG1–MG3 is expensive than the purchasing cost from MG4, MG1–MG3 prefers to buy from MG4. On the other hand, MG4 by comparing the price of DA and RT markets with selling prices just sells the production of its own micro-turbines at time slots t ∈ [10, 24].
Fig. 8.5 Output power of micro-turbines (a) MG1, (b) MG 2, (c) MG 3, and (d) MG 4
216
M. Jalali et al.
8.4 Conclusion In this chapter, an energy trading framework for a networked microgrid framework is presented. In this framework, a third-part economic entity by collecting selling and buying economic values determines deliverable and feasible transactions and corresponding sellers and buyers. Furthermore, physical exchanging power between entities is tackled by the entities’ operators. This aim is achieved by implementing ADMM as a distributed optimization technique. Since the exchange power between entities is determined by TPEE, entities should satisfy the equality constraints of voltage magnitude and voltage angle at the connection points. Therefore, these complicated constraints are relaxed by implementing ADMM. Moreover, the performance of the presented method is studied on the real-time operation of a test system by using updated forecasts of electricity markets’ prices as well as electricity loads. The results show that microgrids select operating their resources when the buying price is high or when their suggested quantity is not available in the market. Moreover, MGs by using updated forecasts of day-ahead and real-time market, at low prices buys from electricity markets and sells to the microgrids in the market.
Nomenclature Sets MT MT i U,i n ,i MT n Bus,i n MG i
Set of micro-turbine Set of micro-turbines within the MG i Indicator of connection node n to the upper grid in MG i Set of connected micro-turbines into bus n within the MG i Sec of connected bussed to bus n within the MG i Set of connected MGs to MG i
Indexes t, k i, j n, m g κ
Time slots Microgrids Buses Micro-turbines Iteration
8 Real-Time Perspective in Distributed Robust Operation of Networked Microgrids
217
Parameters λˆ Pˆ d aMT , bMT SUMT , SDMT UTMT , DTMT PMin , PMax QMin , QMax G, B ξ SLMax VMax
Markets’ price Electricity load demand Micro-turbines cost coefficients Micro-turbines start-up and shut-down costs Micro-turbines minimum up and downtimes Micro-turbines minimum and maximum active power capacity Micro-turbines minimum and maximum reactive power capacity Conductance and susceptance components of the admittance matrix Auxiliary parameter in linearized aperient power flow Maximum aperient power flow Maximum voltage magnitude
Variables PDA ,PRT PMT ,QMT SUCMT ,SDCMT δ MT PInj ,QInj PL,QL,SL V,θ
Purchased power from DA and RT markets Active and reactive generated power of micro-turbine Realized Start-up and shut-down cost Commitment status of micro-turbine Injected active and reactive power Active, reactive, and aperient power flow in lines Magnitude and angle of voltage
References 1. Wang, Z., Chen, B., Wang, J., Begovic, M. M., & Chen, C. (2014). Coordinated energy management of networked microgrids in distribution systems. IEEE Transactions on Smart Grid, 6(1), 45–53. 2. Qiu, J., Meng, K., Zheng, Y., & Dong, Z. Y. (2017). Optimal scheduling of distributed energy resources as a virtual power plant in a transactive energy framework. IET Generation, Transmission & Distribution, 11(13), 3417–3427. 3. Sajjadi, Sayyid Mohssen, Paras Mandal, Tzu-Liang Bill Tseng, and Miguel Velez-Reyes. “Transactive energy market in distribution systems: A case study of energy trading between transactive nodes.” In 2016 North American Power Symposium (NAPS), pp. 1–6. IEEE, 2016. 4. Akter, Most N., Md Apel Mahmud, and Amanullah MT Oo. “An optimal distributed transactive energy sharing approach for residential microgrids.” In 2017 IEEE Power & Energy Society General Meeting, pp. 1–5. IEEE, 2017. 5. Jalali, M., Zare, K., & Seyedi, H. (2017). Strategic decision-making of distribution network operator with multi-microgrids considering demand response program. Energy, 141, 1059–1071. 6. Gao, H., Liu, J., Wang, L., & Wei, Z. (2017). Decentralized energy management for networked microgrids in future distribution systems. IEEE Transactions on Power Systems, 33(4), 3599– 3610. 7. Boyd, S., Parikh, N., Chu, E., Peleato, B., & Eckstein, J. (2011). Distributed optimization and statistical learning via the alternating direction method of multipliers. Foundations and Trends® in Machine learning, 3(1), 1–122.
218
M. Jalali et al.
8. Chitsaz, H., Zamani-Dehkordi, P., Zareipour, H., & Parikh, P. P. (2017). Electricity price forecasting for operational scheduling of behind-the-meter storage systems. IEEE Transactions on Smart Grid, 9(6), 6612–6622. 9. Chitsaz, H., Shaker, H., Zareipour, H., Wood, D., & Amjady, N. (2015). Short-term electricity load forecasting of buildings in microgrids. Energy and Buildings, 99, 50–60. 10. Amjady, N., & Keynia, F. (2008). Day-ahead price forecasting of electricity markets by mutual information technique and cascaded neuro-evolutionary algorithm. IEEE Transactions on Power Systems, 24(1), 306–318. 11. Historic price data,“ http://www.energyonline.com/Data/GenericData.aspx.
Chapter 9
Application of Heuristic Techniques and Evolutionary Algorithms in Microgrids Optimization Problems Amir Aminzadeh Ghavifekr
9.1 Introduction Microgrids can efficiently solve numerous problems of the traditional power systems and provide better management of peak loads, acceptable reliability, and feeder losses reduction [1]. Two primary goals of each commercial microgrid are customer satisfaction and cost-efficiency. Besides all environmental and economic benefits, there are numerous technical challenges in energy management, control, and protection of microgrids that cause designing a cost-effective microgrid system considered as a complicated problem. Every planning process suffers from social, regulatory, geographical, and environmental uncertainties that can be external such as lack of knowledge and the nature of the environment or be internal, which is mostly accrued in the process of identification. Thus, the planning process for microgrids is commonly based on the trade-off in solution searching and can be considered as a multi-objective problem. Despite all achievements of the microgrids, designing a cost-effective structure is a complicated problem due to the different parameters that should be taken into account at any decision level. Different optimization techniques have been presented in the literature to solve microgrids optimization problems. These methods can be compared based on criteria such as calculation complexity, ability to consider predictions, model dependency, and flexibility concerning microgrid expansion. According to the intrinsic features of the optimization problem, each of these algorithms can offer a successful solution. The classical optimization approaches, such as the numerical and analytical methods have been applied to several optimization problems of microgrids. These
A. A. Ghavifekr () Department of Control Engineering, University of Tabriz, Tabriz, Iran e-mail: [email protected] © Springer Nature Switzerland AG 2021 A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems, https://doi.org/10.1007/978-3-030-59750-4_9
219
220
A. A. Ghavifekr
methods provide desirable results when the accurate starting point is set in advance. They provide fast convergence to the best local optimum. However, it is very challenging to escape from the local optimum solution in these algorithms, and the selection of the appropriate starting point is not straightforward. The accurate estimation of this point needs intuition, experience, and lots of theoretical knowledge. Heuristic and metaheuristic algorithms that have been established since the 1980s are nature-inspired methods that include trial and error solutions to find strategies for complex problems. These algorithms refer to calculating the minimum and maximum of a function with systematic inputs chosen through an initial set. It cannot be proved that these methods always find the global optimum, but they can provide sufficiently good results near to the global minimum. These algorithms are best suited for problems with unknown starting points and large optimization space. Although these methods have slower convergence, they can search whole optimization space, which mostly cannot be reached by the classical algorithms. These data-driven methods can cover different challenges of this area, including the sizing and management optimization, predictive maintenance, estimation of exploitable energy, a real-time self-tuning system, siting, operation scheduling, and a variety of other applications. The extension of these results in home microgrids (HMGs), besides improvement in energy performance, can lead to the minimization of the product cost and market clearing price (MCP). Also, the optimization cost function can be formulated for system losses considering the balance of the load and generation. Numerous software tools have been developed for a predefined optimization problem both in real time and also offline. These software tools can model a wide range of nonlinear complicated optimization problems. Also, they can provide precise simulations due to the flexibility in the definition of time intervals. However, these tools still need to be updated for promoted usage in stability analysis and energy management problems. In [2], a detailed discussion and comparison of available software tools for analyzing various energy systems have been presented. Homer is one of these software tools, which was developed by NREL. It can model the hybrid systems and is designed for a large range of generators, loads, and converters. GAMS is another modeling software that can be utilized for a wide range of linear and nonlinear optimization problems. In HYBRID2, which was developed by the RERL, the simulation is very accurate due to the variable time intervals. One of the most downloaded optimization tools is RETSCREEN, which is based on MS Excel. It provides a platform to compare the base case models of classical methods with the financial and technical viability of renewable projects. This chapter will focus on recent progress in the application of computational intelligence and heuristic techniques in microgrids. However, regarding the probabilistic nature of these algorithms, they present variable performance in solving different optimization problems. This is the main motivation of presenting this chapter to give a general view of the recent achievements in solving microgrid optimization problems with evolutionary algorithms (EA). First, a brief review
9 Application of Heuristic Techniques and Evolutionary Algorithms. . .
221
of some recent and promising EAs is presented. In the rest of this chapter, the applications of these algorithms in energy management systems, operation scheduling, voltage and frequency control, and sizing optimization problems are discussed.
9.2 Brief Introduction to Evolutionary Algorithms There are many types of EAs that have been progressed based on the behavior of nature and can be either trajectory-based or population-based algorithms. This section gives an insight into the structure and basis of some EA algorithms which have been applied for microgrid problem optimizations. Some of the well-known EAs and their related techniques are introduced in this section. These methods try to explore the best available answer among a broad set of feasible solutions for objective functions with practical constraints. Genetic Algorithm (GA), Ant Colony Optimization (ACO), Particle Swarm Optimization (PSO), Biogeography Based Optimization (BBO), Harmony Search Algorithm (HSA), Cuckoo Search Algorithm (CSA), Artificial Bee Colonies (ABC), Grey Wolf Optimization (GWO), Firefly Algorithm (FA), and Grasshopper optimization algorithm (GOA) are among popular widely used heuristics methods that will be discussed in this chapter. These methods are employed to solve different optimization problems and energy planning issues in microgrids literature such as the optimal partition of power grids, replacement, maintenance, and monthly income by selling power to the grid. Using these methods leads to a reliable network partitioning with less CPU effort and save the operating costs of the distributed nodes and interruption costs. Also, they provide more mobility to add further restrictions to the issues mentioned earlier, such as sizing and scheduling of powergeneration sources.
9.2.1 Genetic Algorithm (GA) Genetic Algorithm (GA) is one of the well-known and effective techniques of the EAs and is inspired by Darwin’s theorem of survival of the best. The principles of the genetic and nature’s evolution structure have been utilized in this algorithm to eject the weak population and let only the high-quality ones survive the process. The initial population should be considered as chromosomes, and the operators of the algorithm (Crossover and Mutation) should be applied several times to meet one of the convergence criteria. GA can escape from the local minimum; however, its complexity increases with the number of parameters. Simple pseudo code for this algorithm is presented in Fig. 9.1 This algorithm can be used for nonlinear problems with numerous discrete variables such as microgrids reconfiguration [3], which satisfies the operational
222
A. A. Ghavifekr
Fig. 9.1 Pseudo code of GA
constraints and considers load priorities. Also, this algorithm has been proposed to obtain favorable solutions for operation scheduling and energy management problems [4]. The energy management system should solve the optimization problem with the classification of all the necessary information and resend it to each distributed energy resources (DERs). This concept can be extended to the optimal power flow (OPF) problems in a local microgrid setup [5]. Microgrids include several distributed generators (DGs), which are mostly utilized to supply the local loads. GA covers the sizing optimization problem of microgrids. This technique has been applied for hybrid AC–DC microgrids [6], battery energy storage system (BESS) optimal sizing [7], optimization of storage devices [8], combined heat and power (CHP) microgrid systems [9], determination of optimal sizing of combined wind and gas generators [10], and capacitor placement and sizing [11]. Although distribution system protection has technical aspects in microgrids, their insertion leads to violating the relay coordination. GA can be used for optimization problems of the distribution systems to optimize the operating time, optimal DG placement, and maximize the penetration level of DGs [12]. Also, the distribution network can be partitioned to minimize the energy exchange via the microgrids using GA-based methods [13, 14]. Utilizing decentralized controllers provides more reliability and robustness of the microgrid systems. In the case of any failure in one of the controllers, it can be ensured that the frequency regulation can be achieved via others. GA can be used for optimizing the parameters of these controllers. Such an application can also be extended to voltage and load frequency control [15]. In some studies, more than one optimization problem have been taken into account to define an objective function and solve it with GA [16].
9 Application of Heuristic Techniques and Evolutionary Algorithms. . .
223
9.2.2 Particle Swarm Optimization Algorithm (PSO) Particle Swarm Optimization (PSO) is one of the oldest intelligent optimization algorithms. The algorithm has been introduced in 1995 [17] and is inspired by the social behavior of animals such as a school of fish or flock of birds living in small and large groups. In the PSO algorithm, the population members interact straightly with each other by transforming data and recalling valuable experiences of the past. The PSO algorithm is suitable for a variety of continuous and discrete problems and provides efficient solutions to various optimization problems. Since the PSO starts with a random initial population, it is similar to many other EAs, such as the continuous GA. Unlike the GA, PSO has no evolutionary operator, such as mutation and crossover. Each population member is called a particle (which is similar to the chromosome in GA). These particles randomly take an initial value, and position and velocity are defined for each of them. The dimension of the problem is defined according to the number of parameters. Two memories are dedicated to registering the best position of each particle in the past, and the best position for all particles. Velocities and positions of particles are updated in each iteration. It is deduced that PSO has faster and better results compared with the other classical methods. The flowchart of this algorithm has been depicted in Fig. 9.2. PSO has been utilized in optimal energy management and operation scheduling of microgrids to overcome constraints such as environmental aspects [18, 19]. In [20], a comprehensive study has been done on the usage of PSO in choosing the storage characteristics and voltage regulation. PSO can be used for optimization problems on the battery storage systems and distributed battery systems (DBS) to compensate imbalanced active and reactive power flows [21, 22]. Also, this algorithm can be applied for tuning of controller parameters, which are designed to control the reactive power flow between the main grid and the microgrids [23]. PSO can be utilized to find the optimal coefficients for controllers to enhance voltage unbalance factors (VUFs) [24]. Also, it covers the sizing problems and optimal placement of DG units in microgrids [25] as well as the siting and optimization of power-sharing Schemes [26].
9.2.3 Ant Colony Optimization (ACO) Ant Colony Optimization (ACO), which is introduced in [27], is inspired by the social behavior of ants. The ants work together to find the shortest path between the nest and the food source so that they can transport food to the nest in the shortest time. The ants communicate with each other by laying down pheromone trails. If an ant follows a path, it can identify the path of its return by sensing the pheromone it has placed on the ground. Other ants can find such a path, and instead of traveling at random, follow the trail. These pheromones do not always stay on the ground
224
A. A. Ghavifekr
Fig. 9.2 General flowchart of PSO Fig. 9.3 Pseudo code of ACO
and evaporate slowly after a while. Thus, two main operators in this algorithm are pheromone placing and pheromone evaporation. Applying these operators helps to avoid convergence to the locally optimal solutions. The simple pseudo code for this algorithm is presented in Fig. 9.3.
9 Application of Heuristic Techniques and Evolutionary Algorithms. . . Fig. 9.4 Pseudo code of habitat migration in BBO
225
select H i with probability αλi if H i selected for k =1: N if rndreal(0,1)< λi select H j with probability αμ i if H j is selected Randomly select an SIV σ from H j Replace a random SIV in H i with σ end if end if end for end if
The voltage deviations could be minimized through appropriate control methods whose parameters could be optimally tuned by ACO [28]. Also, it can provide a rapid microgrid power management system, including numerous constraints and objectives such as economic, environmental, and fuel availability considerations [29, 30].
9.2.4 Biogeography Based Optimization (BBO) Biogeography based optimization (BBO) was introduced in [31] and inspired by how species are distributed in multiple habitats. This algorithm includes speciation, the migration of species between habitats, and the extinction of species. The Habitat suitability index (HIS) is defined for each island that is affected by parameters such as vegetative diversity, rainfall, topographic diversity, temperature, and so on. The determiner features are called suitability index variables (SIVs) that, unlike HIS variables, are independent parameters. Islands that have high HIS can support many species, unlike low HIS islands. In these islands, many species try to emigrate to nearby habitats due to the large population that causes much competition for recourses. These islands have high emigration and low immigration rates. If N denotes the dimension of the search space, the semi-pseudo code for habitat migration is depicted in Fig. 9.4. This algorithm has been used for designing a Linear Quadratic Regulator (LQR) and minimizing the frequency excursion following a disturbance in microgrids [32, 33].
226
A. A. Ghavifekr
9.2.5 Harmony Search Algorithm (HSA) Harmony is one of the main components of music that is added to enhance the content of music, and its proper design requires knowledge of the principles of (at least empirically) harmony science. Even ordinary music audiences, who do not listen to music professionally, will notice a deficiency in music if the harmony is removed from many pieces of music. Many famous musicians are well known for designing powerful chords and have spent a great deal of time creating and discovering the right harmonies. From the modeling and simulation process that a composer goes through to harmonize a piece of music, an algorithm is extracted, known as Harmony Search (HS) [34]. This algorithm explores all feasible solutions to reach a new one; unlike the genetic algorithm that only explores two-generation vectors. This feature makes the algorithm more flexible to search for various solution spaces. The pseudo code of this algorithm is illustrated in Fig. 9.5. From the application point of view, this algorithm has also been utilized to optimize the droop control coefficients such that the current sharing error and error due to the voltage drop are minimized [35]. HSA can find the optimum locations of energy storage systems (ESS) or distributed energy recourses (DERs) [36]. Also, it can be utilized for optimal operation of the biomass, solar, and geothermal units considering the minimum functional cost of the system [37], and for a day-ahead scheduling model for the optimal operations of microgrids [38]. HSA can also be used to cover the multi-objective sizing optimization of microgrids [39] or to optimally adjust an islanded microgrid performance in terms of frequency and voltage regulation [40].
Fig. 9.5 Pseudo code of HSA
9 Application of Heuristic Techniques and Evolutionary Algorithms. . .
227
9.2.6 Cuckoo Search Algorithm (CSA) The cuckoo search algorithm (CSA) is inspired by the behavior of a particular species of bird called cuckoo. Specifically, the unique features of this bird in laying and reproduction have been used to implement the cuckoo optimization algorithm [41]. Each cuckoo has a habitat for life and lays eggs in their habitat. Cuckoo migrates to areas with the most food and best living conditions for reproduction and lays eggs around the optimal habitats. This procedure allows for more environmentally optimal habitat to be searched for an optimal global response. After several migration cycles, most cuckoo populations converge to the optimal solution of the optimization problem. The evaluation results show that this algorithm has high speed and accuracy in convergence to the optimal solution of the benchmark functions. Even where the target function has a large number of locally optimal solutions, this algorithm is able to converge to a near approximation of the optimal global solution in a few iterations. The cuckoo optimization algorithm can be considered as one of the successful implementations of the nature-inspired process. One of the advantages of this algorithm is finding an accurate value in fewer iterations. The pseudo code of this algorithm is as follows (Fig. 9.6): This algorithm can be used for solving the multimodal optimization problems in microgrid systems and operation scheduling optimization [42, 43]. CSA can also be developed to optimize the performance of the BESS to mitigate the voltage fluctuation in microgrids [44]. It can be utilized to minimize the emission and generation costs of the microgrids while meeting system constraints and its hourly demands [45] as well as solving nonlinear sizing problems [46].
Fig. 9.6 Pseudo code of CSA
228
A. A. Ghavifekr
9.2.7 Artificial Bee Colonies (ABC) Bees are insects that live in relatively large colonies. Besides the benefits of bees in agriculture and horticulture, the regular social behavior of these organisms has always been the source of inspiration for scientific studies. The mathematical model for this algorithm is proposed in [47] based on the intelligent foraging behavior of bee swarms. In a bee colony, adopted by the ABC algorithm, there are three types of bees: employed bees, onlooker bees, and scouts. Half of the colonies contain employed bees, and the other half contains onlooker bees. Employed bees are responsible for exploiting previously discovered food sources, as well as providing data to other onlooker bees in the hive about the quality of the food they are extracting. Onlooker bees stay in the hive and, according to the information shared by the employed bees, decide on a food source to explore. Scouts randomly search the environment to find a new food source based on intrinsic motivation or external evidence. The general flowchart of this algorithm is presented in Fig. 9.7. This algorithm can also be used for solving a variety of optimization problems in microgrids ranging from economic dispatch (ED) and network reconfiguration of microgrids related to planning and sizing problems [48, 49]. In some literature, Fig. 9.7 General flowchart of ABC
9 Application of Heuristic Techniques and Evolutionary Algorithms. . .
229
it has also been utilized to minimize the market cleaning price (MCP), production cost, and provide better utilization of renewable energy resources [50].
9.2.8 Grey Wolf Optimization (GWO) Gray wolves are animals that live in a semi-democracy environment, and each wolf has its position in society. Gray wolves live and hunt collectively. In each group of gray wolves, there are, on average, between five and 12 wolves. The gray wolves first surround the prey and begin to tame it by narrowing the siege, then according to the order of the leader wolf attack the victim and eventually kill it. The aforementioned process is mathematically modeled in [51]. In implementing the GWO algorithm, four types of gray wolves, such as Alpha, Beta, Delta, and Omega, have been used to model a hierarchical gray wolf algorithm, which includes three steps of the search for prey, prey siege, and bait attacks. Alpha pairs, known as group leaders, make decisions about hunting, sleeping time, waking up, and so on. Alpha decisions apply to the whole group. The second class belongs to the Beta wolves, which help alpha wolves in decision making and other group activities. These wolves are the best candidate for substituting when the alphas are very old or dying. The lowest ranked wolves are Omega wolves, which play the role of scapegoat. They must obey all other wolves and are the last groups that are allowed to eat. Wolves not mentioned in the aforementioned groups are called Delta wolves. Delta wolves are more lower order than alpha and beta but superior to omega. The flowchart of this algorithm is depicted in Fig. 9.8. GWO has been frequently used in the literature to handle different optimization tasks in microgrids. Special attention has been made in its application to determine the optimal size of battery energy storage (BES) considering various constraints such as power and energy capacity [52]. Also, GWO has been used to reduce the operating time of directional overcurrent relays (DOCRs) and to do energy management scenarios [53]. This algorithm has also been utilized for the distributed hierarchical control of microgrids to solve the dispatching problem [54].
9.2.9 Firefly Algorithm (FA) The Firefly Algorithm (FA) has been introduced in 2008 [55]. The main idea is the flashing communication between fireflies, where each of them can be characterized by their flashing light produced by bioluminescence or biochemical structure. This population-based algorithm can be considered as Swarm Intelligence, where higher levels of intelligence are created by the cooperation and competition of the simple and the less intelligent members, which is certainly not obtainable by any of the components. The flashing light in this algorithm is the main concept for mating, and the degree of the attractiveness of each firefly is proportional to
230
A. A. Ghavifekr
Fig. 9.8 General flowchart of GWO
the related brightness and intensity. This persuades less bright fireflies to move toward the brighter one. Thus, it can be deduced that more brightness causes less distance between fireflies. This brightness consists of the objective function, and the evolutionary process is performed to optimize it. The simple pseudo code for this algorithm is presented in Fig. 9.9. This algorithm can be used for energy management and operation scheduling of microgrids, such as determining the optimal power output of each generator at minimum cost [56, 57]. Also, it can be utilized for tuning of controller parameters in a decentralized control scheme to mitigate varying load perturbations and improve the frequency and voltage stability of microgrids [15, 58, 59]. FA has been applied for calculation of the payoff function of each coalition group between microgrids and enables them to maximize their utilities [60]. It presents an acceptable performance in energy storage system management to optimize operation systems [61].
9.2.10 Grasshopper Optimization Algorithm (GOA) This optimization algorithm was presented in 2017 [62]. Grasshoppers have usually a destructive role in nature and can cause damage to the agricultural produces and crop productions. The lifecycle of the full-grown adult grasshopper is depicted in Fig. 9.10. The algorithm is inspired by the food searching behaviors of the grasshoppers and their reactions to the surrendering environment. In this algorithm, updating the location of each grasshopper depends on the distance from the entire
9 Application of Heuristic Techniques and Evolutionary Algorithms. . .
231
Fig. 9.9 Pseudo code of FA Fig. 9.10 The lifecycle of the full-grown adult grasshopper
population in the current generation, and the location of the best grasshopper. Besides simplicity in implementation, the main feature of this algorithm consists of only one parameter to adjust. In [62], the swarm behavior of these insects is modeled mathematically for solving a variety of optimization problems. The pseudo code of this algorithm is presented in Fig. 9.11. This algorithm can be used for tuning of controller parameters for load frequency control of interconnected microgrid power systems [63, 64]. Also, it is able to size the autonomous microgrid systems, optimally [65].
232
A. A. Ghavifekr
Fig. 9.11 Pseudo code of GOA
9.2.11 Whale Optimization Algorithm (WOA) This algorithm is inspired by the hunting mechanism of humpback whales in nature [66]. These kinds of whales can recognize the position of prey and surrendered them. Since the optimal design position in the search space is not known, the algorithm considers the current best candidate solution as the target or a close solution to the optimum. After the definition of the best search agent, the other agents will try to update their positions regarding the best search agent. The pseudo code of this algorithm is depicted in Fig. 9.12. This algorithm has been utilized for the optimal tuning of PI controllers in an autonomous microgrid system. This can enhance the flow of active and reactive power during load variation [67]. WOA presents an acceptable performance in solving the combined economic emission dispatch problem that is one of the wellknown problems of the energy management systems [68]. Table 9.1 presents a timeline of the development of the aforementioned naturebased evolutionary algorithms.
9 Application of Heuristic Techniques and Evolutionary Algorithms. . .
233
Fig. 9.12 Pseudo code of WOA Table 9.1 EA optimization techniques development timeline Method Genetic algorithm (GA) [69] Particle swarm optimization (PSO) [17] Harmony search algorithm (HAS) [34] Ant colony optimization (ACO) [27] Biogeography-based optimization (BBO) [31] Firefly algorithm (FA) [55] Artificial bee colonies (ABC) [47] Cuckoo search algorithm (CSA) [41] Grey wolf optimization (GWO) [51] Whale optimization algorithm (WOA) [66] Grasshopper optimization algorithm (GOA) [62]
Developer John. R. Koza James Kennedy and Russel. C. Earhart Z. W. Geem, J. H. Kim, and G. V. Loganathan M. Dorigo, M. Birattari, and T. Stutzle D. Simon
Year 1988 1995 2001
X.-S. Yang D. Karaboga R. Rajabioun S. Mirjalili, S. M. Mirjalili, and A. Lewis S. Mirjalili and A. Lewis
2008 2010 2011 2014 2016
S. Saremi, S. Mirjalili, and A. Lewis
2017
2006 2008
9.3 Illustrative Examples on Application of EAs in Microgrids 9.3.1 Energy Management and Operation Scheduling EA methods play a significant role in the scheduling of microgrids to obtain better quality power at minimum cost. Since the past decade, EA methods have
234
A. A. Ghavifekr
been employed extensively in unit commitment, economic dispatch, and generation scheduling problems [70]. A genetic algorithm has been proposed to obtain favorable solutions for operation scheduling and energy management problems [71]. The energy management system should solve the optimization problem with the classification of all the necessary information and resend it to each distributed energy resources (DERs). This concept can be extended to the optimal power flow (OPF) problems in a local microgrid setup [72]. PSO has been utilized in optimal energy management and operation scheduling of microgrids to overcome the constraints such as environmental aspects [73]. ACO can provide a rapid microgrid power management system, including numerous constraints and objectives such as economic, environmental, and fuel availability considerations [29, 30]. HSA can be utilized for optimal operation of the biomass, solar, and geothermal units considering the minimum functional cost of the system [37]. In [38], the cost function is defined as the sum of the total generation and operation costs of PV arrays, WT, battery, and DGs. This is such a nonlinear mixed-integer programming problem for smart microgrids, which is hard to solve with conventional methods. HSA with modified mutation and selection operators and adaptive parameters is utilized to solve this optimization problem. It can be concluded that for optimal day-ahead scheduling of microgrids, HSA is more reliable under both fault and normal operation situations. CSA can be used for solving the multimodal optimization problems in microgrid systems and operation scheduling optimization [74, 75]. In [76], the improved version of CSA is applied on a microgrid, including numerous renewable and conventional energy power plants such as two wind powers, two diesels, and three fuel-cell plants. The system is presumed to be isolated from the electric network, and the local generation cost optimization problem is solved with CSA. Also, in [74], CSA is utilized for different scenarios, including MG with all sources, all sources without wind energy, all sources without solar energy, and all sources without solar and wind energy. ABC can be utilized to minimize the market cleaning price (MCP), production cost, and providing better utilization of renewable energy resources [77]. The main objective in [78] is providing optimal scheduling for a real system of 24-hours load demand with the least operation cost of a microgrid, which includes PV array, wind turbine, fuel cell, diesel engine, and microturbine. Applying ABC for solving the proposed optimization problem demonstrates that the minimum cost is obtained when the diesel engine shares less power in comparison with the fuel cell, and microturbine. GWO can help to reduce the operating time of directional overcurrent relays (DOCRs) and other energy management problems [79]. In [80], a different cost function has been defined that leads to optimization of the total grid loss, minimization of the system pollution and voltage deviation, and reduction of the total energy costs in the operation stage. Thus, the multi-objective grey wolf algorithm has been utilized for solving this complex problem. It can be extracted that the integration of storage units and DGs in microgrids has been provided economic and environmental advantages besides power loss reduction and voltage deviation improvement. The improved GWO method helps to accurate operation modeling of DGs and storage units.
9 Application of Heuristic Techniques and Evolutionary Algorithms. . .
235
FA can be used for energy management and operation scheduling of microgrids, such as determining the optimal power output of each generator at minimum cost [81]. These studies demonstrate that FA can achieve the lowest operation cost in comparison with all the aforementioned algorithms [56]. Assume that the MG system included various DG units such as FC, WT, PV, and MT. All units are operated in the unity power factor. The main goal is the reduction of operating costs through the optimal operation of a variety of DG units. The objective function can be defined as: Min F =
T
ft + OM DG , t = 1, 2, . . . , T
(9.1)
t=0
where ft = Cgrid,t + CDG,t + SU C F C,t + SU C MT ,t + SDC MT ,t + SDC F C,t (9.2) and Cgrid, t can be defined as:
Cgrid,t
⎧ if Pgrid,t > 0 ⎨ Pgrid,t Bgrid,t = (1 − tax) Pgrid,t Bgrid,t if Pgrid,t < 0 ⎩ 0 if Pgrid,t = 0
(9.3)
Also, we have SU C F C,t = max 0, uF C,t − uF C,t−1 ∗ SU F C SU C MT ,t = max 0, uMT ,t − uMT ,t−1 ∗ SU MT
(9.4)
where OMDG is the total maintenance, and operation cost of DG units connected to the microgrid, CDG, t is the cost of operating power, and fuel of Dg units Cgrid, t is the grid cost at time t, SUCFC, t and SUCMT, t are the start-up costs of FC and MT at time t, respectively. SDCFC, t and SDCMT, t are the shut-down costs of FC and MT at time t, respectively. Pgrid, t is the power of the grid at time t. OMDG can be calculated as OM DG = T ∗ (OM F C + OM MT + OM W T + OM P V )
(9.5)
where OMFC , OMMT , OMWT and OMPV are operation costs of FC, MT, WT, and PV, respectively. T denotes the total time in hours. Now some constraints should be considered for this optimization problem. 1. The load demand balance constraints should be considered as:
236
A. A. Ghavifekr
Table 9.2 Power limitations and bids of grid and DG units Type FC PV WT MT GRID
Min Power (kW) 3 0 0 6 −30
Max Power (kW) 30 25 25 30 30
Bid ($/kW) 0.326 2.87 1.19 0.507 –
OM ($/kW) 0.095 0.231 0.583 0.049 –
SUC/SDC ($/kW) 1.833 0 0 1.06 –
PD,t = PF C,t uF C,t + PMT ,t uMT ,t + PP V ,t uP V ,t + Pgrid,t ugrid,t + PW T ,t uW T ,t (9.6) where Pφ and uφ demonstrate the power output of φ and the operating status at time t, respectively. 2. Constraint of the output power of DG units (PDG, t ) is formulated as PDG,min ≤ PDG,t ≤ PDG,max
(9.7)
where PDG, max and PDG, min are maximum and minimum power limits of DG units, respectively. 3. Constraint of grid output power (PDG , t) can be formulated as: Pgrid,min ≤ Pgrid,t ≤ Pgrid,max
(9.8)
where Pgrid, max and Pgrid, min are the maximum and minimum power limits of the utility grid. The optimization problem has been defined for a typical microgrid system [56] which is connected with other DG units such as FC, PV, WT, and MT. The minimum and maximum power limits, maintenance and operation cost, bidding cost, SDC, and SUC for units are defined in the following Table 9.2. Now the objective function is ready and the FA algorithm can be applied for solving the optimization problem. The number of iterations is defined as 500. The initial parameters of the algorithm are chosen as step size (α) = 0.5, attractiveness (β) = 0.2, and absorption coefficient (γ ) = 1. In [56], the results will be compared with other EA techniques to validate the priority of FA for the cost minimization problem of microgrids. For PSO, the inertia weight is w = 1, and the acceleration coefficients are chosen as C1 = 0.5 and C2 = 1.5. The intensity of attraction for GOA is 0.5 and the minimum and maximum shrinking factors are 0.1 and 0.9, respectively. For GA, the crossover rate, and mutation rate are selected as Pc = 0.9 and Pm = 0.01,respectively. Emigration and immigration rates for BBO are E = 1 and I = 1, respectively. For comparison purposes, the similar swarm size, number of the habitats, and number of particles are chosen for the aforementioned algorithms. The results are given in Table 9.3.
9 Application of Heuristic Techniques and Evolutionary Algorithms. . .
237
Table 9.3 Comparative study of total operation cost in energy management problem Methods PSO GA BBO FA GOA
Total operation cost ($) 1075.576 1157.623 935.412 903.674 929.325
Iteration number of reaching to the min value 420 480 325 285 315
The total operation cost obtained by FA is compared with other EAs in Table 9.3 and demonstrates the better performance of the FA algorithm. If we increase the number of population, the total generation cost will have more reduction. However, the computation time is also growing.
9.3.2 Optimal Placement and Sizing of Energy-Related Devices The sizing optimization approaches can provide the lowest investment with full use of DG units. The type of suitable fuels for the power plant is an important problem and has a critical impact on reliability and cost-efficiency. The microgrid consists of a wind turbine (WT), photovoltaic (PV), diesel generator, and battery storage system (BSS). The main objective is the optimal construction that can satisfy the demand of the residential housing reliability based on the DPSP while minimizing the cost of energy. GA can be used for solving the optimal sizing problems of microgrids. This technique has been applied for hybrid AC–DC microgrids [6], battery energy storage system (BESS) optimal sizing [82, 83], optimization of storage devices [8], combined heat and power (CHP) microgrid systems [9], determination of optimal sizing of combined wind and gas generators [10], and capacitor placement and sizing [11]. GA can be utilized for optimization problems of the distribution systems to optimize the operating time, optimal DG placement, and maximize the penetration level of DGs [12]. PSO can be used for optimization problems on the battery storage systems and distributed battery systems (DBS) to compensate imbalanced active and reactive power flows [84, 85]. Also, it can cover sizing problems and optimal placement of DG units in a microgrid [86]. The approaches based on HSA are presented for multi-objective sizing optimization of microgrids in [87]. CSA is another evolutionary algorithm that can be developed to optimize the performance of the BESS to mitigate the voltage fluctuation in microgrids [44]. It can be utilized for solving nonlinear sizing problems [46]. The problem of economic dispatch (ED) and network reconfiguration of microgrids can be solved by ABC [48, 49]. Recently, it is deduced that GWO produces the most optimal solution for the BESS sizing problem and determines the optimal size of battery energy storage (BES) considering various constraints, such as power
238
A. A. Ghavifekr
and energy capacity of BES [88, 89]. GOA can optimally size the autonomous microgrid systems [65], and comparing the results with all of the aforementioned papers demonstrates that GOA has better performance and very fast convergence as well as a promising balance between exploration and exploitation. We will present it with more details hereafter. The objective functions for sizing problems can be defined based on the cost of energy (COE) and the deficiency of power supply probability (DPSP). Optimization of these parameters would guarantee minimum cost and reliable power supply. COE (in $/kWh) is calculated as: COE = T NP C/
8760
Pl (h) × CRF
(9.9)
t=1
where TNPC is the total net present cost, includes replacement, capital, maintenance, and operation cost. Pl (h)is the hourly consumption of energy and capital recovery factor (CRF) is obtained by: CRF = ri (1 + ro )n /(1 + ri )n − 1
(9.10)
where ri and ro present the real and optimal interest rates and n shows the system life span. Another adjustable factor is DPSP that can be assumed as the reliability index of microgrids that denotes the probability of the power supply leakage for energy demand. It is formulated as: DP SP = (9.11) (Pl − PP V − PW T + PSOC + PDG ) /Pl where Pl , PPV , PWT , PSOC and PDG denote energy demand, PV power, the power output of the wind turbine, power related to minimum battery state of charge, and generated power by a diesel generator, respectively. The constraints of the optimization problem can be defined based on the number of PV and WT as: PV 0 ≤ N P V ≤ Nmax W T WT 0≤N ≤ Nmax
(9.12)
P V and N W T show the maximum number of PV and WTs, and assumed to where Nmax max be 45 and 10, respectively. Also, we have three autonomy days. Now, the optimization problem can be restated in its traditional form, which can be generally formulated as:
min U (x) = [u1 (x), u2 (x), . . . . uk (x)] subj ect to c(x) ≤ 0 and g(x) = 0
(9.13)
where, x is the vector of the search space, U(x)is a vector of objective functions, and g(x) and c(x) present the equality and inequality constraints.
9 Application of Heuristic Techniques and Evolutionary Algorithms. . .
239
Fig. 9.13 A comparison of algorithms used for the cost of energy reduction in microgrids [65]
GOA algorithm is employed to solve this optimization problem and is demonstrated better performance when compared with other well-known methods such as GA and CSA [65]. Like its counterpart methods, it starts with random particles in the space of the landscape with user-defined limitations. The initial values of GOA parameters are chosen as population size = 5, the number of iterations = 100, and intensity of attraction = 0.5. The minimum and maximum shrinking factors are 0.1 and 0.9, respectively. For comparison purposes, the similar swarm size and number of the nest are chosen for PSO and CS algorithms. Also, the maximum generation of PSO equals the number of iterations of CS, and GOA. The abandoned eggs fraction and smart eggs fraction for CS are Pa = 0.2 and Pc = 0.5, respectively. For PSO, the inertia weight is w = 1,and the acceleration coefficients are chosen as C1 = 0.25 and C2 = 1.75. The convergence characteristic of GOA, CSA, and PSO is depicted in Fig. 9.13, and the optimal size of the system is presented in Table 9.4. It is noticeable that, by applying GOA, the optimal configuration includes 26 PV panels and a 4 kW WT. Furthermore, the COE is minimized to 0.365$/kWh which is less than its counterparts, PSO, and CS algorithms. Zero DPSP means that energy demand is fully satisfied. The results demonstrate that GOA minimizes computer memory usage, and reduces the computation time due to the fast convergence. Also, the CS algorithm has better performance than PSO.
240 Table 9.4 Optimal sizing results obtained by applying EA [65]
A. A. Ghavifekr
DPSP Number of WT Number of PV BSS capacity (kW) DG capacity (kW) COE ($/kWh)
PSO 0 7 30 40 4 0.3674
CS 0 6 29 40 4 0.3662
GOA 0 4 26 40 4 0.3656
9.3.3 Microgrid Optimal Voltage and Frequency Control The interconnection of microgrids can be adjusted and controlled to provide them to work in both islanded and grid-connected modes of operation. The control concern in the grid-connected mode is to regulate the reactive and active power flow among DGs connected within MG and between the main grid and MG. In the latter, the frequency and voltage of the system are controlled by the power system, and there is no control objective to optimize. But in the islanding mode, controlling the frequency and voltage of the whole system is a crustal problem. The control system must guarantee that there are not any considerable circulating currents from the micro sources. These currents even can be produced by the small mismatch in voltage and frequency set points. In order to provide smooth operation of microgrids, an optimized control plan is mostly required for islanded operation. Any uncertainties or inappropriate selection for the gain of controller leads to a large variation in frequency and voltage levels in the operation mode. This architecture includes a DC–DC boost converter, two PV panels, a three-phase voltage source inverter (VSI), a coupling inductor, an RLC filter, and a three-phase load. Utilizing decentralized controllers provides more reliability and robustness of the microgrid systems. In the case of any failure in one of the controllers, it can be ensured that the frequency regulation can be achieved via others. GA can be used for optimizing the parameters of these controllers and can be extended to voltage and load frequency control [90, 91]. PSO is another evolutionary algorithm that can be used for tuning of controller parameters, which are designed to control the reactive power flow between the main grid and the microgrids [23]. Using DGs leads to active provide more flexible microgrids. PSO can be utilized to find the optimal coefficients for controllers to enhance voltage unbalance factors (VUFs) [92, 93]. BBO [33], and HSA [40] have promising results in designing a Linear Quadratic Regulator (LQR) and minimize the frequency excursion following a disturbance in microgrids to improve the islanded microgrid frequency and voltage. FA [58, 59] and GOA [63, 64, 94] can be utilized for tuning of controller parameters in a decentralized control scheme and load frequency control to mitigate varying load perturbations and improve the frequency and voltage stability of interconnected microgrid power systems. The control circuit generates the controlled pulses for the VSI for the production of pure sinusoidal voltage waves to provide high-quality power for loads. GOA
9 Application of Heuristic Techniques and Evolutionary Algorithms. . .
241
algorithm is utilized in this step to promote the dynamic output of the microgrid with the efficient tuning of controller gains. As the proposed method in [95] minimize the voltage controller corresponding errors, it is released from parameter optimization for the current controller. However, using constant parameters for controller gains cannot cause an optimal operation. Thus, proper tuning of the parameters can enhance the power quality and performance of the system during the load changes and disturbances. The results demonstrate that GOA provides a minimum output current to optimize frequency and voltage overshoots. ITAE is selected as the optimization index since it allows for the smoother implementation in comparison with its counterparts such as an integral square error and absolute error. ITAE can be defined as: .∞ t |e| dt
I T AE =
(9.14)
0
where e(t) is the error signal that shows the difference between the controlled value and the reference value. Using Eq. (9.14), the fitness function can be assumed as the summation of the frequency and voltage error functions and formulated as: ⎛∞ ⎞ . .∞ Min ⎝ t × |ev | dt + t × ef dt ⎠ 0
(9.15)
0
Minimization of the aforementioned objective function leads to the optimal selection of controller gains, which enhance the dynamic performance of the islanded microgrid system. The GOA algorithm has been employed for parameter tuning of PI controllers and its performance compared with PSO and WOA. The numbers of iterations and particles are set to 50 for each of the algorithms to have a reasonable comparison. The cognitive and social constants of the PSO are selected as C1 = C2 = 2, and inertial weight is C = 0.5. The control parameters of WOA are d = 3, and 0 < r < 1. Where d is the number of variables, and r is the random number. For GOA, the intensity of attraction is 0.5, and the minimum and maximum shrinking factors are 0.1 and 0.9, respectively. The convergence behavior of these algorithms is depicted in Fig. 9.14. Smaller final value and higher convergence rate represent the better performance of the controller. The number of iterations and final optimized values for the proposed objective function is given in Table 9.5. As it is deduced from Table 9.5, the GOA algorithm provides a faster and more accurate optimal solution in comparison with the PSO and WOA. The searching process stops in the given criteria for the objective function or in the given iteration numbers. The final values for the voltage controller gains (Kpv and Kiv ) and frequency controller gains(Kpf and Kif ) after applying the GOA algorithm are given in Table 9.6.
242
A. A. Ghavifekr
Fig. 9.14 The convergence plot of the EA methods [95] Table 9.5 The minimum fitness function and convergence values of proposed EA algorithms [95] EA method PSO WOA GOA
Min value of FF 1.00145 0.87710 0.49635
Table 9.6 Final values of controller gains [95]
Iteration number of reaching to the min value 21 24 16 EA method PSO WOA GOA
Kpv 0.25 0.94 17.09
Kiv 25.64 1.60 27.87
Kpf 0.09 0.04 0.08
Kif 23.64 26.74 12.26
The voltage and frequency regulation comparison between the EA algorithms are depicted in Fig. 9.15. The aforementioned designed controller restores the nominal voltage and frequency of the system as soon as the voltage falls. GOA based controller provides the frequency (50 Hz) and nominal voltage (375 V) as well as minimum overshoot and settling time and can keep the frequency and voltage in the standard limits. Also, it is deduced that the proposed control method can provide a better situation for the high power generation of DGs and is faster and more efficient. Moreover, it is concluded from Fig. 9.15 that, the EA-based control strategy keeps the frequency value within its ±1%tolerance and is able to reach its rated voltage within 0.07 s.
9 Application of Heuristic Techniques and Evolutionary Algorithms. . .
(a)
243
450 PSO
WOA
GOA
400
Voltage (Volt)
350 300 250 200 150 100 50 0
0
0.1
0.2
0.3
0.4
(b)
0.7
0.8
0.9
WOA
PSO
50.3
Frequency (Hz)
0.5 0.6 Time (sec)
1
GOA
50.3 50.1 50 49.9 49.8
0
0.1
0.2
0.3
0.4
0.5 0.6 Time (sec)
0.7
0.8
0.9
1
Fig. 9.15 Voltage and frequency regulations comparison after applying EA methods [95]
9.4 Conclusion This chapter provides an overview of the latest studies concerning the use of EAs in microgrids planning. Approaches for optimal sizing, operational scheduling, and voltage and frequency control of autonomous microgrids based on recently introduced EAs have been presented. Regarding the probabilistic nature of these algorithms, they present different performance in solving optimization problems. GWO produces the most optimal solution for the BESS sizing problem and determines the optimal size of battery energy storage (BES) considering various constraints such as power and energy capacity of BES. On the other hand, GOA has better performance in optimal sizing and optimal control problems of microgrids in comparison with other algorithms. It can minimize computer memory usage, and reduce the computation time due to the fast convergence as well as a promising
244
A. A. Ghavifekr
Table 9.7 Microgrids planning approaches and references using evolutionary algorithms Evolutionary Algorithm Genetic algorithm
Particle swarm optimization
Ant colony optimization
Biogeography-based optimization Harmony search algorithm
Cuckoo search algorithm (CSA)
Artificial bee colonies (ABC)
Grey wolf optimization (GWO)
Firefly algorithm (FA)
Grasshopper optimization algorithm (GWO)
Whale optimization algorithm (WOA)
Related Problem Operation scheduling Scheduling of power generation Sources and sizing BESS and storage devices sizing Siting Voltage and frequency control Operation scheduling BESS and storage devices sizing Scheduling of power generation Sources and sizing Siting Voltage and frequency control Operation scheduling Scheduling of power generation Sources and sizing Voltage and frequency control Siting Operation scheduling Scheduling of power generation Sources and sizing Voltage and frequency control BESS and storage devices sizing Operation scheduling Scheduling of power generation Sources and sizing Operation scheduling Scheduling of power generation Sources and sizing Scheduling of power generation Sources and sizing Operation scheduling Siting Voltage and frequency control Operation scheduling BESS and storage devices sizing Siting Voltage and frequency control
References [4, 71, 96–98] [6, 9–11] [7, 8, 82, 83] [13, 14, 99] [15, 90, 91, 100] [18, 19, 73, 101, 102] [20–22, 84, 85] [25, 86] [26, 103] [24, 92, 93, 104, 105] [29, 30, 106] [28] [32, 33] [35, 36] [37, 38] [39, 87, 107] [40, 108] [44] [42, 43, 74–76] [45, 46] [50, 77, 78, 109] [48, 49] [52, 88, 89] [53, 79, 80] [54] [15, 58, 59] [56, 57, 81] [61, 110, 111] [60] [63, 64, 94]
Scheduling of power generation Sources and sizing Operation scheduling
[65] [68]
Voltage and frequency control
[67]
9 Application of Heuristic Techniques and Evolutionary Algorithms. . .
245
balance between exploration and exploitation. Also, it is deduced that FA has better performance in energy management and operation scheduling of microgrids, such as determining the optimal power output of each generator at a minimum cost. This review of the application of EA techniques for microgrid systems can be instructive for microgrid planners and power system engineers. All reviewed papers are classified in Table 9.7, regarding their optimization algorithm and the related problem.
References 1. Kang, X., Nuworklo, C. E., Tekpeti, B. S., & Kheshti, M. (2017). Protection of micro-grid systems: A comprehensive survey. The Journal of Engineering, 2017, 1515–1518. 2. Connolly, D., Lund, H., Mathiesen, B. V., & Leahy, M. (2010). A review of computer tools for analysing the integration of renewable energy into various energy systems. Applied Energy, 87, 1059–1082. 3. F. Shariatzadeh, N. Kumar, and A. K. Srivastava, ”Intelligent control algorithms for optimal reconfiguration of microgrid distribution system,“ in 2015 IEEE Industry Applications Society Annual Meeting, 2015, pp. 1–8. 4. Leonori, S., Paschero, M., Mascioli, F. M. F., & Rizzi, A. (2020). Optimization strategies for microgrid energy management systems by genetic algorithms. Applied Soft Computing, 86, 105903. 5. V. Suresh and P. Janik, ”Optimal Power Flow in Microgrids using Genetic Algorithm,“ in 2019 2nd International Conference on Power and Embedded Drive Control (ICPEDC), 2019, pp. 483–487. 6. Rousis, A. O., Konstantelos, I., & Strbac, G. (2019). A planning model for a hybrid AC–DC microgrid using a novel GA/AC OPF algorithm. IEEE Transactions on Power Systems, 35, 227–237. 7. Luo, L., Abdulkareem, S. S., Rezvani, A., Miveh, M. R., Samad, S., Aljojo, N., et al. (2020). Optimal scheduling of a renewable based microgrid considering photovoltaic system and battery energy storage under uncertainty. Journal of Energy Storage, 28, 101306. 8. T. Adefarati, S. Potgieter, R. Bansal, R. Naidoo, R. Rizzo, and P. Sanjeevikumar, ”Optimization of PV-Wind-Battery Storage Microgrid System Utilizing a Genetic Algorithm,“ in 2019 International Conference on Clean Electrical Power (ICCEP), 2019, pp. 633–638. 9. J. Zhao, H. Nian, and L. Kong, ”multiobjective sizing optimization for combined heat and power microgrids using a triangular evaluation model and the self-adaptive genetic algorithm,“ in 2019 IEEE Innovative Smart Grid Technologies-Asia (ISGT Asia), 2019, pp. 1840-1845. 10. A. S. Eldessouky and H. A. Gabbar, ”Micro grid renewables dynamic and static performance optimization using genetic algorithm,“ in 2015 IEEE International Conference on Smart Energy Grid Engineering (SEGE), 2015, pp. 1–6. 11. M. Mahdavian, M. H. Kafi, A. Movahedi, and M. Janghorbani, ”improve performance in electrical power distribution system by optimal capacitor placement using genetic algorithm,“ in 2017 14th International Conference on Electrical Engineering/Electronics, Computer, Telecommunications and Information Technology (ECTI-CON), 2017, pp. 749–752. 12. P. P. Bagul and S. M. Akolkar, ”Relay coordination in microgrid,“ in 2017 International Conference on Computing Methodologies and Communication (ICCMC), 2017, pp. 784–789.
246
A. A. Ghavifekr
13. M. Kamal, I. Ashraf, and E. Fernandez, ”Energy Resource Planning for a Rural Microgrid: Comparison of Results using Different Optimization Algorithms,“ in 2019 Second International Conference on Advanced Computational and Communication Paradigms (ICACCP), 2019, pp. 1–6. 14. S. Korjani, A. Facchini, M. Mureddu, and A. Damiano, ”A genetic algorithm approach for the identification of microgrids partitioning into distribution networks,“ in IECON 2017-43rd Annual Conference of the IEEE Industrial Electronics Society, 2017, pp. 21–25. 15. A. Sawarni, K. Dhamane, D. Kumar, and H. Mathur, ”Decentralized Frequency Control for an Isolated Microgrid using Nature Inspired Algorithms,“ in 2019 IEEE International Conference on Electrical, Computer and Communication Technologies (ICECCT), 2019, pp. 1–6. 16. Parol, M., Wójtowicz, T., Ksi˛ez¨ yk, K., Wenge, C., Balischewski, S., & Arendarski, B. (2020). Optimum management of power and energy in low voltage microgrids using evolutionary algorithms and energy storage. International Journal of Electrical Power & Energy Systems, 119, 105886. 17. J. Kennedy and R. Eberhart, ”Particle swarm optimization,“ in Proceedings of ICNN’95International Conference on Neural Networks, 1995, pp. 1942-1948. 18. L. Ravi, C. V. Kumar, and M. R. Babu, ”Stochastic optimal management of renewable microgrid using simplified Particle Swarm Optimization algorithm,“ in 2018 4th International Conference on Electrical Energy Systems (ICEES), 2018, pp. 261–266. 19. Hossain, M. A., Pota, H. R., Squartini, S., & Abdou, A. F. (2019). Modified PSO algorithm for real-time energy management in grid-connected microgrids. Renewable Energy, 136, 746– 757. 20. D. Kaczorowska and J. Rezmer, ”Particle swarm algorithm for microgrid optimization,“ in 2018 Innovative Materials and Technologies in Electrical Engineering (i-MITEL), 2018, pp. 1–4. 21. M. W. Alzahlan, F. S. El-Faouri, M. G. Batarseh, A. Mohammad, and M. E. Za’ter, ”Particle Swarm Optimization of a Microgrid’s Cost Function Involving Distributed Generation and Highly Fluctuating Load,“ in 2019 IEEE Jordan International Joint Conference on Electrical Engineering and Information Technology (JEEIT), 2019, pp. 319–324. 22. Y. Jiang, Y. Yang, S.-C. Tan, and S.-Y. R. Hui, ”Mitigating distribution power losses of standalone AC microgrids using particle-swarm-optimization control for distributed battery systems,“ in 2019 IEEE Applied Power Electronics Conference and Exposition (APEC), 2019, pp. 3224–3229. 23. P. Malathi and M. Saravanan, ”Power management in microgrid using PI controller tuned by particle swarm optimization,“ in 2017 Innovations in Power and Advanced Computing Technologies (i-PACT), 2017, pp. 1–6. 24. S. H. Qazi, M. Mustafa, N. Hussain, and U. Sultana, ”Performance evaluation of PI and PIPSO in improving power quality of an autonomous microgrid,“ 2017. 25. R. R. Nivedha, J. G. Singh, and W. Ongsakul, ”PSO based economic dispatch of a hybrid microgrid system,“ in 2018 International Conference on Power, Signals, Control and Computation (EPSCICON), 2018, pp. 1–5. 26. L. K. Vedulla and M. K. Mishra, ”PSO based power sharing scheme for an islanded DC microgrid system,“ in IECON 2017-43rd Annual Conference of the IEEE Industrial Electronics Society, 2017, pp. 392–397. 27. Dorigo, M., Birattari, M., & Stutzle, T. (2006). Ant colony optimization. IEEE Computational Intelligence Magazine, 1, 28–39. 28. N. Emerson and S. Srinivasan, ”Integrating hybrid power source into islanded microgrid using ant colony optimization,“ in 2015 International Conference on Advanced Computing and Communication Systems, 2015, pp. 1–4. 29. I. N. Trivedi, D. K. Thesiya, A. Esmat, and P. Jangir, ”A multiple environment dispatch problem solution using ant colony optimization for micro-grids,“ in 2015 International Conference on Power and Advanced Control Engineering (ICPACE), 2015, pp. 109–115.
9 Application of Heuristic Techniques and Evolutionary Algorithms. . .
247
30. Marzband, M., Yousefnejad, E., Sumper, A., & Domínguez-García, J. L. (2016). Real time experimental implementation of optimum energy management system in standalone microgrid by using multi-layer ant colony optimization. International Journal of Electrical Power & Energy Systems, 75, 265–274. 31. Simon, D. (2008). Biogeography-based optimization. IEEE Transactions on Evolutionary Computation, 12, 702–713. 32. Mishra, S., Mallesham, G., & Sekhar, P. (2013). Biogeography based optimal state feedback controller for frequency regulation of a smart microgrid. IEEE Transactions on Smart Grid, 4, 628–637. 33. R. H. Kumar and S. Ushakumari, ”Biogeography based tuning of PID controllers for Load Frequency Control in microgrid,“ in 2014 International Conference on Circuits, Power and Computing Technologies [ICCPCT-2014], 2014, pp. 797–802. 34. Geem, Z. W., Kim, J. H., & Loganathan, G. V. (2001). A new heuristic optimization algorithm: harmony search. simulation, 76, 60–68. 35. P. Monica and M. Kowsalya, ”Improved Harmony Search Based Optimization of Droop Control Parameters for Load Sharing in DC Microgrids,“ in TENCON 2019–2019 IEEE Region 10 Conference (TENCON), 2019, pp. 2199–2203. 36. N. A. Ashtiani, M. Gholami, and G. Gharehpetian, ”Optimal allocation of energy storage systems in connected microgrid to minimize the energy cost,“ in 2014 19th Conference on Electrical Power Distribution Networks (EPDC), 2014, pp. 25–28. 37. E. Lazar, A. Ignat, D. Petreus, and R. Etz, ”Energy Management for an Islanded Microgrid Based on Harmony Search Algorithm,“ in 2018 41st International Spring Seminar on Electronics Technology (ISSE), 2018, pp. 1–6. 38. Zhang, J., Wu, Y., Guo, Y., Wang, B., Wang, H., & Liu, H. (2016). A hybrid harmony search algorithm with differential evolution for day-ahead scheduling problem of a microgrid with consideration of power flow constraints. Applied Energy, 183, 791–804. 39. Y. Li, W. Dong, X. Feng, and M. Zhang, ”The Expansion Planning of Wind-Thermal Cogeneration System Based on Harmony Search Algorithm under Smart Grid,“ in 2019 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids (SmartGridComm), 2019, pp. 1–5. 40. Sedhom, B. E., El-Saadawi, M. M., Hatata, A. Y., & Alsayyari, A. S. (2020). Hierarchical control technique-based harmony search optimization algorithm versus model predictive control for autonomous smart microgrids. International Journal of Electrical Power & Energy Systems, 115, 105511. 41. Rajabioun, R. (2011). Cuckoo optimization algorithm. Applied Soft Computing, 11, 5508– 5518. 42. J. Singh, S. Poddar, S. P. Ramalingam, P. K. Shanmugam, and A. Kalam, ”investigation on dynamic economic dispatch problem of micro grid using cuckoo search algorithmgrid connected and islanded mode,“ in TENCON 2019–2019 IEEE Region 10 Conference (TENCON), 2019, pp. 1886-1891. 43. D. Arcos-Aviles, G. García-Gutièrrez, F. Guinjoan, E. V. Carrera, J. Pascual, P. Ayala, et al., ”Adjustment of the Fuzzy Logic controller parameters of the energy management strategy of a grid-tied domestic electro-thermal microgrid using the Cuckoo search algorithm,“ in IECON 2019-45th Annual Conference of the IEEE Industrial Electronics Society, 2019, pp. 279–285. 44. K. Tantrapon, P. Jirapong, and P. Thararak, ”Optimal Operation of Battery Energy Storage System for Mitigating Voltage Fluctuation in Microgrid Using Cuckoo Search Optimization,“ in 2019 16th international conference on electrical engineering/electronics, computer, Telecommunications and Information Technology (ECTI-CON), 2019, pp. 877–880. 45. S. Vasanthakumar, N. Kumarappan, R. Arulraj, and T. Vigneysh, ”Cuckoo search algorithm based environmental economic dispatch of microgrid system with distributed generation,“ in 2015 International Conference on Smart Technologies and Management for Computing, Communication, Controls, Energy and Materials (ICSTM), 2015, pp. 575–580.
248
A. A. Ghavifekr
46. Kumar, J. S., Raja, S. C., Nesamalar, J. J. D., & Venkatesh, P. (2018). Optimizing renewable based generations in AC/DC microgrid system using hybrid Nelder-Mead–cuckoo search algorithm. Energy, 158, 204–215. 47. Karaboga, D. (2010). Artificial bee colony algorithm. scholarpedia, 5, 6915. 48. P. Chaweewat, J. G. Singh, W. Ongsakul, and A. K. Srivastrava, ”Economic and environmental impact assessment with network reconfiguration in microgrid by using artificial bee colony,“ in 2016 International Conference on Cogeneration, Small Power Plants and District Energy (ICUE), 2016, pp. 1–7. 49. L. Ciabattoni, F. Ferracuti, G. Ippoliti, and S. Longhi, ”Artificial bee colonies based optimal sizing of microgrid components: A profit maximization approach,“ in 2016 IEEE Congress on Evolutionary Computation (CEC), 2016, pp. 2036-2042. 50. M. M. Moghaddam, M. Marzband, and F. Azarinejadian, ”Optimal energy management for a home Microgrid based on multi-period artificial bee colony,“ in 2017 Iranian Conference on Electrical Engineering (ICEE), 2017, pp. 1446–1451. 51. Mirjalili, S., Mirjalili, S. M., & Lewis, A. (2014). Grey wolf optimizer. Advances in Engineering Software, 69, 46–61. 52. Miao, D., & Hossain, S. (2020). Improved gray wolf optimization algorithm for solving placement and sizing of electrical energy storage system in micro-grids. ISA Transactions. 53. Y. Cao, T. Chen, X. Qing, J. Zhang, and L. Sun, ”Energy Management for a Microgrid With Different Charging and Discharging Priorities of Batteries Using Modified Grey Wolf Optimizer,“ in 2019 IEEE PES Asia-Pacific Power and Energy Engineering Conference (APPEEC), 2019, pp. 1–5. 54. Zhang, J., Wang, X., & Ma, L. (2019). An optimal power allocation scheme of microgrid using Grey wolf optimizer. IEEE Access, 7, 137608–137619. 55. Yang, X.-S. (2008). Firefly algorithm. Nature-inspired Metaheuristic Algorithms, 20, 79–90. 56. J. D. Vasanth, N. Kumarappan, R. Arulraj, and T. Vigneysh, ”Minimization of operation cost of a microgrid using firefly algorithm,“ in 2017 IEEE International Conference on Intelligent Techniques in Control, Optimization and Signal Processing (INCOS), 2017, pp. 1–6. 57. M. Abdullah, N. L. Abdullah, and J. Jamian, ”Optimal power generation in microgrid system using firefly algorithm,“ in 2017 6th International Conference on Electrical Engineering and Informatics (ICEEI), 2017, pp. 1–6. 58. R. Khan, N. Gogoi, J. Barman, A. Latif, and D. C. Das, ”Virtual Power Plant enabled Coordinated Frequency Control of a grid connected Independent Hybrid Microgrid using Firefly algorithm,“ in 2019 IEEE Region 10 Symposium (TENSYMP), 2019, pp. 795–800. 59. M. Akbari, M. Golkar, and S. M. Tafreshi, ”Firefly algorithm-based voltage and frequency control of a hybrid ac-dc microgrid,“ in 2012 Proceedings of 17th Conference on Electrical Power Distribution, 2012, pp. 1–7. 60. S. Sun, J. Ni, Y. Cheng, P. Yu, Z. Xin, and Q. Ai, ”Firefly algorithm based dynamic coalition formation strategy in micro-grids,“ 2015. 61. Satapathy, P., Dhar, S., & Dash, P. K. (2017). Stability improvement of PV-BESS diesel generator-based microgrid with a new modified harmony search-based hybrid firefly algorithm. IET Renewable Power Generation, 11, 566–577. 62. Saremi, S., Mirjalili, S., & Lewis, A. (2017). Grasshopper optimisation algorithm: Theory and application. Advances in Engineering Software, 105, 30–47. 63. A. Annamraju and S. Nandiraju, ”Frequency Control in an Autonomous Two-area Hybrid Microgrid using Grasshopper Optimization based Robust PID Controller,“ in 2018 8th IEEE India International Conference on Power Electronics (IICPE), 2018, pp. 1–6. 64. Barik, A. K., & Das, D. C. (2018). Expeditious frequency control of solar photovoltaic/biogas/biodiesel generator based isolated renewable microgrid using grasshopper optimisation algorithm. IET Renewable Power Generation, 12, 1659–1667. 65. Bukar, A. L., Tan, C. W., & Lau, K. Y. (2019). Optimal sizing of an autonomous photovoltaic/wind/battery/diesel generator microgrid using grasshopper optimization algorithm. Solar Energy, 188, 685–696.
9 Application of Heuristic Techniques and Evolutionary Algorithms. . .
249
66. Mirjalili, S., & Lewis, A. (2016). The whale optimization algorithm. Advances in Engineering Software, 95, 51–67. 67. S. H. Qazi, M. Uqaili, and U. Sultana, ”Whales Optimization Algorithm Based Enhanced Power Controller for an Autonomous Microgrid System,“ in 2019 8th International Conference on Modern Power Systems (MPS), 2019, pp. 1–7. 68. I. N. Trivedi, M. Bhoye, R. Bhesdadiya, P. Jangir, N. Jangir, and A. Kumar, ”An emission constraint environment dispatch problem solution with microgrid using Whale Optimization Algorithm,“ in 2016 National Power Systems Conference (NPSC), 2016, pp. 1–6. 69. J. R. Koza, ”Survey of genetic algorithms and genetic programming,“ in Wescon conference record, 1995, pp. 589–594. 70. Yamashita, D. Y., Vechiu, I., & Gaubert, J.-P. (2020). A review of hierarchical control for building microgrids. Renewable and Sustainable Energy Reviews, 118, 109523. 71. Askarzadeh, A. (2017). A memory-based genetic algorithm for optimization of power generation in a microgrid. IEEE Transactions on Sustainable Energy, 9, 1081–1089. 72. V. Guliashki and G. Marinova, ”An Accelerated Genetic Single Objective Algorithm for Optimization of Energy Flows in Microgrids,“ in 2018 25th international conference on systems, Signals and Image Processing (IWSSIP), 2018, pp. 1–5. 73. Moghaddam, A. A., Seifi, A., & Niknam, T. (2012). Multi-operation management of a typical micro-grids using particle swarm optimization: A comparative study. Renewable and Sustainable Energy Reviews, 16, 1268–1281. 74. M. Bhoye, S. N. Purohit, I. N. Trivedi, M. H. Pandya, P. Jangir, and N. Jangir, ”Energy management of renewable energy sources in a microgrid using cuckoo search algorithm,“ in 2016 IEEE Students’ Conference on Electrical, Electronics and Computer Science (SCEECS), 2016, pp. 1–6. 75. M. De, G. Das, S. Mandal, and K. Mandal, ”A reliable energy management and generation scheduling model in microgrids using modified cuckoo search algorithm,“ in 2018 Emerging Trends in Electronic Devices and Computational Techniques (EDCT), 2018, pp. 1–6. 76. M. Modiri-Delsha, S. S. Taheri, S. J. Seyed-Shenava, and N. Rahim, ”Optimal generation scheduling in microgrids by cuckoo search algorithm,“ 2014. 77. N. K. Paliwal, N. K. Singh, and A. K. Singh, ”Optimal power flow in grid connected microgrid using artificial bee colony algorithm,“ in 2016 IEEE Region 10 Conference (TENCON), 2016, pp. 671–675. 78. M. Govardhan and R. Roy, ”Artificial Bee Colony based optimal management of microgrid,“ in 2012 11th International Conference on Environment and Electrical Engineering, 2012, pp. 334–339. 79. J. Yi and D. Shi, ”Parameter Optimization of DOCR and FCL in Microgrids Based on Grey Wolf Optimizer,“ in 2018 53rd International Universities Power Engineering Conference (UPEC), 2018, pp. 1–5. 80. M. Moazzami, M. Ghanbari, H. Shahinzadeh, J. Moradi, and G. B. Gharehpetian, ”Application of multi-objective grey wolf algorithm on energy management of microgrids with techno-economic and environmental considerations,“ in 2018 3rd Conference on Swarm Intelligence and Evolutionary Computation (CSIEC), 2018, pp. 1–9. 81. Mohammadi, S., Mozafari, B., Solimani, S., & Niknam, T. (2013). An adaptive modified firefly optimisation algorithm based on Hong’s point estimate method to optimal operation management in a microgrid with consideration of uncertainties. Energy, 51, 339–348. 82. D.-O. Sidea, L. Toma, M. Sanduleac, I.-I. Picioroaga, and V.-A. Boicea, ”Optimal BESS Scheduling Strategy in Microgrids Based on Genetic Algorithms,“ in 2019 IEEE Milan PowerTech, 2019, pp. 1–6. 83. Mahmoud, T. S., Ahmed, B. S., & Hassan, M. Y. (2019). The role of intelligent generation control algorithms in optimizing battery energy storage systems size in microgrids: A case study from Western Australia. Energy Conversion and Management, 196, 1335–1352. 84. Kerdphol, T., Qudaih, Y., & Mitani, Y. (2016). Optimum battery energy storage system using PSO considering dynamic demand response for microgrids. International Journal of Electrical Power & Energy Systems, 83, 58–66.
250
A. A. Ghavifekr
85. Kerdphol, T., Fuji, K., Mitani, Y., Watanabe, M., & Qudaih, Y. (2016). Optimization of a battery energy storage system using particle swarm optimization for stand-alone microgrids. International Journal of Electrical Power & Energy Systems, 81, 32–39. 86. F. Tooryan, E. R. Collins, A. Ahmadi, and S. S. Rangarajan, ”Distributed generators optimal sizing and placement in a microgrid using PSO,“ in 2017 IEEE 6th International Conference on Renewable Energy Research and Applications (ICRERA), 2017, pp. 614–619. 87. Elattar, E. E. (2018). Modified harmony search algorithm for combined economic emission dispatch of microgrid incorporating renewable sources. Energy, 159, 496–507. 88. S. Sukumar, M. Marsadek, A. Ramasamy, and H. Mokhlis, ”Grey Wolf Optimizer Based Battery Energy Storage System Sizing for Economic Operation of Microgrid,“ in 2018 IEEE International Conference on Environment and Electrical Engineering and 2018 IEEE Industrial and Commercial Power Systems Europe (EEEIC/I&CPS Europe), 2018, pp. 1–5. 89. Sharma, S., Bhattacharjee, S., & Bhattacharya, A. (2016). ”Grey wolf optimisation for optimal sizing of battery energy storage device to minimise operation cost of microgrid,“ IET Generation. Transmission and Distribution, 10, 625–637. 90. P. C. Nayak, S. Bisoi, R. C. Prusty, and S. Panda, ”Performance Analysis of PDF+ (1+ PI) Controller for Load Frequency Control of the Multi Microgrid System Using Genetic Algorithm,“ in 2019 International Conference on Information Technology (ICIT), 2019, pp. 448–453. 91. R. Wang, S. Wu, C. Wang, S. An, Z. Sun, W. Li, et al., ”optimized Operation and Control of Microgrid based on Multi-objective Genetic Algorithm,“ in 2018 International Conference on Power System Technology (POWERCON), 2018, pp. 1539–1544. 92. S. F. Eilaghi, A. Ahmadian, and M. A. Golkar, ”Optimal voltage unbalance compensation in a microgrid using PSO algorithm,“ in 2016 IEEE 7th Power India International Conference (PIICON), 2016, pp. 1–6. 93. Al-Saedi, W., Lachowicz, S. W., Habibi, D., & Bass, O. (2013). Voltage and frequency regulation based DG unit in an autonomous microgrid operation using particle swarm optimization. International Journal of Electrical Power & Energy Systems, 53, 742–751. 94. D. K. Lal, A. K. Barisal, and M. Tripathy, ”Load frequency control of multi area interconnected microgrid power system using grasshopper optimization algorithm optimized fuzzy PID controller,“ in 2018 Recent Advances on Engineering, Technology and Computational Sciences (RAETCS), 2018, pp. 1–6. 95. Jumani, T. A., Mustafa, M. W., Md Rasid, M., Mirjat, N. H., Leghari, Z. H., & Saeed, M. S. (2018). Optimal voltage and frequency control of an islanded microgrid using grasshopper optimization algorithm. Energies, 11, 3191. 96. H. Li, C. Zang, P. Zeng, H. Yu, and Z. Li, ”A genetic algorithm-based hybrid optimization approach for microgrid energy management,“ in 2015 IEEE International Conference on Cyber Technology in Automation, Control, and Intelligent Systems (CYBER), 2015, pp. 1474– 1478. 97. P. P. Vergara, R. Torquato, and L. C. Da Silva, ”Towards a real-time Energy Management System for a Microgrid using a multi-objective genetic algorithm,“ in 2015 IEEE Power & Energy Society General Meeting, 2015, pp. 1–5. 98. C. Changsong, D. Shanxu, C. Tao, L. Bangyin, and Y. Jinjun, ”Energy trading model for optimal microgrid scheduling based on genetic algorithm,“ in 2009 IEEE 6th International Power Electronics and Motion Control Conference, 2009, pp. 2136–2139. 99. Kirthiga, M. V., Daniel, S. A., & Gurunathan, S. (2012). A methodology for transforming an existing distribution network into a sustainable autonomous micro-grid. IEEE Transactions on Sustainable Energy, 4, 31–41. 100. A. Singh, ”GA optimized PID controller for frequency regulation in standalone AC microgrid,“ in 2016 7th India International Conference on Power Electronics (IICPE), 2016, pp. 1–5. 101. Li, P., Xu, D., Zhou, Z., Lee, W.-J., & Zhao, B. (2015). Stochastic optimal operation of microgrid based on chaotic binary particle swarm optimization. IEEE Transactions on Smart Grid, 7, 66–73.
9 Application of Heuristic Techniques and Evolutionary Algorithms. . .
251
102. M. A. Djema and M. Boudour, ”Autonomous microgrid power management using Particle Swarm Optimization,“ in 2016 International Conference on Control, Decision and Information Technologies (CoDIT), 2016, pp. 222–226. 103. Oboudi, M., Hooshmand, R., & Karamad, A. (2017). A feasible method for controlled intentional islanding in microgrids based on PSO algorithm. Swarm and Evolutionary Computation, 35, 14–25. 104. Y. Du, D. Gong, K. Ding, R. Chen, H. Wang, W. Xi, et al., ”Research on Autonomous Frequency Modulation and Voltage Regulation Method of Microgrid Based on Immunity Selection-particle Swarm Optimization (ISPSO),“ in 2019 IEEE PES Asia-Pacific Power and Energy Engineering Conference (APPEEC), 2019, pp. 1–5. 105. M. Akbari, M. Golkar, and S. Tafreshi, ”A PSO solution for improved voltage stability of a hybrid ac-dc microgrid,“ in ISGT2011-India, 2011, pp. 352–357. 106. C. Colson, M. Nehrir, and C. Wang, ”Ant colony optimization for microgrid multi-objective power management,“ in 2009 IEEE/PES Power Systems Conference and Exposition, 2009, pp. 1–7. 107. Li, P., Li, R.-X., Cao, Y., Li, D.-Y., & Xie, G. (2017). Multiobjective sizing optimization for island microgrids using a triangular aggregation model and the levy-harmony algorithm. IEEE Transactions on Industrial Informatics, 14, 3495–3505. 108. Sedhom, B. E., El-Saadawi, M. M., Elhosseini, M. A., Saeed, M. A., & Abd-Raboh, E. E. (2019). A harmony search-based H-infinity control method for islanded microgrid. ISA Transactions. 109. Marzband, M., Azarinejadian, F., Savaghebi, M., & Guerrero, J. M. (2015). An optimal energy management system for islanded microgrids based on multiperiod artificial bee colony combined with Markov chain. IEEE Systems Journal, 11, 1712–1722. 110. O. Penangsang, A. Soeprijanto, A. A. Fitriana, and E. S. Ningrum, ”Operation optimization stand-alone microgrid using firefly algorithm considering lifetime characteristics of battery,“ in 2016 International Seminar on Intelligent Technology and Its Applications (ISITIA), 2016, pp. 565–570. 111. Satapathy, P., Dhar, S., & Dash, P. (2016). A mutated hybrid firefly approach to mitigate dynamic oscillations of second order PLL based PV-battery system for microgrid applications. Sustainable Energy Technologies and Assessments, 16, 69–83.
