Control of Smart Buildings: An Integration to Grid and Local Energy Communities (Studies in Infrastructure and Control) 9811903743, 9789811903748

Citation preview

Studies in Infrastructure and Control

Anuradha Tomar · Phuong H. Nguyen · Sukumar Mishra Editors

Control of Smart Buildings An Integration to Grid and Local Energy Communities

Studies in Infrastructure and Control Series Editors Dipankar Deb, Department of Electrical Engineering, Institute of Infrastructure Technology Research and Management, Ahmedabad, Gujarat, India Akshya Swain, Department of Electrical, Computer & Software Engineering, University of Auckland, Auckland, New Zealand Alexandra Grancharova, Department of Industrial Automation, University of Chemical Technology and Metallurgy, Sofia, Bulgaria

The book series aims to publish top-quality state-of-the-art textbooks, research monographs, edited volumes and selected conference proceedings related to infrastructure, innovation, control, and related fields. Additionally, established and emerging applications related to applied areas like smart cities, internet of things, machine learning, artificial intelligence, etc., are developed and utilized in an effort to demonstrate recent innovations in infrastructure and the possible implications of control theory therein. The study also includes areas like transportation infrastructure, building infrastructure management and seismic vibration control, and also spans a gamut of areas from renewable energy infrastructure like solar parks, wind farms, biomass power plants and related technologies, to the associated policies and related innovations and control methodologies involved.

More information about this series at https://link.springer.com/bookseries/16625

Anuradha Tomar · Phuong H. Nguyen · Sukumar Mishra Editors

Control of Smart Buildings An Integration to Grid and Local Energy Communities

Editors Anuradha Tomar Department of Instrumentation & Control Engineering Netaji Subhas University of Technology New Delhi, India

Phuong H. Nguyen Department of Electrical Engineering Eindhoven University of Technology Eindhoven, The Netherlands

Sukumar Mishra Department of Electrical Engineering Indian Institute of Technology Delhi New Delhi, India

ISSN 2730-6453 ISSN 2730-6461 (electronic) Studies in Infrastructure and Control ISBN 978-981-19-0374-8 ISBN 978-981-19-0375-5 (eBook) https://doi.org/10.1007/978-981-19-0375-5 © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 This work is subject to copyright. All rights are solely and exclusively licensed 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 Singapore Pte Ltd. The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721, Singapore

Preface

Energy systems are an essential part of buildings and facilities, which are associated with high costs, and considered key success factor of businesses and services produced from the building or facility. The next-generation electric grid needs to be smart and sustainable to simultaneously deal with the ever-growing global energy demand and achieve environmental goals. Building energy management systems are commonly used to automate all services and functions within the building, which include energy management. This book provides an overview of how efficient building energy management can be done. It also includes the grid-interactive building, their control, energy management, and optimization techniques to promote better understanding among researchers and business professionals in the utility sector and across industries. The experiences and research work shared helps the readers in enhancing their knowledge in the field of renewable energy, power engineering, building energy management, demand, and supply management and learn the technical analysis of the same in an insightful manner. The book is divided into three sections, namely, Building Energy Management, Local Energy Community Energy Management, and Grid Interactive Building Energy Management. Section I deals with the introduction to smart building energy management systems and possible ways to achieve zero-carbon energy systems, energy monitoring, and intelligent smart buildings. It also describes how transmission and distribution networks are modeled together as an integrated network and used to do steady-state operation analysis in order to assess the interaction of these two networks. Furthermore, the influence of the increasing amount of imbalance at distribution level on the transmission network that is evoked by the increase of highly variable resources and loads at distribution level is also investigated. It focuses on solutions and technologies that can monitor and control the energy use of the building. Section I also discusses the demand side management, peak load management, and flexibility assessment. Various demand side management programs and their applications have also been covered in this section. Section II provides an overview and elaborates on the concept of local energy communities, local energy communities, and highlights the impact on consumption and performance. It also discusses identified benefits and challenges/barriers to their v

vi

Preface

further development. It also considers the operational issues and key challenges related to smart buildings integrated with local energy communities. This section includes the energy solutions that can address the key challenges, such as reducing energy consumption during peak loading conditions. This section also provides an overview of intelligent local energy community buildings of the future from a range of perspectives and application for building services with the local energy community. Section III provides an overview of the design, equipment, and control techniques that facilitate the interactions and summarizes the methods by which grid-interactive building load can be made flexible. AI and multi-agent-based optimization techniques have been shown for optimizing the electricity consumption in a grid-interactive building community. This section also describes how grid-interactive efficient buildings can both reduce net demand and benefit the grid through more flexible loads. This book gives invaluable insights to its readers. All the chapters in each section conclude with a case study for better understanding of the reader. It features contributions from key opinion leaders, successful researchers, and academicians. New Delhi, India Eindhoven, The Netherlands New Delhi, India

Anuradha Tomar Phuong H. Nguyen Sukumar Mishra

Contents

1

An Introduction to Smart Building Energy Management . . . . . . . . . . Anu Prakash, Ashish Shrivastava, and Anuradha Tomar

2

The Influence of the Increasing Penetration of Photovoltaic Generation on Integrated Transmission-Distribution Power Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Maria Kootte and Cornelis Vuik

1

15

3

Building Energy Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Nor Azuana Ramli and Mel Keytingan M. Shapi

37

4

Demand-Side Management and Peak Load Reduction . . . . . . . . . . . . Hongming Yang, Jingshu Yang, Sheng Xiang, Yan Xu, and Yibo Wang

75

5

Demand Response in Smart Buildings . . . . . . . . . . . . . . . . . . . . . . . . . . . 121 B. Rajanarayan Prusty, Arun S. L., and Pasquale De Falco

6

Building Services with the Local Energy Community—Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 Y. P. Chawla

7

Energy Solutions for Smart Buildings Integrated with Local Energy Communities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157 Shalika Walker, Pedro P. Vergara, and Wim Zeiler

8

Towards Advanced Technologies for Smart Building Management: Linking Building Components and Energy Use . . . . . 179 Ghezlane Halhoul Merabet, Mohamed Essaaidi, Hanaa Talei, and Driss Benhaddou

9

Applications to Building Services with the Local Energy Community . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203 Minghao Xu, Furong Li, Chenghong Gu, Kang Ma, Renjie Wei, Junlong Li, and Andrew Shea vii

viii

Contents

10 Optimization in Grid-Interactive Buildings . . . . . . . . . . . . . . . . . . . . . . 231 Xiaolong Jin, Xiaodan Yu, Yihan Lu, Hongjie Jia, and Yunfei Mu 11 Cost-Benefit and Short-Term Power Flow Analysis of Grid Integrated Residential Photovoltaic-Battery Energy System . . . . . . . 251 Mohamed J. M. A. Rasul, Naleen de Alwis, and Mohan Lal Kolhe

Editors and Contributors

About the Editors Dr. Anuradha Tomar currently working as Assistant Professor in Instrumentation & Control Engineering Division of Netaji Subhas University, Delhi, India. Dr. Tomar has completed her Postdoctoral research in Electrical Energy Systems Group, from Eindhoven University of Technology (TU/e), the Netherlands and has successfully completed European Commission’s Horizon 2020, UNITED GRID and UNICORN TKI Urban Research projects. She has completed her Ph.D. in Electrical Engineering from Indian Institute of Technology Delhi, India. Her areas of research interest are Operation & Control of Microgrids, Photovoltaic Systems, Renewable Energy based Rural Electrification, Congestion Management in LV Distribution Systems, Artificial Intelligent & Machine Learning Applications in Power System, Energy conservation and Automation. She has authored or co-authored many research papers in Journals/Conferences of repute. She is an Editor for books with International Publication like Springer, Elsevier. Dr. Tomar is Senior member of IEEE, Life member of ISTE, IETE, IEI, and IAENG. Dr. Phuong H. Nguyen is Associate Professor with the research group Electrical Energy Systems at the TU/e department of Electrical Engineering. His research interests include applications of ICT in smart energy systems, distributed state estimation, control and operation of the power system, distributed and computational intelligence, and applications in the future power delivery system. He is Founder of the digital Power & Energy Systems lab (digi-PES) which aims to enable an energy transition from micro-energy grids towards the future integrated energy system. This laboratory environment is a cyber-physical ecosystem for students and researchers to explore innovations in various energy-related aspects of (but not limited to) nano/micro-grids, local energy communities, local flexibility/energy markets, optimal power/energy flow, and congestion management. Hand-in-hand with emerging (big) data and Internet of Things (IoT) domains, such research

ix

x

Editors and Contributors

provides a foundation for comprehensive data-driven and inter-dependency models of energy system integration. Dr. Sukumar Mishra is Professor and Associate Dean R&D at Indian Institute of Technology Delhi and has won many accolades throughout his academic tenure of 27 years. He has been Recipient of Young Scientist award by Orissa Bigyan Academy, INSA medal for young scientist, INAE young engineer award, INAE silver jubilee young engineer award, The Samanta Chandra Shekhar award, Bimal Bose award and NASI-Reliance Platinum Jubilee award. He has been selected as Mission Innovation National Champion (2019) under the mission innovation initiative to accelerate clean energy in India. He is also Fellow of the NASI, INAE, IET, IETE and IE. Professor Mishra is Great Enthusiast of innovation and recently has incorporated a new company named SILOV SOLUTIONS PRIVATE LIMITED. Professor Mishra’s research expertise lies in the field of power systems, power quality studies, renewable energy and smart grid. Professor Mishra has authored more than 80 IEEE transactions/journals, 30 IET journals and 30 other international journal papers.

Contributors Arun S. L. School of Electrical Engineering, Vellore Institute of Technology, Vellore, India Driss Benhaddou Department of Engineering Technology, University of Houston, Houston, TX, USA Y. P. Chawla JERC/MoP Government of India, New Delhi, India Naleen de Alwis Faculty of Engineering and Management, Ocean University of Sri Lanka, Colombo, Sri Lanka Pasquale De Falco University of Naples Parthenope, Naples, Italy Mohamed Essaaidi SSL, IT Rabat Center, ENSIAS—Mohammed V University, Rabat, Morocco Chenghong Gu Faculty of Engineering and Design, University of Bath, Bath, UK Hongjie Jia Key Laboratory of Smart Grid of Ministry of Education, Tianjin University, Tianjin, China Xiaolong Jin Key Laboratory of Smart Grid of Ministry of Education, Tianjin University, Tianjin, China Mohan Lal Kolhe Faculty of Engineering and Science, University of Agder, Kristiansand, Norway Maria Kootte Delft Institute of Applied Mathematics, Delft University of Technology, Delft, Netherlands

Editors and Contributors

xi

Furong Li Faculty of Engineering and Design, University of Bath, Bath, UK Junlong Li Faculty of Engineering and Design, University of Bath, Bath, UK Yihan Lu Key Laboratory of Smart Grid of Ministry of Education, Tianjin University, Tianjin, China Kang Ma Faculty of Engineering and Design, University of Bath, Bath, UK Ghezlane Halhoul Merabet SSL, IT Rabat Center, ENSIAS—Mohammed V University, Rabat, Morocco; Department of Engineering Technology, University of Houston, Houston, TX, USA Yunfei Mu Key Laboratory of Smart Grid of Ministry of Education, Tianjin University, Tianjin, China Anu Prakash Department of Electrical Engineering, Manipal University Jaipur, Jaipur, Rajasthan, India B. Rajanarayan Prusty School of Electrical Engineering, Vellore Institute of Technology, Vellore, India Nor Azuana Ramli Centre for Mathematical Sciences, College of Computing & Applied Sciences, Universiti Malaysia Pahang, Lebuhraya Tun Razak, Gambang, Pahang, Malaysia Mohamed J. M. A. Rasul Faculty of Engineering and Management, Ocean University of Sri Lanka, Colombo, Sri Lanka Mel Keytingan M. Shapi Electrical Engineering Section, Universiti Kuala Lumpur British Malaysian Institute, Gombak, Selangor, Malaysia Andrew Shea Faculty of Engineering and Design, University of Bath, Bath, UK Ashish Shrivastava Skill Faculty of Engineering and Technology, Shri Vishwakarma Skill University, Gurugram, Haryana, India Hanaa Talei School of Science and Engineering, Alakhawayn University, Ifrane, Morocco Anuradha Tomar Department of Instrumentation & Control Engineering, NSUT, New Delhi, India Pedro P. Vergara Delft University of Technology, Delft, The Netherlands Cornelis Vuik Delft Institute of Applied Mathematics, Delft University of Technology, Delft, Netherlands Shalika Walker Eindhoven University of Technology, Eindhoven, The Netherlands Yibo Wang Wuhan Kemov Electric Co, Ltd, Wuhan, China Renjie Wei Faculty of Engineering and Design, University of Bath, Bath, UK

xii

Editors and Contributors

Sheng Xiang School of Electrical and Information Engineering, Changsha University of Science and Technology, International Joint Laboratory of Ministry of Education for Operation and Planning of Energy Internet Based on Distributed Photovoltaic-Storage Energy, Hunan Provincial Engineering Research Center for Electric Transportation and Smart Distribution Network, Changsha, China Minghao Xu Faculty of Engineering and Design, University of Bath, Bath, UK Yan Xu School of Electrical and Information Engineering, Changsha University of Science and Technology, International Joint Laboratory of Ministry of Education for Operation and Planning of Energy Internet Based on Distributed Photovoltaic-Storage Energy, Hunan Provincial Engineering Research Center for Electric Transportation and Smart Distribution Network, Changsha, China Hongming Yang School of Electrical and Information Engineering, Changsha University of Science and Technology, International Joint Laboratory of Ministry of Education for Operation and Planning of Energy Internet Based on Distributed Photovoltaic-Storage Energy, Hunan Provincial Engineering Research Center for Electric Transportation and Smart Distribution Network, Changsha, China Jingshu Yang School of Electrical and Information Engineering, Changsha University of Science and Technology, International Joint Laboratory of Ministry of Education for Operation and Planning of Energy Internet Based on Distributed Photovoltaic-Storage Energy, Hunan Provincial Engineering Research Center for Electric Transportation and Smart Distribution Network, Changsha, China Xiaodan Yu Key Laboratory of Smart Grid of Ministry of Education, Tianjin University, Tianjin, China Wim Zeiler Eindhoven University of Technology, Eindhoven, The Netherlands

Chapter 1

An Introduction to Smart Building Energy Management Anu Prakash, Ashish Shrivastava, and Anuradha Tomar

Abstract In today’s fast growing life, buildings must have an effective energy management system to make compact and convenient atmosphere with low investment and high utilizations. A smart building energy management system (BEMS) would regulate the heating system, boilers, and pumps as a fundamental function, then locally control the thermal regulation to reach the optimal needed room temperature. BEMS would regulate air conditioning in building by using cooling system. Cooling systems distributes cold air throughout the building by using fans and dampers. BEMS can also be used to regulate lights or other energy-consuming devices, as well as to log data from energy meters. Keywords Building Energy Management System (BEMS) · HVAC system · Sensors · Centralized controller · Grid-Interactive Efficient Buildings (GEBs)

1 Introduction A “SMART” building is a building having a brain or an effective controller that can gather information from sensor-based intelligent structures, use less capital for maintenance, and make complex networks simpler [1]. For smart building, architecture is more important than other integrated systems. Before 1993, a building using remote for sensors embedded system and for other integrated systems was known as Smart Building. But Moore in 1993, developed a theory based on energy management for Smart building architectural design with climate or atmospheric change and which makes its own climate. In this climate, climatic arterial conditioning is not required. A. Prakash Department of Electrical Engineering, Manipal University Jaipur, Jaipur, Rajasthan, India A. Shrivastava (B) Skill Faculty of Engineering and Technology, Shri Vishwakarma Skill University, Gurugram, Haryana, India A. Tomar Department of Instrumentation & Control Engineering, NSUT, New Delhi, India © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 A. Tomar et al. (eds.), Control of Smart Buildings, Studies in Infrastructure and Control, https://doi.org/10.1007/978-981-19-0375-5_1

1

2

A. Prakash et al.

Building energy management systems (BEMS) may be defined as computer-aided architecture empowered with an intelligent controller for the purpose of modulating and monitoring the building activities such as proper lighting as per visibility, improvement of air quality as per ventilating air properly, building’s atmosphere control as per heating properly and so on. For these types of arrangements, there are various sensors along with mechanical and electrical machineries are used in building. Computer-aided architecture of building helps in performance of all equipments accordingly. Green energy has replaced fossil fuels as the major energy source in today’s era [2]. For a modernized electrical power system, local energy sources must be added in generating units to deliver safe supply with low losses to prosumers and to get affordable electricity. Pertinent considerations for modernization of power systems must include performance, safety, and economics. So, while designing a new system, a research has been carried out to discover the main cause of equipment failure as well as curative actions to improve system functionality. Existing power systems are easier to examine manually than previous power systems due to their complex network structure. A comprehensive power systems analysis based software program is required to study accurately and quickly, as these systems are very complex in nature. So, a modern power system includes a sophisticated controllerbased computer. Here, energy management system is required towards more creative computational technique-based energy tracking platforms. This system controls, each piece of equipment operated in a smart building. Energy management solutions allow building operated equipment to be powered only when it’s needed. This reduces the wastage of energy especially from that particular area from lighting; temperature controlling or refrigerating sections haven’t been used for 24 h. Thus, optimized control improves the efficiency of mechanical equipments used in smart building. Due to minimizing maintenance expenditures and efficient usage of equipment, either mechanical or electrical, life of smart building equipments is improved by energy management system. On the other hand, recognizing issues in reduction of energy consumptions in buildings, several methods and techniques have been presented in the literature [3–6]. The architecture of buildings plays a vital role to minimize consumption on energy in mechanical and electrical equipments. An energy management system is required in smart building for balancing supply–demand ratio. To design an energy management system, literature survey is one of the most important steps. The penetration of local energy sources at supply side in energy management system increases difficulty from operational efficiency point of view. The architecture of smart building depends on the kind of structure (household, commercial, car manufacturing, etc.) and service demand (lightly or intensively served). A building can be called as from an energy perspective, if it is efficient and prudent in energy utilization. The performance of building is measured in three categories: (a) social impact, (b) environmental impact, and (c) financial impact. But capacity, safety, security, and automation are also important features to make a building “smart”. The influence on energy usage in buildings is strongly connected to major external climatic parameters as well as key inside environmental features.

1 An Introduction to Smart Building Energy Management

3

There are certain suggestions found out through literature survey for establishment of smart building, which are mentioned as: (a) (b) (c)

An arrangement of sensors to supervise equipment behavior according to their energy consumption and climatic variations. A data handling software for balancing supply and demand through handling response via sensors. An interfacing unit for supporting users.

For each component, many solutions provided in the literature have been analyzed. The building’s temperature, humidity, and other parameters are sent to the BEMS through digital and analog input signals to control running of electrical and mechanical equipments. For performance analysis, location where these equipments are set up, is also important. In that case, locations of fan or boiler or pump are also considered as input signal. According to their location, running time and energy consumption, both can be controlled by centralized controller of smart building. The operator can programme when objects turn on and off automatically.

2 Smart Building A smart building is an intelligent building technology system for managing artificial climatic condition automatically by using sensory setups, mechanical and electrical equipments sophisticatedly. Building automation, communication systems, security and safety, facility management systems, and so on are all features of these systems. A smart building with remote control was designed in 1999, and has been described in [7]. Technologies have enhanced day by day, now automated building with sensory infrastructures and computer-aided technologies help to make a building “smart”. There are lots of equipments along with sensory setup used in smart building which makes the network more complex. So, to link equipment with sensor and its smart application is only possible through computational algorithms [8–12]. The primary job in any building is to maintain climate inside the building like heating, conditioning, ventilating, and lighting which needs energy consumption. The control techniques such as automated lighting systems and ventilation control in building may help to increase building energy efficiency [13–18]. Figure 1 shows the basic concept of smart building that includes sensory connection between each and every electrical and mechanical system. Then individual setups with sensors are connected with controller that control input and output of equipments. Hence, environment of building must be comfortable without using air conditions or blowers. Hybrid Electrical Energy Storage (HEES) systems works on shifting peak power demands for each Hybrid Electrical Energy Storage (HEES) element. It converts peak power demands by converting into electrical energy [19]. The main existing EES components are batteries but it is costly, bulky, and having poor performance [20]. Smart buildings also provide a better air quality system by using proper ventilating

4

A. Prakash et al.

Fig. 1 Basic concept of smart building

machineries, temperature controlling equipments for regulating ambient temperature, and humid controlling system by using dampers [21, 22]. Thermal storages are benefitted for commercial and residential buildings due to shifting in peak and energy price responsiveness [23]. Peak power consumption is generally minimized by smart building thermal storage systems [24–27]. The temperature of the outside air also helps to forecast and regulate the heating demand of smart buildings [28]. For solar radiation and conduction accounting through the wall and window in buildings, there are several methods in smart building using computation tools [28–30]. As a result, in order to improve smart building thermal storage, smart gadgets must be associated with mechanical machineries, such as thermostropic windows provide reversible transmission behavior. The most essential aspects of smart buildings are the inhabitants and their perception of comfort. It should be the primary goal of using smart buildings. Therefore, the preferences of the people living in the building are prioritized by automation of the building [31]. Every system of a smart building is unique in its own way. An architectural document is designed by an architecture team. This document contains all simplest and complex network system in details in a single file. Simple networks are electrical connection paths for cables or cabling in every area. Complex networks are machinery setup rooms, controller rooms (system databases), communications networks, and so on. The main aspects of smart building technology is automation, sensors and its networking, analytical software, monitored controller for energy management, and communication protocols are all aspects of smart building technology. These essential building systems and technologies are elaborated in the following sections. It also explains the technological advancements that have resulted in smart technologies systems, costing, growth of energy market, and equipment’s life.

1 An Introduction to Smart Building Energy Management

5

3 Modern Power System In modern power systems, a substation is needed for gathering data from different buildings/sites via wired or wireless communications such as Ethernet or satellite system. Here, advanced electronic gadgets are connected to make convenient and safe supply. So, this system can send notification in any manner like using alarm. When compared to individual controllers for each system, a BEMS can save 10–20% on energy. BEMS is not useful when buildings have wrong architectural system with improper management [32]. It is only useful in controlling of properly managed big complex structures and multi-building complexes. Huge organizations, such as municipal governments, hospitals, hotels, and even remote area’s organizational buildings utilize them to control buildings that are dispersed across large regions. Intelligent substations exist in modern systems, which may be inspect all issues related to the plant and track it down. Wireless communication helps in making the system less complex and economical due to reduction in wiring cost. It might be used to regulate nearly any size of structure, although the benefits of better management are most noticeable in big, dispersed, and complicated sites/buildings. It’s critical to have effective user interfaces while using a BEMS.

4 Features of Energy Management System in Buildings Nowadays, there are many ways to access BEMS via wired or wireless networks through web browsers in computer, laptops, palmtops, and other smart devices like mobile phones [33]. BEMS must be routinely serviced in order to maximize internal conditions and save money in the long run. BEMS settings should be reviewed at least once a month to ensure that they are up to date and match the needs of the building.

4.1 Basic Requirements It is imperative to focus on the following while assessing the system: • Ensuring the integrity of complex network systems at the time of installation and the time of running. • Sensors: Check their accuracy and appropriateness for their intended use. • Actuators: Check the outputs from the controller and ensure equipment’s running capacity with low losses. • Digital inputs: Verify the functioning of all digital inputs. Calibrate or modify switching devices if needed. • Supervisors: Check the battery supply and resume automatically after a power outage.

6

A. Prakash et al.

• Record-keeping: Keep supervisory control over BEMS by tracking variations or modifications to software, networks, problems discovered, and maintenance conducted. Controls are extremely important and are the foundation of a building’s performance. The ideal maintenance schedules must be established at the start of the project. The customer should have the freedom for choosing services as per their requirement.

4.2 Benefits of Energy Management in Buildings There are major benefits of a BEMS system which makes it user friendly with controlled performance of the equipments • Providing low-cost supply to the building. • Stability and Reliability. • Artificial Environment of the building may not require Air Conditioners or Blowers. • Ensuring a comfort zone to the customer with controlled I/O of mechanical or electrical systems. • Energy-saving control features that help to save money on total energy costs (e.g., weather compensation). • Data monitoring and storing previous data for energy management system. • Automatic detection of sudden interruption or failure of equipment through system alarm. • Identifying critical and non-critical loads • Manages neighboring plants, places, or other buildings with ease.

4.3 Liabilities of Energy Management on Buildings/Users The BEMS has some liabilities too, such as: • The effectiveness of a BEMS is determined by the individuals who utilize it. Although the BEMS requires less maintenance but the staff must still be trained properly. All prosumers must be trained how to handle equipments in case of sudden failure. • Mobile access: The consumers must be trained to utilize features of the BEMS wirelessly. This can be aided by ensuring that the consumers have quick access to mobile devices. The better the understanding, the better people will be able to save energy. This will provide instruction on both the BEMS hardware and software. • BEMS is a strong tool for building management, but it’s only as useful as the people that use it! All employees who have access to the BEMS should get expertise controlling the building by utilizing it on a regular basis. Alerts are established in

1 An Introduction to Smart Building Energy Management

7

most BEMS so that the consumers must be aware about their actions after system start alarming. This may frequently lead to the discovery of issues that would otherwise go unreported, such as high energy use or plants that are kept running continually. Energy meters linked to a BEMS provide the output of consumed energy in real time and store data numerically and visually for recording of energy management of the building. As a result, through trend recording performance, BEMS may increase management information and benefit future planning and costing. This may also inspire employees to be more conscious of their energy usage. Improvements of 10–20% in energy efficiency are not uncommon. However, in order to save as much money as possible, it’s critical to determine the appropriateness of existing structures and equipment. According to its design, it receives signal from utility area first, then send it to the controller and the transmit controller’s output to the building energy management system. As per requirement, warming and lighting in any utility area of a smart building, BEMS controls temperature in certain utility areas, lower the intensity of light. BEMS may also gather data from utility and energy resources then monitor, aggregate, and interpret it at low level. Despite its value, BEMS are still largely utilized in a reactive way to resolve issues after they have occurred. For forecasting and optimizing of building activities, these energy management systems can play vital role in future buildings.

5 Computer-Aided Technique-Based Analytics Platform Computer-aided technique-based analytics platform is the newest advancement in smart building sectors. Basically, it works for making pleasant environment in whole building according to the utility’s requirements. So, it can save more energy and charge battery storage system rapidly. Facility managers may use load shedding schedules to actively and deliberately minimize energy consumption (and hence lower utility bills) by understanding where, when, and how their facility consumes energy [34]. Real-time exclusive information of smart building can be obtained by computational method sensors installed throughout the building. These sensors can remotely monitor a variety of processes [35–39], containing: • • • • • • •

Specific mechanism system Lighting systems Heating, cross-ventilation, and air temperature conditioning (HVAC) system Indoor air quality (ventilation) systems Cold storage (Refrigerating) systems Geyser (water heating) systems Heat controlling Pumps, among other things

8

A. Prakash et al.

The computational technique platforms are gathered data from different areas, make co-relation between them, then measuring or monitoring the data to help the end users. Depending on consumer’s purposes, there is some important gathered information like: • Total energy used by electrically linked systems and equipment: Some equipment in smart building systems is always on, while others are just used once in a while. In any case, it’s critical to comprehend both the building’s entire daily electrical usage and the function that specific gadgets use energy in general. [40, 41]. • Behavior of occupants: Occupants’ energy utilization pattern, basic needs pattern like air condition and other equipments that making home comfortable are the factors to consider for measuring energy efficiency, although they are a bigger factor for particular types of buildings and saving techniques [42, 43]. • Patterns of energy consumption: Some cost-cutting techniques require consumers to understand when and how these facilities consumes energy, as well as seeking to change those patterns for benefits. • Time-of-use costs are levied by the utility company: A typical approach to save money is to shift energy consumption away from high-priced utility-set times. Platforms based on computational techniques can aid in identifying the most cost-effective periods for energy consumption. • Factors that change with the seasons: The building’s energy usage may fluctuate in predictable ways over time, which an analytics platform using computational techniques may take into account when providing recommended solutions. • Weather: weather condition forecasting is most important factor for energy generation in smart building. So, utilization of energy source must dependent on weather conditions. • Indoor air quality improvement, ventilation, and indoor ambience conditioning (HVAC) systems: Using HVAC systems enhance usage of electricity. Consumers may reduce the usage of HVAC energy in other days except its very hot or very cold. It is feasible to design a reliable cost-effective energy management system for smart building with the proper data variables—collected, correlated, and evaluated by the computational method. In commercial building, utilization of energy cannot be fixed. Actually, there are dynamic changes in electricity utilization as per consumer’s certain demands. The major objective of a building management system for many facilities managers is to save money on energy costs. BEMSs are unable to monitor and measure some components, such as humidity, CO2 levels, temperature in air and water, luminosity, and conductivity, which are frequently important in maximizing energy efficiency [44–46]. The purpose of smart building is not only to automate the building but also to reduce CO2 emission, promote local energy communities, remove extra burden from grid, provide low cost electricity, and, most importantly, save electricity to reduce losses. Figure 2 shows how smart building works independently. It shows some basic needs of a building such as fire alarm, water management, and so on. So, sensors

1 An Introduction to Smart Building Energy Management

9

Fig. 2 Working of smart building

collect these data and send it to data monitoring system. After processing these, data demand of energy is cleared and types of load identified. Then controlling demand side by distinguishing between critical and non-critical loads, equipments running time, and modifying output as per weather is done by demand controller. There is another main task of demand controller to identify required supply for fulfilling the demand. The whole system is connected though sensory architecture for safety purposes. If renewable energy is unable to fulfill demand, then grid system will be in working mode. It has a central computer that can control all electronic devices utilization timing and save electricity bills and its lives as well. It also connects with mechanical sensors to control climate without using artificial climatic conditioning in building by controlling humidity, temperatures, and other variables. So, basically central computer collects information, monitors the same, and then gives permission to devices. Central computer controls not only supply side only but also the demand side. So, for knowing utilization of electricity, demands must be known [47]. Demand is variable information that changes according to building operators to enhance building performance. So, demand side management is also important for energy management system. An indispensable portion in demand side management (DSM) is to reduce peak load and simplify the networks with back-up power plants. Nowadays, load management plays key role for peak load reduction. Considerable factors are huge when it comes to load management such as household appliances like heater, boiler, and air condition and so on. Additionally, to manage rapidly increasing loads, the ability of energy management system (EMS) should be increased by facility providers. Due to high penetration of local energy sources and EVs as loads, EMS must be ready to manage their loads in order to lower their energy expenses and boost the profitability of the installed DERs from an economic standpoint. Demand Side Management (DSM) promotes load management by offering several tariffs with changeable prices rather

10

A. Prakash et al.

than fixed rates [48]. With the help of power rates, clients can be encouraged to plan their energy consumption more efficiently. Tariffs can be classified according to their usage time: (1) (2) (3)

Real-time use. Critical time use. Time of use.

6 Grid-Interactive Efficient Buildings (GEBs) The electrical system, evolving across the building, to provide green energy in an economical way to the building users, is known as smart building. Due to rapid penetration of local energy sources and increasing sizes, system becomes more complicated to maintain efficiency, stability and reliability. Fortunately, as technology advances, tenants, owners, and the grid will have more flexibility in managing building and facility energy demands. To improve energy-saving strategies by employing sensory set-ups, and smart controls, GEBs best serve the requirements of inhabitants while taking the grid and environmental circumstances into account (such as peak loads and weather) [49, 50]. Greater automated optimization of buildings’ substantial energy demand and supply tasks has far-reaching electrical policy and regulatory consequences for State Energy Offices, Public Utility Commissions, utilities, and building owners and investors. The main advantages are: • Reduce peak loads, regulate demand ramping, and offer grid services via GEBs. • Improve energy efficiency by using distributed energy sources as well as renewable energy sources. GEBs are buildings that integrate and optimize DERs with the electric grid to give benefits to building owners and inhabitants as well as the electricity system’s functioning. The cornerstone of GEB is a high degree of energy efficiency (EE), which includes both passive and active electrical and mechanical components such as heating, cooling, lighting, refrigeration, cooking, and other electrical appliances and equipment. High energy efficiency is beneficial in almost all instances, regardless of the presence or absence of additional DERs or the degree of grid-interactive capabilities. Buildings (and their operators and inhabitants) and the grid are linked by GEBs, which allow for two-way signal transmission. The signals may be used to operate and monitor equipment directly, or they could be used to indicate pricing and grid conditions, causing building automation systems to respond in accordance with economic incentives and consumer preferences. Smart GEBs should use sensors, controls, and analytics to improve a building’s performance to satisfy tenant demands and offer grid services. GEBs must also be flexible, allowing them to swiftly change loads and/or draw on DERs in order to provide optimal performance.

1 An Introduction to Smart Building Energy Management

11

References 1. Ocl J, Issa RRA, Flood I (2016) Smart building energy management systems (BEMS) simulation conceptual framework. Winter Simul Conf (WSC) 2016:3237–3245. https://doi.org/10. 1109/WSC.2016.7822355 2. Attia M, Haidar N, Senouci SM, Aglzim EH (2018) Towards an efficient energy management to reduce CO2 emissions and billing cost in smart buildings. In: Consumer communications & netorking conference (CCNC) 2018 15th IEEE annual, pp 1–6 3. Hannan MA, Faisal M, Ker PJ, Mun LH, Parvin K, Mahlia TM, Blaabjerg F (2018) A review of internet of energy based building energy management systems: issues and recommendations. Access IEEE 6:38997–39014 4. Kim S, Lee T, Kim SM, Lee S, Park S (2019) Design of intelligent energy management system based on user schedule. In: Consumer electronics (ICCE) 2019 IEEE international conference, pp 1–3 5. Vahidinasab V, Ardalan C, Mohammadi-Ivatloo B, Giaouris D, Walker SL (2021) Active building as an energy system: concept challenges and outlook. Access IEEE 9:58009–58024 6. Zakharov AA, Zakharova IG, Romazanov AR, Shirokikh AV (2018) The thermal regime simulation and the heat management of a smart building. In: Tyumen state university herald. Physical and mathematical modeling. Oil, gas, energy, vol 4, pp 105 7. Control Solutions (2015) Ultimate guide to building automation system (BAS), control solutions Inc. blog, January 22, controlyourbuilding.com/blog/entry/the-ultimateguide-to-buildingautomation 8. Aznavi S, Fajri P, Asrari A, Khazaei J (2019) Two-stage energy management of smart homes in presence of intermittencies. In: 2010 IEEE transportation electrification conference and expo (ITEC), Detroit, USA, pp 1–5 9. Arbib MA (2012) Brains, machines and buildings: towards a neuromorphic architecture. Intell Build Int 147–168 10. Wong SL, Wan KKW, Lam TNT (2010) Artificial neural networks for energy analysis of office buildings with daylighting. Appl Energy 87(2):551–557. https://doi.org/10.1016/j.apenergy. 2009.06.028 11. Rehm M, Ade R (2013) Construction costs comparison between green and conventional office buildings. Build Res Inf 41(2):198–208 12. Liu K, Nakata K, Harty C (2019) Pervasive informatics: theory, practice and future directions. J Intell Build Int 13. Yu H, Pan J, Xiang A (2005) A multifunction grid connected PV system with reactive power compensation for the grid. Sol Energy 79(1):101–106 14. Muratori M, Rizzoni G (2015) Residential demand response: dynamic energy management and time-varying electricity pricing. IEEE Trans Power Syst 31(2):1108–1117 15. Richardson P, Flynn D, Keane A (Jun. 2012) Local versus centralized charging strategies for electric vehicles in low voltage distribution systems. IEEE Trans Smart Grid 3(2):1020–1028 16. Burke W, Auslander D (2009) Residential electricity auction with uniform pricing and cost constraints. North American power symposium, Starkville, USA, pp 1–6 17. Waraich RA, Galus MD, Dobler C, Balmer M, Andersson G, Axhausen KW (2013) Plug in hybrid electric vehicles and smart grids: investigations based on a micro simulation. Transp Research Part C: Emerg Tech-Nologies 28:74–86 18. Liu Y, Wu L, Li J (2019) Peer-to-peer electricity trading in distribution systems of the future. Appl Energy 32:2–6 19. IEA (2012) Energy technology perspectives 2012 pathways to a clean energy system. In: Energy technol. perspect. 2012 pathways to a clean energy system, pp 1–12 20. Alam MR, Sthilaire M, Kunz T (2019) Peer-to-peer energy trading among smart homes. Appl. Energy 238:1434–1443 21. Mills E (2011) Building commissioning: a golden opportunity for reducing energy costs and greenhouse gas emissions in the United States. Energy Effic 4(2)

12

A. Prakash et al.

22. Ock J, Issa RRA, Flood I (2016) Smart building energy management systems (BEMS) simulation conceptual framework. In: 2016 winter simulation conference (WSC), pp 3237–3245 https://doi.org/10.1109/WSC.2016.7822355 23. NIST (2012) Framework and roadmap for smart grid interoperability standards, release 2.0, national institute of standards and technology (NIST) 24. Weidlich A, Karnouskos S (2009) Integrating smart houses with the smart grid through web services for increasing energy efficiency. In: 10th IAEE European conference, September 8, 2009 25. Morvaj B, Lugaric L, Krajcar S (2011) Demonstrating smart buildings and smart grid features in a smart energy city. In: 3rd international youth conference on energetics (IYCE ’11) 26. Ray SD, Gong NW, Glicksman LR, Paradiso JA (2014) Experimental characterization of fullscale naturally ventilated atrium and validation of CFD simulations. Energy Build 69:285–291 27. . Patti E, Acquaviva A, Macii E (2013) Enable sensor networks interoperability in smart public spaces through a service oriented approach. In: Proceedings of the 5th IEEE international workshop on advances in sensors and interfaces (IWASI ’13), pp 2–7, IEEE, Bari, Italy, June 2013 28. Strasser T, Andren F, Kathan J et al (2015) A review of architectures and concepts for intelligence in future electric energy systems. IEEE Trans Industr Electron 62(4):2424–2438 29. Lee EK (2016) Advancing building energy management system to enable smart grid interoperation. Int J Distrib Sens Netw Article ID 3295346. https://doi.org/10.1155/2016/329 5346. 30. O’Neill Z, Pang X, Shashanka M, Haves P, Bailey T (2013) Model-based real-time whole building energy performance monitoring and diagnostics. J Build Perform Simul 83–99 31. Oldewurtel F, Parisio A, Jones CN, Gyalistras D, Gwerder M, Stauch V, Lehmann B, Morari M (2012) Use of model predictive control and weather forecasts for energy efficient building climate control. Energy Build 15–27. https://doi.org/10.1016/j.enbuild.2011.09.022. 32. Privara S, Siroky J, Ferkl L, Cigler J (2010) Model predictive control of a building heating system: The first experience. Energy and Build 564–572 33. Zhao P, Suryanarayanan S, Simoes MG (2013) An energy management system for building structures using a multi-agent decision-making control methodology. IEEE Trans Ind Appl 49:1–9 34. Sun B, Luh PB, Jia QS, Jiang Z, Wang F, Song (2013) Building energy management: integrated control of active and passive heating, cooling, lighting, shading, and ventilation systems. IEEE Trans Autom Sci Eng 10:588–602 35. Petrushevski F, Montazer M, Seifried S, Schiefer C, Zucker G, Preindl T, Suter G, Kastner W (2018) Use cases for improved analysis of energy and comfort related parameters based on BIM and BEMS data. Adv Comput Strat Eng 10864:391 36. Hurtado L, Nguyen P, Kling W, Zeiler W (2013) Building energy management systemsoptimization of comfort and energy use. In: 48th international universities power engineering conference, UPEC 37. Manic M, Wijayasekara D, Amarasinghe K, Rodriguez-Andina JJ (2016) Building energy management systems: the age of intelligent and adaptive buildings. IEEE Ind Electron Mag 25–39 38. Dong B, O’Neill Z, Li Z (2014) A BIM-enabled information infrastructure for building energy fault detection and diagnostics. Autom Constr 197–211 39. Sun Y, Wu T, Li X, Guizani M (2017) A rule verification system for smart buildings. IEEE Trans Emerg Top Comput 367–379 40. Kamienski C, Borelli F, Biondi G, Rosa W, Pinheiro I, Zyrianoff I, Sadok D, Pramudianto F (2015) Context-aware energy efficiency management for smart buildings. IEEE 2nd world forum on internet of things (WF-IoT) 41. Bashir MR, Gill AQ (2016) Towards an IoT big data analytics framework: smart buildings systems. In: IEEE 2nd international conference on data science and systems (HPCC/SmartCity/DSS)

1 An Introduction to Smart Building Energy Management

13

42. Shen Z, Zhang T, Jin J, Yokota K, Tagami A, Higashino T (2019) ICCF: an informationcentric collaborative fog platform for building energy management systems. IEEE Access 7:40402–40415. https://doi.org/10.1109/ACCESS.2019.2906645 43. Pérez-Lombard L, Ortiz J, Pout C (2008) A review on buildings energy consumption information. Energy Build 40:394–398 44. Manic M, Wijayasekara D, Amarasinghe K, Rodriguez-Andina JJ (2016) Building energy management systems: the age of intelligent and adaptive buildings. IEEE Ind Electron Mag 10(1):25–39 45. Pan J, Jain R, Paul S, Vu T, Saifullah A, Sha M (2015) An Internet of Things framework for smart energy in buildings: designs prototype and experiments. IEEE Internet Things J 2(6):527–537 46. Shen Z, Yokota K, Jin J, Tagami A, Higashino T (2018) In-network self-learning algorithms for BEMS through a collaborative fog platform. In: International conference on advanced information networking and applications (AINA), pp 1162–1169 47. Faruque MAA, Vatanparvar K (2016) Energy management-as-a-service over fog computing platform. IEEE Internet Things J 3(2):161–169 48. Shiaa M, Falcarin P, Pastor A, Lecue F, Silva E, Ferreira Pires L (2008) Towards the automation of the service composition process: case study and prototype implementations. ICT-MobileSummit 2008 conference proceedings, Stockholm, pp 10–12 49. Goldman C, Murphy S, Hoffman I, Frick NM, Leventis G, Schwartz LC (2018) The future of U.S. Electricity efficiency Program spending and savings projections to2030. In: Electricity markets and policy group, energy analysis and environmental impacts division, Lawrence Berkeley National Laboratory: Berkeley, CA, 2018. 50. Neukomm M, Nubbe V, Fares R (2019) Grid-interactive efficient buildings, United States: N.p.,2019. https://doi.org/10.2172/1508212

Chapter 2

The Influence of the Increasing Penetration of Photovoltaic Generation on Integrated Transmission-Distribution Power Systems Maria Kootte and Cornelis Vuik Abstract Power system simulations should be adapted to be applicable to the trends that are currently evoked by the energy transition. This transition is pushing our power system from a traditional hierarchical system to a modern interactive system. In order to keep the supply and transport of energy safe and reliant, we need to change the way we perform power system simulations. This requires a comprehensive framework in which both transmission and distribution systems are simultaneously analyzed. This chapter describes how transmission and distribution networks are modeled together as an integrated network and used to do steady-state operation analysis in order to assess the interaction of these two networks. Furthermore, we investigate the influence of the increasing amount of imbalance at distribution level on the transmission network that is evoked by the increase of highly variable resources and loads at distribution level. This influence is not taken into account in traditional power system simulations as power networks are analyzed on its own. We show that the hybrid network representation is a powerful tool to analyze modern power systems and that the effects of increased PV penetration under normal operating conditions are limited.

1 Introduction Electrical power systems are responsible for generation, transmission, and distribution of electrical power. In traditional power flow computations, the transmission and distribution networks are modeled and solved separately, where each system operator is responsible for safe operation and planning of their own electrical power system. Both transmission and distribution system operators (TSO and DSOs) use steady-state power flow computations to obtain insight into the state of their netM. Kootte (B) · C. Vuik Delft Institute of Applied Mathematics, Delft University of Technology, Delft, Netherlands e-mail: [email protected] C. Vuik e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 A. Tomar et al. (eds.), Control of Smart Buildings, Studies in Infrastructure and Control, https://doi.org/10.1007/978-981-19-0375-5_2

15

16

M. Kootte and C. Vuik

work [1]. The electrical power system is a large interconnected system, containing one transmission network and multiple distribution networks. In order to separately analyze these networks, TSOs model the distribution network as an equivalent load bus, while DSOs model the transmission network as an equivalent generator bus [2]. These assumptions were justified as bulk power was generated at transmission level and transported hierarchically toward end consumers such as households and industries. However, the electricity landscape is changing. A major driver for this change is the rapid increase of renewable resources such as wind and photovoltaic (PV) power, connected at distribution level. These resources are also called distributed energy resources (DERs) [2]. The changes induce bi-directional power flow between transmission and distribution network and increased imbalance of the electrical power system [3]. Transmission networks are, in general, balanced networks while distribution networks are, in general, unbalanced networks. This imbalance is a result of different phase loading levels and untransposed distribution cables [4]. The main driver for the increase of DERs is the high level of PV power penetration in urban residential areas. More and more households are motivated to place PV panels on their rooftop out of care for the planet, financial incentives, and/or lowering purchase and installation costs [5]. The government supports this trend as it is in line with the global sustainable development goals, but it puts extra pressure on the electricity grid as it will be difficult for grid operators to maintain safe voltage levels. PV panels can be classified as residential and utility panels. Utility panels can be used to replace traditional bulk power plants at transmission level. Residential panels are small and located at distribution level as they are mainly placed on rooftops in urban areas. Residential PV panels produce active power only as voltage regulatory purposes not allow them to produce reactive power [5]. This can lead to a voltage drop in the network to an undesirable level. On top of that, as distribution networks are unbalanced networks, a voltage drop can result in different phase variations. This can again result in a certain degree of imbalance that is harmful for the network [6]. DSOs operate their system between a certain bandwidth of allowed voltage levels and degree of imbalance [7]. Increased PV penetration at distribution level can also affect transmission systems. Traditional steady-state analysis, done on separate electrical power systems, is not able to capture the effects that the networks have on each other [3]. Therefore, there is a need for integrated power flow simulations that is capable of analyzing the effects that these networks have on each other. In this work, we are going to present a hybrid integrated transmission-distribution network model as a single entity. This hybrid network model represents the transmission network as a single-phase network model, while the distribution network is represented by a three-phase network model [8]. The interface description is not the only challenge that should be considered when doing integrated power flow computations. Power flow solvers have been developed so far for separate power flow analysis [4]. These methods cannot one-on-one be translated to run power flow computations on integrated network models. It is important to consider the efficiency and robustness of the existing separate methods when they are used to solve the integrated power flow problem.

2 The Influence of the Increasing Penetration of Photovoltaic Generation …

17

The integrated model of an electrical power system can be used to study the sensitivities toward PV power penetration such as steady-state voltage stability, induced imbalance, and reduced power losses. The results are compared with a full threephase representation of the integrated network. This chapter contains the following contributions: (1) a description of steady-state power flow computations in (balanced) transmission and (unbalanced) distribution networks (Sect. 2); (2) a modeling approach on how these separate networks can be integrated as a hybrid network model and which techniques can be used to solve this system (Sect. 3); (3) a description of photovoltaic power generation and how it is connected to an electrical power system (Sect. 4); and (4) results of the simulation using a varying amount of PV penetration compared to a zero PV load base case and a full three-phase representation of the integrated network (Sect. 5), after which we draw conclusions (Sect. 6).

2 Steady-State Power Flow Computations The steady-state power flow problem is the problem of determining the voltages V in a network, given the specified power1 S = P + ιQ and current I [9]. V and I are related by Ohm’s Law, I = Y V , and S and V are related by S = V I ∗ . Y represents the admittance of a power cable. Power is generated in three phases leading to three sinusoidal functions that describe phase a, b, and c of the voltage and of the current. Current is never specified in an electricity system, and therefore we substitute Ohm’s Law into S = V I ∗ and get a nonlinear equation for S, which is the three-phase nonlinear power flow equation, described as follows: S p = V p (YV )∗p ,

p ∈ {a, b, c}.

(1)

We represent an electricity network as a graph consisting of buses i = 1, . . . , N representing generators, loads, and shunts and branches representing transformers and cables. During steady-state power flow computations we determine the voltages Vi of each bus given the power supply and demand Si of each bus and admittance Yi j of each branch.

2.1 Transmission Networks The transmission network is the high-voltage network, responsible for the transportation of power over large distances. It is a balanced system which means that the magnitude of the voltage (|V |) of the three phases a, b, and c are equal and that the phase shift (φ) between the voltages are equal [9]. This means that |V |a = |V |b = |V |c and φab = φbc = φca = 23 π . To simplify and speed up the computations in the transmis1

We use the subscript ι as the imaginary unit.