Part II
Control of Microgrids
Chapter 10
Conventional Droop Methods for Microgrids Kwang Woo Joung and Jung-Wook Park
10.1 Introduction Recently, renewable energy sources (RESs) have been connected worldwide to power grids in the form of distributed generation (DG) because of environmental and economic reasons. In order to efficiently operate and control DGs in the decentralized small-scale power grid, the concept of a microgrid is used. In a microgrid, the hierarchical control system is required to control various types of generators in parallel. Then, the droop control is widely used for the most local controller, which is also called as primary control [1–5]. In AC microgrid, the conventional droop control is divided into the frequency and voltage droop controls [6, 7]. The characteristic of P–f droop control is based on the relationship between the active power from the conventional synchronous generator and the frequency of the system. In other words, if the amount of P from the DGs is suddenly increased, f is reduced and vice versa. Similarly, the characteristic of Q–V droop control is applied for restoring V to its nominal value by providing the negative Q when the value of V drops. The synchronous generator has these droop features inherently by rotational inertia. To improve its control performance, additional droop control methods are often used. However, the RESs, which are usually based on the inverter, have no inertia. Without grid supporting controls, they will generate constant power output even when the frequency and voltage of the system are changed [8, 9]. Therefore, the droop control is required for the RESs to maintain a stable operation of microgrid while having the effect of inertia [10]. On the other hand, the grid frequency is not considered in DC microgrid, and there is no need for DGs to be synchronized. Thus,
K. W. Joung · J.-W. Park () School of Electrical & Electronic Engineering, Yonsei University, Seoul, South Korea e-mail: [email protected] © Springer Nature Switzerland AG 2021 A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems, https://doi.org/10.1007/978-3-030-59750-4_10
255
256
K. W. Joung and J.-W. Prank
in DC microgrid, the voltage becomes the only major factor in implementing the droop control [11–13]. In general, the voltage–current (V–I) droop control method, which is based on the relationship between line resistance and current, is used. The generators with droop control in both AC and DC microgrids have the following advantages. • The parallel operation among the DGs is achieved. • The power allocation is possible by considering the capacity or characteristics of DGs. • The DG can be controlled as a grid supporting unit enhancing the reliability of the microgrid. • No communication lines are needed. In this chapter, the conventional droop control methods used in microgrids are firstly described. They can be implemented for generators in AC microgrid with the “self-synchronizing” characteristic of a synchronous generator. However, particularly for inverter-based renewables (IBRs) among the RESs, their power outputs can be varied according to environmental conditions. For example, the power of the wind turbine generator and the photovoltaic generator will vary according to wind speed and irradiance, respectively. Therefore, to support a grid with many IBRs is difficult because their maximum power output depends on the external environment. This makes the energy storage system (ESS) necessary in microgrid since it can reduce the variability of RES as a control unit for enhancing the flexibility and stability of microgrid. The main advantage of ESS is that it can mitigate the power fluctuation faster than conventional synchronous generators. Also, unlike other generators, the ESS enables absorbing the active power from the grid by its charging operation. By considering several different features of DGs, the power sharing among DGs will be successfully achieved when the suitable droop methods are applied to each type of generators. Furthermore, to use the IBRs in the DC system, the conventional DC droop control methods can be applied. In this case, the droop coefficient will act as a “virtual resistance” to imitate the voltage drop caused by actual resistance. The mathematical analysis and simulation study are given below for both AC and DC microgrids to verify the effectiveness of existing droop control methods.
10.2 Conventional Droop Method for AC Microgrid 10.2.1 Mathematical Analysis The generation and consumption of active powers must be balanced to maintain the grid frequency within its proper ranges. When this power balance in a microgrid is
10 Conventional Droop Methods for Microgrids
257
broken, the powers from some generators will be increased (or decreased) abruptly. Then, the rotating speed of generators is determined as. 2H d 2 δ 2H dω = Pm − Pe = 2 ωs dt ωs dt
(10.1)
where H is the inertia constant. δ is the power angle of the generator. ωs is the grid frequency (at synchronous speed). ω is the grid frequency. Pm and Pe are the mechanical and electrical powers of the generator, respectively. In Eq. (10.1), whenever the values of Pm and Pe are not the same, the derivative of ωm changes. This means that ω has a deviation from ωs . After the power balance of the microgrid is broken, the outputs from each generator will change until the new equilibrium state is achieved. For example, assume that one of the generators in a microgrid is suddenly disconnected. Then, Pm of the entire system is decreased, and the derivative of ω becomes negative while resulting in the reduction of the grid frequency. Thereafter, the other generators will increase their power outputs to mitigate the power deviations. The droop control is widely used to implement this physical characteristic in order to control synchronous generators more precisely. In other words, the power output deviation of the generator can be determined by using the frequency deviation as. fi − f ∗ = −Rp−f · Pi − P ∗
(10.2)
where fi and f* are the actual and nominal frequencies of ith generator, respectively. Pi and P* are the actual and nominal active power outputs from ith generator, respectively. RP–f is the droop constant of P–f droop control. In general, the droop constant is set by considering the physical limitation of the generator in its output power variations. From Eq. (10.2), it can be shown that the deviations of both active power and grid frequency have a negative proportional relationship. The characteristic of such a droop control method is shown in Fig. 10.1. Assume that the reference values of grid frequency and active power output in a particular operating point are f1 and P1 , respectively. When the grid frequency is reduced from f1 to f2 , the output from the generator will change from P1 to P2 according to its droop characteristic. For the microgrid with the high renewable penetration, however, this droop control method is difficult to be applied. This is because many RESs are controlled
258
K. W. Joung and J.-W. Prank
f Storage P0
RP – f
f1 f2
– Pmax
P1
P2 P max
P
Fig. 10.1 Characteristic of P–f droop control method
to generate the active power close to their power output limit, Pmax . If the droop control of Eq. (10.2) is used, the deviation of grid frequency can be calculated by the droop constant of the entire system as. Δfsys = −Rsys · ΔP
Rsys =
#n i=1 Ri × Pi,cap #n i=1 Pi,cap
(10.3)
(10.4)
where fsys is the deviation of grid frequency for the entire microgrid system. P is the deviation of active power generation caused by a disturbance. Rsys is the droop constant of the entire microgrid system. Ri is the droop constant of ith generator. Pi,cap is the capacity of ith generator. The value of Rsys in Eq. (10.4) is affected by the operating status of RESs, which can generate more outputs only if they have enough reserve power [6]. In other words, the droop control can be only applied to the RESs, which generate less power when compared to their maximum power point (MPP). After a disturbance, the power deviation is mitigated by generators with the droop control according to Eq. (10.2). If the power output of one RES is increased to its maximum (or conversely if its maximum power is reduced due to the change of environmental condition), the RES is unable to participate in mitigating the power deviation anymore because it only generates the constant maximum power. After at least one RES turns into uncontrollable, Eq. (10.3) can be changed as.
10 Conventional Droop Methods for Microgrids
259
Fig. 10.2 The reduction of droop constant due to the change in the operating status of RES
# Δfsys =
⎞ Ri × Pi,cap ⎛ ⎝Pi − # ΔPi,m ⎠ + Δfm Pi,cap
i=G,drp
(10.5)
i=G,drp
i=G,drp
where Pi,m is the deviation of active power before at least one RES becomes uncontrollable. fm is the deviation of grid frequency before at least one RES becomes uncontrollable. The subscript, G, drp indicates the generator, which can be controlled by the droop control method. From Eq. (10.5) the droop constant for the entire microgrid will be reduced after the RES is unable to control, as illustrated in Fig. 10.2. This usually happens to the RESs because their reserve powers are not always determined precisely. Similar to grid frequency, to maintain the voltage stable is also important. Note that the reactive power is generated depending on the voltage difference between buses as.
Q=
V12 V1 · V2 · cos (δ) X12 X12
where Q is the reactive power from the generator. V1 and V2 are the magnitudes of voltage at buses, 1 and 2, respectively. X12 is the line impedance between buses, 1 and 2.
(10.6)
260
K. W. Joung and J.-W. Prank
It is known from Eq. (10.6) that the voltage can be regulated by controlling the reactive power. That is, as the reactive power output becomes greater, the bus voltage is increased more. When the generations from RESs are increased without any voltage control, it might cause the voltage stability problem. In particular, it becomes serious in microgrid because the distance among generators is relatively close when compared to a large-scale power system. Then, this voltage problem can be solved by controlling the reactive power deviation, which is inversely proportional to the voltage deviation as. Vi − V ∗ = −RQ−V · Qi − Q∗
(10.7)
where Vi and V* are the actual and nominal voltages of ith generator, respectively. Qi and Q* are the actual and nominal reactive power outputs from ith generator, respectively. RQ–V is the droop constant of Q–V droop control. The performance of power sharing depends on the droop constant. The use of a high droop constant might still cause to fluctuate the voltage. In general, it is set by considering the maximum reactive power output from the generator and the maximum voltage variations at the bus as.
RQ−V =
Qmax − Qmin Vhigh − Vlow
(10.8)
The characteristic of Q–V droop control method by Eq. (10.7) is shown in Fig. 10.3. Assume that the reference values of bus voltage and reactive power output in a particular operating point are V1 and Q1 , respectively. When the bus voltage is reduced from V1 to V2 , the generator increases the reactive power output from Q1 to Q2 according to its droop characteristic. The combined P–f and Q–V droop control method is shown in Fig. 10.4. The gray area (circle) represents the operating range of the generator with the apparent power, S. The shapes of this area are different depending on the type of generator.
10.2.2 Synchronous Generator In AC microgrid, the diesel synchronous generator is usually used because of its fast ramp-up capability, reliability, durability, and so on. It is the dispatchable source, which can fully control the active and reactive powers with its governor and exciter, respectively. For its governor control, the model of “DEGOV1” as shown in Fig. 10.5 can be used to carry out its P–f droop control. It can be implemented through
10 Conventional Droop Methods for Microgrids
261
V Inductive Q0 V1
RQ – V
V2
–Qmax
Q1
Q
Q2
Qmax
Fig. 10.3 Characteristic of Q–V droop control method
Q
Q
Qmax
Q* Capacitive
V
*
P V
Inductive
–Qmax
f
f* Storage
– Pmax
Genearation
P * Pmax
P
Fig. 10.4 Operating range of generator by the combined P–f and Q–V droop control method in AC microgrid
the droop constant, RP–f in practice. Typically, the value of RP–f is set to 5% by considering the ramp-up speed of power from the diesel generator. On the other hand, the Q–V droop control is implemented by adding the virtual line impedance (the generator will control the virtual bus voltage instead of real terminal voltage). If the value of virtual line impedance is negative, the terminal voltage is compensated
262
K. W. Joung and J.-W. Prank
f
Actuator
f
*
P
*
Tmax
Electric control box
Engine
– (1 + sT3 ) 1 + sT1 + s 2T2T1
K (1 + sT4 ) s (1 + sT5 ) + (1 + sT6 )
e – sTD
P
Tmin Droop Control
RP – f
Fig. 10.5 Governor model of diesel generator (DEGOV1) Comp
V
I
Vc = V – j . X e . I
IEEEVC
Vc
V
I
Vc = V + ( Rc + j . X c ) . I
(a)
Vc
(b)
Fig. 10.6 Voltage droop control model (a) Comp (b) IEEEVC
by controlling the reactive power output. In other words, the capacitive reactive power is released when the terminal voltage is decreased. In contrast, when it is increased, the inductive reactive power from the generator is provided. The model “Comp” or “IEEEVC” in Fig. 10.6 are used to carry out the Q–V droop control. The maximum value of reactive power is determined by the power factor of output from a diesel generator. Typically, the droop constant, RQ–V is set to 4 ~ 6%. The gray area in Fig. 10.7 represents the operating range of diesel generators. Because the diesel generator cannot generate negative active power, its operating range is limited to the right hemisphere area in Fig. 10.4. Also, the diesel generator has some operating power reserve such that the value of output reference is reduced by 5% from the maximum power output. Moreover, the idle operation (or generating too small output) must be avoided because of its poor operating efficiency. Finally, the other physical limits such as field current, armature heating, winding end region heating, and under excitation may exist [7].
10.2.3 Renewable Energy Sources The power output from RES is very sensitive to environmental conditions. It can be divided into two types, which are the IBR and non-IBR depending on whether it is connected to the microgrid through an inverter or not. The IBR is difficult to control because the RES must be synchronized with the main grid. In particular, this becomes the major issue when the penetration ratio of RESs to
10 Conventional Droop Methods for Microgrids
263
Q Field current limit
Qmax,cap
Armature heating constraints
P V
Qmax,ind
Winding end region heating limit Under excitation limit
Pmin
Pmax
Fig. 10.7 Operating range of diesel generator
microgrid is very high. On the contrary, the non-IBR such as hydroelectric, which is physically synchronized with the grid, is easy to control because it is identical to the synchronous generator.
10.2.3.1
IBR
The replacement of diesel generators with the RESs in the form of IBR reduces the system inertia because they do not have any rotational inertia. For example, the photovoltaic (PV) generator uses a DC–AC inverter, and the wind turbine (WT) generator requires a back-to-back converter for connecting it to the microgrid. Then, the power generations from such IBRs are subject to makes the system unstable, when the unbalance between power supply and demand occurs. Therefore, the proper control method for grid support must be applied to keep increasing the penetration level of IBRs. The structure of the power controller applied to the IBR is shown in Fig. 10.8. The output frequency is synchronized by the phase-locked loop (PLL) with the frequency of the main microgrid, and it is first measured. Then, the external power control loop is used to control P and Q to their reference values with its droop control. In other words, it changes P and Q properly based on its droop characteristics. In normal operation, the active power output of IBR is determined by the maximum power point tracking (MPPT) control method. However, when this droop control method is applied to the IBR, the power deviation will be added to
264
K. W. Joung and J.-W. Prank
Inverter Vabc I abc
Inner loops f V
Voltage Control Loop
Droop Control
Current Control Loop
P
Q
PWM
Power Calculation
Fig. 10.8 Structure of control block diagram for IBR RES Characteristics Environmental factors
P*
P
f*
f Droop f
P P
Q*
V*
V Droop V
Q Q
Fig. 10.9 Structure of droop control for IBR
the MPPT signal, as shown in Fig. 10.9. Next, the inner loops with the voltage and current controls are used to regulate the output from the inverter to the d and q-axes reference signals. However, there are many constraints to operate IBRs with droop control in practice. For example, for the P–f droop control, it is difficult to allocate the reserve power to each IBR. As mentioned earlier, MPPT control is mainly used for the IBR due to economic reasons. In this case, the active power from IBR cannot be increased even when the grid frequency is reduced, and therefore the additional active power is required. In other words, it can be only decreased at MPP. Therefore, in order to fully apply the P–f droop control, the IBR must be capable of adjusting
10 Conventional Droop Methods for Microgrids
265
Q Power Factor
Qmax,cap P V
Qmax,ind
Pmin
PMPPT determined by environmental factors
Fig. 10.10 Operating range of IBR
the reference signal such that it can reduce its maximum output while obtaining the reserve power. Also, the IBR is not dispatchable because of the continuous change of MPP. This problem can be solved by forecasting the environmental condition, calculating the probability of power output from RES, and compensating the power fluctuations by the ESS. The reactive power control of IBR is more easily achieved than that of a diesel generator. It is possible to generate its reactive power at its related value simply by the inverter control because it has less physical constraint than a diesel generator. Therefore, it enables to supply the reactive power even when the active power cannot be generated. The Q–V droop control method is easy to be implemented for the IBR. However, the reactive power from IBR must be limited depending on the amount of active power and power factor according to the system condition [8]. The operating range of IBR is shown in Fig. 10.10.
10.2.3.2
Non-IBR
The control of non-IBR is similar to that of diesel generator. For example, the hydro turbine can be connected to a microgrid, and it is synchronized with the grid frequency. Also, it has rotational inertia like a diesel generator. Therefore, the nonIBR can be connected to the grid without considering the careful grid code. Its operating range is the same as that of a diesel generator, as shown in Fig. 10.7.
266
K. W. Joung and J.-W. Prank
10.2.4 Energy Storage System The ESS is important to improve the controllability of microgrid. As mentioned previously, the ESS can be linked with IBR to reduce the variability and uncertainty of output depending on the weather condition. In an emergency, it can be also served as an effective reserve power source by injecting the additional active power to the grid quickly. This type of ESS is called as the renewable integration (RI)ESS. On the other hand, the ESS can be used for frequency regulation (FR). In this case, it operates directly to satisfy the requirement of a microgrid. This type of FR-ESS plays an important role in mitigating power imbalance, and therefore improving the flexibility and stability of microgrid with its charging and discharging operations. The stored energy in ESS is released when the power generated is less than consumption. Otherwise, the surplus energy is saved to the ESS from the grid. Moreover, the ESS has a great frequency control capability, similarly to the IBR for voltage regulation. In other words, it can enhance the stability of the microgrid because it has a fast ramp rate than the diesel generator [9]. This property can be applied to the droop control method in Eqs. (10.2), and (10.7) by setting the droop constant more sensitive than that of other generators. However, the tradeoff between the stability of the system and dependency on the ESS must be also considered. That is, if the droop constant of ESS is very different from that of other generators, most of the power imbalance in the microgrid must be mainly solved by the ESS. Therefore, it is important to set the appropriate droop constant by considering the generation profile of the microgrid. Furthermore, the ESS might be unable to participate in supporting the microgrid with its droop control method if the amount of stored energy is not enough. Thus, the additional control is required to regulate the state-of-charge (SOC) properly while guaranteeing the reliable and safe operation of ESS. Assuming that the SOC is wellregulated, the ESS can be operated in the widest range when compared to all types of generators used in a microgrid, as shown in Fig. 10.11.
10.2.5 Frequency and Voltage Responses To evaluate the performance of droop control, the case studies are carried out by ® using the DIgSILENT PowerFactory software. The conventional droop control method is implemented in the AC microgrid shown in Fig. 10.12. It has one diesel generator, two IBRs, one ESS, and six loads. The diesel generator of 300 kW operates as a slack bus with a droop constant of 5%. Although the rated power of IBR1 (PV generator) is 300 kW, it is assumed that it generates the real power of 250 kW by its MPPT control. Also, the rated power of IBR2 (wind turbine generator) is still 300 kW. However, its actual power output is assumed to be about 200 kW without its MPPT control to have the reserve power
10 Conventional Droop Methods for Microgrids
267
Q Power Factor
Qmax,cap P V
Qmax,ind Pmax,discharge
Pmax,charge
Fig. 10.11 Operating range of ESS Fig. 10.12 AC microgrid with one diesel generator, two IBRs, and one ESS
Main Grid
ESS1
Diesel
L1
L2
L3
IBR1
L4
L5
IBR2
L6
for an emergency. Finally, the ESS of 300 kWh operates in an idle mode unless a disturbance occurs. The total amount of load demand is 600 kW. For Case 1, the amount of P and Q in the load, L5 is increased by 100 kW and 45 kVAR, respectively, at 1 s. The P–f and Q–V droop constants of diesel generator, IBR1, and IBR2 are set to 5%. Because the ESS is used only to regulate the frequency, it has the steeper droop constant, which is 2%. The results are shown
268
K. W. Joung and J.-W. Prank
Fig. 10.13 Results of Case 1 with the ESS when the load, L5 is increased at 1 s: (a) frequency, (b) bus voltage, (c) active power, (d) reactive power
in Fig. 10.13. It is observed that the frequency drops to 59.72 Hz at maximum, and is restored to 59.78 Hz. After the load is increased at 1 s, the active power outputs from a diesel generator, IBR2, and ESS are increased. Among them, the increased amount of active power from ESS is highest because it has the steepest droop constant. The IBR1 still generates the same amount of P, which is 250 kW, due to its MPPT control, as shown in Fig. 10.10. Also, the bus voltage of IBR2 is decreased, and its reactive power output is increased. In contrast, the bus voltage of IBR1 is increased, and its reactive power output is reduced after the load is changed. This is because their reactive powers are overcompensated by too high Q–V droop constant. Then, the remaining reactive power is compensated by the diesel generator since it is a slack generator. For Case 2, the system configuration and load event are the same as those of Case 1. However, it is assumed that the ESS does not operate because its SOC is in the emergency state. The results are shown in Fig. 10.14. It is clearly observed that the frequency stability of AC microgrid becomes worse without the operation of ESS. In other words, the frequency drops to 59.36 Hz at maximum, and is restored to 59.5 Hz. This is because the droop control effect of the overall AC microgrid becomes lower. However, the reactive power outputs are not overcompensated as
10 Conventional Droop Methods for Microgrids
269
Fig. 10.14 Results of Case 2 without the ESS when the load, L5 is increased at 1 s: (a) frequency, (b) bus voltage, (c) active power, (d) reactive power
the Q–V droop constant is reduced. Therefore, all voltages are decreased after the load is increased.
10.3 Conventional Droop Control Method in DC Microgrid 10.3.1 Mathematical Analysis The role of generators in DC microgrid is to supply the active power to loads and to regulate the DC bus voltage at the same time. To operate multiple generators together without any communication links, the droop control method is widely used. In the DC unit circuit shown in Fig. 10.15, the injected active power from the generator is defined by multiplying the injected current and bus voltage. Also, the amount of power injection to the bus, i is determined by the voltage difference between two buses as.
270
K. W. Joung and J.-W. Prank
Vj
Vi
Pi , I i
R Fig. 10.15 Simplified unit circuit between two DC buses
Pi = Ii · Vi Vi − Vj · Vi Pi = R
(10.9)
(10.10)
where Pi is the active power injected at ith bus. Ii is the DC current injected at ith bus. Vi is the DC voltage of ith bus. R is the resistance between the ith and jth buses. Based on the fact that the injected current (or the active power) is proportional to the voltage difference, the V–I droop control for the DC microgrid is carried out as. Vi = V ∗ − RV −I · Ii
(10.11)
where V* is the nominal voltage. RV–I is the droop constant of V–I droop control. The concept of this conventional V–I droop control method is to use the voltage drop property by adding a virtual resistance [12]. Then, the generator will inject more power to the grid when the magnitude of voltage is less than its nominal value, and vice versa. The characteristic of V–I droop control by Eq. (10.11) is shown in Fig. 10.16. Assume that the nominal values of bus voltage and current are V1 and I1 , respectively. According to the V–I droop control, the current is increased from I1 to I2 when the bus voltage is reduced from V1 to V2 .
10.3.2 Converter-Based Generator Unlike the AC microgrid, most converter-based generators (CBGs) are connected to the DC microgrid with DC–DC converter. Then, it may have a connection to
10 Conventional Droop Methods for Microgrids
271
V
V1 RV – I
V2
I
I1
I2
I max
Fig. 10.16 Characteristic of V–I droop method
AC Grid DC Load
PV
ESS
WT
Fig. 10.17 Example of DC microgrid
other AC systems, which require AC–DC inverters. However, most components such as PV, WT, ESS, and DC load are connected to the DC bus by using DC–DC converters, as illustrated in Fig. 10.17. The controller of CBG is shown in Fig. 10.18. The inner current control loop regulates the output current of the converter, which is determined by the voltage control loop. Then, the droop control method can be applied before the voltage control loop by adding the V–I droop characteristic to determine the reference signal
272
K. W. Joung and J.-W. Prank Inner loops
V*
Voltage Control Loop
Current Control Loop
I dc DC/DC Vdc Converter
V
I
Droop
Fig. 10.18 Structure of the control block diagram of CBG
Fig. 10.19 Results in DC microgrid without the V–I droop control: (a) active power and (b) bus voltage
of voltage. If there is some difference in output current, the voltage reference is adjusted by the V–I droop control, and it is provided to inner control loops. To verify the effectiveness of V–I droop control in the DC microgrid of Fig. 10.17, two case studies with and without droop control are carried out. It has three CBGs. Their capacities are all set to 300 kW. The amount of DC load demand is initially 200 kW, and it is increased from 200 kW to 300 kW at 1 s. When the DC microgrid operates by the CBGs without the V–I droop control, the results are shown in Fig. 10.19. The power required for the load is equally shared after the change of load. In other words, it is allocated among three CBGs according to the resistances between each generator and load, which are identical. Also, the bus voltage at the load bus is slightly decreased from 1 pu after the load is increased. Also, when it operates with the V–I droop control applied to the CBGs, the results are shown in Fig. 10.20. The ratio of power sharing among CBGs is determined mainly by their virtual resistances (or droop constants). The ratios of droop constant in CBG1, CBG2, and CBG3 are set to 1, 2, and 4, respectively. Thus, the CBG1 generates the greatest active power, and the CBG3 supplies the smallest active
10 Conventional Droop Methods for Microgrids
273
Fig. 10.20 Results in DC microgrid with the V–I droop control: (a) active power and (b) bus voltage
V
Small droop Large droop
V*
RV – I
V1
I*
I
I1
I2
Fig. 10.21 Characteristic of V–I droop control depending on the droop constant
power. According to the droop characteristic shown in Fig. 10.21, the injected current (or active power) is increased when the reference voltage is reduced. Also, if the droop constant is larger, the injected current is increased smaller (from I* to I1 ). In contrast, the amount of increase in the injected current becomes larger (from I* to I2 ) when the droop constant is smaller.
10.4 Conclusion This chapter describes the conventional droop control methods used in both AC and DC microgrids. For the stable operation of AC microgrid, the P–f and Q–V droop control methods regulate the grid frequency and bus voltage by changing active and reactive power output, respectively. Also, with the droop control, the power
274
K. W. Joung and J.-W. Prank
sharing among the generators can be achieved. Also, the operating range of each generator type is discussed. The diesel generator has the most constraints because of its complex physical structure. The inverter-based renewables (IBR) and energy storage system (ESS) have the wider operating ranges than the diesel generator or non-IBR because they have no rotational inertia and faster ramp-rate capability. These features are verified by simulation results on the case studies. For the DC microgrid, the V–I droop control method is discussed. Here, the droop constant is used as a virtual resistance, which is making the additional voltage drop. The effectiveness of droop control is also verified by a simulation test with and without droop control. The results show that the converter-based generator (CBG) with the droop control enables to adjust the ratio of power sharing among multiple CBGs by changing their droop constants.
References 1. Wu, D., Tang, F., Dragicevic, T., Vasquez, J. C., & Guerrero, J. M. (2015). A control architecture to coordinate renewable energy sources and energy storage systems in islanded microgrids. IEEE Transactions on Smart Grid, 6(3), 1156–1166. 2. Marwali, M. N., Jin-Woo, J., & Keyhani, A. (2004). Control of distributed generation systems - part II: Load sharing control. IEEE Transactions on Power Electronics, 19(6), 1551–1561. 3. Rocabert, J., Luna, A., Blaabjerg, F., & Rodríguez, P. (2012). Control of power converters in AC microgrids. IEEE Transactions on Power Electronics, 27(11), 4734–4749. 4. Joung, K. W., Kim, T., & Park, J.-W. (2019). Decoupled frequency and voltage control for stand-alone microgrid with high renewable penetration. IEEE Transactions on Industry Applications, 55(1), 122–133. 5. Guerrero, J. M., Vasquez, J. C., Matas, J., Vicuna, L. G. d., & Castilla, M. (2011). Hierarchical control of droop-controlled AC and DC microgrids – A general approach toward standardization. IEEE Transactions on Industrial Electronics, 58(1), 158–172. 6. Joung, K. W., Lee, H.-J., & Park, J.-W. (2019). Assessment of maximum penetration capacity of photovoltaic generator considering frequency stability in practical stand-alone microgrid. Energies, 12(8), 1445. 7. Samimi, A., Kazemi, A., & Siano, P. (2015). Economic-environmental active and reactive power scheduling of modern distribution systems in presence of wind generations: A distribution market-based approach. Energy Conversion and Management, 106, 495–509. 8. Yang, Y., Enjeti, P., Blaabjerg, F., & Wang, H. (2015). Wide-scale adoption of photovoltaic energy: Grid code modifications are explored in the distribution grid. IEEE Industry Applications Magazine, 21(5), 21–31. 9. Venkataraman, S., Ziesler, C., Johnson, P., & Kempen, S. V. (2018). Integrated wind, solar, and energy storage: Designing plants with a better generation profile and lower overall cost. IEEE Power and Energy Magazine, 16(3), 74–83. 10. Van de Vyver, J., De Kooning, J. D., Meersman, B., Vandevelde, L., & Vandoorn, T. L. (2016). Droop control as an alternative inertial response strategy for the synthetic inertia on wind turbines. IEEE Transactions on Power Systems, 31(2), 1129–1138. 11. Che, L., Shahidehpour, M., Alabdulwahab, A., & Al-Turki, Y. (2015). Hierarchical coordination of a community microgrid with AC and DC microgrids. IEEE Transactions on Smart Grid, 6(6), 3042–3051. 12. Gao, F., Kang, R., Cao, J., & Yang, T. (2019). Primary and secondary control in DC microgrids: A review. Journal of Modern Power Systems and Clean Energy, 7(2), 227–242. 13. Lee, S. H. (2019). Adaptive droop based virtual slack control of multiple DGs in practical DC distribution system to improve voltage profile. Energies, 12(8), 1541.
Chapter 11
Distributed Control Approaches for Microgrids Tohid Khalili and Ali Bidram
11.1 Introduction In deregulated power systems, several new challenges have emerged to effectively integrate renewable energy sources (RESs) [1] and provide a balance between the generation and the demanded load of the consumers. The RESs, for example, photovoltaic systems and wind turbine generators are intermittent resources where their output is drastically probabilistic and causes some problems for the load dispatching of the power systems [2, 3]. Moreover, inertias of RESs are low and they don’t have enough capability to improve the power quality of the network. Utilization of the microgrids (MGs) as small-scale power systems that can operate in the grid-connected and islanded modes helps with the mitigation of challenges associated with RESs [4]. The unique feature of MGs is their ability to operate in both grid-connected and islanded modes. The MG control system plays a critical role in accommodating its reliable operation during grid-connected and islanded modes. The MG control system deploys a hierarchical control structure including primary, secondary, and tertiary control levels. These control hierarchies are responsible for the voltage and frequency control and regulation and the optimal operation of the MG. In this chapter, these control hierarchies are elaborated. The MG control system can either adopt a centralized or distributed control structure. The distributed control structure has rendered more advantages compared to the centralized one in terms of reliability and resilience [5]. In a distributed control platform, a peer to peer communication network facilitates the communication of information among Distributed Generators (DGs). DGs make consensus-based control decisions to satisfy a specific MG control target. This chapter addresses
T. Khalili () · A. Bidram Department of Electrical and Computer Engineering, University of New Mexico, Albuquerque, NM, USA e-mail: [email protected]; [email protected] © Springer Nature Switzerland AG 2021 A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems, https://doi.org/10.1007/978-3-030-59750-4_11
275
276
T. Khalili and A. Bidram
the distributed control of both AC and DC MGs and covers the distributed control techniques utilized for voltage/frequency control as well as active/reactive power sharing [6–11]. AC MGs are common types of MGs. However, DC MGs are gaining more attention due to their inherent advantages. DC MGs facilitate the integration of RESs with DC nature. The DC MG’s control system is less complex compared to an AC MG’s as there is no requirement for controlling the frequency and reactive power. Moreover, DC MGs have fewer power losses compared to AC ones [12–18].
11.2 Hierarchical Control Structure of AC and DC MGs Considering the fact that MGs could be operated in both grid-connected and autonomous modes, the stable and economical exploitation of the MGs requires appropriate control approaches [19]. In this section, the hierarchical control structure of DC and AC MGs is elaborated.
11.2.1 DC MGs By having two orders-of-magnitude, DC MGs possess high availability against AC MGs; so, they are a good choice for mission-critical applications [13, 20]. In addition, several challenges of the AC MGs could be solved by using the DC MGs such as frequency synchronization, reactive power control, handling unbalanced loads, and so on [21]. A combination of DC loads which are connected to DC power generators by transmission or distribution systems could create a DC MG. considering the fact that most of the demanded electric loads of the consumers are alternative, power generators have to implement the dynamic control to supply more load instantly respect to the voltage limitations’ criteria. Regarding the required load of the consumers, power generators share the demanded load among each other respect to their rated power which is called proportional load sharing. This concept facilitates the prevention of the power generators’ overstressing and increases their lifetime as well in the MGs. Considering the power generators’ voltage as the single variable to control the power flow in the power systems, they have to be regulated to gain the appropriate voltage profile in the system. To manage and control the DC power generators, a three-level control framework is utilized. The highest and the lowest bandwidth are for the primary and tertiary levels, respectively [22]. The primary level of the control for managing the load sharing utilizes the droop mechanism which relates the DC generator output voltage to its current considering the virtual resistance, RD . This virtual resistance does not have any effect on power loss, and it just has several advantages for load sharing. Accordingly, by using the droop mechanism, and the voltage reference, the voltage controller of the sources is calculated as follows:
11 Distributed Control Approaches for Microgrids
vo∗ = vref − RD .io
277
(11.1)
In (11.1), the voltage controller for the inner loop is shown by vo∗ . In addition, the droop coefficient is indicated by RD . Also, the MG’s rated voltage is shown by vref . Furthermore, the symbol of the source’s output current is io . For voltage restoration of the MGs, the secondary control level is presented. The secondary controller checks the voltage and compares it with the considered voltage by the controller. A Proportional-Integral (PI) module is usually used as the controller. Depending on all of the sources the controller corrects the voltage. By utilizing the corrected voltage as the reference voltage, for performing the droop mechanism, and they do not use the previous voltage reference. To control the power flow between the main grid and the MG, the tertiary control is proposed. Regarding the power flow, the tertiary control compares the reference value with the power flow between the two grids. Then it updates the MG’s reference voltage. Commonly, when the MG’s voltage increases, the DC MG transfers the extra power and vice versa.
11.2.2 AC MGs The MG control structure’s main duties are as follows [23–26]: • Managing the restoration and the transients of the considered situations in the switching mode, • Optimizing of the MG’s operating cost, • Controlling the power flow between the MG and the upstream grid, • Resynchronizing the MG with the upstream grid, • Appropriate load sharing and coordination of the distributed generators (DGs), • Regulating the frequency and the voltage for different conditions. These MGs’ significant roles are so important from the time scale viewpoint. Consequently, these duties are addressed by utilizing a hierarchical control method for answering the mentioned requirements in each level of the control. The hierarchical control method of the MGs has three levels shown in Fig. 11.1. The first level of the control is the fastest from the time duration point of view and helps the MG be stable according to its frequency and the voltage profile in the islanding process and switching between the main grid and the MG. Regarding the nonlinear and the linear loads, it is vital for the DGs to have reactive and active powersharing control which is independent. Additionally, unwanted circulating currents are avoided by using the power-sharing control. The first layer of the control has the basic control infrastructures, which is known as zero level. This includes the internal current and voltage control loops in DGs. The second level of the control operates for compensating the frequency and voltage deviations made by the exploitation of the first level of the control and helps the voltage and frequency synchronization.
278
T. Khalili and A. Bidram
Fig. 11.1 AC MG’s hierarchical control levels [8]
The tertiary control level operates at the highest level and least time scale by managing the power flow between the main grid and the MG and helps to have an optimal economic operation [8, 22].
11 Distributed Control Approaches for Microgrids
279
11.3 Distributed Control of DC MGs 11.3.1 DG Model in a DC MG Figure 11.2 shows the model of a DG integrated into a DC MG. Each DG is equipped with the internal droop and voltage control loops to accommodate a proper voltage regulation and power sharing in the DC MG. The droop control presents the primary control of a DC microgrid and relates DG’s voltage, vo , to its output current, io . The droop controller stabilizes the MG voltage by proportionally sharing power among DGs according to their ratings. The DC droop characteristic can be described as [27, 28]. vo∗ = Vn − rd io ,
(11.2)
where v* o is the voltage reference for the converter and is translated to a duty cycle which creates a similar voltage value across DG’s terminal; Vn is the voltage droop reference; rd is the droop coefficient; rd descries DG as a virtual resistance. It should be noted that the droop coefficients, rd , are chosen according to the DGs current ratings, i.e., rd1 i1,max = · · · = rdN iN,max ,
(11.3)
where ik,max is the kth DG’s current rating. Equation (11.3) facilitates the proportional sharing of power among DGs based on their current ratings, that is,
Fig. 11.2 DC MG’s DG model
280
T. Khalili and A. Bidram
io1 ioN = ··· = . i1,max iN,max
(11.4)
The DG’s terminal voltage can be limited to the prespecified threshold vk, max , if the droop coefficients satisfy rdk ≤
Δvk,max . ik,max
(11.5)
11.3.2 Distributed Secondary Control of DC MGs The DC MG primary control can maintain the voltage of MG in a stable range through the local droop controllers. After the primary controller facilitates the MG transition to the islanded mode, its voltage is less than the nominal MG’s voltage. For critical loads, it is of paramount value to operate them at the nominal voltage. To this end, secondary control is required to regulate the voltage of the critical bus back to MG’s nominal voltage. This section of the chapter presents a distributed DC MG secondary control as seen in Fig. 11.3. The secondary control deploys distributed control protocols on each DG. It is assumed that DG’s control units can communicate with each other in a distributed fashion. The goal of distributed control protocols is to regulate the voltage of a critical bus while proportionally sharing power among DGs according to their current ratings. This relationship is shown in (11.4) [29]. In the distributed secondary control, a PI controller is utilized to define a leader voltage, vleader , for the distributed control agents as
Fig. 11.3 DC MG distributed secondary control [29]
11 Distributed Control Approaches for Microgrids
vleader = kp vref − vcrit + ki
281
.
vref − vcrit dt,
(11.6)
where vref is the DC MG’s nominal voltage. vref can be also defined as a reference voltage dictated by MG’s control center. vcrit is DC MG’s critical bus voltage. kp and ki denote the proportional and integral control gains in the PI controller. The vleader only needs to be shared with one DG. The distributed secondary control protocol at each DG calculates the droop reference, Vn in (11.2). These control protocols are extracted by defining first-order droop control dynamics for each DG as d d d ∗ v + rdk ik,max ik,ratio , (Vnk ) = dt dt ok dt
(11.7)
where ik,ratio describes the kth DG’s current ratio as ik,ratio =
iok ik,max
.
(11.8)
The droop reference for kth DG is calculated as [29]. . Vnk =
uvk dt,
(11.9)
where uvk = Ck (δvk + δik ) ,
δvk =
akj voj − vok + gk (vleader − vok ) ,
(11.10)
(11.11)
j ∈Nk
δik =
akj rdk ik,max ij,ratio − ik,ratio ,
(11.12)
j ∈Nk
where δ vk and δ ik are the local voltage and current neighborhood tracking errors, respectively. akj is the communication link gain between kth and ith DGs. Ck is a control parameter. The pinning gain gk ≥ 0 is nonzero for only one DG that possesses the leader voltage information in (11.6).
282
T. Khalili and A. Bidram
11.4 Distributed Control of AC MGs The conventional secondary level of the MGs’ control operates a centralized control model. Central controller rules based on accumulated data from the system and needs a complicated and centralized communication system [30, 31]. All DGs are controlled through a centralized controller. This negatively influences the system’s configurability and flexibility and enhances the concerns regarding the reliability by posing a single point of failure. The single point of failure indicates that the control system can collapse if the mentioned central controller does not work correctly. MGs could be supposed as the multi-agent systems where their DGs are the agents. A distributed model of the considered communications enhances the reliability of the system. In this controlling model, the control procedures are separated among all of the existing DGs. Thus, the need for a centralized controller is removed, and the control system is not impacted after the failure of a single agent. The control model for a voltage source inverter (VSI) may change with respect to the control targets such as frequency and voltage, or reactive and active power. Distributed control models for autonomous MGs with voltage controlled voltage source inverters (VCVSIs) contains internal current and voltage controller loops which help DGs control their frequency and voltage properly. MGs are capable of exploiting both autonomous and grid-connected states. After the islanding operation, MGs miss the frequency and voltage support by the upstream grid and the reactive and active power equilibrium among the whole consumption and power generation of the system. Thus, several DGs are needed to change their mode to the VCVSI model to make up the differences among the consumption and power generation and to have a fast frequency and voltage support [8]. The VCVSIs are reinforced with the primary droop controllers to keep the MG’s frequency and voltage stability. Primary control stops the frequency and voltage instability by preserving these amounts in safe and stable zones. The primary local controllers’ coordinated control could be gained by the frequency and voltage droop methods as * ωi = ωi∗ − DP i Pi , , (11.13) ∗ = Ei∗ − DQi Qi vo,magi where Ei∗ and ωi∗ indicate the references for the primary control. The droop coefficients are indicated by the DQi and DPi . These droop coefficients are chosen regarding each of VCVSIs reactive and active power ratings. Also, Qi and Pi are the obtained reactive and active power considering the DG’s terminal. The VCVSI’s angular frequency is presented by ωi which the primary control determines. v* o,magi is the reference signal for the magnitude of the terminal voltage of inverter which is considered for controlling the VCVSI’s internal voltage controller. Additionally, the primary control accommodates seamless frequency and voltage support for the MGs. Droop controllers can only stabilize the MG voltage and
11 Distributed Control Approaches for Microgrids
283
frequency. They fail to maintain these quantities at their nominal values. To this end, secondary control is needed to force the frequency and the voltage back to their nominal values. The secondary control of AC MGs includes voltage and frequency controls. ωi∗ in (11.13) is selected by the secondary frequency control to synchronize VCVSIs’ angular frequency to MG’s nominal angular frequency, i.e. ωi → ωref . Furthermore, it finds the VCVSI’s output active powers considering P1 Pmax 1
= ··· =
PNV Pmax NV
,
(11.14)
where NV is the VCVSIs’ number in the autonomous MG. The active power rating of ith VCVSI is indicated by the Pmaxi . So, DPi is selected regarding VCVSIs’ active power ratings as DP 1 P1 = · · · = DP NV PNV .
(11.15)
In (11.13), Ei∗ is selected by the secondary voltage control to regulate VCVSI’s terminal voltages to a reference voltage value, that is, vo,magi →vref . If vref is adjusted to the reference voltage of the MG vnominal , the VCVSIs’ output voltage magnitude synchronizes to the MG’s reference voltage. Meanwhile, vref could be selected to synchronize a vital bus of the MG’s voltage magnitude to the vnominal . The other target of the secondary voltage control could be finding the VCVSI’s output reactive powers considering Q1 Q NV = ··· = , Qmax 1 Qmax NV
(11.16)
where NV is the VCVSIs’ number in the autonomous MG. The reactive power rating of ith VCVSI is indicated by the Qmaxi . So, DQi is selected regarding VCVSIs’ reactive power ratings as DQ1 Q1 = · · · = DQNV QNV .
(11.17)
The characteristics of the voltage droop in (11.13) are built by assuming an MG that is totally inductive. Thus, (11.16) and (11.17) could be exactly fulfilled in a totally inductive MG. However, reactive power ratio mismatches could be seen in the noninductive MGs.
11.4.1 Frequency Control This section elaborates on how to attain the two aims of the secondary frequency control by utilizing just a single distributed control procedure for each of DGs. The distributed control procedure is designed to get the DGs’ frequency, ωi ,
284
T. Khalili and A. Bidram
synchronized with the nominal frequency, ωref . Also, it is designed to share the active power fairly among DGs with respect to their related power ratings. Differencing with the frequency-droop characteristic results in ω˙ i∗ = ω˙ i + DP i P˙i = ui ,
(11.18)
where ui is an additional variable to be modeled. In (11.18), a dynamic model for calculating the ωi∗ from ui is presented. The additional control variable has to be modeled subject to synchronizing the frequency of the DGs to the nominal frequency ωref . Regarding (11.18), an MG’s secondary frequency control having N DGs is changed to a synchronization problem which its objective is to synchronize a first-order linear multi-agent system as. ⎧ ω˙ 1 + DP 1 P˙1 = u1 , ⎪ ⎪ ⎪ ⎨ ω˙ 2 + DP 2 P˙2 = u2 , .. ⎪ ⎪ . ⎪ ⎩ ω˙ N + DP N P˙N = uN .
(11.19)
To get synchronized, it is considered that DGs could send and receive information among themselves via the determined connection graph Gr (representing communication network). The additional controls ui are selected respect to the DG’s self-data, and the data of existing nearby DGs in the connection graph as [5, 19]. ⎛ ui = −cf ⎝
⎞ aij ωi − ωj + bi ωi − ωref + aij DP i Pi − DPj Pj ⎠ ,
j ∈Ni
j ∈Ni
(11.20) where the coupling gain is indicated by cf ∈ R. It is considered that the pinning gain bi ≥ 0 is nonzero for just a single DG which has the nominal frequency ωref . Figure 11.4 illustrates the secondary frequency control’s schematic with respect to the distributed consensus-based control. As shown in Fig. 11.4, the ωi∗ , control input, is indicated as ωi∗ =
. ui dt.
(11.21)
cf as the coupling gain has an effect on the convergence speed of frequency restoration.
11 Distributed Control Approaches for Microgrids
285
Fig. 11.4 The schematic of the secondary frequency control in a distributed fashion [5, 19]
11.4.2 Voltage Control The voltage control’s aim is to select suitable Ei∗ , control inputs, in (11.13) to get vo,magi , VCVSIs’ voltage magnitudes, synchronized to the reference voltage vref , and find the VCVSIs’ generated reactive powers with respect to their reactive power ratings, that is, they meet (11.16) and (11.17). It is notable that (11.13)‘s voltagedroop characteristic is modeled considering an MG contains transmission lines that are inductive. Thus, little differences between voltage magnitudes and the reactive power ratios could be seen in MGs that have not fully inductive transmission lines. The VCVSIs’ voltage magnitudes, vo,magi , synchronization is the same as the synchronization of the vodi . For the concurrent reactive power and VCVSIs’ voltage control, the rapid dynamics of internal current and voltage controllers are disregarded. As a result, v* o,magi and vodi can be considered equal. Due to the stated fact, the (11.13)‘s differentiated voltage-droop specification is written as. ˙ i = vvi , E˙ i∗ = v˙odi + DQi Q
(11.22)
where vvi is an additional variable to be modeled. In (11.22), a dynamic system for calculating the Ei∗ from vvi is presented. The additional control variable has to be modeled subject to synchronizing the VCVSI’s voltage magnitudes to the reference voltage vref and satisfying the (11.16). Regarding (11.22), an MG’s voltage control having NV VCVSIs is changed to a problem wherein its objective is to synchronize a first-order linear multi-agent system as. ⎧ ˙ 1 = vv1 , v˙od1 + DQ1 Q ⎪ ⎪ ⎪ ⎨ ˙ 2 = vv2 , v˙od2 + DQ2 Q . ⎪ ⎪ .. ⎪ ⎩ ˙ NV = vvNV . v˙odN V + DQNV Q
(11.23)
286
T. Khalili and A. Bidram
Fig. 11.5 The schematic of distributed secondary voltage control [5, 19]
To get synchronized, it is considered that VCVSIs could send and receive information among themselves via the determined connection graph Gr. The additional controls vvi are selected with respect to the VCVSI’s self-data, and the data of existing nearby VCVSIs on the connection graph as [5, 19]. ⎛ vvi =−cv ⎝
⎞ aij vodi −vodj + bi vodi −vref + aij DQi Qi −DQj Qj ⎠ ,
j ∈Ni
j ∈Ni
(11.24) where the coupling gain is indicated by cv ∈ R. It is considered that the pinning gain bi ≥ 0 is nonzero for just a single VCVSI which has the data of the nominal voltage vref . It could be observed that all VCVSI voltage amplitudes get synchronized to vref by the (11.24). Moreover, (11.24) allocates VCVSIs’ reactive power with respect to their reactive power ratings. The nominal voltage vref could be adjusted to the MG’s reference voltage for synchronizing the VCVSI voltage magnitudes to the reference voltage or could be selected to control the voltage magnitude of a vital bus of the MG. Figure 11.5 shows the schematic of the presented voltage control. As illustrated in Fig. 11.5, the Ei∗ , control input, is indicated as Ei∗
. =
vvi dt.
(11.25)
11.5 Conclusion and Future Trend This chapter addressed the hierarchical control structure of MGs. The primary, secondary, and tertiary control levels were discussed in detail. The distributed
11 Distributed Control Approaches for Microgrids
287
control techniques for both AC and DC MGs were covered. Although the reliability of the MG control system is enhanced through a distributed control structure increases, it is exposed to cyberattacks due to the deployment of communication and control devices. The future trends in the area of MG control will focus on improving the reliability, security, data integrity, and efficiency of the control system. As one of the major challenges, the cybersecurity of the MG control system should be addressed to tackle both denial-of-service and false data injection attacks. Moreover, with the deployment of intelligent electronic devices, the MG communication network transmits a vast amount of data among different entities. To reduce the burden on the communication network, event-triggered distributed control techniques should be deployed to ensure that the data is only transmitted when the control system is triggered by an event.
References 1. Jafari, A., Khalili, T., Ganjehlou, H. G., & Bidram, A. (Feb. 2020). Optimal integration of renewable energy sources, diesel generators, and demand response program from pollution, financial, and reliability viewpoints: A multi-objective approach. Journal of Cleaner Production, 247, 119100. 2. Hagh, M. T., & Khalili, T. (2019). A review of fault ride through of PV and wind renewable energies in grid codes. International Journal of Energy Research, 43(4). 3. Khalili, T., Nojavan, S., & Zare, K. (Mar. 2019). Optimal performance of microgrid in the presence of demand response exchange: A stochastic multi-objective model. Computers and Electrical Engineering, 74, 429–450. 4. Khalili, T., Jafari, A., Abapour, M., & Mohammadi-Ivatloo, B. (Feb. 2019). Optimal battery technology selection and incentive-based demand response program utilization for reliability improvement of an insular microgrid. Energy, 169, 92–104. 5. Bidram, A., Davoudi, A., & Lewis, F. L. (Aug. 2014). A multi-objective distributed control framework for islanded microgrids. IEEE Trans. Industrial informatics, 10, 1785–1798. 6. T. Khalili, A. Jafari, and E. Babaei, “Scheduling and siting of storages considering power peak shaving and loss reduction by exchange market algorithm,” in IEEE Proceedings 2017 Smart Grid Conference, SGC 2017, 2018, vol. 2018. 7. Khalili, T., Hagh, M. T., Zadeh, S. G., & Maleki, S. (Jul. 2019). Optimal reliable and resilient construction of dynamic self-adequate multi-microgrids under large-scale events. IET Renewable Power Generation, 13(10), 1750–1760. 8. Bidram, A., & Davoudi, A. (Dec. 2012). Hierarchical structure of microgrids control system. IEEE Transactions on Smart Grid, 3, 1963–1976. 9. Hatziargyriou, N., Asano, H., Iravani, R., & Marnay, C. (2007). Microgrids. IEEE Power & Energy Magazine, 5, 78–94. 10. R. H. Lasseter, “Microgrid,” in Proc. IEEE Power Eng. Soc. winter meeting, vol. 1, New York, 2002, pp. 305–308. 11. Lopes, J. A. P., Moreira, C. L., & Madureira, A. G. (May 2006). Defining control strategies for microgrids islanded operation. IEEE Transactions on Power Apparatus and Systems, 21, 916–924. 12. Kwasinski, A., & Onwuchekwa, C. N. (March 2011). Dynamic behavior and stabilization of dc microgrids with instantaneous constant-power loads. IEEE Trans. Power Electronics, 26, 822–834.
288
T. Khalili and A. Bidram
13. Kwasinski, A. (2011). Quantitative evaluation of dc microgrids availability: Effects of system architecture and converter topology design choices. IEEE Transactions on Power Electronics, 26(3), 835–851. 14. Sanchez, S., & Molinas, M. (2014). Degree of influence of system state transition on the stability of a dc microgrid. IEEE Transactions on Smart Grid, 30, 2535–2542. 15. Farhadi, M., & Mohammed, O. (2015). Adaptive energy management in redundant hybrid dc microgrid for pulse load mitigation. IEEE Transactions on Smart Grid, 6, 54–62. 16. Inthamoussou, F. A., Queralt, J. P., & Bianchi, F. D. (2013). Control of a supercapacitor energy storage system for microgrid applications. IEEE Transactions on Energy Conversion, 28, 690– 697. 17. Xu, G., Xu, L., Morrow, D. J., & Chen, D. (2012). Coordinated dc voltage control of wind turbine with embedded energy storage system. IEEE Transactions on Energy Conversion, 27, 1036–1045. 18. Tummuru, N. R., Mishra, M. K., & Srinivas, S. (2015). Dynamic energy management of hybrid energy storage system with high-gain PV converter. IEEE Transactions on Energy Conversion, 30, 150–160. 19. Bidram, A., Nasirian, V., Davoudi, A., & Lewis, F. L. (2017). Cooperative synchronization in distributed microgrid control. Springer International Publishing. ISBN:978-3-319-50807-8 (Print) and 978-3-319-50808-5 (Online). 20. H. Ikebe, “Power systems for telecommunications in the IT age,” in Proc. of IEEE INTELEC, 2003, pp. 1–8. 21. Balog, R. S., Weaver, W., & Krein, P. T. (2012). The load as an energy asset in a distributed dc Smartgrid architecture. IEEE Transactions on Smart Grid, 3, 253–260. 22. Guerrero, J. M., Vasquez, J. C., Matas, J., de Vincuña, L. G., & Castilla, M. (2011). Hierarchical control of droop-controlled AC and DC microgrids – A general approach toward standardization. IEEE Transactions on Industrial Electronics, 58, 158–172. 23. Guerrero, J. M., Matas, J., Vicuna, L. G. D., Castilla, M., & Miret, J. (2007). Decentralized control for parallel operation of distributed generation inverters using resistive output impedance. IEEE Transactions on Industrial Electronics, 54, 994–1004. 24. Guerrero, J. M., Vicuna, L. G. D., Matas, J., Castilla, M., & Miret, J. (2005). Output impedance design of parallel-connected UPS inverters with wireless load-sharing control. IEEE Transactions on Industrial Electronics, 52, 1126–1135. 25. Katiraei, F., Iravani, M. R., & Lehn, P. W. (2005). Microgrid autonomous operation during and subsequent to islanding process. IEEE Trans. Power Del, 20, 248–257. 26. Katiraei, F., & Iravani, M. R. (2005). Power management strategies for a microgrid with multiple distributed generation units. IEEE Transactions on Power Apparatus and Systems, 21, 1821–1831. 27. Chen, Y. K., Wu, Y. C., Song, C. C., & Chen, Y. S. (2013). Design and implementation of energy management system with fuzzy control for dc microgrid systems. IEEE Transactions on Power Electronics, 28(4), 1563–1570. 28. Li, Y., Vilathgamuwa, D. M., & Loh, P. C. (2004). Design, analysis, and real-time testing of a controller for multibus microgrid system. IEEE Transactions on Power Electronics, 19, 1195– 1204. 29. Poudel, B., Mustafa, A., Modares, H., & Bidram, A. (2020). Detection and mitigation of cyberthreats in the DC microgrid distributed control system. International Journal of Electrical Power & energy Systems, 120. https://doi.org/10.1016/j.ijepes.2020.105968. 30. Barklund, E., Pogaku, N., Prodanovic´, M., Hernandez-Aramburo, C., & Green, T. C. (2008). Energy management in autonomous microgrid using stability-constrained droop control of inverters. IEEE Transactions on Power Electronics, 23, 2346–2352. 31. Borup, U., Blaabjerg, F., & Enjeti, P. N. (2001). Sharing of nonlinear load in parallel-connected three-phase converters. IEEE Transactions on Industry Applications, 37, 1817–1823.
Chapter 12
On Control of Energy Storage Systems in Microgrids Yu Wang, Sidun Fang, and Yan Xu
12.1 Introduction The traditional energy structure highly depending on fossil fuels such as coal and oil has become a major concern of climate change and air pollution in modern society. These environmental concerns and energy crises of fossil fuels lead to the rapid development of renewable energy technologies [1, 2]. A large number of renewable energy sources (RESs) together with energy storage systems (ESSs) have been penetrated into existing power systems especially distribution sides through power electronics interfaces. The power generation becomes more and more decentralized other than the conventional centralized generation, which results in the concept of distributed generation. In power distribution systems, a cluster of demand-side loads and distributed energy resources can be connected and disconnected from the main grid to operate in grid-connected or islanded mode. These small-scale power systems are named as microgrids. The original idea of microgrids emerges at the beginning of this century, which aims to benefit the integration of distributed generators (DGs) and enhance grid resilience [3]. In grid-connected mode, the microgrid can be viewed as one entity, which exchanges power with the main grid to realize energy trading and provide ancillary services [4, 5]. In the islanded mode, the major target of the microgrid is to maintain local generation/demand balance, while providing stable and high-quality power supply [6]. According to the current flow, microgrids are usually categorized into alternative current (AC), direct current (DC), and hybrid AC/DC microgrids [7, 8]. The residential, commercial, and industry microgrids have been built all over the world to benefit the renewable penetrations, grid resilience enhancement as well as traffic electrification [9–11].
Y. Wang () · S. Fang · Y. Xu School of Electrical and Electronics Engineering, Nanyang Technological University, Singapore, Singapore © Springer Nature Switzerland AG 2021 A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems, https://doi.org/10.1007/978-3-030-59750-4_12
289
290
Y. Wang et al.
Energy storage systems are relatively new units in microgrids or power distribution systems following in the wake of increased installation of renewable energy generation in the twenty-first century. One typical feature of renewable energy generation is the inherent nature of uncertainties. For example, the power generation of photovoltaic (PV) fluctuates along with the daily solar irradiance variations, and the power generation of wind energy is influenced largely by the seasonal variation of wind at the installed location [11, 12]. PV and wind generation cannot ensure constant power supply as their stochastic and intermittent characteristics, which will influence the stability, reliability, and power quality of power systems [13]. As various problems caused by the penetration of solar panels and wind turbines have become very common, ESSs are considered as one promising solution for such problems in microgrid systems [14]. Typically, ESSs can be categorized by the form of stored energy into five groups: mechanical, chemical, electrochemical, electrical, and thermal energy storage [15]. The energy storage devices belonging to each classification are shown in Fig. 12.1. Among all energy storage categories, electrochemical energy storage with different kinds of batteries is the most widely used in low-voltage electrical systems like microgrids. In microgrids, the ESSs can be installed in a centralized way by the utility company at the point of common coupling (PCC) in the substation [16]. Besides, the ESSs can also be integrated in a distributed way such as plug-in electric vehicles (PEV) and building/home ESSs [17, 18]. Depending on the operation modes of microgrids, the ESSs can be operated for various functionalities. In the grid-connected mode of microgrids, the ESSs with various capacities can be applied for different kinds of ancillary services. According to the time dimension, the primary frequency/voltage control and stability enhancement will be concerned
Fig. 12.1 Classification of energy storage technologies according to energy form [15]
12 On Control of Energy Storage Systems in Microgrids
291
Fig. 12.2 Timescales of energy storage systems for grid ancillary services
in the millisecond to second timeframe, the secondary frequency/voltage control, and reserve service will be concerned in the minute to hour timeframe, peak shaving, and energy trading will be concerned in the hour to day timeframe [16–24]. Figure 12.2 shows the timescales of multiple usages of energy storage systems according to timescale. In the islanded mode of microgrids, the primary objective of ESSs is to balance the generation and demand mismatch, which is particularly important for microgrids with high penetration of RESs. Therefore, the SoC balancing control among ESSs becomes a commonly adopted strategy, in order to effectively utilized ESSs capacity and maintain an uninterrupted power supply of the microgrid. Besides, the ESSs can also be utilized for other functionalities in islanded microgrids such as frequency/voltage control, power quality improvement, renewable generation compensation, and economic dispatch.
12.2 Overview of Energy Storage Systems 12.2.1 Characteristics of ESSs In microgrid applications, the main technical characteristics of ESSs include power density, energy density, life cycle (lifetime), energy efficiency, and self-discharge. According to [25, 26], these metrics are further explained as follows: (a) Energy density (Wh/kg) refers to the energy to weight ratio of one energy storage device. Energy density indicates the capability of continuous energy supply over a period of time. The ESS with higher energy density can discharge energy for a longer period. (b) Power density (W/kg) refers to the power to weight ratio of one energy storage device. Power density indicates the capability of ESS to provide instantaneous power. The ESS with higher power density can discharge a larger amount of power when needed. (c) Energy efficiency (%) refers to the ratio of released energy to stored energy. It also indicates the output and input electricity regardless of self-discharge.
292
Y. Wang et al.
(d) Life cycle refers to the number of times where ESS can provide the designed energy level after each recharge. One cycle means one full charge and discharge operation for the battery. (e) Self-discharge refers to the ratio of initial stored energy to dissipated energy during a nonuse time. Batteries and super-capacitors are two typical ESSs in microgrids, which are further discussed in this paper. Compared to super-capacitors, batteries usually have high energy density but low power density. Among all kinds of batteries (i.e., Pb–Acid, Ni–Cd, Ni–MH, Lithium-ion, Li–polymer, NaS, and VRB), Ni–MH has the lowest energy efficiency. Besides, battery ESSs usually have a short life cycle, while Pb–Acid has the shortest. On the contrary, super-capacitors usually have low energy density but high power density. Although super-capacitors have a long lifetime, their self-discharge is quite high. It means super-capacitors are very suitable for short-term storage applications, considering their high self-discharge and high energy efficiency. In some cases, hybrid ESSs are applied which can combine the positive features of batteries and super-capacitors. However, more complicated power electronics interfaces and control systems are required for the power management of hybrid ESSs. Subsequently, the basic power electronic interface for typical battery ESSs and the battery management system is further discussed.
12.2.2 Power Electronic Interface The power converters serve as the interface between the battery ESSs and the microgrid. The interface should ensure grid codes and electrical standards are satisfied while providing ancillary services to the electricity market [27–29]. The charging and discharging of the battery ESS and power flow of the power conversion system are dependent on the microgrid requirement. In the meantime, the operating points (e.g., terminal voltage, charging/discharging current, state-of-charge (SOC)) of the batteries are regulated, which provides efficient charging/discharging and protects the health of the battery ESS. Depending on the application scenarios, various topologies are proposed to connect battery ESS into the microgrids [27–29]. A simple way is to connect the battery bank directly to the DC link of the DC/AC converter among battery ESS and microgrid. However, as the battery voltages vary with the SOC, the DC-link voltage will be influenced. In this condition, the grid interfacing converters should have the capability to provide a wide range of operating DC voltage. In the meantime, the wide range of modulation indices is also required to address the DC voltage variation, which will cause lower efficiency and higher harmonics. One solution to control the DC-link voltage is to add a bidirectional DC/DC converter between the battery bank and the DC link of the grid interfacing converters, as shown in Fig. 12.3. This conversion system design enables that the battery terminal voltage
12 On Control of Energy Storage Systems in Microgrids
293
Fig. 12.3 A general power electronic interface of grid-connected battery ESS
connecting to the dc link can be adjusted flexibly. In addition, the bidirectional DC/DC converters also contribute to eliminating the low-order harmonics current in the battery bank, which will benefit the total lifetime of the battery bank. Particularly, when the battery ESS is connected to a medium voltage (MV) microgrid system, a transformer should be included in between. Battery cells can be connected in parallel and series at the low-voltage side to build up a battery ESS from hundreds of kWs to tens of MWs. The transformer is installed to boost the voltage from hundreds of Volts to tens of kVs.
12.2.3 Battery Management System The battery ESS usually consists of hundreds to thousands of battery cells. The battery management system (BMS) plays a vital role to manage battery ESS [30, 31]. From an electrical perspective, three major objectives of BMS are: (i) monitoring and estimating the battery states; (ii) protecting the batteries; and (iii) managing the charging/discharging of the batteries. Battery monitoring and state estimation. The battery terminal voltage, current, and surface temperature are key parameters that can be directly measured from sensors. The state estimation is needed to estimate other important indices, including SOC or depth of discharge (DOD), state of health (SOH). SOC or DOD of batteries can be estimated according to the operating condition of battery current, voltage, and temperature [32]. SOH can be calculated by the extent of abuse and performance degradation of batteries [33]. Battery protection and safety control. The protection of the battery should consider both thermal and electrical parameters. When the faults occurs in the battery, the system should diagnose the fault and prevent or mitigate the damage or injury to people caused by overcurrent, overvoltage, overcharge, overdischarge, over-temperature, under-temperature, and so on [34]. Battery cell balancing control. BMS should have the cell equalization function including equalizing charging/discharging to maintain the SOC balance among cells
294
Y. Wang et al.
Table 12.1 ESS Control Strategies in Islanded Microgrids Feature Ref.
Multiple ESSs Droop Consensus [37–43] [44–50]
Hybrid ESSs [51–56]
RESs and ESSs Decentralized [57–60]
Hierarchical [61–66]
[35]. This strategy can prolong the entire battery ESS health by avoiding the overusage of certain cells.
12.3 ESS Control Strategies in Islanded Microgrids In islanded microgrids, the major objective is to maintain power supply while enhancing system stability and resilience. The ESSs play key roles in compensating the short-term power mismatch as well as long-term energy management. This section will review the typical control strategies of ESSs in islanded microgrids by functionalities. The terminologies from hierarchical control architecture in microgrids is adopted [36]. This chapter mainly focuses on the system-level, realtime, coordinated control of ESSs. The control strategies of battery cells and power electronics in ESSs, as well as the system-level optimization of ESSs, are not covered. The coordinated control of islanded microgrids with ESSs summarized in this section are listed in Table 12.1 below.
12.3.1 Coordinated Control of Multiple ESSs The energy limitation can be ignored for DG units supplied by fossil fuels, as it is assumed that the fuel supply is of a sufficient amount. However, the energy stored in ESSs (e.g., the SOC of lithium-ion battery) is limited, which should be considered in the microgrid level control system. The SOC balancing becomes a commonly adopted strategy for multiple ESSs in islanded microgrids, due to the following reasons: (1) the power mismatch between RESs and loads can be buffered by an islanded microgrid with balanced SOC among ESSs; (2) the prevention of unintentionally switch-off batteries caused by their energy depletion or saturation; and (3) the overcharged or overdischarged of a certain battery that damage the battery health can also be avoided. Droop based SOC Balancing. The droop based SOC balancing control has been widely discussed in state-of-the-art. The main idea is by involving SOC into the droop coefficients. The droop coefficients will be changed adaptively according to the SOC of each ESS so that SOC levels in the microgrid will be balanced. Sufficient work has been conducted to investigate the droop based SOC balancing methods. In [37], the authors propose an adaptive droop control for SOC balancing
12 On Control of Energy Storage Systems in Microgrids
295
of distributed ESSs in DC microgrids. Later on, this method is improved with a double-quadrant SOC-based droop control method in [38]. In [39], a multifunctional droop control scheme is proposed, which can provide localized SOC balancing. In [40], a decentralized power management strategy is reported for a hybrid ESS with SOC recovery and autonomous bus voltage restoration. Considering the different capacities of multiple ESSs, the authors in [41] proposes an improved droop control. A voltage scheduling droop control for SOC balancing of distributed ESSs in DC microgrids is reported in [42]. Fuzzy droop control is designed for balancing stored energy in distributed ESSs in [43]. The main concern of decentralized SOC balancing control is the lack of global coordination. The system power quality is sacrificed (e.g., frequency and voltage deviations due to droop control). It is hard to design localized control for frequency and voltage restoration without any communication. Consensus-based SOC Balancing. The distributed SOC balancing control has been included in the secondary level of microgrid control due to the limitations above. Based on neighboring communication, the secondary control objectives can be achieved together with SOC balancing. The power quality in the islanded microgrid system is improved. Therefore, researchers are working on SOC balancing among a group of ESSs in the secondary level by distributed control methods [44– 50]. The voltage or frequency deviation problem caused by primary droop control can be effectively avoided. In the early stage, the authors in [44] propose a distributed multi-agent cooperative control for frequency regulation and energy level balancing in AC microgrids. Later, this method is extended for heterogeneous ESS in DC microgrids in [45] and by using sliding mode control for SOC balancing in [46]. In later research, [47] proposes a SOC balancing for ESSs in grid-connected AC microgrids, while [48] proposes a distributed secondary control including SOC balancing in islanded AC microgrids. Recently, researchers are trying to use novel control strategies to further improve the control performance. In [49], a distributed finite-time consensus control for heterogeneous ESSs in AC microgrids is proposed considering cyber-physical implementation. In [50], a distributed cyber-resilient secondary control is presented for SOC balancing among multiple ESSs. Coordination of Hybrid ESSs. Energy storages with different characteristics can collaboratively operate for power management in microgrids. Typically, supercapacitors are units with high power density but low energy density, while lithiumion batteries have low power density but high energy density [50]. The philosophy is to control HESSs to deal with power variation events with different frequency bandwidth. Research efforts have been made on coordinated control strategies of HESSs. In [52], a droop-based decentralized power-sharing strategy is proposed for hybrid ESSs in DC microgrids. In [53], a hierarchical control scheme for hybrid ESSs in DC microgrids is introduced. In [54], a robust frequency regulation with hybrid ESSs in multi-area microgrids is proposed. In recent research, researchers are also making efforts to including SOC balancing using consensus in hybrid ESS in DC microgrids with different topologies [55, 56].
296
Y. Wang et al.
12.3.2 Coordinated Control of RESs and ESSs RESs like PV and wind turbines are widely integrated into microgrids. The coordination problem of RESs and ESSs have raised much attention in the research community. Many research works have been conducted to provide solutions for RESs and ESSs coordination in islanded microgrids. Fully Decentralized Control. Various kinds of droop control curves are designed for localized coordination of RESs and ESSs in microgrids. In [57], a real powersharing control strategy is proposed for islanded AC microgrids by using frequency bus signaling to achieve power allocation in a decentralized manner. In [58], a smooth switching droop control is introduced to coordinate RESs and ESSs in islanded AC microgrids. In [59], a real power-sharing method by the dynamic droop factors to control charge/discharge power allocation between the superconducting magnetic energy storage and the battery is proposed. Decentralized power control in AC islanded microgrids with RESs and ESSs is also reported in [60]. The frequency and voltage restoration are still the major concerns of decentralized coordination. Hierarchical Control. Influenced by the idea of hierarchical control of traditional power grids. Hieratical control has been proposed for the coordination of RESs and ESSs in microgrids. The centralized secondary control can be adopted for frequency/voltage restoration. The master-slave hierarchical control scheme for DG units has been proposed by some research works. In [61], a master-slave coordinated control strategy for DGs and the ESSs for islanded operation is proposed. In [62], a localized model predictive control (MPC) for DG units and rule-based centralized coordination is introduced. On the other hand, the DG units can be governed by a peer-to-peer hierarchical control scheme. In [63], hierarchical power control of various units in DC microgrids is proposed. Hierarchical control of the hybrid ESSs in DC microgrids is proposed in [64]. In [65], a centralized architecture is proposed for real power curtailment of generation to avoid overcharge of the ESSs, load shedding for avoiding deep discharge of the ESSs, and balancing of the SOCs among all ESSs to prevent battery degradation. The major concerns of the centralized architecture are that the central controller will have large computation and communication burdens for large-scale systems, and it is prone to communication failures and central device failures. In [66], a distributed coordinated control strategy is proposed for managing the multiple ESSs to balance the power generation and load demand while minimizing the system power loss during the charging/discharging.
12.4 ESS Control Strategies in Grid-Connected Microgrids In grid-connected microgrids or power distribution networks, the major objective is to maintain acceptable power quality while providing grid ancillary services. The ESSs play key roles in improving the short-term power quality requirement
12 On Control of Energy Storage Systems in Microgrids
297
Table 12.2 ESS Control Strategies in Grid-Connected Microgrids Feature Ref
Voltage regulation Decentralized [75, 76]
Hierarchical [77–81]
Frequency regulation Primary Secondary [85–92] [93–95]
as well as satisfying long-term grid dispatch order. This section presents the control strategies of ESSs for voltage and frequency regulation in grid-connected microgrids and power distribution networks. The literature reviewed in this section is listed in Table 12.2 below.
12.4.1 Voltage Regulation A power distribution network can be viewed as a grid-connected microgrid with RESs, ESSs, and local load demands. Since the last few years, the fast developments of renewable technology have increased the penetration level of DGs in distribution networks. However, it results in two kinds of voltage problems: voltage rise and drop issues. In a community with a high penetration level of rooftop PVs, PV generation in the day time can significantly exceed load demand. The reverse power flow from the customer end to the utility grid can potentially result in voltage rise issues in the distribution network [67–70]. In the meantime, the large share of PEV will cause additional power demand, which will cause voltage drop issues in the distribution network [71–74]. So voltage limits violation could occur during both peak PV generation and load demand periods, which leads to poor power quality and even equipment failure. Decentralized Control. The decentralized/local voltage control by power inverters of ESSs can be usually divided into P-dependent and V-dependent methods [75, 76]. In the P-dependent method, the real/reactive power output of ESSs is dependent on the real power generation of PV units. This control is also known as ramprate control to mitigate voltage fluctuations [75]. In the V-dependent method, the power output from the ESS is dependent on the local bus voltage, that is, voltage droop control [76]. However, the system statuses are changing rapidly considering the high uncertainty of renewables and plug-and-play ability of PEVs, It makes the decentralized control itself hard to deal with the voltage regulation in such a system. Even with the optimized droop settings and set points from upper level optimization, the system voltages can still exceed the operational limits. In this condition, the realtime coordination between local controllers is required. Hierarchical Control. As fully decentralized control cannot provide well coordination of ESSs in microgrids, the hierarchical control has been proposed which incorporates upper level coordination. The centralized coordinated control scheme of distributed ESSs with tap changer transformers to mitigate voltage rise in a system with high PV penetration is addressed in [77]. Many works have been
298
Y. Wang et al.
conducted for the distributed coordination of ESSs in power distribution networks. Multi-agent-based voltage control for clustered active distribution networks by means of ESSs is proposed in [78]. In [79], a local and distributed coordinated control of distributed ESSs is proposed for voltage and SOC management. In [80], a distributed MPC method for battery ESSs is introduced for voltage regulation in a high renewable penetrated distribution network. In [81], a hierarchical control framework is introduced to coordinate several groups of virtual energy storage (thermostatically controlled loads), to control network loading and voltage in power distribution networks.
12.4.2 Frequency Regulation Traditionally, the synchronous generators from bulk power plants function to balance power mismatch between generation and load consumption. However, the system inertia and the frequency reserves are decreased due to the large integration of RESs and subsequent replacement of rotational generators. Generally, when a disturbance occurred, the power system with less inertia and frequency regulation reserves will suffer from more severe frequency deviations. Under this circumstance, ESSs have been viewed as a good candidate for solving such problems. As the capability to contribute to power system operation and control, ESSs have been gradually deployed in modern power systems [82–84]. Primary frequency control. As the fast response speed of ESSs, they have been viewed as suitable candidates for system primary frequency control, from both industry and academic perspective [85–87]. In [88], a fuzzy logic frequency control strategy is presented by utilizing the large capacity distributed PV systems and PEVs. The sizing problem of ESSs for grid inertial response and primary frequency reserve is studied in [89]. From an economic point of view, the planning of battery ESSs for primary frequency control is investigated in [90]. In [91], the battery life issues during the operation of a grid-tied lithium-ion battery ESS for the provision of primary frequency regulation is studied. In [92], a data-driven method is reported to predict the real-time power fluctuations in power systems, which is used by ESSs for power system frequency support. Secondary Frequency Control. In high renewable penetrated power systems, an emerging but important task to be addressed in the increasing requirement of automatic generation control (AGC) capacity [93]. The centralized ESSs installed by the grid operator is able to provide AGC services. However, the limited capacity of utility ESSs makes them more suitable for primary control. In the meantime, the demand-side ESSs such as PEVs and home ESSs have a high potential for the secondary frequency regulation services. Although the demand-side ESSs are usually of small capacity, the number of them will be very large. Herein, it is needed to design a coordinated control scheme to aggregate ESSs. In [94], a coordinated control framework is designed for large-scale PEV charging aggregators and ESSs to provide the frequency regulation. In [95], an MPC is proposed for load
12 On Control of Energy Storage Systems in Microgrids
299
aggregators and utility-scale ESSs for power system frequency regulation. In [18], a distributed finite-time consensus control scheme is proposed to aggregate demandside ESSs for power system secondary control.
12.5 Research Trends and Opportunity Multifunctional ESSs. In recent studies, it is found that the net economic benefit of ESSs is not exploited with only one single grid ancillary service. For example, only 1–50% of a battery’s lifetime capacity will be taken, if it only participates in primary services. Therefore, there are commercial opportunities from the ESS providing additional functions and services. Recent studies have suggested more applications, that is, stack service, would reduce the idling duration of ESS, increasing the economic value and payback period [96]. In future work, multiple time-scale coordinated control scheme of ESSs is to be developed. The multifunctional operation of ESSs without conflicting each functionality is the bottleneck of this design. Large-scale Control and Optimization of ESSs. In recent years, a distributed control framework has proved its merits for real-time coordination without limitations of centralized control. The computation and communication burden are fairly shared by the distributed controllers. It also offers more robustness to single point and communication failures. For large-scale distributed ESSs integration in power distribution level networks, the distributed coordinated control and optimization is recommended. The distributed algorithms considering the dynamics, functions, economics, health, as well as other working conditions, are to be investigated. Both networked dynamic control and networked optimization techniques are highly demanded in order to address large-scale penetration of ESSs in the power networks. In the cyber networks, the distributed algorithms are realized on multi-agent systems (MASs) with embedded systems [97]. The realization of distributed algorithms is also an important issue to promote future development. Cybersecurity of ESSs. Large-scale distributed control and optimization of ESSs highly relies on the performance of communication networks as well as its security. Yet so far, the cybersecurity issues of ESSs have not raised much attention. However, in future power systems with large-scale networked controlled ESSs, the security problem should be taken into account. To protect the security of networked controlled ESSs under lethal cyberattacks, proper defense mechanisms are to be investigated. Three aspects of defense mechanisms can be investigated in future works [98]. (1) The prevention methods to protect the ESS from cyberattacks. (2) The detection and isolation methods to detect the cyberattacks and isolate the corrupted subsystems. (3) The resilience algorithms to overcome the negative impact of an attack and operate the system closer to its normal state.
300
Y. Wang et al.
12.6 Conclusion This chapter introduces the control and application of ESSs in microgrid systems. The characteristics of energy storage techniques, power electronic interfaces, and battery management systems are introduced. A comprehensive review of ESSs in both islanded microgrids and grid-connected microgrids has been conducted. The future research roadmaps of multifunctional ESSs, distributed control of optimization of ESSs, and cybersecurity issues of ESSs are discussed.
References 1. Bose, B. K. (2010). Global warming: Energy, environmental pollution, and the impact of power electronics. IEEE Industrial Electronics Magazine, 4(1), 6–17. 2. Panwar, N. L., Kaushik, S. C., & Kothari, S. (2011). Role of renewable energy sources in environmental protection: A review. Renewable and Sustainable Energy Reviews, 15(3), 1513– 1524. 3. Lasseter, R. H. (2002). Microgrids. Proc. IEEE Power Engineering Society Winter Meeting, 305–308. 4. Mengelkamp, E., et al. (2018). Designing microgrid energy markets: A case study: The Brooklyn microgrid. Applied Energy, 210, 870–880. 5. Wang, Y., Xu, Y., Tang, Y., & Xu, Y. L. (2020). A distributed control scheme of thermostatically controlled loads for the building-microgrid community. IEEE Transactions on Sustainable Energy, 11(1), 350–360. 6. Rocabert, J., et al. (2012). Control of power converters in AC microgrids. IEEE Transactions on Power Electronics, 27(11), 4734–4749. 7. Liu, X., Wang, P., & Loh, P. C. (2011). A hybrid AC/DC microgrid and its coordination control. IEEE Transactions on Smart Grid, 2(2), 278–286. 8. Wang, P., et al. (2013). Harmonizing AC and DC: A hybrid AC/DC future grid solution. IEEE Power and Energy Magazine, 11(3), 76–83. 9. Ustun, T. S., Ozansoy, C., & Zayegh, A. (2011). Recent developments in microgrids and example cases around the world—A review. Renewable and Sustainable Energy Reviews, 15(8), 4030–4041. 10. Wang, Y., Xu, Y., & Tang, Y. (2019). Distributed Aggregation Control of Grid-Interactive Smart Buildings for Power System Frequency Support. Applied Energy, 251(1). 11. Fang, S., Wang, Y., Gou, B., & Xu, Y. (2020). Towards future green maritime transportation: Seaport microgrids and all-electric ships. IEEE Transactions on Vehicular Technology, 69(1), 207–219. 12. Dincer, I. (2000). Renewable energy and sustainable development: A crucial review. Renewable and Sustainable Energy Reviews, 4(2), 157–175. 13. Arul, P. G., Ramachandaramurthy, V. K., & Rajkumar, R. K. (2015). Control strategies for a hybrid renewable energy system: A review. Renewable and Sustainable Energy Reviews, 42, 597–608. 14. Vazquez, S., Lukic, S. M., Galvan, E., Franquelo, L. G., & Carrasco, J. M. (Dec. 2010). Energy storage systems for transport and grid applications. IEEE Trans. Industrial Electronics, 57(12), 3881–3895. 15. Electrical energy storage project team, “Electrical energy storage.” International Electrotechnical Commission, 2011. 16. Gyuk, I., Johnson, M., Vetrano, J., Lynn, K., et al. (2013). Grid energy storage. U.S Department of Energy.
12 On Control of Energy Storage Systems in Microgrids
301
17. Carpinelli, G., Celli, G., Mocci, S., Mottola, F., Pilo, F., & Proto, D. (June 2013). Optimal integration of distributed energy storage devices in smart grids. IEEE Transactions on Smart Grid, 4(2), 985–995. 18. Wang, Y., et al. (July 2019). Aggregated energy storage for power system frequency control: A finite-time consensus approach. IEEE Transactions on Smart Grid, 10(4), 3675–3686. 19. Energy Market Authority, “Energy Storage Programme,” 2015. Available https:// www.ema.gov.sg/Energy_Storage_Programme.aspx. 20. Hill, C. A., Such, M. C., Chen, D., Gonzalez, J., & Grady, W. M. (2012). Battery energy storage for enabling integration of distributed solar power generation. IEEE Transactions on Smart Grid, 3(2), 850–857. 21. Subburaj, A. S., Pushpakaran, B. N., & Bayne, S. B. (2015). Overview of grid connected renewable energy based battery projects in USA. Renewable and Sustainable Energy Reviews, 45, 219–234. 22. Zhang, Y., Ren, S., Dong, Z. Y., Xu, Y., Meng, K., & Zheng, Y. (2017). Optimal placement of battery energy storage in distribution networks considering conservation voltage reduction and stochastic load composition. IET Generation, Transmission & Distribution, 11(15), 3862– 3870. 23. Fang, S., et al. (2019). Optimal hierarchical Management of Shipboard Multibattery Energy Storage System Using a data-driven degradation model. IEEE Transactions on Transportation Electrification, 5(4), 1306–1318. 24. Dong, C., Jia, H., Xu, Q., Xiao, J., Xu, Y., et al. (2018). Time-delay Stability Analysis for Hybrid Energy Storage System with Hierarchical Control in DC Microgrids. IEEE Trans. Smart Grid, 99(99). 25. Grid integration of large-capacity renewable energy sources and use of large-capacity electrical energy storage. International Electrotechnical Commission, 2012. 26. Ibrahim, H., Ilinca, A., & Perron, J. (2008). Energy storage systems-characteristics and comparisons. Renewable Sustainable Energy Reviews, 12(5), 1221–1250. 27. Wang, G., et al. (2016). A review of power electronics for grid connection of utility-scale battery energy storage systems. IEEE Transactions on Sustainable Energy, 7(4), 1778–1790. 28. Yilmaz, M., & Krein, P. T. (May 2013). Review of battery charger topologies, charging power levels, and infrastructure for plug-in electric and hybrid vehicles. IEEE Transactions on Power Electronics, 28(5), 2151–2169. 29. Grainger, B. M., Reed, G. F., Sparacino, A. R., & Lewis, P. T. (June 2014). Power electronics for grid-scale energy storage. Proceedings of the IEEE, 102(6), 1000–1013. 30. Lu, L., et al. (2013). A review on the key issues for lithium-ion battery management in electric vehicles. Journal of Power Sources, 226, 272–288. 31. Xiong, R., Li, L., & Tian, J. (2018). Towards a smarter battery management system: A critical review on battery state of health monitoring methods. Journal of Power Sources, 405, 18–29. 32. Xiong, B., et al. (2014). Extended Kalman filter method for state of charge estimation of vanadium redox flow battery using thermal-dependent electrical model. Journal of Power Sources, 262, 50–61. 33. Wei, Z., et al. (2018). Online state of charge and state of health estimation for a lithium-ion battery based on a data–model fusion method. Energies, 11(7), 1810. 34. Sidhu, A., Izadian, A., & Anwar, S. (2014). Adaptive nonlinear model-based fault diagnosis of Li-ion batteries. IEEE Transactions on Industrial Electronics, 62(2), 1002–1011. 35. Lee, Y., & Cheng, M. W. (2005). Intelligent control battery equalization for series connected lithium-ion battery strings. IEEE Transactions on Industrial Electronics, 52(5), 1297–1307. 36. Guerrero, J. M., et al. (2010). Hierarchical control of droop-controlled AC and DC microgrids—A general approach toward standardization. IEEE Transactions on Industrial Electronics, 58(1), 158–172. 37. Lu, X., et al. (2013). State-of-charge balance using adaptive droop control for distributed energy storage systems in DC microgrid applications. IEEE Transactions on Industrial Electronics, 61(6), 2804–2815.
302
Y. Wang et al.
38. Lu, X., et al. (2014). Double-quadrant state-of-charge-based droop control method for distributed energy storage systems in autonomous DC microgrids. IEEE Transactions on Smart Grid, 6(1), 147–157. 39. Sun, X., et al. (2016). A multifunctional and wireless droop control for distributed energy storage units in islanded AC microgrid applications. IEEE Transactions on Power Electronics, 32(1), 736–751. 40. Xu, Q., et al. (2017). A decentralized power management strategy for hybrid energy storage system with autonomous bus voltage restoration and state-of-charge recovery. IEEE Transactions on Industrial Electronics, 64(9), 7098–7108. 41. Wu, Q., et al. (2018). SoC balancing strategy for multiple energy storage units with different capacities in islanded microgrids based on droop control. IEEE Journal of Emerging and Selected Topics in Power Electronics, 6(4), 1932–1941. 42. C. Li, et al. ”Voltage scheduling droop control for State-of-Charge balance of distributed energy storage in DC microgrids.“ 2014 IEEE International Energy Conference (ENERGYCON). IEEE, 2014. 43. N. Díaz, et al. ”Fuzzy droop control loops adjustment for stored energy balance in distributed energy storage system.“ 2015 9th International Conference on Power Electronics and ECCE Asia (ICPE-ECCE Asia). IEEE, 2015. 44. Morstyn, T., Hredzak, B., & Agelidis, V. G. (2015). Distributed cooperative control of microgrid storage. IEEE Transactions on Power Systems, 30(5), 2780–2789. 45. Morstyn, T., Hredzak, B., & Agelidis, V. G. (2016). Cooperative multi-agent control of heterogeneous storage devices distributed in a DC microgrid. IEEE Transactions on Power Systems, 31(4), 2974–2986. 46. Morstyn, T., Savkin, A. V., Hredzak, B., & Agelidis, V. G. (2018). Multi-agent sliding mode control for state of charge balancing between battery energy storage systems distributed in a DC microgrid. IEEE Transactions on Smart Grid, 9(5), 4735–4743. 47. Cai, H., & Hu, G. (2016). Distributed control scheme for package-level state-of-charge balancing of grid-connected battery energy storage system. IEEE Transactions on Industrial Informatics, 12(5), 1919–1929. 48. Y. Wang, et al. ”Distributed Secondary Control of Energy Storage Systems in Islanded AC Microgrids.“ 2018 Asian conference on energy, power and transportation electrification (ACEPT). IEEE. 49. Wang, Y., et al. (2020). Distributed control of heterogeneous energy storage systems in islanded microgrids: Finite-time approach and cyber-physical implementation. International Journal of Electrical Power & Energy Systems, 119, 105898. 50. Deng, C., et al. (2020, in press). Distributed resilient control for energy storage Systems in Cyber-Physical Microgrids. IEEE Transactions on Industrial Informatics, 1. 51. Jing, W., et al. (2016). Battery-supercapacitor hybrid energy storage system in standalone DC microgrids: A review. IET Renewable Power Generation, 11(4), 461–469. 52. Xu, Q., et al. (2016). A decentralized dynamic power sharing strategy for hybrid energy storage system in autonomous DC microgrid. IEEE Transactions on Industrial Electronics, 64(7), 5930–5941. 53. Xiao, J., Wang, P., & Setyawan, L. (2015). Hierarchical control of hybrid energy storage system in DC microgrids. IEEE Transactions on Industrial Electronics, 62(8), 4915–4924. 54. C. Ju, Y. Tang, Y. Wang, “Robust Frequency Regulation with Hybrid Energy Storage Systems in Islanded Microgrids,” 2018 Asian conference on energy, power and transportation electrification (ACEPT), Oct. 2018. 55. Lin, P., et al. (2019). A semi-consensus strategy toward multi-functional hybrid energy storage system in DC microgrids. IEEE Transactions on Energy Conversion. 56. B. Wang et al., ”Consensus-based Control of Hybrid Energy Storage System with a Cascaded Multiport Converter in DC Microgrids,“ IEEE Transactions on Sustainable Energy. doi: https:/ /doi.org/10.1109/TSTE.2019.2956054. 57. Wu, D., et al. (2014). Autonomous active power control for islanded ac microgrids with photovoltaic generation and energy storage system. IEEE Transactions on Energy Conversion, 29(4), 882–892.
12 On Control of Energy Storage Systems in Microgrids
303
58. Wu, D., et al. (2014). A control architecture to coordinate renewable energy sources and energy storage systems in islanded microgrids. IEEE Transactions on Smart Grid, 6(3), 1156–1166. 59. Li, J., et al. (2017). Design/test of a hybrid energy storage system for primary frequency control using a dynamic droop method in an isolated microgrid power system. Applied Energy, 201, 257–269. 60. de Matos, J. G., Silva, F. S. F. E., & Ribeiro, L. A. D. S. (2015). Power control in AC isolated microgrids with renewable energy sources and energy storage systems. IEEE Transactions on Industrial Electronics, 62(6), 3490–3498. 61. Kim, J. Y., et al. (2010). Cooperative control strategy of energy storage system and microsources for stabilizing the microgrid during islanded operation. IEEE Transactions on Power Electronics, 25(12), 3037–3048. 62. Y. Wang, K. T. Tan, and P. L. So. ”Coordinated control of battery energy storage system in a microgrid.“ 2013 IEEE PES Asia-Pacific power and energy engineering conference (APPEEC). IEEE, 2013. 63. Jin, C., et al. (2013). Implementation of hierarchical control in DC microgrids. IEEE Transactions on Industrial Electronics, 61(8), 4032–4042. 64. Xiao, J., Wang, P., & Setyawan, L. (2015). Hierarchical control of hybrid energy storage system in DC microgrids. IEEE Transactions on Industrial Electronics, 62(8), 4915–4924. 65. Diaz, N. L., et al. (2016). Centralized control architecture for coordination of distributed renewable generation and energy storage in islanded AC microgrids. IEEE Transactions on Power Electronics, 32(7), 5202–5213. 66. Xu, Y., et al. (2014). Cooperative control of distributed energy storage systems in a microgrid. IEEE Transactions on Smart Grid, 6(1), 238–248. 67. Tonkoski, R., Turcotte, D., & El-Fouly, T. H. M. (Jul. 2012). Impact of high PV penetration on voltage profiles in residential neighborhoods. IEEE Transactions on Sustainable Energy, 3(3), 518–527. 68. Y. Wang, T. Zhao, C. Ju, Y. Xu, P. Wang. “Two-level distributed voltage/Var control of aggregated PV inverters in distribution networks,” IEEE Transactions on Power Delivery, Nov. 2019. 69. Wang, Y., Xu, Y., Tang, Y., Syed, M. H., Guillo-Sansano, E., & Burt, G. (Feb. 2019). Decentralised-distributed hybrid voltage regulation of power distribution networks based on power inverters. IET Generation, Transmission & Distribution, 13(3), 444–451. 70. Wang, Y., Syed, M. H., Guillo-Sansano, E., Xu, Y., & Burt, G. (Nov. 2019). Inverterbased voltage control of distribution networks: A three-level coordinated method and power hardware-in-the-loop validation. IEEE Transactions on Sustainable Energy. 71. C. Zhang, C. Chen, J. W. Sun, P. W. Zheng, et al., “Impacts of electric vehicles on the transient voltage stability of distribution network and the study of improvement measures,” in Proc. 2014 IEEE PES eAsia-Pacific power and energy engineering conference (APPEEC), pp. 1–6, Dec. 2014. 72. Luo, X., & Chan, K. W. (2014). Real-time scheduling of electric vehicles charging in lowvoltage residential distribution systems to minimise power losses and improve voltage profile. IET Generation, Transmission and Distribution, 8(3), 516–529. 73. Richardson, P., Flynn, D., & Keane, A. (2011). Optimal charging of electric vehicles in lowvoltage distribution systems. IEEE Trans. Power Systems, 27(1), 268–279. 74. Wang, Y., et al. (2019). A coordinated voltage control scheme using electric vehicle chargers in low-voltage distribution networks. Electric Power Systems Research, 168, 218–227. 75. Alam, M. J. E., Muttaqi, K. M., & Sutanto, D. (2014). A novel approach for ramp-rate control of solar PV using energy storage to mitigate output fluctuations caused by cloud passing. IEEE Transactions on Energy Conversion, 29(2), 507–518. 76. Kabir, M. N., Mishra, Y., Ledwich, G., Dong, Z. Y., & Wong, K. P. (2014). Coordinated control of grid-connected photovoltaic reactive power and battery energy storage systems to improve the voltage profile of a residential distribution feeder. IEEE Transactions on Industrial Informatics, 10(2), 967–977.
304
Y. Wang et al.
77. Liu, X., et al. (2012). Coordinated control of distributed energy storage system with tap changer transformers for voltage rise mitigation under high photovoltaic penetration. IEEE Transactions on Smart Grid, 3(2), 897–906. 78. Bahramipanah, M., Cherkaoui, R., & Paolone, M. (2016). Decentralized voltage control of clustered active distribution network by means of energy storage systems. Electric Power Systems Research, 136, 370–382. 79. Wang, Y., Tan, K. T., Peng, X., & So, P. L. (2016). Coordinated control of distributed energy storage systems for voltage regulation in distribution networks. IEEE Transactions on Power Delivery, 31(3), 1132–1141. 80. Meng, K., et al. (2015). Cooperation-driven distributed model predictive control for energy storage systems. IEEE Transactions on Smart Grid, 6(6), 2583–2585. 81. Meng, K., et al. (2017). Distributed control of thermostatically controlled loads in distribution network with high penetration of solar PV. CSEE Journal of Power and Energy Systems, 3(1), 53–62. 82. Kottick, D., Blau, M., & Edelstein, D. (1993). Battery energy storage for frequency regulation in an island power system. IEEE Transactions on Energy Conversion, 8(3), 455–459. 83. Roberts, B. P., & Sandberg, C. (2011). The role of energy storage in development of smart grids. Proceedings of the IEEE, 99(6), 1139–1144. 84. Wen, S., et al. (2019). Proactive frequency control based on ultra-short-term power fluctuation forecasting for high renewables penetrated power systems. IET Renewable Power Generation. 85. M. Swierczynski, et al. ”Field tests experience from 1.6 MW/400kWh Li-ion battery energy storage system providing primary frequency regulation service.“ IEEE PES ISGT Europe 2013. IEEE, 2013. 86. Y. Wang, H. Li, Y. Xu, and Y. Tang, “System frequency regulation in Singapore using distributed energy storage systems,” 2017 Asian Conference on Energy, Power and Transportation Electrification (ACEPT), Oct. 2017. 87. Oudalov, A., Chartouni, D., & Ohler, C. (2007). Optimizing a battery energy storage system for primary frequency control. IEEE Transactions on Power Systems, 22(3), 1259–1266. 88. Datta, M., & Senjyu, T. (2013). Fuzzy control of distributed PV inverters/energy storage systems/electric vehicles for frequency regulation in a large power system. IEEE Transactions on Smart Grid, 4(1), 479–488. 89. Knap, V., et al. (2015). Sizing of an energy storage system for grid inertial response and primary frequency reserve. IEEE Transactions on Power Systems, 31(5), 3447–3456. 90. Zhang, Y. J., et al. (2016). Profit-maximizing planning and control of battery energy storage systems for primary frequency control. IEEE Transactions on Smart Grid, 9(2), 712–723. 91. Stroe, D. I., et al. (2016). Operation of a grid-connected lithium-ion battery energy storage system for primary frequency regulation: A battery lifetime perspective. IEEE Transactions on Industry Applications, 53(1), 430–438. 92. S. Wen, et al. “Identification of real-time power fluctuations based on LSTM recurrent neural network and a case study on Singapore power system,” IEEE Trans. Industrial Informatics, 2019. 93. Donadee, J., & Wang, J. (2014). AGC signal modeling for energy storage operations. IEEE Transactions on Power Systems, 29(5), 2567–2568. 94. Zhong, J., et al. (2014). Coordinated control for large-scale EV charging facilities and energy storage devices participating in frequency regulation. Applied Energy, 123, 253–262. 95. Wen, G., et al. (2015). Frequency regulation of source-grid-load systems: A compound control strategy. IEEE Transactions on Industrial Informatics, 12(1), 69–78. 96. F. Teng, M. Aunedi, R. Moreira, et al, “Business case for distributed energy storage,” in Proc. 24th International Conference on Electricity Distribution (CIRED), Glasgow, 2017. 97. Wang, Y., et al. (2020). A distributed control scheme of microgrids in energy internet and its multi-site implementation. IEEE Transactions on Industrial Informatics. in press. 98. Dibaji, S. M., Pirani, M., Flamholz, D. B., Annaswamya, A. M., Johanssonb, K. H., & Chakraborttyc, A. (2019). A systems and control perspective of CPS security. Annual Reviews in Control, 47, 394–411.
Chapter 13
Microgrid Stability Definition, Analysis, and Examples Hossein Shayeghi, Hamzeh Aryanpour, Masoud Alilou, and Aref Jalili
13.1 Introduction Microgrids, as a new type of network in power distribution systems, have been developed with the advent of distributed generation to increase system reliability and address economic and environmental issues [1]. To build a microgrid, renewable energy is usually applied as much as possible so inverter interfaced distributed generations are used widely in the microgrid, which makes the operating characteristics of microgrid quite different from the traditional grid. Similar to synchronous generators that play the most key roles in the traditional grid, inverter interfaced distributed generations are also as the same important roles in a microgrid. However, the dynamic process of microgrid stability is more complicated due to the decrease in the inertia of the microgrid in islanded mode. Due to the microgrid operation mode, its stability problems are categorized into grid-connected and islanded stability issues. In the grid-connected mode, the stability issues of the microgrid in transient and small signal studies are focused more on voltage stability. The researches on small signal stability of islanded microgrid have drawn much attention. Because maintaining power supply and load balance are very vital by microgrid itself. In the islanded mode, microgrid stability is categorized into the voltage stability and frequency stability in both the transient and small signal studies. A linearized model of the network is used for the analysis of small signal stability in the microgrid. Also, the time domain and eigenvalue-based analysis and droop gain optimization are the common methods to study smalldisturbance stability.
H. Shayeghi () · H. Aryanpour · M. Alilou Energy Management Research Center, University of Mohaghegh Ardabili, Ardabil, Iran A. Jalili Department of Electrical Engineering, Islamic Azad University, Ardabil, Iran © Springer Nature Switzerland AG 2021 A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems, https://doi.org/10.1007/978-3-030-59750-4_13
305
306
H. Shayeghi et al.
Microgrid distribution systems comprise different distributed energy resources (DERs) and operate in isolation from or while connected to the main grid. In islanded mode, the voltage and frequency of microgrids should be controlled by different DERs [2] given the numerous disturbances and load uncertainties experienced by these systems in the real world. Such problems reduce the stability of voltage and frequency owing to the consequent imbalance between power generation and demand. Voltage and frequency control cannot be guaranteed by traditional controllers in the presence of different DERs, disturbances, and uncertainties because these controllers are not always suitable for all operating conditions. A controller that robustly performs across a wide range of system operating conditions is necessary for an islanded microgrid (a requirement that has been addressed with the development of hierarchical control). As indicated in the IC/ISO 62264 standard [3], the goal of fast-response primary control is to adjust the frequency and amplitude of voltage references against any variations in sources and loads. Secondary control is intended to regulate frequency and voltage to ensure that they are of acceptable levels and to compensate for the imbalance in the two. In tertiary control, realizing acceptable and cost-effective operating conditions necessitates economic dispatch and power optimization. A review of the literature indicated that many droop control structures have been presented for voltage and frequency compensations, as has been done in [4–6]. The imbalance between total demand and generation causes voltage and frequency to deviate from admissible limits. To improve voltage [7] and frequency [8] responses, researchers implement load shedding as a separation scheme. In [9], a method was developed on the basis of a real-time load shedding computation for the simultaneous recovery of voltage and frequency. In [10], a battery inverter with rapid response was examined to control voltage and frequency in an islanded microgrid. Li et al. presented active and reactive power management for an inverterbased microgrid to enhance the system’s transient stability [11]. In [12], the coordinated control of inverter-based distributed generators and distributed energy storage was provided to evaluate dynamic stability in the microgrid. In islanded microgrids, matching demand and generation, regulating voltage and frequency, and sharing power between sources are critical issues. Correspondingly, researchers recommended an approach to operating islanded microgrids based on voltage source inverters that operate in current-regulated mode [13]. To consider nonlinearity and uncertainties in nonconventional energy sources, the researchers provided a robust energy management system that uses predictive control theory as mathematical bases. This method was verified as robust against variations in the wind power generated by islanded microgrids [14]. In [15], a battery energy storage system was suggested to ensure the independence of system frequency from the inertia of synchronous generators, and an active/reactive power droop controller was employed as the voltage controller. In [16], the authors recommended the coordination of DERs and demand responses. In this coordination, loads are classified on the basis of their
13 Microgrid Stability Definition, Analysis, and Examples
307
associated sensitivity values, and total operating cost in grid-connected mode is minimized using a particle swarm optimization algorithm. Tang et al. proposed a hybrid method based on the combination of a voltage/frequency droop controller and an active/reactive droop controller to improve small-signal stability [17]. One of the most important procedures in the simultaneous control of voltage and frequency is the complete modeling of microgrids which facilitates the design of acceptable controllers. The study, in which this modeling was conducted, increases running time because of rising complexity, experts cannot design a controller with good performance. As a solution, Kunjumhummed modeled a power system based on a multi-machine structure by relating system equations to one another using a network impedance matrix [18]. A survey of microgrid modeling approaches was conducted in [19] to detail dynamic models of main microgrid components. The discussion of the models reflected that some adaptive and robust controllers are used to develop secondary voltage and frequency control schemes; these controllers include PI [20], sliding mode [21], robust mixed H2/H∞ [22], robust H∞ and μsynthesis [23] and fuzzy logic [24] controllers. For the islanded operation of microgrids, it is necessary to develop a new simulation model for the simultaneous control of voltage and frequency as a secondary control scheme. Correspondingly, this chapter develops a secondary control simulation model that is based on a multi-machine structure and expresses the relationship between different units through an admittance matrix. The model exhibits a highly accurate and fast simulation and considers all kinds of renewable energy sources and energy storage devices. On the basis of the model, a new adaptive fuzzy proportional integral-derivative (AFPID) controller is developed to ensure effective control in the presence of renewable energy sources and robust performance over various operating conditions. The proposed controller is equipped with two independent parts that work in parallel with others. The first part is a PID controller that guarantees stability and eliminates steady-state errors in nominal operating conditions. The second part is a fuzzy logic-based mechanism that tunes the gains of the PID controller. This fuzzy component increases flexibility against uncertainties and nonlinearity. The design of the AFPID controller is inspired by the tuning of controller output based on nonlinear and voltage responses. However, precisely designing a fuzzy controller is more complex and based on human knowledge. Thus, the nondominated sorting improved differential evolution (NSIDE) algorithm is used to optimally design the controller. The advantages of the NSIDE algorithm include its easy understandability, robustness, and requirement for only a few parameters in promptly adjusting and handling non-differentiable, nonlinear, and multi-model functions. It also presents diverse solutions and effective uniform distributions of Pareto solutions. In the optimization process, the Analytical Hierarchy Process (AHP) mechanism is used to ensure the selection of the best solutions. The AHP is a multi-criteria decision-making method that helps decisionmakers for addressing a complex problem with multiple conflicting and subjective criteria. The method adopted in this work is the combination of the NSIDE algorithm, fuzzy concepts, and the AHP increases convergence speed and optimal search capabilities. The algorithm is employed to avoid the need for trial-and-
308
H. Shayeghi et al.
error processes, augment the performance of the AFPID controller, and accordingly ascertain that it comprises parameters such as PID gains, fuzzy rule bases, and fuzzy membership functions. The AFPID controller has a simple structure and robustly performs against uncertainties and nonlinearity. For the validation of effectiveness, the controller is applied in an actual microgrid equipped with three combined heat and power (CHP) units and three wind units. The controller is installed in the CHP units and the dynamic behavior of the system is tested under different scenarios. Its performance is then compared with that of the conventional fuzzy PID and classical PID controllers based on performance indices. The results show that the proposed controller exhibits robust performance under different load changes and disturbances and is therefore recommended for use in the actual microgrids as a secondary control scheme. • Motivations and contributions The literature review showed that an adaptive method has been implemented by researchers to enable simultaneous voltage/frequency control in an islanded microgrid. Major factors that prompted the present work are the development of a novel model and the design of a simultaneous fuzzy PID controller. Most traditional controllers do not appropriately function against nonlinearity and uncertainty issues. The current scheme involves employing the NSIDE algorithm to design the AFPID controller which demonstrates to be acceptably productive. The main motivation of the study is the design of the AFPID controller as a secondary control scheme for microgrids. The proposed control strategy is applied in an islanded microgrid for simultaneous controlling of voltage and frequency. Therefore, the contributions of this work are summarized below. • The development of a novel multi-machine structure-based simulation model for the study of the dynamic behavior of microgrids. • The design of an AFPID voltage and frequency controller for islanded microgrids. • The use of the NSIDE algorithm as an effective tool for the optimization and automatic design of an AFPID controller for islanded microgrids. • The effective minimization of the integral of squared time multiplied by square error (ISTSE) against uncertainty and nonlinearity.
13.2 Microgrid Modeling In the studies of microgrid, researchers develop many methods and software programs but they also encounter certain problems such as long simulation times and changes to microgrid structures because of expansion initiatives. In this chapter, a multi-machine model for the study of microgrid dynamics is developed. The model is suitable for use in MATLAB/Simulink software and can be considered multiple types of distributed generators, various loading conditions, and different
13 Microgrid Stability Definition, Analysis, and Examples
309
Synchronous generator differential equation Asynchronous generator differential equation
Voltage Transmission network model IDQ=YBUS× VDQ
Reactive compensator differential equation
Current
Other system components differential equation
Fig. 13.1 The main structure of the proposed model
compensators. The model also links various load models for microgrids through a transmission admittance matrix. Figure 13.1 illustrates the main structure of the proposed model.
13.2.1 The Fixed-Speed Wind Turbine Model A fixed-speed wind turbine consists of three parts, namely, a turbine blade, a squirrel-cage asynchronous generator, and a capacitor bank. Turbine blades convert kinetic energy into mechanical energy which is then converted into electrical energy by squirrel-cage asynchronous generators. Capacitor banks are installed near the terminal of asynchronous generators to compensate for the use of reactive power. The establishment of an asynchronous generator model depends on the level of details in the model; that is, the model that needs to be constructed hinges on whether it is of fifth-order, third-order, or one-order type. These models are described in [25, 26]. The present research uses a third-order asynchronous generator model. In the aerodynamic component shown in Fig. 13.2, Cp, i , λi , β i , and Trot, i are the aerodynamic power coefficient, speed ratio, pith angle, and aerodynamic torque, respectively, in a fixed-speed wind turbine i. This component can be modeled according to Eqs. (13.1)–(13.3) [26]. λi =
Ri ωrot,i υ
(13.1)
310
H. Shayeghi et al.
Speed ratio calculation Eq. (7.1)
Aerodynamic power coefficient calculation Eq. (7.2) , , ,
Aerodynamic torque calculation Eq. (7.3)
Mechanical model Eq. (7.4)
,
Asynchronous generator model
Fig. 13.2 Simple model of constant speed wind turbo generator [26]
Cp,i = (0.44 − 0.0167βi ) sin
Trot,i =
π (λi − 3) − 0.0184 (λi − 3) βi 15 − 0.3βi
1 ωrot,i
Cp,i
ρi π Ri2 υ 3 2
(13.2)
(13.3)
where ρ i , Ri , υ, and ωrot, i denote the air density, rotor radius, wind speed, and rotor speed, respectively. The mechanical part of the asynchronous generator is described by Eq. (13.4) as follows: ω0 ωrot,i dωrot,i Trot,i − Te,i − Di = dt 2Hi ω0
(13.4)
Here, ω0 , Te, i , Hi , and Di represent the synchronous speed, electrical torque, inertia, and damping coefficient of the asynchronous generator, respectively.
13.2.2 CHP Model CHP is the most important energy technology with respect to microgrids; it is efficient and presents environmental, economic, and reliability benefits. These advantages have prompted the increased use of this energy technology. In CHP units, gas turbines are mechanically connected to a synchronous generator. In other words, energy conversion in the units is carried out by a synchronous generator. In this chapter, a gas turbine is used in the CHP unit that has been modeled in [27], with some modifications. In a gas turbine, a droop controller is typically used to regulate frequency but such a controller is inappropriate for islanded microgrids. This deficiency requires the development of a new extra control loop through which
13 Microgrid Stability Definition, Analysis, and Examples
311
DTurb
Pref
AFPID
Controller
(Frequency)
min
-
Vmax
+ W
-
1/R
+
Droop
Wref
1
-
T3.s+1
+
K
-
1
1
Control Select
T1.s+1
Pm
+
T2.s+1
Vmin
+ +
AT
Fig. 13.3 Gas turbine model Qg Vref
PF Error Measurement
PF ref
PI
VRmax
PF Control
+ Pg Vt
+ 1
T.s+1
-1
Ka
AFPID (Voltage)
-
Voltage Control
Ta.s+1
+
1 -
Efd
Te.s
VRmin
Control Select Vf
Kfs s+1/Tf1
s+1/Tf3 s+1/Tf3
+ +
Ke Se
Fig. 13.4 Modified AC5A IEEE excitation model
secondary control is implemented. In consideration of this issue, therefore, we incorporated an extra frequency control loop into our gas turbine. Figure 13.3 shows the modified gas turbine structure of the CHP unit. The IEEE’s type AC5A excitation system is used as the excitation system for the synchronous generator [27]. Figure 13.4 shows the excitation system which has an extra voltage controller (AFPID) as a second voltage control tool. The model developed for the CHP unit can be controlled and regulate voltage and frequency deviations during the islanded operation of microgrids. These controllers can function simultaneously as a secondary control measure.
13.2.3 Synchronous Reference Frame Examining multi-machine stability necessitates identifying a relationship between a machine’s terminal voltage and current in terms of transmission admittance. Individual machine coordinates di–qi may be related to common system coordinates D–Q [28] because the transmission admittance matrix in static coordinates can be expressed by rotating coordinates. Details of this reference frame conversion are presented in [29].
312
H. Shayeghi et al.
13.2.4 Reduced Network Model In this chapter, loads, the capacitor, and reactor compensation are modeled as constant admittance. The admittance of each component can be determined using Eq. (13.5). yCi =
Pi − j Qi Vbase
(13.5)
In this equation, Pi , Qi , and Vbase are the active power, reactive power at bus i, and voltage base, respectively. The variable yCi is incorporated into the transmission admittance matrix Ybus . If n is the number of buses, then the dimensions of Ybus are n × n. The size of Ybus can be reduced by adopting only generator buses and disregarding non-generator buses. That is, the effects of non-generator buses are transferred into generator buses. If Ybus is divided into four sub-matrices (YGG , YGN , red. is written as: YNG , and YNN ), then Ybus Ybus =
YGG YGN YN G YN N
red. ⇒ Ybus = YGG − YGN .YN−1N .YN G
(13.6)
13.3 AFPID Controller Simultaneous voltage/frequency control for islanded microgrids is complex because of the presence of many parametric uncertainties, load changes, and nonlinearity. Therefore, a robust controller should be developed. Given that a fuzzy controller considers uncertainties and nonlinearities in implementing voltage/frequency control, this research is designed as an AFPID controller (Fig. 13.5). AFPID-based control is an approach that combines the traditional PID control and fuzzy logic which does not require an exact mathematical model of a controller object and adopts a fast, small overshoot, and short settling time. In this approach, PID parameters can be adjusted in real-time under adaptive fuzzy PID control, thereby enabling updates to control knowledge and enhancements to the electrical behavior of a system. The proposed AFPID controller has two independent parts that act in parallel with others. The first part is a simple PID controller, whose gains are tuned in nominal load conditions as offline. The second is a fuzzy logic-based mechanism intended to update the gains of PID controllers online due to the dynamical behavior of the microgrid. The mechanism has two inputs with coefficients α (j) and β (j) and automatically tunes the gains of the first part (i.e., the PID controller). As indicated in Fig. 13.5, the proposed controller has two inputs and three outputs, whose membership functions are shown in Fig. 13.6. Seven triangular membership functions are used for each input/output. These are the NL (negative large), NM (negative medium), NS (negative small), ZR (zero), PS
13 Microgrid Stability Definition, Analysis, and Examples
313
Fig. 13.5 Structure of the AFPID controller in CHP
Fig. 13.6 Membership functions of the fuzzy PID controller Table 13.1 The fuzzy rules of fuzzy PID controllers for all outputs
de NL NM NS ZR PS PM PL
e NL ZR NS NM NL NL NL NL
NM PS ZR NS NM NL NL NL
NS PM PS ZR NS NM NL NL
ZR PL PM PS ZR NS NL NL
PS PL PL PM PS ZR NS NM
PM PL PL PL PM PS ZR NS
PL PL PL PL PL PM PS ZR
(positive small), PM (positive medium), and PL (positive large) functions. The basic role base for the proposed controller is given in Table 13.1, which is depicted as a 7 × 7 matrix given that the inputs have seven membership functions. The parameter x(k)(h)(j) in Fig. 13.6 denotes interval changes in a membership function and its value should be obtained from optimization. In the notation, k denotes the input/output. If K is the input, then h is an error or Δ error; otherwise, h denotes ΔKp , ΔK i , or ΔK d . In this notation, j denotes the voltage or frequency controller.
314
H. Shayeghi et al.
13.4 Global Design of the Optimization To derive an optimal design for the AFPID controller, the NSIDE algorithm is used as this avoids a trial-and-error design process and enables the simultaneous control of voltage and frequency in islanded microgrids. Optimally designing a fuzzy system requires the appropriate selection of gains, membership functions, fuzzy weighting coefficients, and rule bases. This design also depends on accurate knowledge of a system, previous experiences, and observations. Such complexity renders the simultaneous optimal adjustment of all parameters a highly complex and time-consuming process. Additionally, it is an unreasonable approach because of the availability of numerous parameters. To overcome these problems, the optimization of this study is divided into three cascaded stages: Stage I: Tuning gains and membership functions. Stage II: Tuning fuzzy rule weighting coefficients. Stage III: Tuning fuzzy rule bases. In every optimization stage, the NSIDE algorithm is adopted to achieve optimal performance. The cost function, optimization variables, and NSIDE algorithm must also be refined to realize the aims of optimal design.
13.4.1 Cost Function One of the aims of this study is to minimize voltage and frequency deviations during the islanded operation of microgrids. Therefore, the ISTSE index based on voltage and frequency deviations is regarded as the cost function, which is described as: *. f1 = max
/
(tsim )
t
2
2 VCH P (i)
t
2
2 FCH P (i)
0
*. f2 = max
(13.7)
i = 1, 2 and 3
(13.8)
/
(tsim ) 0
i = 1, 2 and 3
In these equations, ΔVCHP(i) and ΔFCHP(i) are the voltage and frequency changes in CHP (i), respectively.
13 Microgrid Stability Definition, Analysis, and Examples
315
13.4.2 System Optimization Variables 13.4.2.1
Stage I
In this stage, the gains of the controller and the membership functions must be optimized. Figure 13.5 shows that five gains exist for each controller (Kp(j ) , Ki(j ) , Kd(j ) , α (j) , and β (j) ). In the notation, j denotes the voltage or frequency controller. As illustrated in Fig. 13.6, to reduce the number of optimization variables for fuzzy membership functions, each member function is described by x(k)(h)(j) . The linguistic variables are defined in Eq. (13.9). N L = −x(k)(h)(j ) , NM = −
ZR = 0, P S =
13.4.2.2
x(k)(h)(j ) x(k)(h)(j ) , NS = − 3 3
x(k)(h)(j ) x(k)(h)(j ) ,PM = , P L = x(k)(h)(j ) 3 3
(13.9)
Stage II
In this stage, for each rule (n is the number of rules, equal to 49), a weight factor (j ) w(h) (i) (i = 1 : 49) is examined to determine the importance and influence of rules on the controller’s final voltage and frequency responses. To employ this type of coding, we can provide a condition for the removal of ineffective rules. In this case, a number of fuzzy rules become optimal because the zero weighting coefficients of the rules do not affect the final response and can be removed from the set of rules. The importance of this method is that in addition to improving the performance of the fuzzy controller, it eliminates some ineffectual laws, decreases the number of variables in the next step (fuzzy rules set), and advances improved optimization.
13.4.2.3
Stage III
In this stage, the set of fuzzy logic rules need to be encoded, regulated, and then decoded. To encode fuzzy rules, an index (a positive integer number) is assigned to each law. First, any logical expression determined according to Eq. (13.10) is converted into T, thus: n + 1 T = max (mi ) i=1
(13.10)
where T and mi are the maximum number of logical expressions and the logical expressions related to the ith variable, respectively. In this conversion, zero is assigned to the lowest logic expression, and the number of membership functions
316
H. Shayeghi et al.
minus one is assigned to the highest logic expression. Then, the obtained numbers are converted on the basis of a positive integer T according to Eq. (13.11): I=
n+1
Li T (n+1−i)
Li ∈ (0, 1, . . . , mi−1 )
(13.11)
i=1
Here, I is the index that corresponds to a given fuzzy rule and serves as an optimization variable. The decoding of fuzzy rules contrary to the above-mentioned steps is demonstrated in Eq. (13.12): −(i−1) , + n + 1 T L(n+2−i) = I.T −(i−1) − T i = 1 T
(13.12)
In this equation, [.] denotes that only the integer part of the operational result is taken [30]. In the proposed controller, each variable has seven triangular output membership functions. According to the fuzzy rules table, therefore, the number of fuzzy rules for each output variable is 49. If an optimization variable is allocated to each rule, the number of optimization variables for each variable is 147. In the proposed method, however, this number is reduced to 49, resulting in more effective optimization.
13.4.3 NSIDE Algorithm Nowadays, the differential evolutionary (DE) algorithm is regarded as a very effective method for solving complex multi-objective problems. Its advantages include fast convergence rate, robustness, and global optimization. The basic DE algorithm involves mutation, crossover, and selection operations, through which vectors of trial parameters are generated. The implementation of the DE algorithm has been presented in greater detail in [31].
13.4.3.1
Improved Differential Evolution Algorithms
The DE algorithm is improved by implementing changes to the four stages of initialization, mutation, selection, and updating of the mutation coefficient. This improvement is described as follows: I. Initialization: Randomly construct population OP of NP individuals, with the dimension of each vector being n, using the following rule [32]: Xi+N P = V ar min + rand (0, 1) × (V ar max − Xi )
(13.13)
13 Microgrid Stability Definition, Analysis, and Examples
317
where Varmin and Varmax denote the lower and upper bound of the ith component, respectively, and rand(0, 1) is a uniform random number falling between 0 and 1. II. Mutation: Randomly select three distinct individuals Xr1 , Xr2 , and Xr2 from population S and perform mutation using the following formula: Vi,G+1 = Xtb + F × Xr1,G − Xr2,G
(13.14)
Here, Xtb is the best of the three individuals (Xr1 , Xr2 , and Xr2 from population S). III. Selection: Calculate the objective function value at newly generated individuals and select NP individuals from the population of the latest generated individuals, mutation individuals, and crossover individuals to diversify the newly generated ones. IV. Updating the mutation coefficient: After each generation, the mutation rate is updated using the expressions below so that mutation parameter F increases from an initial value F2i to a final value F2f with the iterative progression of the optimization algorithm [33]. F1 (t) = μ × F1 (t − 1) × [1 − F1 (t − 1)] iter + F2i F1 (t) 0 ≤ μ ≤ 4 F (t) = F2f − F2i iter max
(13.15)
(13.16)
In these equations, F1(0) lies between [0,1]. The index “t” is the current iteration, and F1 (t) is the new mutation factor.
13.4.3.2
Nondominated Sorting-Based Multi-Objective Algorithm
A nondominated sorting-based multi-objective algorithm was proposed and used in the nondominated sorting genetic algorithm II by Deb et al. [34]. In this algorithm, every step of the optimization process involves the selection of favorable responses to produce the next generation of individuals. The domination concept can be defined as Eq. (13.17). x dominate ⇐⇒
∀i : xi ≤ yi ∃i0 : xi0< yi0
(13.17)
In this relationship, x dominates y if the two conditions indicated in the equation above are satisfied. Responses are sorted in a Pareto sequence (Pareto I, Pareto II, and . . . ) according to the number of dominations by other responses. For the next generation of a Pareto’s parent, the first priority is the Pareto number (quality
318
H. Shayeghi et al.
criterion). If the parent needs to select the number of Pareto members, the criterion is used to call the discipline. The discipline criterion is determined on the basis of the responses in a Pareto distribution. The discipline criteria are the dispersion of the responses to be satisfied in the search space and increase the chances of reaching an improved response. Crowding distance is used according to Eq. (13.18) to rank the discipline with the benchmark index. k k f n (i+1) − f(i−1) k k CD i = cd i , i = 1, 2, . . . , np, cd i = (k=1) fm ax k − fm ink (13.18) where i is the number of individuals in each A Pareto, k is the number of objective functions, n denotes the number of objective functions, and f represents the objective function value. Finally, the AHP mechanism (see [35] for more details) is used to select the best response from the first Pareto member of the multi-objective optimization as output and complete the setting of parameters for the AFPID controller. Figures 13.7a–c show the flowchart of the three phases of parameter setting for the AFPID controller.
13.5 Simulation Results Figure 13.8 displays the actual microgrid examined to validate the proposed method [36]. The microgrid is part of the Himmerlands Elforsyning power system in Aalborg, Denmark. It encompasses a total of 10 loads, three fixed-speed wind turbine units, and CHP units with three gas turbo generators. The wind turbine generator close to the unity power factor and the capacitor bank used for reactive power compensation is operated in islanded mode. The data on loads and lines are extracted from [36]. Because the wind turbines in the microgrid run at a fixed speed and the CHPs are the main power producers, the AFPID controller is installed in the CHPs. The reduced admittance matrix of the microgrid is equal to: ⎡
red. Ybus
⎢ ⎢ ⎢ ⎢ 2⎢ = 10 ⎢ ⎢ ⎢ ⎢ ⎣
− 0.0573 + 0.0426i −0.0573 + 0.0426i −0.0573 + 0.0426i
⎤ −0.0928 + 0.0721i −0.0573 + 0.0426i −0.0535 + 0.0355i ⎥ − 0.0928 + 0.0721i −0.0573 + 0.0426i −0.0535 + 0.0355i ⎥ ⎥ − 0.0928 + 0.0721i −0.0573 + 0.0426i −0.0535 + 0.0355i ⎥ ⎥ ⎥ +0.4332 − 0.3624i −0.0742 + 0.0785i −0.0742 + 0.0785i ⎥ ⎥ ⎥ − 0.0742 + 0.0785i +0.3983 − 0.3381i −0.1458 + 0.1328i ⎦
− 0.0535 + 0.0355i −0.0535 + 0.0355i −0.0535 + 0.0355i
− 0.0703 + 0.0670i −0.1458 + 0.1328i +0.3824 − 0.3049i
+1.0157 − 1.1498i −0.3961 + 0.4972i −0.3961 + 0.4972i − 0.3961 + 0.4972i +1.0157 − 1.1498i −0.3961 + 0.4972i − 0.3961 + 0.4972i −0.3961 + 0.4972i +1.0157 − 1.1498i −0.0928 + 0.0721i −0.0928 + 0.0721i −0.0928 + 0.0721i
Power flow results for islanded microgrid, the calculated initial conditions for islanded microgrid, and the parameters of the NSIDE algorithm are presented in Tables 13.2, 13.3, and 13.4, respectively.
13 Microgrid Stability Definition, Analysis, and Examples
319
Fig. 13.7 (a) The tuning process of AFPID controller by NSMDE algorithm. Stage 1: Fuzzy coefficients and membership function tuning. (b) The tuning process of AFPID controller by NSMDE algorithm. Stage 2: Fuzzy rule weights tuning. (c) The tuning process of AFPID controller by NSMDE algorithm. Stage 3: Fuzzy rules tuning
Figure 13.9 shows the nondominated responses (Pareto front) of the threestage optimization of the AFPID controller. The objective functions are the ISTSE, applied according to Eq. (13.7) and Eq. (13.8). As indicated in Fig. 13.9, the best solution is selected by the AHP mechanism. The results of the three-stage optimization including optimized coefficients of controllers by NSIDE algorithm, optimized membership function parameters, optimized rules for voltage controller, and optimized rules for frequency controller are presented in Tables 13.5, 13.6, 13.7, and 13.8, respectively. To demonstrate the effectiveness of the proposed controller, four scenarios with different disturbances are considered. The disturbances are a symmetrical three-
320
H. Shayeghi et al.
Fig. 13.7 (continued)
phase fault, load shedding, and decreasing and increasing loads. In all the scenarios, the worst time-domain values for each of the CHPs are selected to examine the worst-case condition. The performance of the proposed controller in all the scenarios is assessed against that of the fuzzy PID controller [37] and the classical PID controller. The fuzzy PID controller does not work online, but its output signal, which was derived after applying coefficients Kp , Ki , and Kd , is used directly in the system. This controller does not adjust the fuzzy PID coefficients but adjusts its output.
13 Microgrid Stability Definition, Analysis, and Examples
321
Fig. 13.7 (continued)
13.5.1 Scenario 1: Symmetrical Three-Phase Fault A three-phase fault over the lines and near the buses affects the admittance matrix. Therefore, the row and column of the admittance matrix of the microgrid experiencing the fault bus must be omitted, after which the lower columns and rows must be pulled one step back. This process generates a small-order admittance matrix. After the fault is cleared, replacing the current admittance matrix with the initial admittance matrix returns the microgrid to its normal state. In this scenario, a symmetrical three-phase fault occurs in t = 1 s at bus 11, and after two cycles of t = 1.04 s, the fault is cleared. The microgrid voltage and frequency responses to this fault are depicted in Figs. 13.10 and 13.11, respectively. The AFPID controller damps the transient responses in a better manner in terms of settling time and steady-state error than do the comparison controllers. The results also indicate that the AFPID controller is considerably more suitable than the two
322
H. Shayeghi et al.
Fig. 13.8 The real microgrid Table 13.2 Power flow results for islanded microgrid
Bus 1 2 3 4 5 6
|Vi | 1 1 1 0.98 0.97 0.97
θi 0 −0.5 −0.5 −11.34 −11.35 −11.35
Pgi 3.07 2.9 2.9 0.1 0.1 0.1
Qgi 0.69 0.67 0.67 0 0 0
others. To confirm the results obtained through the simulation, the values of some response characteristics (e.g., ISTSE, overshoot, undershoot, and settling time) are determined on the basis of Figs. 13.10 and 13.11. The result of voltage and frequency is listed in Table 13.9, which shows that AFPID improvements in the time domain of voltage and frequency in terms of the ISTSE are 76.75% and 72.23% rather than PID respectively, and 62.83% and 37.14% rather than FPID. The table also indicates that the percentage of improvement achieved with the AFPID controller is higher than that realized using the fuzzy PID controller. We can, therefore, conclude that the AFPID controller is the best control option among all the evaluated controllers.
13 Microgrid Stability Definition, Analysis, and Examples
323
Table 13.3 Calculated initial conditions for islanded microgrid Parameter Ia0 0 Eq0 δ0 Id0 Iq0 Vd0 Vq0 E’d0 E’q0 Te0 Tm0
CHP 1 3.147 −0.221 1.386 0.519 −2.121 2.324 −0.496 0.868 0.278 1.069 3.044 3.044
Table 13.4 The parameters of the NSIDE algorithm
CHP 2 2.976 −0.236 1.369 0.499 −1.996 2.208 −0.487 0.874 0.264 1.071 2.971 2.971
CHP3 2.976 −0.236 1.369 0.499 −1.996 2.208 −0.487 0.874 0.264 1.071 2.971 2.971
WTG 1 0.090 −0.198 1.122 0 −0.018 0.088 −0.218 1.088 0.011 0.002 0.122 0.122 Parameter NC NM NW NR Npop OP Iteration
WTG 2 0.901 −0.198 1.122 0 −0.018 0.088 −0.219 1.088 0.011 0.002 0.122 0.122 Value 5 5 50 50 50 50 50
WTG 3 0.090 −0.198 1.122 0 −0.018 0.088 −0.219 1.088 0.011 0.002 0.122 0.122
Parameter F1t F2i F2f μ CR CRmin CRmax
Value 0.8 0.5 1.5 4 0.6 0.5 0.9
13.5.2 Scenario 2: Load Shedding in Bus 8 In this scenario, to ensure the effective performance of the designed controller under load shedding, the load produced by bus 8 (load 08) at t = 1 s is separated from the system. Load shedding causes to happen an imbalance between power generation and consumption and fluctuations in the microgrid voltage and frequency. The microgrid voltage and frequency responses to this condition are demonstrated in Figs. 13.12 and 13.13, respectively. The values of the time domain characteristics of voltage and frequency are presented in Table 13.10. Table 13.10 reflects that the AFPID controller’s AFPID improvements in the time domain characteristics of voltage and frequency in terms of the ISTSE are 76.64% and 72.53% rather than PID, respectively, and 64.54% and 55.99% rather than FPID. The simulation results on voltage and frequency in the time domain suggested that the AFPID controller improves these characteristics to a greater extent than do the comparison controllers. The ISTSE and overshoot improvements in frequency responses are negative because, in this scenario, the fuzzy PID controller cannot exhibit a suitable performance.
324
H. Shayeghi et al. Best paretto of FPID Optimization with NSMDE in 3Stage 15 stage1: Ceoffs & MFs stage2: Weights stage3: Roles
ISTSE (DV)
10 AHP Selected ISTSE(Dv)=5.0606 , ISTSE(Df)=20.3813
5
0
0
10
20
30
40
50
60
70
80
90
100
ISTSE (Df)
Fig. 13.9 First Pareto (nondominated response) for three-stage controller optimization Table 13.5 Optimized coefficients of controllers by NSIDE algorithm
Parameter K p K I K D α β
Voltage Controller PID AFPID 0.4600 0.0717 1 0.0786 0.4905 0 – 0.5091 – 0.8763
Frequency Controller PID AFPID 0.0114 0.0308 0.0501 0.7473 0.0001 0.6587 – 0.9992 – 0.0703
13.5.3 Scenario 3: 30% Load Decrease In this scenario, 30% of load STNO (see Fig. 13.8) at t = 1 s is reduced. Figures 13.14 and 13.15 show the voltage and frequency responses of the grid in this operating condition of the microgrid, respectively. The other simulation results for this scenario, including those on the voltage and frequency responses to load reduction, are presented in Table 13.11. The reduction in active and reactive loads distorts frequency and voltage; after the occurrence of damping oscillations, frequency and voltage slightly increase. The AFPID controller improves the time domain characteristics of voltage and frequency in terms of the ISTSE to 89.53% and 48.34% rather than PID, respectively, and 80.06% and 34.09% rather than FPID. This result indicates that the proposed controller is superior to the fuzzy PID and PID controllers in terms of achieving system stability.
13 Microgrid Stability Definition, Analysis, and Examples
325
Table 13.6 Optimized membership function parameters x(input) (error) (voltage) x(input) (Δerror) (voltage) x(output) KP (voltage) x(output) KI (voltage) x(output) KD (voltage)
0.0001 0.05679 0.0995 0.1024 0.0219
(
AFPID,
x(input) (error) (voltage) x(input) (Δerror) (voltage) x(output) KP (voltage) x(output) KI (voltage) x(output) KD (voltage)
FPID and
0.2216 0.3520 0.5130 0.9253 0.7264
PID)
Fig. 13.10 Voltage responses of islanded microgrid in scenario 1. (AFPID, FPID, and PID)
13.5.4 Scenario 4: 30% Load Increase In this scenario, the effects of a 30% increase in busload STNO (see Fig. 13.8) produced by bus 15 at t = 1 s is examined. The increased load reduces speed because of an increase in the electric generator’s torque. Figures 13.16 and 13.17 demonstrate the voltage and frequency responses of the microgrid. The time-domain characteristics of this scenario are presented in Table 13.12. After microgrid oscillation damping, frequency considerably decreases. With increasing network reactive power, microgrid voltage is disturbed; efforts to reach sustainability slightly decrease the microgrid voltage. The AFPID controller improves the time domain characteristics of voltage and frequency in terms of the ISTSE to 85.55% and 55.30% rather than PID, respectively, and 70.09% and 35.26% rather than FPID. The results for this scenario are similar to those for scenario 3. The superiority of the AFPID controller in terms of fast and convenient damping of voltage and frequency fluctuations is clearly noticeable. In this scenario, the time domain results of all the three controllers showed that the conventional
e NL NM NS ZR PS PM PL
a) ΔKP(voltage) de NL NM NS – PS PS NS ZR PL NM NS ZR NL – – NL NL PS NL ZR PS ZR PM NL
ZR NS ZR PL ZR ZR PL P
PS PL – ZR – NS ZR PS
PM NL NS PL PS PM PM ZR
PL PL PL PM PL PM PS –
Table 13.7 Optimized rules for voltage controller b) ΔKI(voltage) de NL NL NM – PS PS NS PS PL NM NS ZR NL – – NL NL PL NL PS PL NL NS NL NS ZR PL PM ZR PL ZR NL
ZR PL – PM – PS NS PS
PS PS NL PL PL ZR ZR NM
PM PL PL PS PL PM PS –
PL – NS NM NL NL NL NL
c) ΔKD(voltage) de NL NL NM – NM PM NS ZR NS NM NS ZR NL – – NL NL NM NL NM NM NS PL NL
NS ZR NL PM ZR NS PM NS
ZR PL – ZR – PM PM NS
PS ZR NS PL PL PM PL PM
PM PL PL ZR PL PM PS –
PL – NS NM NL NL NL NS
326 H. Shayeghi et al.
e NL NM NS ZR PS PM PL
a) ΔKP(frequency) de NL NM NS NL NM NM NS ZR PS ZR NS NS NL NL NS NS NL – PM NL ZR NL NL ZR
ZR NL PS PS ZR NL – NL
PS PS NL – PS PS PM NM
PM PL – PM PS PM ZR PL
PL PL NM PS NM NS PS –
Table 13.8 Optimized rules for frequency controller b) ΔKI(frequency) de NL NL NM NL PL ZR ZR ZR PS NL NS NM NL NS NS ZR NL – NM NL NM NL NL ZR NS NL NS PS ZR ZR – NL
ZR NM NM – PS NL NL NM
PS PS – NM PL NL ZR NS
PM NM NM ZR NM PM PS –
PL NL ZR NL NL ZR NM NL
c) ΔKD(frequency) de NL NL NM NL ZR NS PS ZR PS NL NS PS NL NS NS NM NL NM PS NL – NL NL PM
NS PS NS PS ZR NM – NL
ZR NM NM – PS NM NM NM
PS – PL PL PM ZR ZR PL
PM NS PS PL PS PS PS –
PL NL PS NL NL NM PS NL
13 Microgrid Stability Definition, Analysis, and Examples 327
328
H. Shayeghi et al.
(
AFPID,
FPID and
PID)
Fig. 13.11 Frequency responses of islanded microgrid in scenario 1. (AFPID, FPID, and PID) Table 13.9 Time domain characteristics of voltage and frequency in scenario 1
Voltage
Frequency
Parameter ISTSE (pu) Overshoot (pu) Undershoot (pu) Settling time (s) ISTSE (pu) Overshoot (pu) Undershoot (pu) Settling time (s)
Value PID (NSIDE) 26.0163 0.2448 −0.0532 7.8250 50.6243 0.459 −0.0387 14.9450
FPID [37] 69.9972 0.2871 −0.0793 12.1054 80.5385 0.0420 −0.0498 18.7554
AFPID (NSIDE) 111.9122 0.3106 −0.0905 12.8424 182.3305 0.0686 −0.0781 19.4224
Table 13.10 Time domain characteristics of voltage and frequency in scenario 2
Voltage
Frequency
Parameter ISTSE (pu) Overshoot (pu) Undershoot (pu) Settling time (s) ISTSE (pu) Overshoot (pu) Undershoot (pu) Settling time (s)
Value PID (NSIDE) 0.7372 0.0083 −0.0413 5.1550 3.7590 0.0155 −0.0161 9.2150
FPID [37] 2.0792 0.0155 −0.0478 7.6905 8.5422 0.0212 −0.0267 10.8905
AFPID (NSIDE) 3.1344 0.0189 −0.0576 7.8413 13.6843 0.0267 −0.0336 12.2713
13 Microgrid Stability Definition, Analysis, and Examples
329
Fig. 13.12 Voltage responses of islanded microgrid in scenario 2. (AFPID, FPID, and PID)
Fig. 13.13 Frequency responses of islanded microgrid in scenario 2. (AFPID, FPID, and PID)
fuzzy PID exhibits the worst operation; the two other controllers perform well because of the appropriate functioning of the NSIDE algorithm in the optimization of these controller’s parameters. Among the three, however, the AFPID controller remains the most robust and dynamic in performance.
330
H. Shayeghi et al.
Fig. 13.14 Voltage responses of islanded microgrid in scenario 3. (AFPID, FPID, and PID)
(
AFPID,
FPID and
PID)
Fig. 13.15 Frequency responses of islanded microgrid in scenario 3. (AFPID, FPID, and PID)
13.6 Conclusion In this chapter, the operation of an AFPID controller was investigated for the simultaneous control of voltage and frequency in islanded microgrids through a newly established system simulation model. The model was based on a multimachine structure, wherein the relationship between different units was expressed
13 Microgrid Stability Definition, Analysis, and Examples
331
Table 13.11 Time domain characteristics of voltage and frequency in scenario 3
Voltage
Frequency
Parameter ISTSE (pu) Overshoot (pu) Undershoot (pu) Settling time (s) ISTSE (pu) Overshoot (pu) Undershoot (pu) Settling time (s)
(
Value PID (NSIDE) 0.0967 0.0034 −0.0166 5.0950 3.4194 0.0112 −0.0111 10.1050
AFPID,
FPID and
FPID [37] 0.4997 0.0083 −0.0246 5.0991 5.1883 0.0126 −0.0133 9.5876
AFPID (NSIDE) 0.9236 0.0111 −0.0339 5.8740 6.6195 0.0144 −0.0150 9.6740
PID)
Fig. 13.16 Voltage responses of islanded microgrid in scenario 4. (AFPID, FPID, and PID) Table 13.12 Time domain characteristics of voltage and frequency in scenario 4
Voltage
Frequency
Parameter ISTSE (pu) Overshoot (pu) Undershoot (pu) Settling time (s) ISTSE (pu) Overshoot (pu) Undershoot (pu) Settling time (s)
Value PID (NSIDE) 0.0485 0.0115 −0.0024 5.0950 2.9553 0.0103 −0.0105 10.035
FPID [37] 0.1622 0.0138 −0.0046 5.0997 4.5649 0.0126 −0.0109 9.5197
AFPID (NSIDE) 0.3349 0.0200 −0.0065 5.1047 6.6123 0.0155 −0.0129 9.6447
332
H. Shayeghi et al.
(
AFPID,
FPID and
PID)
Fig. 13.17 Frequency responses of islanded microgrid in scenario 4. (AFPID, FPID, and PID)
through an admittance matrix. The AFPID controller functions as a secondary control scheme that covers all conditions and damps oscillations in the voltage and frequency of actual microgrids. The proposed control strategy was equipped with a constant-gain PID controller and a fuzzy inference system, which adjusts PID gains. The AFPID parameters, fuzzy rule weights, and rule bases were automatically tuned to minimize design efforts and identify an improved fuzzy system control through the developed NSIDE optimization algorithm. The simulation results indicated that tuning the fuzzy rule weights eliminates some fuzzy rules and reduces the need to optimize fuzzy rule variables. These effects, in turn, advance improved optimization results. The simulation results for different disturbance scenarios reflected the superior performance of the proposed controller over the conventional fuzzy and classical PID controllers. In all the scenarios, the time domain characteristics achieved by the AFPID controller were superior to those generated by the two others. The microgrid modeling based on the multi-machine system modeling showed that applying a variety of changes in microgrid loads and the optimal design of the AFPID controller is easily possible. It also confirmed the accuracy of the proposed method. The advantages of the proposed method are summarized as follows: • The newly developed model of dynamic microgrid behavior considers all kinds of distributed generators and energy storage devices. • Different faults and load changes are easily considerable in system performance analysis using the developed model.
13 Microgrid Stability Definition, Analysis, and Examples
333
• Because of the gain adjustment for the PID controller, the AFPID controller generates suitable simultaneous voltage and frequency responses against different faults and load uncertainties. The results presented in Sect. 5 indicated that the conventional fuzzy and classical PID controllers exhibit unsuitable and unacceptable performance in all the islanded microgrid operation scenarios. • The newly developed controller based on a fuzzy system can satisfactorily function in all operating conditions. The ISTSE performance index was calculated to ensure the accuracy and speed of voltage and frequency responses, thereby reinforcing the validity of the controller.
References 1. Khalghani, M., Khooban, M., Moghaddam, E., Vafamand, N., & Goodarzi, M. (2016). A self - tuning load frequency control strategy for microgrids: Human brain emotional learning. International Journal of Electrical Power Energy Systems, 75(2), 311–319. 2. Sanjarani, M., & Gharehpetian, G. (2014). Game-theoretic approach to cooperative control of distributed energy resources in islanded microgrid considering voltage and frequency stability. Neural Computing and Applications, 25(2), 343–351. 3. Palizban, O., Kauhaniemi, K., & Guerrero, J. (2014). Microgrids in active network management-part I: Hierarchical control, energy storage, virtual power plants, and market participation. Renewable and Sustainable Energy Reviews, 36(7), 428–439. 4. Schiffer, J., Ortega, R., Astolfi, A., Raisch, J., & Sezi, T. (2014). Conditions for stability of droop controlled inverter based microgrids. Automatica, 50(10), 2457–2469. 5. Bidram, A., & Davoudi, A. (2012). Hierarchical structure of microgrids control system. IEEE transactions on smart grids, 3(4), 1963–1976. 6. Lu, L., & Chu, C. (2015). Consnsus based secondary frequency and voltage droop control of virtual synchronous generators for isolated AC micro grids. IEEE Journal on Emerging Selected Topics in Circuits and Systems, 5(3), 443–455. 7. Wang, Y., Pordanjani, I., & Li, W. (2011). Strategy to minimize the load shedding amount for voltage collapse prevention. IET Generation, Transmission and Distribution, 5(3), 307–313. 8. Gu, W., & Zhu, J. (2014). Adaptive decentralized under frequency load shedding for islanded smart distribution networks. IEEE Transactions on Transmission, Sustain Energy, 5(3), 886– 895. 9. Nourollah, S., Piraysh, A., & Aminifar, F. (2016). Combinational scheme for voltage and frequency recivery in an island distribution system. IET Generation, Transmission and Distribution, 10(12), 2899–2906. 10. H. Laaksonen, P. Saari, R. Komulainen, “Voltage and frequency control of inverter based LV network microgrid”, in Future power systems, 2005 international conference on, 2005. 11. Li, Y., & Li, Y. (2011). Power management of inverter interfaced autonomous microgrids based on virtual frequency voltage frame. IEEE transactions on smart grids, 2(1), 30–40. 12. Mari, M. I., & Soliman, M. H. (2013). A coordinated voltage and frequency control of inverter based distributed generation and distributed energy storage system for autonomous microgrids. Electric Power Components and Systems, 41(4), 383–400. 13. B. Shoeiby, R. Davoodnezhad, D. Holms, B. McGrath, “Voltage - frequency control of an islanded microgrids using the intrinsic droop characteristics of resonant current regulators”, in Energy conversion congress and exposition, 2014. 14. Valencia, F., Collado, J., Saez, D., & Martin, L. (2016). Robust energy management system for a microgrid based on a fuzzy prediction interval model. IEEE Transactions on Smart Grid, 7(3), 1486–1494.
334
H. Shayeghi et al.
15. Kim, Y., Kim, E., & Moon, S. (2015). Frequency and voltage control strategy of standalone microgrids with high penetration of intermittent renewable generation systems. IEEE Transactions on Power Systems, 31(1), 718–728. 16. Bayat, M., Sheshyekani, K., Hamzeh, M., & Rzazadeh, A. (2015). Coordination of distributed energy resources and demand response for voltage and frequency support of MV microgrids. IEEE Transactions on Power Systems, 31(2), 1506–1516. 17. Tang, X., Hu, X., Li, N., Deng, W., & Zhang, G. (2016). A novel frequency and voltage control method for islanded microgrid based on multi energy storage. IEEE Transactions on Smart Grids, 7(1), 410–419. 18. Kunjumhummed, L. (2013). A power system dynamics simulation program using matlab/ Simulink. International Journal of Advanced Research in Electrical, Electronics and Instrumentation Engineering, 2(1), 111–121. 19. Schiffer, J., Zonetti, D., Ortega, R., Stancovic, A., Sezi, T., & Raisch, J. (2016). A survey on modeling of microgrids- from fundamental physics to phasors and voltage sources. Automatica, 74(12), 135–150. 20. Hassan, M., & Abido, M. (2014). Real time implementation and optimal design of autonomous microgrids. Electric Power Systems Research, 109(4), 118–127. 21. S. Pati, K. Mohanty, S. Kar, S. Mishra, “A sliding mode controller based STATCOM with battery storage for voltage and frequency stabilization in a micro grids”, in 2016 International conference on circuit, power and computing technologies, 2016. 22. Vachirasricirikul, S., & Nagmaroo, I. (2012). Robust controller design of micro turbine and electrolyzer for frequency stabilization in a microgrid system with plug-in hybrid electric vehicles. International Journal of Electrical Power & Energy Systems, 4(1), 804–811. 23. Bevrani, H., Feizi, M., & Ataeee, S. (2016). Robust frequency control in an islanded micro grid: H and u synthesis approach. IEEE Transactions on Smart Grid, 7(2), 706–717. 24. Ahmadi, S., Shokoohi, S., & Bevrani, H. (2015). Fuzzy logic based droop control for simultaneous voltage and frequency regulation in an AC microgrids. International Journal of Electrical Power & Energy Systems, 64(1), 148–155. 25. A. Perdana, “Dynamic models of wind turbines”, Thesis of PhD., Goteborg, Sweden, 2008. 26. Abbas, F., & Abdulsada, A. (2010). Simulation of wind turbine speed control by MATLAB. International Journal of Computer and Electrical Engineering, 2(2), 912–915. 27. Mahat, P., Chen, Z., & Jensen, B. (2011). Control and operation of distributed generation in distribution systems. Electric Power Systems Research, 81, 495–502. 28. Y. Yu, “Electric power system dynamics”, 1983. 29. Robert, J., Ian, A., & Hiskens, A. (1997). Lyapunov functions for multi machine power systems with dynamic loads. IEEE Transactions on Circuits and Systems I: Fundamental Theory and Applications, 44, 796–812. 30. Zamani, A., Bijami, E., sheikholeslami, F., & jafrasteh, B. (2014). Optimal fuzzy load frequency controller with simultaneous auto-tuned membership functions and fuzzy control rules. Turkish Journal of Electrical Engineering & Computer Sciences, 22, 66–86. 31. Storn, R., & Price, K. (1997). Differential evolution-a simple and efficient heuristic for global optimization over continuous spaces. Journal of Global Optimization, 11, 341–359. 32. A. Musrrat, P. Millie, A. Ajith, “A modified differential evolution algorithm and its application to engineering problems”, in International conference of soft computing and pattern recognition, 2009. 33. Singh, H., & Srivastava, L. (2014). Modified differential evolution algorithm for multiobjective VAR management. Electrical Power and Energy Systems, 55, 731–740. 34. Deb, K., Pratap, A., Agarwal, S., & Meyarivan, T. (2002). A fast and elitist multi objective genetic algorithm: NSGA-II. IEEE Transactions on Evolutionary Computation, 6, 182–197. 35. Handfield, R., Walton, S., Sroufe, R., & Melnyk, S. (2002). Applying environmental criteria to supplier assessment: A study in the application of the analytical hierarchy process. European Journal of Operational Research, 141(1), 70–87. 36. Mahat, P., Chen, Z., & Jensen, B. (2011). Control and operation of distributed generation in distribution systems. Electric Power Systems Research, 81, 495–502.
13 Microgrid Stability Definition, Analysis, and Examples
335
37. Pradhan, P., Sahu, R., & Panda, S. (2016). Firefly algorithm optimized fuzzy PID controller for AGC of multi-area multi-source power systems with UPFC and SMES. Engineering Science and Technology, 19(1), 338–354. 38. Shayeghi, H., & Aryanpour, H. (2016). Robust online fuzzy PID design based on improved differential evolution for islanded microgrid frequency control considering nonlinear factors and uncertainties. Tabriz Journal of Electrical Engineering, 46, 241–256. 39. Ghasemi, A., Shayeghi, H., & Alkhati, H. (2013). Robust Design of Multi machine power system stabilizers using fuzzy gravitational search algorithm. International Journal of Electrical Power & Energy Systems, 51, 190–200. 40. Shayeghi, H., & Alilou, M. (2015). Application of multi objective HFAPSO algorithm for simultaneous placement of DG, capacitor and protective device in radial distribution network. Journal of Operation and Automation in Power Engineering, 3, 131–146. 41. Shayeghi, H., Shayanfar, H. A., & Jalili, A. (2007). Multi stage fuzzy PID load frequency controller in a restructured power system. Journal of Electrical Engineering, 58, 61–70. 42. Shayeghi, H., Ghasemi, A., & Shayanfar, H. A. (2011). PID type stabilizer Design for Multi Machine Power System Using IPSO procedure. Computer Science and Engineering, 1, 36–42. 43. Hashemi, Y., Shayeghi, H., & Moradzadeh, M. (2017). Design of Dual-Dimensional Controller Based on multi-objective gravitational search optimization algorithm for amelioration of impact of oscillation in power generated by large-scale wind farms. Applied Soft Computing, 53, 236–261.
Chapter 14
Voltage Unbalance Compensation in AC Microgrids Shahram Karimi, Mehdi Norianfar, and Josep M. Guerrero
14.1 Introduction Nowadays, the continuity of the electricity supply and the voltage quality with considering the environmental and economic aspects are the important issues related to the power distribution operation. Adopting the distributed generations (DGs) in the electrical distribution networks are the main strategy of the electrical engineers to guarantee the continuous supply of the local and sensitive loads, improving the power quality, decreasing the emission of greenhouse gases, through applying renewable energy resources, and decreasing the power losses. The presence of DGs in electrical distribution systems has introduced a new structure called microgrid (MG) [1]. MGs can operate in grid-connected or islanded mode. In the islanded condition, there is a greater probability of the power quality problems due to the lack of the main grid support. Power quality can be handled for harmonics, voltage unbalance, interruptions, sags, swells, and transients [2]. This chapter focuses on the harmonics and voltage unbalance disturbances as the main problems of power quality in the steady state. Harmonics and voltage unbalance can have negative impacts on the customer equipment and the network performance if they exceed the permissible limit. So, satisfying the constraints of the total harmonic distortion (THD) and voltage unbalance factor (VUF) is noteworthy due to the negative effects of these disturbances on the sensitive loads. The IEEE and IEC standards recommend the limitations for the power quality parameters [3]. An example of these standards is given in Table 14.1. S. Karimi () · M. Norianfar Department of Electrical Engineering, Faculty of Engineering, Razi University, Kermanshah, Iran e-mail: [email protected] J. M. Guerrero Center for Research on Microgrids (CROM), Department of Energy Technology, Aalborg University, Aalborg, Denmark © Springer Nature Switzerland AG 2021 A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems, https://doi.org/10.1007/978-3-030-59750-4_14
337
338 Table 14.1 IEE voltage distortion limits [6]
S. Karimi et al. Bus voltage VPCC V < 69 kV 69 kV ≤ V < 161 kV 161 kV ≤ V
Individual harmonics (%) 3.0 1.5 1.0
THD (%) 5.0 2.5 1.5
The unbalanced voltage and harmonics can be compensated by using equipment such as active power filters, dynamic voltage restorers (DVR), and static synchronous compensator (STATCOM). However, all these solutions impose additional costs. Therefore, researchers used the surplus capacity of the inverter-interfaced DGs, recently, to improve the power quality in order to reduce the additional costs. In fact, the inverter-interfaced DGs are capable of performing multiple tasks at a time, the so-called multifunctional [4, 5]. Indeed, one can use an appropriate control method in the control system of inverter-interfaced DGs to improve the power quality at the consumer’s side. This chapter involves the control methods proposed to compensate unbalanced voltage and harmonics The performance of these methods is shown analytically and its effectiveness and feasibility are illustrated at the end of the chapter by simulating a simple islanded MG.
14.2 Inverter-Interfaced DG In this chapter, we study the general structure of a three-phase inverter-interfaced DG based on voltage source converter (VSC) allowing to compensate the current harmonics and imbalance of a three-phase three-wire electrical network. The block diagram of a grid-connected VSC is shown in Fig. 14.1.
14.2.1 VSC Control As shown in Fig. 14.1, the control part of a VSC generally performs three main functions: • Identification of the reference currents, • Current controller, • DC voltage regulation. One of the control blocks in the VSC control system is the DC voltage regulation which is outside the scope of this chapter. Reference currents need to be well identified to reduce the current harmonics and compensate the voltage imbalance. In the following sections, current identification methods for reference currents generation will be analyzed and described in detail.
14 Voltage Unbalance Compensation in AC Microgrids
339
Electrical network Linear and non-linear load
Output filter
Identification of the reference current
Current controller
Regulation of the DC voltage. Control part
Energy resource Inverter
Power part
Fig. 14.1 General structure of a VSC
14.3 Harmonic Compensation The harmonic loads and the nonlinear nature of the converters are the main factors in producing the harmonic currents. Therefore, it is not possible to avoid harmonic voltage and current in the microgrids. However, as mentioned earlier, THD should be limited in an acceptable range. There are various methods to reduce harmonic distortion. The harmonic distortion caused by the output voltage of the converters can be reduced by choosing a higher switching frequency. But increasing the switching frequency will increase the power losses. As a result, the frequency of the converter, especially for high powers, cannot be selected too large. Another common solution is to use an LC or LCL filter. By using these common methods and since the inverters are used in the microgrid at their linear operating point, the problem of harmonic distortion resulting from them is largely solved. For harmonic compensation due to nonlinear loads, in the simplest case, a low-pass filter can be used. Since harmonics are close to the fundamental component, for appropriate filtering, the filter bandwidth should be small enough, which affects the fundamental component. The following is a present active solution to this issue.
340
S. Karimi et al.
14.3.1 Identification of Reference Currents The quality of the current harmonics compensation strongly depends on the performance of the chosen identification technique. In fact, a control system, even a very efficient one, cannot perform satisfactory filtering if the harmonic currents are badly identified. Hence, various identification techniques have been proposed and extended. They can be grouped according to two approaches [7, 8]. • Identification in the frequency domain. This type of approach uses the fast Fourier transform, for extracting the harmonics of the load current. This approach is particularly suitable for loads that harmonic content varies slowly. It also has the advantage of selecting each harmonic individually and thus only compensates for the preponderant harmonic currents. However, this method requires heavy calculations in order to identify the harmonic currents. For this reason, frequency methods are not used in practice since the harmonic currents can generally vary rapidly over time. • Identification in the time domain. To identify the harmonic currents, various methods in the time domain have been proposed [7, 9–11]. The basic methods will be the subject of the following paragraphs.