18

M. Kootte and C. Vuik

sion network, only Va is calculated. The other two phases are deducted from phase a. The problem of finding the voltages Vi given power Si and Yi j simplifies then to the following single-phase relationship, where a represents the phase: Sia

=

a Via I i

=

Via (Y V)ia

=

Via

N 

a

a

Y ik V k .

(2)

k=1

We use Newton-Raphson power mismatch (NR power) to compute unknown quantities at each bus i [9]. NR power computes Vi using the following power mismatch formulation: ΔSi = Ss,i − S(Vi ) ≈ 0.

(3)

Ss is the specified power, the known information at generator and load nodes: Ss = Sg − Sd , subscript g representing the generator buses and d the load buses. S(V ) is the injected power, S(V ) = V (YV )∗ . The complex power S is split into an active and reactive part and combined to form the power mismatch vector F,    ΔP Ps − P(x) = 0, F(x) = = Qs − Q(x) ΔQ 

(4)

T  where x represents the state variables xi = δi |Vi | which form the voltage in the phasor notation Vi = |V | exp (ιδ)i . We compute V in an iterative manner using the Jacobian J of the power mismatch vector: Δxν = −J −1 (xν )F(xν ), x

ν+1

ν

(5)

ν

= x + Δx ,

(6) 

where the Jacobian is represented as follows: J (x) =

∂P ∂P ∂δ ∂|V | ∂Q ∂Q ∂δ ∂|V |

 .

We repeat this until the norm of the power mismatch vector |F|∞ is lower than a certain tolerance value ε. We choose ε = 10−5 and start with a flat profile as initial guess: V 0 = 0. The Newton-Raphson algorithm is described in algorithm 1.

2.2 Distribution Networks: Three-Phase Representation Distribution systems are unbalanced: the three phases are not equal in magnitude nor in phase shift [10]. This requires us to compute all the three phases in every iteration when solving equation (1). We use Newton-Raphson Three-Phase Current Injection

2 The Influence of the Increasing Penetration of Photovoltaic Generation …

19

Algorithm 1 The Newton-Raphson iterative method 1: 2: 3: 4: 5:

Set ν = 0 and choose appropriate starting value x0 ; Compute F(x ν ); Test convergence: If |F(x ν )|∞ ≤ ε then x ν is the solution, otherwise continue; Compute the Jacobian matrix J(x ν ); Update the solution: Δx ν = −J−1 (x ν )F(x ν ) x ν+1 = x ν + Δx ν ;

6: Update iteration counter ν + 1 → ν, go to step 2.

Method (NR-TCIM) [11] to solve distribution networks. Instead of applying the standard Newton-Raphson method to power mismatches, Ohm’s law is directly used resulting in the current mismatch vector:   Re,abc  Is − I Re,abc (x) ΔI Re,abc (x) = . ΔI I m,abc (x) IsI m,abc − I I m,abc (x)

 F(x) =

(7)

The specified current Is and computed current I (V ) are calculated using the injected complex power and Ohm’s law: Is,i =

Ss V

and I (V )i = YVi .

(8)

i

The Jacobian is formed by taking the derivative of the real and imaginary current mismatch with respect to the real and imaginary voltage.

2.3 Network Components As explained earlier in this section, an electrical power system is represented as a graph containing branches and buses. These branches and buses represent the several elements that are connected in a network. Loads, shunts, and generators are expressed as buses, while general cables and transformers are modeled using equivalent branches. The buses in an electrical power network are either a load bus (PQ bus), a generator bus (PV bus), or a slack bus, depending on the information we know at that point. The loads in a network are modeled as PQ buses, loads consume power and at this bus, the active (P) and reactive power (Q) are specified. Generators are modeled as PV buses, except for the first generator in a network, this bus is modeled as the slack bus. Generators supply power and at this bus, the active power (P) and voltage magnitude (|V |) are specified. At a slack bus, the voltage magnitude and angle are specified. The several bus types and the according information are summarized in Table 1.

20

M. Kootte and C. Vuik

Table 1 Bus types in a network and the information we know and not know at each bus i Bus type Known Unknown P Q bus P V bus Slack bus

Pi , Q i Pi , |Vi | δi , |Vi |

δi , |Vi | Q i , δi Pi , Q i

In separate network analysis, the distribution network is modeled as an equivalent load bus in transmission power computations, while the transmission network is modeled as the slack bus in distribution power computations. The exact configuration for loads, shunt, transformers, and regulators installed along distribution cables can be found in the more extensive version of the authors [12].

3 Integrated Transmission and Distribution Network Model In order to analyze the power system, we need an integrated transmission and distribution network simulation tool. It is not straightforward to integrate these separate domains, because of the unbalanced nature of distribution networks compared to the balanced transmission network. Next to the modeling issues, one should also regard the difficulties that can arise when solving the integrated system. The differences between transmission and distribution power systems have led to the development of different solvers. Especially the high R/X ratio on distribution lines and the radial configuration compared to the meshed configuration of transmission networks have influenced the development of the solvers [4]. As explained in previous section, we use NR power to solve transmission networks and NR-TCIM to solve distribution networks. Integrating these two separate systems into a new domain should consider the integrated design and carefully investigate which solver is most advantageous to solve the integrated power flow problem. In order to integrate the two separate networks we need an interface that is capable of integrating a hybrid network (a single-phase network with a three-phase network) as a single entity [8]. The two networks are physically connected by a substation. The substation transforms high-voltage power of transmission networks to low-voltage power by a couple of in-series transformers. In separate distribution network analysis, this substation is modeled as a single transformer that connects the distribution network from the right-hand side, and where the transmission network is represented as an equivalent generator. This substation transformer is modeled as a branch using an equivalent Π -model. In order to create an integrated network, we need an interface capable of modeling the substation transformer in a hybrid network configuration such that it connects a

2 The Influence of the Increasing Penetration of Photovoltaic Generation …

21

Fig. 1 The substation transformer in a hybrid network connecting single-phase bus k and threephase bus m

single-phase and a three-phase network model. The substation transformer modeled as an equivalent Π -branch connecting the hybrid network is depicted in Fig. 1. It connects bus k of the transmission network with bus m of the distribution network. The MonoTri formulation [13] is an interface that describes how the substation transformer should be modeled in order to integrate the hybrid network as a single entity. It couples the single-phase quantities at the transmission side to the three-phase quantities at distribution side by transforming the nodal admittance matrix Ykm of the substation transformer. The coming subsection explains how this is established.

3.1 The MonoTri Formulation as Interface to Integrate the Hybrid Network Model To establish the connection of bus k and m via the admittance matrix Ykm , we use three transformer matrices: T1 , T3 =

1 1 1 2  2 [1 a a 2 ], T4 = [1 1 1] , and T5 = 1 a a , a = e 3 πι . 3 3 3

This transformation is based on the assumption that the connecting bus k is perfectly balanced. This means that the single-phase and three-phase quantities are related as follows: 

T   Va Vb Vc k = T1 Va k ,  T   Ia k = T3 Ia Ib Ic k ,  T   Sa k = T4 Sa Sb Sc k .

(9) (10) (11)

The exact outlook of the interface depends on whether the unified system is solved using NR-Power or NR-TCIM. Relations (9), (10), and (11) are substituted in the corresponding power flow equations. These outlooks are described in paragraphs given in Sects. 3.1.1 and 3.1.2.

22

3.1.1

M. Kootte and C. Vuik

MonoTri Formulation Using Current Injections

The NR-TCIM method uses Ohm’s law directly. The relation between node k and m is expressed as follows:  I = YV

⇔

    Y11 Y12 Vk Ik = . Im Y21 Y22 Vm

(12)

If node k and m were both modeled in three phase, we know the following: abc abc abc abc Ikabc = Y11 Vk + Y12 Vm , abc abc abc abc abc Im = Y21 Vk + Y22 Vm .

(13) (14)

We now multiply Eq. (13) by T3 to obtain Ika : abc abc abc abc Vk + T3 Y12 Vm . Ika = T3 Ikabc = T3 Y11

(15)

We then substitute Vkabc in Eqs. (15) and (14) by T1 Vka (Eq. 9): abc abc abc T1 Vka + T3 Y12 Vm , Ika = T3 Ikabc = T3 Y11

Imabc

=

abc Y21 T1 Vka

+

abc abc Y22 Vm .

(16) (17)

From (16) and (17), we see that our new nodal admittance matrix becomes

Ykm =

3.1.2

1 3



1

3

abc abc T3 [Y11 ]T1 T3 [Y12 ] abc abc [Y21 ]T1 Y22

 .

(18)

km

The MonoTri Formulation Using Power Injections

We can also start from the power equations. The relation between node k and m is expressed as follows: S = V I∗

 ⇔

    ∗ Vk Ik Sk = . Sm Vm Im

(19)

In the same manner as current injections, we can write this relation in three phase:

2 The Influence of the Increasing Penetration of Photovoltaic Generation …

23

Skabc = Vkabc Ikabc∗ + Vkabc Imabc∗ ,

(20)

Vmabc Imabc∗ .

(21) (22)

Smabc

= ⇔

Vmabc Ikabc∗

+

abc abc ∗ abc abc ∗ Skabc = diag(Vkabc ) · (Ykk Vk ) + diag(Vkabc ) · (Ykm Vk ) ,

Smabc

=

diag(Vmabc )

·

abc abc ∗ (Ymk Vk )

+

diag(Vmabc )

abc abc ∗ (Ymm Vm ) .

·

(23) (24)

We multiply the first line from the left by T4 to obtain Ska : abc abc ∗ abc abc ∗ Vk ) + T4 diag(Vkabc ) · (Ykm Vm ) . (25) Ska = T4 Skabc = T4 diag(Vkabc ) · (Ykk

Then, we substitute Vkabc = T1 Vka (Equation (9)) in Equations (24) and (25) and obtain the following: abc abc abc ∗ T1 Vka )∗ + T4 diag(T1 Vka ) · (Ykm Vm ) , Ska = T4 diag(T1 Vka ) · (Ykk

Smabc

=

diag(Vmabc )

·

abc (Ymk T1 Vka )∗

+

diag(Vmabc )

·

abc abc ∗ (Ymm Vm ) .

(26) (27)

We can rewrite T4 diag(T1 Vka ), the first part of the RHS in (26), as T4 diag(T1 Vka ) ⇔ T4 diag(T1 )diag(Vka ) ⎡ ⎤ 1 0 0   1 1 1 1 ⎣0 a 2 0⎦ diag(Vka ) = 3 0 0 a ⇔

1 2  1 a a diag(Vka ) 3   

(28) (29) (30)

T5

⇔ diag(Vka )T5 .

(31)

This results in the following relations for single-phase and three-phase power: abc abc abc ∗ T1 Vka )∗ + diag(Vka ) · (T5 Ykm Vm ) , Ska = diag(Vka ) · (T5 Ykk

Smabc

=

diag(Vmabc )

·

abc (Ymk T1 Vka )∗

+

diag(Vmabc )

·

abc abc ∗ (Ymm Vm ) .

(32) (33)

Equations (32), (33) yield the following transformed admittance matrix Ykm :  Ykm =

1 3

1

3

abc abc T5 [Ykk ]T1 T5 [Ykm ] abc [Ymk ]T1

abc Ymm

 . km

(34)

24

3.1.3

M. Kootte and C. Vuik

The Full Three-Phase Representation

In order to compare the results of the hybrid network configuration using the MonoTri formulation of the interface, we need an integrated network that is modeled in a full three-phase representation. The full three-phase formulation is able to capture the possibly induced imbalance at transmission level due to the high increase of PV power penetration at distribution level. The hybrid representation of the integrated transmission-distribution network is not capable of showing this imbalance and thus might exhibit errors in the simulation results. This is because the hybrid network only shows one phase and therefore the imbalance between phases is not visible. The full three-phase representation does not require an interface description but a three-phase representation of the transmission network. This description is based on the assumption that the transmission system is in a balanced initial state, and only possible imbalance that can arise is due to the increased level of PV penetration. In a balanced network, the phases b and c are deducted from the first phase a. We transform the voltage Va , the complex power Sa , and the admittance Ya of all the transmission buses and branches to their three-phase equivalents. We use the following transformer matrices for every bus i in the network: T  T1 = 1 a 2 a

 T 2 and T2 = 1 1 1 , a = e 3 πι ,

and identity matrix I3×3 . This results in the following: T  T1 [Va ]i = Va Vb Vc i ,   T  T2 Sa i = Sa Sb Sc i ,



a Y11 a Y21

⊗ ⊗

a I3×3 Y12 a I3×3 Y22

⊗ I3×3 ⊗ I3×3



 = ij

3

(35) (36)

3

3

abc abc Y11 Y12

3

abc abc Y21 Y22

 .

(37)

ij

3.2 Solvers for Integrated Hybrid Systems The MonoTri formulation of the interface connects the two separate network models into one integrated hybrid network model. As a next step, we consider which techniques can be used to solve this integrated hybrid system efficiently. NewtonRaphson solvers have different outlooks when they are used to solve transmission systems and distribution systems. Newton-Raphson using power mismatches is better capable of solving transmission systems, while the Newton-Raphson method using three-phase current injections (NR-TCIM) is better capable of solving distribution systems, as explained in Sect. 2. Nevertheless, the authors of [14] compare

2 The Influence of the Increasing Penetration of Photovoltaic Generation …

25

several formulations of Newton-Raphson solvers for both transmission and distribution systems. Although they show that the NR-TCIM formulation works best for distribution systems, it is also finely capable of solving transmission systems. As general electrical power systems consists of multiple distribution feeders compared to one transmission feeder and knowing the fact that distribution networks are often much larger than transmission networks, the distribution feeder will dominate the design of the integrated system. Using the outcome of [14], we expect that NR-TCIM should be capable of solving the integrated system. Nevertheless, we would like to point out that it is interesting to analyze which solvers are most efficient and robust to solve integrated systems.

4 The Impact of Photovoltaic Penetration on Electrical Power Systems Now that we have defined how we can integrate an electrical power system, we are interested in the effects of increased PV penetration on integrated transmission and distribution networks. For safe operation and planning of an electrical power system, it is important to maintain steady-state voltage stability among all buses in the networks and to limit the amount of imbalance of the network [15]. This means that individual voltage levels may not exceed limits prescribed by system operators. For transmission systems, this is usually between 0.95 and 1.05 pu. The maximum amount of imbalance is usually defined as 5% [7]. A hybrid integrated network model is not capable of showing the amount of imbalance on the transmission part of the network as it only represents one phase of the transmission network. Therefore, we validate our computations using a full three-phase presentation of the integrated network. In this manner, we can see how close the approximations of the hybrid network formulation are. If we see that the amount of imbalance is relatively high on the transmission network, the proposed MonoTri formulation is not suitable to analyze integrated systems. The authors of [16] propose a technique to incorporate the imbalance of the transmission network using the MonoTri interface without running full three-phase computations. The incorporation of this work is subject to future research. Another implication of increased amount of PV penetration is bi-directional power flow on a system that is originally designed for one-way power flow [17]. Necessary regulation measures should be taken to limit this. A positive consequence that can arise due to increased use of DERs is that these are often installed along distribution feeders close to the loads. This can reduce the power losses as the distance between generator and consumer significantly decreases [18]. These are all important effects that could be analyzed using an integrated network model. In this chapter, we focus our study on investigation of steady-state voltage stability and amount of imbalance on distribution and transmission networks.

26

M. Kootte and C. Vuik

4.1 Photovoltaic Panels Two types of PV panels can be connected to an electrical power system: (1) residential rooftop PV systems and (2) utility-scale PV panels. As the name implies, residential rooftop systems can be found on rooftops of households and buildings. This means that these systems are connected at distribution level, closely located to the loads. Utility-scale panels are installed on transmission level and distribution level and— due to the size of these installations—could be used to replace conventional power generators [18]. The focus of this study is on residential rooftop systems, closely connected to loads along the distribution feeder. The location close to the loads might limit the power losses that arise due to the distances that electrical power conventionally have to bridge to get from a transmission power plant toward end consumers at distribution level. On the contrary, as residential rooftops only supply active power, the location close to the loads might lead to insufficient supply of reactive power to the loads [18]. This can cause the voltage to drop. This will not only lead to voltage variations outside the safe operating limits, but also to individual phase variations as the distribution network is unbalanced. This leads to increasing imbalance, which is harmful for the entire electrical power systems [6]. Therefore, care should be taken to investigate the possible effects on transmission level by analyzing integrated power systems with an increasing amount of PV penetration.

4.1.1

PV Power Models

Most of the residential PV systems provide mainly active power and are therefore modeled as a negative load (P Q) bus, containing only negative active power injection. We are going to analyze the steady-state behavior of the integrated network by running simulations with various levels of PV penetration. Several methods exist to define the amount of PV penetration. We use the definition based on the total available generation in the base case [18], defined as follows: PV penetration(%) =

total PV generation · 100% total generation

(38)

We use varying penetration levels between 10% and 50%. Based on the data analysis of [17], we see that the peak irradiation2 of residential PV systems is 7 kW/m2 and that an individual panel with this level of irradiation produced a maximum power of 105 W. In this way, we can calculate how many panels are necessary to obtain the desired amount of PV penetration and whether this number of panels is a realistic amount for a certain neighborhood or residential area.

2

Irradiance is an instantaneous measurement of solar power over some area [19].

2 The Influence of the Increasing Penetration of Photovoltaic Generation …

27

4.2 PV Penetration Scenarios We are going to run power flow simulation on integrated systems using two artificially created integrated networks from available IEEE data that can be used for test simulations. The focus of this study is to simulate the effects of increasing PV penetration on steady-state voltage stability and amount of imbalance on both distribution and transmission networks. We are defining five scenarios with varying levels of PV penetration which we are going to assess: 0. I. II. III. IV.

The base case: No PV penetration. 10% PV penetration. 20% PV penetration. 50% PV penetration. An extreme scenario of 200% PV penetration, only supplied via phase A.

We are going to do steady-state power flow simulations using the Matpower3 library. We use a single core machine with an Intel Core i7-7600 processor, 2.80 GHz CPU, and 8.00 GB memory.

4.3 Test Cases We create integrated test cases from the existing transmission and distribution test cases from the Matpower library and resources page of IEEE Power & Energy Society [21]. We use the 9-bus and 118-bus networks from Matpower as balanced transmission network test cases. We use the IEEE 13-bus and 123-bus data from IEEE P&ES as unbalanced distribution test cases. This results in two integrated test cases: the T9-D13 network and the T118-D123 network. We model the substation transformers in these networks in a Delta-Grounded Wye configuration. We also create an integrated test case of a transmission network with multiple distribution networks in order to increase the effects that distribution networks can have on transmission networks. A country has most of the time one transmission network with several distribution networks connected to it. This is thus a realistic test case. This results in a third test case: T118-10D123, the 118-bus transmission network connected to ten 123-bus distribution networks. In total we have three test cases of a transmission network integrated with one or multiple distribution networks: 1. T9-D13. 2. T118-D123. 3. T118-10D123.

3

MATPOWER is a package of free, open-source MATLAB-language M-files for solving steadystate power system simulation and optimization problems [20].

28

M. Kootte and C. Vuik

The PV panels are connected to these networks at one bus that is modeled as a negative P Q-bus. Figure 2 shows such an integrated network. This is the visualization of the T9-D13 network, where the PV panels are connected to bus 21 of the integrated network.

5 Simulation Results In this section, we are going to analyze the steady-state voltage stability and amount of transmission and distribution imbalance on the three test cases using the five different scenarios. We start with the steady-state voltage stability.

5.1 Steady-State Voltage Stability The steady-state voltages of distribution networks should not exceed certain predefined limits. For the 13-bus distribution network, this limit is defined as 10%. The amount of PV penetration, modeled as a negative P Q-load only supplying active power, can reduce the voltage as the amount of reactive power is reduced. The following figures show the new steady-state voltages supplying the distribution network with different amount of PV penetration. Figures 3, 4, and 5 show that during normal PV penetration levels, the voltages of the T9-D13 test case never exceed its limits. Furthermore, it shows that the voltage profile is staying close to its original profile of the base case (without any PV penetration). Only in the extreme scenario, we see that the levels drop significantly. The results of the bigger test cases, T118-D123 and T118-10D123, are summarized in Table 2. This table shows the minimum and maximum voltage magnitudes of the three separate phases. This table shows that also in the bigger test cases, the voltage magnitudes do not drop on to a certain extend that it exceeds the prescribed operating limits. The test case including multiple distribution networks, the T118D123 test case, is exceeding its maximum safe operating limit. But, this is already happening during the zero PV penetration level. This has more to do with the amount of distribution networks than to the effect of PV penetration. The authors of [17] studied the effects of several levels of PV penetration on distribution networks only. They also concluded that the effects on voltage levels are limited. The results of this study are thus in line with their findings.

2 The Influence of the Increasing Penetration of Photovoltaic Generation …

29

Fig. 2 The T9-D13 test case visualized. The upper part is the 9-bus meshed transmission network. It is connected at bus 9 via the MonoTri interface description to bus 10, which is the original slack bus of the distribution network. The special elements are visualized at their location. The residential rooftop PV panel is located at bus 21 of the integrated network

30

M. Kootte and C. Vuik Phase A

Voltage magnitude

1.1

1.05

0. Base case I. 10% PV II. 20% PV III. 50% PV IV. 200% PV limits

1

0.95

0.9

0

5

10

15

20

25

Bus number

Fig. 3 Steady-state voltage profile of phase A of the T9-D13 network having different levels of PV penetration. The operating limits, which the voltages should not exceed, are given. The black star shows the location of the bus to which the PV panels are connected. The missing information on certain buses is due to the fact that the distribution feeder contains single-, double-, and three-phase laterals Phase B

Voltage magnitude

1.1

1.05

0. Base case I. 10% PV II. 20% PV III. 50% PV IV. 200% PV limits

1

0.95

0.9 0

5

10

15

20

25

Bus number

Fig. 4 Steady-state voltage profile of phase B of the T9-D13 network having different levels of PV penetration. The operating limits, which the voltages should not exceed, are given. The black star shows the location of the bus to which the PV panels are connected. The missing information on certain buses is due to the fact that the distribution feeder contains single-, double-, and three-phase laterals

2 The Influence of the Increasing Penetration of Photovoltaic Generation …

31

Phase C

Voltage magnitude

1.1

1.05

0. Base case I. 10% PV II. 20% PV III. 50% PV IV. 200% PV limits

1

0.95

0.9 0

5

10

15

20

25

Bus number

Fig. 5 Steady-state voltage profile of phase C of the T9-D13 network having different levels of PV penetration. The operating limits, which the voltages should not exceed, are given. The black star shows the location of the bus to which the PV panels are connected. The missing information on certain buses is due to the fact that the distribution feeder contains single-, double-, and three-phase laterals Table 2 The minimum and maximum voltage magnitudes of phases a, b, and c. Test cases T118D123 and T118-10D123 are shown. The boldface printed magnitudes exceeds the voltage magnitude limits of distribution test cases

Scenario

T118-D123 a Vmin

a Vmax

T118-10D123 b Vmin

b Vmax

c Vmin

c Vmax

a b c a b c Vmin Vmax Vmin Vmax Vmin Vmax

0.

0.97 1.08 0.96 1.03 0.97 1.06 0.96 1.12 0.96 1.07 0.96 1.10

I.

0.97 1.08 0.97 1.03 0.97 1.05 0.96 1.12 0.96 1.07 0.96 1.10

II.

0.97 1.08 0.97 1.03 0.97 1.05 0.96 1.12 0.96 1.07 0.96 1.09

III.

0.97 1.07 0.97 1.03 0.97 1.05 0.96 1.12 0.96 1.07 0.96 1.09

IV.

0.97 1.04 0.91 1.01 0.97 1.13 0.96 1.08 0.90 1.05 0.96 1.17

5.2 Amount of Imbalance on Distribution and Distribution Networks The amount of PV penetration can lead to increased imbalance on both distribution and transmission networks. Distribution networks are already unbalanced network, due to the unequal mutual coupling between phases on the lines, different voltage drop of the three phases, and unbalanced loads installed along the distribution feeder. PV power penetration can increase the amount of imbalance. The National Electrical Manufactures Association (NEMA) uses the following definition for the amount of imbalance [7]:

32

M. Kootte and C. Vuik

Table 3 The amount of maximum and average voltage unbalance in percentage of the distribution feeder for the three different test cases and the number of iterations of the three different methods Test case Distribution Transmission 0. I. II. III. IV. 0. III. IV. T9-D13 T118D123 T11810D123

2.33 2.15

2.40 2.19

2.47 2.19

2.69 2.27

11.29 10.99

0.04 0.03

0.04 0.04

0.36 0.27

2.17

2.21

2.21

2.27

11.14

0.10

0.11

0.67

Vunb =

Max Dev f r omVaverage ∗ 100% Vaverage

(39)

The recommended standard under normal steady-state conditions is that the voltage unbalance of distribution systems will not exceed 5% [7], while the amount of imbalance on transmission systems should be as minimal as possible. Table 3 shows the amount of imbalance on distribution networks for the defined scenarios on three test cases. We also show the amount of imbalance on transmission networks, which we can calculate from running full three-phase computations. We only show this amount for the base case, scenario. IV: 50% PV penetration, and the extreme scenario. Table 3 shows that the amount of imbalance on distribution networks is staying within a limit of 5% when penetrating the networks with a standard amount of PV penetration. Also the amount of imbalance on transmission networks is limited. Only when penetrating the networks with an extreme amount of PV penetration, we see that the imbalance increase beyond its limits. Even then, the amount of imbalance on transmission networks is limited. This means that as long the amount of PV penetration is staying within its operating limits, hybrid network formulation is sufficient for doing integrated power flow computations. As soon as the amount of PV penetration reach its critical limits, measures need to be taken at distribution level as the imbalance increases already beyond its limits and probably won’t influence transmission networks. The authors that introduce the MonoTri formulation [16] also assess the amount of unbalance on transmission networks. They show as well that the amount of unbalance is limited under certain levels of PV penetration. The conclusions are thus in line with this work.

6 Conclusion The rapid increase of photovoltaic (PV) power generation at distribution level makes it important to analyze electrical power systems as integrated systems. Integrated systems are necessary to analyze the steady-state voltage levels of transmission and

2 The Influence of the Increasing Penetration of Photovoltaic Generation …

33

distribution networks simultaneously including the interaction that these networks have on each other. As transmission networks are in general balanced networks and therefore modeled in single phase, while distribution networks are unbalanced and modeled in three phase, we need an interface that is capable of modeling this hybrid network configuration as an integrated network. The MonoTri formulation is a description of the interface that is able to integrate these two separate networks. The increasing amount of residential PV panels connected near the loads at the distribution network can lead to several effects which can be harmful for both the distribution and transmission networks. Residential rooftop panels are, in general, not capable of supplying reactive power. The lack of reactive power can lead to a voltage drop of the steady-state voltages and exceeding the limits that are designed on safe operating conditions. Next to that, increased PV penetration can increase the amount of imbalance due to the design of the distribution network. The amount of imbalance can also have harmful consequences for electrical power systems and should therefore not exceed a certain limit. Transmission networks are in general balanced and thus modeled as a single-phase network. They are therefore not able to show the amount of imbalance. The increased PV penetration could eventually lead to induced imbalance on transmission networks, which cannot be shown using a hybrid (single-phase/three-phase) network design. We have run several steady-state power flow simulations on integrated transmission-distribution networks to investigate whether the voltages exceed its safe operating limits and induce imbalance on distribution and transmission networks. We used these simulations to analyze the effects of increasing amount of PV penetration by comparing various levels of PV penetration with a base case of zero PV power. We used standard penetration levels of 10, 20, and 50 %, and an extreme case of 200 %. In the extreme case, PV power is only injected through phase A. This is done to intensify the effects that PV penetration can have on transmission networks. The simulations show that under normal levels of PV penetration, the steady-state voltages of distribution networks slightly drop but never exceed the safe operating limits. The amount of imbalance increases slightly, but not until an extent that would be harmful for distribution networks. Only in the extreme scenario, we see effects on the transmission network. We see a slight drop of voltage magnitude in some buses and a tiny amount of extra imbalance compared to the base case. Yet, this extreme scenario is already so harmful for distribution networks that measures should be taken at this level to prevent harmful amount of imbalance or voltage drop. We expect that these preventive measures will also decrease the effects on the transmission network. This means that the hybrid network representation of the integrated system is sufficient to analyze the steady-state behavior of electrical power systems. To summarize, we can conclude that a hybrid network model is a sufficient tool to analyze the effects of increased PV penetration of integrated electrical power systems. The amount of induced imbalance and the increased voltage drop are still within safety margins according to distribution operating standards and transmission

34

M. Kootte and C. Vuik

networks are not affected during these normal scenarios. The extreme case shows a slight increase in imbalance, but as the effects on distribution networks are much worse, we expect that preventive measures will be taken at distribution level such that this will not influence transmission networks.

References 1. Schavemaker P, Van der Sluis L (2008) Introduction to power system analysis. In: Electrical power system essentials, Sussex. Wiley, United Kingdom 2. Bhatti BA (2020) Department of electrical and computer engineering, virginia tech. In: Broadwater R, Dilek M (eds), Analyzing impact of distributed pv generation on integrated transmission & distribution system - a graph trace based approach, Energies, vol 13, no. 4526, pp 6–9 3. Balasubramaniam K, Abhyankar S (2017) A combined transmission and distribution system co-simulation framework for assessing the impact of volt/var control on transmission system, no. Vvc 4. Bollen M, Hassan F (2011) Integration of distributed generation in the power system 5. Yan R, Saha TK (2012) Voltage variation sensitivity analysis for unbalanced distribution networks due to photovoltaic power fluctuations. IEEE Trans Power Syst 27(2):1078–1089 6. Balamurugan K, Srinivasan D (2011) Review of power flow studies on distribution network with distributed generation. In: IEEE PEDS, pp 411–417 7. Kersting WH (2011) The whys of distribution system analysis. Requirements for a power-flow study. In: IEEE industry applications magazine, pp 59–65 8. Kootte ME, Sereeter B, Vuik C (2020) Solving the steady-state power flow problem on integrated transmission-distribution networks: a comparison of numerical methods. In: 2020 IEEE PES innovative smart grid technologies Europe (ISGT-Europe), pp 899–903 9. Schavemaker P, van der Sluis L (2008) Energy management systems. In: Electrical power system essentials, Sussex. Wiley, United Kingdom 10. Sereeter B, Vuik K, Witteveen C (2017) Newton power flow methods for unbalanced threephase distribution networks. Energies 10(10):1658 11. Garcia PAN, Pereira JLR, Carneiro S, Da Costa VM (2000) Three-phase power flow calculations using the current injection method. IEEE Trans Power Syst 15(2):508–514 12. Kootte ME, Romate J, Vuik C (2011) Solving the power flow problem on integrated transmission-distribution networks: a review and numerical assessment 13. Taranto GN, Marinho JM (2008) A hybrid three-phase single-phase power flow formulation. IEEE Trans Power Syst 23(3):1063–1070 14. Sereeter B, Vuik K, Witteveen C (2017) Newton power flow methods for unbalanced threephase distribution networks. Energies 10:1658 15. Bhatti BA, Broadwater R, Dilek M (2020) Analyzing impact of distributed pv generation on integrated transmission & distribution system - a graph trace based approach. Energies 13(4526):6–9 16. Taranto GN (2018) Simulation of integrated transmission and distribution networks with a hybrid three- phase/single-phase formulation. In: Conference proceedings 17. Ramachandran V, Solanki SK, Solanki J (2011) Steady State Analysis of Three Phase Unbalanced Distribution Systems with Interconnection of photovoltaic cells. In: 2011 IEEE/PES power systems conference and exposition, IEEE, pp 1–7 18. Eftekharnejad S, Vittal V, Heydt GT, Keel B, Loehr J (2013) Impact of increased penetration of photovoltaic generation on power systems. IEEE Trans Power Syst 28(2):893–901

2 The Influence of the Increasing Penetration of Photovoltaic Generation …

35

19. SNL PV Performance modeling collaborative, “Irradiation & Insolation” (2021) 20. Zimmerman RD, Murillo-sánchez CE, Thomas RJ, Fellow L, Atpower AM (2011) MATPOWER: steady-state operations, systems research and education. 26(1), 12–19 21. Schneider KP, Mather BA, Pal BC, Ten CW, Shirek GJ, Zhu H, Fuller JC, Pereira JL, Ochoa LF, De Araujo LR, Dugan RC, Matthias S, Paudyal S, McDermott TE, Kersting W (2018) Analytic considerations and design basis for the IEEE distribution test feeders. IEEE Trans Power Syst 33(3):3181–3188

Chapter 3

Building Energy Management Nor Azuana Ramli and Mel Keytingan M. Shapi

Abstract Statistics show that approximate energy usage in a building is 10–20 times more than residential which is around 70–300 kWh/m2 . The electricity demand is expected to increase triple than current demand in 2030. It is found that total energy demand and produced are not balanced whereby there will be not enough energy to supply for higher demand in the future. This why we need to manage energy properly especially for commercial building. Thanks to technology, now there is no need for building owners to hire energy auditor in order to know how to manage energy in their building. Technology has evolved commercial building into smart building. By installing sensors in the building and make use of Internet of Things technology, the energy can be managed through web or mobile apps. In this chapter, we are going to explain on how building evolved from commercial building to smart building and the development of building energy management by using machine learning and big data analytic approach.

1 Traditional Building Energy Management Electricity is the most important energy in our daily life. The usage of the electricity is widely used in all important sectors, especially manufacturing, residential, commercial building, education and transportation sector. The impact of the growing number of national developments is also increasing the energy demand. The increase of energy demand is somehow terrifying as it contributes significantly to climate change by adding more carbon dioxide to the atmosphere. To simulate sustainable building practice, some countries are practicing a code or standard. For example, in Malaysia, the Code of Practice on Energy Efficiency N. A. Ramli (B) Centre for Mathematical Sciences, College of Computing & Applied Sciences, Universiti Malaysia Pahang, Lebuhraya Tun Razak, 26300 Gambang, Pahang, Malaysia e-mail: [email protected] M. K. M. Shapi Electrical Engineering Section, Universiti Kuala Lumpur British Malaysian Institute, Batu 8, Jalan Sungai Pusu, 53100 Gombak, Selangor, Malaysia © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 A. Tomar et al. (eds.), Control of Smart Buildings, Studies in Infrastructure and Control, https://doi.org/10.1007/978-981-19-0375-5_3

37

38

N. A. Ramli and M. K. M. Shapi

and use of Renewable Energy for Non-Residential Buildings Malaysian Standard MS 1525:2007 are being used. These standards discuss about planning on how to design a peak of the energy efficiency of a building for the engineering, architectural, landscaping and site planning aspects [5]. According to the Efficient Management of Electrical Energy Regulation 2008, the building that received the electrical supply from the supply authority with a total electricity consumption equal to or exceeding to 3,000,000 kWh as be measured at the metering point over any period of the consecutive months must comply with the requirement regulations. The fact is that the increase of energy demand has opened the eye of the government. There are many ways which can be adapted in order to manage energy efficiently. For the traditional building energy management, energy audit and energy modelling are among the most used method in practice. In this section, we are going to discuss further on the energy audit and energy modelling with a case study as an example.

1.1 Energy Audit Generally, energy audit need to be done so that performance of the building can be analysed and any potential of energy saving in the building can be identified. This is the first step in identifying opportunities to reduce energy consumption and achieve energy efficiency. Energy audit involves systematic record of energy data, analysis of energy data and proposal of potential energy-saving measures. This process takes few weeks to several months to complete, depending on the complexity of the building. Usually, energy audit is done by auditor that can be hired through energy services companies (ESCOs), engineering firms or energy consultants. The cost to hire auditor will be added into labour cost which will increase the building operation cost. With the advancement of technology, the future of energy audit will no longer depend on auditor as all the process in energy audit can be done through technology such Internet of Things, artificial intelligence and so on. In Sect. 3.2, we will discuss on how technology like Internet of Things is applied in managing energy consumption for commercial building.

1.2 Energy Modelling Energy modelling is a simulation of a building that focuses on energy consumption, utility bills and many other energy-related items such as air conditioning and lighting. The foundation of energy modelling is building simulation. Building simulation is the process of using a computer to build a virtual construction of a building. Simulation is performed by considering some data as inputs such as the building envelope, orientation and materials with weather data based on building’s location. Usually,

3 Building Energy Management

39

building simulation is divided into two categories which are load design and energy analysis. There are many capabilities of energy modelling such as it can be used to predict the monthly energy consumption, bills and annual energy cost, compare different efficiency options and determine payback from many options of saving measures. The process of energy modelling starts with the construction of a building by using computer software. Then, energy flow of the building can be obtained through computer simulation provided some information such as weather data, location of the building, energy transfer and others. Building Energy Index (BEI) can be determined through the calculation of building energy consumption by using energy flow data. From the existing BEI obtained, the efficiency measures to optimize energy usage are modelled. From the energy audit results, it can be seen that generally the loads or applications that contribute to high energy consumption in the commercial buildings are mainly from heating, ventilation and air conditioning (HVAC) system. High energy consumption from HVAC system is due to the system itself that has high energy use equipment such as chiller, AHU, boilers, pumps and supply fans. To plan and design the efficiency measures that are suitable and applicable for a building with this problem is not easy but it can be done through energy modelling. A case study of a commercial building in Malaysia will be presented here to show how energy modelling method can be applied in order to plan and design the efficiency measures that are suitable with the building to reduce the electricity cost. First, the building was built by using SketchUp software and then the 3D model was developed using PDF construction drawings due to the unavailability of AsBuilt drawings. Figure 1 shows the front view of the building in 3D. There are some

Fig. 1 Building 3D model. Reproduced from [1]

40

N. A. Ramli and M. K. M. Shapi

limitations that need to be considered such as changes to the actual building that have not been documented where this may result in a difference between simulated results and actual onsite results. There is also limited information on the specifications and details of core building elements such as glazing specifications, wall construction and various materials used. To overcome this limitation, assumptions were made referencing typical building construction data and materials for this analysis. The essential data that is considered to create a baseline for the simulation in this analysis are wall insulation, floor insulation, roof insulation, glazing U-value (the level of solar emittance allowed), glazing visible light transmittance, glazing solar heat gain coefficient, infiltration rate, ventilation rate, equipment power density and lighting power density. Then, the next step in this analysis is to calculate overall thermal transfer value (OTTV) before the calculation of BEI can be made. For OTTV, calculation can be made by using these two formulae where these two equations were referred from Malaysian standard MS1525:2014 of Clause 5.2.1 and 5.2.2, respectively: OTTV =

(A1 × OTTV1 ) + (A2 × OTTV2 ) + · · · + (An × OTTVn ) A1 + A2 + · · · + An

(1)

where A1 is the gross exterior wall area for orientation 1 and OTTV1 is the value for orientation 1 from Eq. 2. For a fenestration at a given orientation, the formula is OTTVi = 15α(1 − WWR)UW + 6(WWR)Uf + (194 × OF × WWR × SC) (2) where WWR is the window-to-gross exterior wall area ratio for the orientation under consideration; α is the solar absorptivity of the opaque wall, as in Table 6 in MS1525:2014; UW is the thermal transmittance of opaque wall (W/m2 K); Uf is the thermal transmittance of fenestration system (W/m2 K); OF is the solar orientation factor, as in Table 5 in MS1525:2014 and SC is the shading coefficient of the fenestration system. OTTV is an important parameter that needed to be determined by engineers as it indicates the overall amount of the thermal energy transferred into the building which will affect the operation of cooling system. After OTTV value is obtained, the analysis continued with determining BEI. It is important to determine the BEI for any building as it is used to monitor the performance of energy consumption in building, acts as a reference point that provides the baseline for energy performance comparison and offers the best benchmarking on building energy utilization in order to organize an effective energy efficiency scheme in the future. The formula to calculate BEI is as follows: BEI =

Total annual energy consumption (kWh/year) Total floor area of building (m2 )

(3)

3 Building Energy Management

41

In this analysis, the BEI is generated from the simulation of the baseline data of the building model. The Sefaira web-based analysis is capable of producing the BEI through the analysis engine. Finally, based on the results from the energy modelling method, a cost saving analysis is developed in order to determine the potential of cost saving from the potential energy efficiency measures (EEMs) that is being proposed. Basically, the cost saving analysis is affected by the proposed EEMs from the findings of the highest energy usage contributing factors. Hence, it must be noted that the proposed energy efficiency measures must be accurate and relevant in order to obtain accurate cost saving values. The results from this analysis showed that energy-saving analysis was able to perform and it contributed to 52.82% of saving in terms of BEI and 36.44% of saving in terms of annual cooling energy. This case study had some limitations and flaws but the results of the study are sufficient to show that energy modelling method can be used for better energy management in a building. Although it is believed that a better result can be obtained with accurate data and information about the building. Traditional building energy management is probably easier to be implemented compared to smart building energy management. This is because it only involves energy audit and energy modelling that does not require high-level skills such as Internet of Things, big data, machine learning, sensors and coding. A more complex system which is smart building energy management will be explained in the next section. Although it is more complex it is the future as it involved the fourth industrial revolution (IR 4.0).

2 Smart Building Energy Management Smart building is a cutting-edge technology structure that can improve energy efficiency by simplifying the system connection in using computerized process [6]. In a smart building, all the electrical appliances can be controlled digitally and offer handiness towards consumers in managing their energy management. Thanks to sensors and Internet of Things (IoT) technology, now energy consumption can be monitored through your website or mobile application anywhere anytime and energy efficiency can be applied at the tip of your fingers. Other technology such as machine learning can help the building owner predict energy usage in the building, and hence make it easier for building owner to plan for reducing the energy usage. All these technologies will be discussed in this section with a case study as an example. At the end of this section, readers can see that technology has helped us in obtaining information so much faster than traditional ways although it is a little bit costly and complex.

42

N. A. Ramli and M. K. M. Shapi

2.1 Internet of Things (IoT) A smart building with the help of IoT is the most effective ways in monitoring and managing energy consumption [7]. IoT is also one of the most technology aspects that is required in order to develop a smart management system successfully. Generally, IoT is the connection between the Internet and the machine to make work easier, cheaper and more effective. IoT is a technology revolution that represents the future of computing and communications. It has the capacity to track a large number of variables, trace an item, speed up procedures and reduce errors. Example of a case study that will be shared in this section is a shopping complex building located in Malaysia. The smart building in this study was already equipped with IoT meters which are connected to the power inlet socket at two major tenants of the building. Each tenant is divided into two areas which consists of two IoT meters named tenant A1, A2, B1 and B2. Collected data per minute that mapped into Tenaga National Berhad (abbreviated as TNB hereafter, TNB is one of energy providers in Malaysia) requirements were saved in an open-source web server. User such as building owner or tenant can observe and monitor the collected data internally through an online platform by giving an ID number and password. The online website is really helpful as the users can monitor the building energy consumption data using systematic and manageable online system. It provides graph visualization that shows the data collected for every hour in order to help the users in visualizing the behaviour or trend of the maximum demand of each tenant. For example, Figs. 2 and 3 show

Fig. 2 Hourly consumption of maximum demand graph for Tenant A1 in June 2018. Reproduced from [3]

Fig. 3 Hourly consumption of maximum demand graph for Tenant A2 in June 2018. Reproduced from [3]

3 Building Energy Management

43

Fig. 4 IoT based data collection framework for smart commercial building

the images that represented hourly data of maximum demand (MD) for tenants A1 and A2 collected by the smart meter reading that has been provided via the website in June 2018. Collected data can be extracted manually from the online platform in the form of CSV file. Implementation of IoT in a smart building is not just for monitoring energy consumption but to analyse the data collected involving energy consumption. In the case study stated above, the data was collected through a smart meter for every hour and transmitted to data storage using a Zigbee-based wireless data transceiver through the ASP.NET web server. As per TNB requirement, the smart meter collected the data of kilowatt-hour (kWh), voltage, current, power factor and maximum demand. These collected data were statistically analysed before the study proceeds with predictive model development. The working process is illustrated in Fig. 4. As can be seen in Fig. 4, development of predictive model is a part of smart building energy management. This is due to a statement made by Kaytez et al. [8] where the authors stated that the fundamental task of energy prediction is to provide a benchmark for energy investment planning and hold a crucial position whereby overestimation could cause waste in resources, while underestimation could impact financially to the energy utility provider and also disrupt the existing power system. There are various methods which had been used in predicting energy consumption and one of the methods is machine learning. In the next section, a complete analysis of energy consumption and energy prediction by using machine learning method will be explained in the details based on the same case study.

2.2 The Usage of Machine Learning Method in Predicting Energy Consumption Other than automated control, a smart building also consists of an intelligent system which provides energy consumption forecast as an energy efficiency initiative. This is due to its advantage of yielding economical savings and as a sustainable approach for

44

N. A. Ramli and M. K. M. Shapi

Fig. 5 Process of generating predictive model after data preparation

energy management to minimize energy wastage [9]. One of the famous methodologies applied in predicting energy consumption is machine learning. Before building a predictive model, dataset used in the analysis need to go through pre-processing (details on the pre-processing will not be discussed in this book as we focus more on the machine learning algorithm. Readers can refer to Shapi et al. [4] for further details). Now, by continuing the same case study in this book chapter, the machine learning method will be applied in order to predict energy consumption. The type of machine learning that will be applied in this case study is supervised learning. After pre-processing, the data will be inputted into the learning algorithm. Then, data partitioning was done to separate the data into two groups which are training group and testing group. The predictive modelling for this case study used classification method to predict discrete variables instead of regressive prediction. AzureML was used as the platform while Caret R package was utilized for all prediction to ensure uniform execution. Three types of machine learning algorithm were selected which are k-Nearest Neighbour (k-NN), Support Vector Machine (SVM-RBF) and Artificial Neural Network (ANN-MLP). Figure 5 shows the process after the data preparation until the generation of the predictive model. The first prediction model in this case study is k-Nearest Neighbour (k-NN) method. This algorithm is known for its simplicity and it is frequently used due to that reason. Its forecasting capability on intricate non-linear pattern [10]. k-NN algorithm provides prediction by determining the similar instances between data points in a feature space [11]. For this case study, the method was used to predict maximum demand by using voltage, current and power factor as the features, due to the resultant multiplication of the values will output the electrical power usage (kW). Therefore, the k-NN will have three-dimensional plotting as illustrated in Fig. 6. Based on Euclidean distance function, k-NN method was trained repeatedly up to maximum tuning parameter (k-value) which equals to 43. The resultant model from the training with lowest RMSE value was chosen for prediction. The next prediction method was Support Vector Machine (SVM) with Radial Basis Function (RBF) as its kernel function. If k-NN known for its simplicity, then SVM is known because of maximum margin classifier and it is usually applied to

3 Building Energy Management

45

Fig. 6 Dimensional k-NN plotting

solve problems involving classification and regression for a large dataset [12]. The work process of the SVM method is as illustrated in Fig. 7 based on methodology by Liu et al. [13]. The step starts with transforming the data into a specific format. For SVM method, only discrete and continuous data can be used; hence, any categorical data need to be converted to numeric data in advance. In this case study, all data are numerical, so there was no transformation of data type needed. After that, the data was pre-processed to reduce the prediction complexity.

Fig. 7 SVM structure with Gaussian RBF kernel [13]

46

N. A. Ramli and M. K. M. Shapi

There are various kernel selections available for SVM method such as Gaussian, Sigmoid, Polynomial, linear and Radial Basis Function (RBF). In this case study, RBF was chosen due to the broad and non-linear characteristics of the dataset.   2    (4) K R B F x, x  = exp −γ x − x   where  γ is a gamma parameter to determine the spread distribution of the kernel and x − x   is the Euclidean distance between the set of points. Equation 5 has corresponding definition as represented in Eq. 4 using sigma parameter [14]. 