14.3.1.1
Instantaneous Active and Reactive Power Method
The instantaneous active and reactive power method (generally so-called pq method) was initially extended in [9]. In this method, the instantaneous active and reactive powers are calculated using the Concordia transformation. After transformation, the instantaneous powers include a DC component and AC components. The DC component is related to the fundamental frequency of the load current and the AC components are associated with the harmonic components of the load current. By eliminating the DC component of the instantaneous active power through a low-pass filter (LPF), the harmonic components can be identified. The basis of pq method is presented in the following. Let vs1 , vs2 , vs3 and ic1 , ic2 , ic3 be the phase voltages of a three-phase network without homopolar, and the load currents, respectively. The transformation of Concordia makes it possible to bring this balanced three-phase system to a twophase system that axes are quadrature: ⎡ ⎡ ⎣
iα iβ
⎤ ⎦=
0
2 3
1 −√12 1− √2 0 23 − 23
ic1
⎤
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ic2 ⎥ ⎢ ⎥ ⎣ ⎦ ic3
(14.1)
14 Voltage Unbalance Compensation in AC Microgrids
341
⎡ ⎡ ⎣
vα
⎤
0
⎦=
vβ
2 3
1 −√12 1− √2 0 23 − 23
vs1
⎤
⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎢ vs2 ⎥ ⎥ ⎢ ⎦ ⎣ vs3
(14.2)
The instantaneous active power p and the instantaneous reactive power q are defined by: ⎡ ⎣
p
⎤
⎡
⎦=⎣
−q
vα
vβ
⎤⎡ ⎦⎣
− vβ vα
iα
⎤ ⎦
(14.3)
iβ
The instantaneous active and reactive powers can be written as follows: ⎡ ⎤ ⎡ ∼⎤ p p+p ⎥ ⎣ ⎦=⎢ ⎣ ⎦
(14.4)
∼
q+q
q
∼
∼
wherein p and q are the DC components of p and q, and p and q are the AC components of p and q. From eq. (14.3), we can deduce the expressions of the components of the load current along the axes αβ: ⎡ ⎣
iα
⎤
⎡
⎦=⎣
vα
vβ
⎦
− vβ vα
iβ
⎤ ⎡ ⎤⎡ vα −vβ p 1 ⎦= ⎦ ⎣ ⎣ ⎦⎣ vα2 + vβ2 −q −q vβ vα
⎤−1 ⎡
p
⎤
(14.5)
By replacing (14.4) in (14.5), these currents are expressed along the axes αβ by: ⎡ ⎣
iα iβ
⎤ ⎦=
⎡ 1 ⎣ vα2 + vβ2
vα −vβ vβ
vα
⎤⎡ ⎦⎣
p −q
⎤ ⎦+
⎡ 1 ⎣ vα2 + vβ2
vα −vβ vβ
vα
⎤⎡ ∼ ⎤ p ⎥ ⎦⎢ ⎦ ⎣ ∼ −q (14.6)
Depending on the control objectives, one can simultaneously compensate for the current harmonics and the reactive power, or just the current harmonics compensation is selected as the control objective. If for example the current harmonics and the reactive power compensation are considered as the control objectives, the DC component of p is eliminated by using a LPF. Then, the active power pc (necessary for the regulation of the DC voltage vdc ) is added to the AC component of the instantaneous active power. The reference currents along the αβ axes are obtained as follows:
342
S. Karimi et al.
Regulator
LPF
abc Eq. 3
αβ
−
Eq. 7
αβ abc
abc αβ
Fig. 14.2 Reference currents generation by using the instantaneous powers method
⎡ ⎣
iref α irefβ
⎤ ⎦=
⎡ vα2
1 ⎣ + vβ2
vα −vβ vβ
vα
⎤ ⎤⎡∼ p + pc ⎥ ⎦⎢ ⎣ ⎦ −q
(14.7)
Finally, by the inverse transformation of Concordia, the reference currents along the abc axes are achieved: ⎤ ⎡ iref 1 ⎡ ⎤⎡ ⎤ ⎥ 0 ⎢ 1 0 iref α ⎥ ⎢ √ 2 ⎥ ⎢ ⎢ 1 3 ⎥⎣ ⎦ (14.8) ⎢ iref 3 ⎥ = ⎣−2 2√ ⎦ ⎥ ⎢ 3 3 1 ⎦ ⎣ irefβ −2 − 2 iref 3 Figure 14.2 illustrates the reference currents generation when the instantaneous powers method is used to simultaneously compensate for the current harmonics and the reactive power.
14.3.1.2
Synchronous Reference Frame (SRF) Method
This method, introduced in [12], also exploits the transformation of Concordia but it is applied only to load currents ic1 , ic2 , and ic3 . A second transformation is then carried out to obtain the load currents along the dq axes. This makes it possible to transform the fundamental component of the load current into a DC component and the harmonic components of the load current into AC components. The DC component of the load current can then be eliminated by using a LPF. The advantage of the SRF method compared to the pq method lies in the fact that the voltage harmonics no longer have a significant influence on the identified currents. The principle of the SRF method is described in the following. Let ic1 , ic2 , and ic3 be the load currents of a three-phase system without a zero sequence component. The transformation of Concordia makes it possible to obtain
14 Voltage Unbalance Compensation in AC Microgrids
343
the load currents in the stationary reference frame, as illustrated previously by the relation (14.1). By generating the signals cos θˆ and sin θˆ from the network voltage, using a phase-locked loop (PLL), the load currents in the dq reference frame are obtained: ⎤ ⎡ ⎤ sin θˆ − cos θˆ ⎢ ⎥ iα ⎥⎣ ⎦ ⎣ ⎦=⎢ ⎣ ⎦ iq iβ cos θˆ sin θˆ ⎡
id
⎤
⎡
(14.9)
with θˆ the angular position of the fundamental network voltage, estimated by the PLL. These components can then be expressed as the sum of a DC component and an AC component: ⎡ ⎣
id
⎤
⎡
⎦=⎢ ⎣
iq
∼
id + id
∼ iq + iq
⎤ ⎥ ⎦
(14.10)
∼
∼
with id and iq the DC components of id and iq , and id and iq the AC components of id and iq . From (14.9), the load currents in the stationary reference frame can be expressed as follows: ⎤−1 ⎤ ⎡ ⎡ ⎤ ⎡ ⎤ ˆ sin θˆ − cos θˆ sin θ cos θˆ i d ⎥ ⎥ id ⎢ ⎢ ⎥ ⎣ ⎦=⎢ ⎥⎣ ⎦ ⎣ ⎦=⎢ ⎣ ⎣ ⎦ ⎦ iβ iq iq cos θˆ sin θˆ − cos θˆ sin θˆ ⎡
iα
⎤
⎡
(14.11) Or again: ⎤ ⎤⎡ ⎤ ⎡ ∼ ⎡ ⎤ ˆ sin θˆ cos θˆ sin θ cos θˆ i ⎥ d ⎥ ⎢ id ⎥ ⎢ ⎢ ⎥⎣ ⎦ + ⎢ ⎥ ⎣ ⎦=⎢ ⎣ ⎣ ⎦ ⎦⎣ ∼ ⎦ iβ iq − cos θˆ sin θˆ − cos θˆ sin θˆ iq ⎡
iα
⎤
⎡
(14.12) Considering the simultaneous compensation of the current harmonics and the reactive power, after adding the current icd necessary for the regulation of the DC ∼
voltage vdc , to the AC component id , the Eq. (14.12) becomes:
344
S. Karimi et al.
Regulator
LPF
abc Eq. 9
αβ
−
Eq. 13
αβ abc
PLL Fig. 14.3 Generation of reference currents by the SRF method
⎤⎡ ⎤ ∼ sin θˆ cos θˆ i + i cd ⎥⎢ d ⎢ ⎥ ⎥ ⎣ ⎦=⎢ ⎦ ⎣ ⎦⎣ irefβ iq − cos θˆ sin θˆ ⎡
iref α
⎤
⎡
(14.13)
Then, using (14.8) the reference currents in the abc reference frame are obtained. Figure 14.3 illustrates the generation of reference currents for the simultaneous compensation of the current harmonics and the reactive power by the SRF method.
14.3.2 Performance of Pq and SRF Methods in Ideal and Nonideal Conditions In this subsection, the comparative analytical study of the performance of pq and SRF methods in ideal and nonideal conditions is presented. To conduct this study, we will consider the classic example of a three-phase network connected to a threephase thyristor rectifier bridge. The rectifier current ic can be represented using Fourier series according to the following equation: √ ic = 2I1
1 sin [11 (ωt−α)] + sin (ωt−α) − 15 sin [5 (ωt−α)] − 17 sin [7 (ωt−α)] + 11 1 sin [13 (ωt−α)] − 1 sin [17 (ωt−α)] − 1 sin [19 (ωt−α)] + . . . 13 17 19
(14.14)
with α the thyristor commutation angle and I1 the effective value of the fundamental of current defined by: √ I1 = where Id is the load average current.
6 Id π
(14.15)
14 Voltage Unbalance Compensation in AC Microgrids
345
As seen in (14.14), the harmonic currents are of rank (h = 6 k ± 1) with k integer, and that the effective value of each harmonic current is inversely proportional to its rank (Ih = I1 /h).
14.3.2.1
Case of the pq Method
Ideal Case: Balanced Harmonic Currents and Balanced Sinusoidal Voltages In the ideal case, the nonlinear load currents are balanced and defined by the Eq. (14.14). As for the network voltages, they are assumed to be sinusoidal and balanced: ⎡
⎤
⎡
vs1 ⎥ ⎢ ⎥ √ ⎢ ⎥ ⎢ ⎢ vs2 ⎥ = 2Vs ⎥ ⎢ ⎦ ⎣ vs3
sin (ωt)
⎤
⎢ ⎥ ⎢ ⎥ ⎢ 2π ⎥ ⎢ sin ωt − 3 ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎣ ⎦ sin ωt + 2π 3
(14.16)
By applying the Concordia transformation, the currents and voltages along the axes αβ are obtained: ⎡ ⎣
vsα vsβ
⎡
⎤ ⎡ sin (ωt) √ ⎦ ⎦ = 3Vs ⎣ − cos (ωt) ⎤
(14.17)
⎤ ⎤ ⎤ ⎞ ⎛⎡ ⎡ ⎡ sin (ωt) sin (ωt) sin (ωt) √ ⎣ ⎦ = 3I1 ⎝⎣ ⎦ −1 ⎣ ⎦− 1⎣ ⎦ + ...⎠ 5 7 − cos (ωt) − cos (ωt) − cos (ωt) iβ (14.18) iα
⎤
The instantaneous active power, p, can be calculated from (14.3), (14.17), and (14.18) and is defined by Eq. (14.19): p = 3Vs I1 cos α +
3Vs I1 3Vs I1 cos (6ωt − 5α) − cos (6ωt − 7α) + . . . 5 7 (14.19)
The first term of this equation represents the average active power, noted p, related to the fundamental of the current, while the sum of the other terms represents ∼ an AC power, noted p, generated by the harmonic currents. From (14.19), we can
346
S. Karimi et al.
h 0
1
5
6
7
11
12
13
17 18 19
Harmonic rank Fig. 14.4 Harmonic components of the instantaneous active power in the ideal conditions
distinguish the harmonic components of the instantaneous active power p. Figure 14.4 shows these harmonic components and their origins. A similar figure can be obtained for instantaneous reactive power. As seen in Fig. 14.4, the current harmonics of ranks (h = 6 k ± 1) give rise to alternative powers with multiple pulses of 6. In this case, to generate the reference currents, it suffices to filter the DC components of p and q.
Balanced Harmonic Currents and Unbalanced Sinusoidal Voltages The network voltages are now considered to be unbalanced and defined by the following matrix relation: ⎡
⎤
vs1 ⎢ ⎥ ⎢ ⎥ √ ⎢ ⎥ ⎢ vs2 ⎥ = 2V + ⎢ ⎥ ⎣ ⎦ vs3
⎡
sin (ωt)
⎤
⎢ ⎥ ⎢ ⎥ ⎢ 2π ⎥ ⎢ sin ωt − 3 ⎥ √ − ⎢ ⎥ + 2V ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎣ ⎦ sin ωt + 2π 3
⎡
sin (ωt)
⎢ ⎢ ⎢ ⎢ sin ωt + 2π 3 ⎢ ⎢ ⎢ ⎢ ⎣ sin ωt − 2π 3
⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ (14.20)
where V+ and V− represent the positive and negative sequence components of the network voltage, respectively. The load current has been defined according to (14.14). By applying the Concordia transformation, we obtain along the αβ axes the currents already defined in (14.18) and the following voltages: ⎡ ⎣
vsα vsβ
⎤ ⎤ ⎡ ⎡ sin (ωt) sin (ωt) √ − √ + ⎦ + 3V ⎣ ⎦ ⎦ = 3V ⎣ − cos (ωt) cos (ωt) ⎤
(14.21)
In this case, we will also establish the analytical expression of the active power p and generalize the results obtained to the reactive power. Active power can be calculated from Eqs. (14.3), (14.18), and (14.21) and is expressed according to the relation (14.22):
14 Voltage Unbalance Compensation in AC Microgrids
347
h 0
1
2
4
5
6
7
8
10 11 12 13 14
16 17 18 19
Harmonic rank Fig. 14.5 Harmonic components of the instantaneous active power (Balanced harmonic currents and unbalanced sinusoidal voltages)
+
p = 3V I1 cos α +
3V + I1 3V − I cos (6ωt − 5α) − 5 1 cos (4ωt − 5α) 5 3V + I1 3V − I1 − 7 cos (6ωt − 7α) + 7 cos (8ωt−7α) + . . .
3V − I1 cos (2ωt − α) +
(14.22)
The first term of this equation represents the average active power, while the sum of the other terms represents an AC power originating from the composition of the harmonic currents and the positive and negative sequence voltages. From (14.22), we obtain the harmonic components of the instantaneous active power. Figure 14.5 shows these harmonic components and their origins. In Fig. 14.5, we find in red the contribution of the harmonic currents of the load, composed with the positive sequence voltage (see Fig. 14.4). Furthermore, the composition of the negative sequence voltage with the fundamental of the load current generates an AC power 2ω. In addition, the negative sequence voltage composed with harmonic currents induces alternative powers of (4ω, 8ω, 10ω, 14ω, 16ω, . . . ). An identical result can be obtained for instant reactive power. From these analyzes, it appears that the unbalanced voltages are at the origin of the harmonic component of rank 2 of the instantaneous powers. This can lead to an incorrect value when identifying the reference currents because a LPF (conventionally used in the pq method) is not effective for eliminating this harmonic component close to its cutoff frequency and has a significant residue.
14.3.2.2
Case of the SRF Method
The performance of the SRF method strongly depends on the performance of the PLL implemented and intended to generate sine and cosine signals, synchronous with the network. In this subsection, we will analyze the behavior of the conventional PLL under different conditions: balanced or unbalanced sinusoidal or non-sinusoidal source voltages. The block diagram of conventional PLL is presented in Fig. 14.6.
348
S. Karimi et al. ∧
∧
Regulator
−
⁄
−
∧
∧
abc αβ Fig. 14.6 Structure of PLL
Balanced Sinusoidal Source Voltages In this case, the network voltages along the axes αβ are defined by (14.17). The expression of these voltages along the axes dq is then as follows: ⎤ ⎤ ⎡ cos θˆ sin θˆ ⎢ ⎥ vsα ⎥⎣ ⎦=⎢ ⎦ ⎣ ⎣ ⎦ vsq vsβ − sin θˆ cos θˆ ⎡
vsd
⎤
⎡
(14.23)
From (14.17) and (14.23), we obtain: vsd =
√ 3Vs sin θ − θˆ
(14.24)
with θ = ωt. Knowing that the value of (θ − θˆ ) is small, (14–24) can be approximated by: vsd ≈
√ 3Vs θ − θˆ
(14.25)
The eq. (14.25) shows that in the case where the network voltages are sinusoidal and balanced, an efficient regulator allows the PLL to accurately estimate the angular position.
Unbalanced Sinusoidal Source Voltages In this case, the network voltages along the axes αβ are defined by (14.21). From eqs. (14.21) and (14.23), we get:
14 Voltage Unbalance Compensation in AC Microgrids
vsd =
√
349
√ 3V + sin θ − θˆ + 3V − sin θ + θˆ
(14.26)
The first term in (14.26) represents the influence of the positive sequence component of the voltage, while the second term represents the influence of its negative sequence component. Considering (14.26), we can conclude that vsd , and therefore the angular position estimated by the PLL, will both be affected by the negative sequence component of the voltage.
Balanced Source Voltages Containing Harmonics In this case, we consider that the network voltages contain harmonics. They can therefore be written:
vs1 vs2 vs3 ⎛⎡
sin (ωt) ⎜⎢ ⎜⎢ ⎜⎢ ⎜ ⎢ sin ωt − 2π √ ⎜⎢ 3 = 2Vs ⎜ ⎢ ⎜⎢ ⎜⎢ ⎜⎢ ⎝⎣ sin ωt + 2π 3
⎤
⎡ 1 sin (5ωt) 5 ⎥ ⎢ ⎥ ⎢ + , ⎥ ⎢ 1 2π ⎥ ⎢ ⎥ ⎢ 5 sin 5 ωt − 3 + ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎦ ⎣ + , 1 sin 7 ωt + 2π 5 3
⎤
⎡ 1 sin (7ωt) 7 ⎥ ⎢ ⎥ ⎢ + , ⎥ ⎢ 1 2π ⎥ ⎢ ⎥ ⎢ 7 sin 7 ωt − 3 + ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎦ ⎣ + , 1 sin 7 ωt + 2π 7 3
⎤
⎞
⎥ ⎟ ⎥ ⎟ ⎥ ⎟ ⎥ ⎟ ⎥ ⎟ + . . . ⎥ ⎟ ⎥ ⎟ ⎥ ⎟ ⎥ ⎟ ⎦ ⎠
(14.27)
By applying the Concordia transformation, we obtain the following equation: ⎡ ⎣
vsα vsβ
⎤ ⎤ ⎤ ⎞ ⎛⎡ ⎡ ⎡ sin (ωt) sin (5ωt) sin (7ωt) √ 1 1 ⎦+ ⎣ ⎦+ ⎣ ⎦+...⎠ ⎦ = 3Vs ⎝ ⎣ 5 7 − cos (ωt) cos (5ωt) cos (7ωt) (14.28) ⎤
From (14.28), one can deduce: vsd
√ = 3Vs sin θ − θˆ +
√ √3V 3Vs s ˆ sin 5θ + θ + sin 5θ − θˆ + . . . 5 7 (14.29)
As seen from (14.29), the voltage harmonics induce alternative components that affect the performance of the conventional PLL. Some solutions are introduced to improve the performances of PLL in the nonideal conditions which are beyond the reach of this chapter. However, interested readers can refer to the references [13– 16].
350
S. Karimi et al.
14.4 Imbalance Compensation The voltage unbalance has negative impacts on the equipment such as induction motors, electronic converters, and adjustable speed drives. Additionally, the voltage unbalance causes more power losses in the distribution networks. Therefore, to compensate for the unbalanced voltage, the negative sequence component of the unbalanced loads must be effectively compensated. In order to achieve this purpose, it is necessary to properly extract the negative sequence component of the unbalanced loads. In the literature, various methods have been presented for controlling the power converters under unbalanced conditions, such as resonant controllers [17, 18], hysteresis current controllers [19, 20], direct power control methods [21, 22], and model-based predictive controllers [23]. In this section, however, some methods that are introduced for extracting of the positive and negative sequence components are briefly described. Generally, positive and negative sequence extraction could be grouped into two main approaches: direct extraction methods and indirect extraction methods. In direct extraction methods, the instantaneous positive and negative sequences are directly extracted by using Lyon transform. In indirect extraction methods, the instantaneous positive and negative sequences are estimated by using PLL or Frequency Locked-Loop (FLL).
14.4.1 Direct Extraction Methods 14.4.1.1
In Abc Frame
Direct extraction methods are based on the symmetric components, which are performed in the frequency domain by the Fortescue transform and in the time domain by the Lyon transform [24, 25]. The symmetrical components can decompose the steady-state phasors of a generic three-phase system into the positive, negative, and zero sequence components. Using this principle, the positive, negative, and zero sequence phasors of phase a of unbalanced currents can be calculated by the Fortescue transform as follows: − → I +−0(a)
⎤ ⎡− → → ⎤ ⎤⎡− ⎡ I +(a) I a 1 a a2 − → → → ⎥ ⎥ 1 ⎢− ⎢− = [T+−0 ] I abc ; ⎣ I −(a) ⎦ = ⎣ 1 a 2 a ⎦ ⎣ I b ⎦ 3 − → − → 1 1 1 I 0(a) I c
(14.30)
where a = ej2π /3 = 1 120◦ is Fortescue operator. Lyon extended the work of Fortescue and applied the symmetrical components in the time domain. Using this principle, the instantaneous positive and negative sequence components of the following set of three-phase unbalanced sinusoidal waveforms are given by
14 Voltage Unbalance Compensation in AC Microgrids
351
⎡
iabc
⎤ ia + − 0 = ⎣ ib ⎦ = iabc + iabc + iabc ic
(14.31)
⎡
+ iabc
⎤ ⎤⎡ ⎤ ⎡ ia+ ia 1 a a2 1 = [T+ ] iabc ; ⎣ ib+ ⎦ = ⎣ a 2 1 a ⎦ ⎣ ib ⎦ 3 ic+ ic a a2 1
(14.32)
⎡
− iabc
14.4.1.2
⎤ ⎤⎡ ⎤ ⎡ ia− ia 1 a2 a 1 = [T− ] iabc ; ⎣ ib− ⎦ = ⎣ a 1 a 2 ⎦ ⎣ ib ⎦ 3 ic− a2 a 1 ic
(14.33)
In αβ Frame
Regarding positive and negative sequence components and by using the nonnormalized Clarke transformation, the unbalanced currents in the αβ reference frame can be calculated as follows: ,T +
Iαβ = Iα Iβ = Tαβ Iabc
(14.34)
with
Tαβ
0 1 2 1− −√12 2 √ = 3 0 23 − 23
(14.35)
Hence, the instantaneous positive and negative sequence currents in the αβ reference frame can be obtained by: + iαβ
+
T 1 1 −q = Tαβ iabc = Tαβ [T+ ] iabc = Tαβ [T+ ] Tαβ iαβ = iαβ 2 q 1 (14.36)
− iαβ
−
T 1 = Tαβ iabc = Tαβ [T− ] iabc = Tαβ [T− ] Tαβ iαβ = 2
π
1 q iαβ −q 1 (14.37)
where q = e−j 2 is a phase-shift time-domain operator to obtain in the quadrature version (90◦ -lagging) of an original waveform. It is necessary to mention the
352
S. Karimi et al.
+
-
[
−
]
[
−
]
++ [
]
+
−
+
-
+
−
−
Fig. 14.7 Instantaneous positive and negative sequence calculation in αβ reference frame
homopolar component does not have effect on the instantaneous positive and negative sequence components in the stationary reference frame. Figure 14.7 shows the block diagram of the instantaneous positive and negative sequence calculation in the αβ reference frame.
14.4.1.3
In dq Frame
The positive and negative sequences of the unbalanced currents in the dq reference frame are obtained by using the Park’s transformation: ,T +
= Tdq iαβ idq = id iq
Tdq =
cos θ sin θ − sin θ cos θ
(14.38)
(14.39)
Thus, the instantaneous positive and negative sequence currents in the dq reference frame can be deduced by: + idq
+ + = Tdq+ iαβ = Tdq θ + iαβ =
− vdq
−
− = Tdq− iαβ = Tdq θ − iαβ =
cos θ + sin θ + + i − sin θ + cos θ + αβ
(14.40)
cos θ − sin θ − − i − sin θ − cos θ − αβ
(14.41)
14 Voltage Unbalance Compensation in AC Microgrids
+
353
-
[
]
++ [
]
+
−
−
+
-
+
−
[
]
−
Fig. 14.8 Instantaneous positive and negative sequence calculation in dq reference frame
Figure 14.8 shows the block diagram of the instantaneous positive and negative sequence calculation in dq reference frame.
14.4.2 Indirect Extraction Methods In direct extraction methods, it is supposed that the unbalanced currents are sinusoidal. Therefore, the presence of harmonic components causes inappropriate separation of positive and negative sequences. In the presence of harmonic components, the use of indirect extraction methods is more efficient. Some of these methods are briefly described in the following.
14.4.2.1
Double Synchronous Reference Frame (DSRF)
In the DSRF method, to separate the positive sequence component, the current signal in the αβ frame must be rotated with the system angular frequency ω in the counterclockwise direction by applying the Park’s transformation. Also, for extracting the negative sequence component, the current signal in the αβ frame must be rotated in the clockwise direction (see Fig. 14.9). After rotating, the positive and negative sequence components in the dq frame comprise two components, a DC component and an AC component as follows [26]: , + + + − + + − idq = e−j θ · iαβ = idq + e−j (θ −θ ) · idq 3456 3 45 6 dc term
AC term
(14.42)
354
S. Karimi et al.
−
abc
−
−
−
Fig. 14.9 DSRF structure
, + −j (θ − −θ + ) − − − + · idq idq = e−j θ · iαβ = idq + ee 3456 3 45 6 dc term
(14.43)
AC term
where id+ iq+ id− iq−
= I + cos δ + − ϕ + = I + sin δ + − ϕ + = I − cos δ − − ϕ − = I − sin δ − − ϕ −
(14.44)
θ + = ωt + ϕ + θ − = −ωt + ϕ −
(14.45)
which that
And the park’s transformation is given by + , ej θ =
cos θ sin θ − sin θ cos θ
(14.46)
In these equations, ϕ+ and ϕ− are the initial phases of the positive and negative sequence of the measured voltage, respectively. Also, δ + and δ − are the initial phases of the positive and negative sequence of the current. According to (14.42), (14.43), and (14.45), from the positive sequence reference frame point of view, the negative sequence is rotating at a double frequency and vice versa. Generally, to eliminate this oscillatory component a low pass filter is used. However, the low pass filter cannot completely eliminate the oscillatory component. Some methods, such as decoupled double synchronous reference frame (DDSRF), have been proposed for eliminating the oscillating term and decoupling the positive and negative sequences [26, 27].
14 Voltage Unbalance Compensation in AC Microgrids
+
-
+-
355
′
′
SOGI Fig. 14.10 SOGI scheme [29]
14.4.2.2
DSOGI-FLL
There are various ways for quadrature-signals generation. One of the easy and effective methods is using the second-order generalized integrator (SOGI). The SOGI scheme is shown in Fig. 14.10. The transfer functions of a SOGI are as follows [28]: D(s) =
i (s) ks ω0 s = 2 i(s) s + ks ω0 s + ω0 2
(14.47)
Δ(s) =
qi (s) ks ω0 2 = 2 i(s) s + ks ω0 s + ω0 2
(14.48)
where ω0 and ks set resonance frequency and damping factor, respectively. Transfer functions (14.47) and (14.48) reveal that if i is a sinusoidal signal, i and qi will be sinusoidal as well. Moreover, qi will be always 90-lagging i , independently of both the frequency of i and the values of ω0 and ks . These characteristics make the SOGI ideal for the quadrature signals generation. A dual SOGI (DSOGI) is necessary to perform the positive and negative sequences estimation. On the other hand, only when the ω0 corresponds to the grid frequency, the estimated components will be correct. Therefore, it is necessary for a frequency adaptation system to tune the SOGIs with the grid frequency. For this purpose, the frequency locked-loop (FLL) can be used [29]. DSOGI-FLL has good performance, even under imbalance and harmonic conditions. The basis of the DSOGI is the concept of symmetrical components that was discussed earlier. The DSOGI-FLL diagram is shown in Fig. 14.11. As can be seen in Fig. 14.11, DSOGI-FLL consists of two SOGI and one FLL.
356
S. Karimi et al.
′ ′ ′
+
FLL ′
-
′
++
′
′
′ ′ ′
SOGI-QSG (α) ′ ′ ′
+
′
−′
+
-
SOGI-QSG (β)
−′
+
DSOGI DSOGI-FLL
Fig. 14.11 DSOGI-FLL block diagram
′
+
-
′
[
PI
]
++
′
++
′
′
′
SRF-PLL ′ −
′
−′
+
−
−
+
DSOGI-QSG
-
+
−
[
]
−
Fig. 14.12 Block diagram of the DSOGI-PLL [29]
14.4.2.3
DSOGI-PLL
Figure 14.12 shows the block diagram of DSOGI-PLL. It is clear from Fig. 14.12 that the DSOGI-PLL is similar to its DSOGI-FLL counterpart, except that it is in the rotating reference frame. Hence PLL is used instead of FLL in the DSOGI-PLL structure.
14 Voltage Unbalance Compensation in AC Microgrids
357
14.4.3 Instantaneous Power under Unbalanced Conditions According to instantaneous power theory, complex power injected into the grid is defined and given by: ∗ s = p + j q = V I ∗ = Vsdq Idq
(14.49)
where V = [va . vb . vc ], I = [ia . ib . ic ]T , Vsdq = [vsd . vsq . vs0 ], Idq = [id . iq . i0 ]T . Considering the voltage and current sequence components, (14.49) can be represented as: ∗ + − + − ej ωt Idq s = ej ωt Vsdq + e−j ωt Vsdq + e−j ωt Idq
(14.50)
By separating the real and imaginary parts: ∼
p = real {s} = P + p; p(t) = P + Pc cos (2ωt) + Ps sin (2ωt) ∼
q = img {s} = Q + q; q(t) = Q + Qc cos (2ωt) + Qs sin (2ωt)
(14.51)
(14.52)
where P and Q are the average parts of the active and reactive powers, respectively, ∼ ∼ and p and q are the oscillatory parts of the active and reactive powers, respectively. The average active and reactive power parts are as follows: + + − − + + − − Id + Vsq Iq + Vsd Id + Vsq Iq P = Vsd
(14.53)
+ + − − + + − − Q = −Vsd Iq + Vsq Id − Vsd Iq + Vsq Id
(14.54)
And the oscillatory parts are deduced using the following equations: + − − + + − − + Id + Vsq Iq + Vsd Id + Vsq Iq Pc = Vsd
(14.55)
+ − − + + − − + Ps = Vsd Iq − Vsq Id − Vsd Iq + Vsq Id
(14.56)
+ − − + + − − + Qc = −Vsd Iq + Vsq Id − Vsd Iq + Vsq Id
(14.57)
+ − − + + − − + Qs = Vsd Id + Vsq Iq − Vsd Id − Vsq Iq
(14.58)
358
S. Karimi et al.
As one can see from (14.51) and (14.52), the frequency of the oscillatory terms is two times the grid frequency. In addition, as can be seen from (14.55) to (14.58), there are interactions between positive and negative sequence current and voltages. Since, for compensating the unbalanced load currents, the negative sequence current must be injected by the inverter-interfaced DG, the power oscillation is inevitable, even if the voltage becomes quite balanced.
14.5 Case Study In this section, simulation demonstrations are presented to analyze the characteristics of the approaches previously described. In Fig. 14.13, the structure of the studied MG is shown. The studied MG is operated in islanded mode and comprises two inverter-interfaced DGs. DG#1 acts as a grid-forming converter and DG#2 operates as a grid-feeding converter that provides the ancillary services for power quality improvement. DG#1 regulates the voltage magnitude and frequency, and DG#2 is responsible for compensating the harmonic and imbalanced currents. In addition, DG#2 delivers active power to the MG. The harmonic load LH considered in this simulation is a nonlinear load consists of a three-phase diode bridge rectifier with the constant DC current equal to 80 A. To simulate the three-phases unbalanced load (LUB ), two sinusoidal current sources are considered for phase a and phase b, and phase c is unconnected. The other parameters of the studied MG are listed in Table 14.2. The block diagram of the control system for the grid-forming and grid-feeding converters are illustrated in Figs. 14.14 and 14.15, respectively. In the grid-
PCC
Grid
DG #2 DG #1 Grid-feeding Converter
Fig. 14.13 Studied MG
Grid-forming Converter
14 Voltage Unbalance Compensation in AC Microgrids Table 14.2 System parameters
359
Parameters Rf Lf Cf RL LL f
Value 1.19 m 100 μH 2500 μF 1 m 4 mH 50 Hz
Parameters Vdc vdref vqref LB1 LB2 Pref
Value 1500 V 391 V 0 40 kW 200 kW 100 kW
abc dq
abc dq
Voltage Control Loop
Current Control Loop
Pulses
dq
abc
PWM
Fig. 14.14 The block diagram of the control system for DG#1
dq abc
-
abc
+
PI
⁄
-1
dq Negative Sequence Extraction
+ + +
Pulses Hysteresis
Fig. 3
Fig. 14.15 The block diagram of the control system for DG#2
forming converter control system (Fig. 14.14), the desirable voltage magnitude and frequency are sent to the voltage and current control loops, and then translated using dq/abc transformation to the abc reference frame and given to the PWM block to generate suitable switching signals. The grid-feeding converter control system consists of four main parts (Fig. 14.15). The first and the second parts are the conventional control blocks that are used in the grid-feeding converters, and the third and the fourth parts perform the desired ancillary controls. The first part is used to deliver the specified active power to MG. In this part, the reference power is converted to the reference current in d axis after division by 3/2vdref . The voltage magnitude control is also the task
360
S. Karimi et al.
Fig. 14.16 Unbalanced load current
of the second part. In this part, the reference current in q axis is generated by using the voltage error and a PI controller. The third part deals with the harmonic current compensation caused by the nonlinear load. This part generates the reference currents required to compensate for the harmonic currents, using the scheme shown in Fig. 14.3. The fourth part is responsible for compensating the unbalanced currents caused by an unbalanced load. In this part, using one of the indirect extraction methods, the negative sequence reference currents are generated. Finally, these reference currents are added together and given to the hysteresis controller to generate the suitable switching signals for the inverter-interfaced DG#2. The unbalanced load is connected to the MG at t = 0.25 Sec. The unbalanced load current is shown in Fig. 14.16. The peak of the unbalanced current is 80 A. The nonlinear load also is connected at t = 0.4 Sec. Figure 14.17 depicts the current of this load. Figure 14.18 to Fig. 14.20 illustrate the DG#2 output voltage when DG#2 operates with the conventional control system (without ancillary controls). Figure 14.18 shows the DG#2 output voltage before the unbalanced and nonlinear loads are connected. As one can see, the voltage is properly controlled and the voltage peak is about 391 V. Figure 14.19 and Fig. 14.20 present the DG#2 output voltage after connecting the unbalanced and nonlinear loads, respectively. As can be seen, in the absence of ancillary controls, the voltage becomes unbalanced and distorted. The VUF and THD of the DG#2 output voltage under these conditions are presented in Fig. 14.21. The THD and VUF are calculated using the following formulas: # M T H D% =
2 h=2 Vh
V1
where Vh is the RMS of h-order harmonic.
× 100%
(14.59)
14 Voltage Unbalance Compensation in AC Microgrids
361
Fig. 14.17 Nonlinear load current
Fig. 14.18 Voltage before unbalanced and nonlinear load connection
V UF% =
vd− + vq−
2
vd+ + vq+
2
2 2
× 100%
(14.60)
Figure 14.21 shows that both VUF and THD exceeded their permissible limit, and the VUF and THD reach about 8% in the steady state. It is important to remark that in the distribution networks, the THD and VUF must be limited under
362
S. Karimi et al.
Fig. 14.19 Voltage after unbalanced load connection
Fig. 14.20 Voltage after nonlinear load connection
5% and 2%, respectively. Figure 14.22 presents the results of the FFT analysis of the DG#2 output voltage in the steady state. Active and reactive output power of the DG#1 and DG#2 are depicted in Fig. 14.23 and Fig. 14.24, respectively. As seen, the unbalanced and nonlinear load affects the active and reactive powers and causes power oscillations. In addition, under unbalanced and distorted conditions, the DG#2 does not properly follow its reference to active power. Figure 14.25 to Fig. 14.29 illustrate the results after adding ancillary control. Figure 14.25 shows the DG#2 output voltage after compensation. As seen, the
14 Voltage Unbalance Compensation in AC Microgrids
363
(a)
(b) Fig. 14.21 VUF and THD of the DG#2 output voltage
voltage is appropriately compensated and controlled. The VUF and THD of the DG#2 output voltage after compensation are depicted in Fig. 14.26. As one can see, the mentioned ancillary control can effectively reduce the VUF and THD, and they are less than the permitted amount. Figure 14.27 shows the results of FFT analysis in the steady state under these conditions. This figure illustrates that using the ancillary control the THD in the steady state is 1.45%. Active and reactive output power of the DG#1 and DG#2 after compensation are presented in Fig. 14.28 and Fig. 14.29. As indicated in Fig. 14.28, since the unbalanced voltage is compensated, the power
364
S. Karimi et al.
Fig. 14.22 Voltage harmonic components
oscillations no longer exist in the active and reactive output power of the DG#1. It is important to note that the power oscillations in the active and reactive output power of DG#2 are generated due to the injection of the negative sequence current by DG#2.
14 Voltage Unbalance Compensation in AC Microgrids
(a)
(b) Fig. 14.23 Active and reactive output power of the DG#1
365
366
S. Karimi et al.
(a)
(b) Fig. 14.24 Active and reactive output power of the DG#2
14 Voltage Unbalance Compensation in AC Microgrids
Fig. 14.25 Voltage after compensation
367
368
S. Karimi et al.
(a)
(b) Fig. 14.26 VUF and THD after compensation
14 Voltage Unbalance Compensation in AC Microgrids
Fig. 14.27 Voltage harmonic components after compensation
369
370
S. Karimi et al.
(a)
(b) Fig. 14.28 Active and reactive output power of the DG#1 after compensation
14 Voltage Unbalance Compensation in AC Microgrids
(a)
(b) Fig. 14.29 Active and reactive output power of the DG#2 after compensation
371
372
S. Karimi et al.
References 1. Sabzevari, K., Karimi, S., Khosravi, F., & Abdi, H. (2019). Modified droop control for improving adaptive virtual impedance strategy for parallel distributed generation units in islanded microgrids. International Transactions on Electrical Energy Systems, 113, 758–771. 2. Pal, R., & Gupta, S. (2019). Topologies and control strategies implicated in dynamic voltage restorer (DVR) for power quality improvement. Iranian Journal of Science and Technology, Transactions of Electrical Engineering. https://doi.org/10.1007/s40998-019-00287-3. 3. Guerrero, J. M., Chandorkar, M., Lee, T. L., & Loh, P. C. (2012). Advanced control architectures for intelligent microgrids—Part I: Decentralized and hierarchical control. IEEE Transactions on Industrial Electronics, 60(4), 1254–1262. 4. Naderi, Y., Hosseini, S. H., Ghassem Zadeh, S., Mohammadi-Ivatloo, B., Vasquez, J. C., & Guerrero, J. M. (2018). An overview of power quality enhancement techniques applied to distributed generation in electrical distribution networks. Renewable and Sustainable Energy Reviews, 93, 201–214. 5. Y. Naderi, S.H. Hosseini, S.G. Zadeh, B. Mohammadi-Ivatlo, J.C. Vasquez, J.M. Guerrero, “Distributed power quality improvement in residential microgrids”, 10th International Conference on Electrical and Electronics Engineering (ELECO), 2017, pp. 90–94. 6. IEEE recommended practice and requirements for harmonic control in electric power systems, IEEE Std. 519-2014 (Revision of IEEE Std 519-1992), 2014, pp. 1-213. 7. Singh, B., Al-Haddad, K., & Chandra, A. (1999). A review of active filters for power quality improvement. IEEE Transactions on Industrial Electronics, 46(5), 960–971. 8. Asiminoaei L, Blaabjerg F, Hansen S. “Evaluation of harmonic detection methods for active power filter applications”, Twentieth Annual IEEE Applied Power Electronics Conference and Exposition, (APEC 2005), 2005 (Vol. 1, pp. 635-641). 9. Akagi H. “Generalized theory of the instantaneous reactive power in three-phase circuits”, IEEJ IPEC-Tokyo’83. 1983. 10. Chang, G. W., & Shee, T. C. (2002). A comparative study of active power filter reference compensation approaches. IEEE Power Engineering Society Summer Meeting, 2, 1017–1021. 11. Ordonez R, Karimi C, Oueidat M, Sadarnac D. “Comparison of current reference generation techniques for non-active power compensation under distorted and unbalanced conditions”, 32nd Annual Conference on IEEE Industrial Electronics, (IECON 2006), 2006. 12. Bhattacharya S, Divan DM, Banerjee B. “Synchronous frame harmonic isolator using active series filter”, European conference on power electronics and applications 1992. 13. Freijedo, F. D., Yepes, A. G., Lopez, O., Vidal, A., & Doval-Gandoy, J. (2010). Threephase PLLs with fast post fault retracking and steady-state rejection of voltage unbalance and harmonics by means of lead compensation. IEEE Transactions on Power Electronics, 26(1), 85–97. 14. Campanhol, L. B., da Silva, S. A., & Goedtel, A. (2014). Application of shunt active power filter for harmonic reduction and reactive power compensation in three-phase four-wire systems. IET Power Electronics., 7(11), 2825–2836. 15. Han, Y., Luo, M., Zhao, X., Guerrero, J. M., & Xu, L. (2015). Comparative performance evaluation of orthogonal-signal-generators-based single-phase PLL algorithms—A survey. IEEE Transactions on Power Electronics., 31(5), 3932–3944. 16. Golestan, S., Guerrero, J. M., & Vasquez, J. C. (2016). Three-phase PLLs: A review of recent advances. IEEE Transactions on Power Electronics., 32(3), 1894–1907. 17. Teodorescu, R., Blaabjerg, F., Liserre, M., & Loh, P. C. (2006). Proportional resonant controllers and filters for grid-connected voltage-source converters. Electric Power Applications, IEE Proceedings, 153(5), 750–762. 18. Lascu, C., Asiminoaei, L., Boldea, I., & Blaabjerg, F. (2007). High performance current controller for selective harmonic compensation in active power filters. IEEE Transactions on Power Electronics, 22(5), 1826–1835.
14 Voltage Unbalance Compensation in AC Microgrids
373
19. Stankovic, A. V., & Lipo, T. A. (2001). A novel control method for input output harmonic elimination of the PWM boost type rectifier under unbalanced operating conditions. IEEE Transactions on Power Electronics, 16(5), 603–611. 20. Serpa, L. A., Round, S. D., & Kolar, J. W. (2007). A virtual-flux decoupling hysteresis current controller for mains connected inverter systems. IEEE Transactions on Power Electronics, 22(5), 1766–1777. 21. Serpa, L. A., Ponnaluri, S., Barbosa, P. M., & Kolar, J. W. (2007). A modified direct power control strategy allowing the connection of three- phase inverters to the grid through LCL filters. IEEE Transactions on Industry Applications, 43(5), 1388–1400. 22. Vazquez, S., Sanchez, J. A., Carrasco, J. M., Leon, J. I., & Galvan, E. (2008). A modelbased direct power control for three-phase power converters. IEEE Transactions on Industrial Electronics, 55(4), 1647–1657. 23. Cortés, P., Rodríguez, J., Antoniewicz, P., & Kazmierkowski, M. (2008). Direct power control of an AFE using predictive control. IEEE Transactions on Power Electronics, 23(5), 2516– 2553. 24. Lyon, W. V. (1937). Applications of the method of symmetrical components. McGraw-Hill book Company, Incorporated. 25. Fortescue, C. L. (1918). Method of symmetrical co-ordinates applied to the solution of poly phase networks. Transactions of the American Institute of Electrical Engineers, 37(2), 1027– 1140. 26. Rodríguez, P., Pou, J., Bergas, J., Candela, J. I., Burgos, R. P., & Boroyevich, D. (2007). Decoupled double synchronous reference frame PLL for power converters control. IEEE Transactions on Power Electronics., 22(2), 584–592. 27. Reyes, M., Rodriguez, P., Vazquez, S., Luna, A., Teodorescu, R., & Carrasco, J. M. (2012). Enhanced decoupled double synchronous reference frame current controller for unbalanced grid-voltage conditions. IEEE Transactions on power electronics, 27(9), 3934–3943. 28. Rodriguez P, Teodorescu R, Candela I, Timbus AV, Liserre M, Blaabjerg F. “New positivesequence voltage detector for grid synchronization of power converters under faulty grid conditions”, 37th IEEE Power Electronics Specialists Conference 2006. 29. Rodriguez, A. L., Candela, I., Mujal, R., Teodorescu, R., & Blaabjerg, F. (Jan). Multi-resonant frequency-locked loop for grid synchronization of power converters under distorted grid conditions. IEEE Transactions on Industrial Electronics, 58(1), 127–138.
Chapter 15
WAM-Based Hierarchical Control of Islanded AC Microgrids E. S. N. Raju P and Trapti Jain
15.1 Introduction A microgrid comprises of low voltage distributed systems with DG units, storage devices, loads, and interconnecting switches [1, 2]. Microgrids (MGs) can be operated either in an island mode or in a grid connected mode of operation [1, 2]. Based on the type of grid voltage, MGs can be classified into AC MicroGrids (ACMGs), DC MicroGrids (DCMGs), and Hybrid AC/DC MicroGrids (HADMGs). ACMGs have been proposed to utilize the existing AC grid technologies, protection, and standards. However, power generation from various DGs such as photo-voltaic arrays and fuel cells is DC, which needs to be converted into AC power through power electronic interface for connecting them with the AC utility grid. This AC power is again converted back into DC power required by today’s electrical loads such as Uninterrupted Power Supply (UPS), fluorescent lights, variable motor drives, and hybrid electric vehicles. Thus, an individual ACMG may be less efficient due to more power losses occurring in multiple conversions. Besides this, synchronization, stability, and reactive power requirement are its inherent demerits. DCMGs are emerging as a better alternative due to the abovementioned reasons for renewable energy based DGs. However, power generation from the sources such as diesel generator, small hydro turbine with synchronous generator and photovoltaic panels, etc., as well as the electrical loads are a mix of AC and DC power. Thus, an individual DCMG may not completely eliminate the losses occurring in multiple stage conversions, though the losses occurring in DC/DC conversions are lesser than those occurring in AC/DC or DC/AC conversions.
E. S. N. Raju P () · T. Jain Discipline of Electrical Engineering, Indian Institute of Technology Indore, Indore, Madhya Pradesh, India e-mail: [email protected]; [email protected] © Springer Nature Switzerland AG 2021 A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems, https://doi.org/10.1007/978-3-030-59750-4_15
375
376
E. S. N. Raju P and T. Jain
Fig. 15.1 Typical architecture of hybrid AC/DC microgrid
In an individual ACMG, AC loads require single stage conversion and DC loads require multiple conversions. Similarly in an individual DCMG, DC loads require single stage conversions and AC loads require multiple stage conversions. Therefore, a HADMG, which is a cluster of ACMG, DCMG, bidirectional converters, control equipment, and energy management system, as shown in Fig. 15.1, has been proposed. ACMG consists of DGs such as diesel generators, small hydro turbine with synchronous generator, biomass based power generation, etc., which generate AC power output, AC loads, fly wheel energy storage system with AC/AC interface, and utility grid connection through bidirectional power electronic based switch at the PCC. DCMG includes DGs such as photo-voltaic panels, fuel cell tracks, etc., which produce DC power output, loads requiring DC power input such as UPS, fluorescents lighting, etc., DC energy storage system such as battery, super-capacitor, etc., and hybrid electric vehicles. These components are connected to DC bus through DC/DC buck or boost converter. A bidirectional AC-DC/DCAC converter is required to interface ACMG and DCMG. A back-up converter is also included to avoid any islanding of ACMG and DCMG. The main objective of the bidirectional AC/DC converter is to maintain the smooth power transfer between ACMG and DCMG, to maintain stable voltages of AC bus and DC bus under varying generation and load conditions. Power transfer would take place from ACMG to DCMG if the power generation in ACMG is more than in DCMG and vice versa in grid connected as well as islanding modes of operation. The various microsources in the HADMG are interfaced through five different types of power electronic converters viz. AC/AC, AC/DC/DC, DC/DC, DC/AC, and AC-DC/DC-AC bidirectional converter. These power electronic converters are used to control the microsources in the grid connected as well as in the island mode of operation. Different types of control methods are used, as shown in Fig. 15.2 to
15 WAM-Based Hierarchical Control of Islanded AC Microgrids
377
HADMG Micro Sources Control Methods
ACMG Micro Sources Control Methods Island Mode
PQ control Method
DCMG Micro Sources Control Methods
Grid Mode -Droop Control and -Voltage/frequency control Method
Analog Control Techniques
-Voltage Mode Control and -Current Mode Control
Digital Control Techniques -Current control -Predictive digital current programmed control -Variable frequency predictive control -Sensor less current mode control and -Predictive digital dead-beat control
Fig. 15.2 Microsources control methods in hybrid AC/DC microgrid
control the microsources in the ACMG and those in the DCMG. The microsources in the ACMG are controlled by active and reactive power (PQ) control, droop control, and voltage/frequency control, while the microsources in the DCMG are controlled using analog and digital control techniques. In PQ control method, which is applied in grid connected mode of operation, the reference values for the active and reactive power are given by the utility grid controller to the controllers of microsources in the ACMG. Droop control method and v/f control method, which are employed in island mode of operation, are like primary control and secondary control, respectively, in the transmission network. There are two types of droop control method viz. P-f droop control method and Q-V droop control method. P-f control method controls the frequency by controlling the active power supplied by DGs whereas the QV droop control method controls the voltage magnitude by controlling the reactive power supplied by the DGs. These methods can minimize the fluctuations in the voltage and the frequency for small disturbances only. In order to minimize large fluctuations in the voltage and frequency, v/f control method is used. V/f control method uses PI controllers and PID controllers to damp out the oscillations in the voltage and the frequency. The controllers shall be designed such that in island mode of operation, it turns from PQ control to droop control for small disturbances and from droop control to v/f control for large disturbances. Analog control techniques utilize the comparison of output voltage and output current with the reference value and known as voltage mode control and current mode control, respectively. Analog control techniques are simple and cheap but
378
E. S. N. Raju P and T. Jain
suffer from low flexibility, low reliability, high complexity, and are more sensitive. Digital control techniques are current control method, predictive digital current programmed control method, variable frequency predictive control method, sensorless current mode control method, and predictive digital dead-beat controller for dc-dc converters. Thus, the HADMG seems more beneficial to facilitate the connection of various renewable AC and DC power sources and loads with the power system in order to minimize the conversion losses. However, the practical implementation of the HADMG needs to overcome several technical and economical challenges as listed below. • Building a new DC grid and upgrading the existing AC grid is a long-term process. • Development of appropriate control strategies and protection schemes for safe and reliable operation of the HADMG in grid connected as well as islanded mode of operation. • Designing the control coordination between the various types of power electronic converters for power sharing among various types of DGs under varying operating conditions. • Redesigning of home and office products to remove the embedded AC/DC rectifiers. • Optimal voltage levels need to be determined for easy connection of various types of DC loads. Due to the above technical and economical challenges, ACMGs are still dominant because of their similar intrinsic characteristics to the traditional Low Voltage (LV) as well as Medium Voltage (MV) distribution networks. The main challenges in ACMGs include stability issues and development of control schemes to enhance their stability and dynamic performance. Power system stability is defined as “the ability of an electric power system, for a given initial operating condition, to regain a state of operating equilibrium after being subjected to a physical disturbance, with most system variables bounded so that practically the entire system remains intact.” [3, 4]. It has been considered as an important issue since 1920. Stability of a power system mainly depends on the amount of physical disturbance and operating condition prior to the disturbance. Depending upon the amount of disturbance, viz. small-signal and large-signal, stability can be classified as small-signal stability and transient stability, respectively [3, 4]. The small-signal disturbances are in the form of continuous load variations, whereas, large-signal disturbances are severe disturbances, line outages, and faults. Power system stability is mainly classified into three categories: rotor angle stability, frequency stability, and voltage stability, as shown in Fig. 15.3 [3, 4]. The rotor angle stability and voltage stability are further categorized into smallsignal/disturbance stability and large-signal/disturbance stability, while frequency stability is divided into the short-term and long-term stability. This classification has been done based on the following considerations [3, 4].
15 WAM-Based Hierarchical Control of Islanded AC Microgrids
379
Fig. 15.3 Classification of power system stability [3, 4]
• The physical nature of the resulting mode of instability as indicated by the main system variable in which instability can be observed. • The size of the disturbance considered, which influences the method of calculation and prediction of stability. • The devices, processes, and the time span that must be taken into consideration in order to assess stability. Stability issues in ACMGs become more vital concern due to the presence of multi-energy source based IIDG units, interaction between different types of power electronic converters and energy storage system. Stability analysis in ACMGs follows the similar concepts as in the existing AC grid. The stability issues in ACMGs can be divided into small-signal, transient, and voltage stability. The reasons behind their occurrence, shown in Fig. 15.4, and improvement methods, shown in Fig. 15.5, are briefly given below [5]. • Small-signal stability: The small-signal stability in ACMGs can be analyzed with a linearized model of microsources, network lines, and loads. Small-signal instability in ACMGs may occur due to the feedback controller, continuous load switching, system damping, and power limit of the IIDG units. Smallsignal stability can be improved by adding various supplementary control loops around the existing control loops of IIDG units, using stabilizers with IIDG units, coordinated control of the microsources, and energy management system. • Transient stability: The transient stability analysis of ACMGs can be performed with their nonlinear models [6]. The construction of the Lyapunov function [7] is one of the methods to analyze the transient stability. A fault with subsequent island, loss of IIDG unit, faults in the main grid or in the ACMG, large changes in the load posses most of the transient instability problems. The improvement in the transient stability can be achieved by controlling of Energy Storage System (ESS) to inject active/reactive power during shortage of generation, tripping of IIDG units, load dynamics, and islanding. Load shedding, control of power electronic converters, and adaptive protection devices also help in improving the stability.
380
E. S. N. Raju P and T. Jain
Fig. 15.4 Reasons behind the stability problems of ACMGs
Fig. 15.5 Stability improvement methods of ACMGs
• Voltage stability: The voltage stability problem in ACMGs can be revealed from the P-V and Q-V curves. The P-V curve shows the maximum loadability, while the Q-V curve indicates the necessary amount of reactive power required at the load end for the desired voltage. The main issues in the voltage stability analysis include characteristics of the load, control strategy of reactive power, slow increase in the power demand, and outage of one of the parts of the ACMG network [5]. Reactive power limits/current limiters, load dynamics associated
15 WAM-Based Hierarchical Control of Islanded AC Microgrids
381
with the induction motor loads, tap changers, and voltage regulators create most of the voltage instability problems in ACMGs. Voltage regulation with IIDG units, reactive compensation with distributed FACT devices like DSTATCOM [8], advanced load controller, load shedding, and modified current limiters of the microsources can improve the voltage stability in ACMGs. Stability of ACMGs is not a critical issue in grid connected mode of operation as the stiff grid would be responsible for its stable operation. However, in an island mode of operation, it is an important concern due to the low-inertial nature of power electronics interfaced DER units. At a given steady-state operating condition, the islanded AC microgrids (IACMGs) may be unstable when it is subjected to the large disturbance. However, the IACMG should operate satisfactorily if it is subjected to small-signal disturbance at a given steady-state operating condition. Therefore, small-signal stability enhancement is a fundamental requirement for the satisfactory and reliable operation of IACMGs. Stability of IACMGs feeding passive, active, and dynamic loads has been enhanced by developing secondary controllers [9] such as decentralized [10–13], centralized [14], distributed [15, 16], and hierarchical controllers [17–19]. However, these controllers have been validated only on a particular type of static or dynamic load and provide relatively good performance for those load dynamics. Future microgrids (MGs) are expected to have multiple types of static and dynamic loads such as constant power load (CPL), rectifier interfaced active load (RIAL), and dynamic induction motor (IM) load. Further, the stability analysis of IACMGs feeding these multiple types of static and dynamic loads, simultaneously, addressed in [20], revealed that the presence of these multiple types of loads introduces impedance unbalance caused by the negative incremental impedance characteristics offered by both the CPL as well as RIAL. In addition, the dynamic IM load introduces new low-damping high frequency, interarea, and unstable modes in IACMGs. Moreover, the aforementioned controllers are not found suitable for the stability enhancement of IACMGs when these multiple types of static and dynamic loads were present. Therefore, it is essential to design the controller considering the dynamics of these multiple types of static and dynamic loads in IACMGs. Recent advancements in synchronized phasor measurement technology (SMT) led to developing a wide area measurement system (WAMS) based hierarchical controllers, shown in Fig. 15.6, for the stability enhancement of conventional power system [21–31]. These hierarchical controllers have been proved to be one of the most potentially effective solutions to enhance the stability of a conventional power system. Traditional supervisory control and data acquisition (SCADA)/energy management system (EMS), based on voltage and current measurements from conventional remote terminal units (RTUs), are becoming increasingly unreliable for real-time operations of microgrids because they cannot fully anticipate all the conditions faced by operators. New technologies, which rely on accurate, high resolution, real-time monitoring of actual system conditions using SMT, are needed to support the real-time operations. For these reasons, nowadays, industries are beginning to explore the use of SMT at the distribution level for distribution system
382
E. S. N. Raju P and T. Jain
Fig. 15.6 General structure of wide area measurement system (WAMS) based control system
visibility enhancement, integration of distributed generation, microgrid operation and control, demand response, and power quality applications [32]. Motivated by these factors, PMUs supported two-level hierarchical controller has been explored for the stability enhancement of IACMGs as well [33]. However, delay-free communication channels have been considered during the design procedure, which may not be pragmatic in real WAMS. In WAMS-based stabilization application, transmitting signals from PMUs to the upper-level centralized controller, through the phasor data concentrator (PDC), and then back to generator’s lower-level decentralized controllers involve some time delays [23–31]. These signal transmission time delays mainly depend on the type of communication link used. In the Western Electricity Coordinating Council (WECC) system, these time delays vary from 25 ms (one-way from PMUs to the upper-level centralized controller) for fiber optics cables to 250 ms for satellitebased communication links [23]. Moreover, these time delays can deteriorate the effectiveness of the WAMS-based two-level hierarchical controllers [23–31]. Thus, it is essential to verify the effectiveness of PMUs supported two-level hierarchical controller for a wide range of signal transmission time delays encountered in real WAMS. Therefore, in this chapter, the impact of signal transmission time delays on the performance of PMUs supported two-level hierarchical controller, designed for the stability enhancement of IACMGs with static and dynamic loads, has been investigated.
15 WAM-Based Hierarchical Control of Islanded AC Microgrids
383
Fig. 15.7 Schematic diagram of a studied IACMG system with the proposed hierarchical controller considering signal transmission time delays
The proposed hierarchical controller is composed of a lower-level decentralized controller for each IIDG unit helped by a multi-input-multi-output (MIMO) centralized controller at the upper level, as shown in Fig. 15.7. Both lower-level and upper-level controllers at the two levels work together for the stability enhancement of IACMGs. Further, the proposed hierarchical controller relies on synchronized measurements supplied by PMUs through the WAMS. The modal-based extended linear-quadratic-Gaussian (LQG) approach has been used to design the proposed hierarchical controller and its performance has been compared with the hierarchical controller based on state-based extended LQG approach. The proposed design approach also includes time delays experienced in the transmission of signals from PMUs to the upper-level centralized controller and from the upper-level centralized controller to both the lower-level decentralized controller of each inverter-interfaced
384
E. S. N. Raju P and T. Jain
distributed generation (IIDG) unit as well as local controller of RIAL. Finally, the performance of the proposed hierarchical controller has been assessed for different time delays using both eigenvalue analysis and time-domain simulations performed on the studied IACMG system, as shown in Fig. 15.7.
15.2 Design of WAMS-Based Hierarchical Controller Considering Signal Transmission Time Delays Figure 15.8 shows the proposed hierarchical controller’s design procedure considering signal transmission time delays. It is worth mentioning that the detailed design procedure of the proposed hierarchical controller considering the delay-free communication channels has been produced in [33]. However, the same design procedure of the proposed hierarchical controller considering signal transmission time delays has been discussed in this paper for enhanced readability. In the design procedure, first, a lower-level decentralized controller is incorporated in the IACMG system model. The lower-level decentralized controller adds additional auxiliary
Fig. 15.8 Design procedure of the proposed hierarchical controller considering signal transmission time delays
15 WAM-Based Hierarchical Control of Islanded AC Microgrids
385
control terms to the conventional power sharing/droop controller. The auxiliary control terms are based on the desired power sharing of each IIDG unit and the total active and reactive power generations information acquired from PMUs through ∗ ∗ the WAMS [33]. The control commands, ωi−dec and v0di−dec , from the lower-level decentralized controller modulates the output of conventional droop controller to ∗ generate the modified reference frequency (ωi∗ ) and d-axis voltage (v0dqi ) of i th IIDG unit, as given in (15.1)–(15.3). 5
∗ ωi−dec
63 nG
∗ + KP deci αP i ωi∗ = ωi−dr
4 Pi − Pi
∀i
(15.1)
i=1 ∗ v0di
=
∗ v0di−dr
. + KQdeci 3
αQi
nG
Qi − Qi dt ∀i
i=1
45
∗ v0di−dec
∗ ∗ v0qi = v0qi−dr = 0 ∀i
(15.2)
6 (15.3)
where nG represents total number of IIDG units, αP i and αQi , are the desired real and reactive power sharing by the i th IIDG unit, respectively, KP deci and KQdeci , are the real and reactive power control gains of lower-level decentralized controller, respectively. Therefore, these are the tuning parameters of the lowerlevel decentralized controller of the i th IIDG unit. Furthermore, it is evident that nG nG # # ∗ αP i = 1 and αQi = 1. In (15.1) and (15.2), reference frequency, ωi−dr , and
i=1
i=1
∗ reference voltage, v0di−dr , are set by the conventional droop controller according to the active and reactive power droop characteristics [20, 33, 34]. The main advantage of the lower-level decentralized controller is that the stability and basic operation of the IIMG system are maintained when the centralized controller fails to communicate with the local controllers. In the next step, a small-signal linearized model has been developed to produce frequency and voltage responses in the presence of load disturbances [33, 34]. Then, an optimal set of input/output signals have been obtained using geometric measures approach to reduce the costs associated with the installation of PMUs [21, 22]. It is worth mentioning that the I/O signals are selected from and to all IIDG units in the studied IACMG system due to its small size and for the purpose of demonstrating the design procedure of the proposed hierarchical controller, thoroughly. Thereafter, model-order reduction technique based on the Schur balanced model reduction technique [14, 21, 35] has been used to obtain a reduced-order model and selected input/output signals have been incorporated in the reduced-order model. The most important communication issue coupling with PMUs is a signal transmission time delay. Transmitting signals from PMUs to the upper-level centralized
386
E. S. N. Raju P and T. Jain
controller, through the PDC, and then back to both the lower-level decentralized controller of each IIDG unit as well as local controller of RIAL involves some time delays, as shown in Fig. 15.7. These signal transmission time delays mainly depend on the type of communication link used and could vary from 25 ms (one-way) for fiber optic based links to 250 ms (one-way) for satellite-based links [23]. Basically, time delays involved in the processing and routing of signals at both PMUs and the controller are small (especially in slow communication links) compared to communication link delays and can be neglected without loss of generality [24]. In general, time delays may reduce the effectiveness of the controller, especially, when relatively slow communication links such as satellite-based links are used. Therefore, signal transmission time delays, experienced in practical WAMs, have been incorporated. Padé approximation method is widely used in WAMS-based stabilization application for the modeling of signal transmission time delays [23–28]. Figure 15.9 shows the phase response of 1st and 2nd order Padé approximations, given in (15.4) and (15.5), respectively, and compares them with the exact response of the time delay (e−τ s ) [24]. It is evident that the 1st order Padé approximation provides good accuracy in modeling short time delays (25–100 ms). However, its accuracy is reduced for longer time delays greater than 100 ms. Whereas, 2nd order Padé approximation is more accurate than the 1st order Padé approximation for modeling longer time delays (250 ms) encountered in slow satellite-based communication links. Therefore, in this paper, the 2nd order Padé approximation has been used for the modeling of longer time delays.
Phase response
0
=25 ms =100 ms
-50
=250 ms
-100
e-
s
P2(s) -150 0 10
P1(s) 1
10 Frequency, rad/s
Fig. 15.9 Phase response of 1st and 2nd order Padé approximations
15 WAM-Based Hierarchical Control of Islanded AC Microgrids
−τ s + 2 τs + 2
(15.4)
τ 2 s 2 − 6τ s + 12 τ 2 s 2 + 6τ s + 12
(15.5)
P1 (s) = P2 (s) =
387
where P1 (s) and P2 (s) are 1st and 2nd order Padé approximation functions in Laplace domain, respectively, and τ is signal transmission time delay for one-way communication. It is worth mentioning that time delays are assumed to be identical and fixed for all channels i.e., τ = τin = τout . Next, noise and disturbance models are augmented to the reduced-order model with selected input/output signals to consider the impact of measurement noises and model mismatch, respectively. Then, the augmented model has been used to design the upper-level centralized controller, which generates control commands to the lower-level controllers of IIDG units and RIAL, as shown in Fig. 15.7. It was observed that the robust control approaches, available in the existing literature [21– 31], such as conventional LQG, H2 and H∞ syntheses, and mixed H2 /H∞ synthesis with pole placement constraints [35] are not suitable to stabilize IACMGs when multiple types of static and dynamic loads are present. The reason behind this is that the low-inertial nature of IIDG units and the presence of multiple types of static and dynamic loads make IACMGs more vulnerable to instability [20]. Therefore, the upper-level centralized controller has been designed based on a modal-based extended LQG approach, shown in Fig. 15.10, to enable direct damping of the targeted modes while keeping other modes unaffected. This feature makes it highly suitable to PMUs supported two-level hierarchical controller. In the modal-based extended LQG approach, first, the augmented model is transformed into the modal canonical form using the real Schur decomposition [24]. The modal-augmented model has been used to design the upper-level centralized controller. Next, the separation principle [21] has been used to solve the modalbased extended LQG approach. This principle divides the central control problem into two subproblems: an optimal estimate of the state and optimal output-feedback control problems. The former one is based on the Kalman estimator theory [14, 21, 35] and the latter one is based on the linear quadratic regulator with prescribed
Fig. 15.10 Upper-level centralized controller based on modal-based extended LQG control approach
388
E. S. N. Raju P and T. Jain
degree of stability (LQRPDS) with output weighting [13, 14, 35]. The gain matrix of both the optimal output-feedback control and the optimal state estimator is dependent on their diagonal weighting matrices, represented by QfMG & RfMG and QcMG & RcMG , respectively. Therefore, these are the tuning parameters of the upper-level centralized controller. Further, the prescribed degree-of-stability value, α, can be used to force the proposed hierarchical controller to act faster and then prevent any slowing down of the response caused by the additional supplementary controller. The optimal parameters of both lower-level decentralized as well as upperlevel centralized controllers have been obtained by formulating a bi-objective optimization problem, as given in (15.6). It is desirable that the controller should be able to provide a good transient response while sustaining small parameter perturbations so that the stability of IACMGs is maintained under small-signal disturbances. This can be achieved by minimizing the condition number of the eigenvector matrix of the closed-loop IACMG as well as the second norm of the optimal output-feedback control gain matrix, as given in (15.6). minimize x
f1 (x) = VMG (x)2 VMG −1 (x)2 f2 (x) = KcMG (x)2
(15.6)
subject to xmin ≤ x ≤ xmax where KcMG is the optimal output-feedback control gain matrix, VMG is the eigenvector of the closed-loop IACMG, and x is a vector of the real and reactive power control gains of decentralized controllers of nG number of IIDG units and diagonal elements of QfMG & RfMG and QcMG & RcMG , as given in (15.7).
x = x1 x2 x3
(15.7)
x1 = KPDEC
= KP dec1 KP dec2 · · · KP decnG
(15.8)
x2 = KQDEC
= KQdec1 KQdec2 · · · KQdecnG
(15.9)
x3 = diag(QfMG )diag(RfMG ) diag(QcMG )diag(RcMG )
(15.10)
The two objective functions, given in (15.6), are conflicting in nature [14]. Therefore, a fast and elitist multi-objective non-dominated sorting genetic algorithm (NSGA-II) [36], which produce Pareto-optimal solutions, has been used to solve the above optimization problem. Post-Pareto analysis [14] has been performed to identify the most effective Pareto-optimal solution.
15 WAM-Based Hierarchical Control of Islanded AC Microgrids
389
Fig. 15.11 Block diagram of an IIDG unit with control signals from the centralized controller considering time delay
Fig. 15.12 Block diagram of a RIAL with the control signals from the centralized controller considering time delay
Finally, the control commands generated by the upper-level centralized controller are given to the lower-level modified power-sharing controller of each IIDG unit and AC current as well as DC voltage controllers of the RIAL, as shown in Figs. 15.11 and 15.12. The i th IIDG unit with consideration of the generated control command including signal transmission delays for round-trip communication, e−2τ s ucci , is shown in Fig. 15.11. The control command, ucci , modulates the dq-axis voltage reference output of the modified power-sharing controller, v∗0dqi , to generate a modified dq-axis voltage reference, v∗0dqcci , as given in (15.11). v∗0dqcci = v∗0dqi + e−2τ s ucci
∀i
(15.11)
where ucci = v dqi is a voltage correction (angle and magnitude of voltage) signal, which is responsible to compensate voltage deviations caused by the small-signal disturbances. Further, the modulated voltage signal, v∗0dqcci , is used as setpoint values for the inner voltage controller. This modulates the dq-axis reference current signal, i∗ldqi , of the current controller to generate a modified dq-axis reference
390
E. S. N. Raju P and T. Jain
input voltage signal, v∗idqi , to pulse width modulation (PWM) based voltage source inverter (VSI), as shown in Fig. 15.11. Figure 15.12 shows the AC current and DC voltage controllers of the i th RIAL with consideration of the generated control commands including signal transmission ∗ ∗ delays for round-trip communication, e−2τ s ilqALi and e−2τ s vdcALi , as given in (15.12) and (15.13), respectively. ∗ ∗ ∗ = ilqALi + e−2τ s ilqALi ilqccALi
(15.12)
∗ ∗ ∗ vdcccALi = vdcALi + e−2τ s vdcALi
(15.13)
∗ ∗ and vdcALi are reference current and voltage correction signals where ilqALi generated by the upper-level centralized controller, respectively. Further, these ∗ correction signals modulate the q-axis reference current, ilqALi , of the AC current ∗ controller and DC voltage reference, vdcALi , of DC voltage controller, respectively, to generate a modified dq-axis reference voltage input signal, v∗idqALi , as shown in Fig. 15.12.
15.3 Results and Discussions The robustness of the proposed hierarchical controller has been assessed for different signal transmission time delays (τ ) such as 25 ms (for fiber optics cables) and 250 ms (for satellite-based communication links) for one-way communication, i.e., 50 ms and 500 ms for round-trip communication [23]. Furthermore, due to the wide range of time delays encountered in real WAMS projects [23–31], the effectiveness of the proposed hierarchical controller has also been verified for a/an longer/increased signal transmission time delays such as 500 ms, 525 ms and 530 ms (maximum admissible value) for one-way communication (i.e., 1000 ms, 1050 ms and 1060 ms for round-trip communication). The maximum admissible value (530 ms) was chosen as the time delay for which the stability of the closed-loop IACMG (with the proposed hierarchical controller) system has been deteriorated. The assessment has been carried out on the studied IACMG system, shown in Fig. 15.7, using both eigenvalue analysis and time-domain simulations implemented in MATLAB/SIMULINK environment. The studied IACMG system’s parameters are given in Appendix [20]. The tuned parameters of the proposed hierarchical controller, given in Table 15.1, have been obtained from the design procedure, given in Sect. 15.2, considering the signal transmission time delay of 25 ms for one-way communication. The rationale behind this consideration is that microgrids are assumed to be established with the fiber optics communication links since they cover a smaller area as compared to the conventional power system, which is usually spread over a wide area and may be established with the satellitebased communication links. However, it is possible to retune the parameters of the
15 WAM-Based Hierarchical Control of Islanded AC Microgrids
391
Table 15.1 Tuned parameters of the proposed hierarchical controller for the signal transmission time delay (τ ) of 25 ms Decentralized controller Gain Parameters matrix KPDEC [607.81e−9 675.99e−9 531.23e−9 548.17e−9] KQDEC [30.16e−3 34.24e−3 14.83e−3 33.22e−3] Centralized controller Matrix Parameters Optimal state estimator QfMG [40.46e−3 46.55e−3 49.21e−3 88.98e−3 55.91e−3 39.02e−3 36.83e−3 59.84e−3 63.31e−3 22.38e−3 81.31e−3 61.47e−3] RfMG [29.07e−3 36.08e−3 68.05e−3 20.77e−3 74.78e−3 36.19e−3 81.88e−3 35.47e−3 50.85e−3 34.36e−3 46.83e−3 66.34e−3 66.43e−3 64.51e−3 39.27e−3 54.04e−3 74.81e−3 64.84e−3 51.97e−3 27.98e−3 89.15e−3 28.89e−3 34.53e−3 81.87e−3 13.59e−3 62.02e−3 78.76e−3 25.45e−3 68.14e−3] Optimal state-feedback control QcMG [875.00 226.22 667.02 150.62 687.34 821.94 714.56 472.97 130.42 817.17 338.88 343.30 293.42 342.75 522.58 578.79 688.78 951.79 190.18 788.18 942.00 611.32 151.89 290.82 342.82 744.16 219.41 318.93 631.93] RcMG [59.58e−3 336.72e−3 594.33e−3 563.39e−3 586.98e−3 156.81e−3 996.66e−3 467.93e−3 630.52e−3 616.27e−3 611.45e−3 663.39e−3]
proposed hierarchical controller by considering the different amount of time delays at the design stage and can evaluate its validity for increased time delays. It can be observed from Table 15.1 that the tuned diagonal elements of QcMG are very large as compared to the diagonal elements of RcMG . This indicates that the cost of the proposed hierarchical controller is quite small, i.e., a cheap control strategy [35]. Further, the effectiveness of the proposed hierarchical controller has been verified by comparing the results obtained with those of a hierarchical controller based on state-based extended LQG [14]. In order to have a fair comparison, both the hierarchical controllers are designed with the same tuned parameters given in Table 15.1. It is worth mentioning that the value of the prescribed degree of stability, α, has been chosen as 10 to minimize both the excessive control energy cost or controller complexity cost as well as to force the controller to act faster [35].
15.3.1 Eigenvalue Analysis To test the impact of signal transmission time delays on the performance of the proposed hierarchical controller, an eigenvalue analysis of the open-loop IACMG (without controller) and closed-loop IACMG has been performed, as given in
Without controller With the state-based hierarchical controller S.No Unstable modes Eigenvalue Damping factor Eigenvalue Damping factor Delay-free communication channels (τ = 0) 1 iCP LQ &φAL 62.98 ± j 39.21 −0.8489 −46.71 ± j 23.10 0.8964 Signal transmission time delay (τ ) = 25 ms (one-way communication) 2 iCP LQ &φAL 62.98 ± j 39.21 −0.8489 −36.00 ± j 20.79 0.8660 Signal transmission time delay (τ ) = 250 ms (one-way communication) 3 iCP LQ &φAL 62.98 ± j39.21 −0.8489 −23.70 ± j24.35 0.6975 Signal transmission time delay (τ ) = 500 ms (one-way communication) 4 iCP LQ &φAL 62.98 ± j39.21 −0.8489 4.45 ± j5.41 −0.6355 Signal transmission time delay (τ ) = 525 ms (one-way communication) 5 iCP LQ &φAL 62.98 ± j39.21 −0.8489 8.87 ± j6.61 −0.8022 Signal transmission time delay (τ ) = 530 ms (one-way communication) 6 iCP LQ &φAL 62.98 ± j39.21 −0.8489 17.46 ± j7.92 −0.9101
0.9411 0.8765 0.8422 0.6646 0.4451 -0.6975
−60.01 ± j21.55 −41.88 ± j22.99 −32.82 ± j21.06 −7.06 ± j7.94 −3.19 ± j6.41 4.74 ± j4.87
With the proposed hierarchical controller Eigenvalue Damping factor
Table 15.2 Eigenvalue analysis based performance validation of the proposed hierarchical controller for different signal transmission time delays
392 E. S. N. Raju P and T. Jain
15 WAM-Based Hierarchical Control of Islanded AC Microgrids
393
Table 15.2. It should be noted that the seven (3 complex conjugate and one real) modes lie in the unstable region (right-hand plane (RHP)) for the open-loop IACMG system [33]. However, the most critical complex conjugate mode, affected by the increased signal transmission time delays, has been considered for a better comparison of its performance. It can be observed that an increase in time delay degrades both the degree of stability as well as the damping ratios of the critical complex conjugate mode. Further, it can be seen that the proposed hierarchical controller shifts the critical complex conjugate mode to the stable region for signal transmission time delays less than 525 ms, while the critical mode shifts to the unstable region for signal transmission time delays equal to 530 ms and above (the exact value is 527 ms and above). Whereas, the state-based hierarchical controller shifts the critical mode for signal transmission time delays less than 250 ms (it was found that the exact value is 400 ms and below), while the critical mode shifts to the unstable region for signal transmission time delays greater than 400 ms. Moreover, it can be noticed that the degree of stability and damping ratios of unstable modes stabilized by the proposed hierarchical controller are more as compared to the state-based hierarchical controller. In addition to this, although the damping ratios are degraded by the increased time delays, the proposed hierarchical controller is still effectively mitigating the instability for longer signal transmission time delays up to 526 ms. Thus it can be concluded that the maximum signal transmission time delay that can be tolerated by the proposed hierarchical controller is 526 ms (one-way communication), whereas, the state-based hierarchical controller can tolerate up to 400 ms (one-way communication).
15.3.2 Time-Domain Simulations Time-domain simulation results have also been performed to assess the impact of different signal transmission time delays (25 ms, 250 ms, 500 ms, 525 ms, and 530 ms for one-way communication) on the robustness of the proposed hierarchical controller in the presence of small-signal disturbances. These smallsignal disturbances are associated with the small load disturbances at the individual bus as well as multiple buses at a time. Considering the worst case of adding a step signal to the input load disturbance at all the buses, simultaneously, as shown in Fig. 15.7, time-domain simulations of the open-loop IACMG and closedloop IACMG are shown in Fig. 15.13 and Figs. 15.14, 15.15, 15.16, 15.17, 15.18, respectively. It should be noted that, in order to show the detailed oscillations, the figures show the results only for periods of interest. However, the settling time is given in Table 15.3 to identify the time required for the response to reach and stay within a range of certain percentage (usually 5% or 2%) of the final value. Figure 15.13 shows the dynamic step response of the deviation in output variables, such as frequency and voltage at all buses, of the open-loop IACMG. It can be observed that the step response of the deviation in output variables is
394
E. S. N. Raju P and T. Jain
Fig. 15.13 Dynamic step response of the deviation in the outputs of open-loop IACMG. (a) Frequency. (b) Voltage at buses
Fig. 15.14 Dynamic step response of the deviation in frequency. (a) With the state-based hierarchical controller. (b) With the proposed hierarchical controller
increasing continuously with respect to time, reflecting instability of the openloop IACMG. It is worth reiterating that the detailed information about causes behind the instability of the open-loop IACMG can be found in [20]. Figures 15.14, 15.15, 15.16, 15.17, and 15.18 show the dynamic step response of the deviation in output variables of the closed-loop IACMG, using both the state-based as well as the proposed hierarchical controllers. It can be seen from Figs. 15.14a, 15.15a, 15.16a, 15.17a, and 15.18a that the state-based hierarchical controller has produced the stable response for signal transmission time delays varying from 25–250 ms, while the response is unstable for signal transmission time delays equal to 500
15 WAM-Based Hierarchical Control of Islanded AC Microgrids
395
Fig. 15.15 Dynamic step response of the deviation in voltage at bus 1. (a) With the state-based hierarchical controller. (b) With the proposed hierarchical controller
Fig. 15.16 Dynamic step response of the deviation in voltage at bus 2. (a) With the state-based hierarchical controller. (b) With the proposed hierarchical controller
ms and above (it was found that the exact value is 400 ms and above). Whereas, the proposed hierarchical controller has produced the stable response for signal transmission time delays varying from 25–525 ms, while the response is unstable for signal transmission time delays equal to 530 ms and above, as shown in Figs. 15.14b, 15.15b, 15.16b, 15.17b, and 15.18b. Furthermore, the time-domain simulation specification, i.e., the settling time of the dynamic step response of the deviation in output variables, shown in Figs. 15.14,
396
E. S. N. Raju P and T. Jain
Fig. 15.17 Dynamic step response of the deviation in voltage at bus 3. (a) With the state-based hierarchical controller. (b) With the proposed hierarchical controller
Fig. 15.18 Dynamic step response of the deviation in voltage at bus 4. (a) With the state-based hierarchical controller. (b) With the proposed hierarchical controller
15.15, 15.16, 15.17, and 15.18, is given in Table 15.3 for better performance comparison of the two hierarchical controllers.It can be seen that the settling time increases for an increase in signal transmission time delay, especially, for the longer time delays, i.e., 400 ms for the state-based hierarchical controller, while in case of the proposed hierarchical controller, it increases gradually for the longer time delays of 500 ms and 526 ms. It can also be observed that the proposed hierarchical controller has produced a fast settling time than the
Deviation in S.no. output 1 f (H z) 2 V1 (V ) 3 V2 (V ) 4 V3 (V ) 5 V4 (V )
With the state-based two-level hierarchical controller: settling time (ms) Without delay 25 ms delay 250 ms delay 259 302.1 375.6 168.1 404.0 472.4 295 400.0 459.0 229 359.1 428.5 319 344.7 460.4
With the proposed two-level hierarchical controller: settling time (ms) 400 ms delay Without delay 25 ms delay 250 ms delay 500 ms delay 2360 250 287.1 361.1 1253 3600 129 399.3 466.6 1150 3400 152 392.5 452.3 1088 3250 189 352.3 426.1 1042 3476 314 338.5 453.9 1056
526 ms delay 1830 2100 1990 1970 2020
Table 15.3 Performance comparison of the proposed hierarchical controller with the state-based hierarchical controller for different signal transmission time delays
15 WAM-Based Hierarchical Control of Islanded AC Microgrids 397
398
E. S. N. Raju P and T. Jain
state-based hierarchical controller. The rationale behind this is that the proposed hierarchical controller is based on modal-based extended LQG approach, which enables the addition of damping only to modes of interest while keeping other modes unaffected [24]. Thus, it can be concluded from Figs. 15.14, 15.15, 15.16, 15.17, 15.18 and Tables 15.2, 15.3 that both the proposed as well as state-based hierarchical controllers have an adequate capability to mitigate the undamped oscillations for time delays varying from 25–250 ms due to their optimal Kalman estimator that takes care of the delay in signal transmission. Furthermore, the increase in time delay degrades the damping ratios of the critical modes. In addition, time delay has a significant influence on the boundary of the small-signal stability region of microgrids, especially, when the time lag is large. Moreover, the proposed hierarchical controller is effective in enhancing the stability even with longer signal transmission time delays (i.e., 526 ms (one-way communication)) encountered in real WAMS.
15.4 Conclusion In this chapter, a wide area measurement system (WAMS) based hierarchical control of islanded AC microgrids (IACMGs) with static and dynamic loads has been presented. The proposed controller has a hierarchical (two-level) structure comprising of an upper-level centralized controller built on the top of lowerlevel decentralized/local controllers and incorporates signal transmission time delays encountered in the two-way communication. Finally, the simulation results concluded that the proposed hierarchical controller is effective in enhancing the stability of IACMGs for a wide range of time delays encountered in real WAMS.
Appendix IIDG Units Ratings: IIDG1 -(10 + j6) kVA; IIDG2 -(15 + j9) kVA; IIDG3 -(20 + j12) kVA; IIDG4 -(25 + j15) kVA. Static Active and Reactive Power Droop Gains: mP 1 = 6.28e−4 rad/s/W, mP 2 = 4.18e−4 rad/s/W, mP 3 = 3.14e−4 rad/s/W, mP 4 = 2.52e−4, nQ1 = 1.66e−3 V/VAR, nQ2 = 1.11e−3 V/VAR, nQ3 = 8.33e−4 V/VAR and nQ4 = 6.66e−4 V/VAR. IIDG unit Parameters: Lf = 1.35 mH, Cf = 50 μF, Rf = 0.1 , fsw = 8 kHz, wc = 31.41 rad/s, Kpv = 0.05, Kiv = 390, Kpi = 10.5, Kii = 16e−3, F = 0.75, fnl = 50.5 Hz, Rc = 0.03 , Lc = 0.35 mH. RIAL Parameters: Lf = 2.3mH, Cf = 8.8 μF, Rf = 0.1 , fsw = 10 kHz, wc = 31.41 rad/s, Kpv = 0.5, Kiv = 150, Kpi = 7, Kii = 25e3, Rc = 0.03 , Lc = 0.93 mH. Line Parameters: Line 1: (0.23 + j0.11) , Line 2: (0.35 + j0.58) , Line 3: (0.30 + j0.47) . Load Parameters: Induction Motor Load: 10 hp, 400 V, 50 Hz, rs = 0.7834, Lss = 127.1 mH, rr = 0.7402, Lrr = 127.1 mH, Lm = 124.1 mH, P=4, TL = 47.75 N m; CP L:
15 WAM-Based Hierarchical Control of Islanded AC Microgrids
399
12 kVA, rCP L =13.224 /phase and cosα=0.85; RI AL: 12 kW and RRI AL =40.833 ; R Load: 25 kW, RRLoad =6.347 /phase and VDC = 700 V.
References 1. Parhizi, S., Lotfi; H., Khodaei, A., & Bahramirad. S., (2015). State of the art in research on microgrids: A review. IEEE Access, 3, 890–925. 2. Chowdhury, S., Chowdhury, S. P., & Crossley, P. (2009). Microgrids and active distribution networks. London: Institution of Engineering and Technology. 3. Kundur, P. (1994). Power system stability and control. New York: McGrawHill. 4. Kundur, P., et al. (2004). Definition and classification of power system stability IEEE/CIGRE joint task force on stability terms and definitions. IEEE Transactions on Power Systems, 19(3), 1387–1401. 5. Majumder, R. (2010). Some aspects of stability in microgrids. IEEE Transactions on Power Systems, 28(3), 3242–3253. 6. Xinhe, C., Wei, P., & Xisheng, T. (2010). Transient stability analyses of micro-grids with multiple distributed generations. In Proceedings of the International Conference Power System Technology (POWERCON). 7. Narendra, K. S., & Balakrishnan, J. (1994). A common Lyapunov function for stable LTI systems with commuting A-matrices. IEEE Transactions on Industrial Electronics, 39(12), 2469–2471. 8. Majumder, R. (2013). Reactive power compensation in single-phase operation of microgrid. IEEE Transactions on Industrial Electronics, 60(4), 1403–1416. 9. Olivares, D. E., et al. (2014). Trends in microgrid control. IEEE Transactions on Smart Grid, 5(4), 905–1919. 10. Liang, H., Choi, B. J., Zhuang, W., & Shen, X. (2013). Stability enhancement of decentralized inverter control through wireless communications in microgrids. IEEE Transactions on Smart Grid, 4(1), 21–331. 11. Kahrobaeian, A., & Mohamed, Y. A.-R. I. (2014). Analysis and mitigation of low-frequency instabilities in autonomous medium-voltage converter-based microgrids with dynamic loads. IEEE Transactions on Industrial Electronics, 61(4), 1643–1658. 12. Ovalle, A., Ramos, G., Bacha, S., Hably, A., & Rumeau, A. (2015). Decentralized control of voltage source converters in microgrids based on the application of instantaneous power theory. IEEE Transactions on Industrial Electronics, 62(2), 1152–1162. 13. Raju, P. E. S. N., & Jain, T. (2017). Optimal decentralized supplementary inverter control loop to mitigate instability in an islanded microgrid with active and passive loads. International Journal of Emerging Electric Power Systems, 18(01). 14. Raju, P. E. S. N., & Jain, T. (2017). Robust optimal centralized controller to mitigate the small signal instability in an islanded inverter based microgrid with active and passive loads. International Journal of Electrical Power & Energy Systems, 90, 225–236. 15. Guo, F., Wen, C., Mao, J., & Song, Y. D. (2015). Distributed secondary voltage and frequency restoration control of droop-controlled inverter-based microgrids. IEEE Transactions on Industrial Electronics, 62(7), 4355–4364. 16. Wang, Y., Wang, X., Chen, Z., & Blaabjerg, F. (2017). Distributed optimal control of reactive power and voltage in islanded microgrids. IEEE Transactions on Industrial Electronics, 53(01), 340–349. 17. Zhao, Z., Yang, P., Guerrero, J. M., Xu, Z., & Green, T. C. (2016). Multiple-time-scales hierarchical frequency stability control strategy of medium-voltage isolated microgrid. IEEE Transactions on Power Electronics, 31(8), 5974–5991.
400
E. S. N. Raju P and T. Jain
18. Baghaee, H. R., Mirsalim, M., & Gharehpetian, G. B. (2016). Power calculation using RBF neural networks to improve power sharing of hierarchical control scheme in multi-DER microgrids. IEEE Journal of Emerging and Selected Topics in Power Electronics, 04(4), 1217– 1225. 19. Li, Z., Zang, C., Zeng, P., Yu, H., & Li, S. (2018). Fully distributed hierarchical control of parallel grid-supporting inverters in islanded AC microgrids. IEEE Transactions on Industrial Informatics, 14(2), 679–690. 20. Raju, P. E. S. N., & Jain, T. (2018). Impact of load dynamics and load sharing among distributed generations on stability and dynamic performance of islanded AC microgrids. Electric Power Systems Research, 157, 200–210. 21. Zolotas, A. C., Chaudhuri, B., Jaimoukha, I. M., & Korba, P. (2007). A study on LQG/LTR control for damping inter-area oscillations in power systems. IEEE Transactions on Control Systems Technology, 15(1), 151–160. 22. Surinkaew, T., & Ngamroo, I. (2016). Hierarchical co-ordinated wide area and local controls of DFIG wind turbine and PSS for robust power oscillation damping. IEEE Transactions on Sustainable Energy, 7(3), 943–955. 23. Stahlhut, J. W., Browne, T. J., Heydt, G. T., & Vittal, V. (2008). Latency viewed as a stochastic process and its impact on wide area power system control signals. IEEE Transactions on Power Systems, 23(1), 84–91 24. Almutairi, A. (2010). Enhancement of Power System Stability Using Wide Area Measurement System Based Damping Controller, Ph.D. dissertation, The University of Manchester, Manchester. 25. Cheng, L., Chen, G., Zhang, F., Li, G., & Gao, W. (2014). Adaptive time delay compensator (ATDC) design for wide-area power system stabilizer. IEEE Transactions on Smart Grid, 5(6), 2957–2966. 26. Milano, F. (2016). Small-signal stability analysis of large power systems with inclusion of multiple delays. IEEE Transactions on Power Systems, 31(4), 3257–3266. 27. Beiraghi, M., & Ranjbar, A. M. (2016). Adaptive delay compensator for the robust wide-area damping controller design. IEEE Transactions on Power Systems, 31(6), 4966–4976. 28. Ghosh, S., Folly, K. A., & Patel, A. (2018). Synchronized versus non-synchronized feedback for speed-based wide-area PSS: Effect of time-delay. IEEE Transactions on Smart Grid, 9(5), 3976–3985. 29. Aminifar, F., Fotuhi-Firuzabad, M., Safdarian, A., Davoudi, A., & Shahidehpour, M. (2015). Synchrophasor measurement technology in power systems: Panorama and state-of-the-art. IEEE Access, 2, 1607–1628. 30. Li, J., Chen, Z., Cai, D., Zhen, W., & Huang, Q. (2016). Delay-dependent stability control for power system with multiple time-delays. IEEE Transactions on Power Systems, 31(3), 2316– 2326. 31. Li, M., & Chen, Y. (2018). A wide-area dynamic damping controller based on robust H∞ control for wide-area power systems with random delay and packet dropout. IEEE Transactions on Power Systems, 33(4), 4026–4037. 32. National Academies of Sciences, Engineering, and Medicine and Others (2016). Analytic Research Foundations for the Next-Generation Electric Grid. Washington, DC: National Academies Press. 33. Raju, P. E. S. N., & Jain, T. (2018). A two-level hierarchical controller to enhance stability and dynamic performance of islanded inverter-based microgrids with static and dynamic loads. IEEE Transactions on Industrial Informatics, 15(5), 2786–2797. 34. Raju, P. E. S. N., & Jain, T. (2019). Development and validation of a generalized modeling approach for islanded inverter-based microgrids with static and dynamic loads. International Journal of Electrical Power & Energy Systems, 108, 177–190. 35. Xue, D., Chen, Y. Q., & Atherton, D. P., et al. (2007). Linear Feedback Control: Analysis and Design with MATLAB. New York: SIAM. 36. Deb, K., Pratap, A., Agarwal, S., & Meyarivan, T. (2002). A fast and elitist multiobjective genetic algorithm: NSGA-II. IEEE Transactions on Evolutionary Computation, 06(2), 182– 197.
Part III
Protection of Microgrids
Chapter 16
Fault Ride Through and Fault Current Management for Microgrids Wei Kou and Sung-Yeul Park
16.1 Introduction Considerable distributed energy resources (DERs) integrated into the transmission system poses many questions about the fact that the system cannot be assisted in disturbances. A DER’s response to a voltage disturbance can affect the stability of the system as defined in [1–4]. The development of stringent grid code requirement to maintain security of supply is dictated by the increased DER integration into bulk power systems. Meanwhile, microgrids (MGs) as an essential element of the modern power system have gradually used a mixture of different DERs. Potential applications for MGs can provide service to a single customer, for example, a residential (also called the nanogrid) MG or a campus MG; or to a group of customer, for example, secondary MGs, partial and full feeder MGs, and substation MGs. The industry has meanwhile become more interested in the application of MG to take advantage of the proliferation of DER to address planning objectives, for example, by improving resilience and efficiency. Fault ride through (FRT), which mainly reflects the ability of DERs to be connected during any fault occurrence, is one of the main requirements. Smart inverters play a critical role in enabling the FRT implementation in MGs by providing advanced control capabilities (for example, voltage and frequency control). Unlike the latest FRT technologies used in DERs such as large-scale solar power plants and wind power stations [5], the FRT strategy on MGs has a higher power control criterion. One is to remove power ripples, especially the double-frequency power ripples, created by the unbalanced faults. A high proportion of the doublefrequency power ripples in the power exchange between MGs and the host grid
W. Kou () · S.-Y. Park Department of Electrical and Computer Engineering, University of Connecticut, Storrs, CT, USA e-mail: [email protected] © Springer Nature Switzerland AG 2021 A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems, https://doi.org/10.1007/978-3-030-59750-4_16
403
404
W. Kou and S.-Y. Park
will contaminate the inverter-based device safety, the host grid power quality as well as the MGs local loading system reliability in the event of system failures. Furthermore, MG is run as a group with a range of loads and DGs, and it manages the output power fairly passively. It is unlike solar and wind plants, which could decrease output power rapidly by using the maximum power point tracking (MPPT) control and the wind-blade diverter, respectively. The IB-MG FRT difficulty in managing the power flow between the MG and the host network is responsible for implementing the dynamic energy management on the inverter-based interfaces. The universal FRT power control scheme is based on the dq frame of the symmetrical components of the three-phase injection currents to regulate both active and reactive power. It comprises both positive and negative sequence loops, and each sequence loop has the d- and q-axis frame current control loops [6– 9]. The symmetrical components’ decomposition and the synchronous reference frame transition must be carried out on voltages and currents, and at least four PI controllers have to be controlled in this system. Therefore, the amplitudes and phase angles of the three-phase currents cannot be openly modified with the reference current calculation algorithm based on the double synchronous reference frame. Recently, Kou et al. [10–13] developed a novel FRT strategy for directly regulating active output power in the abc reference frame without any electrical variable transformations. The decomposition and transformations of the three-phase electrical measurements, which are conducted in the double synchronous reference frame, are not needed in this new method. Its effectiveness on power flow balance and power ripple elimination is comparable with the double synchronous reference frame vector control. With the FRT technique being adapted to DER integration at a fast pace, the associated fault current increase due to the DER connections during faults becomes an issue. Especially in places where the old aging grid with fast-growing concentrated loads, the substation circuit breakers in the distribution network will be at the risk of exceeding its short-circuit duty limitation because of the fault current injections from DERs. Even if the increase in fault current does not exceed the protection limits of the installed devices, coordination of the primary and secondary protective devices may be disturbed due to excessive DER fault current. Concerning the effects of DERs on fault current increase, the capacity of DER integration through FRT will be limited, and the reliability of the whole system during fault will be decreased. Overcoming these problems clearly requires fault current management (FCM) to eliminate the effects of the fault current injection from MGs on the distributed networks. FCM is controlling the fault current injection from DERs to eliminate the fault current increase in the main grid, which is applied at the system level. FRT is controlling the three-phase current injections from DERs to maintain the power exchange between DERs and main grid during the fault, which is applied at the device level. To overcome the fault current control conflict between FCM and FRT, FCM should be designed with the compatibility with FRT on each MG. Fault current limiter (FCL) is a mature FCM device, which is deployed at the point of common coupling (PCC) and used to limit the fault current inputs of MGs through a rapid
16 FRT and FCM for MGs
405
increase in impedance [14, 15]. FCL installation provides additional investments in the integration of MGs with energy losses, triggering and retrieval times, steadystate impedance as well as the installation costs of the FCLs, and the system operators have to consider them all. The inverter-based interfaces of MGs that could achieve the same effect as FCLs are recently taken into account by flexibly modifying the fault currents. The magnitudes of the three-phase output currents could be limited to a reasonable value by using a hard limit on current referrals in a dq frame. But in unbalanced conditions of fault, the three-phase fault currents are connected and cannot be separately controlled. It is not a feasible solution. An FCM technique to preserve the present fault current level by adjusting both amplitude and phase angles from the inversion interface output current is proposed by Rajaei et al. [16, 17]. When the output current changes in the same step, the use of this FCM technique on the IB-MG directly results in a sudden power imbalance in the PCC and also affects the IB-MG’s inner device operation. The FCM analysis in this chapter was motivated by the need to encourage IB-MG FRT to provide FCM in different fault situations, especially unbalanced faults.
16.2 MG FRT in Grid Codes The large number and different types of power generation units and their connection complexity makes the transmission system operators form an operative collection of regulations to manage their integration and service in the grid, which is called grid code [18]. The incorporation of DERs within the bulk power network has gained substantial interest from grid operators when building productive distributed networks by environmental, economic, and technological opportunities [19–21]. Some basic specifications, including IEEE 1547, IEEE 2030 [22], VDE-AR-N 4105 [23], C22.2 NO.257, C22.3 NO.9, and NB/T 32015-2013 [24], have been developed to govern the operating process of the integrated DERs. For international grid code, IEEE 1574 is the earliest standard for distributed resource integration, which was published by IEEE in 2003. It can be applied to all distributed resources and has been extended to a series of standards, which cover testing, monitoring, information exchanges and control, etc. IEC/IEEE/PAS 63547-2011, the standard of distributed resources interconnecting with electric power systems, was initially converted from IEEE 1547 but not published under the standards IEC. It was used at last as PAS (Publicly Available Specification), and its specification is close to that of IEEE 1547. In Germany, the guidelines for power generation systems integrated to the medium-voltage and the lowvoltage distribution networks were released in January 2008 and August 2011, respectively. All specifications are essential for the incorporation of DERs, which refer to all power generation systems connected by synchronously motor and asynchronously engaged or transforming devices, such as wind, hydropower, and photovoltaic electricity, to the medium- or low-voltage distribution networks. Two major incorporation standards exist in Canada, for example. The DER services
406
W. Kou and S.-Y. Park
incorporated into networks below 50 kV and within total integration capability of 10 MW are specified by C22.3 NO.9; the requirements for connecting DER based on inverter to low-voltage distribution networks under 0.6 kV are specified by C22.3 NO.9. The Chinese technical standard NB/T32015-2013 was released in 2013. The standard applies to new infrastructure, restoration, and extension of distributed resources linked to power systems no more than 35 kV. While DER penetration has increased in recent years, some specifications in these standards have passive effects on the dynamic operation of the power system. IEEE 1547, the first DER integration standard published in 2003 by IEEE, for instance, has “must trip” requirements of the grid-connected DERs against system voltage deviations, whereas the host grid disturbance is worsened and the safety of power supply is at risk. In 2014, IEEE 1547 was revised to make the “must ride” specifications more flexible. Amendment 1 to IEEE 1547 requires DER to “pass around” grid tension anomalies and “grid operators and DER can agree on other tensile voltage and time settings” [25]. Inverter-based DERs have been developed and used to help maintain the reliability of the bulk power system during faults by FRT technology that allows DER exchange power to the host system under the voltage deviation case [26, 27]. As FRT technology matures, the ability to integrate FRT in DER standards such as VDEAR-N 4105 and NB/T 32015 begins to be required. It shows that DER is expected to support the bulk network in the event of failures, and FRT technology should play a critical role in this dynamic process. DERs are typically grouped into MGs with their local loads and connected to the distribution networks in urban areas. Such DERs also face technical challenges in their adaptability to their FRT capabilities. These problems could be categorized as flexible power flow control, meeting the varied failure scenarios, distributed network reliability criteria, and communication support for timely measurements. DER-related ride-through specifications are laid out in draft IEEE P1547, and they explicitly focus on voltage and frequency ride-through, as well as improvements in the voltage cycle angle and frequency. Abnormal performance requirements for ride-through disturbance are classified into categories I, II, and III. Category I is intended to meet minimum BPS reliability needs and to be achievable by all DER technologies, including rotating machines. Category II is designed to align with the requirement in NERC PRC-024-2 on the ride-through specifications for the whole converter-coupled device and other performance criteria to enable the delayed post-fault stress retrieval at the distribution level (adding some ride-through curve margins). In high DER penetration or low-inertia, distribution-related grids such as California and Hawaii, Category III is developed to align with California Rule 21(19), Hawaii Rule 14(20), and other similar rules. It focuses on the additional requirements for high DER penetration systems such as California and Hawaii, where both distribution network and BPS reliability depend considerably on the performance of DER. FRT requirements of BPS generating resources are also defined in the NERC quality standard PRC-024-2, which describes a “no trip” zone for generator frequency and voltages relay used to trip the applicable generating units. Footnote
16 FRT and FCM for MGs
407
1 of the standard clearly describes that these protective functions are part of “multifunction protective devices or protective functions” within the control systems directly operating or transmitting trip signals to the generator based on the inputs of frequencies and voltages. It is essential to realize that the control actions that perform the protective functions are part of the requirements in PRC-024-2 for BESconnected converter-coupled resources.
16.3 FRT Control in Microgrids MG is an organization of DERs capable of islanding and grid-connected operation [28–30]. AC and DC are two main categories of the microgrid networks. The DER capabilities of AC MG are combined with the AC grid. It is cheaper to construct and can be introduced without modifying the current network. If the sources and loads are both AC natured, AC MGs are more effective. DC MGs shown as in Fig. 16.2 are mostly combined with distributed energy resource and storage systems. DC MGs have the benefits of offering higher reliability, efficiency, and simplified monitoring topology over AC MGs. An IB-MG system consists of the utility, the microgrid, and the inverter-based interface for implementing FRT control. The inverter-based interface is developed to balance the total power flow between the utility and MG in the regular operation and FRT operation, and it is disabled when MG is running in the islanded operation. In this session, FRT controls for IB-MG in both AC and DC microgrids are introduced.
16.3.1 AC Microgrids To achieve a versatile power flow control between the utility and the MG under the usual and even abnormal load, the grid-connected interface of IB-MG adopts a backto-back (B2B) configuration comprised of two traditional pulse-width-modulated (PWM) voltage source converters (VSCs) with their DC sides linked via a DC link capacitor [5, 12]. Figure 16.1 displays the B2B framework and its control implementation. The utility side converter, VSC-1, is the master converter of IBMG FRT, and the MG side converter, VSC-2, is the slave converter operated by the MG frequency regulation. When a failure exists on the utility side, the fault current injected by the utilities Ig , together with the DC-link voltage Vdc , the three-phase voltages uM , and the three-phase currents at VSC-1 side iM , is detected and sent to VSC-1 controller for calculating the FRT reference current iMr ef . The three-phase voltages at MG local load side, uL , is fed in phase-locked loop (PLL) for collecting the control variable inputs, the AC local bus frequency fL , and the voltage angle θL of the dq reference frame current controller of VSC-2. The PI control on frequency variance gives the d-axis part of the corresponding reference currents iL_ref . For generating PWM signals, SM _abc and SL_abc , for both VSC-1
408
W. Kou and S.-Y. Park
Fig. 16.1 Schematic diagram of AC microgrid FRT control
and VSC-2, a deadbeat control algorithm is used [31]. See Appendix for details on the adopted controls of FRT reactive power injection Q∗ in section “FRT Reactive Power Injection,” PLL in section “Synchronous Reference Frame PLL,” and the deadbeat control algorithm in section “Deadbeat Current Control.” Before faults appear in utility, the B2B interface works under the regular operation mode and keeps the power flow balance between the utility side and the MG side. Under the normal operation, VSC-1 is controlled by the DC-link voltage control in the dq reference frame, which adopts the structure of the VSC-2 frequency PI regulator [10, 32].
16.3.2 DC Microgrids Although DC MGs have no power system frequency, the control systems of DC MGs can be studied in a framework similar to AC MGs . The control is divided into two timescales: the primary controls on the order of milliseconds, and the secondary control on the order of a few seconds, or even a minute. The “secondary” controllers typically use relatively slow communication to control power generators, storage, loads, and voltages to minimize losses. The secondary control designs in the AC or DC architectures are not substantially different. The primary control, however, reacts to quick MG transients or other upsets. In DC architecture, voltage replaces frequency as a physical variable, which can be used to communicate the imbalance of generation and burden on the “primary” control timescale [29]. The required speed of the primary response and critical nature of the control makes a communication-based control less reliable. In comparison to AC architecture, voltage is a “local” variable in the DC architectures, i.e., the voltage droop depends on the position of the charge changes relative to the generation location and the measuring point. As for the inverter interfaces connected with DERs in DC MG, in the regular operation conditions, they usually use MPPT to control the output
16 FRT and FCM for MGs
409
DC BUS DC/AC Utility
DC/DC
IB-MG FRT MPPT
DC/DC
I-V Droop Energy Storage
ENABLE
AC/DC Adaptive Droop
DC/AC
MPPT
AC Bus Load
I-V Droop ENABLE
frequecy PI control
Motor
Fig. 16.2 Schematic diagram of DC microgrid FRT control
power from solar or wind power. At the moment of the grid voltage fault, MPPT function is disabled and is replaced with I–V droop control, and the active current injection should be regulated according to the fault severity level, as shown in Fig. 16.2. The generation or storage source response across DC MG is uniform when the voltage is used as an indication of generation/load imbalance. Although it may not be optimal to configure the response in the primary control timescale, it can be modified with a secondary control at a slower time frame. In the faster primary control, however, the reaction comes exclusively from power electronic devices with more or less equivalent dynamics to prevent mixed MG control complications in AC architectures. In Fig. 16.2, the energy storage adopts adaptive droop control by adjusting droop constant rather than merely selecting a constant droop at the cost of achieving voltage regulation and load sharing simultaneously.
410
W. Kou and S.-Y. Park
16.3.3 Traditional Reference Current Control Methods The traditional methods of FRT current control involve both positive and negative sequence loops and the grid voltage synchronous reference system, which is used to regulate the active and reactive power through dq transformation [31]. The control strategy is complicated because both positive and negative frames are necessary to feed the two control loops, which are converted independently from the three-phase voltages and currents [33]. Others [32, 34] obtained the active and reactive power expressions under the framework of d-and q-axis currents and voltages. The current references can then be obtained based on the required power output. Nevertheless, there are common issues with these strategies. First of all, they take dual current control schemes for positive and negative sequences, which increases the difficulty of controls. Second, they require accurate online symmetric component decomposition in time domain, which compromises the solidity and reliability of the FRT. In what follows, two reference current control methods in dq frame are introduced.
16.3.3.1
Balanced Positive Sequence Control Method
Adopting balanced positive sequence control (BPSC) method, the goal is to control grid-connected current injection with the positive sequence components only. Thus the corresponding active power and reactive power reference values Pref and Qref can be simplified as *
Pref = 3V+ ip∗ Qref = 3V+ iq∗
(16.1)
This method is useful for maintaining the quality of the currents injected [35]. Besides, it has less technique demanding on PLL, by which the balanced currents can be generated using simple synchronous controllers, provided that the synchronization system accurately estimates the amplitudes and the phase angles of the positive sequence components of the grid voltages. The BPSC strategy calculates the active and reactive reference currents as follows: ip∗ = iq∗ =
Pref |ν+ |2 Qref |ν+ |2
· υ+ ,
(16.2)
· υ+ ,
(16.3)
16 FRT and FCM for MGs
411
where ⎤ ⎡ √ ⎤ + φ+ ) va+ √ 2V+ sin(ωt 2π v+ = ⎣vb+ ⎦ = ⎣ 2V+ sin(ωt − 3 + φ+ )⎦ , √ vc+ 2V+ sin(ωt + 2π 3 + φ+ ) ⎡
2 2 2 |v+ |2 = va+ + vb+ + vc+ =
16.3.3.2
3 |V+ |2 . 2
(16.4)
(16.5)
Decoupled Double Synchronous Reference Control Method
To minimize double-frequency ripples on DC-link voltage and the three-phase output voltages caused by the unbalanced faults, the negative sequence components of output currents are involved in the decoupled double synchronous reference control method. Using instantaneous power theory and Park’s transformation [36], one can obtain the instantaneous active and reactive power outputs P and Q as P = P0 + Pcos cos(2ωt) + Psin (2ωt),
(16.6)
Q = Q0 + Qcos cos(2ωt) + Qsin (2ωt).
(16.7)
Here, P0 and Q0 are the constant components of P and Q, Pcos and Psin are the amplitudes of real power ripple injection, Qcos and Qsin are the amplitudes of reactive power ripple injection, which can be determined by ⎡ + ⎤ ed P0 ⎢ e− ⎢P ⎥ ⎢ d ⎢ cos ⎥ ⎢ − ⎥ ⎢ ⎢ Psin ⎥ 3 ⎢ ed ⎥= ⎢ + ⎢ ⎢ Q0 ⎥ 2 ⎢ eq ⎢ − ⎥ ⎢ ⎣ eq ⎣Qcos ⎦ Qsin −ed− ⎡
eq+ eq− −eq− −ed+ −ed− −eq−
ed− ed+ −eq+ eq− eq+ ed+
⎤ eq− ⎡ ⎤ + eq+ ⎥ ⎥ id + ⎥ ⎢ +⎥ ed ⎥ ⎢iq ⎥ ⎥⎢ ⎥, −ed− ⎥ ⎣id− ⎦ ⎥ −ed+ ⎦ iq− eq+
(16.8)
where ed+ , eq+ and ed− , eq− are the constant signals being transformed from positive and negative sequence components of three-phase grid-tied voltage by the Park’s transformation. id+ , iq+ and id− , iq− are transformed from three-phase grid-tied current. Equation (16.8) shows that unbalanced sags induce double-frequency ripples of power output. For instance, ed− and eq− would not be zeros under unbalanced sags, which gives a nonzero Pcos . If eliminating the active power fluctuations is the control priority, P0 , Pcos , Psin , and Q0 should be selected as the control priority, and the appropriate dq rotation coordinates should be set as eq+ = 0 and eq− = 0. Therefore, the expression of the dq-axis reference currents could be deduced as
412
W. Kou and S.-Y. Park
⎧ ⎪ ⎪ id+ = ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ + ⎪ ⎪ ⎨ iq =
2 3 2 3
q
3
ed+ + 2 (ed ) +(ed− )2 ed− + 2 (ed ) −(ed− )2 ed− + 2 (ed ) +(ed− )2
⎪ ⎪ id− = 23 ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ 2 − ⎪ ⎩i =
ed+ P∗ + 2 (ed ) −(ed− )2 0
Q∗0 P0∗
Q∗0
−
ed− P∗ + 2 (ed ) −(ed− )2 cos
+
ed− P∗ + 2 (ed ) +(ed− )2 sin
+
ed+ P∗ + 2 (ed ) −(ed− )2 cos
−
ed− P∗ + 2 (ed ) +(ed− )2 sin
(16.9)
.
The iα and iβ could be derived through inverse Park transformation in positive and negative reference frame as ⎡ +⎤ i ⎢ d ⎥ iα cos ωt − sin ωt cos ωt sin ωt ⎢iq+ ⎥ = ⎢ ⎥. iβ sin ωt cos ωt − sin ωt cos ωt ⎣id− ⎦ iq−
(16.10)
Finally the reference currents in αβ-axis are transformed into three-phase reference currents through the inverse Clark transformation. ⎤ ⎡ ⎤ 0 ⎡ 1 √0 ia 3 ⎥ iα 1 ⎣ ib ⎦ = 2 ⎢ ⎣− 2 2√ ⎦ iβ 3 ic − 12 − 23
(16.11)
Figure 16.3 shows the operation procedures of the reference currents in the decoupled double synchronous reference control method.
Fig. 16.3 The decoupled double synchronous reference current calculation block
16 FRT and FCM for MGs
413
16.3.4 Natural Phase-Coordinates Approach Standard regulation involves both positive and negative sequence loops and uses a synchronous grid voltage reference framework to monitor both active and reactive energy by switching to dq-axis [32, 37]. The d-axis average negative sequence reference value is set to zero to eliminate double-frequency ripples on the DClink voltage and the three-phase output voltage induced by unbalanced faults. The control strategy is complicated since two control loops, transformed separately by the three-phase voltages and currents, are supplied both with the positive and negative dq frames. This section presents a new method, natural phase-coordinates (NPC) approach, for calculating reference current under abc phase coordinates rather than with symmetric component decomposition to address these problems that appeared in applying the unbalanced FRT technology. The relationship between active power ripples and current injections has been theoretically established and analyzed under phase coordinates in the unbalanced faults. The reference currents can be derived under abc phase coordinates for regulating the active base power, eliminating the active double-frequency power ripples and stabilizing DC bus voltage. The natural phase-coordinates approach is developed based on the instantaneous power theory [35]. The instantaneous active power of a three-phase system can be expressed as (16.12) p3φ (t) = pa + pb + pc = ua ia + ub ib + uc ic Ui cos(ωt − ψi )Ii cos (ωt − δi ) =
(16.12) (16.13)
i=a,b,c
=
Ui Ii [cos (2ωt − ψi − δi ) + cos(δi − ψi )] 2
(16.14)
i=a,b,c
=
i=a,b,c
P0,i +
P2ω,i .
(16.15)
i=a,b,c
where uabc and iabc are the three-phase instantaneous voltage and current phasors at the PCC of the integrated DER, respectively, under the abc phases. Equation (16.12) could be expanded into the time-domain expression with the amplitude values and the phase angle values, respectively, as shown in (16.13)–(16.15). Uabc and ψabc are the amplitudes and phase angels of uabc . Iabc and δabc are the amplitudes and phase angels of iabc . Under the balanced fault conditions, the amplitudes and phase angles of uabc are usually kept as U a = Ub = Uc , ψa = ψb +
2π 2π = ψc − . 3 3
(16.16) (16.17)
414
W. Kou and S.-Y. Park
If the output current is generated as the same balanced structure, the vector addition of the three-phase double-frequency power components of p3φ (t) is zero while the constant component of p3φ (t) remained as same as PM 0 , which is the active power injected by MG based on its local load balance. p2ω,a + p2ω,b + p2ω,c = 0,
(16.18)
p3φ = P0,a + P0,b + P0,c = PM _0 .
(16.19)
The balanced relationship of uabc in Eqs. (16.16) and (16.17) does not exist under the unbalanced fault conditions, like the single-phase-to-ground or the doublephase-to-ground faults. If iabc stays balanced with the unbalanced vabc , it breaks the balance among three-phase output power in (16.18) and generates the doublefrequency active power components in p3φ . Thus, an unbalanced reference current frame is designed in NPC approach. For a single-phase-to-ground fault, the faulted phase is set as the benchmark phase. Without loss of the generality, here phase A is set as the benchmark phase. The currents on the other two phases, iM _b and iM _c , both have δ phase angle difference to iM _a as shown in Fig. 16.5b. Considering the application field of the three-wire connection systems, the zero-sequence symmetrical components should be avoided in iM _abc . To generate symmetric threephase currents for the balanced loading system, the amplitude ratios among the symmetric three-phase currents should satisfy Icref = Ibref =
1 Ia . 2 cos(π − δ) ref
(16.20)
Thus, the generating structure of iref on the abc frame is designed as ia _ref = Iaref cos (ωt) ,
(16.21)
ib_ref =
1 Ia cos (ωt − δ) , 2 cos(π − δ) ref
(16.22)
ic_ref =
1 Ia cos (ωt + δ) . 2 cos(π − δ) ref
(16.23)
As seen, the phasor function of iref is calculated by δ. In [5], δ is determined with the assumption that uabc remains symmetrical but no longer balanced. Here, two words must be clarified: “balanced” represents the amplitudes and phase angles of the three-phase phasors, which exactly fulfill a relationship given in Eqs. (16.16) and (16.17); and “symmetrical” describes a kind of disarrangement against the “balanced” uabc , which is depicted in Fig. 16.4a, that is in mirror symmetry. After taking into consideration the irregular loading, zero-sequence impedance on transmission lines, as well as the grounding impedance, the three-phase system voltages at the PCC after the unbalanced fault typically will not be symmetrical but be asymmetrical as shown in Fig. 16.4b. In what follows, the symmetric NPC
16 FRT and FCM for MGs
415
Fig. 16.4 Phasor graphs of uabc : (a) symmetrical uabc , and (b) asymmetrical uabc
method for calculating δ under the unbalanced but symmetric faults will be introduced at first. Then, it will be expanded to a generalized NPC method for tackling the asymmetric cases.
16.3.4.1
Symmetric NPC Method
The following discussion is assumed that a single-phase-to-ground fault is happened on phase A. The amplitude ratio of uM _abc after the fault is UM _a = μUM _b = μUM _c ,
(16.24)
where μ ⊂ (0, 1) represents the depth of the voltage sag on phase A. The phase angle differences of uM _abc after the fault is still kept as 23 π as shown in Fig. 16.5a. From Eqs. (16.13)–(16.15), the double-frequency power components of pM _abc could be extracted as follows:
416
W. Kou and S.-Y. Park
Fig. 16.5 Vector diagrams for (a) uM _abc , (b) iM _abc , and (c) p2ω,abc
UM _a IM _a cos(2ωt), 2 2 UM _b IM _b cos 2ωt − π − δ , = 2 3 2 UM _c IM _c cos 2ωt + π + δ , = 2 3
p2ω,a =
(16.25)
p2ω,b
(16.26)
p2ω,c
(16.27)
whose vector diagram is shown in Fig. 16.5c. To eliminate the power ripple, the amplitudes of p2ω_abc should have the following relationship: P2ω,c = P2ω,b =
1 2 cos(− 13 π + δ)
P2ωa ,
(16.28)
to make the phasor sum of p2ω equal to zero. Combining Eqs. (16.24), (16.20), and (16.28), δ is derived as
16 FRT and FCM for MGs
417
Fig. 16.6 The MG FRT reference current calculation block
2μ + 1 + π. δ = tan−1 − √ 3
(16.29)
Here, δ has the same expression as Eq. (16.29) in the double-phase-to-ground fault with the voltage sag ratio μ in both B and C phases of uM _abc and their phase angle offsets against A phase are still even kept as 23 π . The outcome can be applied to any symmetric three-phase fault voltages cases with any predefined phase angle offsets ϕ. Thus, (16.29) could be rewritten as δ = tan
−1
cos ϕ − μ sin ϕ
+ π.
(16.30)
When ϕ = 23 π , Eq. (16.30) has the same expression of δ as in Eq. (16.29) (Fig. 16.6 and 16.15).
16.3.4.2
Generalized NPC Method
For the asymmetric three-phase fault voltages, their amplitudes and phase angles could be expressed as Ua = μb Ub = μc Uc ,
(16.31)
ϕa = ϕb + θb = ϕc − θc ,
(16.32)
where μb and μc represent the voltage sag depth on phases B and C, respectively, as shown in Fig. 16.4b. θb represents the phase angle difference between phase A and B, and θc represents between phase A and C. By applying Eqs. (16.21)–(16.23) into Eq. (16.14), the three-phase double-frequency ripples are expressed as
418
W. Kou and S.-Y. Park
Ua Ia cos(2ωt), 2 Ua Ia cos(2ωt − θb − δ), = 4μb cos(π − δ)
p2ω,a =
(16.33)
p2ω,b
(16.34)
p2ω,c =
Ua Ia cos(2ωt + θc + δ). 4μc cos(π − δ)
(16.35)
In order to satisfy (16.19), the following relationship is built up: P2ω,b cos(π − θb − δ) + P2ω,c cos(π − θc − δ) = P2ω,a ,
(16.36)
and the amplitude values in (16.33)–(16.35) are substituted to yield cos(π − θb − δ) cos(π − θc − δ) + = 1. 2μb cos(π − δ) 2μc cos(π − δ)
(16.37)
Through the trigonometric function transformation, δ is obtained as δ = arctan
cos θb cos θc μb + μc − 2 sin θb sin θc μb + μc
+ π.
(16.38)
By comparing (16.38) with the expression of δ shown in (16.29), it is concluded that the δ calculation for the asymmetrical uabc needs four measurement inputs, θb , θc , μb , μc , which are more than the only one measurement input, μ in the δ calculation for the symmetrical uabc . Utilizing the WAMS system and the PMU devices in modern power systems, these measurements are accessible. When θb = θc = 2π 3 , μb = μc , Eq. (16.38) will degrade to Eq. (16.29) and its compatibility is proven.
16.4 Fault Current Management for Microgrids Due to the fast-growing concentrated loads, IB-MGs usually connect to low-voltage distribution networks to power the loads nearby. The substation circuit breakers of those distribution networks are usually reaching their short-circuit duty limitations. To ride through a grid fault, the current contribution of IB-MGs causes the substation breakers not to cut off the short currents with the large fault currents from the utility side, especially in the first few cycles after the fault happens [38]. MGs with FRT ability can inject significant fault currents beyond current fault limit of the grid equipment and then result in catastrophic damage and possible grid instability [39]. For further analysis, Fig. 16.7 shows an equivalent circuit of a system connected with an IB-MG when a fault occurs in the upstream side of the MG at Bus_4. Utility power system transfers power through network equivalent impedance Znet
16 FRT and FCM for MGs
Z net
HV/MV
T1
Utility
419
1
2
Z1
Z2
3
MV/LV MG 2
I MG
DS substation
Ig
MV/LV
Circuit breaker
Zf
4
If
Fig. 16.7 Short-circuit current injection of IB-MG system
and transformer T1 to the distributed power system at the left side of the substation, Bus_1. The fault currents from the utility side Ig flow through the distribution transmission line with equivalent impedance Z1 and fault impedance Zf . On the right side of Bus_3, IB-MG behaves as a current source to inject IMG into fault at Bus_4 through Z2 . In this system, the three-phase short-circuit current If before implementing DG can be estimated as the fault current from the utility side Ig =
Znet
E E = . + Z1 + Zf Z + Zf
(16.39)
If it is an unbalanced fault, like a single-phase-to-ground fault on phase A, the fault current of A phase is Ig,a =
3E , Z (1) + Z (2) + Z (0) + 3Zf
(16.40)
where Z (1) , Z (2) , and Z (0) are the positive, negative, and zero sequences shortcircuit impedance of transmission line. And the angle difference between steadystate short current and utility source is ϕg,a = tan−1
ωL . R
(16.41)
After adding the injecting current from IB-MG, IM , the fault current can be obtained as [40] If =
Znet + Z1 E + IM . Znet + Z1 + Zf Znet + Z1 + Zf
(16.42)
For the low-impedance fault (Zf Znet + Z1 ), the fault current can be further approximated as [17]
420
W. Kou and S.-Y. Park
Fig. 16.8 Fault current waveforms for IB-MG system. (a) IB-MG supplies power to the utility, (b) IB-MG absorbs power from the utility
If ≈
Znet
E + IM . + Z1 + Zf
(16.43)
Here, IM is positive if IB-MG supplies power to the utility during the fault; otherwise, it is negative. With the incorporation of IM , the magnitude of the shortcircuit current If could be higher than Ig as shown in Fig. 16.8a or lower as shown in Fig. 16.8b. Thus, If is at the risk of breaking the safety limits planned for the protection system. The FCM design for IB-MG interface during FRT is manipulating the current phase angle to satisfy ig + im = ig ,
(16.44)
which is realized by shifting the phase angle of im to eliminate the magnitude increase of if . The magnitude Ig _a and phase angle ϕg _a of ig in the fault phase are known as shown in Fig. 16.9, with a single-phase-to-ground fault occurred in phase A as an example. Along with changing the amplitude of im , the range of ϕM _a variation could be any point on the circle with the radius Ig . Thus, the effect of extra current injection from MG on increasing the short-circuit level of the utility grid could be neutralized. Considering the direction of power flow, ϕM _a belongs to (− π2 , π2 ). Based on the reference current structure of FRT in Eqs. (16.21)–(16.23), a compatible FCM strategy is proposed for single MG FCM in (16.4.1). Then, the FCM strategy illustrated in (16.44) is generalized in (16.4.2) for a multiple MG
16 FRT and FCM for MGs
421
Fig. 16.9 Phasor illustration of FCM strategy on the single-phase-to-ground fault
FCM case: a utility network with multiple MG connections, where im represents the phasor sum of the fault currents from each MG.
16.4.1 Single MG Fault Current Management To achieve Eqs. (16.44), (16.18), and (16.19) at the same time, the structural form of the three-phase reference current, iM _ref , is designed as follows: iref _a = M cos (ωt + R ) ,
(16.45)
iref _b =
1 M cos (ωt − δ + R ) , 2 cos (π − δ)
(16.46)
iref _c =
1 M cos (ωt + δ + R ) , 2 cos (π − δ)
(16.47)
of which δ is given by Eq. (16.29) to determine the relative relationship on amplitudes and phase angles of iM _ref . Based on δ, iM _b and iM _c are changed to iM _b and iM _c , respectively, as shown in Fig. 16.10. With the three-phase reference currents for FRT designed in Eqs. (16.21)– (16.23), the formulation of iM ref for FCM is modified to satisfy Eqs. (16.44) and (16.19) by introducing two controllable variables (R and M ). iM _a and
422
W. Kou and S.-Y. Park
Fig. 16.10 Phasor illustration of the MG FRT reference current construction
iM _bc are expanded by the same amplitude ratio M and are rotated by the same phase angles R . Thus, the active power flow balance described in (16.27) can be represented as
2 1 UM _ b M UM _a cos R + cos R − δ + π 2 2 cos (π − δ) 3 2 UM _c cos R + δ − π = PM _0 . + 2 cos (π − δ) 3
(16.48)
M is usually configured to fulfill (16.19), as a time-varied value regulated by the close-loop DC-link voltage control. Given M , R could be obtained from the FCM strategy in Eq. (16.44). The calculation of R is explained in further detail below for the different kinds of unbalanced faults. We still take phase A as the faulted phase to discuss a single-phase-to-ground fault case. The existing short circuit occurs only in phase A, and phase B and C have no short-circuit injection. The FCM strategy shown in Fig. 16.10 gets M as follows: . R = ϕM _a = π − ϕg _a − arccos
M 2Ig
.
(16.49)
FCM also remains aware of the short-circuit currents in both fault phases for the double-phase-to-ground faults. Suppose phase B and phase C were the failure
16 FRT and FCM for MGs
423
Fig. 16.11 Phasor illustration of FCM for the double-phase-to-ground fault
phases, the total fault current injecting in the fault location is if _b , if _c . The fault currents on the utility side are defined as ig _b , ig _c for the two fault phases with . phase angles ϕg _b , ϕg _c as shown in Fig. 16.11. R = αM bc can be calculated as follows: αM _b + αM _c . , R = ϕM _bc = 2
(16.50)
where αM _b = π − cos αM _c = cos
−1
−1
M 4 (π − cos δ) Ig _b
M 4 (π − cos δ) Ig _c
+ δ − ϕg _b ,
(16.51)
+ δ + ϕg,c ,
(16.52)
to eliminate the increase of the short-circuit current on phase B and C at the same time.
424
W. Kou and S.-Y. Park
Fig. 16.12 Schematic diagram of multiple MGs interconnected to a 27-kV distributed grid
16.4.2 Multiple Microgrids Fault Current Management The current FRT development is focused primarily on the single MG integration scenario. The related fault current increase due to the numerous MG connections during faults is becoming a concern by applying the FRT technique to a power system with the fast-paced MG integration. The FCM organizational architecture should be applied to the multi-MGs deployment scenarios. The pictorial depiction of a distributed network connected to N MGs is provided in Fig. 16.12. The distribution system is a conventional 10-kV radial network in one way and the MG connection is power independent from the other networks; that is, the only power exchange link for its MG is managed by the inverter-based device. If Bus_4 has a fault situated on the top of the MG links, the remote utility source transfers electricity to the distribution grid on the left side of Bus_1 through a transmission network with the lumped impedance Zg to the transformer T1. The short current ig injected by the remote utility grid flows through Z12 and contributes to the fault current if flowing from Bus_2 to Bus_4. On the right side of the Bus_3, MGs serve as numerous current sources, and a total fault current IMG is generated at Bus_3 and injected into the Bus_4 through Z23 and Z24 . With multiple MGs integration, we approximate the short current if as [17] i f ≈ ig +
iMG,k .
(16.53)
The multi-MG FRT strategy is adjusting iMG,i of each MG within its operation range, and the N MGs are cooperating together to eliminate the amplitude change of if , that is, Minimize If − Ig Subjectto iMG,k ⊆ Ik , k = 1, 2, · · · , N,
(16.54)
16 FRT and FCM for MGs
425
where Ik reflects an MG k operational scope under faults. Each MG has its inverterinterfaced FRT control by the NPC method, which produces the fault current seen in Eqs. (16.45)–(16.47). Section 16.4.1 has demonstrated how to achieve δ, which provides the profile information of iref _abc on the amplitudes and phase angles. Here, we rewrite Eqs. (16.45)–(16.47) as follows: ia _ref = IMG cos (ωt + ϕ) ,
(16.55)
ib_ref =
1 IMG cos (ωt − δ + ϕ) , 2 cos (π − δ)
(16.56)
ic_ref =
1 IMG cos (ωt + δ + ϕ) . 2 cos (π − δ)
(16.57)
The other two controllable#parameters, IMG and ϕ, of iref _abc must, therefore, be employed to manage iMG,k to realize the optimization goal function in Eq. (16.54). Thus, by changing IMG and ϕ, the three-phase reference currents iabc_ref are magnified by the same times as IMG and rotated by the same angles as ϕ. It should be noted that each MG has the individual operation condition: the inverter-based interface capacity, time-varied output power, as well as the distance to a specific fault location. It defines the scope of the fault current, iref _abc , that each MG could generate during the FRT operation mode. It is denoted as the “FRT zone” of MG fault current iMG . Let N = {1, 2, · · · , N} be the set of MG indices with FRT ability, C = {C1 , C2 , · · · , CN } be the set of capacity indices of each MG, and P = {P1 , P2 , · · · , PN } be the set of active output power for each MG before a host grid fault. If Pk > 0 (k ⊆ N), MGi is injecting Pk to the connected host network; otherwise, if Pk < 0 (k ⊆ N), MGk is absorbing Pk from the host network. The power balance constraint for each MG is Ub 2 1 IMG,k Ua cos ϕk + cos ϕk + δ − π 2 2 cos (π − δ) 3 Uc 2 cos ϕk − δ + π = Pk + 2 cos (π − δ) 3 (16.58) with the power electronic interface operation constraint: max IMG,k ,
√ 3 IMG,k × UN,k × √ ≤ Ck , 2 cos (π − δ) 2
(16.59)
where UN,k represents the utility side nominal voltage of the bus connected with MG k. The allowable amount of fault current injection from MG k, (IMG,k , ϕk ), which is identified in Eqs. (16.58) and (16.59), is displayed as the related FRT zones in Fig. 16.13. It shows all the phasor positions of iMG,k that MG k could produce under the faults. Those current phasors start at O, end at any point on the
426
W. Kou and S.-Y. Park
Fig. 16.13 FRT zone for the fault current injection
line segments, Ak Bk , and form a triangular track, Ak OBk , which is the visualization of MG k’s FRT zone. For those FRT zones located in the right plane (Ok > 0), the associated MG injects power Pk to the utility during faults and vice versa. The intersection point on IDER cos ϕ axis, Ok , is decided by the exchange power Pk . The distance of OOk is related to the power capacity exchanging between the utility and MG k during the faults. The inverter-based interface capacity, Ck , determines the maximum value of IMGk in the FRT zone.
16.5 Hardware-in-the-Loop Platform for MG FRT Given the fact that setting up the in-field electrical power system test system is a time-consuming and weather-dependent process, which needs high investment and lacks flexibility, it is challenging to meet the experimental requirements of the FRT controller design, upgrades, and certifications test based on the aim of academics or business. Hardware-in-the-loop (HIL) is an emergent research field that provides both component-level and system-level testing a new approach. The physical or rapid control prototyping for the HIL application is linked to, instead of a physical plant, a virtual plant performed on a real-time simulator. Real-time digital simulator (RTDS) is a simulator developed by RTDS Technologies to carry out the real-time power system simulations. It is designed to simulate
16 FRT and FCM for MGs
427
Fig. 16.14 Co-simulation HIL test setup
Fig. 16.15 The power system configuration in RSCAD
in real-time versions of the control network in 50 μs and to simulate electronic devices in the time steps smaller than 2 μs. RTDS is a mix of software and hardware experimental platforms. The architecture of the hardware consists of optical signal processors, RISC processors, I/O cards, and power supply, arranged into single racks. The simulator can communicate with external devices through analog and digital I/O ports. RSCAD, a personalized app package, supports the user-to-user interface. RTDS, OP4510 simulator, level shifting circuit, and DLP750P scope recorder are used for building this co-simulation testbed. RTDS has been used in real-time simulating grid and failure phases. The inverter was powered by OP4510. In the GTAO card of RTDS, the voltage and current measurements are scaled down and used on the OP4510 analog display. Six PWM signals are transmitted from OP4510 to RTDS GTDI. The fault-triggering condition is detected by the GPC pin in RTDS. For voltage and current waveform monitoring, DL750P scope recorder was used. The testbed layout is shown in Fig. 16.14 [13] (Fig. 16.15).
428
W. Kou and S.-Y. Park
Fig. 16.16 The control system diagram in OPAL-RT
On the other side, by using OPAL-RT OP4510 [41], the FRT control device on the HIL testbed has been incorporated. The analog inputs of the FRT control model are the scaled-down voltage and current measurements, Vabc and Iabc , which are within ±20-volt scale. At the same time, it feeds RTDS six PWM signals. Platforms of 16-bit resolution [42] are available on ADC platforms. By using the additional circuit, the PWM signals are created between 0 V and 5 V, and up to 180 kHz. The OP4510 is equipped with the Intel quad-core 3.3 GHz Intel Xeon processor, which can run up to 50 kHz without any overrun. The Opal-RT simulation control models are depicted in Fig. 16.16, and the process flowchart of the RTDS and OPAL-RT co-simulation platform is shown in Fig. 16.17.
16.6 Analysis and Results On the distributed power grid network, as seen in Fig 16.12, the developed MG FRT methods under the utility faults with the unbalanced voltage sag are tested. To have the high switch frequency inverter simulation model, we use the RTLAB real-time simulation OPAL-RT toolbox, which is completely combined with MATLAB/Simulink. The device specifications and parameters of the simulation are listed in Table 16.1.
16 FRT and FCM for MGs
429
Fig. 16.17 Co-simulation workflow Table 16.1 Simulation test configuration Network specifications Utility system voltage and frequency IB-MG interface capacity Line posi-/zero-sequence resistance Line posi-/zero-sequence inductance Length of line 1−2, 2−4, 2−3 Simulation settings Simulation step size IB-MG interface converter PWM frequency
27 kV, 60 Hz 1 MW 0.2153/0.813 /km 1.05e−3/3.02e−3 H/km 20 km, 20 km, 5 km 50 μs 10 kHz
At t = 1 s, Bus_3 induces a single-phase-to-ground fault, and the inverterinterfaced DER device switches its output current control from the standard operating mode to FRT mode by utilizing three separate approaches: (1) the traditional method under the dq frame, (2) the symmetric NPC method, and (3) the generalized NPC method. The comparison results of these three methods on the three-phase output voltages and currents vabc , iabc , the active power p0 , and the double-frequency active power p2ω on Bus_4 are shown in Fig. 16.18. When the fault arises, an asymmetrical vabc deviated from the balanced threephase voltage profile at the Bus_3 has occurred. Apart from the apparent decrease in the amplitude of va , there are different sag ratios, μb = 0.225 and μc = 0.2, in
430
W. Kou and S.-Y. Park
Fig. 16.18 Left: FRT operation under the dq frame; middle: FRT operation with symmetrical NPC method; right: FRT operation with generalized NPC method. (a) vabc . (b) iabc . (c) p0 . (d) p2ω
nondefault phase voltages vb and vc . The discrepancies in phases for vabc are not the same as 120◦ : va and vb have 131◦ phase variation, and vc and vb have 238◦ . Thus, the voltage profile on Bus_3 connected with DER is asymmetrical. During the fault, the output currents of the traditional dq frame method and the symmetrical NPC method are unbalanced, as shown in Fig. 16.18 right(b) and Fig. 16.18 middle(b), while the DER holds the p0 steady, as seen in Fig. 16.18 right(c) and Fig. 16.18 middle(c), following the fast fluctuation in the first two frequency cycles. By comparing to the other two approaches, the efficiency of the generalized NPC approach for the reduction of power rips through unsymmetrical failures is validated. The double-frequency ripples, p2ω , have been eliminated in Fig. 16.18 right(d), while the peak-to-peak value of p2ω is 2.1 × 105 W in the symmetrical NPC as shown in Fig. 16.18 middle(d) and is increased to 4 × 105 W in the traditional dq frame method in Fig. 16.18 left(d). For asymmetrical voltage failures, the generalized NPC method works better at power ripples elimination. The proposed FRT control scheme is also realized on the co-simulated platform consisting of RT-LAB and Opal-RT simulator. Figure 16.19 displays the three-phase output voltages vabc and currents iabc of the inverter-based interface at the PCC to the utility. After the unbalanced failures, the proposed NPC FRT regulation produces the unbalanced three-phase fault currents. Phase A’s magnitude is higher than that of phase B and phase C in the single-phase-to-ground fault case. The magnitude of phase C is higher than that of phase A and phase B in the double-phase-to-ground fault case.
16 FRT and FCM for MGs
431
Fig. 16.19 (a) Real-time HIL simulation results on single-phase-to-ground fault at phase A; (b) Real-time HIL simulation results on double-phase-to-ground fault at phase A and phase B
16.7 Summary This chapter discussed available IB-MG integration requirements and control technologies from FRT capability viewpoint and system aspects. A new FRT control strategy embedded IB-MG to realize flexible NPC current control during faults is proposed. The related FCM strategies for single MG FRT, as well as multiple MGs FRT, are formulated. A co-simulation HIL testbed comprised of OPAL-RT and RTDS real-time simulators is developed. The motivation of this work was to explore DER integration in a bulk power system where MG FRT development plays an essential role in system resilience improvement and coordination management. The real-time experimental results also identify the control performance of the proposed MG FRT on output active power and three-phase currents.
Appendix 16.7.1 FRT Reactive Power Injection In recent years, many utilities and network system operators are posing requirements for large distributed generators to support grid voltage in a specific voltage drop range (the grid voltage sag depth μ ≥10%) [18]. In FRT control, the MPPT feature of DER is disabled and is replaced by the DC-link voltage control. The IB-MG interface has the ability to provide not only the injection of active power but also the contribution of reactive power. This requires the grid-connected MG injecting a reactive current iq∗ in proportion to μ, as shown in (16.60). Under specific severe scenarios (μ ≥50%), the current injection can even be purely reactive.
432
W. Kou and S.-Y. Park
iq∗
⎧ ⎨
0 0.9 μ < 1.1 = 2 · μ · IN 0.5 0.9 μ < 1.1 ⎩ μ < 0.5 IN
(16.60)
16.7.2 Synchronous Reference Frame PLL The PLL based on the synchronous reference frame (SRF-PLL) is the most popular technique used for frequency-insensitive grid synchronization in threephase structures. By using the Park’s transformation, as shown in Fig. 16.20, the standard SRF-PLL converts the three-phase voltage vector from the abc natural reference frame to the dq rotating reference frame. This dq reference frame’s angular position is controlled by a feedback loop, which controls the q element to zero. ⎡ ⎤ ⎡v ⎤ a 1 1
va 2 1− −√2 √2 = Tαβ · ⎣vb ⎦ = · ⎣vb ⎦ 3 3 3 0 2 − 2 vc vc
(16.61)
√ √
1 cos(−2ωt) vd+ = Tdq+ vαβ = 2V+ = + 2V− vq+ 0 sin(−2ωt)
(16.62)
√ √
1 cos(−2ωt) vd− = Tdq− vαβ = 2V− + 2V+ , vq− 0 sin(−2ωt)
(16.63)
vα vβ
vdq+
vdq− = where
Tdq =
Fig. 16.20 Block diagram of the SRF-PLL
cos θ sin θ . − sin θ cos θ
(16.64)
16 FRT and FCM for MGs
433
16.7.3 Deadbeat Current Control The deadbeat controller is part of the predictive controller family. These are based on a common principle: choose the state of the converter, ON–OFF predictive status, or the average voltage generated by the converter (predictive with a PWM) based on foreseeing the evolution of the controlled quantity (the current) [31]. According to the Kirchhoff voltage law of a three-phase symmetrical system, the loop voltage equation is ⎧ di di ⎨ Ua − Ub = −L dta + L dtb + Uuv dib dic U − Uc = −L dt + L dt + Uvw ⎩ b dia c Uc − Ua = −L di dt + L dt + Uwu
(16.65)
The related loop voltage equation is ⎧ di di ⎨ ua − ub = −L dta + L dtb + (du Udc − dv Udc ) dib dic u − uc = −L dt + L dt + (dv Udc − dw Udc ) ⎩ b dia c uc − ua = −L di dt + L dt + (dw Udc − du Udc )
(16.66)
dk = 1 is defined as k-phase upper bridge switch is on and lower bridge switch is off. dk = 0 is defined as k-phase upper bridge switch is off and lower bridge switch is on. We name the control period as T and the grid base period as Tg . Let T Tg to discretize the loop voltage equation, and it can be considered that the three-phase grid voltage and the DC bus voltage are constant within a control period. Then, the loop voltage equation in a control cycle could be deduced as ⎧ ib∗ −ib ia∗ −ia ⎪ ⎨ ua − ub = −L ∗ T + L T + (du Udc − dv Udc ) i −i i ∗ −i ub − uc = −L b T b + L c T c + (dv Udc − dw Udc ) ⎪ ∗ ⎩ i −i i ∗ −i uc − ua = −L c T c + L a T a + (dw Udc − du Udc )
(16.67)
Among them, ia∗ , ib∗ , and ic∗ are three-phase current reference values; du , dv , and dw are the three-phase switching duty cycles of the inverter. Since ua + ub + uc = 0, the system of (16.67) has only two independent equations. At the same time, in a switching cycle, the total on-time of the three switching tubes on the upper arm of the three-phase inverter and the corresponding total on-time of the lower arm are equal. Thus, we could have ⎧ i ∗ −ib i ∗ −ia ⎪ ⎨ua − ub = −L a T + L b T + (du Udc − dv Udc ) ∗ i −i i ∗ −i ub − uc = −L b T b + L c T c + (dv Udc − dw Udc ) ⎪ ⎩ du + dv + dw = 1.5.
(16.68)
The duty cycle of the PWM signal of the three-phase inverter could be solved as follows:
434
W. Kou and S.-Y. Park
⎧ ⎪ ⎪ ⎪ ⎪ du = ⎪ ⎪ ⎪ ⎨
i ∗ −i i ∗ −i i ∗ −i i ∗ −i 1.5Udc +2 ua −ub +L a T a −L b T b + ub −uc +L b T b −L c T c 3Udc i ∗ −i i ∗ −i i ∗ −i i ∗ −i 1.5Udc − ua −ub +L a T a −L b T b + ub −uc +L b T b −L c T c
dv = ⎪ 3Udc ⎪ ⎪ ⎪ ib∗ −ib ib∗ −ib ia∗ −ia ic∗ −ic ⎪ ⎪ −2 u 1.5U − u −u +L −L −u +L −L a c dc b b T T T T ⎪ ⎩d = . w 3Udc
(16.69)
References 1. Ackermann, T. (2005). Wind power in power systems. Hoboken, NJ: Wiley. 2. Jauch, C., Sørensen, P., Norheim, I., & Rasmussen, C. (2007). Simulation of the impact of wind power on the transient fault behavior of the Nordic power system. Electric Power Systems Research, 77(2), 135–144. 3. Ellis, A., & Gonzalez, S. (2014). Implementation of Voltage and Frequency Ride-Through Requirements in Distributed Energy Resources Interconnection Standards 4. Nair, N.-K. C., & Qureshi, W. A. (2014). Fault ride-through criteria development. In Renewable energy integration (pp. 41–67). New York: Springer. 5. Kou, W., Wei, D., Zhang, P., & Xiao, W. (2015). A direct phase-coordinates approach to fault ride through of unbalanced faults in large-scale photovoltaic power systems. Electric Power Components and Systems, 43(8–10), 902–913. 6. Gkavanoudis, S. I., Oureilidis, K. O., & Demoulias, C. S. (2013). Fault Ride-Through Capability of a Microgrid with WTGs and Supercapacitor Storage During Balanced and Unbalanced Utility Voltage Sags (pp. 231–236). 7. Moursi, E., Shawky, M., Xiao, W., & Kirtley, J. L. (2013). Fault ride through capability for grid interfacing large scale PV power plants. Generation, Transmission & Distribution, IET, 7(9), 1027–1036. 8. Rodriguez, P., Timbus, A. V., Teodorescu, R., Liserre, M., & Blaabjerg, F. (2007). Flexible active power control of distributed power generation systems during grid faults. IEEE Transactions on Industrial Electronics, 54(5), 2583–2592. 9. Wang, F., Duarte, J. L., & Hendrix, M. A. M. (2011). Pliant active and reactive power control for grid-interactive converters under unbalanced voltage dips. IEEE Transactions on Power Electronics, 26(5), 1511–1521. 10. Kou, W., Wei, D., Zhang, P., & Xiao, W. (2015). A direct phase-coordinates approach to fault ride through of unbalanced faults in large-scale photovoltaic power systems. Electric Power Components and Systems, 43(8–10), 902–913. 11. Kou, W., & Park, S.-Y. (2017). Generalized direct phase-coordinates control of inverterinterfaced distributed energy resources on fault ride through strategies. In Power Symposium (NAPS), 2017 North American (pp. 1–6). New York: IEEE. 12. Kou, W., & Wei, D. (2018). Fault ride through strategy of inverter-interfaced microgrids embedded in distributed network considering fault current management. Sustainable Energy, Grids and Networks 15, 43–52 13. Kou, W., Rakiul Islam, S. M., & Park, S.-Y. (2018). Co-simulation testbed for unbalanced fault ride through hardware in the loop validation. In 2018 IEEE Electronic Power Grid (eGrid) (pp. 1–5). New York: IEEE. 14. Zheng, F., Deng, C., Chen, L., Li, S., Liu, Y., & Liao, Y. (2015). Transient performance improvement of microgrid by a resistive superconducting fault current limiter. IEEE Transactions on Applied Superconductivity, 25(3), 1–5. 15. Ghanbari, T., & Farjah, E. (2013). Unidirectional fault current limiter: an efficient interface between the microgrid and main network. IEEE Transactions on Power Systems, 28(2), 1591– 1598.
16 FRT and FCM for MGs
435
16. Rajaei, N., Ahmed, M. H., Salama, M., & Varma, R. (2014). Fault current management using inverter-based distributed generators in smart grids. IEEE Transactions on Smart Grid, 5(5), 2183–2193. 17. Yazdanpanahi, H., Li, Y. W., & Xu, W. (2012). A new control strategy to mitigate the impact of inverter-based DGs on protection system. IEEE Transactions on Smart Grid, 3(3), 1427–1436. 18. Tarafdar Hagh, M., & Khalili, T. (2019). A review of fault ride through of PV and wind renewable energies in grid codes. International Journal of Energy Research, 43(4), 1342– 1356. 19. Colombo, E., Bologna, S., & Masera, D. (2013). Renewable energy for unleashing sustainable development (Vol. 10, pp. 978–983). Cham: Springer International Publishing. 20. Guttromson, R., & Glover, S. (2014). The Advanced Micro-Grid Integration and Interoperability. 21. Hossain, J., & Pota, H. R. (2014). Robust control for grid voltage stability: High penetration of renewable energy. New York: Springer. 22. Basso, T. S. (2014). IEEE 1547 and 2030 standards for distributed energy resources interconnection and interoperability with the electricity grid (Vol. 15013). Golden, CO: National Renewable Energy Laboratory. 23. Cr˘aciun, B.-I., Kerekes, T., Séra, D., & Teodorescu, R. (2012). Overview of Recent Grid Codes for PV Power Integration (pp. 959–965). 24. Jiang, L., Wang, C., Huang, Y., Pei, Z., Xin, S., Wang, W., et al. (2015). Growth in wind and sun: Integrating variable generation in China. IEEE Power and Energy Magazine, 13(6), 40–49. 25. IEEE standard for interconnecting distributed resources with electric power systems - amendment 1. IEEE Std 1547a-2014 (Amendment to IEEE Std 1547–2003), May 2014 (pp. 1–16) 26. El-Naggar, A., & Erlich, I. (2017). Short-circuit current reduction techniques of the doubly-fed induction generator based wind turbines for fault ride through enhancement. IET Renewable Power Generation, 11, 1033–1040. 27. Baghaee, H. R., Mirsalim, M., Gharehpetian, G. B., & Talebi, H. A. (2017). A new current limiting strategy and fault model to improve fault ride-through capability of inverter interfaced DERs in autonomous microgrids. Sustainable Energy Technologies and Assessments, 24(C), 71–81. 28. Dahale, S., Das, A., Pindoriya, N. M., & Rajendran, S. (2017). An overview of DC-DC converter topologies and controls in DC microgrid. In 2017 7th International Conference on Power Systems (ICPS), pp. 410–415. New York: IEEE. 29. Backhaus, S. N., William Swift, G., Chatzivasileiadis, S., Tschudi, W., Glover, S., Starke, M., et al. (2015). DC Microgrids Scoping Study. Estimate of Technical and Economic Benefits. Technical report, Los Alamos National Lab (LANL), Los Alamos, NM. 30. Fei, G., Ren, K., Jun, C., & Tao, Y. (2019). Primary and secondary control in dc microgrids: A review. Journal of Modern Power Systems and Clean Energy, 7(2), 227–242. 31. Teodorescu, R., Liserre, M., et al. (2011). Grid converters for photovoltaic and wind power systems (Vol. 29). Hoboken, NJ: Wiley. 32. Junyent-Ferré, A., Gomis-Bellmunt, O., Green, T. C., & Soto-Sanchez, D. E. (2011). Current control reference calculation issues for the operation of renewable source grid interface VSCs under unbalanced voltage sags. IEEE Transactions on Power Electronics, 26(12), 3744–3753. 33. Shawky, M., Moursi, E., Xiao, W., & Kirtley, J. L. Jr. Fault ride through capability for grid interfacing large scale PV power plants. IET Generation, Transmission & Distribution, 7(9), 1027–1036. 34. Eloy-Garcia Carrasco, J., Tena, J. M., Ugena, D., Alonso-Martinez, J., Santos-Martin, D., & Arnaltes, S. (2013). Testing low voltage ride through capabilities of solar inverters. Electric Power Systems Research, 96, 111–118. 35. Akagi, H., Watanabe, E. H., & Aredes, M. (2017). Instantaneous power theory and applications to power conditioning. Hoboken, NJ: Wiley. 36. Song, H.-S., & Nam, K. (1999). Dual current control scheme for PWM converter under unbalanced input voltage conditions. IEEE Transactions on Industrial Electronics, 46(5), 953– 959.
436
W. Kou and S.-Y. Park
37. Camacho, A., Castilla, M., Miret, J., Vasquez, J. C., & Alarcón-Gallo, E. (2012). Flexible voltage support control for three-phase distributed generation inverters under grid fault. IEEE Transactions on Industrial Electronics, 60(4), 1429–1441. 38. Baran, M. E., & Mahajan, N. R. (2007). Overcurrent protection on voltage-source-converterbased multiterminal dc distribution systems. IEEE Transactions on Power Delivery, 22(1), 406–412. 39. Salcedo, R., Bokhari, A., Diaz-Aguiló, M., Lin, N., Hong, T., de León, F., Czarkowski, D., Flank, S., McDonnell, A., & Uosef, R. E. (2016). Benefits of a nonsynchronous microgrid on dense-load LV secondary networks. IEEE Transactions on Power Delivery, 31(3), 1076–1084. 40. Wang, Q., Zhou, N., & Ye, L. (2015). Fault analysis for distribution networks with currentcontrolled three-phase inverter-interfaced distributed generators. IEEE Transactions on Power Delivery, 30(3), 1532–1542. 41. OPAL-RT Technologies (2015). Op4510 RT-LAB-RCP/HIL Systems User Guide. 42. OPAL-RT Technologies (2015). Op5340-1/op5340-2 User Guide: Analog to Digital Converter Module.
Chapter 17
Microgrid Protection Arturo Conde Enríquez, Yendry González Cardoso, and José Treviño Martínez
17.1 Introduction Developments in the field of electricity generation in recent decades have been oriented toward environmental conservation and greater efficiency, due to the increase in fossil energy costs. This has modified the vertical structure of electrical networks mainly affecting medium- and low-voltage networks, resulting in isolated electrical systems or weak connection electrical networks. In recent years, the installation of renewable generation sources (RASs) in medium- and low-voltage networks has undergone a great deal of growth, and the horizontal structure of electrical networks has led to the formation of microgrids (MGs). These electrical networks have operational benefits since the power of the load is local and the problems of network congestion, power losses, demand peaks are mitigated, hence reducing the operation and generation costs. These networks also have the characteristic that they integrate highly dynamic elements with significant limitations on controllability. The connection of generation sources with an intermittent nature determines both the topological and dynamic power conditions of the power grid. The energy supplied by the MG must be reliable, and of sufficient quality that the electrical network can continue to operate under adequate conditions. This has prompted the development of distributed control systems with hybrid architectures combining local and distributed control systems. These can improve the operation of the power grid and the quality of protection. MG protection systems are particularly affected by highly dynamic topologies and operating conditions. The traditional philosophy for adjusting protections considers the worst operating conditions of the electrical network. These adjustments,
A. C. Enríquez () · Y. G. Cardoso · J. T. Martínez Faculty of Mechanical and Electrical Engineering, Autonomous University of Nuevo Leon, Nuevo Leon, Mexico © Springer Nature Switzerland AG 2021 A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems, https://doi.org/10.1007/978-3-030-59750-4_17
437
438
A. C. Enríquez et al.
which are traditionally constant, often result in problems with sensitivity, selectivity, or inadequate operating times. Adaptive relay systems offer an opportunity to increase the protection of MG systems. The online settings of protection schemes can be modified depending on the topological and operational conditions of the electrical system. A multi-agent system (MAS) can be used to improve protection performance for electrical networks under dynamic operating conditions.
17.2 Microgrids MGs show highly dynamic behavior in terms of both their topology and their active elements. The integration of an RAS into the power grid has multiple impacts, as non-dispatchable generation sources with an intermittent nature require both voltage and energy support elements. These sources are decoupled from the frequency of the AC network, and since they do not involve energy storage do not provide inertia to the electricity network; it is, therefore, necessary to supply the low inertia for these systems by other means, for example through the use of storage systems or converters. In addition, the proximity of the loads implies an additional dynamic that must be considered. The reactive power demand of the electronic elements imposes a condition of high reactive power demand, and since this is not provided by the utility company, it must be provided by converters or storage systems. In addition, the connections of the shunt elements must be analyzed to avoid harmonic amplification when generating frequency resonance points. These aspects of operation must be considered in the design of protection systems for the MG. In the operation of the MG, it is necessary to use DMS to implement strategies for the operation of the network from the economic and operational points of view, such as the dispatch of units and energy management. The design of DMS systems should consider the contribution of converters as one of the sources of voltage support, an important aspect of which is the characterization of converters in terms of their diversity of design or operation and their behavior under short circuit conditions. It is also necessary to establish operation criteria in real time, due to the high operational intermittency. This condition also affects protection systems, since conventional criteria do not meet the requirements of the MG in many cases. The use of adaptable online adjustment systems, a centralized control system, and the activation/deactivation of protection schemes will ensure the creation of dynamic protection zones that offer both sensitivity and acceptable operating times.