K R B F x, x





   x − x  2 = exp 2σ 2

(5)

There are two tuning parameters for SVM-RBF which are kernel parameter sigma, σ and cost parameter, C. These two parameters were adjusted for repeated training. The sigma value is vital in order to get a good fit model to the data. If the sigma value is larger than the distance between classes, it will output an almost linear decision boundary, while vice versa will lead it to overfit. Cost parameter is the penalty limit in which the data point is misclassified or overstep maximum margin. If the cost value is large, the classifier will highly prohibit misclassified data, and therefore less error. However, it will increase the complexity of the model [15]. Relatively, if the cost parameter is small, more misclassified data is allowed, thus higher error data point. The effect of sigma and cost parameter is visualized in Fig. 8. The third and final methodology for our prediction was Artificial Neural Network (ANN). The main advantage of using this method is the method’s capability to learn complex behaviour which then makes it widely used for predictions and pattern recognition [16]. This model structure consists of a formation of interconnected neurons that have three main layers which are input layer, hidden layer and output layer as shown in Fig. 9. Adjustment of the synaptic weight of each link that connects between the neurons was made by comparing the initial output with the desired output until the Sum Squared Error (SSE) value is minimized. This would provide regularization for the model [13]. The weight is the representation of the priority or importance of the neuron input. For this case study, a Multilayer Perceptron Model (MLP) type of ANN structure with error backpropagation learning algorithm was used for its network solution structure, which is small and rapid rate of computational process for training with large datasets. Other than that, it automatically generalize its knowledge based on the learning process, which enabled the recognition of data sets [17]. For error backpropagation learning, the backpropagation step provided the adjustment to the weights. From Fig. 9, the input node is from the initial observations, while the output layer will provide the predicted future value. In the hidden layer, it is important to select a correct non-linear transfer function as it will be used to compute the information

3 Building Energy Management

47

Fig. 8 Effect of varied cost and sigma parameter value on SVM-RBF classification boundaries [15]

Fig. 9 Multilayer perceptron model with error backpropagation algorithm [16]

48

N. A. Ramli and M. K. M. Shapi

accepted by the input layer. The ANN model was as shown in Eq. 6: yt = α0 +

n j=1

αj f

m

 βi j yt−i + β0 j

+ εt

(6)

i=1

where m is the number of input nodes, n in the number of hidden nodes, f is the Sigmoid Transfer function, {αi , j = 0, 1, . . . n} ib the vector of weights

from the hidden layer to the output layer and βi j , i = 0, 1, . . . m; j = 0, 1, . . . n is the weight from the input to the hidden nodes. Multilayer perceptron model has several hyperparameters available to be tuned, including the number of hidden layers, number of neurons or hidden units and activation functions. For this case study, the hyperparameter tuned was the number of neurons per layer. This number of neurons denotes the width of the network and its latent space [18]. The number of neurons has a substantial effect on the multilayer perceptron model as it determined the decision boundaries of the model [19]. A low number of hidden units would cause the model to be non-adjustable, while a high number of hidden units cause the model to be overfitting [20]. Another penalizing parameter that was tuned and applied was weight decay. This parameter is a penalizing method to constrain the complexity of the model and to limit the growth of the model’s weight parameter [21]. This penalizing method basically enhances the initial generalization of neural network model [22]. For the analysis in this case study, the value of the weight decay was in logarithmic scale, in which the model training has selected values in between 0 and 0.1. The data were partitioned into two groups before inputting to the machine learning algorithm. For this case study, 70% of the dataset was used for training and the rest of the data was partition as testing data. The training data usually were used to train each machine learning algorithm and generate a prediction model that could output value that matches with the recorded maximum demand data, while the rest of the data was held back to be used to test the trained predictive model. The process is as illustrated in Fig. 10. With AzureML, data partitioning for training and testing would not be a hassle and biased as it has built-in support for data division. The partitioning process was straightforward in which selection was made randomly. This process prevented overfitting, which could cause either underestimation or overestimation of the maximum demand value. The last part in machine learning is to validate our model in order to see its performance and accuracy. Three methods of evaluation were used, which are Root Mean Square Error (RMSE), Normalized RMSE and Mean Absolute Percentage Error (MAPE). The formula for RMSE is shown in Eq. 7 and MAPE in Eq. 8. Comparison of performance between each method to different tenants was made by using a normalized RMSE or also known as Coefficient of Variation RMSE (CV RMSE) as shown in Eq. 9. This metrics removes the scale dependent of RMSE [23].

3 Building Energy Management

49

Fig. 10 Testing of the trained predictive model

 Root Mean Squar e Err or, R M S E =

n t=1

Mean Absolute Per centage Err or, M A P E =

(At − Ft )2 n

n 1 (At − Ft )2 n t=1  n 2 t=1

N or mali se Root Mean Squar e Err or, N R M S E =

(At −Ft ) n

Mean

(7)

(8)

(9)

Results that were obtained from this case study are divided into two parts. First part will be explained on each method while the second part will be explained on the comparison between these three methods after the best results were obtained from the tuning of each method. The first analysis was conducted on k-Nearest Neighbour (k-NN) to find the best k parameter for each tenant. The resampling was iterated 20 times in which k-value ranging from 5 to 43 was selected. Tables 1, 2, 3 and 4 show the performance evaluation of the training based on the range of k-value with the best performing k-value highlighted. Based on the result, it can be deduced that all tenants have different best performing k-value. For department A, the best performing k-value was 11 and 13, respective to Tenant A1 and Tenant A2. The RMSE evaluation result shows that the Tenant A1 model with k-value 11 has an error rate of 5.39198, while Tenant A2 was 3.86651. In department B, the best performing k-value for Tenant B1 was 29 with error rate of 14.97611, while Tenant B2 was 43 with error rate of 0.54824. Understanding of RMSE evaluation, the prediction model was reckoned as the best performing model if the RMSE value was lower. The comparison of RMSE value between tenants was made only to provide a perspective of the RMSE difference and not to provide an inference regarding the accuracy of the method. From the performance evaluation table result, it was observed that Tenant B2 has low RMSE

50 Table 1 Performance evaluation of k-NN model training for Tenant A1

Table 2 Performance evaluation of k-NN model training for Tenant A2

N. A. Ramli and M. K. M. Shapi k-value

RMSE

5

5.57960

7

5.44382

9

5.40786

11

5.39198

13

5.40994

15

5.43031

17

5.45572

19

5.46711

21

5.49447

23

5.52399

25

5.53781

27

5.55795

29

5.58129

31

5.60897

33

5.63802

35

5.66665

37

5.69051

39

5.71114

41

5.73325

43

5.75187

k-value

RMSE

5

3.92161

7

3.87741

9

3.86665

11

3.86815

13

3.86651

15

3.87567

17

3.89211

19

3.90609

21

3.92863

23

3.95779 (continued)

3 Building Energy Management Table 2 (continued)

Table 3 Performance evaluation of k-NN model training for Tenant B1

51

k-value

RMSE

25

3.97728

27

3.99574

29

4.01121

31

4.02655

33

4.04512

35

4.05826

37

4.06722

39

4.07847

41

4.08835

43

4.09735

k-value

RMSE

5

0.58692

7

0.57065

9

0.56575

11

0.56308

13

0.56125

15

0.55832

17

0.55627

19

0.554959

21

0.55336

23

0.55189

25

0.55106

27

0.55057

29

0.55024

31

0.54965

33

0.54913

35

0.54856

37

0.54841

39

0.54837

41

0.54833

43

0.54824

value at 0.54824. After the best performing model k-value was selected, the trained model was used to predict the demand using the hold-out data. Table 5 shows the evaluation of the predicted demand based on test data. The result from the test prediction evaluation does not fall far from the evaluation of model training.

52 Table 4 Performance evaluation of k-NN model training for Tenant B2

Table 5 Performance evaluation of test prediction for all tenants using trained k-NN model

N. A. Ramli and M. K. M. Shapi k-value

RMSE

5

0.58692

7

0.57065

9

0.56575

11

0.56308

13

0.56125

15

0.55832

17

0.55627

19

0.554959

21

0.55336

23

0.55189

25

0.55106

27

0.55057

29

0.55024

31

0.54965

33

0.54913

35

0.54856

37

0.54841

39

0.54837

41

0.54833

43

0.54824

Tenant

k-value

RMSE

MAPE (%)

A1

11

5.0025748

A2

13

3.6548885

9.98

B1

29

14.934312

15.43

B2

43

0.5439403

48.75

3.02

The next analysis was conducted on Support Vector Machine with Radial Basis Function kernel to find the best sigma and cost parameter. The resamplings were as well iterated 20 times, in which the cost parameter was changed in a range from 0.25 to 131,072.00. For sigma value, it was initially selected as a single value as it was the default behaviour of the SVM function in caret package. The value selection was chosen by calculating the mean of sigma values suggested by the dataset through “sigest” function in the package. This mean value was kept at constant throughout each iteration of the cost parameter. In this model training, RMSE was used as the definitive to select the best parameter for the model. Tables 6, 7, 8 and 9 show the performance of SVM model tuning for each tenant.

3 Building Energy Management Table 6 Performance evaluation of SVM model training for Tenant A1

Table 7 Performance evaluation of SVM model training for Tenant A2

53

Sigma

Cost

RMSE

0.7879233

0.25

5.238166

0.5

5.146431

1

5.090147

2

5.050991

4

5.042777

8

5.056961

16

5.071987

32

5.102601

64

5.158422

128

5.233809

256

5.320873

512

5.439078

1024

5.631843

2048

5.881493

4096

6.276858

8192

7.334011

16,384

9.805928

32,768

12.603121

65,536

16.574019

131,072

20.571380

Sigma

Cost

RMSE

1.39528

0.25

3.898670

0.5

3.834173

1

3.785405

2

3.756411

4

3.749604

8

3.764149

16

3.773521

32

3.773971

64

3.787017

128

3.810500 (continued)

54 Table 7 (continued)

Table 8 Performance evaluation of SVM model training for Tenant B1

N. A. Ramli and M. K. M. Shapi Sigma

Cost

RMSE

256

3.860896

512

3.949990

1024

4.105368

2048

4.340844

4096

4.753597

8192

5.475539

16,384

7.682156

32,768

9.485241

65,536

11.374227

131,072

14.361323

Sigma

Cost

RMSE

1.41591

0.25

16.958630

0.5

16.882330

1

16.820560

2

16.762600

4

16.701220

8

16.636800

16

16.580610

32

16.510300

64

16.419470

128

16.357890

256

16.332290

512

16.347500

1024

16.394700

2048

16.479580

4096

16.681790

8192

17.122630

16,384

17.885590

32,768

21.462150

65,536

23.700210

131,072

28.274190

From the table, the sigma parameter during model tuning for Tenant B1 was higher than the other tenants at 1.41591. In contrast, the sigma parameter for Tenants A1, A2 and B2 are 0.7879233, 1.39528 and 1.31358, respectively. By comparing the situation with cost parameter, it was observed that Tenant B1 also has a larger value

3 Building Energy Management Table 9 Performance evaluation of SVM model training for Tenant B2

55

Sigma

Cost

RMSE

1.31358

0.25

0.561659

0.5

0.560519

1

0.559862

2

0.559315

4

0.559200

8

0.559300

16

0.559219

32

0.559615

64

0.561408

128

0.564752

256

0.571516

512

0.584971

1024

0.607109

2048

0.641071

4096

0.679627

8192

0.729509

16,384

0.798584

32,768

0.989541

65,536

1.299583

131,072

1.619222

in comparison to the other tenants at 256, while the other tenants have a similar cost parameter of 4. This observation shows that for the SVM model training for Tenant B2, more feature points near the hyperplane were considered as support vector. Comparing with other tenants, the sigma value also indicates that the vector support area was smaller and has lower flexibility of the hyperplane boundaries [24]. The selection of cost value for the best performing model was assessed based on their RMSE evaluation. Tabulation of the performance evaluation of SVM model shows that Tenants A1, A2, B1 and B2 RMSE values were 5.050991, 3.749604, 16.332290 and 0.559200, respectively. It was observed that Tenant B1 has the highest RMSE value compared to other tenants. Following with the selection of model trained with best performing cost value, the model was used to forecast the energy demand by using the partitioned test data. Table 10 shows the evaluation of the prediction using the trained model. The RMSE of the predicted demand was almost similar with the error rate during the training of the models. From the table, the SVM model trained for B1 has the highest RMSE with 16.0690844. The third analysis was conducted by using Artificial Neural Network (ANN) where the model was determined based on the network architecture and the structure was built upon the transfer function and learning algorithm. For the network structure, the multilayer perceptron model with weight decay and number of hidden units as

56

N. A. Ramli and M. K. M. Shapi

Table 10 Performance evaluation of test prediction for all tenants using trained SVM model Tenant

Sigma

A1

0.7879233

A2 B1 B2

Cost

RMSE

MAPE (%)

4

4.7506789

1.39528

4

3.5898263

9.38

1.41591

256

16.0690844

12.09

1.31358

4

0.5558279

43.97

Table 11 Performance evaluation of ANN model training for all tenants

Tenant

No. of Hidden Units

2.76

Weight Decay

RMSE

A1

9

0

7.47574

A2

37

0

5.00378

B1

35

0

17.73607

B2

35

0

0.54580

its parameter was used. The resampling was iterated for 20 times for both weight decay and number of hidden units. The weight parameter was tuned from 0 until 0.1. Diversely, the number of hidden units was tuned from 1 to 39. The number of hidden layers was kept constant at 1 for each iteration. In overview, 400 ANN-MLP models were created from this tuning, and each of this model was evaluated using RMSE metrics. The best performing model was tabulated in Table 11 and the summary of the model trained with the best performing parameters was visualized in neural network plot, as shown in Fig. 11, 12, 13 and 14 (Table 12). From Table 11, it was observed that ANN model for Tenant A1 has the smallest number of neurons with nine hidden units. Tenants A2, B1 and B2 have number of hidden units more than 30, whereby their sizes were at 37, 35 and 35, respectively. This value shows that the network size for Tenant A1 was smaller than the other tenants. Based on Panchal and Panchal [25] and Gaurang et al. [26], the number of hidden neurons has a significant effect on the performance of the neural network model. A large number of hidden neurons could induce an overfitting model as it introduces inessential neurons to the model. On the contrary, a low number of neurons caused the model to be underfitting, in which the neurons are deficient in manipulating complex data. This statement is also dependent on the type of dataset. In Table 11, the RMSE value for Tenant A2 is lower in comparison to Tenant A1 and Tenant B1. In this model training, the weight decay was tuned to penalize the weight parameter between neurons. From the result of the best performing model for each tenant, the weight decay for each model was zero, which means that no penalty was made for the weight. However, this does not denote that weight decay generalization for this research does not have a significant effect on the model training. Referring to Fischer and Igel [27], the tuning of weight decay parameter, λ, is hard to select. The grid range in this training was also too large for the value of weight decay as it considered

3 Building Energy Management

57

Fig. 11 Trained ANN-MLP model network plot for Tenant A1 with parameter description

the range from 0 to 0.1. The best weight decay would be assumed to fall in between 0 and 0.0001. From Fig. 11 to 14, it can be observed how the weight, bias and activation function manipulated the structure of the backpropagating neural network. The non-linear relationship between the attributes and the targeted output was generated by providing a weight value internally, during the parsing of data between neurons. The bias was connected to remove any problems which arises if the input to the neurons was zero. For this prediction model, the activation functions were the transfer function, which maps the output of the resulting values between 0 and 1, using Sigmoid (Logistic Activation). After the prediction model has been trained, the model was evaluated by using the test data. Table 13 shows the result of the evaluation.

58

N. A. Ramli and M. K. M. Shapi

Fig. 12 Trained ANN-MLP model network plot for Tenant A2 with parameter description

3 Building Energy Management

Fig. 13 Trained ANN-MLP model network plot for Tenant B1 with parameter description

59

60

N. A. Ramli and M. K. M. Shapi

Fig. 14 Trained ANN-MLP model network plot for Tenant B2 with parameter description

0

0.0001 0.00014678 0.000215 0.000316 0.000464 0.000681 0.001 0.001468 0.002154 0.003162 0.004642 0.006813 0.01 0.014678 0.021544 0.031623 0.046416 0.068129 0.1

Grid range, weight decay

Table 12 Range value of weight decay tuning for model training

3 Building Energy Management 61

62

N. A. Ramli and M. K. M. Shapi

Table 13 Performance evaluation of test prediction for all tenants using the trained ANN model Tenant

WD

RMSE

A1

No. of hidden units 9

0

8.874015

MAPE (%) 5.02

A2

37

0

4.540988

14.16

B1

35

0

20.63566

28.00

B2

35

0

0.547152

60.62

From the performance evaluation shown in Table 13, Tenant A1 has the lowest percentage of error at 5.02%. The prediction model developed for Tenant B2 was the worst performing model in which it has an absolute error percentage of 60.62%. For the purpose of evaluating the model in terms of maximum demand, the model was used to predict the energy demand for the aggregate data. Subsequently after model training and testing, the prediction model generated was compared in terms of performance between algorithms for each tenant. Initially, the result of the testing was observed by comparing the performance between method for individual tenants. The compilation of these results is shown in Table 14, in which the evaluation was observed best on the selected testing evaluation metrics, which are RMSE and MAPE. From the compilation of testing evaluation in Table 14, the SVM method utilizing Radial Basis Function kernel has the best performance for Tenants A1 and A2, in which the RMSE values were 4.7506789 and 3.5898263, respectively. Even though SVM has the best performance, the difference between RMSE value in comparison to k-NN method is minor. The MAPE result also indicates that SVM prediction has lower absolute error percentage. For Tenant B1, k-NN was the best performing Table 14 Performance evaluation of test prediction for all tenants using trained k-NN, SVM and ANN model ([4] Reproduced from) Tenant

Method

A1

k-NN

5.0025748

4.06

3.02

SVM

4.7506789

3.85

2.76

A2

B1

B2

RMSE

NRMSE (%)

MAPE (%)

ANN

8.874015

7.19

5.02

k-NN

3.6548885

11.46

9.98

SVM

3.5898263

11.25

9.38

ANN

4.540988

14.23

14.16

k-NN

14.934312

23.87

15.43

SVM

16.0690844

25.69

12.09

ANN

20.63566

32.99

28.00

k-NN

0.5439403

55.87

48.75

SVM

0.5558279

57.09

43.97

ANN

0.547152

56.20

60.62

3 Building Energy Management

63

method in which the RMSE value was comparably smaller than the other method at 14.934312. But its absolute error was higher than SVM absolute error. This event happened similar to Tenant B2. The best performing algorithm was k-NN with 0.5439403 RMSE value, which was smaller compared to the RMSE result for SVM and ANN. However, the MAPE result for SVM was lower than the result for k-NN. In the case for Tenants B1 and B2 in which the respective RMSE was better, and MAPE result was poorer in comparison to other methods, forecast evaluation technique by Tilmann [29] was used. Based on the technique, the evaluation method must reflect the forecast function, in which if forecasting was made based on the median, the absolute error would obtain a significant result. But if forecast was stated in terms of expectation of a value, the square error method is a much better evaluation method. Thus, the k-NN method was denoted as the best performing algorithm for Tenants B1 and B2. Referring to the normalized RMSE result, it can be observed how each of the prediction method performed differently under different datasets. The performance of every method deteriorated from Tenant A1 to Tenant A2, in which Tenant A2 has the worst performance for every method. Under TNB tariff category, the tenants at both departmental lot A and B were categorized as Medium Voltage General Commercial (Tariff C1). This means that the monthly electricity charges were calculated from acquiring the maximum demand of the month and the kWh [28]. For Tariff C1, the off- and on-peak periods were not applied to the billing process. Therefore, to predict thoroughly for energy billing, the maximum energy demand and the kWh need to be determined. For maximum demand, forecasting an expectation of a value can be done to predict the maximum demand of the month. However, for the energy consumption (kWh) prediction, the hourly predicted demand needs to be added up. As the average value shows the characteristics of the whole dataset, the average forecasted consumption was quantified and compared with the actual average consumption. The percentage difference error between the forecasted and actual value would determine the best forecasting method. The comparison between actual and forecasted average consumption was tabulated and visualized in the form of line graph, as shown in Figs. 15, 16, 17 and 18. From Fig. 15, it can be observed that the ANN forecasted average consumption has a stagnant peakness in which the forecasted average was almost the same for every month. The SVM forecasted average consumption, on the other hand, has better forecasted performance in which the line graph was much closer to the actual average consumption line graph. The k-NN forecast has a slightly larger error in comparison to SVM forecasted result. Comparison for Tenant A2 predicted and actual consumption in Fig. 16 shows that both predictions of average consumption made by k-NN and SVM method have better fit to the actual values than ANN method. In November, it can be observed that all methods forecasted a lower consumption for tenant A2 than the actual consumption. In Fig. 17, the forecasted average consumption by k-NN and SVM was better than ANN. It was also perceived that all methods do not take into consideration the individual high consumption in the month of June. The performance of all methods in Tenant B1 was approximately similar as the performance of predicted consumption

64

N. A. Ramli and M. K. M. Shapi

Fig. 15 Comparison between the actual and forecasted average demand for Tenant A1. Reproduced from [4]

Fig. 16 Comparison between the actual and forecasted average demand for Tenant A2. Reproduced from [4]

Fig. 17 Comparison between the actual and forecasted average demand for Tenant B1. Reproduced from [4]

3 Building Energy Management

65

Table 15 Mean Absolute Percentage Error (MAPE) of forecasted method for all tenants Tenant

k-NN

SVM

A1

0.40

0.241318507

ANN 1.108522675

A2

0.942855477

0.666018364

1.841600488

B1

8.596963137

8.174497001

18.03425668

B2

24.61323638

17.78423714

29.0736946

Fig. 18 Comparison between the actual and forecasted average demand for Tenant B2. Reproduced from [4]

in Tenant B2. Identification of the best method to determine the average consumption on monthly basis was made by calculating the MAPE value for every month. This calculated value was tabulated in Table 15, whereas Fig. 19 shows a bar graph that compares the percentage of difference errors. From Fig. 19 and Table 15, a conclusion can be made in which SVM forecasting is the best method to forecast monthly average consumption. The difference between percentage error of SVM method with k-NN was not significant. Based on the graph, average consumption predicted by ANN method has the highest percentage multiple times in every tenant. This denotes that ANN method is inferior compared to SVM and ANN methods. Comparison in terms of the model training time was also performed to determine the method which has the fastest running time. From the table, k-NN method was the fastest to finish training while the SVM method takes the longest time. On the contrary, ANN method was in the middle but it still takes some time to complete the training. From all the comparison, it is safe to state that SVM was the best method to predict the individual peak energy demand for department lot A; on the other hand, k-NN is the best method for department lot B. In terms of average energy consumption, SVM method has proven to be the best method to predict the monthly mean value for energy consumption. However, high training time was required for SVM model training to achieve this high performance. This result further supports the “No Free Lunch”

66

N. A. Ramli and M. K. M. Shapi

Fig. 19 Percentage of difference errors between forecasted and actual average consumption for all tenants. Reproduced from [4]

theorem by Wolpert [30], which was discussed by Stenudd [31]. No Free Lunch stated that many scenarios would determine whether a machine learning method would perform much better than the other. In this research, the scenario would be the dataset distribution and the targeted output (max demand or average consumption). Due to the small difference in performance result between k-NN and SVM methods, both of these methods were concluded as an excellent prediction method. The choosing element would be whether the user wants a slightly better accurate result or a faster training method (Table 16).

3 Building Energy Management

67

Fig. 19 (continued)

Table 16 Model training running time

Tenant

k-NN

SVM

ANN

A1

37.916s

18h 38m 55.324s

4h 39m 30.035s

A2

28.978s

17h 23m 32.637s

5h 14m 22.311s

B1

34.350s

13h 20m 3.722s

5h 13m 9.418s

B2

34.5s

12h 43m 48.147s

6h 34m 48.972s

68

N. A. Ramli and M. K. M. Shapi

The case study presented in this section showed that machine learning method is a powerful tool that can be used to predict energy consumption. Hence, energy consumption can be managed smartly in the commercial building that is equipped with IoT technology. In the last section in this chapter, we are going to present another case study on how to design a smart energy management system by using techniques such as big data analytics.

2.3 Energy Management System The development of Internet of Things (IoT) and machine learning provides a better energy management system as now the building owner or user is able to predict the energy consumption of the building, and hence they can apply energy efficiency in order to reduce the electricity cost. However, dataset collected from sensors or IoT meters not always fits in one Excel file as it depends on the frequency of the data collected. The truth is most of the data collected is big data and proper data management is needed for this case. To propose a better data management system, IoT and big data analytics (IBDA) is proposed in this case study. The structure of this IBDA was adapted from the existing energy management system (EMS) similar to previous case study in the prior section. Data collected will be stored in the cloud instead of using a server. Figure 20 shows a process flow that represents all the procedures that have been developed to analyse the energy consumption and decision for the system to monitor related appliances based on the proposed method. The data collected from the IoT meters were used to test the cloud storage and website energy profiling developed for this case study. When the programme is stable, the IBDA will be developed at the selected building. Figure 21 shows the work flow starting from existing IoT meter environment from a selected building to be completed as IBDA architecture as element of storage and visualization is added. Not all of the elements reviewed on the existing EMS from the EMS provider will be imitated and constructed on the proposed system. Maximum demand and load factor will not be discussed as the amount of electricity used is in monetary value. To be able to map into TNB requirement, the meters only collect necessary data such as current, voltage, energy, power, humidity and temperature. All the collected data were stored in an open cloud source. Anyone that needs the data can extract it as a CSV file through phpMyAdmin. The data were optimized through MySQL indexing and presented in the website developed through HTML, JavaScript and CSS command. An example of the pseudocode for the data generation is shown in Fig. 22. All the data were being stored in the database to be scaling accordingly to its classes by MySQL engine. MySQL engine provided a way to distribute the processing of the analytics algorithms across a large number of systems. This operation did not require any bulk server as it only used cloud system of an open-source cloud provided by 000webhost. The database can be viewed as in Fig. 23. The last part is the website where it serves as the monitoring site for the project. The website provides any kind of visualization where, for example, user can choose

3 Building Energy Management

69

Fig. 20 The design framework algorithm for the proposed big data analytics method. Reproduced from [2]

to see a correlation between time, temperature and energy. To save cost, the website was developed under a free source. The front access is a type of interface framework of bootstrap normally used for user recognition, while framework of the system body is codeIgniter, and the interface using html and JavaScript. Finally, the back end of the system was developed by using PHP and chart.js was utilized for the chart framework. The example given in this case study shows that a simple energy management system can be built using a free database and free website. Of course a more complex and convenient database and user interface can be developed by using various technologies available today. This case study just shows that it is not impossible or hard to implement smart energy management system based on technologies available today.

70

N. A. Ramli and M. K. M. Shapi

Fig. 21 Extension workflow of existing IoT environment. Reproduced from [2]

Fig. 22 From PHP to chart.js coding. Reproduced from [2]

3 Building Energy Management

71

Fig. 23 The phpMyAdmin database interface. Reproduced from [2]

Hence, building owner, engineer or tenants in a commercial building can start to manage the energy consumption of the building today to cut electricity bill and save the environment by reducing carbon emission.

References 1. Hamid MFA, Ramli NA, Syawal Nik Mohd Kamal NMF (2017) An analysis of energy performance of a commercial building using energy modelling. In: 2017 IEEE conference on energy conversion (CENCON), pp 105–110. https://doi.org/10.1109/CENCON.2017.8262467 2. Hashim NHN, Ramli NA (2019) Smart building energy management using big data analytic approach. In: 2019 13th international conference on mathematics, actuarial science, computer science and statistics (MACS), pp 1–6. https://doi.org/10.1109/MACS48846.2019.9024800 3. Mazlan NL, Ramli NA, Awalin LJ, Ismail MB, Kassim A, Menon A (2020) A smart building energy management using Internet of Things (IoT) and machine learning. Test Eng Manag 83:8083–8090 4. Shapi MKM, Ramli NA, Awalin LJ (2021).Energy consumption prediction by using machine learning for smart building: case study in Malaysia. Dev Built Environ 5 5. Ahmad AS, Hassan MY, Abdullah H, Rahman HA, Majid MS, Bandi M (2012) Energy efficiency measurements in a Malaysian public university. In: 2012 IEEE international conference on power and energy (PECon). Kota Kinabalu, Malaysia 6. Akkaya K, Guvenc I, Aygun R, Pala N, Kadri A (2015) IoT-based occupancy monitoring techniques for energy-efficient smart buildings. IEEE Wirel Commun Netw Conf Work 7. Al-Ali AR, Zualkernan IA, Rashid M, Gupta R, Alikarar M (2017) A smart home energy management system using IoT and big data analytics approach. IEEE Trans Consum Electron 8. Kaytez F, Taplamacioglu M, Ertugul C, Hardalac F (2015) Forecasting electricity consumption: a comparison of regression analysis, neural network and lest squares support vector machines. Int J Electr Power Energy Syst. https://doi.org/67.10.1016/j.ipes.2014.12.036

72

N. A. Ramli and M. K. M. Shapi

9. Xu D, Li Z, Yang S, Lu Z, Zhang H, Chen W (2018) A classified identification deep-belief network for predicting electric-power load. In: 2018 2nd IEEE conference on energy internet and energy system integration (EI2), pp 1–6 10. Valgaev O, Kupzog F, Schmeck H (2016) Low-voltage power demand forecasting using knearest neighbors approach. In: IEEE innovative smart grid technologies—Asia (ISGT—Asia), pp 1019–1024. https://doi.org/10.1109/ISGT-Asia.2016.7796525 11. González-Briones A, Hernández G, Corchado J, Omatu S, Mohamad M (2019) Machine learning models for electricity consumption forecasting: a review. IEEE Virtual-LedgersTecnologías DLT/Blockchain y Cripto-IOT. 12. Ben-Hur A, Ong C, Sonnenburg S, Scholkopf B, Ratsch G (2008) support vector machines and kernels for computational biology. (F. Lewitter, Ed.) PLoS Comput Biol 4(10):1–10. https:// doi.org/10.1371/journal.pcbi.1000173 13. Liu Z, Wu D, Liu Y, Han Z, Lun L, Gao J et al (2019) Accuracy analyses and model comparison of machine learning adopted in building energy consumption prediction. Energy Explor Exploit 1–26. https://doi.org/10.1177/0144598718822400 14. Liu Y, He X, Xu B (2010) Evaluation and comparison of compactly supported radial basis function for kernel machine. In: IEEE international conference on intelligent systems and knowledge engineering. Hangzhou, pp 310–314. https://doi.org/10.1109/ISKE.2010.5680863 15. Kuhn M, Johnson K (2013) Applied predictive modelling. Springer, New York. https://doi.org/ 10.1007/978-1-4614-6849-3 16. Karunathilake SL, Nagahamulla HR (2017) artificial neural networks for daily electricity demand predicitons of Sri Lanka. In: International conference on advances in ICT for emerging regions (ICTer), pp 128–133 17. Tamizharasi G, Kathiresan S, Sreenivasan K (2014) Energy forecasting using artificial neural networks. Int J Adv Res Electri Electron Instrum Eng 3(3):7568–7576 18. Weissbart L, Picek S, Batina L (2019) On the performance of multilayer perceptron in profiling side-channel analysis. Cryptology ePrint Archive, Report 2019/1476. Accessed from https:// eprint.iacr.org/2019/1476 19. Moghaddasi H, Rabiei R, Ahmadzadeh B, Faranbakhsh M (2017) Study on the efficiency of a multi-layer perceptron neural network based on the number of hidden layers and nodes for diagnosing coronary-artery disease. Jentashapir J Health Res. In Press. https://doi.org/10.5812/ jjhr.63032 20. Yan H, Jiang Y, Zheng J, Peng C, LI, Q. (2006) A multilayer perceptron-based medical decision support system for heart disease diagnosis. Expert Syst Appl 30(2):272–281. https://doi.org/ 10.1016/j.eswa.2005.07.022 21. Gnecco G, Sanguineti M (2009) The weight-decay technique in learning from data: an optimization point of view. Comput Manag Sci https://doi.org/10.1007/s10287-008-0072-5 22. Zhang G, Wang C, Xu B, Grosse R (2019) Three mechanism of weight decay regularization. ArXiv, abs/1810.12281 23. Botchkarev A (2018) Evaluating performance of regression machine learning models using multiple error metrics in azure machine learning studio. SSRN Electron J 1–16. Evaluating Performance of Regression Machine Learning Models Using 24. Budiman F (2019) SVM-RBF parameters testing optimization using cross validation and grid search to improve multiclass classification. Sci Visualization 11(1):80–90. https://doi.org/DOI: 10.26583/sv.11.1.07 25. Panchal FS, Panchal M (2014) Review on methods of selecting number of hidden nodes in artificial neural network. Int J Comput Sci Mobile Comput 3(11):455–464 26. Gaurang P, KostaYP, Ganatra A, Panchal D (2011) Behaviour analysis of multilayer perceptrons with multiple hidden neutrons and hidden layers. Int J Comput Theory Eng 3(2):332–337. https://doi.org/10.7763/IJCTE.2011.V3.328 27. Fischer A, Igel C (2012) An introduction of restricted boltzmann machines. In: Progress in pattern recognition, image analysis, computer vision, and applications: 17th iberoamerican congress, CIARP 2012. Buenos Aires, Argentina, pp 14–36 28. Tenaga Nasional Berhad (2006) Tenaga Nasional Berhad Tariff Book

3 Building Energy Management

73

29. Tilmann G (2010) Making and evaluating point forecast. J Am Stat Assoc 106(494):746–762. https://doi.org/10.1198/jasa.2011.r10138 30. Wolpert DH (1996) The lack of a priori distinctions btewen learning algorithm. Neural Comput 1341–1390 31. Stenudd S (2010) Using machine learning in the adaptive control of a smart environment. VTT Publications, Vuorimiehentie, Finland, p 751

Chapter 4

Demand-Side Management and Peak Load Reduction Transaction Mode of Multi-user Demand Response Market Based on Controllable Peak Load Capacity Hongming Yang, Jingshu Yang, Sheng Xiang, Yan Xu, and Yibo Wang Abstract The real-time generation/load balance of power system determines the frequency stability of the system, which is the key to ensure the safe and stable operation of the system. In the traditional power system, due to the lack of effective control means for end-user behavior, the system power balance can only be guaranteed by adjusting the generation power. At this time, in the peak period of power consumption, if the user’s power consumption behavior is changed through demand response, the load in the peak period can be reduced, so as to realize the real-time power balance of the system. Therefore, this chapter proposes a transaction mode of the multi-user demand response market based on controllable peak load capacity. Firstly, combined with the user big data generated by the smart grid, this chapter perceives the user’s power consumption behavior and demand response potential. Secondly, in order to maximize the economic, security, and carbon reduction benefits brought by demand response, this chapter proposes a multi-stage market trading mode of medium-term and long-term monthly trading market and special market for peak demand response, studies the market mechanism of demand response, and analyzes the market and technical functions of load aggregators.

H. Yang · J. Yang · S. Xiang · Y. Xu (B) School of Electrical and Information Engineering, Changsha University of Science and Technology, International Joint Laboratory of Ministry of Education for Operation and Planning of Energy Internet Based on Distributed Photovoltaic-Storage Energy, Hunan Provincial Engineering Research Center for Electric Transportation and Smart Distribution Network, Changsha, China e-mail: [email protected] H. Yang e-mail: [email protected] S. Xiang e-mail: [email protected] Y. Wang Wuhan Kemov Electric Co, Ltd, Wuhan, China e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 A. Tomar et al. (eds.), Control of Smart Buildings, Studies in Infrastructure and Control, https://doi.org/10.1007/978-981-19-0375-5_4

75

76

H. Yang et al.

1 Introduction In recent years, the power demand on the demand-side continues to grow, and a large scale of renewable energy is incorporated into the power grid, which makes it more difficult for the power system to maintain the generation/load balance in real time. On the one hand, the power required by users for cooling and heating is increased when facing the extreme weather of high temperature in summer and cold waves in winter, resulting in the widening of peak–valley difference. In 2019, the maximum peak load in a certain region of China is 29926 MW, and the annual maximum peak–valley difference reaches 58%. The continuous hours of 95% and above of the annual maximum load are 53 h, accounting for only 0.6% of the total hours of the whole year. To meet the peak load demand of this part, the power system needs to invest in the corresponding power generation, transmission, transformation, and distribution capacity. The reserve is prepared in proportion. The capacity utilization rate and the utilization hours of units both are low, resulting in a great waste of social resources [30]. On the other hand, wind power, photovoltaic, and other new energies are connected to the grid in a high proportion. The power generation output has obvious randomness and fluctuation, and the output peak is not synchronized with the peak, which leads to the further increase of peak shaving and frequency modulation pressure in the power system. The safe operation of the power system is greatly challenged. Therefore, demand response is actively implemented in power system operation [27]. Through economic or technical means, power grid companies (its main functions are maintaining and dispatching power system and organizing power market transactions) guide power users to change their power consumption behaviors at an appropriate time and provide power regulation equivalent to power generation resources, so as to achieve the goal of shaving peak and filling valley, ensuring system safety, and reducing system cost [12]. In the power system with large peak–valley differences, when the power supply is tight when the power generation output is less than the power consumption output, the power grid company can take orderly power consumption to manage the power consumption of end-users, including peak staggering, peak avoidance, turn off the power in turn, etc. During the peak period, some industrial loads are forcibly cut off to alleviate the contradiction between supply and demand, to ensure the safe and stable operation of the power system. However, this demand-side management method based on orderly power consumption leads to the reduction of power supply service quality and affects the customer income of production and business processes [4]. Therefore, the enthusiasm of end-users to participate in power system dispatching is not high, which increases the difficulty of peak load reduction. In a word, it is of great significance to study the demand response transaction model for power market and deeply tap the controllable potential of users. According to the seasonal peak load characteristics and demand response potential of large industrial users, general industrial and commercial users, and urban and rural residents, and on the premise of ensuring the production and operation income of various end-users, this chapter will propose a multi-stage market transaction mode under the typical peak scenario

4 Demand-Side Management and Peak Load Reduction

77

is adopted to encourage end-users to participate in demand response independently. This not only promotes the transformation from the traditional model of “source following load” to the collaborative model of “source-network-load interaction” but also reduces the load in peak hours through market means [19]. Then, the real-time generation/load balance is maintained and the safety and economy of power system operation are guaranteed.

2 Analysis of Load Characteristics During Peak Hours Since the thirteenth 5-Year Plan, China economy has achieved great development, and the per capita income and quality of life have been significantly improved. With the rapid development of the economy, the electricity demand of end-users is growing rapidly too.

2.1 Overall Characteristic Analysis of Peak Load in Power System A region of China is taken as an example to show the overall characteristic of peak load in the power system. The load demand increases year by year, and the annual load curve presents a “W” shape with double peaks and double valleys, as shown in Fig. 1. The peak load occurs in July and August during summer and December and January during winter. The valley load occurs in April and May during spring and October during autumn. Affected by the shutdown during the epidemic, the average load in spring 2020 reached the lowest in February. In the past 3 years, the variation trend of daily load in summer in this region is the same, maintaining a high level during the period of 10:45–23:30 every day. The peak period in the daytime is 11:15–17:15 every day. During this period, except that 26000 24000 Load (MW)

Fig. 1 Average load curve of annual range from 2018 to 2020

22000 20000 18000 16000 14000

2018

2019

2020

78

H. Yang et al.

Fig. 2 Typical daily load curve in summer 2018–2020

34000

Load (MW)

32000 30000 28000 26000 24000 22000 20000

2018.07.20

2019.08.20

2020.08.04

the load decreases briefly during 11:45–12:45, the load in other periods maintains a stable value and fluctuates in a small range. The peak period at night is 19:45–23:15 every day. During the evening peak period, the load curve presents an “inverted U” shape. The load reaches the highest value during 20:30–21:45, and the maximum load during the peak period at night is higher than the diurnal peak, as shown in Fig. 2. In the past 3 years, the variation trend of daily load in winter in this area is the same, which has peak and valley periods. The peak period is 8:30–22:30. The valley period is from 23:15 of the current day to 7:45 of the next day. During the peak period, the load curve is “m” type, showing obvious noon peak and evening peak. The daily noon peak occurs from 10:45 to 12:00. The evening peak occurs during the period of 16:45–19:45, and the load fluctuates in a small range. Meanwhile, the maximum load during the noon peak is slightly higher than that in the evening peak period, as shown in Fig. 3. The load exceeding 95% of the maximum load in this area mainly occurs in 9:15– 23:30. The load is normally distributed around 11:30, 13:15, 18:30, and 20:30, where 11:30 and 20:30 have higher frequencies than the other two times, as shown in Fig. 4. Fig. 3 Typical daily load curve in winter 2017–2019 Load (MW)

29000 24000 19000 14000

2018.01.26

2018.12.29

2020.01.10

Fig. 4 Duration hours of load exceeding 95% of maximum load in 2018–2019

Number of peak occurrences

4 Demand-Side Management and Peak Load Reduction

79

12 10 8 6 4 2 0

2018

50 Number of peak occurrences

Fig. 5 Duration hours of load exceeding 90% of maximum load in 2018–2019

2019

40 30 20 10 0 09:1510:4512:1513:4515:1516:4518:1519:4521:1522:45 2018 2019

In 2018 and 2019, the load exceeding 90% of the maximum load occurs 10 times and 48 times (at 11:30), 2 times and 27 times (at 13:15), 14 times and 27 times (at 18:30), 7 times and 40 times (at 20:30), respectively. The load exceeding 90% of the maximum load in this area mainly occurs during 10:30–22:45, as shown in Fig. 5. In 2018, the load exceeding 90% of the maximum load occurred once at 10:30–12:00 every day. The load is normally distributed around 18:00 (occurs three times at 18:00 and 18:15). In 2019, the load exceeding 90% of the maximum load occurs more frequently at 11:30 and 21:15 (8 and 12 times, respectively). Near these two moments, the number of occurrences obeys approximately normal distribution. During 12:30–20:00, the occurrence times of loads exceeding 90% of the maximum load are maintained 2–4 times.

2.2 Analysis of the Influence of Severe Cold and Heat on Peak Load In summer, the daily load change trend is consistent with the temperature change trend between 22:00 and 18:45 (the next day). At 18:45–22:00, the load has an

80

H. Yang et al.

34 25000 29

Temperature ( )

Load (MW)

30000

24

20000

2018.07.20 Temperature

2019.08.20 Temperature

2018.07.20 Load

2019.08.20 Load

Fig. 6 Typical daily load and temperature curve in summer 2018–2019

evening peak, which is related to the living habits of residents, but has no obvious relationship with the temperature, as shown in Fig. 6. Take 2019 as an example, during 6:00–15:00 on a summer day, the temperature rises from 24.7 °C to 37.6 °C, with an increase of 34.3%. During the period 5:45– 13:45, the load fluctuates slightly but shows an overall increasing trend. The load increases from 21,036 MW to 29,583 MW, with a load increase of 28.9%. During 15:00–19:00, the temperature decreases from 37.6 °C to 32.8 °C, with a decrease of 12.8%. The load decreases from 29,050 MW to 26,738 MW, with a load decrease of 8.0%. After 19:00, the load does not share the same trend with the temperature, because residential occurrence increases the load demand. In winter, the daily load changes in two periods: peak load and valley load, which are mainly affected by residential activities. The temperature fluctuates in a small range throughout the day, and the changing trend has no obvious relationship with the changing trend of daily load, as shown in Fig. 7. Take 2018 as an example, the temperature fluctuates in a small range from 14:00 to 18:00. During this period, the maximum and minimum temperatures are −1.4 °C and −1.5 °C, respectively, and the maximum fluctuation rate is 6.7%. The load decreases continuously from 14:00 to 15:30, and then increases continuously.

2.3 Distribution Characteristics of Peak Load There have three types of end-users in the power system: large industrial users, general industrial and commercial users, and residential users [13, 30]. The daily load of large industrial users relates to the peak–valley price, and the overall load fluctuation is small. From 23:30 to 7:00 (the next day), the load shows a downward trend, and the price is low. To save the electricity cost, some production plans can be transferred into this period, which helps to increase the valley load. During the lunch

81

31000

4

29000

3

27000

2

25000

1

23000

0

21000 19000

-1

17000

-2

15000

-3

2018.12.29Temperature 2018.12.29Load

Temperature ( )

Load (MW)

4 Demand-Side Management and Peak Load Reduction

2020.01.10Temperature 2020.01.10Load

Fig. 7 Typical daily load and temperature curve in winter 2017–2019

break from 10:15 to 14:00, the industrial load has a decreasing trend and reaches the minimum value at 12:00. During 15:00–22:00, the price is high, and the load continues to decrease. After 22:00, the load increased sharply and returned to the high load due to the low electricity price. The daily load of general industrial and commercial end-users in summer relates to their operating characteristics, which can be divided into two load segments: peak period and valley period. From 22:45 to 8:00 (the next day), these end-users are closed and consume small electricity. During the business hours from 8:30 to 21:30, the load is at a high level, and the load remains relatively stable during 10:30–17:00 and 18:00–21:00. The daily load of urban and rural residents in summer relates to residential occurrence habits. From 5:15 to 21:30, the load generally shows an upward trend and reaches the maximum value at 21:30. Meanwhile, during this period, the load forms a noon peak from 12:00 to 14:00. From 21:45 to 5:00, the load shows a downward trend and reaches the minimum value at 5:00, as shown in Fig. 8. The changing trend of daily load in winter is consistent with that in summer. General industrial and commercial users show a downward trend from 10:30 to 21:00, but the overall change is relatively stable, as shown in Fig. 9. Overall, the load characteristics during peak hours are mainly as follows: a.

Daily load change is mainly related to the temperature. The load shows different trends in winter and summer. The peak load occurs in different periods, but they all show two peak segments: noon peak and evening peak. The summer afternoon peak is 11:15–17:15. The evening peak is from 19:45 to 23:15. The maximum load values of the three users in the evening peak are 3.1%, 1.1%, and 5.9% higher than those in the peak period during the day. The winter noon peak is 10:45–12:00. The evening peak is 16:45–19:45. The maximum load values of

82

H. Yang et al.

Fig. 8 Load curve of each user on typical day in summer

11000

Load (MW)

9000 7000 5000 3000

Large industrial General Industry and Commerce Resident

Fig. 9 Load curve of each user on typical day in winter

20000

Load (MW)

15000 10000 5000 0

Large industrial General Industry and Commerce Resident

b.

c. d.

the three users in the afternoon peak hours are 3.6%, 2.4%, and 1.6% higher than those in the evening peak, respectively. The peak load time covers a wide range, but it is mainly concentrated in the noon peak and evening peak. The occurrence times at other times are relatively small, and the fluctuation of occurrence times is small too. The daily load changes of large industrial users, general industrial and commercial users, and residential users are related to human activities. The electricity price affects the load demand.

2.4 Potential of Multi-user Demand Response The controllable capacity of the load is generated by changing the original demand mode according to the incentive mechanism. In other words, the load can realize the

4 Demand-Side Management and Peak Load Reduction

83

flexible change of demand increase and decrease within a certain range [11]. There are a large number of load resources with controllable capacity potential on the demand side. These resources have large quantities, small capacities, and diverse subjects. Such as smelting furnaces for large industrial users, central air conditioning and electric furnaces for general industrial and commercial users, and electric water heater for residential users. If the controllable potential of these load resources is analyzed and included in the scope of power system regulation, the regulation capacity of the power system will be greatly improved and the peak shaving capacity will be increased. In the incentive-based demand response, the motivation of demand-side agents to reduce load comes from the requirements of power system dispatching or the inductive compensation measures of market operators. This section takes the compensation price as the incentive signal to encourage end-users to provide controllable load and participate in power system dispatching. The price elasticity coefficient is introduced to measure the degree of response, which is defined as the change of demand divided by the compensation price. The greater the price elasticity coefficient of load, the more sensitive the load is to price changes, and the greater the response to incentive-based demand response measures.

2.4.1

Analysis of the Controllable Potential of Large Industrial Users

Typical large industrial users are mainly divided into steel and non-ferrous metal industry, building materials industry, chemical industry, paper industry, and machinery manufacturing industry [22]. According to the characteristics of load equipment in large industrial users, the loads are classified according to the standards of protective load, main production load, auxiliary production load, and unproductive load [26]. Among them, the protective load is the load that will cause dangers after power failure. The main production load and auxiliary production load will not cause danger after short-term power failure and only affect the production progress. Unproductive loads are loads other than those mentioned above. They have high power-off flexibility and will not have a significant impact.