17.2.1 MG Protection Low-voltage electrical systems have traditionally been designed to operate radially, meaning that protection techniques are generally shaped by overcurrent protection. These protection schemes are traditionally adjusted to maximum or minimum
17 Microgrid Protection
439
border values to ensure dependable and/or safe operation, requiring constant regulation, and without monitoring the operational state of the power grid. The topological changes arising from interconnected configurations, bidirectional flows, and the highly intermittent operation of MGs mean that it is difficult for conventional schemes to meet the protection requirements of an MG. We must therefore consider the design of online protection systems with centralized control systems and continuous monitoring of the operational state of the network, with systems for readjustment of the relays and the use of dynamic protection zones. It is important that these protection schemes are consistent with the mode of operation of the MG. When operating in connected mode, the utility company provides the reactive power required for voltage support and the main contribution of the fault current. Although the sensitivity of the relays is not compromised, the location of the fault can be problematic due to the flat values of the magnitude of the fault currents. A detection system for island operation is required if the MG cannot operate while temporarily isolated. In the case of disconnected operation, the need for a protection system is even greater due to the reduced fault current, temporary loss of voltage regulation, and the highly dynamic and intermittent operation of the MG.
17.2.2 Problems and Functional Solutions for Relays The use of communication and measurement channels through μPMU is generally required as part of the design of the protection scheme. The algorithms like Newton Raphson for power flow compute can be omitted and replaced by the phasor measurement. The phasor update must be less than those used to calculate the demand. The fault current must be updated before topological changes are made. In particular, the calculation of n-1 contingencies in flows and faults for the readjustment of relays must be considered. Data updates must consider the latency of the communication system and the dynamics of the variables that will be updated, either due to topological changes or variations in the load current. The functional limitations are different for each aspect of protection, and simpler protections such as overcurrent will be more strongly affected by the dynamic conditions of the MG, while those that require more information about the electrical system with multipoint measurements (such as differential protection) or the measurement of a greater number of signals (such as distance) will be less affected. It is important to evaluate the coordination between the different protection principles. An adjustment and collaboration strategy between relays can be performed in an MAS, where each relay can be an agent. The MAS system can be integrated using a collaborative structure between protection agents and measuring agents, for example. Agents can also be integrated via communication networks to form the protection system of the electrical network. The limitations and areas of opportunity for each aspect of protection are described below.
440
A. C. Enríquez et al.
Overcurrent relays. Directional overcurrent relays (DOCRs) must be used in mesh systems with selective operation against bidirectional currents in the protected line. For protection in networks with a limited current contribution to the fault, voltage retention schemes may be required in order to increase the selectivity. Traditionally, the use of standardized time curves has been a good solution for the adjustment of relays, and the adjustment of the initial current should be carried out via the maximum value of the load current, power transfer scenarios, and n1 contingencies. However, due to the coordination requirements of an MG, these criteria are generally inadequate. The use of digital relays that can change the setting parameters remotely through communication channels, in pre-established periods or when detecting any topological change in the network, allows various adaptable solutions to be incorporated, thus minimizing the limitations and increasing the applications of MGs. The design of nonconventional flexible curves with more degrees of freedom allows a time curve to be designed for each specific coordination problem. In an MG, the time curve for an overcurrent relay may change according to the coordination requirements under these conditions of operation. Likewise, updating the pickup current with the measured load current flow will allow the sensitivity of the relay to be increased and adaptable settings can be updated, to ensure that safety is not compromised. The coordination of DOCRs has been formulated as an optimization problem, and metaheuristic methods have generally been used due to their robustness in finding solutions in schemes lacking an initial condition. In this approach, an objective function is established and used to perform a search, in order to minimize the operation time of the relays and thus ensure that the coordination requirements are met. In this way, equality and inequality restrictions are established to comply with the relay time calculation and the sequence of operation between the primary and backup relays. Heuristic optimization methods can search for solutions over a larger area than deterministic ones, which gives them an advantage in terms of finding better solutions, although they are typically expected to be slower. In the search for new solution methods, approaches such as differential evolution (DE) and gray wolf optimization (GWO) have been proposed, which offer adequate calculation time for online coordination applications. In the weighting of the optimization method used for online coordination, the dispersion of the results obtained must be considered, since heuristic methods will give different results from each execution. Some approaches such as genetic algorithms (GAs) have a high level of dispersion in the results, that is, with substantial differences in each run, while DE gives a small standard deviation, with very similar results in each run. Relay tuning is commonly done off-line and it is a constant under the dynamic operation of the electrical grid. However, in applications for MG protection, due to dynamic operating conditions such as topological changes and the intermittent contributions from renewable sources, applications are needed that give solutions within a few seconds to enable online coordination. A flow chart for online coordination applications is presented in Fig. 17.1. Updating of the relay settings must be carried out in a stable state, so topological changes and the demand currents must be considered when updating the pickup current and the coordination of relays.
17 Microgrid Protection
441
Fig. 17.1 On-line coordination diagram
The use of a micro phasor (μPMU) in the measurement system can simplify these calculations and avoid the need for computation of power flow. It is important to constantly calculate fault currents since the contributions from renewable sources and topological changes have a great impact on the criteria for setting relays. Differential relays. Differential protection is one of the most efficient types of protection; it has existed for more than 100 years and has a wide range of applications. The differential principle has absolute selectivity and has been used
442
A. C. Enríquez et al.
both for the protection of primary equipment and for the protection of transmission lines and distribution feeders. The trip decision is based on a comparison of the input currents and output currents of the equipment to be protected. The operation of the differential protection system can be affected by conditions that increase the error of differential current, and factors such as a change of tap in transformers, unequal saturation of the current transformers, and internal high impedance can compromise its performance. The sensitivity of differential protection in active distribution network applications may be compromised under conditions of high fault impedance and weak source (WS) inputs. In general, under external fault conditions where the flow of current entering and leaving the scheme does not make a significant difference, protection is not affected. However, connection via electronic converters generates nonfundamental frequency components, and together with the connection of nonlinear loads and the presence of shunt capacitors, which are used to compensate for the associated low voltage in this type of network, high-frequency pollution scenarios commonly arise. In feeder schemes, the magnitude of the nonfundamental frequency components changes depending on the impedance of the transformer, generating a differential current of error that is associated with inter-harmonics and subharmonics (Fig. 17.2). Designs for new differential protection schemes based on positive sequence fault components have been reported to be able to mitigate the effect of active distribution networks due to the high penetration of distributed generators (DGs). The design of dynamic protection zones, in which the nodes of the protection zone are modified depending on the topology or operating conditions of the MG, will allow for greater sensitivity. A differential scheme will be designed in a control center, based on the measured synchronized signals. The communication channels may be wired or wireless; this gives great operational flexibility, and the advantages or disadvantages of communication must be evaluated. A further important aspect to consider is that differential protection requires a backup system. Distance relay. The principle underlying this form of protection involves the apparent impedance of the line, which results from the measurement of two input signals, voltage, and current. The apparent impedance calculated by the relay coincides with the impedance of the line when power is supplied to a single source; as the voltage is a function of the fault current, the impedance is independent of the operational state or the Thevenin equivalent. However, when there is a bilateral supply, an offset is generated between the current measured by the relay and the current from the other end, causing an imaginary component in the fault impedance. Relay detection will depend on the characteristics of the design and their adjustment, so the reach for its first zone in short lines with small impedance is difficult to obtain. In addition, the characteristics of the distance relay are affected by the presence of intermediate sources, with underreach or overreach depending on the strongest contribution and the pre-fault status (Fig. 17.3). Thus, operation times may be longer or short than desired. The presence of inter-harmonics or subharmonics, which are generated mainly by wind power plants (WPPs), can cause erroneous impedance signals.
17 Microgrid Protection
443
Fig. 17.2 (a) Differential relay in an MG; (b) external fault current contaminated with subharmonic and inter-harmonic components
In networks where sensitivity to fault detection is lost, the use of distance relays can be an attractive alternative. Their coordination with overcurrent relays has been well studied and is feasible. In this approach, the intermittency of an RAS does not affect the range of the distance relay as in the overcurrent relay, although the effects of intermediate sources and contamination of the input signals to the relay must be evaluated.
17.2.3 Discussion The operation of the MG in protection systems needs to be evaluated. The operation in mode connected to the utility offers operational support to the MG through the
444
A. C. Enríquez et al.
Second zone X
4 Without infeed
3
2
With infeed
R
Fig. 17.3 Infeed effect in distance relays
contribution of reactive power and the contribution of fault current, the intermittent dynamics of the MG are less representative as well as the performance of the protection system. On the other hand, island operation, in which the system is disconnected from the utility or its connection through high-impedance links, imposes greater requirements on the protection system, island detection schemes must be activated in order to switch the settings of the protection schemes. If the fault is in the utility, the protection system may not detect this contribution, and weak source detection is necessary for the separation of the MG. An MAS can improve the function of protection schemes in an MG. Computational platforms must allow for the integration of various dynamics of elements of the electrical network in a centralized way. Through the real-time monitoring of electrical quantities and topological detectors, a multi-agent environment can carry out the different protection actions that are required in the power grid.
17.3 Multi-Agent System Proposals Distributed control is one method that has been proposed to make energy distribution more efficient. The development of distributed control is closely associated with communication, and it is increasingly necessary to use intelligent devices to control or remotely monitor distribution networks. The development of microprocessors, microcontrollers, and programmable logic controllers (PLCs) has resulted in the emergence of distributed control [1].
17 Microgrid Protection
445
Applications currently under investigation for use in distributed control are: • • • •
Restoration of the electrical system; The operation of active distribution networks; Micro network control; and. Control of electrical systems.
In modern intelligent networks, there are ever more types of new elements, making them increasingly complex [2]. Cooperative distribution has been receiving a great deal of attention as a promising technology for the growth of wireless networks [3, 4]. The authors of [2] propose a new protection scheme for distributed control for a distribution system using an MAS with a rapid relay response and high efficiency under different conditions. One of the most common phenomena that occur during abnormalities in an electrical power system, and especially in short circuits, is an increase in current over normal operating values [5]. The operation of a protection system must also be safe and must avoid tripping under normal operating conditions, such as overloads, power transfer due to topological changes and intermittency, and a lack of control over RASs. Due to the significant interest in the use of MAS, the design of solutions the standards and methodologies are considered [6]. MASs are now used in many applications such as diagnosis, monitoring, power system restoration, market simulation, network control, and automation. This technology has matured to the point where the first MAS is being migrated from the laboratory to a physical system, allowing the industry to gain experience in the use of MASs and to evaluate their effectiveness [6]. The IEEE Power Engineering Society (PES) formed a working group in conjunction with the PSACE committee to investigate these questions in relation to MASs. In this section, a comparison is made with existing technologies for the design and implementation of MASs, such as web services and grid computing.
17.3.1 Definition of a Smart Agent There are many definitions of an agent, and although these differ, they share a basic set of concepts: the notions of an agent, its environment, and autonomy. According to Wooldridge [7], an agent is merely “a software entity (or hardware) that is placed in an environment and is capable of reacting to changes in those environments.” The environment is simply defined as everything external to the agent. At least in order to be placed in an environment, one part of it could be observable or that could be altered by the agent. An agent can alter the environment by taking an action: either physical (such as closing a normally open point in a network), or computational (saving diagnostic information in a database). Obtaining an answer back of the agent should not affect the actions that are taken or the goals for which it was designed. Autonomy means that agents “execute control over their own actions,” meaning that they can initiate or schedule certain
446
A. C. Enríquez et al.
actions for execution based on environmental observations. From an engineering perspective, this definition is problematic, as it is not easy to distinguish an agent from another element of the network. For example, protection from an overcurrent relay could be considered as an agent, as it is located within an environment (that is, the power system), and reacts to changes in its environment (that is, changes in voltage or current). It also has a degree of autonomy. An intelligent agent has the following three characteristics: • Reactivity: An intelligent agent is able to react to changes in its environment and take action based on these changes and the function for which it was designed. • Proactivity: Intelligent agents exhibit behavior with targeted goals. This means that an agent can dynamically change its behavior in order to achieve its goals. For example, if an agent loses communication with another agent that was providing a service required to achieve its goal, it can look for another agent that can provide the same service. Wooldrige describes this proactivity as the ability to “take the initiative” [7]. • Social skill: Smart agents are able to interact with other smart agents. Social ability involves more than simply passing data between different entities, and requires the ability to negotiate and interact cooperatively. This is usually performed using an agent communication language (ACL), which allows agents to chat rather than simply passing data.
17.3.2 Multi-Agents in the Electricity Sector Container architecture is proposed in which the participating agents with their different skills are configured in them. Participating agents are grouped into crews to integrate the protection system (Fig. 17.4). Within the JADE platform, there is a flat hierarchy in which each agent communicates with the others to achieve good coordination. The assigned neighborhood for each agent is an electric system component, for example, a source, load, or protection system. Any agent can take a request from another agent within their neighborhood if it does not have the ability to make any changes. • An agent can detect and warn of any change in the power system. • An external agent can block a request. • More than one algorithm can be associated with an agent. Relays can have different functions, such as measuring or relay protection, and a relay agent may have different skills (see Fig. 17.5) depending on the protection request. The proposed scenario is composed of two agents: a relay agent, which runs the relay algorithms, and a measurement agent, with the cluster of measurement signals. Agents coordinate based on their abilities to obtain better results.
17 Microgrid Protection
447
Fig. 17.4 Proposed architecture for agent gang
Fig. 17.5 Proposed architecture
17.3.3 Gang of Agents A measurement agent has a detection system for registering a certain bounded variation in measured signals. It communicates via TCP/IP with the distribution network and another agent, and in the Jade environment, this is achieved through agent communication language (ACL). If a violation is detected, the system can request support or raise an alarm. For example, it ensures that the initial contribution of the wind generator is not considered a fault and that incorrect protection action is not generated. A relay agent contains the relay algorithms. Its response will not only depend on the local signal but will also be a reply to other relays or measurements received from other points of the electrical network. An agent may send a remote trip signal or request re-coordination if necessary.
448
A. C. Enríquez et al.
Specialist agents can be designed to handle special requests from the system. For example, a relay agent can take remedial actions when a low voltage or low frequency is detected, and load trip schemes can be defined for the current conditions of operation of the MG. A reinitiating agent may reset the relay to its initial conditions in the case of intermittency in communication.
17.3.3.1
Agents in JADE
JADE was used as a development and simulation platform for the construction of the agents (Figs. 17.6 and 17.7). As mentioned earlier, TCP/IP is used for communication with the power network, which is modeled in Simulink. The measuring agent continuously monitors the network, and in the event of a change in the micro network, it warns the relay agent to make the necessary adjustments. The relay agent allows the relay to operate normally in the event of an abnormal condition such as a fault and notifies the measuring agent to avoid the readout during a fault condition. Figure 17.7 shows an agent operation request and the different support requests, using Foundation for Intelligent Physical Agents (FIPA) architecture. Communication between the electrical network, the measuring, and relay agents within the JADE tool is shown. Figure 17.8 shows how support requests are made between agents. First, a support request is made by the measurement agent is done if a change is needed. The Fig. 17.6 Activation of the measurement agent in Jade
Fig. 17.7 Gang of agents in Jade
17 Microgrid Protection
449
Fig. 17.8 Support requests between agents
Fig. 17.9 Algorithm for general operation
relay agent accepts the request for support. The exchange between the measurement agent and the relay agent is based on the following general algorithm (Fig. 17.9). The advantage of the use of agents in these algorithms is that they can be replicated with the same characteristics without the need for additional code, simply by invoking the agents necessary to enable efficient monitoring of the assigned neighborhood. Figure 17.10 shows a group of two measuring agents and a relay, which communicate through ASL in JADE to achieve cooperation.
450
A. C. Enríquez et al.
Fig. 17.10 Gang of two measuring agents and a relay agent
17.4 Implementation in JADE Architecture Agent-oriented programming (AOP) is a relatively new software paradigm that introduces concepts from theories related to artificial intelligence to the domain of distributed systems. AOP essentially involves the application of models using a collection of components called agents that are characterized by aspects such as autonomy, proactivity, and capacity for communication [8]. This framework facilitates the development of complete agent-based applications through an execution environment that implements the life cycle support features required by the agents, the central logic of the agents themselves, and a rich set
17 Microgrid Protection
451
of graphic tools. Since JADE is written entirely in Java, it benefits from a wide range of language features and third-party libraries and therefore offers numerous programming approaches and abstractions that allow developers to build multi-agent JADE systems with minimal experience in agent theory. Jade was initially developed by the research and development department of Telecom Italia s.p.a., but is now a community project and is distributed as open-source software under the LGPL license (http://jade.tilab.com/). It is important to mention that in all aspects related to the interoperability of the nucleus, Jade complies with the standards proposed by FIPA [8].
17.4.1 Communication Language Used by Agents The language used by agents involves a number of elements, including: • • • •
The messenger. The receiver. The act of communicating. The content.
The agent communication language (ACL) allows the transmission of a set of knowledge that will be expressed in a content language. The terms of the content language that represent this knowledge will belong to a vocabulary common to the different agents, called an ontology. More information on the FIPA-ACL ontology is given in the Appendix. The content of a message within JADE has the following structure: (request :sender (agent-identifier :[email protected]) :receiver (agent-identifier :[email protected]) :ontology Rele-Setting :language FIPA-SL :protocol fipa-request :content ”“((action (agent-measuring : [email protected]) (book-Agent :service 15/06/2017 :departure 05/07/2017 ... ) ))”“
17.4.2 Structure of the JADE Work Environment The work environment in JADE is handled by containers. JADE has a main container that has several sub-containers. The two most important of these containers are defined in the FIPA standards for the handling of agents as follows:
452
A. C. Enríquez et al.
1. The agent management system (AMS) is the agent that supervises the JADE platform. The AMS controls the cycle from when it is invoked until the agent service ends. The AMS generates an ID for each agent to identify it from the others. 2. The facilitator agent (directory facilitator, DF) is the “Yellow Pages” agent. Each agent must register with the DF to list the service they can provide. An agent may use the DF to seek support from agents with certain characteristics and to work together. In order for the agent to perform various actions and the activities registered in the DF, it uses internal JADE functions, for example, Setup and AddBehavior. The Setup function serves to initialize the conditions, while the AddBehavior function serves to generate the behavior. A simple example of an agent is given below [8]: import jade.core.Agent; public class HelloWorldAgent extends Agent { protected void setup() { // Printout a welcome message System.out.println(”Hello World. I’m an agent!“); } }
To generate the executable from the agent console, the following instruction is used: javac -classpath HelloWorldAgent.java
and to run the agent: java -classpath ;. jade.Boot Peter:HelloWorldAgent
For example, the a1 monitoring agent used in the graphical environment is illustrated in Fig. 17.11. The console indicates that the agent is waiting for the first signal, as shown in Fig. 17.12. The agent code is used to communicate via both Matlab and Simulink. Once the agent is finished, it is invoked in JADE from the operating system console.
17.4.3 Testing and Measurement Tool This tool allows the user to open a new host, create a host server, and receive or send data via a client port, which in this case would be in Simulink. The testing and measurement tool displays the resources (hardware, drivers, interfaces, etc.) accessible from the toolboxes associated with the tool, and allows for configuration and communication with those resources (Fig. 17.13). One of the tools in the Instrument Control toolbox is the Query Instrument block. The Query Instrument block (Fig. 17.14) configures and opens an interface for an instrument, initializes it, and consults it to obtain data. Configuration and
17 Microgrid Protection
453
Fig. 17.11 Monitoring agent
Fig. 17.12 Agent a1 waiting for a message
initialization are carried out at the start of execution, and the block queries the device during execution. The block has no input ports but has an output port corresponding to the data received from the device [9].
454
A. C. Enríquez et al.
Fig. 17.13 Graphic interface tool Fig. 17.14 Block for receiving information
17.4.4 Communication Between Matlab/Simulink and JADE JADE cannot connect directly with Matlab since it does not support parallel operations, which are essential for MASs. MACSimJx is an extension of MACSim (Multi-Agent Control Simulation), a simulator that allows the user to work with Simulink to implement MASs; however, unlike JADE, it does not use the FIPA standard [10]. FIPA was developed by the IEEE Computer Society to promote standards for agent-based technology. To solve this problem, MACSimJx can be used. Based on MACSim and an extension in Java, it allows the user to interconnect with JADE to create agents that can interact with the models in Simulink [11]. One
17 Microgrid Protection
455
disadvantage of this approach is that the MACSim and the Jx extension are only for 32-bit operating systems, and it is difficult to configure them for 64 bits; in addition, they are incompatible with the newer versions of Matlab. A more direct approach is to take advantage of the TCP/IP protocol sockets via which JADE and the Instrument Control Toolbox communicate in Simulink, and this is the method proposed in this work. The authors of [12] analyzed different ways of creating sockets with Matlab/Simulink, and in [13], JADE and PowerWorld were also connected via sockets. To make a connection in JADE, we take advantage of the fact that the code is Java, and the following libraries are added to agents: import java.net.ServerSocket; import java.net.Socket;
A power grid with distributed generation is then represented in Simulink. The multi-agent environment was built in JADE, and in order to communicate with both environments, it was necessary to establish a communication protocol. The TCP/IP protocol was used to achieve this, and the model is illustrated in Fig. 17.15. In this model, a software layer is removed by removing the interface between JADE and Simulink, and connecting JADE directly via TCP/IP, as shown in Fig. 17.15. Remote communication can be implemented with blocks. The TCP/IP architecture is client-based, and client blocks allow data to be sent from Simulink models to an application or other computers using TCP/IP. It is also possible to send data to TCP/IP server blocks. The server block accepts data from the network sockets using the TCP/IP protocol and the block mode. Data are received at fixed intervals based on Simulink scheme cycles. The basic element of the block is the S-function block, which uses a C MEX file [13]. Fig. 17.15 Proposed communication model
456
A. C. Enríquez et al.
17.5 Test Scenarios In this section, we discuss the application of protection schemes to MGs, in which MAS architectures are used to make the proposals feasible. It is assumed that communication between the relay and measurement agents takes place in the different nodes that make up the MG. The latency times of these communication channels are generally low (less than half a cycle), meaning that their application in the proposed protection scheme is feasible. It is assumed that there are communication platforms for remote programming that will allow us to define parameters to adjust the relays. There are currently different packages on the market that offer this possibility.
17.5.1 Detection of Weak Infeed Conditions Energy management systems (EMSs) must be used to control the network since the operation of the electricity grid with unregulated sources is complicated and the efficiency of the network is degraded by the lack of dispatch of these sources. The operation of interconnection relays is based on the protection of unregulated sources to avoid damage during sustained operation in unfavorable conditions. The disconnection criterion is based on the measurement of electrical quantities that are sensitive to an energy imbalance, such as the function of loss of synchronism, frequency variation, and commonly for fault detection, voltage depression, and overcurrent. Remote faults, where the GD makes only a small contribution (weak source), can be very difficult to detect by conventional or interconnection protection systems. The weak and sustained contribution from the network will continue to feed the fault, maintaining the potentials, and avoiding reclosing operations to maintain continuity of the electrical service. Figure 17.16 shows a probable WS condition. Note that the interconnect relay (bus C) may not detect the contribution of the WS to the remote fault F. Due to the architecture of MAS and the relay agents, it is possible to obtain sensitive and fast operation. Agents have particular functions that are integrated into the MG protection scheme. Relay agents can also perform additional functions in response to changes in demand or the topology of the electricity grid. The proposed structure for the detection of weak sources is not affected by the configuration of the power grid, and the affected area can be represented by means of equivalents of the main contributions and through the WS link. Two operational conditions are considered: a stable state and a fault condition. Stable state. The actualization of relays is based on the topological conditions detected by sensors. Several operating conditions can be established depending on the load or generation trip, and the power flow through the link can change direction based on specific coordination requirements. The relay settings will depend on the measured current flow, which is used to identify steady state events (Fig. 17.17).
17 Microgrid Protection
457
Fig. 17.16 Weak source contribution
Fig. 17.17 Dynamic pickup settings of the relay agent
Relay coordination can be difficult due to the low contribution from the WS. The operating criterion for the GD is therefore different from that in power grids. For low-voltage networks, the GD must be disconnected using island detection schemes, while for high-voltage networks, a disconnection criterion must be established based on the utility’s network code. The relay agents must be able to be reconfigured depending on the different topological states of the power grid. The use of an MAS can complement these functions by communicating with neighboring agents to ensure coordination between relays. Failure Status. The low contribution from the WS represents a problem in terms of loss of sensitivity of the protection system. This contribution will depend on the
458
A. C. Enríquez et al.
type of WS. In conventional generators, the contribution to the fault is sustained, meaning that through continuous pre-fault readjustment, detection can be achieved by varying the pickup current and re-coordination between relays. In unconventional generators, the contribution is limited in magnitude and time, and a transferred trip is necessary since the local relays cannot be reset.
17.5.1.1
Small Conventional Generators
The MAS performs actions based on the switch position, which is used to modify the pickup current and verify online coordination with other relays, giving greater sensitivity and a reduction in operating times. In Fig. 17.18, the relay BC has no sensitivity, and source D will continue to contribute to the fault even when the backup AB is operational. Although the reduced source contribution from D does not mitigate damage to the equipment, it prevents the reclosing of the BC line. The pickup current must be updated to
Fig. 17.18 Re-coordination of RDB for a fault on the line BC: (a) electrical network; (b) coordination chart
17 Microgrid Protection
459
improve the sensitivity. When the starting current is modified, relay coordination on the BD line is necessary when the WS is triggered, and on the DB line when the D-load is triggered. Coordination between the relay on the BC line and the relay on the DB line for a fault on the BC line is shown in Fig. 17.18. This coordination is achieved only for the topological scenario described here, and in other scenarios, the relay agent will achieve adjustment via the MAS. In this operational scenario, the requirement is that the relay on the DB line must operate slowly, while the relay on the BC line must be fast. To achieve this, the relay coordination must be as shown in Fig. 17.18b. In the case where the relay on line BC does not operate, the backup of relay AB will have, but a very long backup time of Relay DB will have for the low fault contribution. The WS, therefore, continues to feed the fault, preventing successful closure. The use of nonconventional curves can reduce backup times. The information exchanged between agents is illustrated in Fig. 17.19.
17.5.1.2
DG Generators
In electrical systems with large fault magnitudes, relays generally have adequate levels of sensitivity and allow for proper operation by correctly discriminating between normal operation and failure. The use of DG, wind (WP), or photovoltaic (PV) generators will compromise protection, due to their low contributions [14, 15]. For instance, the fault current for a double-powered induction generator is only three cycles at 3 p.u., while the failure contribution of a PV is only 1.1 p.u. A loss of sensitivity of the protection system is very likely, and the operation of the relay on the DB line cannot be determined for conventional DOCRs.
Fig. 17.19 Information exchanged during online coordination
460
A. C. Enríquez et al.
The proposed MAS enables the execution of a transferred trip that emulates the relay operation on the BC line (Fig. 17.20). An integration process is used to compute the operating time of the overcurrent relay. The coordination time interval (CTI) is added to achieve coordination, and the transferred trip is then performed. The relay on the DB line emulates the operation of the relay on the BC line by measuring the MAS of the measured fault current and uses the CTI to ensure coordination. Throughout this process, it is avoided to transfer scenarios with fixed times, which are large since they are determined for the worst-case scenario, and this affects the WS trigger times. When implementing the detection system, it is necessary to consider the latency times and computational capacity.
Fig. 17.20 Overcurrent relay using MAS: (a) electrical system, (b) transfer trip signal
17 Microgrid Protection
461
17.5.2 Detection of Islanding Condition Island operation results when the MG does not have the capacity to feed the loads in the distribution network without the utility supply. This results in a loss of network voltage support, and reconnection could result from synchronism, causing damage to the RAS and loads; finally, the island results in an isolated system. The MG should therefore be disconnected as soon as possible immediately after the loss of the utility [16]. The IEEE 929-1988 standard specifies the disconnection of the DG once it is on the island, while the IEEE 1547-2003 standard stipulates a maximum delay of 2 s for the detection of non-intentional island operation [17, 18]. Due to the aforementioned factors, there are various works in the literature dedicated to the detection of electric island operation. The development of anti-island protection systems has been widely reported, but only frequency and voltage variation approaches have been implemented using relays with explicit functions for the separation of the MG (Table 17.1). Disconnection of the island represents losses both for the users and for the owners of the MG. The solution to the problems of island operation is based on finding a balance between generation and load, with a range of applications that are limited by the capabilities of the MG but that considerably reduces the number of cases in which the system is completely lost. Island operation can be addressed by keeping the system running until reconnection to the network is possible while giving priority to the supply of energy in ranges of acceptable quality. The period of island operation can be reduced to a few minutes due to the conditions of the primary source and the variation in solar radiation and wind speed.
17.5.2.1
Detection Schemes
The main objectives of anti-island protection are to detect an island condition and to actuate the switch at the point of common coupling (PCC) between the MG and the distribution network. The aims are to avoid damage to the system isolated by the operation due to frequency values that exceed the limits of the system, to allow for restoration without complications from the supply of the main network, and to avoid reconnection arising from synchronism of the two systems, which can cause severe damage to the electrical network, the RAS and the connected loads. In general, island detection methods are classified into two main groups (Fig. 17.21). Firstly, remote methods are based on the use of a communication channel between the main network switches and generators. Local methods are classified into active and passive approaches. The former interact directly with the operation of the power system by injecting disturbances that will result in a significant change in the system parameters when the MG is isolated, but an insignificant change when connected to the network. The method used in [33] involves a frequency shift algorithm in sliding mode, which injects a small disturbance into the phase shift
462
A. C. Enríquez et al.
Table 17.1 Most effective anti-island protection methods Classification Title Passive Frequency change over power [19] [20]
Voltage change rate and power factor
[17]
Voltage imbalance and THD in the current
[21]
ROCOF relay performance for integrated generation applications Over/low-frequency relays
[22]
[23]
Transient signals: (i) design and implementation; (ii) performance evaluation
[24]
Model similar to Thevenin’s method
[25]
ROCOF y THD
[26]
Relay ROCOF active
[27]
WTP in negative voltage sequence
Active [28]
Automatic phase shift method
Year 2001
Method Based on the effect of the frequency change rate on load power 2001 Monitoring of instantaneous voltage and current signals measured 2004 Monitoring of the voltage balance between the phases of the load connected to the DG and the harmonic current of the network 2005 Proposes the concepts of detection time against power imbalance curve 2006 Frequency estimation for protection of generator detection on island and disconnection based on the frequency 20102012 Extraction of the energy coefficients of transient signals of V and C using a wavelet transform 2015 Equivalent linear network model of the system based on local measurements, voltage, and current phasors at the fundamental frequency 2015 Uses the rate of change of frequency and harmonic content of the equivalent reactance seen in the location of the DG 2017 The performance of the existing relay is evaluated and modifications are suggested in relation to the NDZ 2018 Analysis of the negative sequence component in the PCC using a wavelet transform package (WTP) 2003 Based on the phase displacement of the inverter’s output sinusoidal current (continued)
17 Microgrid Protection
463
Table 17.1 (continued) Classification [29]
Title SFS island detection method
Year 2011
[30]
Scheme based on an adaptable threshold
2016
Local hybrid [31]
Power line signaling-based technique for anti-islanding protection Economic dispatch application of power system
2007
[32]
Fig. 17.21 Classification of island detection methods
2016
Method Analyzes the impact of the active power frequency dependence of the load on the performance of the FSS method Parameters are estimated adaptively according to the level of operational penetration Estimation of voltage imbalance given by sequences 1 and 2 and frequency Uses a two-step genetic algorithm; a load-based scheme based on priorities is proposed
Island Detection Remote
Local
Active
Passive
Hybrids
mode. The authors of [34] used a method of active frequency deviation, which works by adding zero amplitude for a short period in the waveform of the inverter-based output current. Passive local methods operate based on the monitoring of signals in the terminals of the MG tp detect island formation without neglecting over/low-frequency conditions. The ROCOF method is presented [22], which uses a relay to monitor the change in frequency with respect to time.
17.5.2.2
Implementation
The model implemented in this work is illustrated in Fig. 17.22. The system includes a wind farm of 20 wind turbines, which add a nominal generation power of 4 MW. The wind farm is connected to the electricity grid at the distribution level via a small ring network that becomes part of the distribution system. The values of the
464
A. C. Enríquez et al.
Fig. 17.22 Model of the MG system Fig. 17.23 Principle of ROCOF anti-island protection
line impedances are typical distribution values with sections 1 km long, the losses in the micro network are considered. The load blocks are distributed to different areas using four buses. Frequency change rate method (ROCOF). The basic principle is the monitoring either of the frequency at the terminals of the MG or the local loads. Figure 17.23 shows the equivalent circuit for a synchronous generator equipped with a ROCOF relay, operating in parallel with the distribution network. The generator feeds a local load, and if the system is disconnected by a fault, the MG is isolated using the CB switch and a power imbalance is generated due to a loss of power of the main network. This imbalance causes the system frequency to change, and the rate of frequency change df/dt can be used for island detection. The ROCOF relay calculates the rate of frequency change using a measurement window of a few cycles on the nodal voltage wave (typically between 2 and 40 cycles). The signal is processed using low pass filters, and the resulting signal is used to detect the island. If the rate of change of the frequency is greater than an adjustment threshold, a trip signal is immediately sent to the generator switch. Typical ROCOF settings installed in 60 Hz systems are between 0.1 and 1.2 Hz/s. The input parameter is the estimated system frequency, where the effective rate of frequency change used by the relay is calculated based on the average value over five cycles using the equation:
17 Microgrid Protection
465
Fig. 17.24 Operation of the ROCOF anti-island protection scheme
1 fi df = dt 5 ti 5
(17.1)
i=1
The results can be seen in Fig. 17.24. The system is simulated for 60 s, and the switch that disconnects the MG from the main system is activated in the second 30 s. The procedure used by the programmed ROCOF scheme is as follows: 1. The signal from the voltage phasor is continuously processed. 2. The time frequency of the signal is estimated using a frequency estimation algorithm that performs a discretization of the waveform of the original signal and uses convolution of two orthogonal components, the sine, and cosine. 3. Once the frequency is estimated, it is processed again using the ROCOF algorithm, which calculates the frequency change for every five values of the signal using Eq. 17.1. If the disturbance exceeds a limiting value for the adjustment, the magnitude of the voltage is verified. There will be a considerable dejection to shoot by the island. If the frequency disturbance is due to a fault or the starting of a generator, the trip signal will be blocked. Voltage vector phase shift method. The vector shift relay is a passive local island detection technique that can detect an island operating condition by monitoring the voltage signal at the MG terminals. The principle of operation is based on the phase shift in the voltage signal monitored by the relay in relation to a reference
466
A. C. Enríquez et al.
Fig. 17.25 Voltage drop arising from the current demanded by the local load
Fig. 17.26 Internal and terminal voltage before and after opening the breaker
signal. This phase shift can constantly give slight increases or decreases due to the variable behavior of the loads since it is directly proportional to the voltage drop caused by the current demanded by the local load flowing from the MG. However, the island condition causes a change in the phase. Figure 17.25 shows a synchronous generator operating in parallel with a distribution network. It can be seen that a voltage drop ΔV arises between the voltage at the terminal VT and the internal voltage of the generator EI , due to the current demanded by the local ISG load passing through the reactor reactance Xd . This voltage drop causes a displacement in the phase of the voltage difference between the terminal and reference voltages. This can be seen in the phasor diagram shown in Fig. 17.26a. If the CB switch is opened due to a disturbance, the system consisting of the generator and the load will be isolated; at that moment, the generator will feed a greater or lesser load, since the current flow ISYS to or from the network is abruptly interrupted. Consequently, the angular difference between VT and EI is suddenly increased or decreased, and the phasor voltage at the terminals changes direction, as shown in Fig. 17.26b. When this phenomenon is analyzed in the time domain, the instantaneous value of the voltage at the terminals jumps to another value, and the position of the phase changes, as seen in Fig. 17.26. Point A indicates the instant of network loss, and it
17 Microgrid Protection
467
can be seen that the cycle time also changes. The voltage vector offset relay is based on this phenomenon. The relays, which are designed to provide anti-island protection based on this phenomenon, monitor the voltage waveform at the generator terminals and make incremental measurements of the duration of two complete cycles. The duration of both cycles will have to be equal or less on the island condition. This variation in the cycle time results in a proportional variation in the voltage angle at terminals Δθ, and this final value forms the input parameter to the relay. A predetermined adjustment threshold value in the range 2◦ to 20◦ is used as a reference signal, and if the measurement at each cycle exceeds this value, a trip signal is sent directly to the MG switch. In the same way, the voltage vector shift trigger algorithm is programmed using the simulated system signals. This algorithm works with the same voltage input signal. The results of this scheme show the correct operation in Fig. 17.27. The procedure used by the VSR scheme is as follows: 1. The signal from the voltage phasor is continuously processed. 2. The phase angle of the signal over time is estimated. To do this, the scheme measures the duration of each cycle of the signal. The time obtained for each cycle is compared with the duration of the previous cycle, and the difference is directly proportional to an offset angle. This offset angle forms the input to the relay that processes the operation algorithm based on the change in voltage vector. 3. The offset angle obtained in this way is compared with an adjustment value, and when a disturbance occurs, this value exceeds a certain limit. If the signal indicates island operation, the magnitude of the voltage at the terminals of the load will undergo a considerable change, which is the safety condition and the trip of the switches is sent. If the disturbance in the voltage phase is due to a fault or the starting of a generator, the trip signal will be blocked by the voltage verification.
17.5.3 Coordination of DOCRs One of the problems to consider in the implementation of an MG [35] is the design of the protection Scheme [36]. The scheme must be able to meet the basic protection requirements such as selectivity, sensitivity, and reliability, both while connected to the main network and in island operation mode [37]. The protection scheme must be adaptable [38], should take into account the operational modes of the MG, and should provide an online coordination system that also incorporates the necessary configurations of the protection relays depending on load variations, penetration of renewable generation [39], topology changes and WSs, resulting in better fault detection, reduction of false trips and shorter operating times.
468
A. C. Enríquez et al.
Fig. 17.27 Operation of the VSR anti-island protection scheme
A protection scheme based on DOCRs [40] is proposed. DOCRs are used to protect the lines during the operation connected to the main network and in island mode. The relay coordination problem is formulated as a non-linear programming problem [41]. From the characteristic inverse time equation, standardized by the IEEE, it is sought to provide protection with greater flexibility, the use of nonconventional inverse time curves is presented [42], considering each adjustment of the equation as a degree of freedom. The optimal configuration of the relays is determined using the GWO [43] and DE [44] algorithms. DOCRs coordinate with each other to ensure the selectivity and reliability of the protection scheme.
17.5.3.1
Optimization Problem
A. Objective function. The objectives of the coordination problem are to minimize the operation time of each relay and ensure compliance with the coordination time interval between the primary relay and its backup.
17 Microgrid Protection
469
⎫ ⎛ ⎞ ⎞ NCP NCP ⎪ ⎪ # # ⎪ ⎜ ⎜ tb ⎟ tp ⎟ ⎪ ⎛ ⎞⎪ ⎜ ⎪ ⎟ ⎟ ⎜ ⎪ NCP ⎜b=1⎟ ⎬ ⎜p=1⎟ # NV ⎜ ⎟ ⎟ ⎜ ⎝ ⎠ OF = min + ω1 ⎜ CT I pb ⎟ + ω2 ⎜ ⎟ + ω3 ⎜ NCP ⎟ ⎪ ⎜ NCP ⎟ ⎪ N CP ⎪ ⎪ ⎜ ⎟ ⎟ ⎪ ⎜ ⎪ pb = 1 ⎪ ⎪ ⎪ ⎪ ⎝ ⎠ ⎠ ⎝ ⎪ ⎪ ⎪ ⎪ ⎭ ⎩ ⎧ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎨
⎛
(17.2) where NCP is the number of coordination pairs present in the system, NV is the number of coordination pairs violated, tp is the operating time of the primary, tb is the operating time of the backup relay, and CTIpb is the coordination time interval (tb − tp ≥ CTI). The variables ω1, ω2, and ω3 are the Pareto front weights, which are obtained from variations in Ipickup and Isc . B. Time function. The operating time for each relay is defined with nonconventional curves (NCCs), taking as a reference to the curves of the IEEE standard [45]: T p, b (s) =
A
p I sc P SM∗I load
−1
+ B ∗ T DS
(17.3)
where TDS is the time dial setting, Isc is the coordination current, PSM is the plug setting multiplier, Iload is the load current, and A, B, p are characteristic curve parameters. The time curve model considers each coefficient of the vector as a degree of freedom to obtain the generalized formulation of the overcurrent relay, and the design obtained in this way is different for each relay. The resulting adjustment, therefore, increases the possibility of complying with the restrictions of the relay. This implementation allows us to obtain greater flexibility for the generation of the time curves. C. Coordination constraint. The use of conventional curves (CCs) is initially used to find the solution, and standardized IEEE (VI) curve parameters are used. The inequality restrictions for this case are given by: P SM min ≤ P SM ≤ P SM max
(17.4)
T DS min ≤ T DS ≤ T DS max
(17.5)
470
A. C. Enríquez et al.
Table 17.2 Time curve parameter intervals
Parameter TDS PMS A B p
Minimum range 0.5 1.2 0.05 0.11 0.02
Maximum range 3 1.6 28.2 0.49 2.00
NCCs are also used; in this case, the parameters A, B, and p are variable and begin to be part of the inequality restrictions, defining the minimum and maximum ranges that will be assigned to the relay. Amin ≤ A ≤ Amax
(17.6)
Bmin ≤ B ≤ Bmax
(17.7)
pmin ≤ p ≤ pmax
(17.8)
These variables determine the shape of the time curve. Based on its variability (from moderately inverse to extremely inverse), it is possible to obtain a wide diversity of time curves with the allowed parametric variation. Table 17.2 shows the intervals considered for each of the parameters that make up the time function of the relay. To achieve good coordination, the primary relay must operate after a fault occurs in its protected zone, but a sufficient time delay is necessary in order to ensure backup for the adjacent line in the forward direction when the primary relay does not operate. The coordination must be evaluated for a minimal fault to line-end because crossing curves can be present. Hence, coordination must be achieved in both fault locations. D. Sensitivity analysis. A sensitivity filter detects coordination pairs that cannot be coordinated. The backup relay must be able to detect minimal faults at the remote end protected by the primary relay. The sensitivity is determined as follows:
Sb =
I sc2Ø ¿1.5 P SM ∗ Iload
(17.9)
where Isc2φ is the minimum fault detected by the backup relay, and PSM * Iload is the pickup current of the backup relay. A lower limit of 1.5 is considered as among the current multiples of 1 to 1.5, the times resulting from the operation of the relays are large, and are thus useless for protection purposes.
17 Microgrid Protection
471
Fig. 17.28 Classification of wolves based on their distance from the prey
17.5.3.2
Optimization Algorithms
Gray wolf optimizer. The GWO is a bio-inspired algorithm that imitates the hunting mechanism of gray wolves. It is used in various engineering problems [43]. The algorithm achieves optimization in three stages: (1) hunting; (2) surrounding; and (3) attacking the prey. (1) Hunting process. The hunting process is led by the alpha wolf, while the other wolves (e.g., beta and delta wolves) help by recognizing the prey and surrounding it (Fig. 17.28). In an optimization problem, a knowledge of the location of the dam is established like the best solution of the wolves. Other wolves (search agents) will carry out an exploration in the solution plane and are able to obtain better solutions. The best positions are updated to give the best solution. Each wolf contains the possible solution for the relay coordination. (2) Surrounding the prey. The final positions of the wolves X α, β, δ define − → − − → → the estimated area of the prey. Wolves X α, X β and X δ then surround the prey. Surrounding is modeled mathematically as follows: − − → − → − → → D α,β,δ = C 1,2,3 ∗ X α,β,δ − X
(17.10)
− → − → − → − → X 1,2,3 (t) = X α,β,δ − A 1,2,3 ∗ D α,β,δ
(17.11)
− → − → where X α,β,δ are the positions of the gray wolves, X is the prey, t is the current − → − → iteration, and the coefficient vectors A and C are used to avoid falling into local minima and to define the surrounding of the prey. − → → − − → − → → → A = 2− a ∗ r1 − − a , C = 2 ∗ r2
(17.12)
The hunting strategy is implemented using variables r1 and r2 (Fig. 17.29). (3) Attacking the prey. If | A > 1 | and C > 1, GWO executes a search, while if | A < 1 | and C < 1, GWO attacks the prey. Gray wolves end a hunt by attacking the prey when it stops moving (Fig. 17.30).
472
A. C. Enríquez et al.
Fig. 17.29 Updating of positions in GWO
Fig. 17.30 Attacking prey
A specific gene mutation (SGM) can be used to improve the exploration in the search space. Better search solutions and a reduction in the algorithm execution time are obtained (Fig. 17.31). (4) Updating of the wolves’ positions. The best candidate solution is obtained based on the distance from the best solutions: −→ −→ −→ X1 (t) + X2 (t) + X3 (t) − → X (t + 1) = 3
(17.13)
17 Microgrid Protection
473
Fig. 17.31 Flow chart for gray wolf optimization
The stop criterion is the maximum defined iteration and indicates that the wolves have converged and attacked the prey. Differential evolution algorithm. The DE algorithm is a type of evolutionary algorithm and is a search-based algorithm based on natural evolution and gene selection. The DE uses genetic operators in a similar way to other evolutionary algorithms (Fig. 17.32); however, unlike traditional evolutionary algorithms, it randomly disturbs the members of the current generation population. There is therefore an additional distribution of probability in terms of generating offspring. This feature means that the algorithm involves fewer mathematical operations and consequently has a shorter computational runtime than comparable algorithms. (1) Initial population. An initial population is created in which the genes of all individuals are initialized to some number in a range that is feasible with respect to
474
A. C. Enríquez et al.
Fig. 17.32 Main stages of the DE algorithm
Fig. 17.33 Mutation in DE
the relay settings. Each row represents an individual, and each column represents a gene/variable/relay setting. The population size is (NP, D*NR) where NP represents the number of individuals, D the number of control variables, and NR the number of relays. For example, in a system with 12 relays and two degrees of freedom (dial and k), the population size for 60 individuals will be (60,24). The initial population is represented as follows: ⎡
dial (1,1) · · · ⎢ .. .. P =⎣ . . dial (N P ,1) · · ·
dial (1,N R) .. . dial (N P ,N R)
k(1,N R+1) · · · .. .. . . k(N P ,N R+1) · · ·
k(1,N R∗2) .. .
⎤ ⎥ ⎦
k(N P ,N R∗2) (17.14)
(2) Mutation. The mutation operator is based on the arithmetic difference between pairs of randomly selected vectors (Fig. 17.33). A mutant vector is calculated as the scaled difference between two vectors (r1 , r2 ) of three randomly selected vectors, added to the value of the third vector (r0), which is called the base vector. F is a user-defined parameter.
17 Microgrid Protection
475
Fig. 17.34 Crossing in DE
− → − → V i,g = X r
0,g
+F
− → Xr
1,g
− → − Xr
2,g
(17.15)
r0 = r1 = r2 = I
(17.16)
02
(17.17)
(3) Crossing. To enrich the diversity of the population, the crossover operation is applied after generating the mutant vector by the middle of the mutation. The − → mutant vector exchanges its components with the parent vector X i,g to form a child vector. Crossing is carried out for each variable D if a randomly generated number in the range [0, 1] is less than or equal to the Cr value. In this case, the number of parameters inherited from the mutant vector has an almost binomial distribution. This scheme can be expressed as follows: Uj,i,G =
Vj,i,G Xj,i,G
if randi,j [0, 1] ≤ Cr or j = jrand otherwise
(17.18)
where rand i, j (0,1) is a uniformly distributed random number calculated for the j-th component of the i-th individual. j rand ∈rand [1, D] is a random position of − → − → the vector, which ensures that U i,g obtains at least one component of V i,g . This operation is performed once for each individual in each generation (Fig. 17.34).
476
A. C. Enríquez et al.
(4) Selection. To keep the population size constant in subsequent generations, the next step of the algorithm carries out a selection process to determine if the parent vector or the child vector survives to the next generation, referred to as G = G + 1. The selection operation can be represented as: ⎧− ⎨→ U i,G − → X i,G+1 = − → ⎩ X i,G
− − → → U i,G ≤ f X i,G − − → → if f U i,G > f X i,G if f
(17.19)
− → where f ( X is the ability to minimize the function. Hence, if the new child vector has an aptitude function value that is equal to or less than the parent vector, it replaces the parent vector in the next generation; otherwise, the parent vector persists in the population. The population may therefore improve (with respect to the minimization of the fitness function) or retain the same fitness status, but does not get worse. A flow chart for this process is shown in Fig. 17.35.
17.5.3.3
On-Line Implementation
The optimization algorithms selected for online execution are sufficiently robust to achieve coordination. The adjustment update scheme must be implemented in centralized mode, with continuous real-time measurement processing and monitoring of the system topology. Updating settings almost in real time before any topological change must take place, and updating based on variation in demand should be carried out at intervals of no more than 15 min, a similar length to the intervals used to calculate the demand. The latency of the communication available in the power grid must be considered. Ipickup is updated in a stable state within the time interval considered, and the Isc is updated in the case of a topological change or connection of a generation source. The use of phasor measurement is recommended, for example, μPMU and the amount of calculation can be greatly reduced.
17.5.3.4
Test Systems
The proposed test system shown below allows us to study and simulate the behavior of the protection system in an MG, and the performance of the algorithm is analyzed in terms of the coordination between relays. The system consists of 13 phase relays with a total of 12 coordination pairs (Fig. 17.36). The number of relays is obtained after excluding the relay protecting the transformers. The relay names consist of two digits, the first of which represents the nearby bus and the second represents the remote bus. The voltages are adjusted to 34.5 kV.
17 Microgrid Protection
477
Fig. 17.35 Optimized flow chart for differential evolution
The termination criterion is 1000 iterations and the system is simulated with 60 individuals. The maximum faults at the near end and the minimum faults at the far end were obtained by opening the remote end.
478
A. C. Enríquez et al.
Fig. 17.36 Test system (microgrid)
17.5.3.5
Results and Discussion
The GWO with 60 individuals and DE with 100 individuals were used to solve the coordination problem. Twenty algorithm repetitions of 1000 iterations each were carried out to obtain the results. The results were evaluated using the standard deviation (Std) to indicate the similarity of the results at each execution and to analyze the execution time as an indication of the viability of the coordination for online applications. Test case I: An MG connected to the main network and operating in island mode was simulated. The maximum generation conditions were considered and a conventional IEEE very inverse (VI) curve is used. Based on the settings (dial, fk) established in the algorithm, the primary time (tp) and the backup time (tb) could be obtained. We also aimed to establish correct coordination between the times of each relay, accomplish the time interval (CTI). The results are shown in Figs. 17.37 and 17.38. The contribution of Isc of the utility to the MG (Fig. 17.37) means that it is not possible that the system is properly coordinated with conventional curves. There are
17 Microgrid Protection
479
Fig. 17.37 Settings and parameters obtained (microgrid connected to the utility)
Fig. 17.38 Adjustments and parameters obtained (operating in island mode)
at least two instances of violation or loss of coordination. From Fig. 17.38, with the MG operating in island mode, it can be seen that the value of Isc decreases considerably, and therefore the loss of coordination will be more frequent. In this case, four coordination pairs are affected. Correct operation and loss of coordination are shown in Figs. 17.39 and 17.40, respectively. Test case II: Nonconventional time curves are used with the aim of ensuring coordination between all devices in the network. In these simulations, we consider both an MG connected to the main network and operating in island mode. The results show no loss of coordination (Figs. 17.41 and 17.42). With the use of nonconventional curves (NCCs), correct coordination between all the elements of the MG is guaranteed. Coordination between relays R12-R31 and R31-R53 in both directions is achieved. The coordination is different for each
480
A. C. Enríquez et al.
Fig. 17.39 Pair of correctly coordinated relays
mode of operation of MG, that is, connected (Fig. 17.43) and island modes (Fig. 17.44). Figure 17.45 shows the performance of the algorithm using both conventional curves (CCs) and NCCs. The advantages of NCCs can clearly be seen. Test case III: Comparison between GWO and DE, using nonconventional time curves. Simulations were carried out with the two algorithms in parallel, in order to check their robustness in terms of solving the protection coordination problem for both mesh and MG networks. Figure 17.46 shows the fitness of the GWO and DE algorithms, and the average values for each algorithm are presented in Table 17.3.
17 Microgrid Protection
481
Fig. 17.40 Pair with loss of coordination
17.6 Conclusions The dynamic conditions of the MG mean that new strategies are required to enable the electricity network to remain reliable and to deliver energy to users. In this work, the use of interconnection relays at the base of MAS is proposed to support the reconfiguration of the relays, making them more sensitive to the changes in current that is observed in the energy inputs from sources.
482
Fig. 17.41 Microgrid connected to the utility
Fig. 17.42 Microgrid operating in island mode
A. C. Enríquez et al.
17 Microgrid Protection
483
Fig. 17.43 Coordinated pair (connected to utility)
The proposed MAS is feasible for use in detecting the power supply conditions of a WS through the online settings of the relays, considering topological changes and operations due to faults. The relay settings were actualized through the continuous detection of the switch positions and measurement of the demand current. WS detection was achieved using a different strategy depending on the type of source, for small conventional generators and DG. The performance of the protective relays for the detection of a WS condition was improved, and an increase in sensitivity and better coordination were obtained. Non-standardized inverse time curves are used for MG protection, allowing for a reduction in the operating times of each relay and providing the necessary flexibility that the time is within preset ranges for the particular problem to be analyzed or represented in the electricity grid. The results obtained from the test systems were acceptable, as they were within predefined intervals with zero violations.
Fig. 17.44 Coordinated pair (island mode)
Fig. 17.45 Performance of the type of time curve
17 Microgrid Protection
485
Fig. 17.46 Performance of algorithms: (a) GWO and (b) DE Table 17.3 Average values of the results for each algorithm Average GWO Tp Tb 0.552 0.866
CTI 0.313
Fit 17.302
NV 0
Average DE Tp Tb 0.493 0.841
CTI 0.347
Fit 13.489
NV 0
References 1. Ferrari, J. P. (2005). Sistemas de Control Distribuido. Universidad Nacional de Rosario. 2. K. H. Kato T., S. Y. and T. Funabashi, “Multi-agent based control and protection of power distributed system-protection scheme with simplified information utilization,” IEEE Proceedings of the 13th International Conference on Intelligent Systems Application to Power Systems, pp. 49-54. November 2005. 3. Lanemen, J., Tse, D., & Wornell, G. (2004). Cooperative diversity in wireless networks: Efficient protocols and outage behavior. IEEE Transactions on Information Theory, 50(2), 3062–3080. 4. Sendonaris, A., Aazhang, B., & Erkip, E. (2003). User cooperation diversity, part I: System description. IEEE Transactions on Communications, 51(11), 1927–1938. 5. Lotfi-fard, S., Faiz, J., & Iravani, R. (April 2007). Improved overcurrent protection using symmetrical components. IEEE Transactions on Power Delivery, 22(2), 843–850. 6. McArthur, S. D. J., Davidson, E. M., Catterson, V. M., Dimeas, A. L., Hatziargyriou, N. D., Ponci, F., & Funabashi, T. (2007). Multi-agent systems for power engineering applications-part ii: Technologies, standards, and tools for building multi-agent systems. IEEE Transactions on Power Systems, 22(4), 1753–1759. 7. Wooldridge, M. (2009). An introduction to multiagent systems. John Wiley & Sons. 8. Bellifemine, F., Caire, G., & Greenwood, D. (2007). Developing Multi-Agent Systems with JADE. Hoboken, NJ: Wiley series in agent Technology. 9. Patel, R., Bhatti, T. S., & Kothari, D. P. (2002). MATLAB/Simulink-based transient stability analysis of a multimachine power system. International Journal of Electrical Engineering Education, 39(4), 320–336. 10. Robinson, C. R., Mendham, P., & Clarke, T. (2010). MACSimJX: A tool for enabling agent modelling with simulink using JADE. Journal of Physical Agents, 4(3).
486
A. C. Enríquez et al.
11. Silva, A. C. R. D., & Grout, I. A. (2011). MS2SV: Tradução de modelos em Simulink para VHDL-AMS. IEEE Latin America Transactions. 12. MathWorks, Instrument Control Toolbox 2.7, Mathworks Inc., Natick, USA, 2008. 13. MathWorks, Simulink Simulation and Model-Based Design, MathWorks, Inc., Natick, MA, 2004. 14. J. Barsch, G. Bartok, G. BenmouyaI, O. Bolado, B. Boysen, S. Brahma and J. Burnworth, “Fault current contributions from wind plants,” Report to the T&D Committee, IEEE PES Electric Machinery Committee and Power System Relaying Committee, 2012. 15. J. Keller, B. Kroposki, R. Bravo and S. Robles, “Fault current contribution from single-phase PV inverters,” IEEE 37th Photovoltaic Specialists Conference (PVSC), pp. 1822-1826, June 2011. 16. S. Chowdhury, C. F. Ten and P. Crossley, “Islanding protection of distribution systems with distributed generators: A comprehensive survey report,” IEEE Power and Energy Society General Meeting, pp. 1–8, IEEE, 2008. 17. Jang, S. I., & Kim, K. H. (2004). An islanding detection method for distributed generations using voltage unbalance and total harmonic distortion of current. IEEE Transactions on Power Delivery, 19, 745–752. 18. IEEE 929-2000 - IEEE Recommended Practice for Utility Interface of Photovoltaic (PV) Systems. 19. Pai, F.-S., & Huang, S.-J. (2001). A detection algorithm for islanding-prevention of dispersed consumer-owned storage and generating units. IEEE Transactions on Energy Conversion, 16(4), 346–351. 20. Salman, S. (2001). New loss of mains detection algorithm for embedded generation using rate of change of voltage and changes in power factors. In 7th international conference on developments in power systems protection (DPSP2001) (pp. 82–85). 21. Affonso, C., Freitas, W., Xu, W., & daSilva, L. (2005). Performance of ROCOF relays for embedded generation applications. IEE Proceedings—Generation, Transmission and Distribution, 152(1), 109. 22. Vieira, J., Freitas, W., Xu, W., & Morelato, A. (2006). Efficient coordination of ROCOF and frequency relays for distributed generation protection by using the application region. IEEE Transactions on Power Delivery, 21, 1878–1884. 23. Lidula, N. W. A., & Rajapakse, A. D. (2012). A pattern recognition approach for detecting power islands using transient signals part II: Performance evaluation. IEEE Transactions on Power Delivery, 27, 1071–1080. 24. Fusco, G., DiFazio, A. R., & Russo, M. (2015). Islanding detection method based on a Thevenin-like model. IET Generation, Transmission & Distribution, 9, 1747–1754. 25. Merino, J., Mendoza-Araya, P., Venkataramanan, G., & Baysal, M. (2015). Islanding detection in microgrids using harmonic signatures. IEEE Transactions on Power Delivery, 30, 2102– 2109. 26. Gupta, P., Bhatia, R. S., & Jain, D. K. (2017). Active ROCOF relay for islanding detection. IEEE Transactions on Power Delivery, 32, 420–429. 27. Gupta, N., & Garg, R. (2018). Algorithm for islanding detection in photovoltaic generator network connected to low-voltage grid. IET Generation, Transmission & Distribution, 12, 2280–2287. 28. Holdsworth, L., Wu, X., Ekanayake, J., & Jenkins, N. (2003). Comparison of fixed speed and doubly fed induction wind turbines during power system disturbances. IEE Proceedings: Generation, Transmission & Distribution, 150(3), 343. 29. L. W. Arachchige and A. Rajapakse, “A pattern recognition approach for detecting power islands using transient signals—Part I: Design and implementation,” 2011 IEEE power and energy society general meeting, 2011. 30. R. C. Dugan, J. A. Taylor and D. Montenegro, “Energy storage modeling for distribution planning,” 2016 IEEE Rural Electric Power Conference (REPC), 53, pp.12–20, IEEE, 2016. 31. W. Xu, G. Zhang, C. Li, W. Wang, G. Wang and J. Kliber, “A power line signaling based technique for anti-islanding protection of distributed generators: Part I: Scheme and analysis,” 2007 IEEE power engineering society general meeting, Jun 2007.
17 Microgrid Protection
487
32. Yan, N., Xing, Z. X., Li, W., & Zhang, B. (2016). Economic dispatch application of power system. IEEE Transactions on Applied Superconductivity, 26(7), 1–5. 33. Smith, G., Onions, P., & Infield, D. (2000). Predicting islanding operation of grid connected PV inverters. IEE Proceedings—Electric Power Applications, 147(1), 1. 34. Ropp, M., Begovic, M., & Rohatgi, A. (1999). Analysis and performance assessment of the active frequency drift method of islanding prevention. IEEE Transactions on Energy Conversion, 14(3), 810–816. 35. J. A. Ocampo-Wilches, A. J. Ustariz-Farfan and E. A. Cano-Plata, “Modeling of a centralized microgrid protection scheme,” 2017 IEEE Workshop on Power Electronics and Power Quality Applications (PEPQA), pp. 1-6, May 2017. 36. Ali Memon, A., & Kauhaniemi, K. (2015). A critical review of AC microgrid protection issues and available solutions. Electric Power Systems Research, 25, 23–31. 37. A. Sharma and B. Panigrahi, “Optimal relay coordination suitable for grid-connected and islanded operational modes of microgrid,” 2015 Annual IEEE India Conference (INDICON), 2015. 38. Awaad, M., Mekhamer, S., & Abdelaziz, Y. (2018). Design of an adaptive overcurrent protection scheme for microgrids. International Journal of Engineering, Science and Technology, 10(1), 1–12. 39. Jain, D. K., Gupta, P., & Singho, M. (November 2015). Overcurrent protection of distribution network with distributed generation. In 2015 IEEE Innovative Smart Grid Technologies-Asia (ISGT ASIA) (pp. 1–6). 40. Sahoo, A. K. (September 2014). Protection of microgrid through coordinated directional over-current relays. In 2014 IEEE Global Humanitarian Technology Conference—South Asia Satellite (GHTC-SAS) (pp. 129–134). 41. H. Zeineldin, E. El-saadany and M. Salama, “Protective relay coordination for microgrid operation using particle swarm optimization,” 2006 Large Engineering Systems Conference on Power Engineering, pp. 152-157, July 2006. 42. Arreola, O., Conde, A., & Trujillo, L. A. (2014). Overcurrent relay with unconventional curves and its application in industrial power systems. Electric Power Systems Research, 110, 113– 121. 43. Mirjalili, S., Mirjalili, S. M., & Lewis, A. (2014). Grey wolf optimizer. Advances in Engineering Software, 69, 46–61. 44. Storn, R., & Price, K. (1997). Differential evolution—A simple and efficient heuristic for global optimization over continuous spaces. Journal of Global Optimization, 11(4), 341–359. 45. IEEE C37.112: IEEE Standard Inverse-Time Characteristic Equations for Overcurrent Relays, 2018.
Chapter 18
A New Second Central Moment-Based Algorithm for Differential Protection in Micro-Grids Ernesto Vázquez, Héctor Esponda, and Manuel A. Andrade
18.1 Introduction Although the literature presents different definitions for Smart Grid, the characteristics remain the same: it can skillfully integrate the actions of all stakeholders connected to it—generators, consumers, and those that do both—in order to deliver sustainable, economic, and reliable electricity supplies efficiently. A reliable and secure Smart Grid requires fast and accurate fault localization and intelligent automation functions to take care of restoring the power to healthy parts of the network. Smart grids typically consist of various amounts of local generation, as well as energy storage systems and dynamic loads (see Fig. 18.1). The increased penetration of DG challenges conventional protection schemes because fault currents now come from both traditional sources and inverted-driven sources. However, many DG technologies are based on DC/AC converters, making conventional protection schemes infeasible. Moreover, storage systems operating in discharge mode are, from the grid point of view, similar to DG [11]. The paradigm of protection for Smart Grid should be adapted to the flexibility inherent to the Smart Grid concept. An essential requirement is “adaptability”: a protection system needs to adapt to the changes in the primary system. The protection system for Smart Grids should consider the amount and type of DG (and for some units, its capacity of short-circuit current) and the amount of power from the primary system. If part of the Smart Grid DG is capable of supply the load, it can be disconnected in case of a fault occurs. However, intermittency may force the DG to be disconnected to avoid damage. Thus, protection schemes
E. Vázquez · H. Esponda · M. A. Andrade () Universidad Autónoma de Nuevo León, School of Mechanical and Electrical Engineering, Monterrey, Nuevo León, Mexico © Springer Nature Switzerland AG 2021 A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems, https://doi.org/10.1007/978-3-030-59750-4_18
489
490
E. Vázquez et al.
Residential load
Industrial load Grid controller
Hydro generation
Diesel generation
Solar and wind generation
Fig. 18.1 Smart Grid with distributed generation components feeding power to local loads
are needed to ensure the faults that are external to the island, and in some extents to the faulted equipment, do not cause unnecessary disconnection of the DGs and transformers, so the Smart Grid has possibilities of remain energized. Fault protection methods must incorporate more than one approach to coping with the issues arising in Smart Grid protection schemes. Adaptive protection schemes change its response according to the system dynamics. The most straightforward approach is to have two sets of relay settings (for islanded and grid-connected operating modes). Although this is a simple approach, the protection system must cope with changes in fault current levels by shifting through characteristic curves [17]. Adaptability also requires having a piece of prior knowledge on the configuration of the Smart Grid to update the settings for each possible scenario. Another approach to Smart Grid protection is the differential scheme, mostly in scenarios with low levels of fault current. Traditional differential protection cannot be used as a complete solution because it is better suited to detect downstream earth faults and requires additional techniques to detect symmetrical faults and faults without ground component [4]. The classic overcurrent protection is not well suited to cope with DG varying fault levels and intermittency. Traditional overcurrent schemes have problems identifying not-fault conditions as inrush currents and may require directional units. Adaptive protection systems have evolved to wide-area schemes where supervisory systems (SCADA) and IED are integrated to measure over broad areas and
18 A New Second Central Moment-Based Algorithm. . .
491
perform protective actions based on a high volume of data [20] over intensive communication systems. Sometimes it is necessary to use distributed intelligence in the form of multi-agent systems based on fuzzy networks, decision trees, support vector machines, machine learning, etc. [3, 7, 11, 12, 16]. Having such independent agents, with awareness about their component or substation states through sensor links, permits the agents to perform various functions that are not implemented by either the protection systems or the central control systems. To add smartness to a power system, independent protective functions in each component are a must [13], and these protective elements need to adapt itself to the dynamics of the power system. The Smart Grid adaptive protection also requires to be smart. Further sections will discuss some approaches to implement flexible, adaptive protection systems for Smart Grid components. In this chapter, the Smart Grid requirements for the protection system are reviewed from the point of view of differential protection used in different elements of electricity distribution systems such as busbar, transformer, generator, reactor, and the impact of the high harmonic content due to the non-linear loads. The present method is based on the analytical second central moment (SCM) to characterize differential current’s patterns to detect faults occurring within the differential zone. The method calculates the SCM magnitude on the base of waveforms of a half and full sinusoidal signal to define the similitude between fault current and any other current due to transient phenomena. The method applies a filtering and normalization stage to eliminate redundant information and to highlight any change in the differential current signals. On the other hand, a universal threshold was proposed and set up to identify the faults currents. This limit is a universal unique threshold for any element in a smart grid (busbar, transformer, generator, reactor, capacitor bank, motors, and transmission lines), and it is based on the maximum magnitude of the SCM that it can achieve in a half-sinusoidal waveform. The method offers a straightforward implementation with a low computational cost; detecting faults faster than traditional differential protections. In the case of transformers, the method carries on turn-to-turn short-circuits detection even during energization cases.
18.2 Differential Protection Based on SCM Differential protection schemes have been used for 20 years to protect generators, transformers, motors, reactors, buses, lines, and any other equipment where inputs and outputs current are compared. This scheme is selective, and it only responds to faults within protected elements. Nevertheless, the leading difficulty this scheme suffers is to distinguish between inrush current and fault currents correctly. Inrush currents are transient phenomena arising when the transformer is being energized. Although inrush currents have large magnitudes, they do not reflect in the secondary of the transformer. This phenomenon may contribute to the misoperation
492
E. Vázquez et al.
of differential elements because of the false differential current generated. Thus, the reliability of the scheme could be compromised. Diverse methods—most of them based on the harmonic content and the waveform of the signal—have been presented in the literature to overcome this problem. Nonetheless, some conditions as current transformers (CT) saturation, changes in the transformer’s parameters, overexcitation, and frequency variations may impair the performance of these methods.
18.2.1 Differential Protection Principle Differential protection is based on Kirchhoff’s current law [21]. The principle of operation calculates the sum of all currents flowing from the input to the output terminals of the protected zone. The CTs delimit the protected area. Under ideal conditions, steady-state or faults outside of the protected zone, the differential current will be zero as is shown in Fig. 18.2a. On the other hand, if a fault occurs inside of the protection zone as is shown in Fig. 18.2b, the differential current will be different from zero, and the relay will send a trip signal to disconnect the power transformer.
18.2.1.1
Percentage Differential Protection
Although differential current principle, in theory, works well, the error introduced by the mismatch of CT ratios, tap changers, and other factors such as the power transformer phase-shift, may cause a false differential current that must be considered to define the settings of the differential protection scheme. Therefore, percentage differential protection was introduced to decrease the magnitude of differential current in steady-state and gives a high sensibility when CTs are saturated. This scheme adds a restraint current Irest defined as Irest = k (|IW1 | + |IW2 |) ,
(18.1)
Irest = max (|IW1 |, |IW2 |) ,
(18.2)
or
where IW1 and IW2 are the currents flowing through the CT secondary at the high and low voltage windings, respectively. The restraint current is compared to the differential current Idiff defined as Idiff = |IW1 + IW2 | .
(18.3)
18 A New Second Central Moment-Based Algorithm. . . Fig. 18.2 Differential protection principle. (a) During an external fault. (b) During an internal fault
493
Fault 1
2
Protected element 1
2
Relay
Δ =
1
+
2
=0
a Fault
1
2
Protected element 1
2
Relay
Δ =
1
+
2
b The relay will operate if the differential current is greater than a percentage of the restraint current defined by a slope setting (SLP) according to Idiff > SLP × Irest .
(18.4)
Graphically, the operation characteristic of the percentage differential protection is shown in Fig. 18.3. This characteristic consists of a straight line with a slope equal to SLP1 . The intersection point of this characteristic with the vertical axis Iop defines the minimum pickup current of the relay Ipu min . Also, a dual-slope characteristic (SLP2 ) can be added to increase the relay safety in conditions where the CTs experience a heavy saturation as is shown in Fig. 18.3. The SLP values are common expressed in percentage and is defined as
494
E. Vázquez et al.
Fig. 18.3 Relay characteristic with double slope
2
1
op 0
ret 0
SLP =
Idiff . Irest
ret 1
ret 2
(18.5)
Some typical values for SLP are 30% and 80% [21]. Figure 18.4 shows an example of a differential protection performance in a power transformer during steady-state and a fault inside of the differential protection zone, respectively. It was used a dual-slope characteristic with the SLP1 and SLP2 set in 0.3 and 0.6, respectively. Figure 18.4a shows the trajectory of the operation current which is on the non-operation zone and it means the transformer is working correctly. On the other hand, Fig. 18.4b illustrates the current operation crossed the SLP characteristic from the non-operation to the operation zone. This behavior implies that the transformer must be disconnected to avoid damage to the equipment.
18.2.1.2
Negative-Sequence Differential Principle
Recently, the negative-sequence differential protection (87Q) [10, 19] has been introduced to avoid unnecessary relay tripping during external faults. This method has an excellent performance against external faults and turn-to-turn faults where high sensitivity is required. Using the power system as is shown in Fig. 18.5, it is possible to explain the principle of the negative-sequence protection. This system comprises a transformer without phase shifting and equal ratio on both sides. If a fault occurs outside the differential protection zone, a fictitious negative-sequence source can be placed at the fault point as shown in Fig. 18.5a. Therefore, the negative-sequence currents in each side of the transformer will flow in the same direction having a phase-shift of 0 degrees. On the other hand, if a fault occurs inside the differential protection zone as is shown in Fig. 18.5b, the negative-sequence currents will flow in opposite directions with a phase-shift of 180 degrees. These
18 A New Second Central Moment-Based Algorithm. . . Fig. 18.4 Differential protection performance. (a) Steady-state condition. (b) Fault inside the differential protection zone
495
1
op
(pu))
1.5
0.5
0
0
1
2 ret (pu)
3
4
2 ret (pu)
3
4
a
1
op
(pu))
1.5
0.5
0
0
1 b
characteristics have been useful to implement a method to discriminate between an internal and an external fault using the magnitude and the phase shifting of the negative-sequence currents. The negative-sequence method calculates the negative-sequence currents from the primary and secondary side of the element protected. After that, both negativesequence currents are compared with a threshold defined as minimum pickup Ipu min Q . If either or both are greater than this threshold, the method will compare the phases of both negative-sequence currents. The resulting phase will place in an α plane as shown in Fig. 18.6. If the phase falls between ±60 degrees (also called
496
E. Vázquez et al.
Fault 2S
2S
2R
2S
2R
2R
+ −
2
Relay
a Fault
2S
2S
2R
+ −
2S
2
2R
2R
Relay
b Fig. 18.5 Direction of negative-sequence currents. (a) External fault. (b) Internal fault
the relay operating angle (ROA) zone), the relay will send a trip signal to disconnect the element. Otherwise, the event will be determined as a steady-state condition or external fault. To show the performance of the differential negative-sequence scheme, an external three-phase fault, and a turn-to-turn fault at 10% of the winding shortcircuited were evaluated as are shown in Fig. 18.7. During the external fault, the magnitude of negative-sequence current did not enter to the ROA zone. On the contrary, at the time the internal fault occurred from a steady-state condition, the magnitude of the negative-sequence current crossed to the ROA zone. Traditional percentage differential protection and negative-sequence method usually have a good performance discriminating transient events and internal faults. However, these conventional methods may malfunction during the inrush current, remanent flux in the core of the power transformer, CT saturation, incipient faults, non-linear loads, etc. The next section will address the main issues the percentage differential protection faces.
18 A New Second Central Moment-Based Algorithm. . . Fig. 18.6 Negative-sequence alpha plane
497
90
120
60
150
30 No trip zone
ROA
180
0 mín
330
210 240
120
90
120
60
150
30
300
270 90
60
150
30 Fault
10 20 30 40
180
210
0
330 240
270
300
a
2
180
3
210
4
5
0
330 240
270
300
b
Fig. 18.7 Performance of the differential negative-sequence method. (a) External fault. (b) Internal fault
18.2.1.3
Current Transformer Saturation
Current transformers usually are exposed to be saturated when a fault occurs. The high levels of DC component when a fault happens, they do saturate the core of the CTs distorting the waveform of the CT secondary current. The saturation of the CTs introduces an error in the differential currents that may cause the misoperation of the differential elements. The CT saturation will produce an increase in the differential current once the fault already had occurred, even though, the fault was
a
Time
b
Current
Time
Current
E. Vázquez et al.
Current
498
Time
c
Fig. 18.8 Impact of CT saturation in different levels: primary current (solid lines) and secondary current (dashed lines). (a) Light saturation. (b) Normal saturation. (c) Heavy saturation
outside the differential protection zone. Also, the CT saturation introduces harmonic content in the differential signals that may cause a time delay in the operation of the differential elements. Figure 18.8 illustrates how the CT secondary currents are distorted in comparison with the CT primary currents when the core of the CTs is under saturation [6]. Also, it is seen the saturation level directly will impact on the CT secondary current distortion. Figure 18.9 shows an example of the performance of the differential elements during a fault inside the differential protection zone. It is seen the CT saturation caused time delay of the differential elements because of the harmonic content.
18.2.1.4
Methods Summary
All different methods before mentioned have shown different advantages and disadvantages depending on the power system conditions such as inrush current, CT saturation, incipient faults, degradation of the transformer parameters, etc. Also, the different kind of input signals used for the methods such as current, voltage, phasors, symmetrical components, etc., may impact on the security, speed, and reliability of the differential protection. Therefore, to bring a better understanding of what are the events and the operation-basis of the methods, Tables 18.1 and 18.2 are introduced.
18.3 Second Central Moment Applied to Differentiate Inrush Currents from Internal Faults The proposed method is based on the magnitude of the second central moment from the differential currents to differentiate internal faults from transient conditions as well as inrush currents. Figure 18.10 shows the flow chart of the proposed method which is divided into different stages. The method calculates the SCM magnitude of each differential current previously normalized and filtered to identify the kind of event. The SCM magnitude is compared to an identification threshold based on the maximum variance that a half-sinusoidal signal could achieve. If the
18 A New Second Central Moment-Based Algorithm. . .
499
CT saturation Current (A)
1 0.5 0 −0.5 0.4
0.45
0.5 Time (s) a
0.55
0.6
0.65
12 Misoperation
(pu)
8
op
10
6 4
Impact of CT saturation
2 0
0
5
10
20 15 (pu) ret
25
30
b Fig. 18.9 Misoperation of conventional differential protection in presence of heavy CT saturation. (a) CT secondary current. (b) Differential relay performance
SCM magnitude is higher than the limit, the method identifies a fault condition. Otherwise, if the SCM magnitude is lower or equal to the threshold, the method determines a transient event or steady-state condition. Table 18.1 presents which are the power conditions where each method correctly operates. All different scenarios were selected from the most representative cases that may lead to the misoperation of the differential elements. On the other hand, Table 18.2 exposes which kind of signal is used for the method such as voltage, current, or both of them. Also, Table 18.2 shows if the methods use empirical thresholds to detect internal faults and if they are either online or offline versions. As the presented method is intended to be implemented in a digital relay, it forms the differential currents from the CT secondary currents. Accordingly, the CT secondary currents are angle shift-compensated using a set of compensation matrices adapted depending on the transformer vector group [1, 5]. In this case, for the primary side, the matrix 0 was selected. For the secondary side, the matrix
method does not detect inrush currents b The method does not detect residual flux
a The
Method Buchholz relaya,b Harmonic restraint Harmonic cross-blocking Negative sequence Transformer parameters Two-terminal network Differential power PCA Differential current gradient Empirical Fourier transform Non-saturation zone Linkage-flux High-order statistics Wavelet transform Current/voltage ratio Superimposed differential Kalman filter Park’s vector Frequency responsea,b
System condition Non-linear CT saturation loads
Recovery inrush
Table 18.1 Events where each method shows a correct operation
Sympathetic inrush
Cross-country faults
Incipient faults Degradation Over excitation
500 E. Vázquez et al.
Phasors
Instantaneous values
not based on voltage/current signals b Offline method
a Method
Method Buchholz relaya Harmonic restraint Harmonic cross-blocking Negative sequence Transformer parameters Two-terminal network Differential power PCA Differential current gradient Empirical Fourier transform Non-saturation zone Linkage-fluxa High-order statistics Wavelet transform Current/voltage ratio Superimposed differential Kalman filter Park’s vector Frequency responseb
Operation base
Symmetrical components
Superimposed quantities
Table 18.2 Operation base and the kind of measurements each method uses to operate
Statistical distribution
Sensors
Empirical threshold
18 A New Second Central Moment-Based Algorithm. . . 501
502
E. Vázquez et al.
Fig. 18.10 Flow chart of the proposed method’s algorithm to identify transient events from internal faults
Start Data acquisition from CTs
Data window Differential current Normalization [−1, 1]
Delta filter Calculation of SCM
F
SCM > 0.25 T Trip
End
1 was chosen. A modification on the transformer vector group requires updated compensation matrices. The angle-compensated differential currents are ⎡
⎤ ⎡ ⎤⎡ ⎤ ⎡ ⎤⎡ ⎤ Idiff A 1 0 0 Ip A 1 −1 0 Is A ⎣ Idiff B ⎦ = ⎣ 0 1 0 ⎦ ⎣ Ip B ⎦ + aT √1 ⎣ 0 1 −1 ⎦ ⎣ Is B ⎦ , 3 0 0 1 −1 0 1 Idiff C Ip C Is C Idiff ABC = I∗p ABC + I∗s ABC ,
(18.6) (18.7)
where aT is a phase compensation parameter for tap changers [9] and I∗ are the angle-compensated CT secondary currents. The introduced method utilizes a sliding window (see Fig. 18.11) to form a matrix Idiff ABC ∈ M m,n (R) with the phase differential currents:
18 A New Second Central Moment-Based Algorithm. . .
503
Fig. 18.11 Example of sliding window
⎤ Idiff A1 Idiff B1 Idiff C1 .. .. ⎥ ⎢ .. ⎢ . . . ⎥ ⎥ ⎢ ⎢ = ⎢Idiff A32 Idiff B32 Idiff C32 ⎥ ⎥, ⎥ ⎢ . . . .. .. ⎦ ⎣ .. Idiff A64 Idiff B64 Idiff C64 ⎡
Idiff ABC
(18.8)
where m = 64 corresponds to 64 samples/cycle for n = 3 differential currents: Idiff A , Idiff B , and Idiff C . The method does not need a minimum pickup value to start. Once the differential currents matrix is formed and organized in a sliding window, these currents are filtered using a filter named Delta filter. This filter aims to eliminate all irrelevant data and periodicity from differential currents to identify variations. After applying Delta filter on differential currents, the incremental differential currents can be represented as Idiff ABC (i) = Idiff ABC (i) − Idiff ABC (i − nT ) ,
(18.9)
where i denotes the actual sample, and nT is n periods of the primary frequency. The number of periods selected was nT = 1. An example of the application of the
504
E. Vázquez et al.
Current (A)
4 2 0 −2 −4 0
0.05
0.1
0.2 0.15 Time (s)
0.25
0.3
0.25
0.3
Current (A)
a
2 0 −2 0
0.05
0.1
0.2 0.15 Time (s)
b Fig. 18.12 Example of Delta filter application. (a) Original signal. (b) Delta filtered signal
Delta filter to a sinusoidal signal is illustrated in Fig. 18.12. The sinusoidal signal was sampled at 32 samples of one cycle of 60 Hz, and the number of periods chosen was one. Figure 18.12a shows Delta filter eliminated the redundant information after the first cycle. Besides, if any alteration occurs during the filtering as it happened in the signal at ∼0.116 s (see Fig. 18.12b) the filtering will also highlight this change. Details of Delta filter can be found in [2, 8]. Once incremental differential currents are calculated, these currents must be normalized to apply the method to any power transformer independent on the transformer parameters. The normalization will scale the signals to have amplitudes between [−1, 1] dividing each sample by the maximum absolute value in each window as IN (i) =
Idiff ABC (i) , |max (Idiff ABC )|
(18.10)
where IN are the normalized differential currents, i is the actual sample, Idiff ABC are the differential currents, and |max (Idiff ABC )| is the maximum absolute value in each window.
18 A New Second Central Moment-Based Algorithm. . . Table 18.3 Thresholds criteria
Magnitude of SMC 0 ≤ SCM ≤ 0.25
SCM > 0.25
505 Events No fault condition inrush current load Turn-to-turn or inter-turn-to-ground faults
Once the differential currents are formed and normalized, they serve as input to determine the second central moment for each window data according to (18.10) as 1 2 ¯ , (IN (i) − x(i)) n n
SCM =
(18.11)
i=1
where x¯ is the average of x obtained as n #
x¯ =
IN (i)
i=1
n
,
(18.12)
where n represents the samples per window, and x is the actual sample. For this application n = 64; i. e., the selected sample rate was 64 samples/cycle. A threshold value, ε = 0.25 is compared with the SCM magnitude to identify the kind of event. The identification criteria are summarized in Table 18.3. If either one sample of the SCM is in the interval [0, 0.25], the event is identified as a steady-state condition or an inrush current, and the method obtains a new sample re-starting the algorithm. Otherwise, the event is detected as an internal fault, and a trip signal is sent. This kind of data processing does not require a pickup value to start.
18.3.1 Computational Complexity The complexity of the second central moment algorithm was estimated employing the online One-Pass method [18]. The total computational complexity of the SCM method is O (N ) where N is the window’s length. Phasor-based differential protection uses discrete Fourier transform (DFT) with an associated complexity of O N 2 [15]. Further, the fast Fourier transform (FFT) has a computational cost of O N log2 (N ) . For N = 64, the SCM method requires O (64) = 64 computations. In contrast, the DFT and FFT need O 642 = 4096 and O 64 log2 (64) = 384 calculations, respectively. Hence, the SCM method is more computationally efficient than the DFT and FFT. These results make the real-time implementation of the method feasible using a sampling period based on a multiple of 60 Hz.
506
E. Vázquez et al.
18.4 Results This section introduces the results of the differential protection based on the SCM. The method was tested in different elements, e.g., power transformers, generators, power reactors, and busbar. Furthermore, there were considered two different cases focused on how the harmonic content impacts the differential relay due to the power electronics used in renewable energies. In this section, the results are organized as follows: each figure shows the CT secondary currents from the high and low side, normalized differential currents, normalized incremental differential currents, and the magnitude of the SCM for each phase.
18.4.1 Power Transformer Protection A demonstrative example of the discrimination between transient events and internal faults using the second central moment as a basis in a power transformer is addressed. The example is divided into two different cases: a transformer energization and a fault within the differential zone. The study case setup consists of a -Y connected, 100 MVA, 230/115 kV, 60 Hz power transformer. The selected CTs were 500 : 5 and 1000 : 5 for the high and low voltage side, respectively. Figure 18.13 shows the power transformer energization at t = 1 s. Figure 18.13e shows that the magnitude of the SCM remained below the 0.25 threshold level (dashed line). Hence, the method recognized the event as a no fault event and blocked the operation of differential elements. For the second case, Fig. 18.14 shows the inception of a B-g single-phase fault inside the differential protection zone on the high side at t = 0.5 s. Although the CTs experienced saturation when the fault occurred, the SCM magnitude did cross the identification threshold at 14.6 ms as is shown in Fig. 18.14e. Therefore, the method correctly identified the event as an internal fault, and a trip signal must be sent. Similar tests such as the previously presented example were conducted to demonstrate the capability of the SCM method to differentiate transient conditions and internal faults. Results showed that the method could be used to protect the power transformer accurately.
18.4.2 Generator Protection Synchronous generators (SG) are probably the most critical elements in a power system. High repair costs and the undesirability of loss of generation are among the main reasons to protect this element adequately. Generally, the SG is protected with percentage differential protection relays to face internal phase-faults inside the differential protection zone. Other schemes usually implemented for the protection
18 A New Second Central Moment-Based Algorithm. . .
Current (A)
PA
PB
507
PC
5 0 −5 1
1.1
1.2
1.3
1.4
1.5
1.4
1.5
1.4
1.5
1.4
1.5
a Current (A)
SA
SB
SC
5 0 −5 1
1.1
1.2
1.3
diff B
diff C
b Current (A)
diff A
1 0.5 0 −0.5 1
1.1
1.2
1.3
c Current (A)
Δ
diff A
Δ
diff B
Δ
diff C
1 0.5 0 −0.5 1
1.1
1.2
1.3
d Magnitude
SCMA
SCMB
SCMC
0.25 0
1
1.1
1.2 Time (s)
1.3
1.4
e Fig. 18.13 Method performance during a power transformer energization at time t = 1 s. (a–e) Identification threshold (dashed line)
508
E. Vázquez et al.
Current (A)
PA
PB
PC
50 0 −50 −100 0.5
0.7
0.6
0.8
0.9
1
0.8
0.9
1
0.9
1
0.9
1
0.9
1
a Current (A)
SA
SB
SC
5 0 −5
0.5
0.7
0.6 b
Current (A)
diff A
diff B
diff C
0.5 0 −0.5 −1 0.5
0.7
0.6
0.8
c Current (A)
Δ
diff A
Δ
diff B
Δ
diff C
1 0.5 0 −0.5 −1 0.5
0.7
0.6
0.8
d Magnitude
SCMA
SCMB
SCMC
0.5 0.25 0
0.5
0.7
0.6
0.8
Time (s) e Fig. 18.14 Method performance during a fault inside the transformer differential protection zone at time t = 0.5 s. (a–e) Identification threshold (dashed line)
18 A New Second Central Moment-Based Algorithm. . .
509
of the SGs are time-overcurrent relays, impedance relays, negative-sequence relays, over-flux relays, loss of field relays, and stator earth-fault relays. However, these schemes have issues detecting TTFs in the stator as well as phase-to-ground faults adjacent to the neutral. Accordingly, a variation of the SCM method based on differential residual currents 3I0H = IHA + IHB + IHC
(18.13)
3I0h = Iha0 + Ihb0 + Ihc0
(18.14)
is proposed for identifying turn-to-ground and turn-to-turn faults within the SG. Two cases are studied to demonstrate the performance of the SCM-based method. For the first case, Fig. 18.15 shows a turn-to-ground fault from steady-state occurring at time t = 0.1 s. When the fault occurred, phase A current magnitude increased as illustrated in Fig. 18.15a. Figure 18.15d shows how the Delta filter highlighted the internal fault. Eventually, the magnitude of the SCM crossed the threshold recognizing the internal fault as illustrated in Fig. 18.15e. The proposed method detected the internal fault 9 ms after the fault inception. The second case does correspond to a turn-to-turn fault from steady-state. Fig. 18.16 shows the fault inception at time t = 0.105 ms. Figure 18.16a and b show the internal fault is reflected on the currents of both sides of the transformer. However, as Fig. 18.16c shows, the residual currents evidenced an internal fault occurred. Moreover, the SCM magnitude exceeded the threshold when the internal fault occurs (see Fig. 18.16e). The SCM method identified the internal fault correctly, 10 ms after the fault had happened.
18.4.3 Reactor Differential Protection Voltage control on transmission lines is sometimes achieved through high voltage (HV) shunt reactors connected on the receiving end of transmission lines to absorb reactive power. The continuous change in line loading implies that HV reactors are switched several times per day. These switching events generate inrush currents with decaying DC component appearing in one or more phases. Reactor and transformer inrush current share characteristics such as harmonic content and highly non-linear magnetization current. Conventional protection schemes (namely, overcurrent, differential, REF (restricted-earth-fault), and distance relays) typically used to protect reactors are useful to detect high-magnitude faults. However, these protections are blind to turn-to-turn faults. The SCM method performance was assessed on a reactor energization event. Figure 18.17 shows a reactor being energized at time t = 0.2 s . For this case, the algorithm outlined in Sect. 18.3 was evaluated on the residual currents shown in Fig. 18.17c. As mentioned above, the residual currents of reactors share distinctive
510
E. Vázquez et al.
Current (A)
PA
10 5 0 −5 −10
0
0.1
PB
PC
0.2
0.3
0.4
0.5
0.4
0.5
0.4
0.5
0.4
0.5
0.4
0.5
a Current (A)
SA
SB
SC
5 0 −5 0
0.1
0.2
0.3 b
Current (A)
res prim
res sec
1 0 −1 0
0.1
0.2
0.3 c
Current (pu)
Δ
N res prim
Δ
N res sec
1 0 −1 0
0.1
0.2
0.3
Magnitude
d SCMprim
SCMsec
0.2
0.3
0.5 0.25 0
0.1
Time (s) e Fig. 18.15 Method performance during a turn-to-ground fault into the generator differential protection zone at time t = 0.1 s. (a–e) Identification threshold (dashed line)
18 A New Second Central Moment-Based Algorithm. . .
Current (A)
PA
511
PB
PC
20 0 −20 −40 0
0.1
0.2
0.3
0.4
0.5
0.4
0.5
0.4
0.5
0.4
0.5
0.4
0.5
a Current (A)
SA
SB
SC
40 20 0 −20 0
0.1
0.2
0.3 b
Current (A)
res prim
res sec
1 0 −1 0
0.1
0.2
0.3 c
Current (pu)
Δ
N res prim
Δ
N res sec
1 0 −1 0
0.1
0.2
0.3
Magnitude
d SCMprim
SCMsec
0.2
0.3
0.5 0.25 0
0.1
Time (s) e Fig. 18.16 Method performance during a turn-to-turn fault into the generator differential protection zone at time t = 0.105 s. (a–e) Identification threshold (dashed line)
512
E. Vázquez et al.
Current (A)
PA
PB
PC
10 5 0 −5 0.15
0.2
0.3
0.25
0.4
0.45
0.35
0.4
0.45
0.35
0.4
0.45
0.4
0.45
0.35
a Current (A)
SA
5 0 −5 −10 0.15
0.2
SB
SC
0.3
0.25 b
Current (A)
res prim
res sec
0 −0.5 0.15
0.2
0.3
0.25 c
Current (pu)
Δ 0.5 0 −0.5 −1 0.15
0.2
Δ
N res prim
N res sec
0.3
0.25
0.35
d Magnitude
SCMprim
SCMsec
0.25
0
0.2
0.3
0.25
0.35
0.4
Time (s) e Fig. 18.17 SCM method performance during a power reactor energization. (a) CT primary currents. (b) CT secondary currents. (c) Residual currents. (d) Filtered and normalized residual currents. (e) SCM magnitude (solid lines) and identification threshold (dashed line)
18 A New Second Central Moment-Based Algorithm. . .
513
features with transformer inrush currents. After a filtering stage (see Fig. 18.17d) the method identified the event as a non-fault condition. Figure 18.17e shows that the magnitude of the SCM did not exceed the threshold of 0.25. Figure 18.18 presents a second case where a turn-to-turn fault occurred at time t = 0.1 s. Because the reactor is a serial element, differential relays have problems detecting turn-to-turn faults. Nevertheless, the SCM method detected the residual current change, as shown in Fig. 18.18c. Moreover, the Delta filter highlighted the internal fault waveform (see Fig. 18.18d) allowing the method to identify the event as an internal fault correctly: the magnitude of the SCM exceeded the threshold of 0.25 as shown in Fig. 18.18e.
18.4.4 Busbar Differential Protection Busbar protection could be considered one of the most critical elements of protection. The high fault current levels and the costly damages to the busbar, to all associated elements, and the rest of the system in case of fault are the main reasons. Moreover, the clearance of a fault in a busbar reduces the power transfer capability. This is undesirable effect where the system capability is at limit. Busbar protection depends on several factors (e.g., the rated voltage or the complexity of network configuration such as the number of bays, sections, tiebreakers, and disconnects). Further, the availability of CTs must be considered. The most typical schemes used in busbar protection are overcurrent, percent differential, partial differential, high-impedance, directional comparison differential, and timecoordinated. However, the critical issue busbar protection schemes face is the heavy saturation occurring in current transformers. Usually, current transformers have similar CT ratios but are insufficiently rated. CT saturation may lead to the misoperation of protection relays because of residual currents seen as differential currents by the relays. Hence, the security of protection schemes is compromised. On this basis, the SCM method performance was evaluated in two cases: a fault inside and a fault outside the protection zone. Figure 18.19 shows the results for a fault inside the differential protection zone occurring at time t = 1 s from steadystate. The CTs experienced heavy saturation (see Fig. 18.19b) because of the high DC decay component when the fault occurred. However, Fig. 18.19c shows an increase in the magnitude of the differential currents. As seen in Fig. 18.19e, the magnitude of the SCM exceeded the 0.25 limit of 0.25 classifying the event as an internal fault 26 ms after the fault occurred. Protection of busbar elements against external faults is one of the critical scenarios because of the high CT saturation level that CTs are exposed. The CT saturation will distort the waveform of the CT secondary currents leading the misoperation of the protection elements. Consequently, the SCM method was assessed for a fault outside the differential zone. Figure 18.20 shows an ABC-g three-phase external fault occurring at time t = 0.5 s. Even as CTs are heavily saturated (see Fig. 18.20a and b), the magnitude of the SCM stayed below the 0.25
514
E. Vázquez et al.
Current (A)
PA
PB
0 −5 0
0.05
0.1
0.15
0.2 a
Current (A)
SA
SB
Current (A)
0.35
0.4
0.45
0.3
0.35
0.4
0.45
0.3
0.35
0.4
0.45
0.35
0.4
0.45
SC
0 −5 0.05
0.1
0.15
0.2 b
res prim
0.25 res sec
1 0 −1 0
0.05
0.1
0.15 Δ
Current (pu)
0.3
0.25
5
0
0.2 c
0.25 Δ
N res prim
N res sec
1 0 −1 0
0.05
0.1
0.15
0.2 d
SCMprim Magnitude
PC
5
0.25
0.3
SCMsec
0.5 0.25 0
0.05
0.1
0.15
0.2
0.25 Time (s)
0.3
0.35
0.4
e Fig. 18.18 SCM method performance identifying a turn-to-turn fault in an HV reactor. (a) CT primary currents. (b) CT secondary currents. (c) Residual currents. (d) Filtered and normalized residual currents. (e) SCM magnitude (solid lines) and identification threshold (dashed line)
18 A New Second Central Moment-Based Algorithm. . .
Current (A)
PA
PB
515 PC
20 0 −20 0.9
1
1.1
1.2
1.3
1.4
1.5
1.4
1.5
1.4
1.5
1.4
1.5
a Current (A)
SA
SB
SC
20 0 −20 0.9
1
1.1
1.2
1.3
diff B
diff C
1.2
1.3
b Current (A)
diff A
1 0.5 0 0.9
1
1.1 c
Current (A)
Δ
diff A
Δ
diff B
Δ
diff C
50 0 0.9
1
1.1
1.2
1.3
d Magnitude
SCMA
SCMB
SCMC
0.5 0.25 0
1
1.1
1.2 Time (s)
1.3
1.4
e Fig. 18.19 SCM method performance during a fault inside the busbar differential protection zone. (a) CT primary currents. (b) CT secondary currents. (c) Residual currents. (d) Filtered and normalized residual currents. (e) SCM magnitude (solid lines) and identification threshold (dashed line)
516
E. Vázquez et al.
level as seen in Fig. 18.20e. Therefore, the SCM method identified the fault as an external fault condition.
18.4.5 Power Transformer Energization with Harmonic Content As high harmonic distortion produced by non-linear loads may cause the malfunction of the differential relays, the SCM method was tested for a power transformer supplying an IGBT-based non-linear load [14]. Figure 18.21a and b show CT secondary currents for the HV and LV sides, respectively, during a transformer energization occurring at time t = 0.3 s. Figure 18.21a shows that the magnitude of the differential currents is affected negatively by the IGBT-based load. Despite the effects of the non-linear load, the filter stage removed most of the errors (see Fig. 18.21d). Figure 18.21e shows that the magnitude of the SCM remained below the 0.25 threshold for the entire duration of the event.
18.4.6 Power Reactor Energization with High Harmonic Distortion For the sake of assessing the robustness of the SCM in the presence of highly harmonic contaminated currents, it is considered the case of a reactor energization with harmonic content. Figure 18.22a shows the reactor energization occurring at time t = 0.1 s. The currents measured by the CT display a high harmonic content (1, 4, 61, 80, 120, 160.3, 180, 200, 210, 240, 300, 340, and 360 Hz) as seen in Fig. 18.23. Notwithstanding, the filtering stage removes the periodicity present in the current signals (see Fig. 18.22d), allowing the SCM method to identify the event correctly. Figure 18.22e shows that the SCM magnitude did not cross the threshold.
18 A New Second Central Moment-Based Algorithm. . .
Current (A)
PA
517
PB
PC
100 50 0 −50 0.45
0.5
0.55
0.6
0.65
0.7
0.65
0.7
0.65
0.7
0.65
0.7
a Current (A)
SA
SB
SC
20 0 −20 0.45
0.5
0.55
0.6 b
Current (A)
diff A
diff B
diff C
1 0.5 0 0.45
0.5
0.55
0.6 c
Current (A)
Δ
diff A
Δ
Δ
diff B
diff C
1 0 −1 0.45
0.5
0.55
0.6 d
Magnitude
SCMA
SCMB
SCMC
0.25
0
0.5
0.55
0.6
0.65
Time (s) e Fig. 18.20 Method performance during an external three-phase fault of the busbar differential protection zone. (a) CT primary currents. (b) CT secondary currents. (c) Residual currents. (d) Filtered and normalized residual currents. (e) SCM magnitude (solid lines) and identification threshold (dashed line)
518
E. Vázquez et al.
Current (A)
PA
PB
10 0 0.3
0.32
0.34 a
Current (A)
SA
SB
0.32
0.34 b
Current (A)
res prim
0.38
0.4
0.38
0.4
0.38
0.4
0.38
0.4
0.38
0.4
SC
0.36 res sec
1 0.5 0 −0.5 0.3
0.32
0.34 c Δ
Current (A)
0.36
4 2 0 −2 −4 0.3
N res prim
0.36 Δ
N res sec
1 0.5 0 −0.5 0.3
0.32
0.34 d SCMprim
Magnitude
PC
0.36 SCMsec
0.25
0
0.3
0.32
0.34
0.36 Time (s)
e Fig. 18.21 Method performance during a transformer energization event supplying a non-linear load. (a) Primary side currents. (b) Secondary side currents. (c) Differential currents. (d) Delta filtered differential currents. (e) SCM magnitude (solid lines) and identification threshold (dashed line)
18 A New Second Central Moment-Based Algorithm. . .
Current (A)
PA
519
PB
PC
10 0 −10
0.1
0.2
0.15
0.25
0.3
0.25
0.3
0.25
0.3
0.25
0.3
a Current (A)
SA
SB
SC
10 0 −10 0.1
0.2
0.15 b
Current (A)
res prim
res sec
5 0 −5 0.1
0.2
0.15 c
Current (pu)
Δ
N res prim
Δ
N res sec
1 0 −1 0.1
0.2
0.15 d
Magnitude
SCMprim
SCMsec
0.25
0
0.1
0.2
0.15
0.25
Time (s) e Fig. 18.22 Method performance during a reactor energization with high harmonic currents. (a) CT primary currents. (b) CT secondary currents. (c) Residual currents. (d) Filtered and normalized residual currents. (e) SCM magnitude (solid lines) and identification threshold (dashed line)
520
E. Vázquez et al.
120 Hz
FFT spectrum amplitude
8,000
6,000
1 Hz
80 Hz
4,000
180 Hz 160 Hz
4 Hz
2,000
210 Hz 200 Hz
61 Hz
240 Hz
0
0
50
100
200 150 250 Frequency (Hz)
300 Hz
300
360 Hz
350
Fig. 18.23 Fourier spectrum from a reactor energization with high harmonic content
References 1. Behrendt, K., Fischer, N., & Labuschagne, C. (2011). Considerations for using harmonic blocking and harmonic restraint techniques on transformer differential relays. Journal of Reliable Power, 2(3), 36–52 (2011). 2. Benmouyal, G., & Roberts, J. (1999). Superimposed quantities: Their true nature and application in relays. In 1999 26th Annual Western Protective Relay Conference (pp. 1–18). 3. Chaitanya, B. K., Soni, A. K., & Yadav, A. (2018). Communication assisted fuzzy based adaptive protective relaying scheme for microgrid. Journal of Power Technologies, 98(1), 57–69. 4. Che, L., Khodayar, M. E., & Shahidehpour, M. (2014). Adaptive protection system for microgrids: Protection practices of a functional microgrid system. IEEE Electrification Magazine, 2(1), 66–80. 5. Edwards, B., Williams, D. G., Hargrave, A., Watkins, M., & Yedidi, V. K. (2017). Beyond the nameplate – Selecting transformer compensation settings for secure differential protection. In 2017 70th Annual Conference for Protective Relay Engineers (CPRE) (pp. 1–23). 6. Hargrave, A., Thompson, M. J., & Heilman, B. (2018). Beyond the knee point: A practical guide to CT saturation. In 2018 71st Annual Conference for Protective Relay Engineers (CPRE) (pp. 1–23). 7. He, M., Zhang, J., & Vittal, V. (2013). Robust online dynamic security assessment using adaptive ensemble decision-tree learning. IEEE Transactions on Power Systems, 28(4), 4089– 4098. 8. Hensler, T., Pritchard, C., Fischer, N., & Kasztenny, B. (2018). Testing superimposedcomponent and traveling-wave line protection. In 2018 72nd Annual Georgia Tech Protective Relaying Conference (pp. 1–12).
18 A New Second Central Moment-Based Algorithm. . .
521
9. IEEE guide for the application of current transformers used for protective relaying purposes (2008). IEEE Std C37.110-2007 (Revision of Std C37.110-1996) (pp. 1–90). 10. Kasztenny, B., Fischer, N., & Altuve, H. J. (2015). Negative-sequence differential protection – Principles, sensitivity, and security. In 2015 68th Annual Conference for Protective Relay Engineers (pp. 364–378). 11. Kauhaniemi, K., & Voima, S. (2012). Adaptive relay protection concept for smart grids. In: Proceedings of the Renewable Efficient Energy II Conference (pp. 1–10). 12. Lin, H., Sun, K., Tan, Z., Liu, C., Guerrero, J.M., & Vasquez, J.C. (2019). Adaptive protection combined with machine learning for microgrids. IET Generation, Transmission Distribution, 13(6), 770–779. 13. Massoud Amin, S., & Wollenberg, B. F. (2005). Toward a smart grid: Power delivery for the 21st century. IEEE Power and Energy Magazine, 3(5), 34–41. 14. Mishra, M. K., & Karthikeyan, K. (2009). An investigation on design and switching dynamics of a voltage source inverter to compensate unbalanced and nonlinear loads. IEEE Transactions on Industrial Electronics, 56(8), 2802–2810. 15. Proakis, J. G., & Manolakis, D. G. (1996). Digital signal processing: Principles, algorithms, and applications (3rd ed.) Upper Saddle River, NJ: Prentice Hall. 16. Samantaray, S., Mishra, D., & Joos, G. (2018). A combined wavelet and data-mining based intelligent protection scheme for microgrid. In: 2018 IEEE Power Energy Society General Meeting (PESGM) (pp. 1–1). 17. Senarathna, T. S., & Udayanga-Hemapala, K. T. M. (2019). Review of adaptive protection methods for microgrids. AIMS Energy, 7(5), 557–578. 18. Welford, B. P. (1962). Note on a method for calculating corrected sums of squares and products. Technometrics, 4(3), 419–420. 19. Zacharias, D., & Gokaraju, R. (2016). Prototype of a negative-sequence turn-to-turn fault detection scheme for transformers. IEEE Transactions on Power Delivery, 31(1), 122–129. 20. Zhang, Y., Huang, T., & Bompard, E. F. (2018). Big data analytics in smart grids: A review. Energy Informatics, 1(8), 1–24. 21. Ziegler, G. (2012). Numerical differential protection: Principles and applications (2nd ed.) Erlangen: Publicis Publishing.
Chapter 19
Microgrid Protection with Conventional and Adaptive Protection Schemes Aushiq Ali Memon, Hannu Laaksonen, and Kimmo Kauhaniemi
19.1 Introduction Microgrid is considered as the building block for the future smart grids, therefore a properly designed microgrid is essential for the proper functioning of the entire smart grid. Microgrid has many definitions but in general, it is a well-designed local distribution system of electricity that facilitates the local integration of many small-scale renewable energy sources and energy storage systems to meet the local energy demand of consumers in a smart, secure, efficient, controlled, protected, and managed environment. Unlike the traditional distribution systems of electricity, microgrids can operate in islanded mode when the main grid is disconnected due to faults. The islanded mode operation of the microgrid will enhance the reliability level of the existing distribution system if some sections of distribution networks are planned as microgrids. However, the islanded mode operation of microgrid has many challenges and one important challenge is related to the design of the protection scheme. This chapter addresses the issues related to protection schemes in a microgrid, gives an overview of the existing and new requirements of protection schemes, and analyses the potential of the existing and adaptive protection schemes of a microgrid.
A. A. Memon () · H. Laaksonen · K. Kauhaniemi School of Technology and Innovations, University of Vaasa, Vaasa, Finland e-mail: [email protected] © Springer Nature Switzerland AG 2021 A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems, https://doi.org/10.1007/978-3-030-59750-4_19
523
524
A. A. Memon et al.
19.2 Microgrid Protection Issues Microgrid protection issues can be classified into two broad categories depending on its operational modes [1]: (1) Microgrid protection issues in grid-connected mode (2) Microgrid protection issues in islanded mode. For the grid-connected mode of microgrid, the faults inside the microgrid as well as faults outside the microgrid are considered. In the grid-connected mode protection schemes of microgrids should not operate unnecessarily for faults outside the microgrid for example faults upstream of the circuit breaker (CB) at the point of common coupling (PCC). All faults insides the microgrid should be detected and selectively isolated for the minimum interruption to other parts. During external faults at the main grid, microgrid should be able to disconnect quickly to protect its loads and start operating in islanded mode and as soon as the external fault is removed it should be reconnected to the main grid. The external faults at the main grid could be close-in faults or other far-end faults resulting in the loss of mains, hence both types of faults should be differentiated. In the grid-connected mode fault contribution from both the main grid and microgrid sources is being considered. If microgrid has the majority of synchronous generator-based distributed energy resources (DERs), then fault contribution from the main grid could be reduced and overcurrent relays may experience selectivity issues like more time to issue a trip command. On the other hand, if the majority of converter-based DERs are connected inside the microgrid, then it will not have significant issues during the grid-connected mode since fault contribution from the converter-based DERs could be limited to twice the rated current capacity or even less depending on the converter settings or rating. For the islanded mode of operation, the faults inside the microgrid are only considered and fault contribution from local DERs and energy storage systems is taken into account. Microgrid protection should detect and isolate the faults selectively even in an islanded mode of operation. During island operation, for example, in low-voltage (LV) microgrids, large fault currents from the upstream power system grid are not available. In addition, a large share from the DER units in LV microgrids will be inverter-based with low fault currents. Therefore, traditional one-directional protection schemes, assuming a large difference between fault and load currents, are not applicable during island operation. These traditional protection methods could also have slower fault clearing time and reduced sensitivity and selectivity. This means that the system reliability is expected to decrease if the protection schemes are not adapted [2]. For these reasons, conventional overcurrent (OC) protection based on fuses and one setting group will not be able to guarantee selectivity during different types of possible faults. Therefore, LV network conventional protection will not be compatible with island operated LV microgrids and new protection schemes with adaptivity must be created. On the other hand, the new LV microgrid protection scheme needs to be economical and simple [3, 4]. For medium-voltage (MV) microgrids after isolation from the
19 Microgrid Protection with Conventional and Adaptive Protection Schemes
525
utility grid, local DERs are the only fault current sources in the electric island, and the fault current level depends on the types, sizes, and locations of the DER. However, it is generally lower than the fault current from the utility grid [5]. The reduced fault current contributions from microgrid DERs require revised protection settings with the reduced pickup or threshold values of currents for islanded mode or protection schemes based on other protection principles like differential current [6], symmetrical components and residual current based [7], voltage based [8], harmonic content based [9], or a suitable combination of them should be employed to detect and clear the fault in islanded mode. An adaptive protection scheme [10] using highspeed communication links and numerical directional overcurrent protection relays could also be a suitable protection scheme to change protection settings adaptively according to the grid-connected mode or islanded mode operations. In the creation of a new protection system for microgrids, multiple issues need to be taken into account, like • The number of zones for protection, • Operation speed specifications for the different operation modes and configurations of microgrid and. • Protection methods for microgrid normal grid-connected and islanded operation [4]. The created microgrid protection system also needs to be compatible with the microgrid operation and control solutions. Some key issues related to the LV microgrid protection are briefly reviewed based on [11] from which more detailed information can be found. The extent and number of microgrid protection zones will determine the required number of protective devices (PDs) for microgrid protection. However, the protection system simultaneously needs to fulfill customer requirements and be economically feasible [4]. The essential structural choices will define the operation speed needs and principles for the protection of LV microgrid and correspondingly the operation speed requirements will determine some of the structural choices required to fulfill the operation speed needs. The main reasons for the operation speed requirements of LV microgrid protection are stability and customer sensitivity. The stability needs to be maintained after fast disturbances like after islanding due to fault in the upstream network during normal grid-connected operation or after a fault in the LV microgrid during islanded operation. One important issue related to the operation principles of LV microgrid protection is the fault behavior of the converter-based DER units. The fault behavior needs to be compatible with the developed microgrid protection Scheme [4]. As stated in [12], the microgrid protection issues cannot be solved without a complete understanding about the microgrid dynamics before, during, and after islanding or fault. Related to this, for example, directly connected rotating generators or motors are very sensitive from the stability viewpoint in voltage dips caused by faults during microgrid island operation, and so they may endanger the stability of the microgrid. Therefore, if directly connected rotating machines are connected to a microgrid, protection should operate rapidly during all kinds of faults. For
526
A. A. Memon et al.
example, if microgrid customers have fuses with high nominal currents, there can be a risk that customer protection may operate too slowly during island operation due to small fault currents and that may lead to instability after clearance of the fault [4]. In cases where overcurrent based protection is utilized during island operation, the protection and control functions of IEDs (intelligent electronic devices) in microgrids may need real-time information about network topology, the status of DER units (on or off), the state of charge of storage systems, and also number and size of loads connected to the microgrid. These conditions have to be updated and checked continuously in order to guarantee that protection settings are suitable for the actual configuration [3, 4]. Based on the above and as mentioned in [12], the high-speed operation of the protection devices is very crucial for reliable operation of the microgrid protection system. Utilization of high-speed telecommunication is expected to be an essential part of future smart grid protection systems to achieve fast and selective protection both in grid-connected and islanded modes of operation. The same communication protocols and standards used in HV/MV network can be applied directly to the LV microgrids. However, due to the smaller scale of LV microgrids, the costs of protection devices must also be lower than the cost of devices used in the HV/MV network [4]. One important issue, which is required to enable a stable transition from the normal grid-connected operation to island operation, is the coordination of IED protection settings with DER unit fault-ride-through (FRT) requirements (especially low-voltage-ride-through, LVRT). In order to avoid unwanted tripping, faulty lines must be disconnected first by the protection system and after that, the DER units should be disconnected according to their FRT or LVRT functionality. Rapid protection operation is needed especially if there are many protection zones. Therefore, communication-based protection methods and schemes are often required to achieve selectivity [13]. Some further discussion about issues related to the protection of microgrids can be also found in [2, 4].
19.2.1 DER Unit Fault Behavior and Effect on Microgrid Protection In the future, proper coordination between distributed generation (DG) unit grid codes and distribution network protection schemes will be increasingly important during both grid-connected and islanded operation of microgrids. The operation time settings of short-circuit and earth-fault protection must be selective with DER unit fault-ride-through settings during normal operation. In MV network shortcircuit protection operation time delays have traditionally been dependent on the fault-current magnitude or measured impedance with fixed time delays or inverse time curves. In the future, MV networks will be increasingly divided into multiple
19 Microgrid Protection with Conventional and Adaptive Protection Schemes
527
protection zones to improve supply reliability. Therefore, short-circuit protection operation times may become too long if high-speed communication based, and for example, IEC 61850 GOOSE signals-based interlocking/blocking schemes are not utilized. On the other hand, the communication may fail or is not available, and therefore also grid code compatible protection schemes that are not based on high-speed communication are needed in the future at least as a backup for communication-based schemes. Regarding the fault behavior required from DER units during faults, it is important to ensure that the fault behavior is compatible with the developed LV microgrid protection scheme and considers also the including FRT needs. This means that when the protection for an island operated microgrid is determined, one of the most important considerations is related to the fault current contribution of the converter-based DER units [4, 14]. Fault behavior and fault current feeding capability of a DER unit is also highly dependent on the type of the DER unit. For example, a synchronous generator is usually able to feed prolonged fault current (about 200 to 400% of nominal current). An induction generator, in the initial stage, feeds almost as big a fault current as a synchronous generator, but the feeding is reduced quickly. Fault current fed by the inverter-connected generating unit is typically limited to 1.2–1.3 p.u. and highly dependent on the control system and control principles as well as grid code requirements regarding fault-ride-through and reactive current feeding requirements. During normal grid-connected operation of microgrid with different types of DER units, the grid codes can require FRT capability from the DER units in terms of frequency (f ), rate-of-change-of-frequency (df/dt), voltage (U), and voltage support. Traditionally, in grid-connected operation, larger DER units have more FRT and frequency as well as voltage support related requirements in the grid codes. For example, regarding voltage support, the FRT capability of DER units is defined with a voltage-against-time-profile (LVRT curve). In addition, also additional voltage support by capacitive/reactive, positive sequence, current injection during faults is required from MV and HV network connected converter−/doubly-fed-inductiongenerator (DFIG)-based DER units during grid-connected operation. Synchronous generator DER units naturally provide voltage support during faults by feeding reactive current. The required dynamic response of the reactive current feeding is usually also defined in the grid codes for the converter-based DER units. However, LVRT capability and additional reactive current feeding of the converter-based DER units, like wind turbines, is not just a control issue. It also requires suitable technology to be applied in the DC-link of the converter, like for example, DC-link chopper or supercapacitor and crowbar possibly as a backup. Most grid codes for the grid-connected operation of DER units do not include any special requirements for the supply of negative sequence current. In synchronous generators, the negative sequence current is fed naturally and there are no effective measures to influence it. In contrast, with voltage source converters (VSCs) it is possible to individually control positive and negative sequence quantities. The main reason for the minimization of negative sequence current has been the impact of
528
A. A. Memon et al.
asymmetrical voltages on the DER unit (e.g., wind turbine) without considering the impact of it on the network voltages or network protection. The negative sequence may be controlled to mitigate twice the fundamental frequency oscillations appearing in the converter/inverter DC-link during asymmetrical faults. Therefore, in many applications the negative sequence component injection has been reduced partially or fully. However, full negative sequence current reduction control also reduces the line-to-line (2-phase) short circuit current to the level of the load current or even to zero, which means that it may prevent the correct operation of network protection in 2-phase short-circuit faults. To overcome this problem, extensions to grid code have been proposed, which would require DG to inject a clearly defined level of negative sequence current. This is also a requirement already in Germany. From the network voltages point of view, the effect of negative sequence fault current feeding during asymmetrical faults is beneficial because it reduces the negative sequence voltage, improves the voltage phase symmetry, and reduces the overvoltages in the healthy phases during 2-phase short-circuits. However, a negative sequence current injection will limit the control capability of the DG in the positive sequence. It can be expected that the future grid codes will increasingly specify requirements for asymmetrical/imbalanced negative sequence current injection during asymmetrical faults. Usually, the control mode change of one or more DER units connected to the distribution network is required after changing from the normal grid-connected operation to island mode. Traditionally, this means that under normal operation the DER unit uses active(P)/reactive(Q) power control and after islanding, the control mode is changed to voltage(U)/frequency(f ) control (or voltage/speed control). However, control schemes that do not require changing after transition to/from island operation have also been proposed. For example, in [15], an enhanced control strategy was proposed that improves the performance of a DER unit under network faults and transient disturbances, in a multiunit microgrid setting. The proposed control strategy does not require the detection of the mode of operation and switching between different controllers (for grid-connected and islanded) modes, and it enables the adopted DER units to ride through network faults, irrespective of whether they take place within the host microgrid or impact the upstream grid [15]. LVRT, high voltage ride-through (HVRT) as well as f and df/dt related FRT capabilities are required during LV microgrid island operation also from the smallscale DER units, which typically is not the case during grid-connected operation. Typically, in grid-connected operation the small-scale DER units, like PV units, are only required to support frequency stability during over-frequency situations by their active power-frequency (Pf )-droop control. From the island operated LV microgrid protection viewpoint, it is important to know exactly how the converterbased DER units behave during the faults and what kind of fault current (active, reactive, positive sequence, negative sequence, etc.) they will feed. Therefore, the so-called microgrid grid codes for island operated networks are necessary for the development of future smart grid protection solutions to reduce complexity and to avoid the need for too many case specific alternatives [4, 14].
19 Microgrid Protection with Conventional and Adaptive Protection Schemes
529
Based on the simulations done in [14], the increased reactive power feeding with the converter-based DER units was found to be beneficial for the possible overcurrent-based protection in LV microgrid. However, it did not significantly reduce the usability of undervoltage-based protection either due to the resistive characteristics of LV lines. On the other hand, the reactive power feeding during the fault did not significantly reduce the magnitude of the voltage dip, that is, support microgrid voltage during the fault. In the end, the excessive reactive power feeding of the converter-based DER units during the fault in island operated LV microgrid was not justified based on the simulations in [14]. Therefore, it was suggested that during faults in LV island operated microgrid the fault current fed by the converterbased DER units is recommended to be active instead of reactive if possible. In addition, the control of the converters during possible faults was not recommended to change due to the increased possibility of instabilities after fault clearance.
19.2.2 Example—Microgrid Transition to Islanded Operation In the following, the example from [16] is used to define the protection needs (functions, time selectivity) when intentional island operation and especially successful transition to island operation are considered. From the island operation perspective (in addition to grid code FRT requirements from DG units during the normal grid-connected operation) it is required that the wind farm (green) in Fig. 19.1 as well as DER (1) and (2) have sufficient FRT capabilities. Figure 19.1 shows possible intended islands and MV feeder shortcircuit protection at CB1-CB4 which is assumed to be directional like the earth-fault protection. In the following, the time selectivity issues (Fig. 19.2) are discussed, with different fault scenarios (faults A-E in Fig. 19.1), regarding the successful transition to island operation. In Fig. 19.2 protection time selectivity issues, general time delay setting principles, and the role of high-speed communication is shown when (a) islanding is not allowed and (b) islanding is possible. Figure 19.2 also illustrates the role of highspeed communication-based interlockings/blockings (as well as transfer trip-based islanding detection) in the reliable and selective operation of future distribution networks with many sequential protection zones and the possibility for the intended island operation. In Fig. 19.1 and 19.2, the idea is that the possible operation principles of directional short-circuit protection in the forward direction can be 1. Directional overcurrent protection with a fixed time delay (and high-stage/lowstage settings). 2. Distance protection with a fixed time delay (in a forward direction). Similarly, in Fig. 19.1 and 19.2 the possible operation principles of directional protection in the reverse direction (for intentional islanding) can be
530
A. A. Memon et al.
Fig. 19.1 Possible intended islands 1.-4. b (see also Fig. 19.2) [16]
1. Undervoltage with a fixed time delay (and high-stage/low-stage settings) AND current direction detection in the reverse direction. Function pick-up/start is only based on undervoltage (i.e., not in overcurrent, because fault current levels of inverter-based DER units can be fairly low as discussed in previous chapters). 2. Distance protection with a fixed time delay (in the reverse direction).
19 Microgrid Protection with Conventional and Adaptive Protection Schemes
531
Fig. 19.2 Protection time selectivity issues, setting principles, and the role of high-speed communication when (a) islanding is NOT allowed and (b) islanding IS possible (see also Fig. 19.1) [16]
532
A. A. Memon et al.
From Fig. 19.1 and 19.2 it can be seen that selectivity problems are possible if communication-based interlockings/blockings, for example, are NOT used (Fig. 19.2a), because coordination between LVRT curve of DG units (defined by grid codes) and required time differences between CB2 and CB3 in the forward direction may be hard to achieve. This naturally depends on the number of consecutive protection zones and the allowed time difference between the operation time delays of CB2 and CB3. Transition to intentional island operation IS only possible (Fig. 19.2b) if active and reactive power unbalance at CB1, CB2, CB3, or CB4 is small enough (or enough, rapidly controllable active, and reactive power units exist in the possible island (1.-4.b in Fig. 19.1) before the protection start/operation of CB1-CB4 in the reverse direction). If this is not the case transition to island operation should not be allowed. Here it is worth mentioning that the recent grid codes enable/support transition to intentional island operation because of the P/f -droop control requirements of DER units during over-frequency situations (under-frequency based load shedding schemes could have a similar kind of effect) and possibly also due to voltage control (Q/U-control) requirements. In the above discussion and Fig. 19.2, only short-circuit protection has been considered, but naturally also earth-fault protection principles and settings must be proper during both normal and island operation. Therefore, it should be noted here that after opening CB2, CB3, or CB4, the MV network neutral earthing method may change, for example, from compensated to isolated and MV feeder IEDs earth-fault protection settings and protection principles also need to adapt to these changes. In the following, the protection operation principles during different faults A-E in the example network shown in Fig. 19.1 are described. It is assumed that high-speed communication is available/possible, and islanding IS possible (Fig. 19.2b) using time selectivity (with the communication) if power generation and consumption are close to each other behind the possible island connection point CB. In case of fault A in Fig. 19.1 (see also Fig. 19.2): • CB1 will operate in a reverse direction and disconnect the LV microgrid intentionally from the utility network. – If active and reactive power unbalance at CB1 is small enough (i.e., stable transition to island operation is possible) as stated earlier. – CB1 could also send signals to the LV microgrid DER units to change their control mode, and so on after the operation. • CB2 will operate in a forward direction. – CB2 sends a simultaneously interlocking signal to CB3 (and CB4) to prevent their false operation and. – CB2 can also send a communication-based transfer trip (faulty island) disconnection signal to the DER unit (2). Wind farm and DER unit (1) will remain connected according to the LVRT curve of DG units (Fig. 19.2).
19 Microgrid Protection with Conventional and Adaptive Protection Schemes
533
In case of fault B in Fig. 19.1 (see also Fig. 19.2): • CB1 will operate in a reverse direction and disconnect the LV microgrid intentionally from the utility network. – If active and reactive power unbalance at CB1 is small enough. – CB1 could also send signals to the LV microgrid DER units to change their control mode, and so on after the operation. • CB2 may also operate in a reverse direction and disconnect part of the MV feeder intentionally from the utility network to island operation. – If active and reactive power unbalance at CB2 is small enough. – CB2 could also send a signal to the DER unit (2) to change the control mode, and so on after the operation. • CB3 will operate in a forward direction. – CB3 sends a simultaneously interlocking signal to CB4 to prevent false operation and. – CB3 can also send a communication-based transfer trip (faulty island) disconnection signal to the DER unit (1). Wind farm will remain connected according to the LVRT curve (Fig. 19.2). In case of fault C in Fig. 19.1 (see also Fig. 19.2): • MV busbar fault. • CB1 will operate in a reverse direction and disconnect the LV microgrid intentionally from the utility network. – If active and reactive power unbalance at CB1 is small enough. – CB1 could also send signals to the LV microgrid DER units to change their control mode, and so on after the operation. • CB2 may also operate in a reverse direction and disconnect part of the MV feeder intentionally from the utility network to island operation. – If active and reactive power unbalance at CB2 is small enough. – CB2 could also send a signal to the DER unit (2) to change the control mode, and so on after the operation. • CB3 may also operate in a reverse direction and disconnect either part of the MV feeder (beginning of the feeder, 3. in Fig. 19.1) OR the whole MV feeder (2. and 3. in Fig. 19.1) to island operation. – Depending on the active and reactive power unbalance at CB3 and CB2. – Coordination with islanding of part of MV feeder by opening CB2 may be beneficial/required. – CB3 could also send a signal to the DER unit (1) (AND/OR DER unit (2) depending on the power balance situation) to change control mode, and so on or to disconnect. • CB4 will operate in a forward direction.
534
A. A. Memon et al.
The wind farm will be disconnected according to LVRT curve (Fig. 19.2). Also, directional short-circuit protection in the reverse direction could be included in CB3c in order to disconnect the wind farm more rapidly in case of busbar faults like fault C (Fig. 19.1). However, in case of upstream faults (like fault E and F in Fig. 19.1), this reverse direction protection at CB3c should be blocked by communication to enable fault-ride-through support from the wind farm according to grid codes (e.g., LVRT curve). In case of fault D in Fig. 19.1 (see also Fig. 19.2): – Parallel MV feeder fault. – CB3b will operate in a forward direction (Fig. 19.1) and simultaneously send the interlocking signal to other CBs (e.g.) to avoid unnecessary islanding and to ensure selectivity. DER units (1) and (2) as well as the wind farm will remain connected according to the LVRT curve (Fig. 19.2). In case of fault E in Fig. 19.1 (see also Fig. 19.2): • HV/MV transformer fault => intentional islanding can take place. – Possible indication about intentional islanding possibility from HV/MV transformer protection IED to MV feeder IEDs. • CB1 will operate in a reverse direction and disconnect the LV microgrid intentionally from the utility network. – If active and reactive power unbalance at CB1 is small enough. – CB1 could also send signals to the LV micro-grid DER units to change their control mode, and so on after the operation. • CB2 may also operate in a reverse direction and disconnect part of the MV feeder intentionally from the utility network to island operation. – If active and reactive power unbalance at CB2 is small enough. – CB2 could also send a signal to the DER unit (2) to change the control mode, and so on. • CB3 may also operate in a reverse direction and disconnect either part of the MV feeder (beginning of the feeder, 3. in Fig. 19.1) OR the whole MV feeder (2. And 3. in Fig. 19.1) to island operation. – Depending on the active and reactive power unbalance at CB3 and CB2. – Coordination with islanding of part of MV feeder by opening CB2 may be beneficial/required. – CB3 could also send a signal to the DG unit (1) (AND/OR DER unit (2) depending on the power balance situation) to change the control mode, and so on. or to disconnect. • Alternatively, CB4 may also operate in reverse direction and disconnect. – All MV feeders (4.b in Fig. 19.1) to island operation.
19 Microgrid Protection with Conventional and Adaptive Protection Schemes
535
– Some of the MV feeders (e.g., (4.a in Fig. 19.1)) to island operation. Simultaneously, the disconnection signal from CB4 is sent to MV feeder CBs (e.g., (CB3c in Fig. 19.1)) which cannot be included into an intentional island. Wind farm remains connected according to the LVRT curve (Fig. 19.2) unless the disconnection signal to CB3c (Fig. 19.1) is sent by CB4. – This intentional island scheme is somewhat more complex due to the increased number of possible island sizes. – Size of the island depends on the active and reactive power unbalance at CB4, CB3, and CB2 and. Needs (central) coordination with islanding of part of or whole MV feeder by opening CB3 or CB2. Signals to DER unit (1) and (2) to change the control mode, and so on after the operation or to disconnect will be sent based on the planned island size (which depends from the power balance situation before fault C). In case of fault F in Fig. 19.1 (see also Fig. 19.2): • HV network fault = > intentional islanding should not take place (if there is an alternative HV network supply route available), MV and LV network connected DER units should ride-through HV network faults and possibly also support the HV network according to grid code requirements (e.g., by reactive power injection). – Possible indication about HV network fault could be sent from the HV/MV transformer protection IED to MV feeder IEDs to indicate that in this case intentional islanding is not possible/allowed. Instead, FRT and HV network support of DER units is preferred. – However, if very sensitive customers (sensitive to voltage dips) are connected to MV or LV network then similar actions could take place as described earlier for the other faults.
19.3 Protection Requirements The traditional protection scheme requirements include sensitivity, selectivity, and reliability. However, the capability of a microgrid to work in an islanded mode demands the additional requirement of adaptivity for the protection scheme. Moreover, microgrid transition from the islanded mode to the grid-connected mode or even from the isolated mode to the islanded mode requires re-synchronization function to avoid wrong protection tripping during transition periods. Since the transitions from the grid-connected to the islanded mode and vice versa are mainly depen-
536
A. A. Memon et al.
dent on the connection and disconnection status of microgrid switches/breakers, therefore the selection of proper switching technology is also important. In this section, the existing and new protection requirements of a microgrid are discussed.
19.3.1 Sensitivity Sensitivity is one of the essential requirements of a protection scheme. Any protection scheme should be sensitive enough to sense and detect the abnormality or fault in the power system components and respond to clear the fault as soon as possible. Sensitivity can be defined as “the ability of the protection system to detect even the smallest faults within the protected zone” [17]. Sensitivity is related to the minimum pickup or threshold value of measured or sensed quantity which is by some margin above or below the boundary of normal operating value. Different protection methods have different sensitivity levels depending on operational characteristics or settings and the magnitude of electrical or physical quantity on which the protection scheme is working. Fuses, for example, are the traditional and simple overcurrent protection devices that depend on the magnitude of the current flowing through the fusible element which blows or melts due to thermal effects produced by the current. Fuses are both sensing and interrupting devices with inverse operational characteristics, which means less time of fusing operation at a high magnitude of current and more time of fusing operation at a low or minimum magnitude of a current flowing through the fusible element. The operational characteristics of fuses greatly depend on the material of the fusible element through which electric current flows during the fault. Therefore, operational characteristics of fuses can be changed only by changing the material of the fusible element, but once the fuse type is selected and installed the operational characteristics can no longer be changed, this makes fuse a non-resettable device. Different types of fuses have different time-current characteristics and therefore different levels of operating sensitivities. The other protection device working on the overcurrent principle is the overcurrent relay. The sensitivity of definite time overcurrent relay depends on its capability or settings to detect the lowest possible magnitude of the current flowing through the circuit during a fault. Usually, the lowest possible magnitude of the current is observed in case of single-line to ground (SLG) short circuit fault with high fault impedance and the maximum current is observed during three-phase (LLL) short circuit fault. In conventional radial distribution systems with power flow in one direction, overcurrent protection schemes for the detection of SLG faults need to be more sensitive in comparison with overcurrent protection schemes for the detection of three-phase faults. However, with the increasing connection of small-scale DER units particularly the converter-based DERs result in very reduced three-phase fault current levels which greatly affect the sensitivities of existing overcurrent relays and both definite-time and inverse-time overcurrent relays may experience the blinding of overcurrent protection. The blinding of protection scheme occurs when a fault exists but the protection scheme is unable to detect the fault due to wrong or
19 Microgrid Protection with Conventional and Adaptive Protection Schemes
537
less sensitive settings. Moreover, transient events like the starting of motor loads, the energization of transformers, and switching of capacitor banks may affect the sensitivity of overcurrent protection and result in wrong or unnecessary trips due to more sensitive settings. Additional filtering, signal processing algorithms, or even changed protection philosophies or functions can be employed to avoid these situations. Microgrids require different levels of sensitivities in overcurrent protection schemes to differentiate between faults in the grid-connected and islanded mode of operation. In the grid-connected mode, less sensitive settings are required to avoid the unnecessary trips of protection schemes, and in the islanded mode more sensitive settings are necessary to avoid the blinding of protection schemes. Other protection functions based on the measurement of voltage, frequency, or impedance may have different levels of sensitivities in different operational modes and different fault categories, and these functions should be carefully selected and configured. For example, the measurement of fault resistance (RF ) coverage is used as a means of evaluating sensitivity of directional overcurrent, distance, and differential protection for SLG faults [17].
19.3.2 Selectivity Selectivity or coordination is an important requirement of the traditional protection system which ensures that only the section of the power system close to the fault is isolated and the minimum portion of the power system is interrupted. For a complete selective or a coordinated protection scheme, primary protection operates first after fault detection inside the protection zone and backup protection operates only after primary protection fails to detect and isolate the fault after a predetermined time delay. For definite time OC relays definite time-based coordination is done which starts, for example, from the extreme load end of the feeder toward the source side of the feeder or substation. It means for faults at the extreme end of the feeder on the load side, OC relay near the load operates first and then a coordination interval of 0.2 s or so is used between each upstream OC relay toward the substation end for the backup protection of downstream relays. Usually, fast acting relays and breakers or instantaneous adjustable devices like miniature circuit breaker (MCB) or molded case circuit breakers (MCCB) are used on the load side for short circuit protection. If due to any problem, load side primary protection fails, the first upstream relay near the load will operate as a backup after a coordination interval of 0.2 s, the same is valid for all subsequent upstream faults. This definite time-based coordination is commonly used for three-phase faults in radial distribution feeders. Definite time relay coordination is simple, independent of fault current magnitude, and it provides complete coordination as long as each relay is able to detect the fault. The only drawback for definite time coordination is longer tripping times for faults near the source or substation side of the feeder. The inverse-time OC relays, commonly known as inverse definite minimum time (IDMT) relays have operating times that are inversely proportional to the magnitude of the current, the higher the fault
538
A. A. Memon et al.
current magnitude, the faster the IDMT relay operates. The IDMT relays have several families of inverse characteristic curves with different degrees of inversity like standard inverse, very inverse, and extremely inverse characteristics. For a given pickup current above the minimum pickup value and an identical time-dial setting, an extremely inverse IDMT relay operates faster than a very inverse relay and a very inverse relay operates faster than the standard inverse. The IDMT relays with an inverse or very inverse characteristics are the most commonly used types and ideally, the relays with the same inverse characteristics are used throughout the system [18]. The IDMT relays operate faster for faults near the source or substation and slower for faults away from the substation. Compared with the coordination of definite time relays, the coordination of IDMT relays is a much complicated and time intensive job, however, it all depends on the method used. The general methods used for the coordination of IDMT relays include trial and error, curve fitting, and optimization techniques using different algorithms [19]. In traditional radial distribution systems with power flow in one direction, the coordination of OC relays may not be affected, however, when a considerable amount of DERs are connected the coordination is either altered or completely lost depending on the capacity, type, and location of DERs [20, 21]. For microgrids with nearly 100% DER supply with the majority of the converter-based DERs, the loss of protection coordination will result in the reduced security of supply to consumers due to the disconnection of large portions, which is not in line with the very fundamental purpose of microgrids. Therefore, microgrid protection must be coordinated in both the grid-connected and islanded mode of operation. This could be done by the separate coordination study and settings of grid-connected and islanded mode protections or by providing sources of high fault current also in islanded mode. A case study of protection coordination in the grid-connected and islanded mode of microgrid using definite-time and inversetime OC relays is presented in Sect. 4 for further understanding about selectivity issues.
19.3.3 Reliability Reliability is the ability of a protection scheme to operate correctly and is usually defined in terms of dependability and security of relay operations. Dependability is defined as the measure of certainty that the protection relay will operate and trip for all faults for which it is designed and security is the measure of certainty that the protection relay will not operate and trip incorrectly. Traditionally, protection schemes have been designed to provide high dependability at some degree of compromise of security that may increase the false operations of protection schemes resulting in the unwanted trippings of power system elements. Traditional large interconnected power systems provide some degree of redundancy due to many alternative paths of power flow and therefore the loss of a generator or a line (n1 criterion) due to a false trip is less objectionable in comparison with the sustained fault which may damage the faulty component [17, 22, 23]. But, the false trips of
19 Microgrid Protection with Conventional and Adaptive Protection Schemes
539
a distribution line due to the unsecured protection scheme in radial, grid-connected microgrids is less acceptable because it may jeopardize the stability of microgrid. The false trips of the breaker at the microgrid connection point to the main grid may result in unwanted islanding, increased re-synchronization operations for the restoration of grid-connected operations, unwarranted outage to non-priority loads, and microgrid exposure to power quality problems [12, 24]. In an islanded mode of operation, a false trip of a grid-forming generator may result in a complete blackout due to the consequent tripping of grid-following generators. The reliability of modern multifunction numerical relays using communication links is not only dependent on hardware and software-based failures but also the dependability and security of the communication system. To analyze the dependability and security of protection schemes different fault trees can be used. Fault tree analysis is a useful tool for the comparison of the relative reliability of protection schemes. The construction of a fault tree starts with the identification of component failures which may cause a failure to trip (a dependability problem) or an unwanted trip (a security problem). The AND, OR, or other gates are used to represent the combinations of failure rates. The idea behind using the OR gate is that any of several failures can cause the protection scheme to fail, whereas, the AND gate expresses the idea that all component failures happen simultaneously to cause a protection scheme to fail. Various combinations of protection schemes using fiber-optic channels have been analyzed for dependability and security using fault trees [17]. Such type of reliability analysis will be useful for protection schemes in microgrids.
19.3.4 Adaptivity The adaptivity of the microgrid protection scheme is the new requirement that is the ability of the protection scheme to adapt its settings according to changing operational modes from the grid-connected to an islanded mode and vice versa. An adaptive protection is defined as an online activity that changes the preferable response of protection devices according to changing states of the system or its requirements. Adaptive protection is usually automated, but some necessary human interventions can also be included. An adaptive relay is a protection device or relay that includes different setting groups, characteristics, or logic functions that can be altered or changed online very quickly by using external signals or control commands [25]. The modern numerical relays, also called intelligent electronic devices (IEDs), not only provide various protection functions (overcurrent, over/under voltage, etc.) integrated into a single physical device, but also offer various settings groups for each of the available protection functions. These setting groups can be changed in an adaptive manner using the communication link between IEDs and IEDs and circuit breakers (CBs). Adaptivity of protection scheme in a microgrid is mainly required due to different magnitudes of fault current sensed by OC relays in grid-connected and islanded modes and due to the connection and disconnection of DERs. Further detailed discussion on adaptive protection is given in Sect. 5.
540
A. A. Memon et al.
19.3.5 Re-Synchronization Re-synchronization is the ability of a well-planned microgrid to reconnect back to the main grid soon after the clearance of faults on the main grid side. The availability of re-synchronization means at the microgrid point of connection to the main grid is necessary for the smooth transition of microgrid from the islanded mode to the grid-connected mode. Re-synchronization is the process of connecting islanded microgrid back to the main grid after checking or measuring the voltage, frequency, and phase angle of both systems and closing the breaker contacts for parallel operation only if these parameters are within acceptable limits as per Table 5 of IEEE Std 1547–2003 [26]. Three types of synchronization schemes have been mentioned in [27]: active, passive, and open transition synchronization. In active synchronization, there is a control mechanism that can be used to match voltage, frequency, and phase angle of an islanded microgrid to the main grid before closing the breaker contacts. Active synchronization requires collection or sensing of conditions for both the main grid and islanded microgrid and then communicating this information to the control mechanism. Passive synchronization uses traditional synchrocheck relay for closing breaker contacts if the voltage, frequency, and phase angle of both the main grid and islanded microgrid reach within specified limits. Passive technique also requires sensing of conditions for both the main grid and islanded microgrid, however, it is slower than active synchronization. Open transition synchronization requires the disconnection of loads and DGs inside the microgrid before reconnection and does not require any sensing or measurement of conditions. Both active and passive synchronization methods maintain high reliability of microgrid as no load or DG disconnection is required. The same procedure of synchronization is applicable for the reconnection of any synchronous DG, the converter-based DG, or isolated zones with one or more DGs back to the large portion of the islanded microgrid in islanded mode.
19.3.5.1
LV Microgrid Synchronized Reconnection
One important issue in enabling future Smart LV Grids with island operation capability is that challenges related to the synchronized reconnection (i.e., resynchronization) of island operated microgrid are solved. Islanded microgrid may be synchronized with the utility system directly after islanding, but later the synchronism is lost due to generation and load variations inside the microgrid. This means that the voltage phase angle difference across the microgrid interconnection switch or circuit-breaker will change. Microgrid re-synchronization or synchronized reconnection of microgrid means that the voltage angle difference between the utility grid and microgrid needs to be minimized before reconnection [4]. In the HV network where line reactance X is much larger than line resistance R, the active power P depends mainly on load angle δ and reactive power Q depends mainly on voltage difference. This means that the active power P control directly
19 Microgrid Protection with Conventional and Adaptive Protection Schemes
541
controls the load angle δ and frequency f. Generators in the HV network are typically directly connected with synchronous generators (SGs) which can be controlled, for example, during synchronization of separate power system areas. In HV network synchronism check relays have typically such settings that frequency difference over open CB needs to be less than 55 mHz and phase difference 20◦ –45◦ before CB connection [28]. On the other hand, in LV microgrids a large share of the DER units are connected through the inverter- or converter-based interfaces and microgrid synchronized reconnection can be done by the control of these DER units. In LV networks the line resistance R is much larger than the line reactance X and therefore, the active power P depends mainly on voltage difference, while the load angle δ and frequency is mainly dependent on reactive power Q. This means that one possibility to manage the phase difference across microgrid interconnection CB could be coordinated reactive power control of the DER units. The chosen strategy is dependent on the chosen microgrid concept. In [29], it was proposed that the control of the grid-forming energy storage unit (master unit) could slowly shift the microgrid frequency reference closer to the utility grid frequency before reconnection. However, for example with P/f -droop controlled DER units re-synchronization requires the coordinated control of all DER units. This coordination should be done by an external central controller, for example, microgrid management system or controller, which manages all the DG units during the synchronization process [30, 31]. Voltage imbalance due to asymmetrical loads and single-phase DG units affects the voltage phase difference across open microgrid interconnection CB so that the phase difference deviation can be different in phases A, B, and C. This asymmetry between phases may also require to be reduced before microgrid re-synchronization. Active components in the connection point of microgrids, such as microgrid interconnection switch, central energy storage unit, and microgrid management system, are responsible for microgrid synchronized reconnection. In [32] it was proposed that synchronous island operation could be done using a reference signal with phase and frequency information to the microgrid master unit. The phase difference before re-synchronization should be within acceptable levels, for example, less than 60◦ [32]. Based on [33], the microgrid re-synchronizing function has to meet a more stringent requirement than the one defined by IEEE 1547 which requires that the phase difference between a microgrid and the utility grid needs to be smaller than 20◦ before closing the interconnection CB. In [34], LV microgrid re-synchronization was studied and different synchronized reconnection enabling functionalities were developed and simulated. In the simulations [34] either 1. Master unit voltage phase angle or. 2. Reactive power output of DG units was modified to enable synchronized reconnection and in addition also. 3. Controllable single-phase loads were used in for phase asymmetry compensation at MV/LV distribution substation.
542
A. A. Memon et al.
The simulation results in [34] showed that both re-synchronization functions 1. Voltage phase angle adjustment by master unit control and. 2. DER unit reactive power feeding can be utilized to enable successful LV microgrid re-synchronization. However, the voltage phase difference deviation or asymmetry between phases A, B, and C across the microgrid interconnection switch still existed with these re-synchronization functions. If this phase difference deviation is too large, it must be compensated before LV microgrid re-synchronization. In simulations of [34], this phase difference deviation was well corrected by the connection of resistive or capacitive single-phase loads at MV/LV distribution substation. However, quite large frequency and voltage oscillations after the connection of single-phase capacitive loads were detected when compared to the connection of purely resistive loads. In general, the simulation results of [34] clearly showed that re-synchronization is not necessarily a significant issue with small, for example, 2i , ∀k ∈ Nk , p =
s=i i 2
s=i
(20.17)
Ct,f,i,j,k ≥ cdf t,f,i,j,k
total Tf,i
1 − Xf,p
− Yf,p
−
∀t ∈ Nt , f ∈ Nf , ∀i ∈ Ni , ∀j < 2i , ∀k ∈ Nk , p =
i−1 #
Xf,s −
s=2j i 2
i−1 #
Yf,s
s=2j
(20.18) total + T repair Ct,f,i,j,k ≥ cdf t,f,i,j,k Tf,i f,l ∀t ∈ Nt , f ∈ Nf , ∀i ∈ Ni , j = 2i , l = i, ∀k ∈ Nk
(20.19)
20.2.3 Simultaneous Placement of FIs and RCSs To enhance the automation ability of the MG, RCSs are installed in some sections of the feeders. They are classified as sectionalizing switches that are able to be monitored and controlled by the OMS. This capability paves the way for OMS to increase the speed of the FLISR operation either in the case of fully automated or near-automated actions. This rapidness can be achieved in fault isolation and service restoration phases [11]. After inspecting the faulty section by OMS based on the received signals from FIs, the isolation process can be remotely performed by the RCSs. This process is accelerated by remotely opening the nearest RCS to the fault in upstream of the faulty section. Then, healthy consumers in the upstream network can be restored by closing the circuit breakers which were tripped. Due to the higher cost of these switches in comparison with manual switches (MSs), the suitable analysis should be conducted to evaluate the value of the investment cost of them for improving the reliability indices. The main feature of the RCSs is that they have the functionality of both FI and switch as the fault indicator and fault isolator, respectively. This capability makes RCSs comparable to FIs and MSs. Following the model for FI placement in the previous section, authors in [12] proposed a model for simultaneous placement of the FIs, RCSs, and MSs in an MG. This model aims to reduce the fault locating and isolation time by appropriate placement of FIs and RCSs besides reducing the restoration time by appropriate placement of MS and RCS. Therefore, the RCSs are helpful in both phases. Here, the previous model is developed to include the sectionalizing switches. The model is demonstrated by (20.20)–(20.35).
590
A. Yazdaninejadi et al.
C inst =
Minimize C inst + C main + C int
(20.20)
FI RCS MS MS Xf,s CI Ff,sI + Xf,s CI RCS f,s + Xf,s CI f,s
(20.21)
f ∈Nf s∈Ns
C main =
t∈Nt f ∈Nf s∈Ns
1 FI FI RCS RCS MS MS X MC + X MC + X MC f,s f,s f,s f,s f,s f,s (1 + DR)t (20.22)
C int =
t∈Nt f ∈Nf i∈Ni j ∈Nj k∈Nk
1 λt,f,i Lot,f,j,k Ct,f,i,j,k (1 + DR)t
Ct,f,i,j,k = cdf t,f,i,j,k (T )
ε−
1 1 bf,i,l ≤ Lf,i,l ≤ 1 − bf,i,l + ε ζ ζ
bf,i,l =
l−1
FI Xf,s +
s=i
bf,i,l =
l−1 s=i
l−1
(20.25)
(20.26)
RCS Xf,s ∀i > l
(20.27)
s=i
FI Xf,s +
l−1 s=i
Lef,l pre Lf,i,l + Tf ∀f ∈ Nf , ∀i ∈ Ni Vf,l
l∈Nl
(20.24)
RCS Xf,s ∀i < l
bf,i,l = 1 ∀i = 1
loc Tf,i =
(20.23)
(20.28)
(20.29)
20 Fault Identification, Protection Schemes, and Restoration Requirements. . .
591
pre
where Tf , cdft, f, i, j, k are the preparation time for the crew to reach the faulty section of the feeder and Customer damage function, respectively [13]. aut Ct,f,i,j,k ≥ cdf t,f,i,j,k Tf,i,j ∀t ∈ Nt , f ∈ Nf , ∀i ∈ Ni , ∀j ∈ Nj , ∀k ∈ Nk (20.30)
loc Ct,f,i,j,k ≥ cdf t,f,i,j,k Tf,i
1−
s=i
∀t ∈ Nt , f ∈ Nf , ∀i ∈ Ni , ∀j >
Ct,f,i,j,k
2j# −1
i 2 , ∀k
RCS Xf,s
(20.31)
∈ Nk
i−1 # RCS loc ≥ cdf t,f,i,j,k Tf,i Xf,s 1− s=2j
(20.32)
∀t ∈ Nt , f ∈ Nf , ∀i ∈ Ni , ∀j < 2i , ∀k ∈ Nk
Ct,f,i,j,k
2j# −1 rep loc RCS MS Xf,s + Xf,s ≥ cdf t,f,i,j,k Tf,i + Tf,l 1− s=i
∀t ∈ Nt , f ∈ Nf , ∀i ∈ Ni , j >
Ct,f,i,j,k
i 2, l
= i, ∀k ∈ Nk
i−1 # rep loc + T RCS + X MS Xf,s ≥ cdf t,f,i,j,k Tf,i 1− f,l f,s s=2j
∀t ∈ Nt , f ∈ Nf , ∀i ∈ Ni , j
CTI CTI
CTI
>CTI
t min
Distance R B Fault
RC
Fault
Fault
Fault
Fig. 20.15 Simple example of coordination tasks with double-inverse characteristic
⎛ do,f t2
=
T DS do 2
⎞
⎜ A ⎜ ⎜ ⎝ I F do,f B Ip2do
−1
⎟ min 0, I F do,F 3 − I F do,f ⎟ + C⎟ × ; ⎠ I F do,F 3 − I F do,f (20.41)
do,f
t do,f = t1
do,f
+ t2
;
(20.42)
606
A. Yazdaninejadi et al.
where tdo, f reveals the operation time of overcurrent relays with double-inverse characteristics. Moreover, IFdo, F3 is the current seen by relay for fault at F3. When do,f the fault current IFdo, f is more than IFdo, F3 , t1 is zero and when the current is do,f is zero. In this manner, the two inverse curves intersect at less than IFdo, F3 , t2 fault point F3 which is presented in Fig. 20.14. Normally, the operation time of an overcurrent relay is formulated by the first term of Eq. (20.40). The second term in this equation acts as a binary variable. When the fault current is bigger than the current at F3, it is zero. Likewise, for the other faults, it is one. The second term in Eq. (20.41) works in the opposite. Due to providing faster protection schemes, double-inverse characteristics are deployed in [22] for meeting the transient stability of distributed generation. Each section of a multi-inverse characteristic is controlled by separate and independent variables. Hence, the number of coordination variables and coordination points are increased which may lead to a complicated optimization problem. Hence, a proper process is required to find optimal set points to determine the best number of inverse curves in a multi-inverse characteristic. Another key point to notice is the multi-inverse characteristics that should demonstrate the decreasing fashion. Therefore, the proper constraint should be considered in intersection points like F3.
20.3.3.3
Overcurrent Protection with Multi-Function Characteristic
A DOCR detects the direction of faults and operates just for faults in the forward directions. These relays are blocked for the faults in reverse directions. This is while; these relays can be developed with another characteristic in reverse directions of DOCRs as shown in Fig. 20.16. By doing so, each DOCR can act as two independent DOCRs in different directions. This feature is realized based on programming code on numerical DOCRs. These multi-function relay with two independent characteristics are called dual-setting DOCR [23, 24].
Normal inverse Very inverse Extremely inverse
Time (s)
Reverse direction
Forward direction
Multiple of pickup current Fig. 20.16 Dual-setting directional overcurrent relays characteristics
20 Fault Identification, Protection Schemes, and Restoration Requirements. . . Primary Substation
607
DOCR
Protection area HV/MV Transformer
R3
R2
R1
Line A
Fault
(a)
Synchronous-based DER
Dual-setting DOCR
Primary Substation
Low bandwidth communication link
Protection area HV/MV Transformer
R3
R2
R1
Line A Fault
Synchronous-based DER
(b) Fig. 20.17 Protection scheme based on (a): conventional, and (b): dual-setting characteristics
Application of these relays increases the number of protective relays. Therefore, satisfying CTI in the coordination process leads to a complicated problem. This is while; the reverse side characteristic of these relays can be employed instead of backup protection for the primary relay in the same direction. In this case, the conventional backups can be blocked through the communication links which helps to relax the conventional selectivity constraints and hence, alleviates the complexity of the coordination process. Figure 20.17 illustrates the protection scheme with conventional and dual-setting DOCRs. In the case of a conventional scheme, for a fault at Line A, R1 is the primary relay which is backed up with R3. The concept of this scheme is depicted in Fig. 20.17a. However, in the case of employing a multi-function relay and for the same fault, the forward direction of R1 is the primary protection which is backed up with the reverse direction of R2. In this case, the forward direction of R3 as the conventional backup is blocked through the communication link which is shown in Fig. 20.17b. Deploying multi-function relays helps to diminish the tripping time of relays. Therefore, dual-setting schemes
608
A. Yazdaninejadi et al.
Time (s) Reverse direction
Forward direction
Fig. 20.18 Dual-setting DOCRs with double-inverse characteristics
are employed for meeting transient stability requirements of MGs in [20]. As shown in Fig. 20.18, the aforementioned scheme can be developed by multi-inverse characteristics to provide further alleviation in the optimization complexity and hence reduces operation time of relays. It is worth noting that forward direction and reverse direction characteristics of dual-setting relays acts irrespective of each other. Consequently, the number of inverse curves in the forward direction can be different from the reverse direction.
20.3.3.4
Time-Current-Voltage Characteristics
As discussed earlier, the conventional overcurrent relaying operates based on conventional inverse-current characteristics. According to [25], microprocessorbased DOCRs facilities devising new tripping characteristics based on voltage index. Typically, a DOCR determines directions of the occurred faults by using the phase of the voltage and current. In this regard, in addition to the three existing current transformers; these relays also deploy three voltage transformers to measure voltage and current of three-phases [25]. Therefore, based on these relays, it is possible to deploy a voltage index in tripping characteristics. In the developed characteristics, by increasing voltage drop, the operation time of relays should be decreased. Therefore, the conventional characteristic should be multiplied to decreasing function as follow: t=
1 e1−V
K ×
T DS × A B −1
If ault Ip
(20.43)
where the first term reduces the operation time of relays with an increasing voltage drop. Here, V is the bus voltage and K is a constant coefficient. This function is called time-current-voltage characteristic and is plotted in Fig. 20.19. As can be seen, the voltage index decreases the characteristic by keeping inverse behavior
20 Fault Identification, Protection Schemes, and Restoration Requirements. . .
609
1.8 1.6
Operation time (s)
1.4 1.2 1.0 0.8 0.6 0.4 0.2 0 0
10
20
1
0.8
0.6
0.4
0.2
0
Vf (p.u)
Fig. 20.19 Time-current-voltage characteristics
which facilities the way for using this characteristic in the coordination process. The conducted study in [26] shows that this characteristic offers a suitable scheme for meeting the FRT requirement.