Steel Industry The steel industry is a typical high energy-consuming industry. Generally, the 24h three-shift continuous working system is adopted. There is lots of continuous production equipment, such as a sintering machine, coke oven, blast furnace, etc. In addition to the maintenance time, this equipment usually operates at full load, with a high load rate and high requirements for power quality. The load fluctuation throughout the day is small, and there are no obvious peaks and valleys, as shown in Fig. 10. The main production loads of the steel industry include electric arc furnace, blast furnace, converter, continuous caster, sintering machine, etc. These loads account

84

H. Yang et al.

Fig. 10 Daily load characteristic curve of typical steel enterprises

300000

Load (kW)

250000 200000 150000 100000 50000 00:00 01:30 03:00 04:30 06:00 07:30 09:00 10:30 12:00 13:30 15:00 16:30 18:00 19:30 21:00 22:30

0

Steel Industry

for more than 75% of the total load. Auxiliary production load includes water pump, hydraulic pump, and axial flow fan, etc. These loads account for about 8% of the total load. The protective load includes exhaust gas and dust recovery suction fan, circulating cooling water pump, public protective electrical equipment, etc. These loads account for more than 10% of the total load. Unproductive loads include office electrical equipment, central air conditioning, etc. These devices account for 2–5% of the total load. Due to the continuity of production load in the iron and steel industry, the regulation mode is user-independent load control. The regulation time is related to the characteristics of user equipment. Unproductive load accounts for a relatively small proportion. The regulation mode of this kind of load is direct control of the power system. The preparation and recovery time of load control can reach the second level, and the response time is 0.5–2 h. The controllable proportion of productive load is 19%. The controllable proportion of Unproductive load is about 1%.

Cement Industry The production process of the cement industry usually includes ore mining, ore transmission, ore crushing, rotary kiln firing, clinker grinding, and packaging. To save energy costs, the cement industry mainly produces in flat and low electricity price periods. Therefore, the peak–valley load inversion of the industry is obvious, as shown in Fig. 11. The cement industry peaks occur at 12:00–17:00 and 21:00–8:00. The load is high and stable at night. The operation cycle of production equipment in cement industries is long, the load demand is large, and the requirements of power supply reliability are high. The main production facilities of the cement industry include the raw mill, cement mill, ball grinding mill, etc. These facilities are used for raw material grinding, raw material firing, and cement grinding. These facilities account for 55–60% of the total electricity consumption.

4 Demand-Side Management and Peak Load Reduction Fig. 11 Daily load characteristic curve of typical cement enterprises

85

25000

Load (kW)

20000 15000 10000 5000 00:00 01:30 03:00 04:30 06:00 07:30 09:00 10:30 12:00 13:30 15:00 16:30 18:00 19:30 21:00 22:30

0

Cement Indistry

Auxiliary production facility includes the transmission fan, transmission belt motor, etc. These facilities are used for the transmission of raw materials, the delivery of finished cement products, the manufacture of intermediate products, etc. These facilities account for 15–20% of the total electricity consumption. Protective facilities include the cooling water pump, lubricating oil pump, etc. These devices are used for cooling of the cement production process, lubrication of rotating equipment, etc. The load accounts for 8–15% of the total electricity consumption. Unproductive loads include office electrical equipment, central air conditioning, etc. The load accounts for 2–5%. The production period of the cement industry is not limited by process technology. The purpose of continuous production is to improve production efficiency. Therefore, the production period of the cement industry can be transferred, the cement industry has interruptible potential. The cement industry is suitable to participate in demand response. When the production plan can be flexibly adjusted, the controllable load accounts for about 24% of the total electricity consumption.

Non-ferrous Metal Industry The non-ferrous metal industry is generally a 24-h working system with three shifts. Electrolytic aluminum enterprises are taken as an example in this section. The electrolytic direct current system and important auxiliary production system are Class I loads, which account for about 95% of the total electricity consumption in the plant. These Class I loads require high reliability of power supply. The typical daily load characteristic curve of the electrolytic aluminum enterprise is shown in Fig. 12. The main production load of electrolytic aluminum enterprises includes aluminum electrolytic cells, casting furnaces, casting machines, etc., which accounts for more than 75% of the total load. Auxiliary production load includes the functional crown block, air compressor station, water pump station, draft fan, etc., which accounts for 5–10%. The protective load includes wastewater and waste residue treatment unit, flue gas recovery unit, etc., which accounts for 3–10%. The unproductive load

86 6000 5000 Load (kW)

Fig. 12 Daily load characteristic curve of typical electrolytic aluminum enterprises

H. Yang et al.

4000 3000 2000 1000 00:00 01:15 02:30 03:45 05:00 06:15 07:30 08:45 10:00 11:15 12:30 13:45 15:00 16:15 17:30 18:45 20:00 21:15 22:30 23:45

0

Non-ferrous Metal Industry

7000 6000 Load (kW)

Fig. 13 Daily load characteristic curve of typical general and special manufacturing enterprises

5000 4000 3000 2000 1000 00:00 01:15 02:30 03:45 05:00 06:15 07:30 08:45 10:00 11:15 12:30 13:45 15:00 16:15 17:30 18:45 20:00 21:15 22:30 23:45

0

Manufacturing Industry

includes office electrical equipment, central air conditioning, etc., which accounts for 1–5%. The production period of electrolytic aluminum enterprises is not limited by process technology. Therefore, the production period can be transferred and interrupted. The enterprise is suitable to participate in demand response. The main production load of electrolytic aluminum enterprises accounts for 75% of the total load demand. The controllable load in the main production load accounts for 18% of the total controllable load. When the production plan can be adjusted flexibly, the controllable load accounts for about 22% of the total production load.

Manufacturing Industry The manufacturing industry includes a wide range, most enterprises adopt a discontinuous production system. The main production time is 8:00–21:00, as shown in Fig. 13. The peak–valley difference is large. The load curve of a small number of continuous production users is relatively stable, with a fluctuation of about 20%. The main production loads of manufacturing industries include heat treatment furnace, melting furnace, high-frequency furnace, casting machine, electric welding

4 Demand-Side Management and Peak Load Reduction 5000 4500 4000 3500 3000 2500 2000 1500 1000 500 0

00:00 01:15 02:30 03:45 05:00 06:15 07:30 08:45 10:00 11:15 12:30 13:45 15:00 16:15 17:30 18:45 20:00 21:15 22:30 23:45

Load (kW)

Fig. 14 Daily load characteristic curve of typical textile enterprises

87

Textile Industry

machine, wire drawing machine, etc., which account for more than 60% of the total load demand. Auxiliary production loads include air conditioners and cooling pumps for production equipment, which account for more than 5%. Protective loads include computer numerical control machine tools, workshop lighting, public protective electrical equipment, etc., which account for more than 5% of the total load demand. Unproductive loads include office electrical equipment, central air conditioning, etc., which accounts for 5–10%. Manufacturing enterprises have the highest controllable capacity from the main production load, accounting for 16% of the total controllable load. Controllable equipment includes melting furnace, blower, dryer, etc. The regulation mode of the melting furnace is mainly the user-independent control mode. Blower and dryer can be directly controlled by the power system. The controllable capacity of auxiliary load and unproductive loads are 3% and 1%, respectively. The regulation mode of the auxiliary load is user-independent control. The regulation mode of the unproductive load is directly controlled by the power system. The maximum controllable load of manufacturing enterprises accounts for about 20% of the total electricity consumption.

Textile Industry As a typical high energy-consuming industry, the textile industry generally adopts a 24-h three shift continuous working system. The load fluctuates slightly throughout the day, most devices are working in the continuous production mode. In addition to the maintenance time, these devices usually operate at full load, with a high load rate and high requirements for power quality, as shown in Fig. 14. The main production load of the textile industry includes spinning machines, texturing machines, hot boxes, air transformers, etc., which accounts for more than 60% of the total load demand. Auxiliary production load includes the workshop ventilation, axial flow fan, etc., which accounts for more than 5%. Protective load includes the exhaust gas and dust recovery suction fan, circulating cooling water

88

H. Yang et al.

pump, public protective electric equipment, etc. Unproductive loads include office electrical equipment, central air conditioning, etc., which accounts for 5–10%. The main load of the textile industry is concentrated in spinning machines, texturing machines, weaving equipment, and other main production link equipment. These loads account for a high proportion, and most of them have controllable potential. The controllable load accounts for about 31% of the overall load. The auxiliary production load is mainly ventilator and workshop lighting. Controllable load accounts for about 3%. The unproductive load is mainly air conditioning system, office lighting, domestic power, etc. These loads have great controllable potential and account for about 1%. The controllable load of the textile industry can account for up to 35% of the total electricity consumption.

2.4.2

Load Demand Response Modeling of Large Industrial Users

Large industrial user load has the characteristics of large regulation capacity and fast response speed, so it is often used as an important controllable resource on the demand side. However, large industrial users include different types of industrial enterprises with different production processes. Therefore, the power consumption characteristics of various industrial enterprises are also different. This section takes the controllable smelting load (ferroalloy, silicon carbide, etc.) as the research object. By analyzing the relationship between the production process and power consumption, the adjustable power of industrial loads is obtained. The main electrical equipment for smelting load is the smelting furnace. According to the operation form of the equipment, the smelting furnace includes three states: start–stop, production, and furnace drying (maintaining furnace temperature and waiting for production). Thereinto, the start–stop state of the smelting furnace has a high cost and long start–stop time, and the smelting furnace is regarded as uncontrollable at the start–stop state. The controllable power of the process is not considered. When the furnace is at the drying state, reducing the power will make the furnace temperature lower than the production process requirements, resulting in the reduction of product quality. Therefore, the furnace is uncontrollable at the drying state. Hence, the controllable potential of smelting furnace is from the production state, and can be written as PD R,decr eased = xn Pyield

(1)

where PD R,decr eased is the controllable load potential of the smelting furnace; Pyield is the rated power when the smelting furnace is in the production state; and xn is the state variable of the smelting furnace, 1 indicates that the smelting furnace is in the production state, and 0 indicates that the smelting furnace is in the furnace drying state. The duration of furnace participation in demand response ranges from 30 min to several hours. The specific duration depends on the production plan of large industrial users.

4 Demand-Side Management and Peak Load Reduction Fig. 15 Hotel daily load curve

89

8000 7000 Load (kW)

6000 5000 4000 3000 2000 1000 00:00 01:30 03:00 04:30 06:00 07:30 09:00 10:30 12:00 13:30 15:00 16:30 18:00 19:30 21:00 22:30

0

Hotel Users

High energy consumption loads similar to smelting furnaces can participate in incentive-based demand response. The user judges whether to reduce the active power of the smelting furnace according to the compensation price. The provided demand response can be written as Q r eal = ωπ D R

(2)

0 ≤ Q r eal ≤ PD R,decr eased t D R

(3)

where Q r eal is the actual demand response of users, it could not exceed the controllable potential of users; π D R is the compensation price provided by the market operator; and ω is the price elasticity coefficient of the user.

2.4.3

Analysis of Controllable Load Potential for Commercial Users

Hotel Users The load change trend of the hotel is dependent on human behaviors, which can be divided into two load segments: high load segment between 10:00 and 17:15 and low load segment between 22:30 and 6:00. During the two periods, the load remains unchanged and fluctuates only in a small range, as shown in Fig. 15. The power load of the hotel consists of five categories: air conditioning system, lighting system, elevator, dynamical system, and domestic electricity. Thereinto, the load of the air conditioning system accounts for a high proportion, which takes up about 72% of the total load demand. The proportions of the lighting system and domestic electricity are approximately equal, accounting for about 6.5%. The elevator load accounts for the least, only about 1% of the total load, as shown in Fig. 16.

90

H. Yang et al.

Fig. 16 Load proportion in the hotel

air conditionin g system, 71.76%

80% Load proportion (%)

70% 60% 50% 40% power system, 14.04%

30% 20% 10%

lighting system, 6.56%

elevator, 1.17%

domestic electricity, 6.47%

0% Load category

Load (kW)

Fig. 17 Load curve of hotel air conditioning system

180 160 140 120 100 80 60 40 20 0

Air conditioner 1 Chilled water pump

Cooling water pump Air conditioner 1

Air conditioning systems include air conditioning equipment and water pump equipment. Thereinto, the air conditioning equipment accounts for about 75% of the system, and the water pump equipment accounts for about 25%. Affected by the diversity of the hotel business, there are some differences in the load characteristics of different air conditioning devices. According to the different operation curves of air conditioning devices, the air conditioning can be divided into load-changing air conditioning and load-fixed air conditioning. The load-changing air conditioner is mainly the split air conditioner of the customer box. The load change trend of this kind of equipment follows human activities. The peak period of equipment operation is concentrated during the daytime, as air conditioner 1 in Fig. 17. Fixed load air conditioning is mainly the air conditioning equipment required for the 24-h operation of the hotel. Such air conditioning load remains unchanged throughout the day, as air conditioning 2 in Fig. 17. Water pump equipment includes the cooling water pump and chilled water pump. Under normal operation, the load remains stable throughout the day. The load of the air conditioning system is affected by the outside air temperature and is related to the maintained room temperature. The air conditioning system is

4 Demand-Side Management and Peak Load Reduction

91

Table 1 Comparison of effects of various air conditioning load regulation strategies (Unit: KW) Strategy

Maximum load reduction

Average load reduction

Load reduction ratio (%)

User feeling

Close the fresh air unit

174.48

121.89

9.03

Almost none

Adjust the outlet 450.50 water temperature of the main engine

372.12

27.56

Almost none

Turn off the fan coil unit

408.80

326.24

24.17

Yes

Global temperature control

282.63

218.70

16.20

Almost none

adjustable. And the room temperature has a certain thermal inertia. The air conditioning system can quickly respond to the power system dispatching by changing its load. In a short time, the change range of room temperature is small, and the normal operation of commercial activities can be maintained. Therefore, the air conditioning system has a great potential to participate in short-term demand response. The loadchanging air conditioning is volatile and its controllable capacity depends on the load in the response period. The load-fixed air conditioning is adjustable and kept stable throughout the day. The air conditioning system is a complex self-balancing system. Local control of the system will lead to load redistribution. When the operation parameters of the central air conditioning system are changed to provide demand response services, the air conditioning shows nonlinear characteristics. When changing the parameter values of different equipment in the central air conditioning system, the actual response power and adjusted effect are shown in Table 1. It can be seen from the table that the overall regulation effect is the best when the air conditioning load is changed by adjusting the outlet water temperature of the main engine. The maximum load reduction ratio reaches 28% of the total load. This process almost has no impact on end-users. The chilled water pump and cooling water pump can be adjusted by frequency conversion control. The maximum controllable load accounts for 20% of the total load of the water pump. The dynamical system load includes six types of load equipment: kitchen, agricultural materials, laundry, electric auxiliary, frequency conversion machine room, and fire pump. Thereinto, the average load of kitchen and agricultural materials accounts for about 29% and 57%, respectively. The proportion of other loads is relatively small, floating between 2% and 5%, as shown in Table 2. In the dynamical system, the daily load peak of laundry mainly occurs at 7:00– 19:00. At 10:00 and 12:00, the load reaches the full day peak of 15 kW. From 21:00 to 22:00, the laundry stops working and the load is 0. The load of the frequency conversion machine room remains stable throughout the day, as shown in Fig. 18. The load of the laundry is entirely determined by the working hours. And the independent adjustment of working hours has a slight impact on the laundry task.

92

H. Yang et al.

Table 2 Hotel dynamical system load (Unit: KW) Equipment type

Maximum load

Kitchen

72

Minimum load

Average load

Average proportion (%)

2.25

15.99

28.89

Agricultural materials, 32.96

29.36

31.3

56.56

Laundry

16.64

0

3.21

5.80

Electric auxiliary

9.64

0

1.2

2.17

Frequency conversion machine room

4.41

0

0.44

0.80

Fire pump

1.28

1.12

1.2

2.17

Other

4.88

1.36

2

3.61

Load (kW)

Fig. 18 Hotel dynamical system load curve

40 35 30 25 20 15 10 5 0

Laundry

Frequency conversion machine room

Table 3 Domestic electricity load of the hotel (Unit: KW) Equipment type

Maximum load

Minimum load

Average load

Average proportion (%)

Restaurant

75.27

0.15

11.49

45.09

Foot bath

32.8

0.24

7.82

30.67

Dormitory

12.08

0.28

4.2

16.47

8.32

0.2

1.33

7.80

Canteen

During 13:00–14:00 in the noon peak, the laundry will shift its working task to the valley period to provide demand response for the power system. Domestic electrical equipment includes four types of electrical equipment: restaurant, foot bath, dormitory, and canteen. Therefore, the average load of restaurants and foot baths providing entertainment activities accounts for about 45% and 31%, respectively. In contrast, the average loads of dormitory and canteen account for about 16% and 8%, respectively, as shown in Table 3. The load of the canteen is determined by human diet habits. The peak load occurs at 11:00–15:00 (lunch meal) and 19:00–23:00 (dinner meal), as shown in Fig. 19. During lunch meal, the load shows a low–high–low trend, which is uncontrollable. During dinner meal, the load remains unchanged and has strong stability. The noon

4 Demand-Side Management and Peak Load Reduction Fig. 19 Load curve of canteen of hotel

93

14

Load (kW)

12 10 8 6 4 2 0

Canteen

Fig. 20 Lighting load curve

20

Load (kW)

15 10 5 0

Lighting load

peak and evening peak appear in the middle of the meal periods. Meal load is an important part of the peak load in the power system. However, due to the impacts of the normal operation and basic benefits of the hotel, the controllable potential of the canteen load is difficult to tap. The lighting equipment fluctuates irregularly throughout the day. However, the load is always low during 0:00–8:00 and remains relatively high in other periods. During 11:00–12:00 in the noon peak, the load reaches the maximum value of 18.7 kW. On the premise of ensuring visibility, lighting load participates in demand response by turning off unnecessary lighting equipment. The lighting load curve is shown in Fig. 20. The daily average load of the lighting equipment is about 11 kW. The average lighting load is about 16 kW during the noon peak period and is about 11 kW during the noon peak period. To sum up, the operation status of similar equipment is different during the noon peak and evening peak. The controllable load of the hotel is divided into two periods: noon peak and evening peak. The controllable load of the hotel accounts for 25% of the hotel load during the noon peak period and accounts for 21% during the noon peak period.

94

H. Yang et al. 5000 4500 4000 3500 3000 2500 2000 1500 1000 500 0 00:00 01:15 02:30 03:45 05:00 06:15 07:30 08:45 10:00 11:15 12:30 13:45 15:00 16:15 17:30 18:45 20:00 21:15 22:30 23:45

Load (kW)

Fig. 21 Daily load curve of typical commercial complex users

Commercial complex users

Commercial Complex Users The daily load of commercial complex users can be divided into two periods according to business hours. The daily 9:00–21:00 is the high load period. The low load period is from 22:00 of the current day to 8:00 of the next day. The load remains relatively stable in the same period, as shown in Fig. 21. A large commercial complex is a multifunctional, efficient, complex, and unified urban building complex. The main functions include business, office, residence, hotel, exhibition, catering, entertainment, and other functions. The main load comes from the air conditioning system, lighting system, dynamical system, etc. Thereinto, the proportion of air conditioning load reaches more than 50% in summer. The lighting load is relatively stable, accounting for about 20%. The power load mainly comes from the elevator, water pump, etc. This load accounts for about 15%. The refrigeration load of cold storage accounts for about 10%. Lighting loads are mostly non-operating loads, and the control mode is relatively simple. Local line control through disconnection can make the lighting load meet the requirements of load control. During emergency conditions, 100% response from lighting load can be achieved. The cold storage device can be reduced by about 25% by changing the operation mode. The electric boiler can be reduced by 50% through power regulation. Air conditioning load includes central air conditioning, split air conditioning, and varied refrigerant volume (VRV) air conditioning. The central air conditioning has a high proportion of power consumption and good control effect, which is the primary controllable load in buildings. The load regulation time of the central air conditioner generally lasts about 30 min. The duration can maximize the thermal insulation effect of the building body and reduce the impact on end-users. The split air conditioning reduces its load demand by changing the setup temperature. In summer, the split air conditioning is in the cooling mode. When the setup temperature increases by 1 °C, the load can be reduced by about 20%. In winter, the split air conditioning is in the heating mode. When the setup temperature decreases by 1 °C, the load can be reduced by about 10%. The VRV air

4 Demand-Side Management and Peak Load Reduction Fig. 22 Daily load characteristic curve of a typical office building

95

300000 Load (kW)

250000 200000 150000 100000 50000 00:00 01:30 03:00 04:30 06:00 07:30 09:00 10:30 12:00 13:30 15:00 16:30 18:00 19:30 21:00 22:30

0

Office building users

conditioning controls its load through advanced frequency conversion technology and high intelligence, the load can be reduced by about 28%. The demand response from commercial complex users is mainly provided by the air conditioning load and lighting load. These loads can be directly controlled with an excellent regulation performance. The flexible regulation process minimizes the impact on end-users. The controllable load of large commercial complexes accounts for about 36%. Thereinto, the controllable load of air conditioning load in summer is the largest, accounting for about 19% of the total load demand. The total controllable load of the lighting, cold storage, and electric heating boiler reaches up to 13%.

Office Building Users Office building users generally work in one-shift operations. The peak period of equipment operation is about 8:00–20:00, as shown in Fig. 22. The load mainly comes from the air conditioning system, lighting system, office equipment, dynamical system, etc. The air conditioning system can be the central air conditioning or the VRV air conditioning. This part of the load accounts for about 45% of the total load. As office buildings need high-quality office lighting, the load proportion of the lighting system can be as high as 40%. Air conditioning equipment, office lighting, and electric boilers account for large proportions of the load. According to the load characteristics of office buildings, the regulation of these three types of loads will not cause great potential safety hazards. Hence, these loads have high controllable potentials. The air conditioning load can be directly controlled via adjusting the parameters of air conditioning, which makes the control actions can be generated and applied in one minute. The lighting load can be adjusted by cutting off the load power supply to reduce the load, which can be used for regulation services. The electric boiler participates in the demand response by cutting off the power supply and adjusting the load. In summary, the overall controllable load of office buildings accounts for about 32%.

96

H. Yang et al.

2.4.4

Load Demand Response Modeling of Commercial Users

The controllable load of commercial users mainly includes the central air conditioning, electric boiler, etc. These loads account for a large proportion of the total load demand. The controllable potential models of the central air conditioning and electric boiler are as follows.

Controllable Potential of Central Air Conditioning a.

Heating condition

It is assumed that the central air conditioning in the building has been in stable operation before the implementation of demand response. That is, the heating capacity of the central air conditioning matches the heat consumption of buildings. The basic parameters such as building heating area, average floor heigh, are considered in this section. According to the current indoor temperature and its lower limit, the reducible active power of the central air conditioning and the corresponding duration can be obtained. When the duration is determined, the maximum reducible active power can be PD1 R,decr eased × t D1 R × C O P = c × S × H × ρ × (Tset − Tlimit )

(4)

where PD1 R,decr eased is the active power that can be reduced by the centralized air conditioning system; t D1 R is the demand response duration; C O P is the heating energy efficiency ratio of the centralized air conditioning system; c is the specific heat capacity of air, the default value is 1.005 kJ/(kg*k) at the temperature of 300 K; S is the heating area of the building; H is the average floor height of the building; ρ is the density of air, the default value is 1.177 kg/m3 at the temperature of 300 K; Tset is the heating temperature set by the end-use.; and Tlimit is the lower limit of the indoor temperature, which is lower than the heating temperature set by the end-user. b.

Refrigeration condition

It is assumed that the central air conditioning in the building has been in stable operation before the implementation of demand response. That is, the cooling capacity of the central air conditioning matches the cooling capacity consumption of buildings. The basic parameters such as building heating area, average floor height, are considered in this section. According to the current indoor temperature and its upper limit, the reducible active power of the central air conditioning and the corresponding duration can be obtained. When the duration is determined, the maximum reducible active power can be PD1 R,decr eased × t D1 R × E E R = c × S × H × ρ × (Tlimit − Tset )

(5)

4 Demand-Side Management and Peak Load Reduction

97

where E E R is the refrigeration energy efficiency ratio of the central air conditioning system; and Tlimit is the upper limit of the indoor temperature, which is higher than the refrigeration temperature set by the end-user.

Controllable Potential of Electric Boiler The calculation of demand response capacity of the electric boiler needs to consider whether the electric boiler has a tank. a.

Tankless electric boiler

Firstly, the application of electric boiler in building heating is considered. It is assumed that the electric boiler in the building has been in stable operation before the implementation of demand response. That is, the heating capacity of the electric boiler matches the heat consumption of buildings. The basic parameters such as building heating area, average floor height, are considered in this section. According to the current indoor temperature and its lower limit, the reducible active power and the corresponding duration can be obtained. When the duration is determined, the maximum reducible active power can be PD2 R,decr eased × t D2 R × C O P = c × S × H × ρ × (Tset − Tlimit )

(6)

where C O P is the heating energy efficiency ratio of the electric boiler; Tset is the heating temperature set by the user; Tlimit is the lower limit, which is lower than the heating temperature set by the user. Secondly, the application of electric boiler in hot water supply is considered. At this time, the load is affected by the water supply temperature, water supply flow, operation efficiency, and other factors of the electric boiler. A tankless electric boiler works when the end-user needs hot water and stops immediately when the activity is ended. Hence, the boiler is uncontrollable to avoid affect the end-user comfort. The demand response capability of this part is not considered temporarily. b.

Tank electric boiler

The electric boiler participates in peak shaving by changing the electric boiler from the high-power operation mode to the low-power operation mode. PD2 R,decr eased = Phigh − Plow

(7)

where PD2 R, decr eased is the reducible active power; Phigh is the real-time power of the electric boiler in the high-power operation mode; Plow is the real-time power of the electric boiler in the low-power operation mode. The duration t D2 R of the electric boiler participating in demand response ranges from 30 min to several hours. The specific value depends on the application scenario of the electric boiler and end-user energy consumption behavior.

98

H. Yang et al.

To sum up, the demand response model of commercial users can be described as follows: Q r eal = επ D R

(8)

0 ≤ Q r eal ≤ PD1 R,decr eased t D1 R + PD2 R,decr eased t D2 R

(9)

where Q r eal is the actual demand response of the end-user, and its value shall not exceed the potential; π D R is the compensation price provided by the market operator; and ε is the price elasticity coefficient of the user.

2.4.5

Analysis of Controllable Load Potential for Residential Users

The load curve of residentials is related to human behaviors. Generally, residential users start to use electricity from 7:00 to 8:00. The load reaches its peak at 11:00– 13:00 and 19:00–23:00, and the load in the evening peak is normally higher than that in the noon peak, as shown in Fig. 23. In the morning, due to the low temperature and going out to work, the residential load is low. At noon, due to the rising temperature and electricity consumption for cooking, the residential load goes high. At night, the load rate of residential electrical equipment is high. Therefore, the load peaked in the evening and began to decline in the early morning. The main load equipment of residential users includes the air conditioning, water heater, lighting, etc. And the electric equipment in summer and winter are approximately the same. The air conditioning load accounts for about 40–50% of the peak load of residents; the water heater accounts for about 10–30%; the refrigerator accounts for about 1–5%; the lighting load accounts for about 10–20%. The air conditioning load accounts for a high proportion during the peak power consumption in summer and winter, and the load maintains a high-speed growth trend. The water heater reaches its peak during the evening peak. The refrigerator is small 4000 3500 3000 2500 2000 1500 1000 500 0 00:00 01:15 02:30 03:45 05:00 06:15 07:30 08:45 10:00 11:15 12:30 13:45 15:00 16:15 17:30 18:45 20:00 21:15 22:30 23:45

Load (kW)

Fig. 23 Daily load characteristic curve of urban and rural residents

Urban and rural residents

4 Demand-Side Management and Peak Load Reduction

99

and remains relatively stable throughout the day. According to the comprehensive analysis of end-users’ data, the overall controllable load potential of residential users is about 50% of the total load demand.

2.4.6

Load Demand Response Modeling for Residential Users

The residential controllable load includes the split air conditioning, water heater, etc. The controllable potential of the split air conditioning and water heater can be [14]:

Controllable Potential of Split Air Conditioning a.

Constant frequency air conditioning

PD1 R,decr eased = Px − Py

(10)

where PD1 R,decr eased is the reducible active power of constant frequency air conditioning; Px is the rated power of the constant frequency air conditioning under the working modes of refrigeration, heating, dehumidification, electric auxiliary heating, etc.; and Py is the rated power of constant frequency air conditioner under ventilation, standby, and other working modes. b.

Variable frequency air conditioning

The variable frequency air conditioning reduces the active power according to the formula: PD1 R,decr eased = Px − Py

(11)

where PD1 R,decr eased is the reducible active power of variable frequency air conditioning; Px is the active power before the refrigeration temperature of the variable frequency air conditioning is increased or the heating temperature is reduced; and Py is the active power after the refrigeration temperature of the variable frequency air conditioning is increased or the heating temperature is reduced.

Controllable Potential of Water Heater The water heater has two working modes: fast heating, slow heating, and heat insulating. The peak shaving response can be realized by switching the water heater from the fast or slow heating mode to the heat-insulating mode. The reducible active power and corresponding duration can be obtained

100

H. Yang et al.

PD2 R,decr eased = χ P1 + (χ − 1)P2 − P3

(12)

where P1 is the real-time power of the water heater in the fast heating mode. P2 is the real-time power of the water heater in the slow heating mode. P3 is the real-time power of the water heater in the thermal insulation mode. χ is the operation mode of the water heater. When χ is 1, the water heater is in fast heating mode. When χ is 0, the water heater is in slow heating mode. The default demand response duration t D R of the split air conditioning and water heater in this section is 1 h. The residential demand response capacity can be Q real = ωπ D R

(13)

  0 ≤ Q real ≤ PD1 R,decr eased + PD2 R,decr eased t D R

(14)

where Q real is the actual demand response of end-users, which must not exceed the controllable potential; S is the compensation price provided by the market operator; and ω is the price elasticity coefficient of the users.

3 Design of Market Operation Mode of Load Aggregator Considering Multi-user Demand Response 3.1 Definition of Load Aggregator 3.1.1

Load Aggregator with Multi-user Participation

At present, the mining of controllable capacity in the demand side is not deep enough. The degree of the end-user response is not high enough [20]. To integrate demand response resources, developed countries propose a new professional demand response provider–load aggregator. The market entity enables idle small and medium-sized loads with their controllable capacity to participate in the market [14, 23]. In Europe, America and other developed countries, load aggregators are defined as integrators of demand response resources. The specific function is to evaluate the demand response potential of end-users through professional technology and integrate decentralized demand response resources to participate in system operations [1]. On the one hand, load aggregators provide opportunities for small and mediumsized end-users to participate in the market. On the other hand, load aggregators fully explore demand response resources through professional technical means and provide auxiliary services to the power system.

4 Demand-Side Management and Peak Load Reduction

101

The research on load aggregator business has been relatively mature in developed countries. Many countries and regions have application examples of load aggregators. The load aggregator has developed from the original power purchasing agent to a demand response resource integrator providing a variety of technologies and services [3, 6]. Relatively speaking, China has done a lot of research on the basic theory of load aggregators. However, the operation mechanism of load aggregators in the demand response market is still in the exploratory stage. In March 2015, Several Opinions of the CPC Central Committee and the State Council on Further Deepening the Reform of the Electric Power System (Zhong Fa [2015] No. 9, hereinafter referred to as document No. 9) were issued. Therefrom, China has started a new round of power system reform. No. 9 civilization pointed out the need to actively carry out power demand-side management and implement demand response. On the premise of ensuring the safety of human beings and production, the flexible controllable ability of the system is increased through market means. The role of the power system has changed from the management type to the service type [21]. Therefore, it is of great significance to study the demand response model based on China’s national conditions and the existing dispatching system. In recent years, Shandong, Zhejiang, Hunan, and other provinces have successively announced power demand response work plans. The concept of “load aggregator” is mentioned. In 2021 summer, a demand response pilot project had been implemented in Hunan Province. This pilot project can be taken as an example in this section. The implementation measures specify that the participants in the power demand response market are direct demand end-users and load aggregators. Among them, the load aggregator is defined as a new service enterprise with the qualification for electricity selling. The main responsibility of load aggregators is to provide end-users with professional demand response technology and efficient consulting services. Load aggregators obtain revenue by aggregating demand response resources and participating in demand response and electric energy bidding. If the end-users meet the demand and respond to the market requirements, they can choose the power selling company acting as their power energy bidding agent as the load aggregator. If the power selling company does not register as a load aggregator on the demand response management platform, the end-user can select other registered load aggregators.

3.1.2

Market Function Positioning of Load Aggregators

With the gradual deepening of the new round of power system reform, China’s power supply has gradually changed from the planned distribution to the market transaction mod [31]. The construction of China’s power market has ushered in a period of rapid development. However, the coexistence pattern of the planned distribution and the market transaction will exist for a long time [25]. At present, Guangdong, Zhejiang, Shanxi, Gansu, and other provinces in China distribute part of the base electricity to the generator units participating in the market in the form of government-authorized

102

H. Yang et al.

contracts. The base electricity refers to the part of the electricity generated by the market or non-market unit that is priced by the government. The mode of partial electricity entering the market is easier to connect with the current situation and reduce the risk of market-oriented reform. Especially in the provinces with tight supply–demand relationships and low proportions of planned power liberalization, this model is conducive to ensuring the smooth start of the power market [9]. Under the background of this “plan plus market” dual-track system, end-users are divided into non-market-oriented end-users and market-oriented end-users. Nonmarket-oriented end-users are mainly residential end-users and cannot participate in power market transactions. Such end-users usually purchase the planned electricity by the load aggregator and settle the electricity charge according to the benchmark electricity price. Market-oriented end-users are mainly industrial and commercial users. Such end-users participate in the electricity market bidding transaction and settle the price according to the clearing price. Generally, the power exchange organizes the wholesale market and retail market on time according to the rules of the power market. Large industrial users directly participate in the power purchase in the wholesale market, or through a load aggregator. Commercial users must purchase electricity through load aggregators in the retail market. The wholesale market refers to the power trading activities carried out by generators and load aggregators or large industrial users in a market-oriented manner. The market includes medium-term and long-term monthly electricity market transactions, spot electric energy market transactions, and special transactions for peak demand response. At present, China has started the pilot work of spot market reform in some regions. However, the operation experience is not mature, and there has a distance to the open and mature power market. The market-oriented transaction in most provinces or regions still stays in the medium term and long term [5, 28]. Therefore, the markets that large industrial users and load aggregators can participate in are mainly medium-term and long-term monthly electricity markets and special markets for peak demand response. The definition of medium-term and long-term monthly electricity market can be summarized as follows: Electricity trading conducted by market entities through centralized bidding, bilateral negotiation, and other market-oriented methods with electricity as the trading target. Market entities mainly refer to generators, load aggregators, and end-users that meet the access conditions. After the end of the medium-term and long-term electricity market, the power exchange predicts the power supply and demand balance of the next day. Then determine whether to start the special market for peak demand response. The definition of a special market for peak demand response can be summarized as follows: in case of short supply, load aggregators participate in bidding and provide interruptible or transferable load capacity. On the response day, the load aggregator reduces the power demand or transfers the power demand from the peak period to the standard or valley periods, and obtains the response compensation. The retail market refers to the power transaction between load aggregators and end-users in a market-oriented way. Generally, load aggregators sign agency

4 Demand-Side Management and Peak Load Reduction

103

Fig. 24 Load aggregator operations

agreements with end-users through bilateral negotiations and sell electricity to end-users. As shown in Fig. 24, as a business agent between producers and end-users, the market functions of load aggregators should be summarized as follows: a.

b.

First, act as an agent for non-market-oriented end-users to purchase electricity. Load aggregators sign wholesale power purchase contracts with planned power producers according to the benchmark price. Then, the load aggregator will sign a retail sales contract with non-market-oriented end-users according to the tariff. Second, act as an agent for market-oriented end-users to participate in power purchase and sale transactions. In the wholesale market, load aggregators participate in the medium-term and long-term monthly transactions and sign monthly power transaction contracts with generators. After the trading results are announced, the load aggregators can participate in the special market for peak demand response. The subject matter of the transaction is the responsive load capacity and compensation price of end-users under the jurisdiction of the load aggregator. In the retail market, the load aggregator receives the dispatching instructions from the dispatching center, and the load aggregator receives the dispatching instructions. The load aggregator combines the price type [16] or incentive type [29] demand response management methods to guide end-users to change their power consumption patterns. Finally, the load aggregator obtains the income generated by the price difference between purchase and sale, and the special transaction also provides compensation for the aggregator.

104

3.1.3

H. Yang et al.

Technical Function Analysis of Load Aggregators

Smart grid technologies are important means to regulate the generation/load balance. The application of smart grid technologies in the power market improves the efficiency of end-users and changes their power consumption patterns [2, 17]. To this end, China has organized a series of research and pilot projects on system protection and load control to accumulate practical experiences. Load aggregators can aggregate controllable loads through advanced measurement equipment and communication networks. If the load aggregator is introduced into the load control system, the operation efficiency and accuracy of the system can be effectively improved. Taking the load control system in a region of China as an example, the system consists of a load control master station, load aggregator substation, and load terminal. In practice, the power dispatching organization monitors the power consumption on the demand-side in real time through the load control system. The master station controls 110/35/10/6 kV circuit breakers (load switch) that can interrupt the load on the demand-side quickly and accurately, the specific technical functions are shown in Fig. 25, the detail descriptions are as follows: Assuming the load control system in this area is configured as follows: 1 master station, 3 load aggregator substations, and terminal equipment required by endusers. The master station is set in a 500 kV substation. The station receives the summary information of the interruptible local load and the dispatching instruction of the dispatching center. Finally, the master station decomposes the load interruption and sends it to each load aggregator substation. Three load aggregator substations

Fig. 25 Load control system

4 Demand-Side Management and Peak Load Reduction

105

are, respectively, set in 220 kV substations in blocks A, B, and C. The load aggregator substation divides its terminals into several levels according to the controllable load potential. The substation summarizes the real-time interruptible load by potential levels, and uploads the information to the master station. In addition, the load aggregator substation accepts the master station command and sends the interrupt command to each terminal. Finally, terminals are arranged in major user stations in zones A, B, and C. The terminal collects the information of total interruptible load through electric quantity and voltage transformer. After receiving the load interruption command, the terminal accurately executes the command content in real time. To sum up, the technical functions of the load aggregator can be summarized as follows: First of all, the load aggregator assists end-users to realize metering transformation. The collected information includes end-user load, temperature, start– stop control signal, demand response intention, etc. Secondly, the load aggregator excavates and evaluates the demand response potential contained in the information. Finally, the load aggregator receives the dispatching instructions in real time to help end-users execute the response.

3.2 Demand Response Market Transaction Mode for Load Aggregators The sustainable growth of the load demand increases the peak–valley difference. The development of renewable energy increases the inverse peak regulation characteristics of wind power, photovoltaic, and renewable generations [24]. The situation of power supply shortage during peak periods is more severe. Therefore, the power system urgently needs to encourage end-users to participate in demand response through market-oriented means. This will assist the stability of the system and reduce the peak load [18]. In terms of market functions, load aggregators provide power purchase and sale services. In terms of technical functions, the load aggregators accurately monitor the load data of end-users and receive the system dispatching in real time. Combined with the market and technical functions of load aggregators, this section discusses the optimal decision of load aggregators to participate in the monthly market transaction mode and the special market transaction mode for peak demand response.

3.2.1

Demand Response Optimization Decision of Load Aggregators Participating in the Medium-Term and Long-Term Centralized Bidding

The medium-term and long-term monthly transactions are cleared through high–low matching and bilateral centralized bidding. The clearing principle is “price priority,

106

H. Yang et al.

Fig. 26 Medium-term and long-term monthly transaction process

time priority, and environmental protection priority.” This section assumes that all power selling companies in the region are registered as load aggregators. Market entities include power generators, load aggregators, power trading centers, and power dispatching institutions. The specific transaction organization process is shown in Fig. 26.

Power Generators Declare Electric Quantity and Price. It is assumed that M power producers participate in the medium-term and longterm market transactions. Considering their own power generation cost and bidding strategy, power generators declare electric quantity and price through the power exchange.

Demand-Side Declares Power Demand and Electricity Price. It is assumed that all market-oriented end-users purchase electricity through load aggregators. And there are N load aggregators.

4 Demand-Side Management and Peak Load Reduction

107

Transaction Clearing The power exchange ranks the generators according to the declared price from low to high. The sorting result is Sor t{m 1 , m 2 , · · · , m m }. The electricity quantity and price declared by the generator j are Q sp,bd, j,t and πsp,bd, j,t , respectively. The load aggregators are sorted from high to low. The sorting result is Sor t{n 1 , n 2 , · · · , n n }. The power demand and electricity price declared by the load aggregator i are Q bp,bd,i,t and πbp,bd,i,t , respectively. If the electricity prices are the same, load aggregators shall be sorted according to the principle of the earlier of the final declaration time. If the price and time are the same, multiple declared electricity quantities will be combined temporarily. After the transaction result comes out, the transaction power will be distributed to the generator and load aggregator according to the proportion of declared power. The power exchange takes the top declaration data from the generator queue and the demand-side queue, respectively. These data are matched one by one. According to the declaration data of the generator m 1 and demand side n 1 , the corresponding price difference pair πbd,n1,m1,t is formed as follows: πbd,n1,m1,t = πbp,bd,n1,t − πsp,bd,m1,t

(15)

If πbd,i, j,t < 0, the transaction is unavailable. If πbd,i, j,t ≥ 0, the transaction is available, then the matching price πbar,n1,m1,t and matching power Q bar,n1,m1,t can be calculated: πbp,bd,n1,t + πsp,bd,m1,t 2   = min Q bp,bd,n1,t , Q sp,bd,m1,t

πbar,n1,m1,t = Q bar,n1,m1,t

(16) (17)

  The unmatched remaining power  Q bp,bd,n1,t − Q sp,bd,m1,t  enters the front of the corresponding queue for the next group of matching. Based on the above clearing principles, the declaration information of the buyer and the seller forms a matching set O. The power exchange constructs a mathematical model to maximize social benefits. The social benefits generated by this matching pair are [10]   Wn1,m1,t = πbp,bd,n1,t − πbd,n1,m1,t Q bar,n1,m1,t   + πbd,n1,m1,t − πsp,bd,m1,t Q bar,n1,m1,t   = πbp,bd,n1,t − πsp,bd,m1,t Q bar,n1,m1,t

(18)

Here, the social benefit refers to the benefit generated by the purchase price minus the sale price. According to the principle of maximizing social benefits, the clearing model of the medium-term and long-term transaction can be written as  (ni,m j)∈O

Wni,m j,t =

   πbp,bd,ni,t − πsp,bd,m j,t Q bar,ni,m j,t (ni,m j)∈O

(19)

108

H. Yang et al.

The power exchange calculates the declaration information of the medium-term and long-term monthly transactions. After the unconstrained transaction result is formed, the power exchange submits the result to the dispatching center for security verification.

Security Verification The power dispatching center checks the security of unconstrained transaction results. Form binding transaction results according to constraints: a.

Generation/load balance constraints 

Q bar, j,t =

m j∈M O



Q bar,i,t

(20)

ni∈N O

where M O and N O are the total number of generators and load aggregators in the set O, respectively; Q bar, j,t is the transaction power of the generator j in the set M O; Q bar,i,t is the transaction power of the load aggregator i in the set N O. b.

Market liberalization proportional constraints 

Q bar,i, j,t ≤ ηQ all

(21)

i ∈ NO j ∈ MO where Q all is the total demand electric quantity; η is the market liberalization ratio, in other words, the ratio of market electric quantity to total demand electric quantity. In the process of market transaction, the final matching electric quantity of the buyer and seller shall not exceed the market electric quantity. c.

Generator set output constraint

0 ≤ Q bar, j,t ≤ Q max, j,t

(22)

where Q max, j,t is the upper limit of the generation capacity of the generator j in the set M O.

4 Demand-Side Management and Peak Load Reduction

d.

109

Network security constraints

−P l ≤ Pl (t) ≤ P l , l = 1, 2, · · · , L

(23)

where Pl (t) is the active power of line l at the time t; P l is the upper limit of the active power of line l; L is the total number of lines. e.

Declared electric quantity constraints

πbp,bd,i,min ≤ πbp,bd,i,t ≤ πbp,bd,i,max

(24)

πsp,bd j,min ≤ πsp,bd, j,t ≤ πsp,bd, j,max

(25)

where πbp,bd,i,max and πbp,bd,i,min are the upper and lower limits of the declared price of the load aggregator, respectively; πsp,bd, j,max and πsp,bd j,min are the upper and lower limits of the declared price of the power generator, respectively. Finally, the power dispatching center and the power exchange determine the final clearing result. The power exchange will publish the results to all participants.

3.2.2

Optimization Decision of Load Aggregators Participating in the Peak Demand Response Market

The special market for peak demand response is a trading session for peak load periods. The special market adopts unilateral centralized bidding transactions and a unified clearing algorithm for clearing. The clearing principle is “price difference first, time first, and environmental protection first.” The transaction organization process is shown in Fig. 27.

Special Startup Conditions and Trading Targets Power exchange forecasts the load curve of the next day. Compare the curve with the peak load out of limit value and out of limit period threshold to judge whether to start the trading session. a.

Peak load out of limit value δ

The power exchange sets a δ value, such as 95% of the annual maximum load. The peak period occurs when the load curve p(t) is greater than δ during the existing period. The judgment method is as follows:

110

H. Yang et al.

Fig. 27 Peak demand response special transaction process

p(t1 − 1) ≤ δ, p(t1 + 1) > δ

(26)

p(t2 − 1) > δ, p(t2 + 1) ≤ δ

(27)

The peak period can be determined when the segment (t1 , t2 ) satisfies formulas (26) and (27). b.

Out of limit period threshold ε

In addition to the peak startup period, the startup conditions of the special market also need to consider the constraint of over-limit time. The power exchange sets the ε value to determine whether the peak period meets the minimum limit of special opening time: t 2 − t1 ≥ ε

(28)

When the segment (t1 , t2 ) meets the formula (1.29), the peak period meets the minimum limit of special opening time. The power exchange launches a special market for peak demand response and publishes the subject matter of the special session. The trading targets of the special market include the electricity quantity Q D R of supply gap during the trading period, the upper limit πds,max , and the lower limit πds,min . If the power exchange takes the period (t1 , t2 ) as the special trading period, the electricity quantity of the supply gap can be calculated:

4 Demand-Side Management and Peak Load Reduction

111

t2 QDR =

p(t)dt

(29)

t1

The power exchange announces the trading information of the special trading session at 10:00 a.m. the day before. That is the trading period, the electricity quantity of supply gap, the upper limit, and the lower limit.

Unilateral Bidding on the Demand Side The load aggregator declares the response electric quantity and compensation price through the power exchange platform. It is assumed that a total of K load aggregators participate in the trading of the peak period (t1 , t2 ). The compensation electricity price and response electricity quantity declared by the load aggregator k are πds,bd,k and Q ds,bd,k , respectively. During the special trading period, the maximum load available on the generation side cannot meet the power demand on the demand side. The load aggregator uses the supply function to bid for competition [15]. πds,bd,k = bk Q ds,bd,k

(30)

where bk is the bidding strategy of load aggregator k in the period (t1 , t2 ). The supply function describes the compensation price declared by the load aggregator for the best benefit.

Special Market Clearing The power exchange sorts the load aggregators until the responsive power reaches or exceeds the supply gap. The sorted result is Sor t{k1 , k2 , · · · , kl }. The ranking principle is “declaration price from low to high, responsive electric quantity from high to low, and declaration time from early to late.” Finally, the special market forms an unconstrained trading result. The clearing price is the declared price of the bid winner. In the process of clearing, the power exchange constructs a mathematical model to maximize the end-user profit. The profit on the demand side is affected by two factors: user special market revenue and response cost [29]. Therefore, the objective function can be expressed as max F = E D R − C D R

(31)

where F is the profit of the demand side in the special market; E D R is the income of the end-user participating in the special market; C D R is the response cost of the demand side participating in the special market.

112

H. Yang et al.

Revenue E D R on the demand side includes two aspects: power purchase cost saved by end-users due to reducing power consumption. Electricity compensation is received by end-users in the special market, which can be expressed as EDR =

K 

Q ds,bd,k πlm,bar,k +

k=1

K 

Q ds,bd,k πds,bd,k

(32)

k=1

The response cost C D R on the demand side is the quantification of the degree of dissatisfaction when end-users reduce their power consumption. Generally, the function has the characteristics of monotonicity and concavity concerning the response electric quantity. That is, the dissatisfaction of users will increase sharply with the reduction of demand electric quantity. The response cost of end-users can be modeled by quadratic function [7, 8]: CDR =

K

  2 γ Q ds,bd,k + β · Q ds,bd,k

(33)

k=1

where γ and β are constant coefficients of the response cost function. Based on the above model, the power exchange calculates the unconstrained trading results. Finally, the results are submitted to the dispatching center for security verification.

Security Verification The dispatching center checks the security of unconstrained transaction results to form constrained transaction results. Specific constraints are as follows: a.

Power balance constraint M 

PG,m (t) =

m=1

K 

Pds,i (t), t ∈ (t1 , t2 )

(34)

k=1

where PG,m (t) is the active power of the generator m at the time t; Pds,k (t) is the active power of the load aggregator k at the time t. b.