20.3.3.5
Dual Time–Current–Voltage Characteristics
As explained in previous section, time-inverse characteristics can be developed based on numerical relays as a multi-function scheme like the dual-setting approach. In this manner, time-current-voltage characteristics of numerical DOCRs can be also developed as a dual-setting approach with the following formulations: tf w =
K
e1−V
trv =
1
1 e1−V
K
T DSf w × A × If ault B −1 Ipf w
(20.44)
T DSrv × A × If ault B −1 Iprv
(20.45)
where fw and rv indicate forward and reverse directions, respectively. Deploying dual characteristics and communication links helps to relax the conventional constraints in the coordination process which yields fast response protection.
610
A. Yazdaninejadi et al.
VI
EI
Time
Fig. 20.20 Simple piece-wise linear characteristic [27]
NI
Piece-wise linear
Multiple of pickup current
The developed scheme in [26] based on dual time-current-voltage characteristics considers FRT requirements in transmission level interconnected wind parks. The proposed scheme not only reduces the operation time of relays but also eliminates miscoordinations. These schemes can be also enhanced by a user-defined strategy. To do so, in addition to TDS and Ip; the other coefficients A, B, and K should be also considered in the optimization process as new variables.
20.3.3.6
Piece-Wise Linear Characteristic
The typical characteristics for overcurrent relaying are standard (normal) inverse characteristic, very inverse characteristic, and extremely inverse characteristic. However, numerical relays allow the user to follow arbitrary characteristics. This arbitrary characteristic can include some linear lines that are so-called piece-wise linear characteristic. This feature helps to obtain a characteristic with a positive property of standard inverse, very inverse, and extremely inverse characteristics like the characteristic that is portrayed in Fig. 20.20. This figure shows that this kind of characteristic helps to reduce the operation time of relays. These characteristics are suitable for radial networks [27]. As can be seen, the presented characteristic in Fig. 20.20 is piece-wised on the fault currents. Moreover, fault currents belong to parameters set. Therefore, this characteristic cannot be used in the coordination process of looped MG as a linear approach and suitable optimization model would be of interest.
20 Fault Identification, Protection Schemes, and Restoration Requirements. . .
611
20.4 Reconfigurable Topology of MG and Enhancing the Restoration Capability The FLISR task of the OMS ensures a fast fault locating, isolating, and restoration phases. However, it may be possible to restore more interrupted consumers by some switching actions, until the repair operations are accomplished. For this task, it is necessary to consider the technical and operational constraints of the MG. Efficient feeder reconfiguration of the MG which ensures the operation constraints like feeder congestion, voltage security, and radiality of the grid, can enhance the restoration capability of the OMS. In this field, the linear model proposed by the authors in [28, 29], is developed to make it compatible with the restoration requirement of the grid. The final model is represented as below: Minimize C EN S = K EN S
NS Pi,t
(20.46)
i∈B t∈T
in NS D + Pi,t − Pi,t = Pi,t
L Pi,j,t
(20.47)
QL i,j,t
(20.48)
j ∈B,j =i
NS D Qin i,t + Qi,t − Qi,t =
j ∈B,j =i
L Pi,j,t = −Gij 2Vi,t − 1 + Bij θij,t + Gij Vi,t + Vj,t + ωi,j,t − 2 ξi,j,t + γi,j,t (20.49) QL i,j,t = Bij 2Vi,t − 1 + Gij θij,t − Bij Vi,t + Vj,t + ωi,j,t − 2 ξi,j,t + γi,j,t (20.50)
θij,t = −θj i,t
(20.51)
−M 1 − ξi,j,t ≤ γi,j,t ≤ M 1 − ξi,j,t
(20.52)
612
A. Yazdaninejadi et al.
−Mξi,j,t ≤ γi,j,t −
1 Vj,t − Vi,t + 2 ≤ Mξi,j,t 2
−Mξi,j,t ≤ θij,t ≤ Mξi,j,t
ξi,j,t ≤ 2 N Bus − 1 ∀t
(20.53)
(20.54)
(20.55)
i,j ∈B,i=j
ξi,j,t = ξj,i,t
ξi,j,t = 0 ∀ (i, t) | i ∈ F B, t < t rep
(20.57)
(20.58)
j ∈B,j =i
The connectivity of a section which can be controlled by the RCSs is modeled by the binary variable ξ i, j, t . In the last equation, the set FB, are the adjacent buses of the faulty feeder section which should be remained de-energized until the repair process of this section is completely accomplished in trep . This model can be further developed to include other reliability indices, too. The operation constraints that should be considered in the model are as follow: V ≤ Vi,t ≤ V
(20.59)
G2 + B 2 i,j i,j sqr L L + Pj,i,t Ii,j,t = Pi,j,t Gi,j
(20.60)
2 sqr Ii,j,t ≤ I i,j
(20.61)
Note that the linearity of the model is preserved.
20.5 Conclusion This chapter tailored the fault identification, protection requirements, and restoration of MGs. In this way, the essence of meeting transient stability constraints and FRT requirements by protection systems besides the necessity of enhancing
20 Fault Identification, Protection Schemes, and Restoration Requirements. . .
613
fault detection and isolations solutions were explored. It was shown that meeting transient stability and FRT of DERs paves the way for preventing the unintentional disconnection of DERs which calls the need for providing selective and fastresponse protection schemes. In this regard, the main issues were introduced, some suitable solutions were reported, and in particular, a general research line on eminent protection issues was established. Moreover, it was shown that conventional protection devices besides equipment for FLSRT can be jointly used for fast restoration of MGs, fast restoration of DERs, improving equipment health, and enhancing the reliability of MG. Thereby, the developed model of fault detection and isolation techniques along with conventional protections were also discussed.
References 1. Martins, V. F., & Borges, C. L. (2011). Active distribution network integrated planning incorporating distributed generation and load response uncertainties. IEEE Transactions on Power Systems, 26(4), 2164–2172. 2. Le, D., Bui, D., Ngo, C., & Le, A. (2018). FLISR approach for smart distribution networks using E-Terra software—A case study. Energies, 11(12), 3333. 3. Andishgar, M. H., Fereidunian, A., & Lesani, H. (2016). Healer reinforcement for smart grid using discrete event models of FLISR in distribution automation. Journal of Intelligent & Fuzzy Systems, 30(5), 2939–2951. 4. Dimitrijevic, S., & Rajakovic, N. (2015). Service restoration of distribution networks considering switching operation costs and actual status of the switching equipment. IEEE Transactions on Smart Grid, 6(3), 1227–1232. 5. Abniki, H., Taghvaei, S. M., & Mohammadi-Hosseininejad, S. M. (2019). Reliability improvement in smart grid through incorporating energy storage systems in service restoration. International Transactions on Electrical Energy Systems, 29(1), e2661. 6. Reliability improvement form the application of distribution automation technologies- initial results, Available www.smartgrid.gov. 7. Teng, J. H., Huang, W. H., & Luan, S. W. (2014). Automatic and fast faulted line-section location method for distribution systems based on fault indicators. IEEE Transactions on Power Systems, 29(4), 1653–1662. 8. Lwin, M., Guo, J., Dimitrov, N., & Santoso, S. (2018). Protective device and switch allocation for reliability optimization with distributed generators. IEEE Transactions on Sustainable Energy, 10(1), 449–458. 9. Fault location, isolation, and service restoration technologies reduce outage impact and duration, Available www.smartgrid.gov. 10. Farajollahi, M., Fotuhi-Firuzabad, M., & Safdarian, A. (2016). Deployment of fault indicator in distribution networks: A MIP-based approach. IEEE Transactions on Smart Grid, 9(3), 2259– 2267. 11. Izadi, M., & Safdarian, A. (2018). A MIP model for risk constrained switch placement in distribution networks. IEEE Transactions on Smart Grid. 12. Farajollahi, M., Fotuhi-Firuzabad, M., & Safdarian, A. (2018). Simultaneous placement of fault indicator and sectionalizing switch in distribution networks. IEEE Transactions on Smart Grid, 10(2), 2278–2287. 13. Billinton, R., & Wangdee, W. (2005). Approximate methods for event-based customer interruption cost evaluation. IEEE Transactions on Power Systems, 20(2), 1103–1110. 14. Teimourzadeh, S., Davarpanah, M., Aminifar, F., & Shahidehpour, M. (2016). An adaptive auto-reclosing scheme to preserve transient stability of microgrids. IEEE Transactions on Smart Grid, 9(4), 2638–2646.
614
A. Yazdaninejadi et al.
15. Yazdaninejadi, A., Nazarpour, D., & Golshannavaz, S. (2017). Dual-setting directional overcurrent relays: An optimal coordination in multiple source meshed distribution networks. International Journal of Electrical Power & Energy Systems, 86, 163–176. 16. IEEE. (2000). IEEE recommended practice for utility interface of photovoltaic (PV) systems. IEEE. 17. Papaspiliotopoulos, V. A., Korres, G. N., Kleftakis, V. A., & Hatziargyriou, N. D. (2015). Hardware-in-the-loop design and optimal setting of adaptive protection schemes for distribution systems with distributed generation. IEEE Transactions on Power Delivery, 32(1), 393–400. 18. Hooshyar, A., & Iravani, R. (2017). Microgrid protection. Proceedings of the IEEE, 105(7), 1332–1353. 19. Jia, K., Chen, J., Xuan, Z., Wang, C., & Bi, T. (2019). Active protection for photovoltaic DCboosting integration system during FRT. IET Generation, Transmission & Distribution, 13(18), 4081–4088. 20. Yazdaninejadi, A., Nazarpour, D., & Talavat, V. (2018). Optimal coordination of dual-setting directional over-current relays in multi-source meshed active distribution networks considering transient stability. IET Generation, Transmission & Distribution, 13(2), 157–170. 21. Aghdam, T. S., Karegar, H. K., & Zeineldin, H. H. (2018). Variable tripping time differential protection for microgrids considering DG stability. IEEE Transactions on Smart Grid, 10(3), 2407–2415. 22. Yazdaninejadi, A., Nazarpour, D., & Talavat, V. (2019). Coordination of mixed distance and directional overcurrent relays: Miscoordination elimination by utilizing dual characteristics for DOCR s. International Transactions on Electrical Energy Systems, 29(3). 23. Sharaf, H. M., Zeineldin, H. H., & El-Saadany, E. (2016). Protection coordination for microgrids with grid-connected and islanded capabilities using communication assisted dual setting directional overcurrent relays. IEEE Transactions on Smart Grid, 9(1), 143–151. 24. Yazdaninejadi, A., Golshannavaz, S., Nazarpour, D., Teimourzadeh, S., & Aminifar, F. (2018). Dual-setting directional overcurrent relays for protecting automated distribution networks. IEEE Transactions on Industrial Informatics, 15(2), 730–740. 25. Saleh, K. A., Zeineldin, H. H., Al-Hinai, A., & El-Saadany, E. F. (2014). Optimal coordination of directional overcurrent relays using a new time–current–voltage characteristic. IEEE Transactions on Power Delivery, 30(2), 537–544. 26. Saleh, K. A., El Moursi, M. S., & Zeineldin, H. H. (2015). A new protection scheme considering fault ride through requirements for transmission level interconnected wind parks. IEEE Transactions on Industrial Informatics, 11(6), 1324–1333. 27. Ojaghi, M., & Ghahremani, R. (2016). Piece–wise linear characteristic for coordinating numerical overcurrent relays. IEEE Transactions on Power Delivery, 32(1), 145–151. 28. Hamidi, A., Golshannavaz, S., & Nazarpour, D. (2017). D-FACTS cooperation in renewable integrated microgrids: A linear multiobjective approach. IEEE Transactions on Sustainable Energy, 10(1), 355–363. 29. Jianfeng, D. A. I., Yi, T. A. N. G., Yuqian, L. I. U., Jia, N. I. N. G., Qi, W. A. N. G., Ninghui, Z. H. U., & Jingbo, Z. H. A. O. (2019). Optimal configuration of distributed power flow controller to enhance system loadability via mixed integer linear programming. Journal of Modern Power Systems and Clean Energy, 7(6), 1484–1494.
Chapter 21
Real-Time Testing of Microgrids A. S. Vijay and Suryanarayana Doolla
21.1 Introduction Power electronic (PE) converters and equipment used for the conversion, control, and conditioning of electric power are ubiquitous in our everyday lives. PE converters are also increasingly being used in the integration of renewable sources [1] and in electric vehicles [2]. The advancements in semiconductor technology and microelectronics/ computational capability have been the primary cause for this. In today’s competitive market scenario, designers of such equipment need to ensure that the equipment are rigorously tested before field deployment, to meet various performance standards, at the same time ensuring minimum time to market the products. Therefore, the increasing need arises for evolving novel techniques for testing, which has been actively taken up by various research labs around the world. The testing methodology developed has to provide accurate and reproducible results for the complex equipment being tested, and yet at the same time should be economical, scalable, and easy to implement. Safety, testing time, and space requirements typically pose constraints and demand careful consideration as they may be in conflict with each other [3]. Important aspect to the testing process is the degree of detail of the model of the system under test: the model should be detailed enough so that the dynamic metrics of interest are captured accurately through the testing process, and at the same time be simple enough to enable quick testing and meet real-time (RT) constraints. The very nature of PE converters makes them nonlinear due to the switching of semiconductor devices and this poses enormous challenges in their testing. The modern distribution system faces the increasing penetration of these low inertia and nonlinear PE equipment, which
A. S. Vijay · S. Doolla () Department of Energy Science and Engineering, IIT Bombay, Mumbai, India e-mail: [email protected] © Springer Nature Switzerland AG 2021 A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems, https://doi.org/10.1007/978-3-030-59750-4_21
615
616
A. S. Vijay and S. Doolla
makes the modeling, design, and testing of these systems a nontrivial task. The smart active distribution networks and microgrids are undergoing major technological transformations, and in this context, the system level interactions between the various PE converters and the controllers and the communication networks used for their management pose uphill challenges for their modeling and testing. The typical stages in the design and validation of a system involve: (1) stating down the objectives/requirements, (2) analysis and system design, (3) modeling, (4) prototyping, (5) testing, (6) validation, and then (7) iterations depending on the results. The testing of the product development can be done at various stages of the process depending on the complexity and the degree of detail required. This chapter will cover the topics related to the testing of microgrids. Section. 2 will outline the various RT testing methods and the categories of simulations, while Sect. 3 highlights the concepts in digital RT testing. Sections. 4 and 5 elaborate on the popular real-time emulation and HIL techniques. Section 6 presents the novel state of the art and hybrid testing approaches which are becoming increasingly popular and Sect. 7 presents the challenges and conclusions in the aspects of microgrid testing.
21.2 Real-Time Testing Methods Approaches for testing power electronic equipment include simulations (off-line, online, and real-time), model in the loop (MIL), processor in the loop (PIL), rapid controller prototyping, software in the loop (SIL), hardware in the loop (HIL)—both controller HIL (CHIL) and power HIL (PHIL), RT power level emulation, hardware test beds (HTBs) or platforms and newer hybrid or combined approaches. Of these, off-line simulations, MIL, PIL, and SIL are purely software-based approaches and typically based on a fixed time step and need not be carried out in RT. Online simulators are transient stability simulators that typically interact and exchange data with an online system (such as supervisory control and data acquisition (SCADA), energy management systems (EMS), state estimators, etc.) and these may also participate in decision-making process and control the real/physical electrical network. The integration of hardware in the testing process demands a RT simulation. For large systems such as smart distribution networks, off-line digital simulations are commonly performed to verify the performance. Previously large analog simulators were built for carrying out tasks and verifying the system performance. The rapid growth of computational power and increase in processing speed [4] had given rise to digital simulation which made data analysis, storage, and processing extremely convenient. Off-line digital simulation is carried out through an iterative process of solving the differential equations of the system through suitable numerical integration techniques. Typically, this is implemented by the use of a high-level language-based code on a computer. Normally, single processors were sufficient to carry out the task, but with the increasingly complex systems
21 Real-Time Testing of Microgrids
617
of today, newer software having multiple processors (cores) make possible faster testing and provide graphical user interface (GUI) blocks to assemble the system components and carry out the simulation task. Any approach to testing involves the modeling of the system as the first step. Model-based design (MBD) is a popular approach in the industry which prioritizes the model of the system in the design process. Model-based Systems Engineering (MBSE) and simulations are becoming increasingly popular to cope with more and more complex systems. Off-line simulation permits the modeling to be carried out for any depth of detail, but at the expense of the computational burden and simulation time. This is proven to be the cheapest. Issues related to safety are least concern in this type. This time step (fixed or variable) is very critical and depends on objective of the simulation [5]. Also critical is the integration method. The time required to simulate larger systems is very high. The time taken by the processor to completely solve one integration instance is known as the simulation step time or execution time (or the time step). In off-line simulation, the execution time is more than the numerical integration time step [6]. In the context of PE converters, capturing the fast switching transients demands appropriate modeling of the semiconductor devices and needs smaller time steps for the simulation (of the order of nanoseconds) which can be achieved with modern field programmable gate arrays (FPGAs). Typically, off-line simulations form the first step in the design and development stage. The next step in the testing process would be involving hardware in the testing process. But this requires the simulation to be carried out in real-time (RT), as the hardware has to interact with the simulation as if it were in the actual world. Thus, RT simulation tries to capture real-world behavior in the process of testing. Realworld time constraints impose stringent conditions on the time sequence of events and demand the use of a simpler model (lesser degree of detail) giving rise to a tradeoff between accuracy and hardware integration. In RT simulation, the time taken for execution is less than the integration step size. RT simulations can be classified into three types depending on the applications: (a) Hard RT applications, which use a fixed time step since the real-time constraint must strictly be met in every simulation time step: cases in which the hardware integration interface sampling period is comparable (small) to the minimum simulation time step, (b) Firm RT applications, wherein a variable simulation time step can be used with restriction: cases in which the interface sampling period is larger than the minimum simulation time step, yet lower than a certain value (such as 1 ms), and (c) Soft RT applications which employ variable time-stepping fully (minimum restrictions): cases with no hardware interfacing (or a very large interface sampling period). In digital real-time simulators (DRTS), a large system is broken down into smaller subsystems and simulation is performed (simultaneous) on parallel processors. This enables the accurate replication of the time scale of events [7]. The digital real-time simulation can be classified into two main categories: (1) a fully software based DRTS and (2) HIL RT simulation or HILS. Figure 21.1 portrays the various types of simulation approaches. In some studies, the EUT is required to be connected to
618
A. S. Vijay and S. Doolla
Fig. 21.1 Classification (types) of simulations
electric drive system or a distribution network making. These are very large systems (especially the distribution network). Since this may not be practically realizable due to economic and space constraints, the sections of hardware are connected to much more complex systems by the RT simulation of these systems. HILS can be categorized into CHIL (or software HIL (SHIL)), PHIL, and mechanical HIL (used in the context of electric drives). The difference between CHIL and PHIL approaches is primarily on the voltage and current values at the interface. This interface is for the RT simulation and EUT. In CHIL, the power level at the interface is very small (in the order of mW to W), where as in PHIL the power levels can easily be in the range of tens to even hundreds of kW. PHIL makes possible the easy and flexible realization experiments which may otherwise be risky, difficult, and expensive. Figure 21.2 shows interconnection of various systems in HILS. The next stage in the development process involves the experimentation on a hardware prototype of the system. The prototyping stage gives conclusive test results. This prototyping stage serves as final validation for the system under consideration. For prototype testing, test conditions are close to the real-world conditions. Table 21.1 summarizes the comparative analysis of the various testing techniques against different metrics such as flexibility, cost, ease of implementation, degree of detail, safety, stability, and space requirements. RT emulation holds the promise for providing the realistic environments for testing systems. An emulator consists of a PE converter, which is controlled such that it mimics a desired source or load. Since the hardware is fixed and the software (model) is completely reconfigurable, it is possible to study the effect of aging components like sensors, noise issues, tolerances, etc. This is not possible with
21 Real-Time Testing of Microgrids
619
Fig. 21.2 Concept of HILS: hardware (device or equipment) interfaced to an RT simulator
Table 21.1 A comparison of various testing approaches [1] Parameter Degree of detail Ease of implementation Flexibility/scalability Safety Stability Space requirements Cost
Off-line simulation Highest (I) Easiest (I) Highest (I) Safest (I) High (I) Least (I) Least (I)
RT simulation Medium (IV) Easy (II) Medium (III) Safe (II) Medium (II) Less (III) Medium (II)
HIL High (III) Medium (III) Medium (III) Medium (III) Least (IV) Less (III) Medium (II)
Emulation Higher (II) Tough High (II) Medium (III) Medium (II) Medium (II) High (III)
simulation unless we model them in detail. For testing large systems, parallel use of small emulators is a good option. Hardware test benches and a combination of the abovementioned testing techniques (hybrid approaches) have also been reported in the literature for modeling testing and analyzing the control algorithms and complexities associated with microgrid systems. Yet another way of classifying simulations would be as multi-domain and cooperative/co-simulation [8]. The final stage of testing could involve a fully functioning interactive virtual prototype, which allows for an effective testing, and verification on both the component and system levels and may even involve virtual test beds (VTB) based platforms.
620
A. S. Vijay and S. Doolla
21.3 Digital Real-Time Testing: Concept In off-line simulations, detailed modeling of each and every component of the system is possible and hence the output is highly accurate. It is known to consume more time if the system is large and complex. Real-time simulation with hardware uses a reduced model. Thus, there exists a tradeoff between the simulation accuracy and run time. With today’s state-of-the-art processors it is possible to create a balance between these two aspects [9]. The time scale of the simulation is crucial to preserve the dynamic characteristics of the system. The difference between the two types of simulations: off-line and RT, is highlighted in Fig. 21.3. For the off-line
Fig. 21.3 Types of simulation. (a) Off-line accelerated simulation (faster than real-time), (b) offline simulation (slower than real-time), and (c) RT simulation
21 Real-Time Testing of Microgrids
621
simulation case, the time for execution for each integration step can either be lesser or (but is normally) greater than the integration time (RT clock or wall clock time) [6]. Respectively these are either faster than RT simulation—known as accelerated simulation, or slower than RT simulation as illustrated in Fig. 21.3a, b. As can be seen from the figures, idle time and over runs occur in the faster and slower cases, respectively [10]. The RT simulator should reproduce the internal variables and outputs of the simulation accurately, within the same duration of time that the actual physical system would do. This is a necessary condition if we want an RT simulation to be valid. Therefore, for RT simulations the execution time should always be lesser than the integration time step and hence the processor will always have some idle time as is shown in Fig. 21.3c. This time permits the processor to perform all necessary operations such as driving I/O to and from externally interfaced equipment and devices. The RT simulator waits until the clock ticks to the next time step to start the computation, as against accelerated simulation in which the idle time is used to compute the states at the next time step. Hence, RT simulation aims to replicate the dynamics of the simulated system at the natural time scale of occurrence of the events. However, due to the presence of nonlinear events (such as switching) in the RT simulation, numerical instability occurs. This can be attributed to the nature of the discrete time step solvers. As we all know successfully bringing about RT synchronously is an uphill task. The real-time simulators available in the market come with advanced I/O cards. The sampling rates of these system are much higher than the fixed-step simulations. This helps in overcoming the challenges. For a detailed comparison on various commercial RT platforms one can refer to [7]. RT simulation platforms are commonly used today to obtain closer to real-world results for the designed systems. In the context of PE applications and power systems, smart grid technologies and its development has been summarized in [11], which highlights that the way to address the challenges is through RT testing, probably involving the combination of HILS and RT simulations. Due to the immense computing power of today’s DSCs (digital signal controllers) and FPGAs (field programmable gate arrays), computationally intensive studies involving nonlinearities and magnetic saturation, etc. in the area of PE and electric machines have also been carried out [12].
21.4 Hardware in the Loop Testing A hardware interfaced with real-time simulator is termed as hardware in the loop simulation (HILS). CHIL, PHIL, and MHIL are three classifications in HIL. The controller hardware in the loop simulation involves the testing of the controller board while the plant is modeled and simulated in the real-time simulator. Thus, only signal level power exists at the interface (feedback signals from the plant and command signals from the processor—which complete the loop). In the past, HILS was mainly used as a testing method for electrical drives in the aircraft industry.
622
A. S. Vijay and S. Doolla
Fig. 21.4 CHIL and PHIL: (a) signal level exchange in CHIL, (b) power level exchange through the interface in PHIL
Nevertheless, now it has become an industry practice for power electronic and power system apparatus testing. HILS has become a popular approach in microgrids too [13]. A detailed comparison of simulation and HIL alternatives is presented in [14]. The work emphasizes that in digital simulations, the simulation run time is the main constraint. This is critical for systems involving high-frequency switching. These systems demand use of small time steps, thus increasing the computational burden on the processor. PHIL involves actual power transfer at the interface through a power electronic converter, unlike its counterpart (CHIL). The differences between the two types of HILS are as indicated in Figure 21.4. The RT simulator consists of the simulation of the rest of system (ROS) which may have a human machine interface (HMI) as well. Employing PHIL would reduce the cost and late changes to design specifications [3]. It is because realistic and detailed testing environments can be provided through PHIL, which can permit real-world scenario studies before the equipment is installed in the field. The details of various applications, including high-speed generator testing and fault current limiter testing, are provide in the work. Also available are experimental applications pertaining to ship propulsion and rotor heating in motors (superconducting). All these applications use the PHIL approach. Research presented in [15] summarizes the modeling and simulation aspects for PHIL-based testing. The issue of stability in PHIL testing and the relevant analysis of the same has been discussed widely in the literature. The ideal interface will have unity gain, infinite bandwidth, zero delay, and will be linear over the complete range of
21 Real-Time Testing of Microgrids
623
inputs. The methodology or approach setup for the exchange of signals between the EUT and the RT simulator (or through the interface) is known as the interface algorithm (IA). Two popular algorithms typically used in PHIL are compared in terms of the accuracy and stability in [16]. Attempts have been made to address the stability issues in the PHIL systems using virtual impedance implementation in software [17], impedance measurement using online wideband system [18], etc. Stability issues arising out of the type of interface (voltage or current) are discussed extensively in [19]. Issues arising out of sensor noises and bandwidth time delay are also highlighted. Other known issues of numerical instability caused by splitting of system into multiple subsystems for simulation are discussed. The impedances of the equipment under test and the virtual aspects of it play a critical role in determining the stability.
21.5 Real-Time Emulation The block diagram of a general-purpose real-time emulator is shown in Fig. 21.5. The power electronic converter or emulator converter presents itself as source or load to rest of the system. The load emulation is done by current generation through the real-time simulation in closed loop control while source emulation
Fig. 21.5 Generalized emulator (source/load)
624
A. S. Vijay and S. Doolla
is through voltage generation. The characteristics of the required load or source are implemented through differential /difference equations, which describe the mathematical model of the particular entity. Load emulation involves the tracking of currents while its dual-source emulation involves the tracking of voltages. Emulators have predominantly been used for the testing of individual components like sources, loads, impedances, etc.; however recent applications in PE testing include converter impedance control emulation, energy storage emulation, and even the emulation of nonlinearities in the system [20–22]. Separate load and source emulation is very well known and accepted by the industry. A new area of system emulation through power electronic interface has gained recent attention. For microgrid and distribution system studies, having a hierarchical system in laboratory scale is very challenging. Same is the case with distributed controls. It is almost impossible to put together all of them on a single system. If we like to reproduce the hierarchical layers of the control scheme, proper emulation methods need to be employed which may not be easily achieved on off-line simulators. A detailed review of renewable energy system emulators is reported in [23]. The work presented in [24] reports a microgrid system emulator based on a single bidirectional converter. Another recent area where emulation is being widely used is in the testing of the drive trains of electric vehicles [25, 26].
21.6 Novel/Hybrid Testing Approaches A combination of real-time, PHIL, and emulation approaches to enhance the testing performance is gaining attention of both industry and academia. Software in the loop (SIL) is another approach which refers to the software-only simulation of two or more subsystems on the same RT simulator. Closed loop simulation of two or more subsystems on different real-time simulation platforms is known as Co-operative simulation or Co- Simulation [27]. It refers to the case in which the subsystems (of a larger complex system) with different modeling and run-time environments are solved using a coordinated approach. This technique is employed for the simulation of complex systems involving multi-domain and multi-rate components (the power and communication layers) such as in smart grids [28]. The driving force behind co-simulation is due to the fact that the subsystems individually employ numerous established technologies and simulation tools. Depending on the framework used the co-simulation employed can be further categorized as (1) generic co-simulation and (2) specific co-simulation. If the model of the system and the run-time environment/solver are integrated on the same platform, the simulation is said to be the classical simulation. However, if the model of the complete system is developed on one platform but executed on different solvers, it falls under the category of parallel or distributed simulation. An alternative to this approach is the hybrid or mixed simulation approach in which separately developed representations (models) of various subsystems are solved on the same solver. Apart from the abovementioned methods, hardware test benches/beds and high-power
21 Real-Time Testing of Microgrids
625
Fig. 21.6 Illustration of a generalized testing platform for a microgrid, using a combination of testing approaches
level prototypes are also used extensively to test designed microgrid controllers and equipment [14, 29]. Figure 21.6 depicts a scenario in which a large system is tested using a combination of methods such as co-simulation, RT emulation, and HIL [30, 31]. With the increased smartness in the system due to intelligent electronic devices (IED) and information and communication technologies (ICT), the testing framework needs to account for the interaction between various communication layers and associated latencies, which render complexities. In the Internet of Things (IoT) scenario, smart microgrids are becoming ever more intelligent and self-resilient compared to the conventional system [32]. A new concept called the internet of energy (IoE) has emerged [33], which is a subsector of the internet of things and involves the use of advanced digital controllers, sensors, meters, and actuators with the ability of information exchange through IT networks. However, these result in increased structural and communication network complexities. To address the microgrid management and protection objectives an efficient and highly reliable communication network is necessary. The latency requirements and security of the communication network are vital aspects that need to be addressed. Artificial
626
A. S. Vijay and S. Doolla
Intelligence (AI) and deep learning algorithms and fog computing/ clouds are being explored to constantly enhance existing services, and technologies such as edge computing and smart systems have already been proposed in literature to improve the system functionality and power quality. Transaction Management Platforms (TMP) based on the Internet of Things (IoT) is a recent approach wherein individual prosumers can engage in interactions, negotiate with each other, enter agreements, and make proactive run-time decisions—individually and collectively depending upon the energy demands and environmental conditions. The challenges of privacy, trust and providing resilience is challenging—and rigorous algorithms which seek to provide solutions need to be extensively tested, which demand that the testing platforms provide both the cyber physical and power level capability and flexibility. Another vital issue that needs to be addressed is that of the vulnerabilities caused by cyber-attacks and data corruption, which again demand sophisticated communication layers that need to be carefully modeled and emulated and validated for their robust performance. To achieve a safe, resilient, efficient, and dispatchable energy network, the cyber-security solution may need both hardware and softwarebased mechanisms providing several layers of defense against cyber intrusions. The use of blockchain-based mechanisms as a distributed solution for the management of energy transactions in modern distributed systems is another recent trend [34, 35]. Blockchains are distributed databases that maintain time stamped lists of records and permit transactions among peers (P2P) without intermediary or central institutions. Using crypto-currencies for monetary transactions is one of the several possibilities offered by blockchain in the energy field. The key properties of the blockchain include the maintenance of the consensus mechanism throughout the network, storage of the data as a ledger into blocks, synchronization of the whole ledger throughout the network, and provision of decentralized data. The blockchain technology provides good cyber resilience to the microgrid, with its auditable and robust data management characteristics and elimination of monopoly and providing transparency and security. The testing of these management algorithms also needs highly flexible reconfigurable platforms, which can capture the interaction between the communication and hardware worlds. Another trend which is quite popular in industry is a platform involving a realtime digital replica of the product with HIL capability. This type of system, which is known as a “Digital Twin,” is an approach compatible with model-based design. Digital twin is a significant step toward the achievement of smart manufacturing and provides a new paradigm for test validations and fault diagnosis. It simulates and maps the complete product life cycle through realistic and dynamically optimized simulation models [36, 37], thus reducing the cost of the design and maintenance cycles. However, manufacturing companies face an uphill challenge to incorporate digital twins due to the absence of verified and validated simulation models. The main advantage of carrying out “virtual validation and commissioning” is reduced physical design and commissioning time. The information generated by the digital twin can be used to improve future predictive simulations, and the obtained analytics can be used to carry out predictive and condition-based maintenance. The digital twin is constantly optimized during the product design, development, and
21 Real-Time Testing of Microgrids
627
production phases. These experiment-able Digital Twins (EDTs) have a promising potential to ease the model generation process for a complex system of systems such as a smart microgrid and may offer the required degree of interoperability and flexibility necessary for models spanning different disciplines and models. Agent-based modeling (ABM) and AI-based deep learning algorithms may be used to enhance the optimization of the digital twins. Thus, the analysis and design phases are mixed with the anticipatory predictive diagnostics and this could well hold the key for the efficient and effective testing of the complex smart microgrid subsystems.
21.7 Conclusions Microgrid systems and their applications have undeniable importance in the modern smart distribution network which is becoming ever more decentralized, digitized, decarbonized, and distributed. The testing of such systems is important in their design and development cycle. A very flexible and accurate testing environment is provided by both HIL and RT emulators. Figure 21.7 depicts the various approaches adopted for the testing of microgrids around the world. Some of these systems are widely deployed in academic institutions and research labs. However, issues related to bandwidth, time step, stability, sensors and their interface have yet to be
Fig. 21.7 Testing platforms for microgrids around the world
628
A. S. Vijay and S. Doolla
overcome. As microgrids continue to evolve into much faster systems with growth in size and complexity, and with the ever-evolving stringent standards governing their control, various events and modifications may occur of a sudden, and if a conventional centralized control is used, it may hit the limits. A possibility in the area lies in the development of approaches to remotely perform the hardware in the loop testing. This could be addressed from more research and exploration on distributed HIL testing based on model reference adaptive control (MRAC). The degree of detail required in the final results, constrain the testing method. This would indirectly effect the cost of the system. Choice of a method is case specific and hence a general comment of the superiority of one approach over the other cannot be made. The RT simulation methods discussed in this chapter are typically used independently for development and testing. These could then be combined together to form a powerful tool for the testing of integrated systems. Interoperability and intellectual property sharing of the models developed is necessary to overcome the limitations. A thorough analysis of the models, the assumptions involved in their simplification (model order reduction) and the limitations or uncertainty quantification is the need of the hour. Therefore, a collective effort between different research facilities and manufacturers would prove beneficial. This will possibly accelerate the development of these testing technologies, reducing the costs and time scales involved. More focus on the interoperability and standardization of various tools and methods with attention to their possible combinations for optimizing the performance of the overall testing of complex systems is required. As microgrids and active distribution systems continue to evolve into more efficient, smart, and complex systems, the multidimensional challenges which arise must be handled using new methods and comprehensive tools. Acknowledgments The authors would like to thank the support from India-UK Centre for Education and Research in Clean Energy (IUCERCE), a project funded by Department of Science and Technology, India.
References 1. Vijay, A. S., Doolla, S., & Chandorkar, M. (2017). Real-time testing approaches for microgrids. IEEE J. Emerg. Sel. Topics Power Electron., 5(3), 1356–1376. https://doi.org/10.1109/ JESTPE.2017.2695486. 2. Adzic, E. M., Adzic, M. S., Katic, V. A., Marcetic, D. P., & Celanovic, N. L. (2013). Development of high-reliability EV and HEV IM propulsion drive with ultra-low latency HIL environment. IEEE Trans. Ind. Informat., 9(2), 630–639. https://doi.org/10.1109/TII.2012.2222649. 3. Steurer, M., Meeker, R., Schoder, K., & McLaren, P. (2011). Power hardware in- the-loop: A value proposition for early stage prototype testing. In Proceedings of the IEEE international 37th annual conference of industrial electronic society (IECON ‘11), November 2011, IEEE (pp. 3731–3735). 4. Issacs, A. (2017). Simulation technology: The evolution of power system modelling. IEEE Power and Energy Magz., 15(4), 88–102. https://doi.org/10.1109/MPE.2017.2690527.
21 Real-Time Testing of Microgrids
629
5. Siqueira, J. C. G., Bonatto, B. D., Marti, J. R., Hollman, J. A., & Dommel, H. W. (2015). A discussion about optimum time step size and maximum simulation time in EMTP-based programs. Int. J. of Electric Power and Energy Syst., Elsevier., 72, 24–32. https://doi.org/10.1016/j.ijepes. 2015. 02.007. 6. Popovici, K., & Mosterman, P. J. (2013). Real-time simulation Technologies: Principles, methodologies, and applications (p. 645). Boca Raton: CRC Press. https://doi.org/10.1201/ b12667-1. 7. Faruque, M. D. O., et al. (2015). Real-time simulation technologies for power systems design, testing, and analysis. IEEE Power Energy Technol. Syst. J., 2(2), 63–73. https://doi.org/ 10.1109/JPETS.2015.2427370. 8. Lauss, G. F., Faruque, M. O., Schoder, K., Dufour, C., Viehweider, A., & Langston, J. (2016). Characteristics and design of power hardware-in-the-loop simulations for electrical power systems. IEEE Transactions on Industrial Electronics, 63(1), 406–417. https://doi.org/10.1109/ TIE.2015.2464308. 9. Chen, Y., & Dinavahi, V. (2013). Multi-FPGA digital hardware design for detailed large-scale real-time electromagnetic transient simulation of power systems. IET Genr. Transm. Distrib., 7(5), 451–463. https://doi.org/10.1049/iet-gtd.2012.0374. 10. Marti, J. R., & Lin, J. (1989). Suppression of numerical oscillations in the EMTP power systems. IEEE Transactions on Power Apparatus and Systems, 4(2), 739–747. https://doi.org/ 10.1109/59.193849. 11. Dufour, C., & Belanger, J. (2014). On the use of real-time simulation technology in smart grid research and development. IEEE Transactions on Industry Applications, 50(6), 3963–3970. https://doi.org/10.1109/TIA.2014.2315507. 12. Chen, Y., & Dinavahi, V. (2012). Digital hardware emulation of universal machine and universal line models for real-time electromagnetic transient simulation. IEEE Transactions on Industrial Electronics, 59(2), 1300–1309. https://doi.org/10.1109/TIE.2011.2157296. 13. Panwar, M., Lundstrom, B., Langston, J., Suryanarayanan, S., & Chakraborty, S. (2013). An overview of real-time hardware-in-the-loop capabilities in digital simulation for electric microgrids. In Proceedings of the IEEE North Amer. Power Symp. (NAPS ‘13), September 2013, IEEE (pp. 1–6). 14. Sanchez, A., Castro, A., & Garrido, J. (2012). A comparison of simulation and hardware-inthe- loop alternatives for digital control of power converters. IEEE Trans. Ind. Informat., 8(3), 491–500. https://doi.org/10.1109/TII.2012.2192281. 15. Edrington, C. S., Steurer, M., Langston, J., Mezyani, T. E., & Schoder, K. (2016). Characteristics and design of power hardware-in-the-loop simulations for electrical power systems. IEEE Transactions on Industrial Electronics, 63(1), 406–417. https://doi.org/10.1109/ TIE.2015.2464308. 16. Hatakeyama T, Riccobono A, Monti A. Stability and accuracy analysis of power hardware in the loop system with different interface algorithms. In: Proceedings of the IEEE 17th workshop control modeling power electron. (COMPEL ‘16), June 2016, IEEE; 2016. p. 1–8. 17. Yin, C., et al. (2016). Virtual impedance method of the power hardware-in-the loop simulation to improve its stability and accuracy. In Proceedings of the IEEE 8th international power electron. Motion control conference (IPEMC-ECCE Asia ‘16), may 2016, IEEE (pp. 2752– 2758). 18. Liegmann, E., Riccobono, A., & Monti, A. (2016). Wideband identification of impedance to improve accuracy and stability of power-hardware-in the loop simulations. In Proceedings of the IEEE international workshop Appl. Meas. Power Syst. (AMPS ‘16), September 2016, IEEE (pp. 1–6). 19. Dargahi, M., Ghosh, A., & Ledwich, G. (2014). Stability synthesis of power hardware-inthe-loop (PHIL) simulation. In Proceedings of the IEEE PES general meeting conference exposition, July 2014, IEEE (pp. 1–5). 20. Daniil, N., & Drury, D. (2016). Improving the stability of the battery emulator - pulsed current load Interface in a power hardware in-the-loop simulation. In Proceedings of 42nd annual conference of the IEEE industrial electronics society (IECON ‘16), October 2016, IEEE (pp. 2064–2069).
630
A. S. Vijay and S. Doolla
21. Li, S., Qi, W., Tan, S. C., Hui, S. Y., & Tse, C. K. (2018). A general approach to programmable and reconfigurable emulation of power impedances. IEEE Transactions on Power Electronics, 33(1), 259–271. https://doi.org/10.1109/TPEL.2017.2663424. 22. Shen, Z., & Dinavahi, V. (2017). Dynamic variable time-stepping schemes for real-time FPGA-based nonlinear electromagnetic transient emulation. IEEE Transactions on Industrial Electronics, 64(5), 4006–4016. https://doi.org/10.1109/TIE.2017.2652403. 23. Araujo, E. P., Rosell, P. O., Mane, M. C., Robles, R. V., & Bellmunt, O. G. (2015). Renewable energy emulation concepts for microgrids. Renew. Sustain. Energy Rev. Elsevier., 50, 325–345. https://doi.org/10.1016/j.rser.2015.04.101. 24. Vijay, A. S., Chandorkar, M. C., & Doolla, S. A system emulator for AC microgrid testing. IEEE Trans. Ind. Electron. 2019; IEEE Early Access. https://doi.org/10.1109/ TIA.2019.2942275. 25. Fajri, P., Lee, S., Prabhala, V. A. K., & Ferdowsi, M. (2016). Modeling and integration of electric vehicle regenerative and friction braking for motor/dynamometer test bench emulation. IEEE Transactions on Vehicular Technology, 65(6), 4264–4273. https://doi.org/ 10.1109/TVT.2015.2504363. 26. Rabia, S., & Tang, S. (2013). Control Laws for the emulation of an electric vehicle drivetrain by two electric machines. In Proceedings of the IEEE vehicle power and propulsion conference (VPPC ‘13), October 2013, IEEE (pp. 1–7). 27. Alvarez, A. F., Garcia, M. P., Valderas, M. G., Lopez, J., & Sanz, M. (2017). HW/SW co-simulation system for enhancing hardware in the loop of power converter digital controllers. IEEE J. Emerg. Sel. Topics Power Electron., 5(4), 1779–1786. https://doi.org/10.1109/ JESTPE.2017.2739710. 28. Lundstrom B, Chakraborty S, Lauss G, Brundlinger, Conklin R. evaluation of systemintegrated smart grid devices using software- and hardware-in-the-loop. In: Proceedings of the IEEE power energy society innovative smart grid technology conference (ISGT ‘16), September 2016, IEEE; 2016. p. 1–5. 29. Caldognetto, T., Santa, L. D., Magnone, P., & Mattavelli, P. (2017). Power electronics based active load for unintentional islanding Testbenches. IEEE Transactions on Industry Applications, 53(4), 3831–3839. https://doi.org/10.1109/TIA.2017.2694384. 30. Technologies, O., & Electrification Magazine, I. E. E. E. (2017). Vol., 2, 5. 31. Donahue, E. J. (2019). Microgrids: Applications, Solutions, Case Studies, and Demonstrations, InTech Open Access Chapter. https://doi.org/10.5772/intechopen.83560. 32. Shahinzadeh H, Moradi J, Gharehpetian G B, Nafisi H, Abedi M, Internet of Energy (IoE) in Smart Power Systems In: Proceedings of 5th Conference on Knowledge Based Engineering and Innovation (KBEI ‘19) 2019. p. 627–636. 33. Moness, M., & Moustafa, A. M. (2016). A survey of cyber-physical advances and challenges of wind energy conversion systems: Prospects for internet of energy. IEEE J Int. of Things., 3(2), 134–145. https://doi.org/10.1109/JIOT.2015.2478381. 34. Di Silvestre, M. L., Gallo, P., Ippolito, M. G., Sanseverino, E. R., & Zizzo, G. (2018). A technical approach to the energy Blockchain in microgrids. IEEE Trans. Ind. Informat., 14(11), 4792–4803. https://doi.org/10.1109/TII.2018.2806357. 35. Musleh, A. S., Yao, G., & Muyeen, S. M. (2019). Blockchain applications in smart grid–review and frameworks. IEEE Access, 7, 86746–86757. https://doi.org/10.1109/ ACCESS.2019.2920682. 36. Xu, Y., Sun, Y., Liu, X., & Zheng, Y. (2019). A digital-twin-assisted fault diagnosis using deep transfer learning. IEEE Access, 7, 19990–19999. https://doi.org/10.1109/ ACCESS.2018.2890566. 37. Khan A, Dahl M, Falkman P, Fabian M, Digital Twin for Legacy Systems: Simulation Model Testing and Validation In: Proceedings of IEEE 14th International Conference on Automation Science and Engineering (CASE ‘18) 2018. p. 421–426.
Index
A ABC, see Artificial bee colonies (ABC) ACMGs, see AC microgrids (ACMGs) AC microgrids (ACMGs), 407–408 DGs, 376 power losses, 375 stability analysis, 379, 380 ACO, see Ant colony optimization (ACO) AC solid state circuit breakers (AC-SSCB), 543 Adaptive directional overcurrent protection, 557–558 Adaptive fuzzy proportional integral-derivative (AFPID) controller, 312–313, 319, 322 design, 307 NSIDE algorithm, 308 renewable energy sources, 307 voltage and frequency, 322, 325 Adaptive protection for AC microgrids, 553, 555–571 Adaptivity, protection scheme, 539, 592 Advanced metering infrastructure (AMI), 68 AFPID controller, see Adaptive fuzzy proportional integral-derivative (AFPID) controller Agent-based modeling (ABM), 627 AI, see Artificial intelligence (AI) AI-based deep learning algorithms, 627 AMI, see Advanced metering infrastructure (AMI) Ant colony optimization (ACO), 221, 223–225, 244 Artificial bee colonies (ABC), 221, 228–229, 234, 244, 513
Artificial intelligence (AI), 626 Automation devices and level of automation, 584, 585 Average energy not supplied (AENS), 591
B Balanced positive sequence control (BPSC) method, 410–411 Battery energy storage (BES), 98, 136–137, 143, 144, 222, 229, 237–238, 407, 552 Battery–flywheel, 143, 145 Battery–fuel cell, 141–142 Battery supercapacitor, 141, 142 BBO, see Biogeography based optimization (BBO) BES, see Battery energy storage (BES) Biogeography based optimization (BBO), 221, 225, 233, 236, 237, 240 Blockchain-based mechanisms, 626 Blockchain technology, 626 BPSC method, see Balanced positive sequence control (BPSC) method
C CAES, see Compressed air energy storage (CAES) Centralized communication-based adaptive protection, 555–556 Central moment-based algorithm adaptive protection schemes, 490 DC/AC converters, 489 differential protection, 492–498
© Springer Nature Switzerland AG 2021 A. Anvari-Moghaddam et al. (eds.), Microgrids, Power Systems, https://doi.org/10.1007/978-3-030-59750-4
631
632 Central moment-based algorithm (cont.) power system, 491 Smart Grids, 489, 490 CERTS, see Consortium for Electric Reliability Technology Solutions (CERTS) Circuit breaker (CB) technology, 542–543, 584 Closed loop simulation, 624 Communication-based adaptive protections for AC microgrids, 556–561 centralized, 555–556 decentralized, 556 Communication-less adaptive protection, 568–569 Communication networks, 616 Compressed air energy storage (CAES), 132, 134–136 Consortium for Electric Reliability Technology Solutions (CERTS), 3 Constraints ADMM, 211 charge and discharge rate limit, 48 in coordination process, 601 distributed generation, 110 economic and space, 618 energy storage system, 110–111 frequency, 92 generation capacity limit, 48 NMCs, 25–26 objective functions (see Objective functions) power balance, 48, 109–110 reserve, 49–50 Control in MGs AC MGs, 407–408 DC MGs, 408–409 natural phase-coordinates approach amplitudes and phase angles, 413–414 generalized NPC method, 417–418 phase-coordinates approach, 413 phasor graphs, 415 symmetric NPC method, 415–417 three-phase currents, 414 traditional reference current control methods BPSC, 410–411 decoupled double synchronous reference control method, 411–412 Controller hardware in the loop (CHIL), 616, 618 signal level exchange, 622 Control mode change, DER units, 528 Conventional droop methods AC microgrid ESS, 266
Index frequency and voltage responses, 266–269 mathematical analysis, 256–260 RESs, 262–265 synchronous generator, 260–262 DC microgrid converter-based generator, 270–273 mathematical analysis, 269–270 Conventional protection of microgrids, 543–553 Conventional service restoration procedure, 582, 583 Co-operative simulation, 624 Co-simulation, 624 Cost-benefit analysis, 591 Critical clearing times (CCT), 593–594 Crypto-currencies for monetary transactions, 626 Cuckoo search algorithm (CSA), 221, 227, 233, 234, 239, 244 Current symmetrical components based protection, 558–559 Customer damage function (CDF), 592 Cyber-attacks, 626 Cyber resilience, 626
D Data corruption, 626 DCMGs, see DC microgrids (DCMGs) DC microgrids (DCMGs), 375–377, 408–409, 571–574 DCS, see Distributed control strategy (DCS) Deadbeat current control, 408, 433–434 Decentralized communication-based adaptive protection, 556 Decision-making system assessment tools load flow, 65 load forecast/estimation, 65 network and component modeler, 64 remedial action scheduling tool, 64–65 security assessment tool, 64 short circuit calculation, 65 state estimation, 65 uncertainty assessment, 65 optimization tools decision making under uncertainty, 66 economic dispatch, 66 network reconfiguration tool, 66 optimal power flow, 66 unit commitment, 66 restoration tools, 67 UC and ED, 69
Index Decision procedure decision algorithm, 193–196 DR cost modeling, 191–193 microgrid cost modeling, 189–191 Decoupled double synchronous reference control method, 411–412 Deep learning algorithms, 626 Demand response (DR) programs, 50–51, 100–101 cost modeling IDR-based, 193 PDR-based, 191–192 decision algorithm, 193–196 with DR program, 120–124 economic problem, 181 green systems, 180 mathematical model, 180 numerical studies commercial load, 198–199 final deduction, 201 hospital load, 199–201 industrial load, 197–198 price-based, 99 types IDR strategies, 183–184 PDR strategies, 181–183 programs and microgrid operation, 184 without DR program, 116–120 See also Decision procedure Department of energy (DOE), 584–585 DERs, see Distributed energy resources (DERs) DER unit fault-ride-through (FRT) requirements, 526 DG units, see Distributed generation (DG) units Differential protection busbar differential protection, 513, 515–516 factors, 442 multipoint measurements, 439 principle current transformer saturation, 497–498 negative-sequence, 494–497 percentage, 492–494 reactor, 509, 513, 514 variable tripping time, 604 Digital real-time simulators (DRTS), 617 HIL RT simulation, 617, 618 software based, 617, 618 Digital real-time testing, 620–621 Digital signal controllers (DSCs), 621 Digital twin, 626 Directional overcurrent protection, 529, 557–558
633 Directional overcurrent relays (DOCRs) coordination on-line implementation, 476 optimization problem, 468–477 results and discussion, 478–485 test systems, 476–478 dual-setting, 608 energy management scenarios, 229 Distance protection, 559 with a fixed time delay, 529, 530 Distributed control, 9–10 AC MGs frequency control, 283–285 voltage control, 285–286 applications, 445 DC MGs DG model, 279–280 distributed secondary control, 280–281 MCA and MCB, 30 MG control system, 275 in NMCs, 35–36 single system, 624 SOC balancing, 295 Distributed control strategy (DCS), 32, 35–36 Distributed energy resources (DERs), 581 and BESS, 552–553 best utilization of, 21 converter-based, 536 energy users, 14 fault current injections, 404 flexibility, 151 FRT technique, 404, 581 grid-connected mode, 548, 550 integration, 151 inverter-based, 406, 598–600 KKT conditions, 206 methodology description, 206–207 MGs, 70 on-site generation sources, 185 optimal switch allocation, 582 optimal target configuration, 582 optimization problem, 234 power grid, 185 systems, 4 problem formulation distributed energy management, 211 microgrids’ real-time energy management, 208–210 numerical studies and result analysis, 211–215 protection coordination grid-connected mode with DERs, 548–551
634 Distributed energy resources (DERs) (cont.) grid-connected mode without DERs, 546–548 islanded mode with BESS, 552–553 relay-relay coordination, 595 and storage systems, 407 synchronous-based, 593, 601 transient stability, 581 unit fault behavior, 526–529 utilization of, 21 Distributed generation (DG) units active and reactive output power, 364–366, 370, 371 capacity optimization, 98 constraint, 110 DC MG, 279–280 economic and environmental issues, 305 and ESS, 106 fault behavior and effect, 526–529 generators, 459–460 grid codes and distribution network protection schemes, 526 inverter-interfaced, 338, 339 network multiple microgrids, 21 nonrenewable units fuel cell, 102–104 micro turbine, 102 power systems, 70 renewable units photovoltaic panel, 104 wind turbine, 104–106 Simulink, 455 Distributed HIL, 628 Distributed simulation, 624 Distribution management system (DMS), 29–31, 34, 37, 67, 438, 583 Distribution network protection schemes, 526 Distribution networks, 581 DMS, see Distribution management system (DMS) DOCRs, see Directional overcurrent relays (DOCRs) Double synchronous reference frame (DSRF), 353–354 Drive trains of electric vehicles, 624 DR programs, see Demand response (DR) programs DSOGI, see Dual SOGI (DSOGI) DSRF, see Double synchronous reference frame (DSRF)
Index Dual SOGI (DSOGI) DSOGI-FLL, 355–356 DSOGI-PLL, 356 Dual time–current–voltage characteristics, 609–610 E EA, see Evolutionary algorithms (EA) Economic dispatch (ED), 23, 25, 28, 34, 39, 66, 69, 93, 180, 228, 237, 306 Electrification process, 581 EMS, see Energy management systems (EMS) Energy management systems (EMS), 616 centralized, 69 control system, 63–64 decentralized, 69 decision-making system, 64–67 functions, MG scheduling microgrid’s hierarchical scheduling, 72–74 system operation strategies, 74–76 interaction with other systems AMI, 68 bid/offer interface, 68 DMS, 67 electricity market, 68 maintenance scheduling system, 68 OMS, 68 weather forecasting system, 68 mathematical modeling MG’s components, 76–84 system security, 85–92 and MGs, 62, 63 monitoring system, 63 NMCs (see Networked microgrid clusters (NMCs)) scheduling of MG, 70 structures, 39–40 studies, strategies, 42–43 time framework, 37, 38 Energy not supplied (ENS) index, 583 Energy storage systems (ESS), 106, 266 AC and DC, 289 applications in microgrids load leveling, 143–144 power quality, 145 BMS, 293 characteristics, 291–292 classification, 133, 290 constraint, 110–111 energy/power buffer, 127 in grid-connected microgrids, 296–299
Index hybrid energy storage technologies battery–flywheel, 143 battery–fuel cell, 141–142 battery supercapacitor, 141 in islanded microgrids, 294–296 land-based microgrids, 128–130 load demand types, 132 microgrids classification, 128 mobile microgrids, 130–131 operation modes, 131 opportunity, 299 PCC, 290 power electronic interface, 292–293 generation of DGs, 21, 289 research trends, 299 single energy storage technologies BES, 136–137 CAES, 135–136 FBES, 137–138 FES, 138 fuel cell, 140–141 PHS, 133–135 SES, 139, 140 SMES, 139 timescales, 291 See also Microgrids (MGs) Enhanced MV microgrid protection scheme 1—HIF detection, 566–568 ESS, see Energy storage systems (ESS) Evolutionary algorithms (EA) ABC, 228–229 ACO, 223–225 application energy management, 233–237 energy-related devices sizing, 237–240 frequency control, 240–243 microgrid optimal voltage, 240–243 operation scheduling, 233–237 optimal placement, 237–240 BBO, 225 cost-effective microgrid system, 219 CSA, 227 FA, 229–230 GA, 221–222 GOA, 230–232 GWO, 229 heuristic and metaheuristic algorithms, 220 HSA, 226 microgrids planning approaches and references, 244 PSI, 223 WOA, 232–233 Experimentable Digital Twins (EDTs), 627
635 F FA, see Firefly algorithm (FA) Fast-response protection in restoration, 600–602 Fault behavior and fault current feeding capability of a DER unit, 526–529 Fault current management (FCM) angle difference, 419 IB-MG interface, 420 multiple microgrids, 424–426 phasor illustration, 421 short-circuit currents, 418, 419, 422 single MG, 421–423 waveforms, 420 See also Fault ride through (FRT) Fault indicators (FIs), 581 allocation, modeling process, 586 fault detective devices, 587 fault locating process, 588 feeders and laterals, 586 placement, 585–589 and restoration devices, 584–585 simultaneous placement, 589–592 Fault location isolation and service restoration (FLISR), 582 and restoration requirements, 582–592 Fault ride through (FRT) analysis and results, 428–431 capabilities and requirements, 552, 569 control in MGs, 407–418 deadbeat current control, 433–434 DER integration, 404 FCM, 418–426 hardware-in-the-loop platform, 426–428 implementation, 403 large-scale solar power plants, 403 MG FRT in grid codes, 405–407 power control scheme, 404 reactive power injection, 431–432 requirements, 526 SRF-PLL, 432 voltage profile, 601 FBES, see Flow battery energy storage (FBES) FC, see Fuel cells (FC) FCM, see Fault current management (FCM) FES, see Flywheel energy storage (FES) Field programmable gate arrays (FPGAs), 617, 621 Firefly algorithm (FA), 98, 221, 229–230 Firm RT applications, 617 Flow battery energy storage (FBES), 132, 137–138
636 Flywheel energy storage (FES), 132, 138, 143 Fog computing/clouds, 626 FPGAs, see Field programmable gate arrays (FPGAs) Frequency locked-loop (FLL), 350, 355–356 FRT, see Fault ride through (FRT) Fuel cells (FC), 6, 14, 52, 102–104, 124, 140–141, 234, 375, 376
G GA, see Genetic algorithm (GA) Generalized emulator (source/load), 623–624 Generator-based distributed energy resources (DERs), 524 Generic co-simulation, 624 Genetic algorithm (GA), 221–222, 226, 234, 317, 440 Global design of the optimization cost function, 314 NSIDE algorithm improved differential evolution algorithms, 316–317 nondominated sorting-based multiobjective algorithm, 317–318 system optimization variables stage I, 315 stage II, 315 stage III, 315–316 GOA, see Grasshopper optimization algorithm (GOA) Graphical user interface (GUI) blocks, 617 Grasshopper optimization algorithm (GOA), 230–232 Grey wolf optimization (GWO), 99, 111–112, 120, 124, 221, 229, 244 Grid codes, 569–571 Grid-connected microgrids DERs, 548–551 ESS control strategies frequency regulation, 298–299 voltage regulation, 297–298 frequency regulation, 298–299 islanded mode, 4, 289 MG stability, 10, 11 microgrid protection, 524 OC relay protection, 546 photovoltaic-battery hybrid system, 98 protection coordination, 546–548 reactive and active power flow, 240 re-synchronization, 540
Index structure mode, 40 voltage regulation, 297–298 Grid-connected mode of AC microgrid with single setting OC relay protection, 545, 546 GWO, see Grey wolf optimization (GWO)
H Hard RT applications, 617 Hardware in the loop (HIL), 616 Hardware in the loop RT simulation (HILS), 621–623 CHIL, 618 comparative analysis, testing approaches, 618, 619 electrical drives in aircraft industry, 621 hybrid approaches, 619 interactive virtual prototype, 619 interconnection of various systems, 618, 619 mechanical HIL, 618 multi-domain and cooperative/cosimulation, 619 PHIL, 618 prototyping stage, 618 Hardware test beds (HTBs), 616 Harmonic compensation pq method, 345–347 reference currents identification instantaneous active, 340–342 reactive power method, 340–342 SRF method, 342–344 SRF method, 347–349 Harmony search algorithm (HSA), 226, 233, 244 HCS, see Hierarchical control structures (HCS) Hierarchical control structures (HCS) AC MGs, 277–278 DC MGs, 276–277 MG control system, 275 primary control, 35 Hierarchical system in laboratory scale, 624 High-speed telecommunication, 526 High voltage ride-through (HVRT), 528 HSA, see Harmony search algorithm (HSA) HV network fault, 535 Hybrid AC/DC microgrids (HADMGs), 6, 222, 237, 289, 375–377 Hybrid simulation approach, 624 Hybrid testing approaches, 624–627
Index I IACMGs, see Islanded AC microgrids (IACMGs) IBRs, see Inverter-based renewables (IBRs) IDR, see Incentive-based DR (IDR) IED, see Intelligent electronic devices (IED) Imbalance compensation direct extraction methods in abc frame, 350–351 αβ reference, 351–352 dq frame, 352–353 indirect extraction methods DSOGI-FLL, 355–356 DSOGI-PLL, 356 DSRF, 353–354 instantaneous power under unbalanced conditions, 357–358 Incentive-based DR (IDR), 183–184, 193–195, 197–201 Information and communication technologies (ICT), 625 Intelligent electronic devices (IED), 490, 526, 534, 556, 558, 625 Internet of energy (IoE), 625 Internet of things (IoT), 625 Inverter-based renewables (IBRs), 256, 262–266, 274 Inverter-interfaced DG switching signals, 360 VSC control, 338, 339 Islanded AC microgrids (IACMGs) ACMGs, 379–381 DCMGs, 375 dynamic step response, 394 HADMG, 376, 378 hybrid AC/DC microgrid, 376 microsources control methods, 377 MIMO, 383 open-loop, 393 PDC, 382 power system stability, 378, 379 Q-V droop control method, 377 results and discussions eigenvalue analysis, 391–393 time-domain simulations, 393–398 signal transmission time delays, 383 small-signal disturbance, 381 SMT and WAMS, 381 stability enhancement, 382 WAMS, 382 Islanded microgrids calculated initial conditions, 323 coordinated control of of multiple ESSs, 294–295
637 RESs and ESSs, 296 double-sided auction, 155 ESS control strategies, 294 power flow, 322 proposed control strategy, 308 voltage/frequency control, 312 voltage responses, 325 Islanding detection during island operation of nested microgrid, 569
J JADE agents, 448–450 architecture communication language used by agents, 451 flat hierarchy, 446 Matlab/Simulink, 454–455 testing and measurement tool, 452–454 work environment structure, 451–452
L Land-based microgrids, 143, 145 industrial, 129, 130 residential, 128–129 Line differential protection based on current measurements, 560 Load emulation, 623–624 Local markets benefits, 154 definition, 153–154 key elements, 156, 158 market settlement approaches, 161–169 models, 158–159 objectives, 154–155 services, 155–156 trading approaches hybrid, 161 pool-based, 159, 160 P2P trading, 160 value streams for microgrids, 156, 157 Low voltage ride-through (LVRT), 528 LV microgrid protection, 525 LV microgrid synchronized reconnection, 540–542
M Manual switches (MSs), 589 Market settlement approaches auction-based approach, 162–163 case examples, 170–172
638 Market settlement approaches (cont.) in local markets, 170–172 the Monash microgrid, 169–170 optimization-based approach decentralized clearing, 166–169 distributed clearing, 164–166 and pricing, 158 MAS proposals, see Multi-agent system (MAS) proposals Mathematical modeling, MG’s components CHP and boiler, 80–82 DG, 77–79 electrical network linear distflow, 83–84 SOCP, 82–83 energy exchange, 84 storages, 79–80 loads, 76–77 reactive power resources, 80 renewable-based units with MPPT, 79 Maximum power point tracking (MPPT), 79, 263–266, 268, 404, 408, 409, 431 MCA, see MG A controller (MCA) MCB, see Miniature circuit breaker (MCB) Medium-voltage (MV) microgrid protection, 564–569 microgrids after isolation, 524–525 MEG, see Microgrid Exchange Group (MEG) MG A controller (MCA), 29, 30 MGs, see Microgrids (MGs) Microelectronics/computational capability, 615 Microgrid cost modeling installation, 189–190 maintenance, 190 operation, 190 start-up, 190–191 Microgrid Exchange Group (MEG), 3 Microgrid protection issues, 524–535 in grid-connected mode, 524 in islanded mode, 524 number of zones for protection, 525 operation speed specifications, 525 protection methods, 525 reduced fault current contributions, 525 Microgrids (MGs), 99–100 advantages, 12–13 CERTS, 3 challenges economical, 14 marketing, 14 regulation, 14 technical, 13 classification
Index application, 6 campus/institutional sector, 186 characteristics/properties, 7 commercial sector, 186–187 configuration, 6–7 healthcare sector, 187 industrial sector, 185–186 military sector, 186 operation mode, 6 other microgrids, 188 remote/rural microgrids, 188 residential sector, 187 size, 6 type, 5 combined heat and power model, 52–53 components, 4–5 control coordination centralized, 9 distributed, 9–10 hybrid, 10 demand response programs, 98 discussion, 443–444 DOCRs coordination on-line implementation, 476 optimization algorithms, 471–477 optimization problem, 468–471 results and discussion, 478–485 test systems, 476–478 ESS (see Energy storage systems (ESS)) fuel cell model, 52 hierarchical control primary, 7 secondary, 8 tertiary, 8 islanding condition detection schemes, 461–463 implementation, 463–468 micro-turbines model, 52 multi-agent system proposals, 444–450 multi-objective function, 99 objective functions (see Objective functions formulation) operation management (see Operation management) optimal management, 97 problems and functional solutions, 439–443 protection, 12, 438–439 unit commitment, 98 weak infeed conditions detection DG generators, 459–460 small conventional generators, 458–459 Microgrid transition to islanded operation, 529–535 MIMO, see Multi-input-multi-output (MIMO)
Index Miniature circuit breaker (MCB), 29, 30, 537, 542 Mixed simulation approach, 624 Model-based design (MBD), 617 Model-based systems engineering (MBSE) and simulations, 617 Model in the loop (MIL), 616 Model reference adaptive control (MRAC), 628 The Monash microgrid, 169–170 MPPT, see Maximum power point tracking (MPPT) MRAC, see Model reference adaptive control (MRAC) Multi agent-based protection, MGs, 602–604 Multi-agent system (MAS) proposals decentralized decision making, 69 distribution system, 445 electricity sector, 446, 447 first-order linear, 285 gang of agents, 448–450 IED, 556 overcurrent relay, 460 power supply conditions, 483 smart agent definition, 445–446 switch position, 458 Multi-domain and cooperative/co-simulation, 619 Multi-input-multi-output (MIMO), 383
N Networked microgrid clusters (NMCs) ancillary services improvement, 22 best utilization, DERs, 21 bilateral and out of market transactions, 23 challenges disallowed transactions, 25 privacy of MGs, 24 protection coordination, 24 stability of the system, 23–24 threat of cyberattack, 24–25 compare EMS structures, 39–44 control strategy DCS, 35–36 hierarchical, 33–35 master-slave, 33 peer-to-peer, 32–33 and DS, 18 DS structure, 19 EMS, 37–39 energy management modeling and solution methods, 40, 45 improvement
639 reliability, 22–23 resiliency, 22 objectives and constraints, 25–30 physical and control layers, 20 reduction of overall cost, 21 structure, 18 typical architecture of interconnected MGs, 31–32 multiple distribution feeders, 31–32 parallel MGs, 30–31 serial MGs, 26, 29–30 single distribution feeder, 26, 29–31 Networked microgrids (NMGs) energy trading framework, 216 KKT conditions, 206 methodology description, 206–207 multiple MGs, 424 nominal capacity, 22 problem formulation distributed energy management, 211 microgrids’ real-time energy management, 208–210 numerical studies and result analysis, 211–215 resilience benefits, 20 RT markets (see Real-time (RT) markets) seller and buyer, 205 See also Networked microgrid clusters (NMCs) NMCs, see Networked microgrid clusters (NMCs) NMGs, see Networked microgrids (NMGs) Novel/hybrid testing approaches, 624–627 Nuisance tripping of protective devices fuse—re-closer, 596–598 relay-relay, 594–596
O Objective functions formulation and constraints (see Constraints) cost operation modeling, 47 defined, 235 DRP, 50–51 economic–environmental, 116 MGDC, 107–109 MGs, 51–53 numerical result, 53–57 optimization algorithm, 111–113 pollution emission modeling, 47–48, 109 problem constraints charge and discharge rate limit, 48 generation capacity limit, 48 power balance constraint, 48
640 Objective functions formulation (cont.) reserve, 49–50 self-sufficiency, 49 Off-line accelerated simulation, 620 Off-line digital simulation, 616 Off-line simulation, 617 OMS, see Outage management system (OMS) Online simulators, 616 Operation management distribution systems, 205 NMCs/NMGs (see Networked microgrid clusters (NMCs)) objective functions (see Objective functions formulation) Optimal dispatch, 124, 144 Optimization algorithm, 471–476 fuzzy method, 112–113 GOA, 230–232 multi-objective grey wolf, 111–112 NSIDE, 332 PSO, 223 WOA, 232–233 Outage detectors, 584–585 Outage management system (OMS), 68, 583, 584, 589, 611 Overcurrent (OC) protection, 524, 525 definite-time vs. inverse time, 543–553 with multi-function characteristic, 606–608 multi-inverse characteristic, 604–606
P Parallel simulation, 624 Particle swarm optimization (PSO), 28, 42, 58, 98, 221, 223, 224, 234, 237, 239–242, 244, 306 PCC, see Point of common coupling (PCC) PDR, see Price- based DR (PDR) Peer-to-peer (P2P) trading, 152, 160–162 Phase-locked loop (PLL), 343, 347–350, 356, 407, 408, 410, 432, 542 Piece-wise linear characteristic, 610 Point of common coupling (PCC), 4, 18, 30, 31, 34, 39, 290, 376, 413, 461, 548, 552 Pool-based trading approaches, 159–161 Power electronic (PE) converters and equipment, 615 applications, 624 Power hardware in the loop (PHIL), 616, 618 actual power transfer, 622 ideal interface, 622 modeling and simulation aspects, 622 power level exchange, 622
Index stability in, 622 stability issues, 623 virtual impedance implementation, 623 P2P trading, see Peer-to-peer (P2P) trading Price- based DR (PDR), 191–192 costs, 198, 199 PDR-based cost, 191–192 strategies, 181–183 Processor in the loop (PIL), 616 Protection based on voltage and directional overcurrent, 561 Protection coordination in a grid-connected mode with DERs, 548–551 in islanded mode with DERs and BESS, 552–554 without DERs, 546–548 Protection of LV AC microgrids, 561–564 Protection operation principles, 532–535 Protection scheme requirements, MG, 535–543 adaptivity, 539 circuit breaker (CB) technology, 542–543 reliability, 538–539 re-synchronization, 540–542 selectivity or coordination, 537–538 sensitivity, 536–537 Protection schemes of MG, 592–610 Protection time selectivity issues, 531, 532 PSO, see Particle swarm optimization (PSO)
R Rapid controller prototyping, 616 RBTS-bus4 test system, 591 Real-time (RT) constraints, 615 emulation, 623–624 markets EMS, 69 energy management, 208–210 integration platform, 170 load shedding computation, 306 RTS, 72 power level emulation, 616 simulators, 621 testing centralized control, 628 digital, 620–621 distributed HIL, 628 emulation, 623–624 HILS, 621–623 methods, 616–619 MRAC, 628
Index novel/hybrid testing approaches, 624–627 PE, 615 platforms around the world, 627 Reconfigurable topology of MG, 611–612 Rectifier interfaced active load (RIAL), 381, 384, 386–389 Reliability, protection scheme, 538–539 Remote fault indicators, 584 Remotely controllable switches (RCSs), 584 simultaneous placement, 589–592 Renewable energy sources (RESs) autonomous operation, 128 DC MG’s control system, 276 diesel generator-based systems, 188 and energy storage systems, 523 global warming, 4 IBR, 263–265 MG control system, 275 non-IBR, 265 optimal unit commitment, 98 small-scale, 523 Renewable energy system emulators, 624 RESs, see Renewable energy sources (RESs) Rest of system (ROS), 622 Restoration capability, MG, 611–612 Re-synchronization, protection scheme, 540–542 RIAL, see Rectifier interfaced active load (RIAL)
S SCM, see Second central moment (SCM) Second central moment (SCM) computational complexity, 505 delta filter application, 504 differential protection (see Differential protection) events, 499, 500 flow chart, 502 magnitude, 491 misoperation, 499 operation, 499, 501 results busbar differential protection, 513, 515–516 generator protection, 506, 509–511 power reactor energization, 516, 519, 520 power transformer energization, 516, 518 power transformer protection, 506–508
641 reactor differential protection, 509, 513, 514 sliding window, 503 thresholds criteria, 505 Second-order cone programming (SOCP), 82–83 Selective and fast protection schemes, 592–610 Selective protection, MGs nuisance tripping of protective devices fuse—re-closer, 596–598 relay-relay, 594–596 protective devices blinding, inverter-based DERs, 598–600 selective dead time in reclosing scheme, 593, 594 Selectivity/coordination, protection scheme, 537–538 Semiconductor technology, 615 Sensitivity, protection scheme, 536–537 SES, see Supercapacitor energy storage (SES) Signal transmission time delays design procedure, 384 IIDG unit, 385, 389 lower-level decentralized controller, 384 LQG approach, 387 noise and disturbance models, 387 optimal parameters, 388 phase response, 386 RIAL, 386, 389 round-trip communication, 390 small-signal linearized model, 385 Simulation results active and reactive output power, 362, 363, 365, 366, 370, 371 AFPID controller, 319 block diagram, control system, 358, 359 CHP units, 318 frequency controller, 327 30% load decrease, 324–325, 331, 332 load shedding, 323–324, 329 nondominated responses, 319, 320 nonlinear load current, 360, 361 proposed controller, 320 studied MG, 358 symmetrical three-phase fault, 321–323, 325, 328 time domain characteristics, 331 unbalanced load current, 360 and nonlinear load connection, 360, 361 voltage after compensation, 362, 367 after nonlinear load connection, 360, 362
642 Simulation results (cont.) after unbalanced load connection, 360, 362 controller, 326 harmonic components, 362, 364 harmonic components after compensation, 363, 369 VUF and THD, 361, 363, 368 Simulation step time or execution time, 617 Smart relays, 584 SMES, see Superconducting magnetics energy storage (SMES) SMT, see Synchronized phasor measurement technology (SMT) SOC, see State of charge (SOC) SOCP, see Second-order cone programming (SOCP) Soft RT applications, 617 Software in the loop (SIL), 616, 618, 624 Source emulation, 623–624 Specific co-simulation, 624 SRF, see Synchronous reference frame (SRF) Stability AFPID, 307, 308 cryogenic liquid, 139 DERs, 306 grid-connected and islanded, 305 MG stability, 11 islanded MG stability, 11 microgrid modeling CHP model, 310–311 fixed-speed wind turbine model, 309–310 reduced network model, 312 synchronous reference frame, 311 motivations and contributions, 308 power distribution systems, 305 of the system, 23–24 voltage and frequency compensations, 260, 306, 307 State of charge (SOC), 63, 73, 110, 111, 266, 292–295, 298 Supercapacitor energy storage (SES), 132, 139–141 Superconducting magnetics energy storage (SMES), 132, 134, 139 Supervisory control and data acquisition (SCADA), 583, 616 Switch placement problem, 592 Synchronized phasor measurement technology (SMT), 381 Synchronous reference frame (SRF)
Index case of, 347–349 DSRF, 353–355 pq method, 345–347 System average interruption duration index (SAIDI), 582–583 System operation strategies economic aspects, 74–75 technical aspects, 75–76 System security HAS, 85–89 RTS, 87, 90–92
T Techno-economic analysis of MG planning, 585 THD, see Total harmonic distortion (THD) Third-part economic entity (TPEE), 206–207, 210, 213, 214–216 Time-current-voltage characteristics, 608–609 Total harmonic distortion (THD), 337, 339, 360, 363, 462, 559, 569 TPEE, see Third-part economic entity (TPEE) Transaction management platforms (TMP), 626
U UC, see Unit commitment (UC) Undervoltage-based protection, 529 Undervoltage with a fixed time delay, 530 Uninterrupted power supply (UPS), 138, 291, 375, 376 Unit commitment (UC), 66 and battery charge/discharge status, 53 demand response program, 98, 100–101 DG units (see Distributed generation (DG) units) numerical results with DR program, 120–124 hourly ability, 115, 116 IBC-SOLAR Company, 114 micro turbine, 113 without DR program, 116–120 UPS, see Uninterrupted power supply (UPS)
V Vacuum circuit breaker (VCB), 543 Virtual test beds (VTB) based platforms, 619 Virtual validation and commissioning, 626 Voltage based protection schemes, 559–560 Voltage imbalance due to asymmetrical loads and single-phase DG units, 541
Index Voltage source converter (VSC) control, 338, 339 DC link capacitor, 407 neutral-point-clamped, 572 positive and negative sequence quantities, 527 Voltage unbalance factors (VUFs), 223, 240, 307, 308, 337, 360–363, 368 VUFs, see Voltage unbalance factor (VUFs)
643 W WAMS, see Wide-area measurement system (WAMS) Whale optimization algorithm (WOA), 232–233, 244 Wide-area measurement system (WAMS), 381–383, 385, 386, 390, 418 WOA, see Whale optimization algorithm (WOA)