Compensation price constraint πds,min ≤ πds,bd,k ≤ πds,max

(35)

4 Demand-Side Management and Peak Load Reduction

c.

113

Controllable load constraints

In order to ensure the normal power consumption of the end-user, the response electric quantity shall be within the range of the controllable load potential: 0 < Q ds,bar,k ≤ σk Q bp,bar,k

(36)

where Q bp,bar,k is the trading electric quantity of the load aggregator k in the mediumterm and long-term market; σk is the controllable potential coefficient of the load aggregator k. Based on the controllable potential analysis in Sect. 2.4, the coefficient is determined by the type and number of end-users under its jurisdiction. The greater the value of the coefficient, the greater the controllable potential of the load aggregator, and the value is (0, 1). d.

Network security constraints −P l ≤ Pl (t) ≤ s P l , l = 1, 2, · · · , L

(37)

where Pl (t) is the active power of the line l in the period (t1 , t2 ); P l is the upper limit of the active power of the line l.

4 Case Study In this section, a simulation is carried out concerning the actual power consumption data of users in a region in southern China in the summer. As shown in Fig. 28, the out-of-limit period threshold ε is 60 min, and the peak out-of-limit value δ is 95% of the maximum load. The annual maximum load in this area is 33211.9 MW, so δ is 31551.3 MW. From 20:00 to 23:00, the demand-side load exceeds δ, and the duration meets the threshold ε. Suppose that the peak period is divided into three small periods. And Fig. 28 Identification of summer peak period in a region of southern China

34000

Load (WM)

32000 30000 28000 26000 24000 22000

Daily load curve

Maximum load 95%

114

H. Yang et al.

Table 4 Power supply gap of each trading period (Unit: MWh) Trading period

20:00–21:00

21:00–22:00

22:00–23:00

The electric quantity of supply gap

1013.39

1550.32

802.01

these periods are used as three trading periods to organize bidding. The power supply gap of each period is shown in Table 4. It is assumed that the three load aggregators participate in the medium-term and long-term centralized bidding and special market for peak demand response. Users under the jurisdiction of each load aggregator have the same price elasticity coefficient and response cost coefficient, as shown in Table 5. It is assumed that in the power wholesale market, the clearing price of the load aggregator in the mediumterm and long-term centralized bidding transaction is 329 yuan/MWh. In the three trading periods of the special session, the compensation price of the load aggregator is 1900 yuan/MWh, 2000 yuan/MWh, and 1800 yuan/MWh, respectively. In the retail market, the electricity selling price negotiated bilaterally between the load aggregator and the end-user is 502 yuan/MWh. When a load aggregator participates in a special transaction on behalf of a user, it can obtain a certain agency fee. The agency fee usually does not exceed 20% of the compensation income. This section proposes two proxy schemes: first, the load aggregator takes 15% of the compensation income as the agency fee; second, the load aggregator takes 10% of the compensation income as the agency fee. See Table 6 for details. The analysis shows that the peak demand response reduces the peak load. The greater the price elasticity coefficient of the end-user is, the higher the power that the response capacity. The higher the compensation price obtained by end-users, the greater the enthusiasm of users to participate in the special market. Therefore, in the special transaction, the power exchange can effectively improve the peak shaving ability of the demand side by adopting reasonable incentive means. Finally, this section calculates the profits of load aggregators and end-users in special sessions when load aggregators adopt different agency schemes. As shown in Figs. 29, 30, 31, and 32. Table 5 Value of user demand response coefficient under the jurisdiction of load aggregator

Price elasticity coefficient ω

Response cost coefficient γ

Response cost coefficient β

Users under load aggregator A

0.05

0.1

0.1

Users under load aggregator B

0.07

0.1

0.05

Users under load aggregator C

0.1

0.15

0.1

4 Demand-Side Management and Peak Load Reduction

115

Table 6 Electric quantity of load aggregator response Special market clearing compensation price (yuan/MWh)

1900

Agency fee

15%

10%

15%

10%

15%

10%

Compensated electricity price obtained by users (yuan/MWh)

1615

1710

1700

1800

1530

1620

Electric quantity (MWh) of load aggregator A (ω=0.05) response

80.75

85.5

85

90

76.5

81

Electric quantity (MWh) of load aggregator B (ω=0.07) response

113.05

119.7

119

126

107.1

113.4

Electric quantity (MWh) of load aggregator C (ω=0.1) response

161.5

171

170

180

153

162

Total response electric quantity

355.3

376.2

374

396

336.6

356.4

Fig. 29 Special market revenue of load aggregator at 15% agency fee

2000

1800

25000

Profit (yuan)

20000 15000 10000 5000 0 Period1 Period2 Load aggregator A Load aggregator B

6000 5000 Profit (yuan)

Fig. 30 Special market revenue of load aggregator at 10% agency fee

Period3 Load aggregator C

4000 3000 2000 1000 0 Period1 Load aggregator A

Period2 Load aggregator B

Period3 Load aggregator C

Through the analysis, it can be seen that the load aggregator and end-users obtain certain compensation income by participating in the special market. Under the scheme of 15% agency fee, load aggregators and end-users get higher income. However, the income difference in different trading periods is large, and there are economic risks caused by uncertain factors. Under the scheme of 10% agency fee, the income of load aggregators and users in each trading period is relatively stable,

116

H. Yang et al.

Fig. 31 Revenue of users under the jurisdiction of load aggregator at 15% agency fee

400000

Profit (yuan)

350000 300000 250000 200000 150000 100000 50000 0

Period1 User of load aggregator A User of load aggregator C

Profit (yuan)

Fig. 32 Revenue of users under the jurisdiction of load aggregator at 10% agency fee

450000 400000 350000 300000 250000 200000 150000 100000 50000 0

Period1

User of load aggregator A User of load aggregator C

Period2 Period3 User of load aggregator B

Period2

Period3

User of load aggregator B

but the income is relatively low. Therefore, load aggregators and users can analyze their demand response capacity and risk tolerance. Through strategic participation in special market bidding, load aggregators and users can improve their income. To sum up, the transaction mode and load aggregator oriented to power demand response can be introduced into the power market. On the one hand, the transaction mode can expand the dispatching scope of the power system during peak hours to fill the gap between supply and demand. This will improve energy utilization and save power system construction costs. On the other hand, this model encourages end-users to reasonably plan the power consumption period and improve the power consumption efficiency. At the same time, end-users obtain certain economic benefits through the special market. Therefore, the establishment of a perfect trading mechanism and rule system is conducive to the development of China’s power marketization.

4 Demand-Side Management and Peak Load Reduction

117

5 Conclusion With the steady progress of China’s power market construction, the proportion of market-oriented electricity is gradually increasing. The proportion of all marketoriented end-users purchase on the demand side is increasing, and the types of users are gradually diversified. The demand response behavior of the large industry, commercial, and other users has a greater and greater impact on the power market. Therefore, China urgently needs to take market measures to guide users to change their habitual power consumption mode. This can reduce or transfer the load during peak periods to achieve the purpose of stabilizing the operation of the power system. The multidimensional space–time load distribution characteristics are analyzed in the region. Consider the load proportion and average peak–valley difference level of large industrial users, commercial users, and residential under different time scales of year, quarter, month, and day. The typical load characteristic model of multi-users in the peak period is constructed. Analyze the production characteristics of large industrial users, the operation characteristics of commercial users, and the living habits of residents. Explore the response degree and response potential of end-users as the necessary information for peak shaving. Secondly, under the background of the dual-track system of “plan plus market,” the calculation power analysis takes the Hunan power market in China as an example. This section designs a special trading mode for power demand response in peak periods and defines the market and technical functions of load aggregators. As a power demand side management service organization, load aggregators aggregate users with demand response capacity. This will provide the power market with better load response characteristics than a single user. Acknowledgements This work was supported in part by the National Natural Science Foundation of China (No. 71931003, No. 72061147004, No. 72171026), the Science and Technology Projects of Hunan Province and Changsha City under (No. 2021WK2002, No. 2019GK5015 and No. 2020GK1014), the Hunan International Scientific and Technological Cooperation Base of Energy Storage Interconnection (No. 2018WK4010), the Big Data-driven Energy Internet International Science and technology Cooperation base of the Ministry of Science and Technology (No. 2017D01011), the University–Industry Collaborative Education Program, and the Special funding for the construction of innovative Hunan province (No. 2019RS1045).

References 1. Anderson CL, Cardell JB (2014) A decision framework for optimal pairing of wind and demand response resources. IEEE Syst J 8(4):1104–1111. https://doi.org/10.1109/JSYST.2014.232 6898 2. Asrari A, Ansari M, Khazaei J et al (2020) A market framework for decentralized congestion management in smart distribution grids considering collaboration among electric vehicle aggregators. IEEE Trans Smart Grid 11(2):1147–1158. https://doi.org/10.1109/TSG.2019.293 2695

118

H. Yang et al.

3. Bessa RJ, Matos MA (2013) Optimization models for EV aggregator participation in a manual reserve market. IEEE Trans Power Syst 28(3):3085–3095. https://doi.org/10.1109/TPWRS. 2012.2233222 4. Clement-Nyns K, Haesen E, Driesen J et al (2010) The impact of charging plug-in hybrid electric vehicles on a residential distribution grid. IEEE Trans Power Syst 25(1):371–380. https://doi.org/10.1109/TPWRS.2009.2036481 5. Chatzigiannis DI, Biskas PN, Dourbois GA (2017) European-type electricity market clearing model incorporating PUN orders. IEEE Trans Power Syst 32(1):261–273. https://doi.org/10. 1109/TPWRS.2016.2542823 6. Du C, Wang X, Wang X et al (2016) A block-based medium-long term energy transaction method. IEEE Trans Power Syst 31(5):4155–4156. https://doi.org/10.1109/TPWRS.2015.249 7965 7. Deng R, Yang Z, Chow MY et al (2015) A survey on demand response in smart grids: mathematical models and approaches. IEEE Trans Inform 11(3):570–582. https://doi.org/10.1109/ TII.2015.2414719 8. Fahrioglu M, Alvarado FL (2001) Using utility information to calibrate customer demand management behavior models. IEEE Trans Power Syst 16(2):317–322. https://doi.org/10.1109/ 59.918305 9. Guo Z, Pinson P, Chen S et al (2021) Online optimization for real-time peer-to-peer electricity market mechanisms. IEEE Trans Smart Grid 12(5):4151–4163. https://doi.org/10.1109/TSG. 2021.3075707 10. Guo H, Chen Q, Xia Q et al (2019) Market power mitigation clearing mechanism based on constrained bidding capacities. IEEE Trans Power Syst 34(6):4817–4827. https://doi.org/10. 1109/TPWRS.2019.2913334 11. Kondoh J, Lu N, Hammerstrom DJ (2011) An evaluation of the water heater load potential for providing regulation service. IEEE Trans Power Syst 26(3):1309–1316. https://doi.org/10. 1109/TPWRS.2010.2090909 12. Logenthiran T, Srinivasan D, Shun TZ et al (2012) Demand side management in smart grid using heuristic optimization. IEEE Trans Smart Grid 3(3):1244–1252. https://doi.org/10.1109/ TSG.2012.2195686 13. Lu T, Wang Z, Wang J et al (2019) A data-driven stackelberg market strategy for demand response-enabled distribution systems. IEEE Trans Smart Grid 20(3):2345–2357. https://doi. org/10.1109/TSG.2018.2795007 14. Lu X, Chan KW, Xia S et al (2021) An operation model for distribution companies using the flexibility of electric vehicle aggregators. IEEE Trans Smart Grid 12(2):1507–1518. https:// doi.org/10.1109/TSG.2020.3037053 15. Li N, Chen L, Dahleh MA (2015) Demand response using linear supply function bidding. IEEE Trans Smart Grid 6(4):1827–1838. https://doi.org/10.1109/TSG.2015.2410131 16. Muratori M, Rizzoni G (2016) Residential demand response: dynamic energy management and time-varying electricity pricing. IEEE Trans Power Syst 31(2):1108–1117. https://doi.org/ 10.1109/TPWRS.2015.2414880 17. Muthirayan D, Kalathil D, Poolla K et al (2020) Mechanism design for demand response programs. IEEE Trans Smart Grid 11(1):61–73. https://doi.org/10.1109/TSG.2019.2917396 18. Ma O, Alkadi N, Cappera P et al (2013) Demand response for ancillary services. IEEE Trans Smart Grid 4(4):1988–1995. https://doi.org/10.1109/TSG.2013.2258049 19. Nekouei E, Alpcan T, Chattopadhyay D et al (2015) Game-theoretic frameworks for demand response in electricity markets. IEEE Trans Smart Grid 6(2):748–758. https://doi.org/10.1109/ TSG.2014.2367494 20. Nunna KHSVS, Doolla S (2014) Responsive end-user-based demand side management in multimicrogrid environment. IEEE Trans Inform 10(2):1262–1272. https://doi.org/10.1109/ TII.2014.2307761 21. Nosair H, Bouffard F (2016) Energy-centric flexibility management in power systems. IEEE Trans Power Syst 31(6):5071–5081. https://doi.org/10.1109/TPWRS.2015.2512990

4 Demand-Side Management and Peak Load Reduction

119

22. Roos JG, Lane IE (1998) Industrial power demand response analysis for one-part real-time pricing. IEEE Trans Power Syst 13(1):159–164. https://doi.org/10.1109/59.651628 23. Rashidizadeh-Kermani H, Vahedipour- Dahraie M, Shafie-Khah M et al (2020) A regret-based stochastic bi-level framework for scheduling of DR aggregator under uncertainties. IEEE Trans Smart Grid 11(4):3171–3184. https://doi.org/10.1109/TSG.2020.2968963 24. Rahimi F, Ipakchi A (2010) Demand response as a market resource under the smart grid paradigm. IEEE Trans Smart Grid 1(1):82–88. https://doi.org/10.1109/TSG.2010.2045906 25. Tafreshi SMM, Lahiji AS (2015) Long-term market equilibrium in smart grid paradigm with introducing demand response provider in competition. IEEE Trans Smart Grid 6(6):2794–2806. https://doi.org/10.1109/TSG.2015.2413951 26. Tong X, Kang C, Xia Q et al (2016) Smart metering load data compression based on load feature identification. IEEE Trans Smart Grid. 7(5):2414–2422. https://doi.org/10.1109/TSG. 2016.2544883 27. Wang K, Ouyang Z, Krishnan R et al (2015) A game theory-based energy management system using price elasticity for smart grids. IEEE Trans Inform 11(6):1607–1616. https://doi.org/10. 1109/TII.2015.2426015 28. Wang J, Zhou H (2013) Conceptual design and the future development for operation smart system in China southern power grid. IEEE Trans Smart Grid 4(3):1621–1629. https://doi.org/ 10.1109/TSG.2013.2252634 29. Wang Y, Chen Q, Kang C et al (2015) Load profiling and its application to demand response: a review. Tsinghua Sci Technol 20(2):117–129. https://doi.org/10.1109/TST.2015.7085625 30. Yu M, Hong SH (2016) A real-time demand-response algorithm for smart grids: a stackelberg game approach. IEEE Trans Smart Grid 7(2):879–888. https://doi.org/10.1109/TSG.2015.241 3813 31. Zhang Z, Jiang Y, Lin Z et al (2020) Optimal alliance strategies among retailers under energy deviation settlement mechanism in China’s forward electricity market. IEEE Trans Power Syst 35(3):2059–2071. https://doi.org/10.1109/TPWRS.2019.2955479

Chapter 5

Demand Response in Smart Buildings B. Rajanarayan Prusty, Arun S. L., and Pasquale De Falco

Abstract The demand-side management scheme, one of the smart grid’s distribution side features, is well accepted due to its role in controlling energy consumption in residential buildings by ensuring sustainability, leading to the concept of smart buildings. The consumers’ appropriate response without compromising the comfort, referred to as demand response, helps decrease their electricity consumption bill. Further, the flexibility and controllability in the power consumption patterns of the end user due to demand response has gained tremendous research interest in the smart grid research domain. Smart appliances’ demand pattern alternation, battery back-up during peak load period, and partial load demand fulfilment through renewable generations are instrumental in reducing the electricity bill. Besides, short-term forecasting plays a vital role in power system scheduling and management. This chapter attempts to summarize the concept of demand response strategies in smart buildings. Further, it discusses the importance of adapting point and probabilistic forecasting methods to yield efficient demand response strategies in a smart building environment.

Abbreviations ANN CF DL DR DSO

Artificial Neural Network Compact Fluorescent Lamp Deferrable Load Demand Response Distribution System Operator

B. R. Prusty (B) · Arun S. L. School of Electrical Engineering, Vellore Institute of Technology, Vellore, India e-mail: [email protected] Arun S. L. e-mail: [email protected] P. De Falco University of Naples Parthenope, Naples, Italy e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 A. Tomar et al. (eds.), Control of Smart Buildings, Studies in Infrastructure and Control, https://doi.org/10.1007/978-981-19-0375-5_5

121

122

DSM HEMS IDL ISWP LED NDL NIDL PCA PDB QR RER SVM

B. R. Prusty et al.

Demand-Side Management Home Energy Management System Interruptible Deferrable Load Improved Sliding Window Prediction Light Emitting Diode Non-Deferrable Load Non-Interruptible Deferrable Load Principal Component Analysis Portrait Dataset Based Quantile Regression Renewable Energy Resources Support Vector Machine

1 Introduction Electric power generation and its supply to end users through transmission and distribution corridors are significant tasks of the traditional energy sector. A limited number of power plants (thermal, hydro, and nuclear) are located at various remote places to supply the required power to maintain the generation-demand balance. The generated voltage levels are altered through transformers to minimize transmission losses and enhance the reliability of the power supply system. Finally, the end consumer is supplied with the power at the required voltage levels. The traditional energy sector follows the unidirectional power flow. Due to rapid growth in energy demand and a decline in fossil fuel resources, the conventional energy sectors are experiencing various operational difficulties. Besides, increasing the number of traditional power plants and enhancing the strength of the transmission systems to reach a high generation-to-demand ratio are challenging for power system engineers due to ecological constraints, rapid progress in population, and economic and political reasons. Therefore, the traditional grid is upgraded as a smart grid to overcome the above challenges. Smart grid is a reliable, clean, secure, resilient, sustainable, and efficient power system [1, 2]. Considering various technologies introduced in the smart grid paradigm, DSM is a reliable approach adopted by many utilities to control the enduser electricity consumption. The vital aspect of DSM is the cost-effective implementation compared to establishing new generation units or massive energy storage systems [3]. DSM technique includes (i) energy-efficient programs, (ii) energy conservation programs, (iii) demand management programs, and (iv) DR programs [4]. The modifications done on the equipment are categorized under the energy-efficient program. For example, replacing old incandescent lamps with LEDs or CFLs and improving the physical properties of the systems [5]. An energy conservation program is also considered part of energy-efficient programs, emphasizing end users’ behavioral changes to make more efficient energy consumption [5]. The demand management program mainly includes direct control and energy pricing techniques

5 Demand Response in Smart Buildings

123

Fig. 1 Two-way cyber-secure communication framework

to reduce peak demand. In the former, the service providers directly regulate the operating pattern of certain end users’ appliances, whereas in the later the electricity bill of the end users is calculated based on their energy consumption pattern. The utilities currently follow various pricing strategies while seeing the demand variation in end-user premises and generation costs. Simple pricing, flat-rate pricing, block rate pricing, demand-based pricing, day-ahead pricing, critical peak pricing, time of use pricing, and real-time pricing [6] are a few pricing strategies followed by utilities. Besides, few utilities introduce a time-dependent demand limit for end users’ demand to enhance the peak-to-average ratio of the distribution sector. In this scheme, the end users are penalized when the user’s net demand exceeds the predefined utility limit. Hence, the end users are intended to alter their demand pattern to lessen electricity bill. Finally, DR programs refer to the activities performed by the end users in response to utility DSM strategies without compromising the user’s comfort. DSOs effectively manage the DSM strategies in a way beneficial to utility and end users. In the smart grid paradigm, the active end users are directly monitored by secondary DSOs. The end users have two-way cyber-secure communication with secondary DSOs and transfer the real-time energy consumption data through smart meters. Finally, the secondary DSOs share information, e.g., electricity price, DSM incentives, and other utility parameters, with the end users as instructed by the primary DSO. The entire framework is portrayed in Fig. 1. Nowadays, residential buildings are getting smart by actively participating in utility DSM schemes. The intelligent HEMS developed in these smart buildings significantly reduces the user’s electricity bill by altering the power consumption pattern of household appliances. Presently, many utilities are imposing a real-time pricing strategy to improve their operational and economic benefits. Further, the util-

124

B. R. Prusty et al.

Fig. 2 Residential DR architecture

ities fix high electricity costs during peak intervals to meet the additional operation cost for peak power plants. In addition to this, the utilities are also charging more penalties for exceeding the demand limit. Hence, the subscribers are forced to schedule most of their demand to non-peak or mid-peak intervals by compromising their comfort. However, few users install energy storage devices to fulfill their essential demand with a minimum electricity bill during peak intervals. Among various available energy storage techniques, battery storage systems are highly attracted by the residential user because of their admiring features. A typical smart residential building equipped with different household appliances, battery units, and RER is shown in Fig. 2. Various communications among units are represented with solid and dotted arrow indicated lines. Advancement in the renewable energy system and affordable investment cost motivates the end user to install small-scale renewable generation setups such as

5 Demand Response in Smart Buildings

125

rooftop solar PV panels and low-rated wind turbines. The end users effectively utilize the generated renewable energy, resulting in a considerable reduction in electricity bills payable to the utility. Further, the smart buildings use the battery banks to store the excess energy generated by the RER during non-peak and utilize during peak intervals, further reducing electricity bills. Besides, few utilities allow the subscribers to share their excess generation at utility buying cost. Those subscribers are generally termed prosumers. The prosumers utilize the maximum amount of in-house generated renewable energy and improve their profits by actively participating in the energy trading with utility. Consequentially, the prosumers-based utilities are experiencing many operational challenges due to the significant penetration of low-rated in-house RERs. The utilities are forced to implement the time-varying limit for power injection to the prosumers to ensure reliable operation. The prosumers can improve their trading profit without exceeding this injection limit. The excess generation beyond this limit is stored in a battery bank for upcoming intervals or dissipated in the dump load [7].

2 Demand Response in Smart Buildings The end users are improving their savings in electricity bills with the help of the DR technique. The DR program is mainly classified into forced DR and volunteer DR. In the former, the end users respond immediately without any incentives to the utility DSM signals, whereas in the latter end users may react to utility DSM signals to decrease the electricity bill [8–10]. As part of this, utilities are also introducing attractive DSM schemes with significant incentives to increase end-user participation in energy society. The end user adopts suitable DR techniques based on affordable investment cost and smartness in the existing household appliances. Based on the end users’ contribution, these techniques are classified into three categories. In the first category, the end user is expected to respond manually due to the utility DSM signals, e.g., reducing the demand during peak intervals by postponing appliances manually to non-peak intervals. In this technique, the end users have economic benefits with zero investment. However, the end user must have adequate knowledge of utility DSM schemes. In the second category, the end users receive alert messages and suggestions to reduce the electricity bill. The recent advancement in smart meter infrastructure provides a viable solution to control energy consumptions. Based on the smart meter instruction, the end users are expected to respond manually. In the third category, the buildings are expected to be equipped with a dedicated EMS to automatically control household appliances’ operation. This system is usually synchronized with the smart meter to receive real-time updates from the utility. Based on the requirements and comfort of the user, the system optimally schedules the operation of smart appliances. With this advanced technique, the end user has the maximum economic benefit.

126

B. R. Prusty et al.

2.1 Smart Building Components Recently, residential buildings have been equipped with various smart household appliances to make life easier. These appliances are divided into NDLs (or noncontrollable loads) and DLs (or controllable loads or flexible loads) based on their functioning. NDLs are critical loads that perform the task instantaneously whenever energized. NDLs are classified into essential and thermal NDLs. The appliances like lights, fans, computers, mobile phones, laptop chargers, televisions, speakers, all-electric kitchen appliances, and security systems are categorized under essential NDLs. The operation of these appliances is solely based on the user’s interest; time scheduling of these demands shall cause inconvenience to the user. On the other hand, temperature-regulating loads like refrigerators, electric water heaters, air conditioners, and space heaters are categorized under thermal NDLs. The dynamics in the power consumption pattern of these loads are highly influenced by user behavior and climatic conditions. During operation, thermal NDLs maintain the temperature at the predefined value set by the user but within the tolerance limit set by the manufacturer. The thermal NDL is operated in standby mode when the operating temperature is maintained within the tolerance limit. However, it moves to run mode when the temperature exceeds the tolerance limit. Time of operation of DLs can be flexibly scheduled in any given period. Considering the nature of the operation, the DLs are further categorized into NIDLs and IDLs. The NIDLs are operated continuously to complete the task. The appliances like cloth washing machines, dryers, and food grinders are classified under this category. The IDLs, on the other hand, may operate continuously or discontinuously in a given period to finish the task. The appliances like an electric vehicle, well pump, and dishwasher are classified under this category. The HEMS is expected to consider the following constraints while optimally time scheduling the operation of DLs: (i) The DL should be scheduled only within the user-predefined time span. (ii) The number of intervals in which the DL is scheduled should be equal to the number of intervals required by the DL to complete the task. (iii) The operation of NIDLs should be continuous. Nowadays, the end users prefer battery banks to have an uninterruptible power supply. The battery banks also minimize the electricity bill by meeting part of the demand during peak intervals. Further, charging the battery banks during non-peak intervals (less energy price intervals) reduces the users’ electricity bill. Hence, controlling the battery operation is essential to have more savings in electricity bills. The operating mode (floating mode/charging mode/discharging mode) and power exchanging of battery are taken to be the controllable factors of battery. In general, the battery banks are presumed to be additional DLs during the charging mode of operation. In contrast, they are taken to be an added resource during the discharging mode of operation. The HEMS is expected to consider the following constraints while optimally time scheduling the operation of the battery:

5 Demand Response in Smart Buildings

127

(i) For any scheduling interval, the scheduled mode of operation of the battery should be unique. (ii) The battery operational parameters such as state of charge, charging, and discharging powers must be maintained within the boundary limits predefined by the manufacturer to extend the battery life. Recently, the residential user’s increased interest in renewable-based generation leads to reduced utility grid dependency. However, these generations are highly intermittent and depend on the installation site. The unpredictability in renewable generations causes additional operational challenges to the utility. The smart building is mainly equipped with rooftop solar PV-based power generation and the surrounding areas could be equipped with lower rating wind power generations. While, on the one hand, PV generation is influenced by solar irradiation and atmospheric temperature wind power generation, on the other hand depends upon wind speed.

3 Management of Prosumers’ Generation and Demand In recent times, short-term forecasting has played a significant role in managing and scheduling smart grid power systems [11, 12]. The development of models for forecasting load powers and renewable generations to help propose efficient DR strategies for aggregate prosumers’ load have recently gained research attention. In a smart building environment, reliable forecasts for load power and renewable generations are essential. Although load prediction of the entire grid can be attained with relatively high accuracy, the dynamics change radically for buildings. The volatility effect is more significant in the latter case compared to the former. However, compared to renewable generation time series, the building load consumption time series is relatively stable. The presence of high volatility effects in the renewable generations [13] requires suitable adjustments to original building load forecasting algorithms. On this note, several classical time-series models and machine learning forecasting models (such as multiple linear regression, random forests, quantile regression, ANN, SVM, etc.) have been applied in the literature. A detailed discussion of various short-term forecasting methods suitable for the DR environment can be found in [14].

3.1 Challenges in Smart Building Prosumers’ Aggregate Load Forecasting The accuracy of the forecast models’ output strictly depends on the accuracy of the recorded time-series dataset [15–18]. Unfortunately, raw data is not used directly for forecasting model development due to missing samples and outliers [19]. Data preprocessing (or data cleaning or data cleansing) refers to adopting a set of steps for

128

B. R. Prusty et al.

obtaining a suitable dataset for forecasting model development. Among the various, two essential and typical steps in data preprocessing are (i) resolving inconsistencies by imputing missing values and (ii) detection of outliers and their apt correction [15–18]. Therefore, the two significant steps include data preprocessing and a suitable forecasting model for forecasting. Instruments used in the field to acquire data often fail due to various reasons resulting in gaps or missing values in time series. Therefore, resolving inconsistencies becomes an important step. A straightforward strategy frequently adopted in the literature is to replace the missing samples with the recent previous values to retain the time-series property of the dataset. On the other hand, samples with unexpected spikes or dips in a time series that are unusual and inconsistent with the rest of the dataset are referred to as outliers [15]. The presence of outliers leads to biased parameter estimation of forecasting models and therefore biased forecast output. Thus outliers in time-series data should be detected in the first step and corrected for improved forecasting. Outlier detection is typically interested in local subsections of a time series instead of in global properties. It involves determining the location, type, and magnitude of outliers present in a time series. Outlier correction or outlier replacement is an equally important task like detection. There have been numerous methods proposed in the literature for outlier detection and correction [15–18]. Statistical methods have gained overwhelming interest in power system applications; mentioning a few includes B-spline smoothing-based method [17], PDB method [16], ISWP method [18], etc.

3.2 Integration of DR and Short-term Forecasting This section elaborates the adaption of point and probabilistic forecasting methods to yield efficient DR strategies in smart building environment. The role of forecasting in DR was well recognized in early 2010 [20]. Various load forecasting challenges are detailed. The use of stochastic optimization along with short-term load forecasting resulted in realistic solutions to manage dispatchable DR and non-dispatchable DR [20]. A multiple load power short-term forecasting framework using ANN and SVM is presented in [21]. Further, it experiments that the combination of anthropologic and structural data helps improve forecasting accuracy. Ref. [22] proposed a robust DR model to adjust the hourly load power level to the hourly electric price. ANN with confidence interval is used to forecast the unknown electricity price. The authors in [23] used feedforward ANN to yield accurate forecast results while handing large, nonlinear, multivariate time-series data. A real-time weather forecast integrated optimal hybrid renewable energy system model is suggested in [24]. Load and electric price forecasts, essential for designing optimal DR strategies, are coupled via an optimization framework to assist automatic DR [25]. Such a framework significantly lessens the human intervention in DR. A QR-based probabilistic model

5 Demand Response in Smart Buildings

129

to forecast aggregate smart households’ DR capacity in the day-ahead market is proposed in [19]. The attractive feature of the above model is that it embodies the uncertainty associated with aggregated DR capacity realistically. The least-squares SVM-based forecasting model is established by extracting the load characteristics from the massive load power time series [26]. The adoption of the data-driven concept has resulted in the reduction of forecasting error. The problem of insufficient training data is well recognized in the incentive-based DR forecasting problems. The authors in [27] applied transfer learning technique to overcome the above problem of target customer. Data of electricity consumers, including industrial clients, residential, and commercial customers with power per customer ranging from 3 kW to 8 kW, and PV generation are considered for assessing short-term demand and intricate PV generation variability [14]. Further with a detailed case study summarized the importance of executing DR actions to reduce net consumption. Besides, the stable accuracy of an SVM-based forecasting model using PCA for available aggregate DR is verified in [28].

4 Conclusion Energy-related issues are the major concerns for society’s sustainability. In recent years, sustainability has been more customer centric. Notably, the aspects such as prosumers’ net demand prediction, energy management set barriers to new markets’ effective deployment. Firstly, this chapter highlights the importance of DR in the present-day smart grid. Particular emphasis is given to the significance of DR strategies in the smart building by comprehensively elaborating various components. Then, the importance of short-term forecasting in the management of prosumers’ net demand is discussed. It is worth noting that forecasting model development becomes more challenging with the assets’ aggregation level decreasing. The adaption of the point and probabilistic forecasting methods for renewable generation and load power for executing effective DR strategies is summarized. The important proven methods/models for DR are also highlighted to instill in novice readers the advancements in the related literature.

References 1. Fang X, Misra S, Xue G, Yang D (2011) Smart grid–the new and improved power grid: A survey. IEEE communications surveys & tutorials 14(4):944–980 2. Sheen J (2005) Economic profitability analysis of demand side management program. Energy Convers Manage 46(18–19):2919–2935 3. Palensky P, Dietrich D (2011) Demand side management: demand response, intelligent energy systems, and smart loads. IEEE Trans Industr Inf 7(3):381–388 4. Ramanathan B, Vittal V (2008) A framework for evaluation of advanced direct load control with minimum disruption. IEEE Trans Power Syst 23(4):1681–1688

130

B. R. Prusty et al.

5. Pedrasa MAA, Spooner TD, MacGill IF (2009) Scheduling of demand side resources using binary particle swarm optimization. IEEE Trans Power Syst 24(3):1173–1181 6. Centolella P (2010) The integration of price responsive demand into regional transmission organization (rto) wholesale power markets and system operations. Energy 35(4):1568–1574 7. Arun SL, Selvan MP (2017) Dynamic demand response in smart buildings using an intelligent residential load management system. IET Gener Transm Distrib 11(17):4348–4357 8. Arun S, Selvan M (2017) Intelligent residential energy management system for dynamic demand response in smart buildings. IEEE Syst J 12(2):1329–1340 9. Arun S, Selvan M (2019) Smart residential energy management system for demand response in buildings with energy storage devices. Frontiers Energy 13(4):715–730 10. Rao BH, Arun SL, Selvan MP (2020) Framework of locality electricity trading system for profitable peer-to-peer power transaction in locality electricity market. IET Smart Grid 3(3):318– 330 11. Bracale A, Carpinelli G, De Falco P (2016) A bayesian-based approach for the short-term forecasting of electrical loads in smart grids.: Part i: theoretical aspects. In: 2016 International Symposium on Power Electronics, Electrical Drives, Automation and Motion (SPEEDAM), IEEE, pp 121–128 12. Tripathy DS, Prusty BR, Jena D (2020) Short-term pv generation forecasting using quantile regression averaging. In: 2020 IEEE International Conference on Power Systems Technology (POWERCON), IEEE, pp 1–6 13. Prusty BR, Jena D (2018) Preprocessing of multi-time instant pv generation data. IEEE Trans Power Syst 33(3):3189–3191 14. Ruiz-Abellón MC, Fernández-Jiménez LA, Guillamón A, Falces A, García-Garre A, Gabaldón A (2020) Integration of demand response and short-term forecasting for the management of prosumers’ demand and generation. Energies 13(1):11 15. Ranjan KG, Prusty BR, Jena D (2021) Review of preprocessing methods for univariate volatile time-series in power system applications. Electr Power Syst Res 191:106885 16. Tang G, Wu K, Lei J, Bi Z, Tang J (2014) From landscape to portrait: a new approach for outlier detection in load curve data. IEEE Trans Smart Grid 5(4):1764–1773 17. Chen J, Li W, Lau A, Cao J, Wang K (2010) Automated load curve data cleansing in power systems. IEEE Trans Smart Grid 1(2):213–221 18. Ranjan KG, Tripathy DS, Prusty BR, Jena D (2021) An improved sliding window predictionbased outlier detection and correction for volatile time-series. Int J Numer Model Electron Networks Devices Fields 34(1):e2816 19. Xiang B, Li K, Ge X, Zhen Z, Lu X, Wang F (2019) Day-ahead probabilistic forecasting of smart households’ demand response capacity under incentive-based demand response program. In: 2019 IEEE Sustainable Power and Energy Conference (iSPEC), IEEE, pp 1078–1083 20. Luh PB, Michel LD, Friedland P, Guan C, Wang Y (2010) Load forecasting and demand response. In: IEEE PES General Meeting, IEEE, pp 1–3 21. Huang L, Agustin M (2012) Demand response forecasting in practice: challenges and opportunities. In: 2012 IEEE Power and Energy Society General Meeting, IEEE, pp 1–3 22. Jiang H, Tan Z (2012) Load forecasting in demand response. In: 2012 Asia-Pacific Power and Energy Engineering Conference, IEEE, pp 1–4 23. Schachter J, Mancarella P (2014) A short-term load forecasting model for demand response applications. In: 11th International Conference on the European Energy Market (EEM14), IEEE, pp 1–5 24. Wang X, Palazoglu A, El-Farra NH (2014) Operation of residential hybrid renewable energy systems: integrating forecasting, optimization and demand response. In: 2014 American Control Conference, IEEE, pp 5043–5048 25. Alamaniotis M, Tsoukalas LH (2016) Implementing smart energy systems: integrating load and price forecasting for single parameter based demand response. In: 2016 IEEE PES Innovative Smart Grid Technologies Conference Europe (ISGT-Europe), IEEE, pp 1–6 26. Luo F, Yang X, Wei W, Lu H, Zhang T, Shao J (2020) Data driven load forecasting method considering demand response. In: 2020 IEEE Power & Energy Society General Meeting (PESGM), IEEE, pp 1–5

5 Demand Response in Smart Buildings

131

27. Cai L, Wen H, Gu J, Ma J, Jin Z (2020) Forecasting customers’ response to incentives during peak periods: A transfer learning approach. Int Trans Electr Energy Syst 30(7):e12251 28. Wang F, Xiang B, Li K, Ge X, Lu H, Lai J, Dehghanian P (2020) Smart households’ aggregated capacity forecasting for load aggregators under incentive-based demand response programs. IEEE Trans Ind Appl 56(2):1086–1097

Chapter 6

Building Services with the Local Energy Community—Applications Y. P. Chawla

Abstract The concept of local energy community (LEC) has been well received and caught on globally in various segments of the population from Rural setting–Remote villages, to Blue Economies—getting integrated with Green Power and to Urban housing societies as well, safeguarding from the present non-resilient energy infrastructure as of now. Demand-side management, reduction in energy consumption by intelligent governance of intelligent buildings has also caught society’s attention. Surplus Power generated in the buildings at different times led to the development of grid virtual power plants. Community energy storage battery, topping on the Concept of LEC grid-connected or even off-grid with smarter buildings occupied by the smarter occupants, thus creating a ‘Human in the Loop system’- (Comptes Rendus Physique Volume 18, Issues 7–8, September–October 2017, Pages 428– 444) rather than all auto-controlled. With higher per capita energy consumption in countries, the importance of intelligent buildings establishes still higher priority considering heating loads and other luxury requirements. Further, the buildings are now considered energy-positive to generate power from solar rooftops for building’s consumption and sharing with the community. Energy flow is two-way between the community and the buildings through micro-grids or nano-grids, thus offering flexibility and energy demand response while acting as a significant reserve. The intelligent behaviour of the structures on getting integrated with the intelligent grid helps make the LEC smarter. The wise governance of the buildings by optimising the power generation management, its storage for meeting the self-load demand that anticipates and matches the load for optimisation and then shares the energy under Peer-to-Peer (P2P) trading in the local community. Keywords Smart buildings · Smart grids · Smart energy consumption · Energy-positive buildings · Community approach · Peer-to-peer energy trade

Y. P. Chawla (B) JERC/MoP Government of India, New Delhi, India © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 A. Tomar et al. (eds.), Control of Smart Buildings, Studies in Infrastructure and Control, https://doi.org/10.1007/978-981-19-0375-5_6

133

134

Y. P. Chawla

1 Introduction The energy sector has never been under so much pressure as it is post-corona.1 A blue and clean sky, fresh air and refreshing the commitment made by 197 countries—nearly the whole world on the Paris 2015 accord on climate control. War-torn Syria is the last country endorsing the agreement, which has now become an International Law. One hundred ninety countries have further solidified their support as of February 2021.2 Iran, Iraq and Turkey are yet to be a party to the Agreement formally. The Agreement on Climate control got established by May 2021 Dutch Court ruling against Royal Dutch Shell, ordering Royal Dutch to reduce emissions by 45% from 2019 levels by the year 2030. Each country has made Nationally Determined Commitment (NDC) towards reduced emission targets. For reduced emissions, each sector and sub-sector using energy has been made answerable to contribute towards consuming the energy intelligently. The fury of climate change reflected by rains, floods, landslides, tsunamis, heat waves, extremely low temperatures, melting of the polar ice has scared humanity of the further consequences of climate change unless immediate corrections are applied. Pandemic has delayed the November 2020 meeting for furtherance on the subject to November 2021. Earlier during the liberalisation of the Electrical Sector in the 1990s, environmental concerns saw local energy communities’ rise. However, where the community leaders were less aware, the idea is delayed and is catching up these days. The community’s role (human-in-the-loop) and the application of energy services to the building also takes a vital role as any other sector/sub-sector as an energy consumer. As a result, the society members as energy consumers got a changed status of Prosumers (Producers and the Consumers) are now Prosumanagers (Prosumers and Managers). In the USA, concerns raised focus on the Policymakers ensuring that the building practitioners’ practices should care for increase climate risks and use better land-use principles. 3 As a result, the concept of building services and a twin to community energy has caught on. The footprint of the LEC approach is getting extended to the regional process. Other than two facets indicated the other facet of community power generation as distributed generation and contribution to the community. So far, the building sector has been a consumer of the power utilities; however, got graduated to a power generation asset through the aggregation of power from each energy-positive building at any time of the day and supporting as Energy-Efficient Resource Grid (EERG) buildings. Credit goes to the innovative buildings developers leaders and becoming a part of the climate control solution [13, 14], thus enhancing the footprint of the standalone concept of the local energy community. The subject of LEC is generating much interest amongst researchers, as depicted hereunder. Figure 1 displays the Research Data on Energy Communities and Distributed Energy 1

Comptes Rendus Physique Volume 18, Issues 7–8, September–October 2017, Pages 428–444. https://www.nrdc.org/stories/paris-climate-agreement-everything-you-need-know. 3 https://www.cfr.org/in-brief/what-coronavirus-pandemic-teaches-us-about-fighting-climatechange. 2

6 Building Services with the Local Energy Community—Applications

135

Research data on Energy Communi es and Distributed Energy Resources 600

Distribution Genera o Projects

400

23 364

200

220 142

0

246 15

9 2010

11

26500

26

34 606

34 28 29

519 27

24

435

302

266

19 14

10

9 2011

2012

2013

2014

2015

2016

2017

2018

2019

Fig. 1 Research and review papers on local energy community. Data Source “Sustainability 2019, 11, 3493: https://doi.org/10.3390/su1123493”

Sources. However, the COVID has put some brakes on the subject as humanity got busy with other anxieties.

2 The Chapter’s Scope The chapter explores the contribution that local energy communities can make in reducing the peak loads on the primary electrical grid and creating a possibility of share by the Energy positive buildings amongst the local energy community4 . The local energy communities have already established respective contributions globally. However, the local energy communities set-up was under compulsions because there was no connection to the primary power grid, or it was economical to ensure reliable power. The issue involves many facets—energy Efficiency in the buildings and the local community approach to acting as energy community being most important.

3 Concerns About the Subject The lack of awareness, virtual net metering, group net metering concept, electricity regulatory hurdles for the local electricity communities, Social obligations versus Return on Investment, lack of LEC set-up, peer-to-peer (P2P) trade of power are some of the concerns that need attention.

4

http://doi.org/10.3390/su13094944; https://www.mdpi.com/journal/sustainability.

136

Y. P. Chawla

4 Materials and Method Review of literature on how the LEC can lead to a sustainable pathway and energy transition covering the content review followed by an argumentative review of stakeholder strategies and solutions in a specific community or geographic context.

5 Local Energy Communities/Local Building Communities—Showcasing Experience The local energy communities has received much focus in Europe. The USA delayed adaptation is due to the top-down approach of the LEC concept. The success of the local energy communities did not get equal importance in many other parts of the globe. The distributed generation with renewable Energy and its storage has changed the landscape of the power sector and energy usage in the building sector and the role that local energy communities play. Local community projects5 are diversified, ranging from energy generation by an individual for sharing with the community to a group of building owners or owners of different floors in a building or building in the near vicinity (Community). These local energy communities also cover the urban cooperatives set up to manage small- or medium-sized projects of energy generation and further distribution. These local energy concepts have extended their footprint to electric vehicle charging or car-pooling. The common philosophy of these projects is the sharing eco-systems of energy generated and go in for a drive in energy democratisation. The goal of the local community’s in the long term is to store energy in the home batteries. Thus, it is sharable between the community members, even to those without any solar rooftops. Therefore, a micro-grid acts as a free-standing grid connecting a couple of the LEC buildings. The micro-grid then extends the ower to a suburb or even a town if required. The paper “Sustainability 2021.”6 defines the principles of LEC. The power supply and generation primarily by solar rooftop and supplemented by the fossil-fired power supplied through the primary grid is only for a backup. Figure 2 depicts a schematic of the LEC network (Fig. 2). The success stories are too many, and writing the same runs into volumes; a selected few cases are hereunder Modified from https://www.abc.net.au/news/rural/ 2019-12-03/microgrids-set-to-transform-how-we-use-energy/11756672.

5 6

https://www.mdpi.com/2071-1050/13/4/2128/htm. https://www.mdpi.com/2071-1050/13/4/2128/htm.

6 Building Services with the Local Energy Community—Applications

137

Fig. 2 Grid-connected local community, solar power supported (Picture modified and adopted From “Micro-grids and neighbourhood power-sharing set to transform how we use energy” [3])

5.1 Australia Progresses on the Bill7 in the Parliament on Community Ownership The bill presented to the Australian Parliament, having sought for mandating the Community Ownership of regional renewable energy projects, has progressed. The bill seeks funding of $467millions to support regional communities for developing renewable energy projects for 50 regional communities. However, about 100 LECs do exist in Australia. A guide8 to set up LECs is explaining to develop a business plan on various technologies explains it very well. The guide talks of LEC for building structures and building (creating LEC). Some examples of diversified technologies are hereunder.

5.1.1

Yackandandah Town—North Victoria, Australia

Three community micro-grids connected various houses in the community, all with solar rooftops intended to be on 100% renewable energy. Even 14 houses have formed a LEC and maintained their building services on the micro-grid. 9 7

https://www.pv-magazine-australia.com/2021/06/29/bill-mandating-community-ownership-ofregional-renewable-projects-progresses/. 8 https://www.environment.nsw.gov.au/resources/communities/cpa-community-energy-how-to. pdf. 9 https://www.abc.net.au/news/rural/2019-12-03/microgrids-set-to-transform-how-we-use-energy/ 11756672.

138

5.1.2

Y. P. Chawla

University of Queensland—Australia

A first amongst major universities is the local energy community by itself. With 3.3 MW solar farm at Gatton, 160 GWh Warwick Solar (~powering 27000 homes, reducing the carbon footprint of 60,000 Tonnes/year), 1.2 MW solar rooftop at St. Lucia Camp, another 64 MW solar at Warwick are energy surplus. With 1.1 Mw Tesla Battery at St Lucia Campus, it saved $75,000 of electricity costs during the initial 3 months of operation and now trades Power (two-way) with the grid operator.

5.1.3

Middelgrunden Wind Farm—Australia

20 MW Wind Farm a local joint venture of 8500 Members.

5.1.4

Hepburn Community Wind Farm—Australia

4.2 MW Wind.

5.2 Europe Nearly 3250 LECs10 based on REs, involving about one million people.

5.2.1

Jühnde Bioenergy Village—Germany

700 kW Electricity and 700 kW Heat.

5.2.2

Torrs Hydro—UK

63 KW Hydro in English Town New Mills.

5.2.3

Alpine Region/South Tyrol-Italy

Three hundred small cooperatives and municipal-owned companies are operating as LECs. 10

https://www.google.com/search?rlz=1C1RXQR_en-GBAU935AU935&sxsrf=ALeKk01Xh rTMxB--x-h9JH-VZ_LolrxbeA:1629172823038&q=Number+of+Energy+Communities+in+Eur ope&sa=X&ved=2ahUKEwjN4JPblbfyAhXQfX0KHd_9DVAQ3rMBegQICBAC&biw=1536& bih=664.

6 Building Services with the Local Energy Community—Applications

5.2.4

139

EcoPower Cooperative

EcoPower Cooperative buys and builds renewable electricity units, such as wind and water turbines, in Belgium. Also gone and set up projects in Chile.

5.2.5

Germany

Germany11 has shown significant growth in local energy communities after the renewable energy sources made an appearance that required comparatively lesser technological inputs during installations and easily handled by local entrepreneurs. According to a study, the number of such local energy communities or energy cooperatives (Energiegenossenschaften) generating power from photovoltaic power mode rose from 45 numbers in the year 2008 to 200 numbers in the year 2011 and going down to a smaller number of 25 in the year 2016. Furthermore, 40,800 MW of PV Solar installations ownership belonged to Farmers (16%); Industry (24.4%); and Individuals sharing 33.1% of the total 40,800 MWe. The renewable energy total installations of 100.3 GWe got shared by Farmers (10.5%), 31.5% (Private) and 13.4% by trade. The decline in the growth of these local energy communities in Germany is due to regulatory hurdles. In June 2018, the Federal Parliament of Germany decided to suspend some of the privileges until 2020. This suspension came after criticism of a legal definition of “Citizen Energy” considered flawed. However, these privileges are being re-considered after a scary Climate Change report AR-6 dated 9 August 2021. Another report (source not visibly apparent) from Germany puts the numbers “Citizen form Cooperatives to drive the energy transition” from 66 in 2001 to 888 in 2013.

5.3 Scottish Model Community BENfits (COBEN)12 projects are delivering the benefits of LECs.

5.4 Enercoop French Enercoop French went from 3,000 clients in 2009 to 10,000 in late 2011. Enercoop has 20,000 clients (2014).

11 12

https://www.cleanenergywire.org/factsheets/citizens-participation-energiewende. https://www.localenergy.scot/what-is-local-energy/local-energy-plans/.

140

Y. P. Chawla

5.5 WindShare WindShare is a cooperative working wind power cooperative launched in February 2002 in Toronto, Ontario, Canada. Toronto Renewable Energy Co-operative (TREC), a non-profit incorporated in 1998, created this group. The TREC remains to exist as a separate non-profit entity.

5.6 Sweden Taxation based on Energy Efficiency 5.7 Case of the United States of America Over 100 LECs called Green Power Communities13 across 14 States and DC generates more than 8.4 billion kWh of green power in aggregate for local government use, residential use and the local business per a survey in 2020.

5.8 Emerging Economies Embracing LEC Concepts India started very late in adapting LEC but is catching fast. Forty-Six countries of sub-Sahara Africa has understood its importance in recent years.

5.8.1

Solarising Delhi

Initiative has led any Group housing societies to set up the common solar rooftop owned by the community (members and owners of the houses in the Group housing society), predominantly in the city of Dwarka—the SW of Delhi. The water pumping, sewage pumps, common lighted areas in parks and walkways, lighting in stairs are examples of using power from the LEC’s plant.

5.8.2

Chandigarh, A Union Territory of India

The group solar projects on the standard flat roofs of the markets are getting ready to get on the concept of LCE. Chandigarh has also adapted virtual net-metering by putting up the solar plants at a location away from their place of work or living wherever adequate space is available, feeding the energy generation to the grid and utilising at the home of their consumption. The issue rights of the roof ownership

13

https://www.epa.gov/greenpower/green-power-communities.

6 Building Services with the Local Energy Community—Applications

141

of a multistoreyed building by raising the solar plants above 7 feet through raised structures.

5.8.3

Puducherry—Another Union Territory of India

Group net metering by Aurobindo Ashram by setting up a solar plant at the roof of one of their building and feeding to the grid to be utilised at other buildings owned by Aurobindo Ashram is another example of LEC.

5.8.4

Cooperatives in India for Electricity for Local Electricity Community (Villages)

Anakapalle Rural Electric Society (Andhra Pradesh), Sircilla (Andhra Pradesh), Hukeri (Mysore), Kodinar (Gujarat), Mula-Pravara (Maharastra) and Lucknow (Uttar Pradesh) Started in 1974.

5.8.5

Desi14

Power (Decentralised Energy Systems India PVT. Ltd.) Renewables and Environment Araria, Bihar, Decentralised Energy Systems (India) Pvt. Ltd., more commonly known as DESI Power, is an independent rural power produce.

5.8.6

JUSCO

Jamshedpur Utilities and Services Company Ltd in Water and Electricity Distribution.

5.8.7

SFC Energy

SFC Energy is providing Power to local communities in remotes of North East of India by Hydrogen cells, supported by Solar Power and Li-Ion Battery.

5.8.8

Mera Gaon Power (MGP)

An entrepreneur has established Uttar Pradesh, India, the local energy community network to provide affordable and reliable power supply, including also adopting 14

Paul A, Tyagi Ruchika,WATER and ENERGY INTERNATIONAL September VOLUME 64/RNI, NO. 6 September 2021ISSN: 0974–4711.

142

Y. P. Chawla

enhanced payment collection mechanism while realising the full potential of microgrids.

5.8.9

Gram Power—Rajasthan, India

Under this model of local community energy, the model adopted is “pay-as-you-go” on a prepaid basis, circumventing the line losses and theft. Smart Prepaid metering and analytics are integrated with the help of big data and using easy-for rural folks to understand voice-enabled messages customised towards customer service.

5.8.10

Husk Power—Bihar, India

Supported on a cloud-based platform for ease in management from remote to monitor the performance at the site and the usage patterns of electricity on a real-time basis. Theft-proof system for reliable voltage supply with stable frequency. Big data analysis as a part of customer care and regular analysis of business operations. Waste monetisation through rice char, a waste from gasifier to create green incense sticks, gives local women employment.

5.8.11

Market Associations

In Chandigarh.

5.8.12

Gurgaon Societies

The above are a few examples in India.

5.9 Community Battery The power distribution companies take proactive steps to get closer to the consumers and not let their consumers out of their services, thus bringing in community batteries (Fig. 3).15 The results of Tesla battery installations in Lake Macquarie and Canterbury Banks town in Sydney—New South Wales, Australia, are to be seen. It took about 2 years after the decision to implement such a system, thus offering shared storage in areas with high uptake of the solar, facilitating the consumer getting a higher bang on the 15

https://reneweconomy.com.au/ausgrid-installs-first-of-many-community-battery-installationsin-sydney-network/.

6 Building Services with the Local Energy Community—Applications

143

Fig. 3 2 × 130 KW tesla power packs + 232 KWh Storage + MTU energy pack QS = 320 kW/550 kWh

solar earnings without each individual going in for an individual battery. This system extends the LEC footprint but with an external player that is the distribution company jumping in. The distribution or even the transmission companies’ strategies to soak up excess solar during the day, share it with the energy-deficit areas and release the stored energy after the sun-down. Consumers are back home needing more air conditioners to run. The community batteries will also act smoothening the demand curve and avoid peak power costs for the community’s benefit with the implementation of the community battery concept in 600 homes in Western Australia. Beacon Hill at Sydney, as in Fig. 4 is having a battery storage configuration of 150 kW/267 kWh Energy Pack QS. Residents with solar rooftops near the Northern Beaches of Sydney are getting covered for participation by Ausgrid (A short form of Australia Grid). The Ausgrid allows storage of the Power up to 10 kWh/day of excess solar can be covered at no cost to the prosumers. Similarly, 2 × 130 kW batteries with 232 kWh storage alongside MTU Energy Pack QS, thus providing 320/550 kWh storage capacity. The residents will receive credits on the excess energy stored, which is limited to 10 kWh/day. The savings expected are $50–$250/year (0.5–2%) per electricity bill (Fig. 5). Power Bank 3 is the latest edition of the community battery bank and is the largest in the scheme to date, covering up to 600 houses in Western Australia, enabling them to leverage the battery storage technology. The 600 household scheme follows the

144

Y. P. Chawla

Fig. 4 Beacon area Sydney—400 houses for community battery connection

installation of 9 numbers of 116 kW batteries across the state of Western Australia late last year. Melbourne, another city of Australia, is going in for putting community batteries in the suburbs of the inner-city in a joint venture with M/s Yarra Energy Foundation—a not for profit organisation. One in every 6 households in Australia have solar, but only 1 in 60 have battery storage. The solar power aggregation in the community battery acts as a drop box for the energy units and is drawable when needed by the individual consumer (Figs. 6 and 7). System of Virtual Power Plants—aggregating the surplus energy from the building were energy-positive and transferred to an Energy-deficient part of the community are available. The energy so aggregates act as a supply source to the energy-deficit area. Aggregation of extra energy from many building structures builds a source of energy as a power plant is called a virtual power plant. The community batteries also aggregate the surplus power. This way, we have graduated from adjusting the power peaks of individual consumers in the local energy community to a larger geographical area. By changing and meeting the power, peaks become a case of moving peak power load to a different time as is being done in Australia and is a valuable deployment experience in New Delhi and other places. The shifting of a load of the individual consumer by keeping them in the loop helps improve the economics by computing synergy with the energy within the community

6 Building Services with the Local Energy Community—Applications

145

Fig. 5 Bank town area—250 houses for community battery connection

(local or larger community or a state. The load curve gives the opportunity to shift electrical load to high RE time slots, ADR and BDR in residential, commercial and industrial and EV consumers. Balancing RE and BESS together with DR will be needed where technology plays a key role. The peak load curves and solar curves for New Delhi are as in Fig. 8. And the energy aggregation geographic footprint can extend and enhances the role of the distribution supply operator (Fig. 9). Solar energy is under-utilized at the time of generation; in the community sharing directly or stored in the community, battery thus gets pooled and used in the community is accounted for by two-way metering in an urban setting or rural setting. Earlier, when the local energy communities were of smaller scale and absence of metering, the accounting was manual, when the energy generation and consumption within the local community. With the advent of technology, it is possible to trade the Solar power P2P— Pear to Peer (at the time of generation or with the BESS—Battery Energy Storage System). The accounting is done on the electronic ledger and settled. The electricity regulation on P2P was brought in during the year 2020 by the Uttar Pradesh Electricity Regulation—UPERC. And the Government of India has allowed consumers with

146

Y. P. Chawla

Fig. 6 Virtual power plant power aggregated from nearby sources—concept capture from a case study of Australia

more than 100 kW to buy renewable energy in open access. The provisions are adopted to encourage the generation and consumption of power. A cyber-based block chain connects to the physical power generation and distribution system, enabling a trade environment among various stakeholders in a community on a micro-grid covering urban and rural set-ups.

6 Investments Appetite for Smart Buildings, Energy Efficiency and the Power Sector Value Chain The buildings and the household are giving importance to energy efficiency more than industry and services. Europe on energy efficiency trends are: Europe registered a higher saving of 50 Mtoe (4.1 Mtoe because of Structures/Buildings) in 2019 compared to 2014 in all sectors indicated above. (Energy Efficiency Trends in Buildings in Europe | Policy brief | ODYSSEE-MURE, 2016). With the rise of electricity cost of supply, the new buildings are to be energyefficient—thus, drawing the specifications and building properties accordingly marked. The building getting refurbished are also getting more intelligent and required to be so. The niche building sectors that are important in growth areas

6 Building Services with the Local Energy Community—Applications

147

Fig. 7 Virtual power plants—another layout of the same concept. The distribution system operator functions similarly to the regional system operator (at typical setup)

Fig. 8 Matching the renewable energy to meet peak loads

are helpful to target local energy community concept owned or leased by small or big organisations (Fig. 10). The more intelligent buildings using communication and technologies (ICT), interacting these with local energy communities is easier for operations and control and intelligent equipment in the buildings are capable of enhancing comfort and productivity with lesser energy. The energy saved is thus easily sharable with the community for drawing the benefits of electricity economies—a step towards affordable and clean Power. Intelligent technologies have penetrated the existing buildings

148

Y. P. Chawla

Fig. 9 The additional role of the distribution system operator on aggregating the surplus of the local energy community (ies) Concept sourced from Abhishek Ranjan BSES presentation

but are growing in all types of buildings. European Union has taken initiatives for making the buildings energy-efficient (Table 1). Learning from the lessons of the EU, Indian new upcoming buildings coming up now can embark upon electricity consumption reduction in domestic and commercial buildings by LECs which hovers around 24% and 8% respectively as of now (Figs. 11 and 12). The Power sector is also on the move as per the investment trend hereunder. Energy Cloud Eco-systems (ECE) are available for transforming the energydeficient building sector from Energy Cloud Ecosystem (ECE) that transforms the building sector from Energy to Energy-Efficient Grid Resource (EEGR), thus able to share energy in LEC. EERC developed for buildings each of 100 suites, 175 comfort spaces with a peak load of approx. 350 kW can provide a scalable capacity reserve of 70 kW/building with real-time analyses of various energy consumption points. (SPRY 2021).

7 Changing Landscape of the Power Utilities The investments committed in the power sector are coming with a changing landscape. The power sector companies are now working on social equity in the energy sector as the carbon-free future requires more and more social equity, and consumers

6 Building Services with the Local Energy Community—Applications

149

Fig. 10 Investments in building sectors where energy efficiency and LEC concepts can be deployed

are rightfully demanding the same. However, it is not an option by the power sector companies (Fig. 13). With the distributed energy resources increasing on individual consumer premises or the LEC, the role of the retail power distribution company will be challenging on the operation side and reaching a status of energy services company as shown in the top right-hand corner of the future construction of the electricity utility various role are required. The additional roles cover energy efficiency advisory to the consumer/prosumer, incorporating Distributed Energy Resources (DER) and Distribution into integrated resource planning. Now understand consumer

150

Y. P. Chawla

Table 1 List of the green recovery measures adopted by EU during 2020–2030 Name of measure Immediate launch of grants for energy renovations of buildings from the unused budget of 2020–21 New grant scheme for energy renovations of existing buildings, 2021–27 Grants for energy renovations of buildings under construction for an upgrade to Near-Zero Energy Buildings

Investment in decade 2020–2030 (Me) 30 140 70

Source https://www.odyssee-mure.eu/publications/policy-brief/covid-19-green-economic-rec overy.html

Fig. 11 Electricity consumption in domestic and commercial sector in India where LEC can be deployed

6 Building Services with the Local Energy Community—Applications

151

The Power sector is also on the move as per the investment trend hereunder

Paper On Building Services With The Local Energy Community 12102021 11/16 Fig. 12 The renewable energy takes a significant share along with distribution network establishes the importance of local energy community handling these

High Future of Electric U lity - Constructs

Opera onal Control of Distributed Energy Genera on

ESCO- Company Providing Energy Systems Pla orm- Distribu on System Resources

Provider- Distributed Energy Resources

Wires Only Content and Carriage Low

Interconnec ng and Integra ng

Concept Adapta on: Smart Energy Power Alliance (SEPA) 2019

Deployment Control of Distributed Energy Genera on

High

Low

Fig. 13 The changing pathway and future construct of a electric utility—Concept adapted from Smart Electric Power Alliance (SEPA)-2019

preferences, integrate heat and power requirements, gather sentiments on natural monopoly/private monopoly status, foster consumer choices.

152

Y. P. Chawla

8 Acceptability of the Local Energy Communities (LEC)/Cooperatives See Table 2. Table 2 Legal Status and acceptability of LECs/cooperatives—global snapshot # Organisation set-up

Autonomy of the LEC/cooperatives Tariffs

1. India Allowed with regulatory permissions

Independent—no formal political interference Top-down approach

By LEC but not more than regulatory authorities

2. The United States of America Renewable energy associations, set Politically and financially up by communities, as the independent borrower, organisation set up by the elected board of directors

By board of directors

3. Bangladesh Palli Biyut Samities (PBS) set up by Rural Electricity Board (REB) and Communities, Board of PBS is democratically elected and supervising Mangers are deployed by REB

PBS supposedly independent and privately owned. Financially monitored by REB. Economically dependent on REB

By PBS approved by REB

Financial independence, No assertion of direct political support seemed, NACEUN—association of such cooperatives advocate their Policy interests at National Level

By the local energy community within the limits set by Nepal electricity authority

1980s Political dependence and corruption within NEA affected the setups of LECs. Reforms led to greater Political independence. Today LECs/Cooperatives are Financially independent only. In August 2019, 121 Cooperatives worked for 13 Mn Homes

Tariffs by the Regulatory Commission only

4. Nepal Community initiative, board democratically elected

5. Philippines National Electricity Authority (NEA) established a Cooperative scheme and Funds. Cooperatives came up with their respective Boards, Supervising Managers by NEA

6 Building Services with the Local Energy Community—Applications Zone 1

Zone 2

Poly Centrism

153

Zone 3

Communi es Establishing Iden es

Communi es of Interests

Zone 2 Zone 3

Environment ConcernsHighlighted a er COVID -19 and importance of SDG Goals / Paris Agreement, AR6 Report Zone 2

Distributed Genera on

Local Energy Communi es Zone 4 Civic CulturePeer Pressure

Autonomy, Freedom, Independence Zone 3

Entrepreneurial Culture Electraprenuer

Contrasts and Linkages

Fig. 14 Adapted from—present trends of technologies, contrasts and linkages and inevitable trade-offs [8]

9 Road to the Future The last para shows us the path of the electric utility for the future. The local energy communities have triggered the changing landscape of energy systems by harnessing renewable energy and making a consumer to prosumer and prosumanagers thus actively participating in the energy transition for enjoying more incredible benefits. (Modified from European Commission’s statement).16 The LECs are now signing Power Purchase agreements with power generators to buy when tariffs are lower and sell the local energy produced when the tariffs are higher. LECs are making their members towards energy-efficient equipment and lighting (LEDs) and white goods. LECS are also working on car-pooling and transitioning to a vehicle fleet with low emissions, including electric vehicles. Zero waste and recycling, including sustainable food practices support, is on their agenda as they have professionalised the process as a business. The LECs are extending their domain of services (Fig. 14). The stakeholders in power sector and consumers opt for connection in micro-grid, which has shifted its base from rural to urban settings. The power sector bound by the regulatory environment needs supportive regulations for LEC. Co-design of the strategies by Discoms and LEC help to engage the stakeholders and manage the local energy projects.

16

https://ec.europa.eu/energy/topics/markets-and-consumers/energy-communities_en.

154

Y. P. Chawla

10 Conclusions The global lessons from Europe, the USA and Australia on energy communities gets extended to the globe. The utilities are getting reshaped and are offering different service packages. Utilities are helping the consumers integrate ever-expanding services and created a catalogue of such services providing distributed energy technologies in consumers’ life and the grid for carbon reduction while improving the grid. Consumers are getting changed to prosumers and prosumanagers. Climate control; pressures are leaving no options but adopting harsh measures on using carbon-free energy and using energy efficiently. Behind the meter, services have also taken a vital business model by the utilities. Planning the future of distributed energy Vol II (Tanuj Deora) as a sequel to Vol I by Black and Veech’s work. For a (net-zero energy) green building requirement, power control through automated systems makes the best quality power available. It utilises any generation and storage sources within the building. These hybrid-like systems allow optimum utilisation of the grid’s sources and varied loads to give the most stable system with quality power control. Per capita consumption is increasing, leading to increased power consumption, with diversified changes happening simultaneously. It is time to realign our actions by the local energy communities and building (creating) services new LECs for effective operations. The solar electric power associations turning into an association status (11 April 2016) to get more stakeholders involved.

11 Over to You After reading the contents of this book and chapter specifically, it is time to take up the work in this sector (“Author” also becomes a part of you after the writing part is over). Action has to start now as an individual or a business manager in any domain in any organisation. IPCC’s report AR6 sternly indicates that the time for talks is over, and action needs to start achieving net-zero emissions by building (creating) the local energy communities. In the changing climate change scenario incorporating features of reliable power to resilient power, the role of local energy community leaders has changed drastically (Fig. 15). The leaders must know their goal change and keep it updated by understanding the data that matters and looking beyond the technology by becoming a champion in energy, including strengthening the team while learning, teaching and evolving. All of us can take the concept of LEC further to have a climate-resilient power sector, with sustainable and robust features and yet giving us an affordable power to the society.

6 Building Services with the Local Energy Community—Applications

Dynamic shi of Finish Line

Data Ma ers

Be a Champion and build a team

Update Learn, Teach and Evolve

155

Looking Beyond Technology Concept Source : h ps://be erbuildings solu oncenter.energy. gov/5 -habits -energy efficiency-leaders

Fig. 15 Leadership of local energy communities

References 1. Ambole A, Koranteng K, Njoroge P, Luhangala DL (2021a) A review of energy communities in sub-saharan africa as a transition pathway to energy democracy. En.x-Mol.com; en.x-mol.com. https://en.x-mol.com/paper/article/1362182812497907712 2. Ambole A, Koranteng K, Njoroge P, Luhangala DL (2021) A Review of energy communities in Sub-Saharan Africa as a transition pathway to energy democracy. Sustainability 13(4):2128. https://doi.org/10.3390/su13042128 3. Davis J (2019) As home solar increases, the grid is becoming an antiquated technology. So what will replace it? ABC News-Print 4. Filho WL, Fedoruk M, Zahvoyska L, Ávila LV (2021) https://www.researchgate.net/public ation/351124445_ identifying and comparing the obstacles and incentives for the implementation in projects in energy saving in Eastern and Western European Countries an exploratory study/figures?lo=1. Research Gate; Research Gate. researchgate.net 5. G, J., Jun 24, hiok / T. / U., 2021, & Ist, 12:03. (2021, 24th June) Delhi yet to meet its 2022 solar power target, 7% achieved so far | Delhi News—Times of India. The Times of India. https://timesofindia.indiatimes.com/city/delhi/delhi-yet-to-warm-up-to-solar-energy/art icleshow/83791161.cms 6. Gasiorowski-Denis +41 22 749 03 25, E. (2016) Why investing in energy-efficient buildings pays off. ISO. https://www.iso.org/news/2016/11/Ref2140.html 7. HJ (2020) Energy communities. Energy—European Commission. https://ec.europa.eu/energy/ topics/markets-and-consumers/energy-communities_en 8. https://www.eca.europa.eu/Lists/ECADocuments/SR20_11/SR_Energy_efficiency_in_buil dings_EN.pdfc. (n.d.) 9. Odyssee-mure. eu (2016) Energy efficiency trends in buildings in Europe |Policy brief| ODYSSEE-MURE. Odyssee-Mure.eu; Odyssee-mure. eu. https://www.odyssee-mure.eu/pub lications/policy-brief/buildings-energy-efficiency-trends.html

156

Y. P. Chawla

10. PwC (2020) Real Estate Emerging Trends Asia Pacific 2020. https://www.pwc.in/assets/pdfs/ trs/emerging-trends-in-real-estate-2020.pdf 11. SEPA (2019) New Smart Electric Power Alliance report demystifies transactive energy. SEPA. https://sepapower.org/knowledge/new-smart-electric-power-alliance-reportdemystifies-transactive-energy/ 12. Smart Electric Power Alliance (SEPA) (2019a) Understanding and evaluating the potential models for the future electric power utility. www.sepapower.org; Smart Electric Power Alliance (SEPA). https://www.ourenergypolicy.org/resources/understanding-and-evaluatingpotential-models-for-the-future-electric-power-utility/ 13. SPRY G (2021) EEGR buildings target 58% of all energy wasted each year. Energy Central. https://energycentral.com/o/sensorsuite/eegr-buildings-target-58-all-energy-wasted-each-year 14. Spry G (2021) Energy Central; Energy Central USA. https://energycentral.com/o/energy-cen tral/special-edition-energy-efficient-grid-resource-buildings-glen-spry-ceo-president?utm_ source=browser&utm_medium=push_notification&utm_campaign=vwo_notification_162 8207340&vwo_powered=1 15. Vorrath S (2021) Ausgrid installs “first of many” community batteries on the Sydney network. RenewEconomy. https://reneweconomy.com.au/ausgrid-installs-first-of-many-community-bat tery-installations-in-sydney-network/ 16. Zachariadis T (2021) Green economic recovery—energy efficiency measures |Policy brief| ODYSSEE-MURE. https://www.odyssee-mure.eu/publications/policy-brief/covid-19-greeneconomic-recovery.html

Chapter 7

Energy Solutions for Smart Buildings Integrated with Local Energy Communities Shalika Walker, Pedro P. Vergara, and Wim Zeiler

Abstract The energy neutrality of buildings is subject to different interpretations and in the context of a building or neighborhood (or, equivalently, district), it can be understood to mean “the generation of equal electricity, as it consumes”. Nevertheless, it is found out that even though buildings can achieve energy neutrality annually, the energy self-sufficiency achieved is not satisfactory. Most of the locally produced electricity is fed back to the grid because of the mismatch of supply and demand. However, nearly self-sufficient buildings and neighborhoods will be a future requirement for newly arising smart buildings and neighborhoods with the electrification of energy systems. Therefore, in this chapter, with the aim of improving energy selfsufficiency, an analysis was conducted for a renovated smart neighborhood. In this context, a smart neighborhood is a group of highly performing houses, equipped with enhanced energy efficiency measures. These houses include improved insulation and high-efficiency heat pumps, electric water heaters, domestic appliances, and photovoltaic panels. In this neighborhood, to tackle the challenges associated with multiple distributed energy resources, and to design (size) and control them simultaneously, an optimization problem is developed and partitioned into sub-problems in a distributed manner. The study showed with the introduction of different storage systems and optimal smart control of these systems, the energy self-sufficiency of the neighborhood can be increased from the current 10% value to 71.5%. Keywords Energy self-sufficiency · Distributed optimization · Energy communities · Hydrogen storage systems · Thermal storage systems · Heat pumps

S. Walker (B) · W. Zeiler Eindhoven University of Technology, PO Box 513, 5600 MB Eindhoven, The Netherlands e-mail: [email protected] P. P. Vergara Delft University of Technology, Postbus 5, 2600 AA Delft, The Netherlands © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 A. Tomar et al. (eds.), Control of Smart Buildings, Studies in Infrastructure and Control, https://doi.org/10.1007/978-981-19-0375-5_7

157

158

S. Walker et al.

1 Introduction Along with the uprising regulations on climate change, governments in the European Union are to provide alternatives to the existing fossil fuel-based energy systems [1]. Among them, the replacement of natural gas-based heating systems has become a top priority in the Netherlands [2]. This transformation directly affects the building sector as a result of a large amount of natural gas used for space heating and hot water requirements [3]. Therefore, buildings are rapidly undergoing a physical and digital transformation to address various operational gaps that contribute to high energy use [4] and the replacement of gas-based heating systems [5]. With this digital and physical transformation, buildings are upgraded to ‘smart buildings’ with high efficiency and intelligent energy systems. In the future, a collection of these smart buildings or so-called intelligent communities equipped with cutting-edge systems and appliances will give customers more control over their building features and energy use. The community or neighborhood may contain different definitions or boundaries with the included number of buildings [6]. This selection is dependent on the geographical area as well as stakeholder inputs and building owners’ interest and willingness to participate. An illustration of a neighborhood boundary selection is presented in Fig. 1. When buildings and neighborhoods become smarter as well as the energy systems used inside the buildings are electrified (e.g. Heat Pumps (HPs) for space heating instead of gas boilers), clear evidence indicating the benefits of such systems must be outstretched first. Sources from which electricity is generated need to be taken into consideration. Indeed, should the electricity come from coal and gas power plants,

Fig. 1 Illustration of neighborhood boundary [7]

7 Energy Solutions for Smart Buildings Integrated with Local Energy …

159

Fig. 2 Trias Energetica and 5-step strategy [7]

such a transformation would not ultimately be sustainable and question whether the transition would be a real sustainable success. As a response to the above-mentioned concerns, ‘Trias Energetica’ was developed which is a three-step top-down strategy [3] to achieve an energy-efficient design by reducing energy consumption, local sustainable energy production, and using the most efficient conversion technologies. This applies to both new constructions and renovations. However, in the process of replacing fossil fuels entirely and achieving energy neutrality of buildings, a coherent mix of complex integrated energy systems involving renewable energy sources and energy storage options are needed [3, 8]. The major challenge is to create an optimally controlled energy system without compromising user comfort as well as adjust to the unpredictable behavior of the occupants. Therefore, the ‘Trias Energetica’ strategy was developed into 5 steps [7] including occupant behavior and storage systems and transforming it into a bottom-up approach (Fig. 2). Even though the 5-step approach demonstrated steps in realizing an energyneutral design, the intermittent behavior associated with distributed renewable energy sources has created a negative impact on the reliability of power supply and reinforcement of the grid [9, 10]. PV generation for all users aiming to reduce CO2 is not a straightforward solution because PV penetration is also limited by technical constraints on the electricity grid [11]. Most popularly in the electricity distribution grid, PVs and wind turbines create voltage fluctuations, excess production of energy, and power fluctuations [9]. Moreover, PV production and building demand patterns contradict each other, meaning, PV production occurs during the daytime and higher energy demand (for dwellings) occurs during the nighttime. Usually, it is possible to achieve energy neutrality of individual buildings annually (annual electricity consumption of buildings = annual electricity production of local renewable energy sources). However, self-sufficiency which is the ability of the building to function with its local energy sources without the need for external energy imports is surprisingly low [3, 7]. That means a higher percentage of the produced

160

S. Walker et al.

electricity is fed back to the grid without consuming it locally [12]. This study is motivated by this factor. Proper integration and collaboration of different storage systems (short-, medium-, and long-term storage) could effectively improve the self-sufficiency of buildings and neighborhoods. Storage units are generally classified into Electrical Energy Storage Systems (EESS) and Thermal Energy Storage Systems (TESS). Even though they are not extremely popular worldwide, electrical batteries have already been introduced at the building level [13] with renewable energy sources and with Electrical Vehicles (EVs). These EESS could be seen in elevated quantities at the neighborhood level in the future as the prosumers increase in number and with the persistently reducing costs of storage units [14]. Therefore, in this chapter, energy solutions for smart buildings integrated with local energy communities are discussed with the introduction of different storage systems at building and neighborhood levels. By combining these storage options along with an optimal sizing and operation problem in a distributed manner [15] using the energy hub concept, the improvement of buildings’ energy efficiency and self-sufficiency has discoursed.

2 Method and Models After introducing the case study buildings in the selected neighborhood, the modeling of the distributed optimization problem is introduced in this section.

2.1 Case-Study Neighborhood For this study, a recently renovated neighborhood is used which has no natural gas connection and is provided with PV systems, heat pumps, and high levels of insulation. There are 70 dwellings in the neighborhood. When considering a full year, the dwellings are meant to be ‘net-zero-energy’ (i.e. yearly consumption = yearly production). Specification of the houses as per the provider of the data set is shown in Table 1. The collected smart meter data from the dwellings are smart meter (SM) consumption, heat pump (HP) consumption, and PV production. SM consumption data does not include the electricity consumed by the HP. It is measured separately. Hourly consumption and production data for two consecutive years (2016–2017) has been used in this study, i.e. 17,520 data points per house. For the sake of the quality and reliability of the proposed method, a preprocessing step has been performed to clean the collected data by modifying the outliers and filling the missing data points [16]. The provided data contained some unrealistic outliers, valued a hundred times larger than the mean. The data is inspected visually to see where the threshold for a so-called outlier should be. According to this visual inspection, a threshold for

7 Energy Solutions for Smart Buildings Integrated with Local Energy …

161

Table 1 Technical specifications of the dwellings Specification

Value

Living surface

85–110 m2

RC -score roof

6 m2 ·K/W

RC -score walls

3.5 m2 ·K/W

U-value windows

1.1 W/m2 ·K

Hot water buffer

150 l = (0.15 m3 )

Heat pump power

1.2 kW (SPF 3.9)

PV installation capacity

5–8 kWp

Step-1 • Fill with previuos timestamp values

Step-2 • Fill with previuos day same timestamp values

Step-3 • Smoothen the data set with mooving average

Fig. 3 Missing value filling process

outliers is determined for each data set. Every value above this is replaced by a realistic replacement which is its preceding value. The whole process followed to fill the missing data points is illustrated in Fig. 3. After the preprocessing step, the evaluation of the optimal combination of energy systems was conducted which is described in Sect. 2.2.

2.2 A Distributed Energy Hub System To tackle the challenges associated with an optimization problem where multiple distributed energy resources (e.g. storage systems and HPs) are controlled simultaneously, the problem is partitioned into sub-problems in a distributed fashion [17]. This allows scalability and modularity of the overall system as well as increased resilience should one of its components fail. Additionally, the computational burden of the problem is alleviated by splitting the initially larger problem into multiple smaller ones [17]. However, coordination among the resulting sub-problems needs to be tended for in order to secure system stability and avoid conflicting local actions. To this end, this study considers a sequential distributed solving approach, also followed by Stadler and Ashouri [18]. The approach allows the exchange of information between the different sub-systems while keeping the complexity inherent to each model to itself [19]. This results in an information-optimized communication system where local energy flows exchanged through the distribution network

162

S. Walker et al.

are transferred from one optimized system to the next. Distributed optimization problems contemplate a few distinct control strategies, ranging from cooperative to competitive cases, depending on shared or conflicting goals the sub-systems might include. Sub-systems can be treated as different players interacting around a game (the optimization problem), where each individual action influences the game: a process known as Game theory [20]. To optimally increase the self-sufficiency of the residential district, a cooperative optimization is chosen, where all sub-systems share a common objective function. The operating limits of the energy system considered in this study are typically represented by the energy hub concept [21].

2.2.1

Energy Hub Concept

The concept of Energy Hub can be defined as a flexible interface [22] where multiple energy carriers meet each other and these energy flows can be converted, conditioned, stored, and finally distributed according to the demand requirements in an optimal manner [21, 23]. The modeling concept of an energy hub describes the relationship between input and output energy flows and can be used to optimize energy consumption during planning and operation. Figure 4 illustrates an energy hub composite of different energy carriers which optimally coordinates between supply and demand. This concept can be designed in different spatial scales according to the resources available and the level of complexity needed. In this context, complexity refers to the

Fig. 4 Illustration of an energy hub [24]

7 Energy Solutions for Smart Buildings Integrated with Local Energy …

163 Buildings

Hydrogen Storage System

Battery

Electricity Hot Water Tank Electricity Heat-pump Heating Boiler

Fig. 5 Energy hub district system

number of different energy systems attached. The following subsection describes the evaluated energy hub through two distinct scales, namely the grid and the building systems, and presents the distributed sub-problems considered for the optimization.

2.2.2

District Scale

The district system exemplifies the energy hub concept with both heating and electric distribution networks linking the building community to a shared set of utilities with the external energy grid (Fig. 5). In this setting, the hub allows the optimization model to consider a community heat pump, two short-term storage systems (a battery and a thermal storage tank), as well as a long-term hydrogen storage system composed of an electrolyzer, a hydrogen tank, and a fuel tank set up in series [25]. The aggregated set of buildings are equipped with solar panels, which are allowed to feed the larger, and more efficient, community storage systems. A similar setup is considered for the heat network of the hub, here taking into account heat losses from the network insulation.

2.2.3

Building Scale

While the district system considers the buildings as an aggregated community composed solely of input/output energy flows, the buildings’ sub-systems are each endowed with unique energy demand profiles resulting from the stochasticity inherent to occupant behavior. Each building is defined as an independent sub-system

164

S. Walker et al.

Photovoltaics

Battery Building

Electricity Hot Water Tank

Heating

Heat-pump

Fig. 6 Energy hub building system

connected to the above district system energy distribution network. It allows consideration of conversion (i.e. photovoltaics, heat pump) and storage units (battery, thermal storage tank) in its optimization scheme, as illustrated in Fig. 6.

2.2.4

Optimization Sub-problem

Setting up the distributed optimization with building and district systems treated separately would find suboptimal convergence in the operation utilities. The notable (and necessary) variation brought to the considered system for the optimal design of the energy utilities is to solve individual buildings and district-scale utilities conjointly (Fig. 7). This allows distributed problems to gather information on both available utility scales, with their costs and respective efficiencies to balance out. The solver can, in this way, choose an appropriate sizing of building scaled utilities and larger district scaled ones in a distributed fashion, with both scales incorporated into the problem’s state space.

2.3 Model Formulation The energy hub concept yields the basis for the model formulation, where both individualistic and aggregated utilities may be optimally sized and operated over a year’s horizon to increase the district’s self-sufficiency. In addition to individual energy systems, for residents with installed PV panels, this study discusses the optimal combination of community-scale and distributed utilities desirable to maximally utilize the PV-produced electricity on a global neighborhood scope. In the following sections, the presented model relies on the later formulation [26], where a scalar variable is represented by an italic letter, symbolizing the design value of a device. A bold letter stands for a vector, usually, time-dependent, which

7 Energy Solutions for Smart Buildings Integrated with Local Energy …

165

Fig. 7 Overall illustration of the district and building systems

points to the operational use of a particular variable. Finally, the superscript d denotes an independent decision variable. A Python-based implementation of the optimization problem is proposed using Pulp coupled to the Gurobi solver [27]. The below subsections present the models of the different utilities considered in the energy hub framework.

2.3.1

Heat Pump

The operation of the heat pump is modeled by considering an ideal Carnot heat pump, and a Carnot factor. By using this method, the effect of varying COP values is incorporated. Equation (1) describes the Carnot COP for a heat pump. Thigh is taken as the output temperature of the HP, being 328 K (55 °C). This temperature was chosen because the thermal grid supposedly provides the hot water demand. To avoid bacteria (legionella) conditions for hot water, at least a temperature level of 55 °C is required. Tlow is taken as the air temperature for air-source and ground temperatue for ground-source heat pump, varying throughout the year. Ambient data provided by Koninklijk Nederlands Meteorologisch Instituut (KNMI) is used in this study [28]. C O P H P,car not =

Thigh Thigh − Tlow

(1)

166

S. Walker et al.

The real performance of the heat pump can be described in terms of the Carnot factor, Carnot , and presented by Eq. (2). Typical values for Carnot vary between 0.3 and 0.5 for small electric heat pumps [29]. In this research, a value of 0.5 is used for  Carnot. The seasonal performance factor of the current heat pumps is known to be 3.9. The value chosen for the Carnot factor aligns when calculating the seasonal performance factor of the current heat pumps. car not =

C O P r eal C O P car not

C O P r eal = car not .

(2)

Thigh Thigh − Tlow

The constraints used in the modeling of the heat pump are represented by Eq. (3). In the equation, QHP denote the heat pump output energy and C HP the capacity limit: 0 ≤ Q dH P ≤ C Hd P .T ; T = 1 hour

2.3.2

(3)

Storage Systems

To account for cyclic daily patterns of residential buildings, this study considers two short-term storages in its optimization scheme, i.e. a battery and a thermal storage tank. Water-based thermal storage tanks and typical Red-Ox batteries were considered suitable systems for community and distributed use. These systems are widely employed in such settings and possess attractive economical features. Additionally, a long-term community storage system is added with a hydrogen storage system. Both long- and short-term storage systems are modeled using similar linear equations. Equations (4)–(7) describe the storage model used. E t = E t−1 .(1 − ∝decay ) + ηCharge .E dcharge,t −

1 η Discharge

.E ddischarge,t

(4)

d 0 ≤ E t ≤ C ST O

(5)

d max 0 ≤ C ST O ≤ C ST O

(6)

d 0 ≤ E dcharge/discharge,t ≤ C ST O .δCharge/Discharge

(7)

At each time interval t, the storage level is represented by Et in kWh with the variable Echarge/discharge representing input and output energy flows. Storage system parameters can be found in Table 2. The notation E is used for electric storage, while

7 Energy Solutions for Smart Buildings Integrated with Local Energy …

167

Table 2 Storage systems parameters Symbol

Description

Unit

Utility Battery

Water tank

Hydrogen storage system

∝decay

Decay rate Small (building system) Large (grid system)

–

0.01 0.003

0.01 0.005

– 0

ηCharge

Charging efficiency Small (building system) Large (grid system)

–

0.9 0.95

0.9 0.9

– 0.58

η Discharge

Discharging efficiency Small (building system) Large (grid system)

–

0.9 0.95

1 1

– 0.52

max C ST O

Maximum capacity Small (building system) Large (grid system)

kWh

15 1000

300 3000

– 5000

δCharge

Charging rate

–

1

0.25

1

δ Discharge

Discharging rate

–

1

0.20

1

for thermal storage, notation Q is used. The Q denoted equations are not shown but carry the same format.

2.3.3

Hydrogen Boiler

A straightforward energy convection model is considered to model the boiler, which transforms stored hydrogen into heat through combustion. d 0 ≤ Qdboiler ≤ Cboiler .T

(8)

d max 0 ≤ Cboiler ≤ Cboiler

(9)

Qdboiler = ηboiler .Qdhydr ogen

(10)

The sizing constraints (Eqs. (8)–(10)) restrain the utility capacity to a maximum, and the boiler efficiency is taken into account through the ηboiler parameter.

2.3.4

Energy Networks

This study considers both an electric and heat grid for district energy optimization in order to consider which optimal setup would be the most advantageous for the community. Indeed, enabling buildings to exchange energy between one another creates a larger state space resulting in more ways to optimally manage the energy

168

S. Walker et al.

of a particular system. The system modeling is carried out on two different scales, echoing the distributed optimization layout of the system evoked in Sect. 2.2, i.e. the building and the district scales. Building: The building system energy balance block considers a total of 4 balance equations: the intrinsic heat and electric balance equations of the system (Eqs. (11), (12)), and its connections to the district energy grids (Eqs. (13), (14)). E dblg,in + E P V + E dB AT,blg,out = E dblg,out + E dH P,blg,in + E dB AT,blg,in + E demand (11) Q dH P,out + Q dT S,blg,out + Q dblg,in = Q demand + Q dT S,blg,in + Q dblg,out E dblg,in/out [bi ] +



E dblg,in/out [b] = E dgrid2blg,in/out

(12) (13)

b∈B\{bi }

Q dblg,in/out [bi ].σloss +



Q dblg,in/out [b] = Q dgrid2blg,in/out

(14)

b∈B\{bi }

Parameters Edemand and EPV represent the electric energy demand and PV production of the building, respectively, while E(blg,in/out) presents the energy consumption of the building block or its feedback to the grid. E(BAT,blg) is the electrical energy flow entering or exiting the battery associated with the building at hand, and E(HP,blg,in) is the energy consumed by the heat pump. Equations (13) and (14) represent the electric and heat grid connecting the buildings together to the grid system. It should be noted that the heat distribution system incorporates energy losses from the network’s insulation: σloss . This factor is fixed at 5% for building systems and 10% for the later grid system. District: The district network is quite straightforwardly composed of two main balance equations, relative to the heat and electric grid considered. E dgrid,in + E dgrid2b lg,out + E dH 2,out + E dB AT,grid,out = E dgrid,out + E dH P,grid,in + E dH 2,in + E dgrid2b lg,in + E dB AT,grid,in

(15)

Q dH P,out .σloss + Q dT S,grid,out .σloss + Q dgrid2blg,out = Q dT S,grid,in + Q dgrid2blg,in (16) Both grids allow the possibility to include community storage systems for heat and electric energy carriers. This scale of the model plays a linking role between the building systems and the high voltage electrical distribution system, from which the energy is purchased. 2.3.5

Objective Function

The objective of this problem is to optimally operate and design energy systems while maximizing their self-sufficiency. The self-sufficiency of the neighborhood

7 Energy Solutions for Smart Buildings Integrated with Local Energy …

169

exercised in this study can be described with the below equation.  self − sufficiency =

t

E P V − E dgrid,out  d t E grid,in

To maximize its value, the objective function presented in Eq. (17) minimizes economic costs resulting from utility investments, maintenance as well as energy purchasing costs (grid feed-in). The particularity of distributed optimization problems is that all problems may have independent and/or common objectives, which results in systems functioning either in cooperative or competitive ways. Considering the global objective this study undertakes, a cooperative distributed optimization is preferred, where all systems share the same objective function—therefore, converging to a common goal. U 

CiI nvestment + CiO M F + CiV

(17)

C I nvestment = Cost unit .Cap unit .AC F unit

(18)

Obj : min

i=1

AC F unit = 

r



(19)

C O M F = O M F unit .Cap unit

(20)

CV =



1−

1 (1+r )

Li f etime

E dgrid,in .celec

(21)

t

Total costs comprises investment cost (CInvestment ), fixed operation and maintenance costs (COMF ) and variable costs (CV ) of energy systems (e.g. HP, thermal storage), summed over every considered utility device (unit). Annualized cost factor (ACF) is calculated based on the estimated lifetime of the energy systems and discount rate (r). Applicable parameters linked to the objective function are presented in Table 3.

3 Simulation Results and Discussion Following the method and models formulated in Sect. 2, the simulations were conducted, and the results are discussed accordingly. As discussed in the objective function, the overall aim of the study is to increase the energy self-sufficiency of the neighborhood so that the energy consumed from the external electricity grid is minimized and consumption from the locally produced energy is maximized. This was

170 Table 3 Objective function parameters

S. Walker et al. Symbol

Description

Value

Reference

r (%)

Discount rate

6

[25]

celec (e/kWh)

Electricity tariff

0.22

[30]

Estimated lifetime (yr)

Thermal storage Battery storage Heat pump H2 boiler H2 storage system Fuel cell

20 12 20 12 22 15

[25, 31] [25] [25] [32] [24] [24]

Investment cost (e/kWh)

Thermal storage Small tank (10 m3 ) Battery Small Large Heat pump H2 boiler H2 storage system

50 10 700 578 1600 (e/kW) 3200 (e/kW) 20

[33] [34] [24] [24] [35] [25]

OMF Operation and maintenance -yearly (e/kWh)

Battery Small Large Thermal Storage Small tank (10 m3 ) Heat Pump H2 boiler H2 storage system

14 11.5 0.3 0.5 5.4 (e/kW) 0 (e/kW) 0

[36] [33] [25] [35] [25]

achieved by introducing different types of storage systems at the building level and district level, and discovering the optimal combination of different storage systems at the building and district levels. The simulation study shows the effectiveness of the proposed algorithm in limiting the aggregated power flow of fed-back electricity. In order to emphasize the effectiveness of the cooperative approach, a comparison has also been performed between cooperative and individualistic (non-cooperative) behaviors of buildings as energy players [37]. The difference in the cooperative and non-cooperative approaches appears in the objective function. In the cooperative case, district system and building systems minimize one global objective function while the other case, local objective functions. Figures 8 and 9 present the aggregated power flow from the external grid toward the district level (EGrid,In ) and vice versa (EGrid,Out ) with the cooperative approach and a non-cooperative scheme. From these figures, it

7 Energy Solutions for Smart Buildings Integrated with Local Energy …

171

Fig. 8 Electric energy flow from the external electricity grid toward the district system

Fig. 9 Electric energy flow from the district system toward the external grid

is visible, especially during summertime, the cooperative approach performs better than the non-cooperative case. It emerges that the distribution algorithm converges in eleven iterations (see Fig. 10, Illustration of algorithm convergence). In order to confirm the cooperative distribution approach converges to a global optimal rather than a suboptimal state, the distributed approach is compared with a centralized optimization using a few houses. This proof of concept is demonstrated in Appendix A. Table 4 reveals the optimal combination of different energy systems that were obtained as results of the cooperative optimization problem. This analysis showed that incorporating electric battery storage either at the building or district level is not economical when H2 and thermal water storage are available. However, this can be because of the battery storage associated costs that were used in the optimization problem. In the future, if the battery storage is cheaper than the water storage or H2 storage, and/or when the feed-in tariff and electricity tariff change, the results

172

S. Walker et al.

Fig. 10 Illustration of algorithm convergence

Table 4 Optimal combination of energy systems at building and district levels Level

HP

H2 boiler

H2 storage

Thermal water storage

Battery storage

District level

×

×

✓

✓

×

Building level

✓

Not considered

Not considered

×

×

could take a different turn. Captivatingly, the results show that aggregated level H2 storage with the electrolyzer configuration brings value to the neighborhood in terms of increasing energy self-sufficiency.

3.1 Grid System This section discusses the district-level behavior of the different storage systems. In Fig. 11, an illustration of the energy flows at the district level is represented. The notations corresponding to Fig. 11 are elaborated in Table 5. In Fig. 11, it is possible to observe that during the winter and autumn periods 268 MWh of energy is drawn from the external grid toward the neighborhood. However, during the summertime because of the introduced H2 storage and electrolyzer system, PV-produced electricity fed back to the grid from the neighborhood has drastically reduced. Before introducing the storage systems, only with HP operation, the fed-back electricity of the neighborhood is counted to be around 2000 MWh. After introducing different storage systems at district and building levels and the optimal scheduling of these systems with distribution optimization, the fed-back electricity has reduced up to 50 MWh, thereby increasing the self-sufficiency of the neighborhood from 10 to 71.5%. In order to acquire this increment, according to

7 Energy Solutions for Smart Buildings Integrated with Local Energy …

173

Fig. 11 Electric energy flow of the district-level energy systems

Table 5 Elaboration of the notations Notation

Explanation

EGrid,In

Electric energy flow from the external electricity grid toward the district level

EBattery,Grid,In

Electric energy input to the district-level battery storage system

EBattery,Grid,Out

Electric energy output from the district-level battery storage system

EH2,In

Electric energy input to the district-level H2 storage system

EH2,Out

Electric energy output from the district-level H2 storage system

EHP,Grid,In

Electric energy input to the district-level HP

EGrid,Out

Electric energy flow from the district system toward the external grid

the optimization problem, a thermal water storage capacity of 685 kWh and an H2 storage capacity of 695 kWh are required.

3.2 Building System The distributed optimization results for each of the 70 building systems showed that no storage systems are acknowledged in the considered building systems. But, HPs are shown to be favorable for houses to provide the required heating demand. Figure 12 shows the obtained optimal HP capacities of the 70 houses in descending order so that the required heating demand is fulfilled. The HP capacities range from 1.5 to 4.5 kW.

174

S. Walker et al.

Fig. 12 Optimal HP capacities required to provide the heat demand

4 Conclusion In this chapter, a distributed control of a local energy community using the energy hub concept has been developed to improve energy self-sufficiency. The study was theorized with the fact that the current local consumption of PV electricity is not enough when the energy systems associated with the grid electrify. For the simulation, several storage options have been used at the building and community levels. The cooperative distributed algorithm has been implemented in order to converge toward equilibrium while ensuring that the global constraint shared by all the buildings is respected. The distributed control leads to the maximization of local energy exchanges via the storage systems. The studied cooperative optimization problem proved to have better performances. Traditionally, the control of generation units is performed in a centralized manner. This centralized methodology requires a high bandwidth communication infrastructure and a high level of connectivity. However, when the number of participants and distributed resources increases in number, the centralized approach can limit its implementation. In contrast, distributed approaches offer features that make them scalable and adaptable while decreasing the computational time needed for convergence. These characteristics make the distributed methods robust, allowing them to respond to changes in the number of operating units, unexpected increases in the renewable generation, or load consumption. While discussing the possibility of improving energy self-sufficiency, this study proves the potential of applications of distributed control on the local energy community level. However, for the practical realization of such proposed neighborhood energy management systems, the traditional energy market strategies need to be redesigned. In conclusion, this chapter showed the clever utilization of locally produced renewable electricity for a better alignment of supply and demand to confront a future-proof smart energy system at the neighborhood level.

7 Energy Solutions for Smart Buildings Integrated with Local Energy …

175

Acknowledgements The research work is funded by NWO Perspective program TTW Project B (14180)—“Interactive energy management systems and lifecycle performance design for energy infrastructures of local communities” (https://ses-be.tue.nl/). We wish to express our gratitude to all the organizations which have been a part of this program. A special thanks is given to Julien le Prince for helping to complete the simulation. We express our gratitude to BAM Bouw en Techniek for providing data.

Appendix A: Proof of Concept This section validates the utilized cooperative distributed optimization by comparing it with a centralized optimization scheme using five dwellings. In Fig. 13 after eleven iterations, the cooperative optimization problem reaches a convergence. The objective value corresponds to the convergence point equals e 7641.397. For the centralized optimization, the obtained objective value equals e 7564.226 making the difference between the two methodologies about 1%. Therefore, the used cooperative distribution approach is proven to be converged to the global optimal state.

Fig. 13 Illustration of algorithm convergence for proof of concept

176

S. Walker et al.

References 1. Hansen K, Connolly D, Drysdale D, Thellufsen JZ (2016) Heat roadmap Europe: identifying the balance between saving heat and supplying heat. Energy 115:1663–1671. https://doi.org/ 10.1016/J.ENERGY.2016.06.033 2. Rijksoverheid: Energierapport, transitie naar duurzaam (2016) 3. van Leeuwen R (2017) Towards 100% renewable energy supply for urban areas and the role of smart control. http://purl.org/utwente/doi/https://doi.org/10.3990/1.9789036543460 4. Lund H, Werner S, Wiltshire R, Svendsen S, Thorsen JE, Hvelplund F, Mathiesen BV. (2014) 4th Generation district heating (4GDH). Integrating smart thermal grids into future sustainable energy systems. Energy. 68:1–11. https://doi.org/10.1016/j.energy.2014.02.089 5. Fawcett T (2011) The future role of heat pumps in the domestic sector 6. Monti A, Pesch D, Ellis KA, Mancarella P (2017) Introduction. In: energy positive neighborhoods and smart energy districts. Elsevier, pp 1–5 7. Walker S (2021) Sustainable energy transition scenario analysis for buildings and neighborhoods: data-driven optimization. https://research.tue.nl/en/publications/sustainable-energy-tra nsition-scenario-analysis-for-buildings-and 8. Lund H, Mathiesen BV (2009) Energy system analysis of 100% renewable energy systems— the case of Denmark in years 2030 and 2050. Energy 34:524–531. https://doi.org/10.1016/j. energy.2008.04.003 9. Passey R, Spooner T, MacGill I, Watt M, Syngellakis K (2011) The potential impacts of grid-connected distributed generation and how to address them: a review of technical and nontechnical factors. Energy Policy 39:6280–6290. https://doi.org/10.1016/j.enpol.2011.07.027 10. Jarnut M, Wermi´nski S, Wa´skowicz B (2017) Comparative analysis of selected energy storage technologies for prosumer-owned microgrids 11. Vergara PP, Salazar M, Mai TT, Nguyen PH, Slootweg H (2020) A comprehensive assessment of PV inverters operating with droop control for overvoltage mitigation in LV distribution networks. Renew Energy 159:172–183. https://doi.org/10.1016/j.renene.2020.05.151 12. Rae C, Bradley F (2012) Energy autonomy in sustainable communities—a review of key issues. Renew Sustain Energy Rev 16:6497–6506. https://doi.org/10.1016/J.RSER.2012.08.002 13. Cox R, Walker S, van der Velden J, Nguyen P, Zeiler W (2020) Flattening the electricity demand profile of office buildings for future-proof smart grids. Energies 13:2357. https://doi.org/10. 3390/en13092357 14. Hoppmann J, Vollanda J, Schmidt T, Hoffmann V (2014) The economic viability of battery storage for residential solar photovoltaic systems—a review and a simulation model. Renew Sustain Energy Rev 39:1101–1118. https://doi.org/10.1016/J.RSER.2014.07.068 15. Vergara PP, Rey JM, Shaker HR, Guerrero JM, Jorgensen BN, Da Silva LCP (2019) Distributed strategy for optimal dispatch of unbalanced three-phase islanded microgrids. IEEE Trans. Smart Grid. 10:3210–3225. https://doi.org/10.1109/TSG.2018.2820748 16. Walker S, Bergkamp V, Yang D, van Goch TAJ, Katic K, Zeiler, W (2021) Improving energy self-sufficiency of a renovated residential neighborhood with heat pumps by analyzing smart meter data. Energy 229:120711. https://doi.org/10.1016/j.energy.2021.120711 17. Costanzo GT, Gehrke O, Bondy DEM, Sossan F, Bindner H, Parvizi J, Madsen H (2013) A coordination scheme for distributed model predictive control: Integration of flexible DERs. In: 2013 4th IEEE/PES innovative smart grid technologies Europe, ISGT Europe 2013 18. Stadler P, Ashouri A, Marechal F (2016) Distributed model predictive control of energy systems in microgrids. In: 10th annual international systems conference and proceedings, ( SysCon), pp 1–6. https://doi.org/10.1109/SYSCON.2016.7490607 19. Stephant M, Abbes D, Hassam-Ouari K, Labrunie A, Robyns B (2021) Distributed optimization of energy profiles to improve photovoltaic self-consumption on a local energy community. Simul Model Pract Theory 108:102242. https://doi.org/10.1016/j.simpat.2020.102242 20. Yang B, Johansson M (2010) Distributed optimization and games: a tutorial overview. Lect. Notes Control Inf. Sci. 406:109–148. https://doi.org/10.1007/978-0-85729-033-5_4

7 Energy Solutions for Smart Buildings Integrated with Local Energy …

177

21. Geidl M, Andersson G (2007) Optimal coupling of energy infrastructures. Presented at the (2007) 22. La Scala M, Vaccaro A, Zobaa AF (2014) A goal programming methodology for multiobjective optimization of distributed energy hubs operation. Appl Therm Eng 71:658–666. https://doi. org/10.1016/j.applthermaleng.2013.10.031 23. Geidl M, Koeppel G, Favre-Perrod P, Klöckl B, Andersson G, Fröhlich K (2007) Energy hubs for the future. IEEE Power Energy Mag 5. https://doi.org/10.1109/MPAE.2007.264850 24. Walker S, Labeodan T, Maassen W, Zeiler W (2017) A review study of the current research on energy hub for energy positive neighborhoods. In: Energy procedia. https://doi.org/10.1016/j. egypro.2017.07.387 25. Murray P, Orehounig K, Grosspietsch D, Carmeliet J (2018) A comparison of storage systems in neighbourhood decentralized energy system applications from 2015 to 2050. Appl Energy 231:1285–1306. https://doi.org/10.1016/J.APENERGY.2018.08.106 26. Ashouri A, Fux SS, Benz MJ, Guzzella L (2013) Optimal design and operation of building services using mixed-integer linear programming techniques. Energy 59:365–376. https://doi. org/10.1016/J.ENERGY.2013.06.053 27. Gurobi The fastest solver—Gurobi, https://www.gurobi.com/ 28. KNMI—Uurgegevens van het weer in Nederland—Download, https://projects.knmi.nl/klimat ologie/uurgegevens/selectie.cgi 29. Marini D, Buswell RA, Hopfe CJ (2019) Sizing domestic air-source heat pump systems with thermal storage under varying electrical load shifting strategies. Appl Energy 255:113811. https://doi.org/10.1016/j.apenergy.2019.113811 30. Netherlands household electricity prices (2020) Statista, https://www.statista.com/statistics/ 418106/electricity-prices-for-households-in-netherlands/ 31. Lund H, Mathiesen B, Østergaard P (2019) Book of abstracts. In: 5th International conference on smart energy systems 32. Boiler Manufacturers Call for Hydrogen Boilers by 2025 Boiler Guide, https://www.boiler guide.co.uk/hydrogen-ready-by-2025 33. Lund H, Østergaard PA, Connolly D, Ridjan I, Mathiesen BV, Hvelplund F, Thellufsen JZ, Sorknses P (2016) Energy storage and smart energy systems. Int J Sustain Energy Plan Manag 11:3–14. https://doi.org/10.5278/ijsepm.2016.11.2 34. Jülch V (2016) Comparison of electricity storage options using levelized cost of storage (LCOS) method. Appl Energy 183:1594–1606. https://doi.org/10.1016/j.apenergy.2016.08.165 35. Hydrogen powered boiler developed by Giacomini, https://www.boilerguide.co.uk/hydrogenpowered-boiler-developed-by-giacomini 36. Larsson P, Börjesson P Cost models for battery energy storage systems 37. Altman E, Avrachenkov K, Miller G, Prabhu B (2007) Discrete power control: Cooperative and non-cooperative optimization. In: Proceedings—IEEE INFOCOM, pp 37–45

Chapter 8

Towards Advanced Technologies for Smart Building Management: Linking Building Components and Energy Use Ghezlane Halhoul Merabet, Mohamed Essaaidi, Hanaa Talei, and Driss Benhaddou Abstract With over 4.4 billion people living in urban areas, buildings are currently the greatest energy consumers on a worldwide scale, accounting for 40% of the global energy consumption and using 50% of the world’s final electrical energy. The built environment, as being the focus of numerous types of research, is the key platform for introducing technologies and regulations that can contribute to decreasing energy consumption and improving the energy efficiency of accessible services. Technological advancements have made it possible to build more autonomous and efficient buildings capable of managing their infrastructures and services, such as lighting or airconditioning systems, using a number of predefined patterns via software as a decision-making basis. Indeed, the diagnosis of the new applications allowed by the advancement of research in Information and Communication Technologies, and especially in the area of the Internet of Things (IoT) makes it possible to observe that the concept of “Smart Building” has also evolved. Hence, through this chapter, we examine how the introduction of the new technologies in the building sector can be used to improve the requirements of environmental quality and energy use. In this context, we included a case study with the purpose to show the importance of using building-related data analysis and Machine Learning techniques to define the energy inefficiencies in highly efficient Leadership in Energy and Environment Design (LEED), Energy Star, and Net Zero-certified building in Houston, Texas, USA. Keywords Smart building · Advanced control techniques · Environmental comfort · Energy efficiency · Energy management G. H. Merabet (B) · M. Essaaidi SSL, IT Rabat Center, ENSIAS—Mohammed V University, 713 Rabat, Morocco e-mail: [email protected] H. Talei School of Science and Engineering, Alakhawayn University, 1005 Ifrane, Morocco G. H. Merabet · D. Benhaddou Department of Engineering Technology, University of Houston, Houston, TX 77204, USA © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 A. Tomar et al. (eds.), Control of Smart Buildings, Studies in Infrastructure and Control, https://doi.org/10.1007/978-981-19-0375-5_8

179

180

G. H. Merabet et al.

Abbreviations AI AmI ANN ASHRAE BACnet BEMS BMS EE FLC GA HCS HMI HVAC IAQ ICT IoT IPCC MACES MAS MDP ML PID PMV RBF SBS SOAP WHO XML

Artificial Intelligence Ambient Intelligence Artificial Neural Networks American Society of Heating, Refrigerating, and Air-Conditioning Engineers Building Automation and Control networks Building Energy Management System Building Management System Energy Efficiency Fuzzy Logic Control Genetic Algorithm Human-Centric Systems Human–Machine Interface Heating, Ventilation, and Air Conditioning Indoor Air Quality Information and Communication Technologies Internet of Things International Panel on Climate Change Multi-Agent Comfort and Energy System Multi-Agent System Markov Decision Problems Machine Learning Proportional–Integral–Derivative Predicted Mean Vote Radial Basis Function Sick Building Syndrome Simple Object Access Protocol World Health Organization Extensible Markup Language

1 Introduction Building environmental effect has been gaining traction in the agendas of a number of cities and countries throughout the world since the 1990s. Indeed, the usage and occupation of buildings account for approximately 38% of global final energy consumption growth between 2015 and 2050, with developed (or industrialized) countries accounting for the majority of this demand, as shown in Fig. 1. As a result, the buildings sector was recognized a world pioneer in CO2 emissions in the International Panel on Climate Change’s fifth report (IPCC) [1]. However, given

8 Towards Advanced Technologies for Smart Building Management …

181

Fig. 1 Final energy consumption by region. From [2]

design opportunities, technical advancements, and user behavior, the same report identifies the sector as having the most potential for lowering CO2 emissions. Indeed, it is commonly acknowledged that building energy use varies substantially between regions and countries, depending on environmental situations, economic strength, accessible technology, and cultural patterns. Figure 2 shows that, by splitting the building into commercial and residential, the demand for space cooling and /heating in residential buildings in the United States, in 2010, was around 43% followed by water heating at 13%.

Fig. 2 Primary energy consumption in US residential and commercial buildings in 2010. From [3]

182

G. H. Merabet et al.

It is critical to adopt initiatives and solutions capable of decreasing energy usage in order to meet this growth in energy demand. Energy savings, through technological substitution and process improvement, are part of the scope of strategies taken by the majority of industrialized and developed countries when planning for energy efficiency. Progress tends to bring new technologies to the market that are capable of providing the same service as previous ones but using less energy. Over the last decade, several research works have been performed on the application of Information and Communication Technologies (ICT) in buildings. These technologies, according to research works done by the European Union, will allow for a 15% decrease in the global energy consumption in the building sector in the future years [4]. The expected benefit is associated with the introduction of new technologies or solutions that can provide users with comfort, security, and resources in a more efficient and intelligent manner than existing systems. Furthermore, as a result of technological developments, significant population growth, and shifting/fluctuating resource prices, there is a greater need for thermal and visual comfort, as well as indoor air quality. In this regard, efforts are currently directed at fulfilling the building energy demands; by ensuring operational needs while optimizing their resources: maximum efficiency vs. minimum cost and optimum indoor environmental quality. Smart objects, embedded systems with local processing, data acquisition, provision of actions and interactions, and above all, the ability to make decision are among the technologies used in buildings. Smart objects have evolved simultaneously with the progress of the Artificial Intelligence (AI) and the Internet of Things (IoT) concepts. The IoT refers to the ability of objects (or devices) to interact with each other, share information across a dynamic infrastructure, and open up new possibilities for virtual and physical world interactions [5]. When used in buildings, these technologies allow equipment and systems to be connected via a communication network, allowing information to be exchanged between components and improving the overall performance of the system. Whereas, AI has been used to control both conventional and bioclimatic buildings via computational models of intelligence. To overcome the non-linearity of the Heating, Ventilation, and Air Conditioning (HVAC) systems, as well as the latency and uncertainty of the systems; advanced paradigms underlying AI, such as Artificial Neural Networks (ANNs) and Genetic Algorithms (GAs), were developed. In this context, various AI-based tools for energy and environmental comfort inside buildings have been introduced [6]. The remainder of this chapter first covers the importance of Energy Efficiency (EE) in buildings while presenting different ways of EE assessment (cf. Sect. 2). Section 3 depicts the accessible services within smart buildings; whereas Sect. 4 provides advanced solutions used to control smart building services as well as energy consumption. Section 5 shows the importance of these advanced control techniques, such as data analysis and ML, in evaluating the energy inefficiencies in a Platinum Leadership in Energy and Environment Design (LEED) building case study, and finally, Sect. 6 concludes the chapter.

8 Towards Advanced Technologies for Smart Building Management …

183

2 Energy Efficiency in Buildings 2.1 The Importance of Energy Efficiency Currently, energy consumption has become an indicator of a country’s activity and development. However, this perception has been extrapolated from consumption and development to all industrial and social sectors, ignoring the parameters of energy consumption efficiency, and is only valued by short-term profitability indices, resulting in a situation in which the actual Energy Efficiency (EE) measures have less impact than desired. A system that rates EE exclusively in terms of short-term advantages can lead to a conception of immediate well-being, which ignores the continuity of social comfort in the long- and medium-terms. Whereas an action-oriented on maintaining comfort and activity indices employing efficient generation, transportation, and energy consumption procedures can maintain and even increase both societal comfort of the society and the industrial productivity while reducing the number of the Megawatts consumed. Reducing consumption through EE implies allowing more people access to the desired level of comfort while reducing production costs, enhancing industry competitiveness, and making it a driving force in a country’s economy, which is why the building sector has enormous potential in the field of EE [7]. EE in a building is recognized as the adaptation of the building to its surroundings to reduce its energy demand, as well as the use of solar or supplementary energy to meet the energy requirements of buildings in terms of heating, cooling, and lighting with the goal of effectively reduce conventional energy use. By reducing energy consumption, CO2 emissions and other pollution agents are consequently lowered. Hence, the following are the goals to be achieved with Energy Efficiency: • Providing appropriate conditions for more energy-efficient buildings in both new and existing buildings, while considering the surrounding climate without being extraneous to the building’s architecture; • Promoting the use of renewable natural resources for building conditioning, also known as natural conditioning techniques, while considering the building’s components, construction techniques and location; • Integrating active solar systems for thermal heating or energy generation as a building component. In order that buildings work efficiently, we must understand how a building works. To do so, we should first understand the components that hold the building’s minimum EE criteria, as it is built for the operation to which it will be applied. EE can be described as any activity that aims to optimize energy consumption. Energy undergoes a more or less lengthy transformation process before being changed into heat, cold, movement or light, during which a portion is wasted and the other, which reaches the user, is not always fully used, resulting in damaging waste to the environment.

184

G. H. Merabet et al.

Some of these losses are unavoidable and result in physical issues, while others are due to poor system performance and lack of optimization. Because the energy is transformed later, when it is consumed, EE necessitates the applications of strategies and methods to combat energy waste throughout the process. However, energy loss occurs not only during the transformation and conversion phases but also during the consumption phase. In this context, the EE of buildings attempts to provide optimum comfort to users while consuming the fewest resources possible. Based on the quality of the construction, with an appropriate selection of materials, and the use of renewable energy resources, with a special attention to the bioclimatic architecture techniques. As a result, when a building provides the same climatic conditions with less energy usage, it is more energy-efficient than others. The transformation of this paradigm requires a high level of social awareness, investment, and innovation, as well as the adoption of an adequate and balanced political policy that promotes the sustainable improvement of building energy efficiency. These behavioral adjustments do not need to cause discomfort, deprivation, or resource reductions; rather, they should promote access to knowledge about how much energy is spent and wasted so that smarter solutions can be chosen.

2.2 Evaluating Energy Efficiency The Energy Efficiency of a building is determined by calculating or measuring the energy consumption required to meet the building’s annual energy demand under normal operating and occupancy conditions, and is expressed qualitatively or quantitatively using indicators, indices, and ratings, or letters on a conventionally determined scale ranging from higher to lower efficiency. The energy evaluation is expressed through several indicators that allow explaining the reasons for good or bad energy performance of the building and provide useful information on the aspects to be considered when proposing recommendations to improve such performance. These indicators, which refer to the unit of the useful surface area of the building, on an annual basis, shall be obtained from the energy consumed by the building to be satisfied under certain climatic conditions. The requirements associated with normal operating and occupancy conditions, which shall include the energy consumed in heating, cooling, ventilation, domestic hot water production, and, where appropriate, lighting; to maintain thermal and lighting comfort conditions as well as the indoor air quality. The main or global EE indicators are: • The annual CO2e 1 emissions; • The annual consumption of non-renewable primary energy. 1

CO2e : CO2 equivalent. In the context of energy certification, references to CO2 emissions correspond to CO2e emissions.

8 Towards Advanced Technologies for Smart Building Management …

185

These main indicators include the impact of the services of heating, cooling, domestic hot water production, and in uses other than residential buildings (housing), lighting, as well as the reduction of emissions or consumption of non-renewable primary energy derived from the use of renewable energy sources. The complementary EE indicators are: • The annual heating energy demand; • The annual energy demand for cooling; • The annual non-renewable primary energy consumption disaggregated by services; • The annual CO2e emissions disaggregated by services; • The annual CO2e emissions disaggregated by electricity consumption and other fuels. The services considered in the complementary indicators are those of heating, cooling, domestic hot water production, and in buildings used for purposes other than private residential use, also lighting. The units used to express these indicators will be kW h per m 2 of the usable floor area of the building, for demand or consumption values, and kg C O 2e per m 2 of useful area of the building, for emissions values.

3 Managing Services Present in Different Parts of the Smart Building 3.1 The Notion of Service The capabilities of the systems that exist in a building are evaluated by the function they perform. These functions, which can be extremely diverse, have specific characteristics (such as nature, scope, or objectives) that allow them to be classified into sets. In this context, the concept of Service is introduced, and it refers to a group of functions that, by their nature, close interrelation and/or dependence, sharing or intervention on common information, and association with the same type of physical equipment, justify their arrangement in an individualized entity [8]. The concept of Service is fairly broad, and it may be applied to a wide range of disciplines, not just automation and building management. A service’s functions, on the other hand, may not always have to be associated with physical equipment (including interactions with sensors and actuators). Some of the services that can be provided in a smart building are Miscellaneous Services (inventory and asset management, vehicle parking control, elevators, irrigation control, management, and administration system, people localization and equipment); Energy (energy management, lighting, HVAC); Maintenance (fault diagnosis and system maintenance, cabling management, building maintenance);

186

G. H. Merabet et al.

User interactivity (presence management); Security (emergency detection, information, surveillance, and intrusion detection, access control, communications, audio, and video distribution, ordinance support). All services allow for configuration and management of the service, allowing for its adaptation to each case (it is possible, for example, to define which equipment is associated with the service and what its type, to define its identification, the building locations in which it is installed). It is also possible to monitor and test the operational state of mechanical devices and their control equipment in order to detect failures and record operating times (this information will be very useful for carrying out maintenance actions). In the context of smart buildings, there is a large number of applications that can be automated, such as HVAC systems, lighting, energy control, and so forth [9]. All of these can be grouped into services such as air quality, lighting, thermal comfort, energy control, occupant’s control, access control, etc. Hence, we will go over a set of smart building services and describe a summary of their functions. It is intended to approach the smart building from a functional perspective, providing a diversified and comprehensive vision that encompasses multiple areas of interest and not just those traditionally associated with Technical Management.

3.2 Building Utilities One of these specialized systems is the Building Management System (BMS), which includes building utilities such as electrical, hydraulic, and lighting control, as well as transportation systems such as elevators and escalators, and also water, gas, and energy consumption measurements. The BMS offers real-time information and status of all devices to enterprise operational managers, primarily to improve system maintenance through historical events, management reports, and the interactivity between the various components that constitute the system. One of the BMS modules, known as Building Energy Management System (BEMS), is a control system that monitors and controls the mechanical and electrical equipment in a building, including HVAC systems, lighting, power systems, fire systems, and security systems. A BEMS consists of an integrated software and hardware platform. The BEMS is often configured hierarchically, utilizing protocols such as C-bus and Profibus, whereas BEMS systems combine Internet protocols and open standards such as DeviceNet, SOAP, XML, BACnet, LonWorks, and Modbus are also available on the market. Furthermore, a BEMS can increase the building’s performance and ease of operation throughout its life cycle. A building’s major purpose with BEMS is to minimize the long-term costs of facility ownership for owners, occupants, and the environment. All components of such buildings are integrated to operate together, improving operational performance, occupant comfort, and satisfaction, and providing the owner with systems, technologies, and tools to manage and minimize energy use [10]. Owning a building with a BEMS means having access to [11]:

8 Towards Advanced Technologies for Smart Building Management …

187

• Practical information regarding the performance of building systems and facilities; • Proactive monitoring and detection of errors or deficiencies in building systems; • A level of business system integration that delivers real-time reports on management operations, energy consumption, and occupant comfort; • Tools, technology, resources, and practices that contribute to energy conservation and environmental sustainability.

3.3 Occupant’s Control As the occupants are the main agents in the activities that take place within the place under study, it is crucial to understand their demands and to know whether or not they are present. The activities of the occupants generate heat release into the surrounding space, and the amount of heat released varies depending on the type of action performed [12]. Methods such as timing, detection, or prediction can be used to determine the number of occupants. Scheduling is the simplest approach to use; however, it also has the highest risk of error, because it is based on the inhabitant’s day-to-day behavior, and if they change their behavior for whatever reason, the equipment will start working in an unjustified manner. Besides, detection uses sensors and other equipment to offer real-time information on the presence of people. Presence sensors are classified as passive infrared, microwave, or dual-technology sensors [13]. Sensors have some restrictions in terms of operation, which can result in false or undetected presences. A false presence occurs when the sensor detects someone who is not present in the space, while a non-detected presence occurs when the sensor detects no inhabitants in the space [14]. Dual technology sensors can reduce this error, as both have to be triggered before sending information to the controller. Another approach is described in [14], which combines a passive infrared sensor with a magnetic blade that detects the opening and closing of doors, and can also reduce the associated inaccuracy, with the detection of inhabitants based on the opening of the door and the infrared detection. Prediction is a more complex way of determining when the occupant is on-site, and it relies on historical data to forecast future events. One of the methods used for predicting events is ANNs, which, based on a set of data, can learn and anticipate when the next entry will be and how many people will be on-site. In terms of predicting, the concern arises as to what happens if occupants decide not to go home as usual on a given day, as well as the costs of keeping the system working when no one is present. Prediction, on the other hand, is better suited to places with a high number of frequencies, and situations where occupants are unable to anticipate entries.

188

G. H. Merabet et al.

3.4 Environmental Control—HVAC Systems The HVAC control systems, which are generally automatic, are designed to maintain a consistent indoor climate throughout the building—i.e., they regulate the indoor climatic conditions to give optimal comfort. They also allow: • Each occupant to adjust the temperature of his workspace or department, always within the restrictions set according to the location of the building; • Monitoring and adjusting the temperature in each space according to predefined values; • Adjusting the indoor air quality based on occupancy of spaces or existing requirements; • Adjusting humidity, temperature, and airflow velocity; • Allowing to predict and use air volume distributions in any space. Despite its evolution, HVAC equipment still accounts for over half of all energy consumption in buildings [15], therefore thorough HVAC control is crucial for a considerable gain in building EE [14, 15]. The majority of HVAC studies are based on models that try to optimize the quality of data on occupants—i.e., estimating if there are occupants, how many there are, and where they are using it, in order to make more efficient use of this resource [16]. Since HVAC is the equipment that allows to actively change the interior temperature of a room, a concept that specifies the intended comfort temperature inside, also known as the reference of Set-point temperature, must be introduced. The Set-point is determined independently for each space and based on the demands of the inhabitants [14]. It is worth noting that HVAC equipment uses more energy when it has to create greater temperature variations in the room, therefore, one of the goals of improving EE and lowering consumption should be gradual rather than abrupt temperature changes. Some behaviors can improve HVAC efficiency that can be overcome by setting constraints in place so that it only turns on when these are met. In [15], they describe an inadequate behavior: “[...]when the HVAC is on, the doors and windows are not closed or did not close immediately[...]”. This behavior can be avoided by implementing the following constraint: “If the doors and windows are closed, then the HVAC system turns on”. Whereas, in [14] the HVAC control used the occupancy of the room as an input—i.e., the equipment is turned ON or OFF depending on whether there are people in the room, in which case the temperature change could be abrupt. This information is relayed to a central system, which automatically changes the HVAC status of a specific zone to “ON” or “OFF” based on the information received.

8 Towards Advanced Technologies for Smart Building Management …

3.4.1

189

Indoor Air Quality

The connection between using a building as a workplace or residence and the manifestation of discomfort and symptoms that fulfill the definition of a disease, in some situations, is no longer debatable. The primary cause is the pollution of various forms in the building, which is commonly referred to as “poor indoor air quality.” Many people are affected by the negative impacts of poor indoor air quality since studies reveal that urban populations spend between 58 and 78% of their time in an indoor environment that is contaminated to some degree. It is a problem that has been worsened by the development of buildings that are more airtight and recycle air with a reduced amount of fresh air coming from outside to improve EE. Buildings lacking natural ventilation are now widely acknowledged to be at risk of contamination. Indoor air is the most commonly used to describe non-industrial indoor environments such as office buildings, public facilities (such as schools, hospitals, theaters, restaurants, etc.), and private houses [17, 18]. The concentrations of contaminants in the indoor air of these structures are often similar to those found outdoors, and far lower than those found in the industrial environment, where air quality is assessed using reasonably well-known standards. Indoor air quality depends on some parameters: • Biological, such as mold and bacteria; • Chemicals, such as CO2 and aero-disperse; • Physical, such as temperature, humidity, and airspeed. The behavior of the Indoor Environmental Quality (IEQ) of buildings has an impact on the health, productivity, and well-being of the building’s occupants—i.e., Indoor Air Quality (IAQ), along with living costs and energy consumption [19]. Given the negative effects of poor IAQ on humans, the World Health Organization (WHO) defines “Sick Building Syndrome” (SBS) as a situation of perception on the part of building occupants about several criteria of their state of health. SBS is a term used to describe situations in which at least 20% of the users of a given building experience negative health and comfort effects, which usually tend to disappear after short periods away from the affected people, such as mucosal irritation, neurotoxic effects, respiratory and skin symptoms, and altered senses [20]. Characteristics commonly used for IAQ assessment include the quantities/concentrations of pollutants in the air, as well as temperature and humidity. Other parameters used, but less frequently, are pollutants on internal surfaces, illumination, and the acoustical conditions of the environment. The rate of outdoor air supply to a building—i.e., the air renewal rate, is handled as an IAQ condition because it has a direct impact on the concentration of indoor air pollutants.

3.4.2

Thermal Comfort

Besides ensuring the stability and safety of its occupants, one of the most significant characteristics of a building is that it provides good thermal comfort conditions while

190

G. H. Merabet et al.

Table 1 The Seven-point thermal sensation scale. Adapted from ISO 7730-2005 [26] Sensation

Cold

Cool

Slightly cool

Neutral

Slightly warm

Warm

Hot

Vote

−3

−2

−1

0

+1

+2

+3

respecting the temperature and the environment in which it is located [21]. Thermal comfort in the built environment is a tough topic to define. A considerable variety of indices have been developed in recent decades for climate analysis and HVAC control system projects [22–25]. For example, Fanger proposed a model [25], known as the Predicted Mean Vote (PMV) index, for estimating thermal comfort that includes four physical parameters (air temperature, relative humidity, air velocity, and mean radiant temperature) and two individual variables (clothing insulation and metabolic rate) based on the thermal equilibrium approach (cf. Eq. 1). The closer the PMV value is to zero, the better the occupants’ thermal comfort sensation. M − W = E + RH E + K + C + S

(1)

• Whereas M is the metabolic rate (W/m2 ); W is the mechanical work done by the body (W/m2 ); E is the evaporative heat gain or loss (W/m2 ); RHE is the radiant heat exchange (W/m2 ); K is the conductive heat transfer (W/◦ C.m2 ); C is the convective heat transfer (W/m2 ); and S is the net heat storage (W/m2 ). Subsequently, the American Society of Heating, Refrigerating, and AirConditioning Engineers (ASHRAE) established a thermal comfort scale (cf. Table 1). The PMV index was then adopted by the ISO 7730 standard, which specifies that the PMV be maintained at level 0, with a tolerance of 0.5 as the optimal level of thermal comfort. Controlling HVAC systems has been the subject of numerous studies. However, this issue is only addressed in terms of temperature regulation, ignoring the impact of other factors on thermal comfort. This line of research is still under development. However, even when this strategy is used in new commercial buildings with sophisticated HVAC systems, individuals may be dissatisfied with their thermal comfort. Furthermore, direct control of comfort is an alternative to temperature control (which implies the temperature). This kind of method was previously controlled using conventional control techniques such as ON/OFF control. The main limitation of these controllers, however, is that they do not consider energy savings and they have poor temperature regulation. Otherwise, ultimate control involves sophisticated techniques that combine various forms of action mechanisms, such as shading devices, automatic opening/closing of windows, and so forth [27, 28], with the primary purpose of reducing the usage of air conditioning systems and hence energy consumption [29].

8 Towards Advanced Technologies for Smart Building Management …

191

3.5 Intelligent Lighting Control One of the most important systems of a smart building is lighting control, which is extremely flexible and simple to use, and allows for the creation of quiet and comfortable conditions while also helping to save electrical energy [30]. The fundamental notion of lighting control is based on the functionality and features that are performed in a specific location (reception halls, living room, bedrooms, offices, convention centers or events, etc.), as well as the time of day, time of year, or the ambient light intensity. Certain illumination levels are encoded into the controls, which can be controlled either automatically, via dimmers, or central control system. The energy savings are derived from the detection of occupancy of a space; sensors are placed in rooms, stairs, corridors, and other areas; when they detect movement, they send a signal that causes the lighting to turn on for a set period (motion sensors); after this time, the area is censured to detect movements or some presence again. If there are no sensors, the system will automatically turn off the lamps in that area, only to turn them back on when there is a new movement [31]. Due to the automatic activation of the lights when detecting a presence in a certain sector, lighting plays a critical role in security, detecting intrusions in sectors that should not be occupied at certain times. To summarize, lighting control systems enable to: • Turn ON/OFF automatically the lights using sensors or computers; • Modify illumination levels using presence detection, natural light, window proximity, or events or scenarios that occur at a specific location (conferences, meetings, etc.); • Alter the lighting brightness by the users via a computer or even a phone, depending on their preferences; • Control the entire system from a central computer, which may be used to man-age certain sectors or the entire building, as well as perform activities such as turning off, turning on, and adjusting the brightness level of the lights; • Manage the energy usage by occupation or sector, as well as lighting adjustments based on occupant preferences or predetermined levels; • Obtain a history of consumption and energy savings to establish, maintain or change policies in this field.

4 Advanced Techniques for Smart Building Services and Energy Use Management As a result of technological advancements, there is a compelling need to design and build intelligent systems that not only reduce the number of resources wasted but also allow for the replacement of existing systems with newer and more efficient ones. Users’ abilities and knowledge have been transferred to systems that can control

192

G. H. Merabet et al.

equipment and make decisions automatically. These systems are becoming increasingly self-contained and can be used in a variety of settings. This section discusses some of the proposed solutions for integrating intelligent systems into buildings and related equipment. They can be classified into (i) fuzzy logic control, (ii) learning approaches, and (iii) distributed Artificial Intelligence (AI) and multi-agent systems.

4.1 Fuzzy Logic Control A Fuzzy Logic Controller (FLC), unlike traditional control techniques, is more commonly utilized in a set of complicated operations that can be managed by a fully trained human operator with little knowledge of the underlying dynamics. The main idea behind FLC is to include an Expert Experience of the human operator into the control of a process whose input–output relationship is specified by a set of fuzzy control rules (e.g., IF–THEN), rather than a sophisticated dynamic model. In [32], Guo and Zhou looked at how fuzzy logic may be utilized to control HVAC systems, while the authors of [33] and [34] developed a fuzzy-PID (Proportional– Integral–Derivative) control mechanism based on the deployment of a virtual PMV sensor. As a result, the fuzzy logic technique can be used to simulate building user behavior and produce language descriptions of thermal comfort sensations that approximate the PMV model and simplify control system calculations. The performance outcomes of an HVAC system controlled by fuzzy logic by Lea et al. [35] have been explored when applied to HVAC systems. They brought up a point about the performance of systems concerning people’s sensitivity. Hamdi et al. [36] have made a significant investment in this direction, developing the notion of comfort conditions control based on human sensitivity, rather than maintaining a constant indoor temperature but rather a continuous interior comfort. The findings showed that it is possible to integrate people’s comfort with resource conservation. Fuzzy logic approaches are not limited to a single field of application. For example, in [37], fuzzy logic was used to regulate the environmental conditions (thermal, visual, and IAQ) within a building. These strategies can also be used to solve optimization challenges. While in [38], they introduced a conventional hierarchical threelayer control strategy based on fuzzy logic, intending to increase comfort while improving EE in a built environment. Furthermore, by combining the concepts of the IoT with fuzzy logic, the authors suggested a new approach for autonomous temperature regulation in [39]. Fuzzy logic facilitates the association of multiple types of variables. In this regard, Javaid et al. [40] developed a globally adaptable thermostat for regulating energy usage in hot and cold climates by incorporating factors such as external temperature, human presence, energy cost, and interior temperature set-point using fuzzy logic. The results revealed that using fuzzy logic instead of traditional models lowered energy consumption with improved efficiency. This new generation of intuitive and judgment-based control systems attempts to offer a simple, adaptable, and efficient control without relying on a structural

8 Towards Advanced Technologies for Smart Building Management …

193

description of the process. Their performance is typically compared to that of conventional PID/PI/PD controllers, with the advantage that they consider knowledge of the system’s behavior (represented in fuzzy natural language or assimilated through neural network learning methods), or a degree of optimality (e.g., Genetic Algorithms (GAs)). These techniques have been used for temperature control in buildings since the 1990s.

4.2 Learning Approaches Machine Learning (ML) is a term used to describe a variety of self-learning methodologies. They were created to allow users to manage their equipment according to their preferences. These methods include Q-learning, Markov Decision Problems (MDP), Linear Programming (LP), Evolutionary Algorithms (EA), Neural Networks (NN), Dynamic Programming (DP), and so on [41]. Neural network models have been used to tackle challenges in the field of thermal buildings, such as HVAC system control, in a number of studies [42–44]. A neuron, on the other hand, is a single-output information processing unit with a large number of inputs. The functionality of these networks is based on learning methods that enable them to recall and categorize data, and they are frequently used to handle recognition and classification difficulties. In situations when the PMV index could not be computed, these models were developed to define the concept of thermal comfort. To demonstrate the method’s efficacy, an experiment was carried out in an air-conditioned room. As a result, both works [45, 46] look into the evaluation of individual thermal comfort. Modeling the thermal and energy characteristics of an enclosure also uses neural networks. Such a model can be used, for example, to estimate the temperature of a greenhouse or the energy use of a passive single-zone solar building, the methodologies for which are discussed in [47, 48]. In addition, neural networks have been utilized to modify the three PID parameters of an air-conditioning controller online in order to improve control performance [49]. In [50], they present a study of two types of feed-forward-type learning algorithms for a comparable system. Another application of neural networks was proposed in [51]: calculating the best moment to start heating after a period of non-occupancy. The inside temperature, outdoor temperature, and their gradients are the network’s input variables in this approach. The drawback of this technique is that it requires a large amount of data for the learning operation in order to provide precise results. In some studies, neural networks are used in conjunction with GAs. In [52], for instance, an ANN is used to simulate the behavior of the building and then combined with a multi-objective evolutionary algorithm for system optimization. Whereas in [53], Radial Basis Function (RBF) networks were coupled with a multiobjective optimization-GA to develop prediction models for two variables required for assessing thermal comfort and air temperature, as well as relative humidity. Likewise, a reinforcement learning-based control method was implemented in [54]

194

G. H. Merabet et al.

by using the plug-and-plug technique and learning the system dynamics from the acquired environmental data. In [55], a Deep Reinforcement Learning (DRL) strategy was used to optimize the HVAC control.

4.3 Distributed Artificial Intelligence and Multi-Agent Systems In monitoring complex systems, designers are required to develop and implement real-time control systems that incorporate a set of controllers rather than a single controller system. Furthermore, the human component is part of any control system, whether it acknowledges or rejects a specific control approach (learning via reinforcement). These systems are referred to as Human-Centric Systems (HCS) [56]. Currently, the focus of control engineering is to structure the problem by breaking it down into a series of manageable sub-problems. The multi-controller system is currently being developed, and it will be deployed as part of a broader controlleragent architecture. These controller agents are driven by a coordinating agent for optimal utilization in this architecture [57]. The use of intelligent systems in smart buildings allow to monitor the environmental comfort while optimizing the energy consumption. Using multi-objective Markov Decision Problems (MDP), Klein et al. [58] proposed Multi-Agent Comfort and Energy System (MACES), a framework for coordinating equipment and inhabitants of a building using a multi-agent system for energy management in Domotics. This system was installed and assessed buildings in terms of thermal zones, temperature, occupants’ preferences, and occupants’ schedules using four control strategies: baseline, reactive, proactive, and proactive MDP. Hierarchical multi-agent systems, on the other hand, are addressed in [59–62] for application in smart and energy-efficient buildings for comfort and energy control. In [63, 64] the authors described a strategy based on a multi-agent system for the control and real-time learning of smart home features. Minar et al. [63] approached the construction of a distributed architecture for ubiquitous computing agents from a wider perspective. In [64], a parallel architecture was proposed, however, the latter takes a different approach and concentrates on the implementation details of the agents, largely to fulfill the demands of compactness, whereas the first does not. Multi-Agent Systems based on orientated approach interactions for optimizing comfort management and HVAC operations were developed in [65], by coordinating the HMI (Human–Machine Interface) and MAS Communities. Another agent-based system control was developed by Barakat and Khoury [66], which investigates multicomfort (visual, thermal, and acoustic) level control to minimize energy usage. An agent-based model was proposed in [67] to mimic the impact of human behavior on comfort and energy consumption within a residential building. The developed model provided an accurate estimate of energy consumption levels.

8 Towards Advanced Technologies for Smart Building Management …

195

Other studies have used fuzzy logic to create new types of intelligent agents that incorporate more and more humans into dynamic and embedded systems such as smart buildings, while considering the criteria: high speed, reliability, as well as meeting environmental changes. In this regard, Naji et al. [68] established a simulator based on a multi-agent system that assures a balance of work among the different equipment within a building, as well as occupant comfort, and energy usage control.

5 Case Study: Energy Efficiency in Platinum LEED Building The case study presented in this section shows the importance of using building related data, data analysis, and ML in identifying energy inefficiencies even in a highly efficient Leadership in Energy and Environment Design (LEED), Energy Star, and Net Zero-certified building in Houston, Texas, USA. The building is managed by Houston Advanced Research Center building (HARC) and is an 18,600 square foot office building. HARC was certified by the International Living Future Institute (ILFI) as zero energy which made the building the first and the only net zero energy office in Texas and one less than 55 office buildings in the USA. Net zero energy simply means that more renewable power was produced on-site than the power consumed by the building over the past 12 months. As part of EE analysis, the data acquisition system in the HARC’s building collects 1 min power consumption data at different building levels including lighting, plug loads, HVAC system, and others. For a better understanding of the collected datasets, it is important to explore data as it reveals new paths for discovery which help in identifying future problems. The sampling rate was changed from 1 min to 1 h for a better prefiguration of data as presented in Fig. 3 which presents the change of

Fig. 3 Meters consumption—hourly

196

G. H. Merabet et al.

the meters power consumption hourly and unveiled some opportunities for energy savings. Given that Houston weather has a humid subtropical climate, the HVAC meter revealed a higher consumption than other meters, as shown in Fig. 4, prompting the conclusion that the study presented in this chapter should be limited to HVAC meter data only. Building energy managers may generally improve HVAC operations by adjusting the thermostat temperature based on the number of people in a building and occupied space, as well as setting up a schedule based on the business operation times: day vs. night, and weekdays vs. weekends. However, HVAC systems can still be inefficient since some areas in the building is not being occupied during business hours, and the number of people fluctuates throughout the building’s operation. At HARC, the energy manager is using a strategy to change the set point, generally, twice a day at 06:00 AM and 05:00 PM which is clearly noticeable in Fig. 5 as the HVAC consumption exhibit a clear change at these time slots. The temperature in Houston typically begins to increase about 10:00 AM, reaching a peak around 03:00

Fig. 4 August 2019 power consumption

Fig. 5 HVAC consumption change throughout the day

8 Towards Advanced Technologies for Smart Building Management …

197

Fig. 6 Correlation heatmap matrix: HVAC consumption, Number of users and Outdoor temperature

PM, and then drops afterwards to settle around 08.00 PM. These temperature patterns are clearly depicted in the Fig. 5. In addition to using occupancy data, the analysis presented in this chapter is using outdoor temperature data collected from WILLIAM P. HOBBY AIRPORT STATION weather station, which provides hourly temperature data. In order to investigate the change of HVAC consumption in relation to the number of users and the outdoor temperature, the correlation heatmap matrix was used to display the relationships between the three features, as shown in Fig. 6. Unpredictably, the correlation between the HVAC consumption and the outdoor temperature is relatively not very high which implies the use of other techniques to better analyze/explain such behavior. ML is a field of AI aiming at designing algorithms that learns from data and adapt to new data without human intervention. In data analysis, ML is considered as a subprocess meant for building a data model automatically; classifying or clustering data is a ML task but, at the same time, a core task of data analysis as well. Given its simplicity, which is a main requirement of big data analytics, K-Means was used to cluster the data used in this case study which managed to generate three groups of data with different characteristics: • Group1: aggregated data with high HVAC consumption and many numbers of users and high outdoor temperature; • Group 2: grouped data with low HVAC consumption, low number of users and outdoor temperature; • Group 3: clustered data with high consumption while the number of users or/and outdoor temperature are low. K-Means was able to spot a cluster with time slots characterized with energy inefficiencies explained by HARC energy manager to areas that did not switch to unoccupied mode properly, or a day that was input as a vacation day in the system, etc. Data Analysis is a key to a better understanding of building related data and clustering is instrumental in helping building managers identify time slots with potential energy savings if the HVAC system is configured properly.

198

G. H. Merabet et al.

6 Conclusion The purpose of this chapter was to provide not only an overview of the utility of advanced techniques such as ML (and hence, AI) in controlling smart building services, but also an approach to intelligent spaces designed to improve people’s quality of life and modernize buildings, which included technologies, design techniques, and sustainability concepts. Nowadays, the concept of intelligent building covers not only home automation but also a wide range of fields of knowledge that must be brought together in order for it to be realized. Furthermore, the concept of sustainability should be considered, since it is becoming increasingly demanded by the market, as more and more buildings seek sustainable shapes and models. As technology progresses, advanced technological solutions become more popular, and the majority of people use them. Buildings of the future will incorporate technology that makes daily living easier and improves the quality of the environment, whether it is domestic, commercial, or industrial. In the residential sector, these technologies will not only interact with inhabitants but will also gather data and transmit it to other centers for information sharing and system improvement. The integration will be done in such a way that traffic systems, energy, and other systems, will become intelligent, expanding the Smart City concept. Technology continues to advance in buildings, and today thousands of devices inside an interface and transfer data to a computerized plant that automates processes and improves operational and building management efficiency. As a result, the concept has grown in popularity; numerous trade shows across the world have showcased new features, and many companies believe this market to be one of the most promising in the coming years.

References 1. Woodward A, Smith KR, Campbell-Lendrum D, Chadee DD, Honda Y, Liu Q, Olwoch J, Revich B, Sauerborn R, Chafe Z et al (2014) Climate change and health: on the latest IPCC report. Lancet 383:1185–1189 2. IEEJ Outlook (2018) Prospects and challenges until 2050—Energy, environment and economy. The Institute of Energy Economics, Japan, Tokyo, Japan 3. Ürge-Vorsatz D, Eyre N, Graham P, Harvey D, Hertwich E, Jiang Y, Kornevall C, Majumdar M, McMahon JE, Mirasgedis S, Murakami S, Novikova A, Janda K, Masera O, McNeil M, Petrichenko K, Herrero ST (2012) Energy End-Use: Buildings. Global Energy Assessment: Toward a Sustainable Future 649–760 4. ICT for a Low Carbon Economy—Smart Buildings (2009) European Commission DirectorateGeneral Information Society and Media, Brussels 5. Borgia E (2014) the internet of things vision: key features, applications and open issues. Comput Commun 54:1–31. https://doi.org/10.1016/j.comcom.2014.09.008 6. Halhoul Merabet G, Essaaidi M, Ben Haddou M, Qolomany B, Qadir J, Anan M, Al-Fuqaha A, Abid MR, Benhaddou D (2021) Intelligent building control systems for thermal comfort

8 Towards Advanced Technologies for Smart Building Management …

7. 8.

9.

10.

11. 12. 13.

14.

15. 16.

17. 18. 19.

20. 21. 22. 23.

24. 25. 26.

27.

199

and energy-efficiency: a systematic review of artificial intelligence-assisted techniques. Renew Sustain Energy Rev 144. https://doi.org/10.1016/j.rser.2021.110969 Berardi U, Asdrubali F (2018) Handbook of energy efficiency in buildings—a life cycle approach, 1st edn. Butterworth-Heinemann Facility management and smart buildings: smart data, insights and integration. In: i-SCOOP. https://www.i-scoop.eu/internet-of-things-guide/facility-management-iot-smart-bui ldings/. Accessed 8 Aug 2021 Verbeke S, Ma Y, Tichelen PV, Bogaert S, Uslar M, Schulte J, Bettgenhäuser K, John A, Hermelink A, Offermann M, Groezinger J (2018) Support for setting up a Smart Readiness Indicator for buildings and related impact assessment—Second progress report. European Commission DG Energy Manic M, Wijayasekara D, Amarasinghe K, Rodriguez-Andina JJ (2016) Building energy management systems: the age of intelligent and adaptive buildings. IEEE Ind Electron Mag 10:25–39. https://doi.org/10.1109/MIE.2015.2513749 Smart Buildings Center—Taking building energy efficiency to the next level. In: Smart Buildings Center. https://www.smartbuildingscenter.org/. Accessed 16 Aug 2021 ASHRAE Handbook—Fundamentals (SI Edition) (2009) American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc. (ASHRAE) Different Types of Motion Sensors with Applications (2014) In: Buy Electronics & Electrical Projects in Europe. https://www.efxkits.co.uk/motion-sensors-detectors-with-applicati ons/. Accessed 11 Aug 2021 Weng T, Agarwal Y (2012) From buildings to smart buildings—sensing and actuation to improve energy efficiency. IEEE Des Test Comput 29:36–44. https://doi.org/10.1109/MDT. 2012.2211855 X. Du, X. Qi, C. Wang (2009) Determination of effective energy in buildings. In: 2009 IEEE International Conference on Automation and Logistics. pp 56–61 Erickson VL, Carreira-Perpiñán MA, Cerpa AE (2011) OBSERVE: Occupancy-based system for efficient reduction of HVAC energy. In: Proceedings of the 10th ACM/IEEE international conference on information processing in sensor networks. IEEE, Chicago, IL, USA, pp 258–269 US EPA O (2015) Indoor air quality by building type. https://www.epa.gov/indoor-air-qualityiaq/indoor-air-quality-building-type. Accessed 12 Aug 2021 Ventilation for acceptable indoor air quality—ANSI/ASHRAE Standard 62.1 (2019). ANSI/ASHRAE, Atlanta, GA Heinzerling D, Schiavon S, Webster T, Arens E (2013) Indoor environmental quality assessment models: a literature review and a proposed weighting and classification scheme. Build Environ 70:210–222. https://doi.org/10.1016/j.buildenv.2013.08.027 WHO guidelines for indoor air quality: selected pollutants (2010) World Health Organization. Regional Office for Europe, DK-2100 Copenhagen Ø, Denmark Nicol F, Humphreys M, Roaf S (2012) Adaptive thermal comfort: principles and practice. Routledge Sherman M (1985) A simplified model of thermal comfort. Energy Build 8:37–50. https://doi. org/10.1016/0378-7788(85)90013-1 Gagge AP, Fobelets AP, Berglund LG (1986) A standard predictive index of human response to the thermal environment. In: ASHRAE Trans.; (United States). Research Org.: John B. Pierce Foundation Lab., New Haven, CT American Society of Heating R and A-CEngineers (1993) 1993 ASHRAE handbook : fundamentals. ASHRAE, Atlanta, GA Fanger PO (1972) Thermal comfort: analysis and applications in environmental engineering. McGraw-Hill, New York ISO E, (2005) 7730: 2005 (2005) Ergonomics of the thermal environment—analytical determination and interpretation of thermal comfort using calculation of the PMV and PPD indices and local thermal comfort criteria. Beuth, Berlin Oldewurtel F, Gyalistras D, Gwerder M, Jones CN, Stauch V, Lehmann B, Morari M. Increasing energy efficiency in building climate control using weather forecasts and model predictive control. 8

200

G. H. Merabet et al.

28. Gwerder M, Gyalistras D, Oldewurtel F, Lehmann B, Wirth K, Stauch V, Tödtli J (2010) Potential assessment of rule-based control for integrated room automation. In: 10th REHVA World Congress, “Sustainable energy use in buildings”—Clima 2010. Antalya, Turquia, p 8 29. Dounis AI, Caraiscos C (2009) Advanced control systems engineering for energy and comfort management in a building environment—a review. Renew Sustain Energy Rev 13:1246–1261. https://doi.org/10.1016/j.rser.2008.09.015 30. Abdennadher I, Khabou N, Rodriguez IB, Jmaiel M (2015) Designing energy efficient Smart Buildings in ubiquitous environments. In: 2015 15th International conference on intelligent systems design and applications (ISDA). IEEE, Marrakech, Morocco, pp 122–127 31. Marcus JS (1983) The ‘intelligent’ buildings. The New York Times 00001 32. Guo W, Zhou M (2009) Technologies toward thermal comfort-based and energy-efficient HVAC systems: a review. In: 2009 IEEE International conference on systems, Man and Cybernetics. IEEE, San Antonio, TX, pp 3883–3888 33. Gouda MM, Danaher S, Underwood CP (2001) Thermal comfort based fuzzy logic controller. Build Serv Eng Res Technol 22:237–253. https://doi.org/10.1177/014362440102200403 34. Kolokotsa D, Niachou K, Geros V, Kalaitzakis K, Stavrakakis GS, Santamouris M (2005) Implementation of an integrated indoor environment and energy management system. Energy Build 37:93–99. https://doi.org/10.1016/j.enbuild.2004.05.008 35. Lea RN, Dohmann E, Prebilsky W, Jani Y (1996) An HVAC fuzzy logic zone control system and performance results. In: Proceedings of IEEE 5th International fuzzy systems, vol 3, pp 2175–2180 36. Hamdi M, Lachiver G (1998) A fuzzy control system based on the human sensation of thermal comfort. In: 1998 IEEE International conference on fuzzy systems proceedings. IEEE World congress on computational intelligence (Cat. No.98CH36228). IEEE, Anchorage, AK, USA, pp 487–492 37. Dounis AI, Caraiscos C (2007) Intelligent coordinator of fuzzy controller-agents for indoor environment control in buildings using 3-D fuzzy comfort set. In: 2007 IEEE International Fuzzy Systems Conference. IEEE, London, UK, pp 1–6 38. Yu Z, Zhou Y, Dexter A (2007) Hierarchical fuzzy rule-based control of renewable energy building systems. 6 39. Meana-Llorián D, González García C, Pelayo G-Bustelo BC, Cueva Lovelle JM, GarciaFernandez N, (2017) IoFClime: The fuzzy logic and the Internet of Things to control indoor temperature regarding the outdoor ambient conditions. Futur Gener Comput Syst 76:275–284. https://doi.org/10.1016/j.future.2016.11.020 40. Javaid S, Javaid N, Iqbal S, Mughal MJ (2017) Controlling energy consumption with the worldwide adaptive thermostat using fuzzy inference system in smart grid. In: 2017 International conference on information and communication technology convergence (ICTC). IEEE, pp 66–71 41. Shaikh PH, Mohd Nor NB, Nallagownden P, Elamvazuthi I (2014) Optimized intelligent control system for indoor thermal comfort and energy management of buildings. In: 2014 5th International conference on intelligent and advanced systems (ICIAS). IEEE, Kuala Lumpur, Malaysia, pp 1–5 42. Pombeiro H, Machado MJ, Silva C (2017) Dynamic programming and genetic algorithms to control an HVAC system: maximizing thermal comfort and minimizing cost with PV production and storage. Sustain Cities Soc 34:228–238. https://doi.org/10.1016/j.scs.2017.05.021 43. Wei T, Wang Y, Zhu Q (2017) Deep reinforcement learning for building HVAC control. In: Proceedings of the 54th annual design automation conference 2017—DAC ’17. ACM Press, Austin, TX, USA, pp 1–6 44. Auffenberg F, Snow S, Stein S, Rogers A (2017) A comfort-based approach to smart heating and air conditioning. ACM Trans Intell Syst Technol 9:1–20. https://doi.org/10.1145/3057730 45. Liu W, Lian Z, Zhao B (2007) A neural network evaluation model for individual thermal comfort. Energy Build 39:1115–1122. https://doi.org/10.1016/j.enbuild.2006.12.005 46. Liang J, Du R (2008) Design of intelligent comfort control system with human learning and minimum power control strategies. Energy Convers Manage 49:517–528. https://doi.org/10. 1016/j.enconman.2007.08.006

8 Towards Advanced Technologies for Smart Building Management …

201

47. Ferreira P, Ruano A (2002) Choice of RBF model structure for predicting greenhouse inside air temperature. IFAC Proc Vol 35:91–96 48. Kalogirou SA, Bojic M (2000) Artificial neural networks for the prediction of the energy consumption of a passive solar building. Energy 25:479–491. https://doi.org/10.1016/S03605442(99)00086-9 49. Zaheer-Uddin M, Tudoroiu N (2004) Neuro-PID tracking control of a discharge air temperature system. Energy Convers Manage 45:2405–2415 50. Henze GP, Hindman RE (2002) Control of air-cooled chiller condenser fans using clustering neural networks. ASHRAE 108:232–244 51. Yang I-H, Yeo M-S, Kim K-W (2003) Application of artificial neural network to predict the optimal start time for heating system in building. Energy Convers Manage 44:2791–2809. https://doi.org/10.1016/S0196-8904(03)00044-X 52. Magnier L, Haghighat F (2010) Multiobjective optimization of building design using TRNSYS simulations, genetic algorithm, and artificial neural network. Build Environ 45:739–746. https:// doi.org/10.1016/j.buildenv.2009.08.016 53. Ruano AE, Ferreira PM, Mendes H (2010) MOGA design of temperature and relative humidity models for predictive thermal comfort. IFAC Proc Vol 43:116–121. https://doi.org/10.3182/201 00329-3-PT-3006.00023 54. Marantos C, Lamprakos CP, Tsoutsouras V, Siozios K, Soudris D (2018) Towards plug&play smart thermostats inspired by reinforcement learning. In: Proceedings of the workshop on intelligent embedded systems architectures and applications—INTESA ’18. ACM Press, Turin, Italy, pp 39–44 55. Zhang Z, Lam KP (2018) Practical implementation and evaluation of deep reinforcement learning control for a radiant heating system. In: Proceedings of the 5th conference on systems for built environments—BuildSys ’18. ACM Press, Shenzen, China, pp 148–157 56. Pedrycz W (2005) From granular computing to computational intelligence and human-centric systems. IEEE Connect 3:6–11 57. Van Breemen A, de Vries TJ (2001) Design and implementation of a room thermostat using an agent-based approach. Control Eng Pract 9:233–248 58. Wang Z, Wang L, Dounis AI, Yang R (2012) Multi-agent control system with information fusion based comfort model for smart buildings. Appl Energy 99:247–254. https://doi.org/10. 1016/j.apenergy.2012.05.020 59. Wang Z, Yang R, Wang L (2011) Intelligent multi-agent control for integrated building and micro-grid systems. In: ISGT 2011. IEEE, Anaheim, CA, USA, pp 1–7 60. Wang Z, Yang R, Wang L (2010) Multi-agent control system with intelligent optimization for smart and energy-efficient buildings. In: IECON 2010 - 36th Annual conference on IEEE industrial electronics society. IEEE, Glendale, AZ, USA, pp 1144–1149 61. Merabet GH, Essaaidi M, El Brak M, Benhaddou D (2017) Agent based for comfort control in smart building. In: 2017 International renewable and sustainable energy conference (IRSEC). IEEE, pp 1–4 62. Shaikh PH, Nor NMohd, Nallagownden P, Elamvazuthi I, (2013) Intelligent optimized control system for energy and comfort management in efficient and sustainable buildings. Proc Technol 11:99–106. https://doi.org/10.1016/j.protcy.2013.12.167 63. Minar N, Gray M, Roup O, Krikorian R, Maes P (1999) Hive: Distributed agents for networking things. In: Proceedings. First and third international symposium on agent systems applications, and mobile agents. IEEE, pp 118–129 64. Clarke G, Colley M (2002) lncremental synchronous learning for embedded agents operating in ubiquitous computing environments. Soft Comput Agents New Perspect Dyn Inf Syst 83:25 65. Kan EM, Yadanar K, Ling NH, Soh Y, Lin N (2015) Multi-agent control system with intelligent optimization for building energy management. In: Proceedings of the 18th Asia Pacific Symposium on intelligent and evolutionary systems, vol 2. Springer, pp 505–518 66. Barakat M, Khoury H (2016) An agent-based framework to study occupant multi-comfort level in office buildings. In: 2016 Winter simulation conference (WSC). IEEE, Washington, DC, USA, pp 1328–1339

202

G. H. Merabet et al.

67. Alfakara A, Croxford B (2014) Using agent-based modelling to simulate occupants’ behaviours in response to summer overheating. In: Proceedings of the symposium on simulation for architecture & urban design. Society for Computer Simulation International, San Diego, CA, USA 68. Naji HR, Meybodi MN, Falatouri T (2011) Intelligent building management systems using multi agents: Fuzzy approach. Int J Comput Appl 14:9–14

Chapter 9

Applications to Building Services with the Local Energy Community Minghao Xu, Furong Li, Chenghong Gu, Kang Ma, Renjie Wei, Junlong Li, and Andrew Shea

1 Introduction The energy consumption from buildings accounted for 30% of the global energy consumption in 2019 and is forecasted to increase by 50% in the next 30 years [1, 2], posing a tremendous challenge on the global efforts towards Net-Zero emissions. Electrification and intellectualisation of buildings through the installation of smart and low-carbon devices, such as heat pumps (HPs), electric vehicles (EVs) and other intelligent appliances, are expected to help address the challenge. In the UK, National Grid ESO predicts that the stock of EVs nationwide could reach as high as 36 million by 2040 [3], and the domestic heat pumps could reach 2.5 million by 2030 [4]. The worldwide domestic Wi-Fi devices are expected to reach 17 billion units by 2030 [5]. With the potential move towards local energy communities (LECs) and the integration of smart and low-carbon technologies, how to provide energy-efficiency, cost and carbon saving and user-friendly building services have aroused great concern. In this Chapter, Sect. 2 discusses (1) the dominant needs of buildings, (2) services to meet the needs and (3) devices utilised to provide these services. In Sect. 3, three case studies are used to demonstrate the potentials and benefits of utilising intelligent control to manage building services. Section 4 concludes the chapter and provides an outlook to the future.

M. Xu · F. Li (B) · C. Gu · K. Ma · R. Wei · J. Li · A. Shea Faculty of Engineering and Design, University of Bath, Bath, UK e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 A. Tomar et al. (eds.), Control of Smart Buildings, Studies in Infrastructure and Control, https://doi.org/10.1007/978-981-19-0375-5_9

203

204

M. Xu et al.

2 Basic Concepts of Building Services 2.1 Dominant Needs of Buildings and Building Services Various needs of buildings have to be met and optimised, and they are not all energyrelated. For example, security is an essential need of smart buildings, involving fire detection, intrusion detection and access control [6]. This chapter only focuses on the energy management system (EMS) in smart buildings, and four dominant energy-related needs of smart buildings have been summarised in Fig. 1. Firstly, Sufficient Electricity is a basic need of smart buildings. All the appliances, digital devices, electrified equipment and control systems are supplied by electricity. Thus, sufficient electricity is the top priority which guarantees the normal operations of smart buildings. Secondly, people inside smart buildings require a Comfortable Environment, including proper temperature and humidity. This needs to rely on the intelligence and efficiency of heating, ventilation and air conditioning (HVAC) systems. Thirdly, for energy consumers in smart buildings, the aim of investing an EMS is to save energy costs and improve user experience, which leads to the last two needs, Energy Cost Savings and Minimal Manual Control. To meet the building needs in an optimised manner, the controllable building services need to coordinate with each other to respond to energy prices and clean energy generation. To meet the dominant needs of the buildings, the services below have been identified and can be provided by smart buildings [7–9]: • HVAC Control: The HVAC control directly determines the environment comfort level of the inside smart buildings. With an intelligent HVAC control system, people can enjoy the proper temperature with less energy cost. In addition, predictive maintenance approaches in the HVAC control system can save the inconvenience of manual works as much as possible [9]. • Lighting Control: With sensors, such as photosensitive sensors and motion sensors, the lights inside the buildings can be automatically turned on/off. In this way, the energy cost and manual works can be largely saved. • EV Economic Charging/Discharging: EVs are gradually replacing the traditional fossil-fuelled vehicles, which requires more EV charging stations to be installed as part of the building services. On the one hand, this is expected to help society move towards carbon neutrality. On the other, the EV charging cost

Sufficient Electricity

Energy Cost Saving

Comfortable Environment

Minimal Manual Control

Fig. 1 Dominant energy-related needs of smart buildings

9 Applications to Building Services with the Local Energy Community

205

is likely to constitute a significant part of the total energy bills for the adopters. Thus, EV economic charging services and Vehicle-to-Grid (V2G) or Vehicle-to-X (V2X) charging (where “X” means anything, such as a building, a battery or the grid) will be critical for smart buildings to save energy costs. • General Energy Management: General EMS is essential as it needs to provide sufficient electricity while keeping the energy cost as low as possible.

2.2 Future Smart Building Technologies To provide these services, intelligent coordination and control between energyrelated devices and systems are required. With the accelerating electrification and intellectualisation of smart buildings, various energy-related intelligent devices and systems have been wildly deployed into smart buildings. Some of the core devices and systems are presented in Fig. 2, which can be divided into 3 types according to their relations with energy: (1) Distributed Generation: clean energy resources; (2) Storable Loads: energy storage system (ESS) and EV charging station; (3) Shiftable & Flexible Loads: Controllable domestic appliances, lighting system and HVAC system. The smart meter and energy hub are the central controls to coordinate the above devices and subsystems to provide the smart-building services. Among these energy-related intelligent devices, EV and HP (a core device in HVAC systems) are expected to be widely installed worldwide. The benefits of the

Smart Building Clean Energy Resource HVAC System

Lighting System

Energy Storage System

Domestic appliances EV Charging Station Smart Meter / Energy Hub

Fig. 2 Energy-related intelligent devices and systems in smart buildings

206

M. Xu et al.

introduction of them are twofold: (1) reducing carbon emissions by electrifying the traditional transportation and heating/cooling system and (2) providing flexibilities on the demand side [10]. Global EV sales have escalated from less than 10,000 in 2010 to 774,000 in 2016, surpassing 2 million cumulative sales [11]. This brings increasing needs for EV charging services to smart buildings in the future. While challenging, the benefits are clear too. As storable loads, V2G or V2X enabled EVs can be utilised in demandside response within LECs to reduce peak load and energy costs by responding to energy price variations and local clean energy generation curves [12]. As for HPs, compared with conventional HVAC systems, they are typically more efficient, hence come with lower energy costs [13]. In 2015, about 800,000 HP units were sold in Europe and a growing trend in sales is estimated for the coming years [14]. In addition to its high efficiency, within the comfortable temperature range, intelligence-enabled HP can adjust its power according to the energy price variations and local clean energy generation curves, which will further save energy costs [15]. In the following sections of the Chapter, several case studies will demonstrate the capabilities and benefits of EVs and heat pumps as the core building services.

3 Methodology 3.1 Overview of Methodology To achieve optimal energy usage in the smart home/building, an energy management system (EMS) is developed. The optimal control aims to minimise the cost of electricity. This is achieved through managing the building services, such as EV charging or discharging, which, in our model, are incentivised by (1) TOU tariffs of electricity from the main grid; and (2) low-price PV electricity that is generated at LEC and can be purchased through P2P trading. An overview of the EMS and its optimal control is shown in Fig. 3. As illustrated, the EMS will collect information as input data, such as the local PV generation and price from the P2P market, EV states and TOU tariffs. Considering the device physical constraints and local system constraints, the EMS will guide the electricity purchasing and the EV charging and discharging to minimise the electricity cost.

3.2 Modelling of Building Needs and Services The core needs and services in our in our proposed system/building environment are introduced below, along with their mathematical modelling.

9 Applications to Building Services with the Local Energy Community

207

Fig. 3 Overview of the EMS and the optimal control

A Power Demand of the Building The power consumption of the building is divided into two different parts: power for heating and the baseload power, which is given by Q demand,t = (P baseload,t + Pheating,t ) ∗ t The baseload power means the power consumption except the heating and the EV charging, e.g. lights, fans and televisions. The power for heating depends on the consumer heating mode. If the consumer chooses the gas heating mode, the power for heating is regarded as a part of baseload power. If the consumer chooses the electricity heating mode, the power for heating is isolated with the baseload power. B PV Energy Output Model Combined with the LEC, for the future smart building, PV energy is one of the important energy sources besides traditional fuel fossil energy. In this chapter, for the PV model, it is assumed that during each time slot the PV output power is stable and the smart building uses the average value of the output power for a certain time slot. C EV Model In this section, an EV model is presented, which models the control decisions of the EVs (e.g. the charging power and the discharging power), the EV states and their transitions over time. For EV i, the arriving time, the departure time and the initial state are uncertain variables. In this book, these uncertain variables are gathered from historical data and survey. For any time slot t, St denotes the state of the EV at stage t, given by

208

M. Xu et al.

  S Ev,t = qt , SoC t , Pcharge,t , Pdischarge,t where qt is a discrete state: qt = 1 means that the EV is charging at time t and the charging power is Pcharge,t ; qt = −1 means that the EV is discharging the energy stored in the battery to the building and the discharge power is Pdischarge,t ; qt = 0 means that the EV keeps idling at time t. SoC t means the state of charge (SoC) of the EV at time t. However, for the EV, considering the battery energy loss, the battery charging and discharging are not 100% efficient process. The efficiency of charging is ηcharge and the efficiency of discharging is ηdischarge . The charging and discharging process of the battery are given by Q charge,t = ηcharge ∗ Pcharge,t ∗ t Q discharge,t = Pdischarge,t ∗ t/ηdischarge D Heat Pump Model In the future smart grid, to achieve Net-Zero emission, the heat pump is considered to replace the traditional heating devices to decarbonise the energy supply. In this paper, the heat pump is regarded as a choice to provide heating for consumers. The state of the heat pump is given by   Sheat pump,t = COP, Q heat pump,t , Pheat pump,t where COP is the coefficient factor of performance, Q heat pump,t means the heating energy the heat pump needs to transfer at time t and Pheat pump,t means the power of the heat pump at time t.

3.3 Problem Formulation of the Proposed EMS The implementation of the EMS is formulated as an optimisation problem with the following objective and constraints. A Objective Function The objective function of the smart building model is the minimisation of electricity cost which includes two parts: the electricity purchased from the grid using TOU tariff and the electricity generated by local PV at an agreed price via P2P trading. The objective function for this model is thus given by

9 Applications to Building Services with the Local Energy Community

min f =

N 

Q grid,t ∗ pgrid,t +

t=1

N 

209

Q P V,t ∗ p P V,t

t=1

where Q grid,t is the electricity delivered from the grid and Q P V,t is the electricity delivered from the PV generation. pgrid,t denotes the price of the electricity from the grid at time t and p P V,t means the price of the electricity from the PV at time t. B Constraints: The following constraints are identified in our problem and should be satisfied: (1)

Power balance constraint: The power consumption should be balanced for every time slot. All the purchased electrical energy is used for three different parts: the baseload power consumption, the heat power consumption and the EV charging. Also, the EV can discharge to supply the building demand. The energy transfer function is given by N 

Q grid,t +

t=1

+

N 

Q P V,t =

t=1 N 

N  t=1

Q baseload,t +

N 

Q heat,t +

t=1

N 

qt ∗ Q charge,t

t=1

qt ∗ Q discharge,t

t=1

(2)

where qt is the state variable mentioned in Sect. 3.2 (C). Q baseload,t denotes the base load consumption and Q heat,t denotes the heat consumption. Q charge,t and Q discharge,t is mentioned in Sect. 3.2 (C). EV final SoC state constraints In this paper, it is assumed that although the initial SoC of the EV is uncertain, the final SoC before leaving the smart building should be full state. It is given by SoC t ( f inal) = 1

(3)

EV battery charging/discharging power constraints The charging/discharging power of the EV should be lower than the maximum value of the charging/discharging power. The two maximum power are defined by the EV specification ∀t, Pcharge,t < Pchmax ∀t, Pdischarge,t < Pdischmax

210

(4)

M. Xu et al.

EV SoC requirements The EV SoC can only vary from the lower bound to its upper bound to avoid over charging or over discharging. ∀t, SoC min < SoC t < SoC max

C Optimal Control Strategy Based on the smart building model formulated as above, the optimal strategy can be solved by utilising mixed-integer linear programming (MILP). The details of MILP will not be discussed here due to its irrelevance to this chapter’s main goals.

4 Case Studies This section provides case studies to demonstrate how different technologies and building services can be utilised to save consumer energy bills and to support LEC. Three case studies will be given: 1. 2. 3.

Benchmark: conventional homes without any smart building services nor control. EV homes: V2X-enabled EV homes, incentivised by Time of Use (TOU) tariff and local P2P trading. Heat pump and EV adopters: homes adopting V2X-enabled EVs and heat pumps, incentivised by TOU tariff and local P2P trading.

4.1 Traditional Homes Demand characteristics, e.g. peak demand and time of peak demand, are traditionally unavailable until the recent roll-out of smart meters. The attained visibility on household’s demand characteristics shows great volatility for individual consumers and variability across different consumers. Better understanding and utilisation of these characteristics would help consumers save energy bills and choose better tariffs. In this case study, real consumer demand data from London, England, will be segmented and presented [16]. Typical consumers from the segmented groups will be selected to implement the proposed EMS. The data set is in half-hourly resolution and consists of over 1,000 domestic consumers’ smart metering data for a year. Figure 4 is an illustration of what do their normalised daily average demand curves look like. Each of the coloured line in Fig. 4 represents one consumer’s daily average demand, which has been normalised by their peak value.

9 Applications to Building Services with the Local Energy Community

211

Fig. 4 Normalised daily average demand of all consumers

To reveal the underlying demand patterns, a generic X-means clustering is conducted on the data, where the consumers are divided into different groups based on the similarity of their normalised demand profile shape. Four groups of consumers have been identified and plotted in Fig. 5, where the black line represents the centroid of each group (group average) and the coloured lines represent different consumers’ demand within each group. The first and the third groups/clusters represent general consumers who have a small demand peak in the morning and a larger peak in the evening after work. The difference between the two is that consumers in Group 1 have a slightly larger peak/trough demand ratio and a higher evening peak demand than those in Group 3. Group 2 represents households that consume energy at a relatively constant rate across that day. Group 4 represents consumers who are active in energy consumption from midnight to 3:00 a.m. in the early morning. From each of the four groups, one consumer is selected as the typical consumer of the corresponding group. Seven consecutive days’ load profiles for all of the four typical consumers are given in Figs. 6, 7, 8 and 9, representing consumer Group 1, Group 2, Group 3 and Group 4, respectively. In contrast to residential consumers, commercial electricity consumers generally consume more energy. The load profile shown in Fig. 10 is sampled from a small commercial consumer from Bristol, UK [17]. Same as the residential consumers’ example load profiles, the commercial load profile is also in half-hourly resolution and spans over seven consecutive days.

212

Fig. 5 Plots of all four groups of consumers

Fig. 6 Load profile of one typical consumer of Group 1 from 1 to 7 January

M. Xu et al.

9 Applications to Building Services with the Local Energy Community

Fig. 7 Load profile of one typical consumer of Group 2 from 1 to 7 January

Fig. 8 Load profile of one typical Consumer of Group 3 from 1 to 7 January

213

214

Fig. 9 Load profile of one typical consumer of Group 4 from 1 to 7 January

Fig. 10 Load profile of one commercial consumer from 1 to 7 February

M. Xu et al.

9 Applications to Building Services with the Local Energy Community

215

Fig. 11 Overview structure of smart home/building with EV

4.2 EV Connected Homes A Overview It is estimated that a total number of 49.0 million EVs will be needed to meet the UK’s 2050 Net Zero target [18]. The UK government is strongly supporting this, announcing that all new petrol and diesel cars and vans will be phased out by 2030. EVs are expected to be an integral part of future homes and buildings. The structure of the designed smart home/building with EV is shown in Fig. 11. It will consist of the following: V2X-enabled EV, PV power generation from LEC, which can be purchased via P2P trading at a relatively low price, and loads. With the help of EV, local PV output can be better used to fulfil high demands during peak periods; in this way, consumers can take advantage of TOU tariffs to save their electricity bills and to save their carbon footprint. To achieve optimal energy usage in the smart home, an energy management system (EMS) is developed, as formulated in Sect. 3, which takes the input data of forecasted consumer load data, forecasted available PV output and price at the P2P market in LEC and TOU tariffs.

216

M. Xu et al.

Fig. 12 TOU tariff

The developed EMS is validated against the typical consumer of Group 1, Group 2, Group 3 and Group 4 and the commercial consumer, which are introduced in Sect. 3.1. The assumptions in our modelling are summarised as follows: 1. 2.

3.

4. 5. 6. 7.

EV battery capacity is 50 kWh for both the residential consumers and the commercial consumers. Both the residential and the commercial consumers adopt TOU tariffs to purchase electricity from the grid. The adopted TOU curve, as shown in Fig. 12, is proposed by [19]. In this book, it has been rescaled to match the current electricity price in the UK market. Both the residential consumer and the commercial consumer participate in the local P2P market, where they can directly purchase electricity generated by PV within LEC at a fixed price of £0.07/kWh. The available PV output in the P2P market is assumed to be known by the consumer for simplicity and is plotted in Fig. 13. The typical consumer of Group 1 and Group 4 are assumed to work during the day and would only connect their EVs to the home after work in the evening. The typical consumer of Group 2 is assumed to stay at home most of the day and their EV is hence connected to the home across most of the day. The typical consumer of Group 3 is assumed to work during night and their EV is connected to the home during daytime. The commercial consumer would only connect their EV to the building at work during daytime.

9 Applications to Building Services with the Local Energy Community

217

Fig. 13 PV output in the P2P market

B Results The optimal EV charging/discharging curves and the resultant demand and PV curves for the typical consumers of Group 1, Group 2, Group 3, Group 4 and the selected commercial consumer are stack plotted in Figs. 14, 15, 16, 17 and 18, respectively. In Fig. 14a, the coloured area above the x axis means the EV is charging and vice versa. Different colours specify the sources of the charged electricity and the end use of the discharged electricity. More specifically, red means the electricity is charged from the grid, whereas green means it is from the local PV generation. As for the yellow area, which is always below 0 in the figure, means the EV is discharging to supply the consumer’s demand. The blue line depicts the SoC variations of the EV across the day. It should be noted that there are times where the EV SoC drops to 0. This is only due to that EVs are away from home and the SoC is unknown and does not reflect the actual SoC of the EV during these periods. By comparing Fig. 14a with Fig. 12, it can be observed that the EMS schedules the EV to charge excess electricity during low-price times in the early morning and around midnight and uses the stored electricity to supply demand during night while the tariff is high. Figure 14b shows how the demand changes over the day and the compositions of the demand during each period, i.e. whether if the demand is supplied by the grid, local PV or EV through discharging. Different electricity sources are assigned with different colours, i.e. red represents the portions of demand supplied by the grid; green represents the portions of electricity supplied by PV generation; and yellow accounts for the electricity powered by the EV. In addition, the hatched area within the red and yellow areas represents the power that is used for charging EVs. The net

218

M. Xu et al.

Fig. 14 EMS results for Group 1 typical consumer

demand seen by the grid is depicted by the black dotted line in the figure. As can be seen, for consumer 1, though the EV is not at home during daytime to utilise the cheap PV generation for charging, a great portion of the PV generation is consumed by the consumer, nonetheless. The reason for this is that the consumer’s demand during the day matches the PV generation fairly well, particularly in the afternoon, in terms of the consumption patterns and magnitudes. Figure 14c illustrates how the local PV generation is utilised. The black line shows the total available PV generation in LEC during each half-hour period across the day. The end uses of the PV generation are differentiated by varying colours in the plot. Specifically, the green area in the stack plot shows how much PV power is used to charge the EV, and the yellow area represents the PV power that is used to supply the non-EV demand at home. Lastly, for the red area, it shows how much PV power is unused. The blue dotted line shows how much PV power has been used in total. For this consumer, it can be observed that about half of the PV generation in the morning is not used due to the absence of EV and the low demand requirement of the consumer during that period. For the typical consumer for Group 2, as shown in Fig. 15, the household consumes energy at a relatively constant level and the EV is connected to the home across the

9 Applications to Building Services with the Local Energy Community

219

Fig. 15 EMS results for Group 2 typical consumer

whole day. The increased availability of EV at home greatly reduces the power reliance on the grid. It can be observed that the EV is charged using the cheap PV electricity during the day and is discharged, using the stored the energy, to supply the consumer while the price of electricity is high and the PV is not generating power. Figure 16 shows the EMS results for the typical consumer for Group 3. The consumer is on the night shift and the EV is only connected to the home during daytime. We can see that the EV is constantly charging during the daytime, from the PV generation and the grid when the TOU is low. It is until around 4:00 p.m., when the TOU is the highest during the day, that the EV starts to supply the household demand by discharging its battery. As for the last residential consumer, the typical consumer for Group 4, the consumer is on the day shift and the EV is only connected to the home at night, after 7:00 p.m., as shown in Fig. 18. The working pattern of this consumer is similar to the Group 1 typical consumer. However, the energy consumption patterns of the two are drastically different. The Group 4 consumer consumes minimal demand during the day and actively consumes energy after getting home, even after midnight in the early morning. For this consumer, the majority of the matched PV generation

220

M. Xu et al.

Fig. 16 EMS results for Group 3 typical consumer

in LEC is unused and wasted as the demand during daytime is low and the EV is not at home to utilise the cheap PV generation for charging. One common issue for all the typical residential consumers, except for the typical consumer for Group 3, is the newly created peak demand due to EV charging, which is at least eightfold the original peak demand. This is because the EV can only be charged during certain period of the day. And when the electricity price is low, the EV would charge at a high-power rate to take in as much energy as possible. Figure 18 shows the EMS results for the commercial consumer, which are quite similar to the results for the typical consumer of Group 3. The PV generation is well used for charging the EV and supplying the building demand. During the testing period, 86.46%, 66.82%, 99.98%, 55.45% and 100% of the total available PV output has been locally consumed by the typical consumers for Group 1, Group 2, Group 3, Group 4 and the commercial consumer, respectively. This leads to carbon emission saving of about 0.83 kg each day or 303.24 kg for a whole year for consumer 1; about 0.64 kg each day or 234.35 kg for a whole year for consumer 2; about 0.96 kg each day or 350.65 kg for a whole year for consumer 3; about 0.53 kg each day or 194.46 kg for a whole year for consumer 4; and lastly, about 0.961 kg each day or 350.72 kg for a whole year for the commercial consumer.

9 Applications to Building Services with the Local Energy Community

221

Fig. 17 EMS results for Group 4 typical consumer

4.3 Heat Pump and EV Adopters A Overview In the UK, heating contributes to about 30% of the UK’s total greenhouse emissions with half of this from heating homes. As the country sets its plan to achieve Net Zero by 2050, the UK Government announced a ban on gas boilers in newly built homes from 2025. This accelerated the consensus that heat pumps will be a major source of clean heat in the UK. In this section, the EV home in Sect. 4.2 is further extended to include heat pumps in the buildings. The structure of the extended smart home/building with EV and heat pump is shown in Fig. 19. It’s very similar to the structure in Fig. 11, except that heat pumps are introduced as the alternative heating technology to gas boilers. The developed EMS considering heat pumps is validated against the same consumers as in Sect. 4.2, i.e. the typical consumers of Group 1, Group 2, Group 3, Group 4 and the commercial consumer. The modelling follows the same assumptions as introduced in Sect. 4.1 A, while some new assumptions are made as well: 1.

The coefficient of performance (COP) of the installed heat pump is 3.5.

222

Fig. 18 EMS results for commercial consumer

Fig. 19 Overview structure of smart home/building with EV and heat pump

M. Xu et al.

9 Applications to Building Services with the Local Energy Community

223

Fig. 20 Residential consumer heat pump demand

2.

The residential consumer follows the heat pump demand profile as shown in Fig. 20, while the commercial consumer follows Fig. 21.

B Results The objective of the EMS remains to optimise the power usage with minimum electricity cost under TOU tariffs and P2P trading. The difference lies in that the deployment of heat pumps into the buildings increases the electric demand as the original heating demand is electrified. The increased electricity demand will also increase the consumption of local PV generation which will be demonstrated in the following figures. The optimal EV charging and discharging curves and the SoC curve for the typical consumers for Group 1, Group 2, Group 3, Group 4 and the commercial consumer are plotted in Figs. 22, 23, 24, 25 and 26 respectively. The plots follow the same format as those in Sect. 4.2 B. By comparing these plots with the results before the heating is electrified with heat pumps in the previous section, it can be noted that the charging and discharging behaviours of the EVs follow very similar patterns for the same consumer, before and after installing heat pumps. However, the increased electricity demand due to introduction of heat pumps, increases the consumption of the local PV generation as the consumers have more demand to be supplied by the PV generation particular when the EV is not at home. During the testing period, 98.14%, 100%, 100%, 92.46% and 100% of the total available PV output has been locally consumed by the typical consumers for Group 1, Group 2, Group 3, Group 4 and the commercial consumer respectively. This leads

224

M. Xu et al.

Fig. 21 Commercial consumer heat pump demand

to carbon emission saving of about 0.94 kg each day or 344.18 kg for a whole year for consumer 1; about 0.961 kg each day or 350.72 kg for a whole year for consumer 2, consumer 3 and the commercial consumer; and about 0.89 kg each day or 324.26 kg for a whole year for consumer 4. The PV consumption by consumer 1, consumer 2 and consumer 4 has been greatly increased due to the electrified heating. While no such improvement is seen for consumer 3 and the commercial consumer, as the PV generation has already been nearly 100% consumed locally before the introduction of heat pumps.

9 Applications to Building Services with the Local Energy Community

225

Fig. 22 EMS results for Group 1 typical consumer with heat pump

5 Conclusions and an Outlook to the Future 5.1 Conclusions This chapter provides representative yet concrete examples to show the capability and potential, at the micro level, of smart building services in managing building energy, EVs and heat pumps in the context of LEC. Though the proposed EMS is greatly simplified, the core attributes of the system and technologies are all captured. The results show that the adoption of EVs and heat pumps, enabled by intelligent control, could help consume LEC PVs, save consumer energy bills and also carbon emissions. The results also highlight that the performance of the combined technology and intelligence can be greatly affected by consumers’ energy behaviours and EV driving behaviours.

226

M. Xu et al.

Fig. 23 EMS results for Group 2 typical consumer with heat pump

5.2 Outlook to the Future Accounting for a great proportion of the total energy consumption, buildings are expected to be more intelligent and electrified to reduce their carbon emissions to achieve Net Zero by 2050. As a viable solution for electrification in buildings, transportation and LECs, EV and HP have received increasing attention and are being widely promoted worldwide. In the UK, there are over 300,000 EVs at the end of May 2021, and more than 600,000 plug-in models if including plug-in hybrids (PHEVs). To meet the Net-Zero target, the number of EVs is estimated to reach 36 million by 2040, then 49 million by 2050 [4]. This significant growth is also expected in the installation of HPs. With around 95% of the heat pump manufacturing market share, the heat pump association estimate that about 67,000 HP units have been delivered by their supply chains in 2021. The UK government also plans to install 600,000 heat pumps a year by 2028. In Future Energy Scenarios (FES), the number of domestic heat pumps is expected to reach 2.5 million by 2030 [4]. To realise Net Zero in 2050, 28% of the district heating

9 Applications to Building Services with the Local Energy Community

227

Fig. 24 EMS results for Group 3 typical consumer with heat pump

sources in the UK is expected to be provided through heat pumps (as compared to 5% today). This dramatic increase of EVs and HPs leads to gigantic electricity demand, which brings a serious challenge to the power system. EVs could consume approximately 65–100TWh electricity in 2050, when the system-wide demand is forecasted to be 600–900TWh. As for HPs, the UK government’s heat pump sales target could add 14 TWh to electricity demand by 2030. These rapidly increasing electricity demands lead to increasing pressure for the network operators and generators. Topdown regimes that heavily rely on investment are not always cost-effective and may slow down the uptake of new technologies. Mechanisms that instead incentivise bottom-up optimisation and coordination from buildings, LECs to cities and nations offer an alternative path to low-carbon transition and realising the Net Zero.

228

Fig. 25 EMS results for Group 4 typical consumer with heat pump

M. Xu et al.

9 Applications to Building Services with the Local Energy Community

229

Fig. 26 EMS results for commercial consumer with heat pump

References 1. Yu L, Qin S, Zhang M, Shen C, Jiang T, Guan X (2021) A review of deep reinforcement learning for smart building energy management. IEEE Internet Things J 2. The Global Alliance for Buildings and Construction (GABC) (2020) The Global Status Report 2020 3. Future Energy Scenarios (2018) National Grid 4. Future Energy Scenarios (2019) National Grid ESO 5. Mercer D (2019) The wireless home: assessing the scale of the global home Wi-Fi device market 6. CONSTRUCTIONS—SMART BUILDING MANAGEMENT SOLUTIONS. http://www.ele ctrocombd.com/construction.php 7. Serving People. Building Technology. https://www.sbsmi.com/ 8. Smart Building Energy Solutions. https://www.acesolutionsgroup.ca/services.html 9. Efficiency is Everything. https://macmiller.com/services/optimize/ 10. Ghazvini MAF, Lipari G, Pau M, Ponci F, Monti A, Soares J, Castro R, Vale Z (2019) Congestion management in active distribution networks through demand response implementation. Sustain Energ Grids Networks 17:100185 11. Global Plug-in Sales for 2016 (2017). http://www.ev-volumes.com/news/global-plug-in-salesfor-2016/

230

M. Xu et al.

12. Gomez-Gonzalez M, Hernandez JC, Vera D, Jurado F (2020) Optimal sizing and power schedule in PV household-prosumers for improving PV self-consumption and providing frequency containment reserve. Energy 191:116554 13. de Tapia CD, Harper CM (2021) Comparing geothermal heat pump systems to conventional HVAC systems in hot and humid climates. J Civ Eng Constr 10(2):84–100 14. Thomas NP (2015) European heat pump market and statistics report 2015, Technical report for The European Heat Pump Association AISBL (EHPA) 15. Bee E, Prada A, Baggio P (2018) Demand-side management of air-source heat pump and photovoltaic systems for heating applications in the italian context. Environments 5(12):132 16. Schofield JR, Carmichael R, Tindemans S, Bilton M, Woolf M, Strbac G (2015) Low carbon London project: data from the dynamic time-of-use electricity pricing trial, 2013. UK Data Service, SN 7857 (2015):7857–7851 17. Kaushik S, Dale M, Aggarwal R, Smyth A, Redfern M, Waite K (2014) Project SoLa BRISTOL migration from” ecohome” to” integrated homes”. In: 2014 49th International Universities Power Engineering Conference (UPEC), IEEE, pp 1–5 18. Wills T (2020) The UK’s transition to electric vehicles. Clim Change Committee 19. Zhao C, Dong S, Gu C, Li F, Song Y, Padhy NP (2018) New problem formulation for optimal demand side response in hybrid AC/DC systems. IEEE Trans Smart Grid 9(4):3154–3165. https://doi.org/10.1109/tsg.2016.2628040

Chapter 10

Optimization in Grid-Interactive Buildings Xiaolong Jin, Xiaodan Yu, Yihan Lu, Hongjie Jia, and Yunfei Mu

Abstract This chapter proposes an optimal scheduling approach for multi-energy grids (MEGs) integrated with the grid-interactive buildings. To optimally coordinate the grid-interactive buildings and the MEGs, this chapter develops a bi-level optimization method. The MEGs operator is able to activate the heating demand response (HDR) from grid-interactive buildings, while the grid-interactive buildings can enjoy heating cost saving with the proposed bi-level optimization method. At the upper level, the MEG operator optimizes the heating sale price (HSP) to the buildings and the energy schedules by dispatching the multi-energy devices. In the lower level, in order to reduce consumers’ heating costs, indoor radiators’ water flow rates are optimally adjusted according to the HSPs. To efficiently solve the bi-level optimization problem, we reformulate the original problem as a mixed-integer linear programming (MILP). We also use the piecewise linearization method to treat the nonlinearity of constraints of the heating distribution network. Case study results demonstrate that the proposed method can optimally coordinate the grid-interactive buildings and the MEGs. Consequently, the flexibility of the grid-interactive buildings can be fully used in the optimization of the multi-energy grids. Moreover, the heating costs of the buildings can be significantly reduced with the proposed bi-level optimization method. Keywords Bi-level optimization · Grid-interactive buildings · Optimal heating sale price · Multi-energy grids · Thermal dynamics

1 Introduction Building’s energy consumption accounts for roughly 40% of the global energy consumption [1]. Therefore, energy efficiency improvement in buildings is becoming more and more significant. Nowadays, the multi-energy grid (MEG) [2, 3] is widely employed to provide multiple energy (e.g., electricity and heating) for buildings. The X. Jin (B) · X. Yu · Y. Lu · H. Jia · Y. Mu Key Laboratory of Smart Grid of Ministry of Education, Tianjin University, Tianjin 300072, China e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 A. Tomar et al. (eds.), Control of Smart Buildings, Studies in Infrastructure and Control, https://doi.org/10.1007/978-981-19-0375-5_10

231

232

X. Jin et al.

MEG operator can optimize the energy schedules by dispatching the multi-energy devices [4]. In this way, the buildings can enjoy economic and efficient energy utilization solutions provided by the MEG operator. Thanks to buildings’ thermal inertia, consumers in buildings are able to optimize their indoor temperatures dynamically within their comfort ranges. Therefore, there is flexibility in buildings’ heating loads, which enable consumers in buildings to participate in the heating demand response (HDR) program triggered by the MEG [5]. In this context, buildings to energy grid (B2EG) integration is beneficial for both the consumers and the MEG operator. Many studies about the optimal scheduling methods for B2EG have been carried out. References [6, 7] model buildings as thermal storages and involve the buildingbased thermal storages for the optimization of the heating networks. The studies demonstrate that building-based thermal storages can enlarge the heating production range of the combined heat and power (CHP) units. An optimal dispatch method for grid-interactive buildings in the MEGs is proposed in [8]. The results show that the flexibility of buildings due to the thermal dynamics can further increase the operational flexibility of the MEGs. An optimal scheduling approach for an office building integrated electricity distribution system is developed in [9]. The proposed method can minimize the electricity cost of the office buildings and enhance the security of the electricity distribution system. The congestion problems of the heating networks are addressed in [10] by adjusting the water flow rates in the heating distribution networks. A distributed optimization method is proposed in [11] for optimal scheduling of an office building integrated electricity distribution system. The proposed method can protect the private information of both the office buildings and the electricity distribution system. Great contributions have been made to the optimal scheduling methods for B2EG by the above studies. However, the above studies treat the buildings as passive consumers rather than grid-interactive buildings. The demand response potential of buildings are ignored and their flexibility is not used. To address this issue, this chapter proposes a B2EG framework, where the MEG operator can trigger the HDR from the grid-interactive buildings, while the grid-interactive buildings can enjoy cost saving for heating. In addition, the heating sale price (HSP) is optimized by the MEG operator and the grid-interactive buildings together with the B2EG framework to optimally coordinate the MEG operator and the buildings. The MEG operator and the consumers in buildings are different entities. It means that their interests are heterogeneous or even conflicting. Consumers in buildings tend to reduce their energy costs as much as possible while the MEG operator tends to increase its profit as much as possible. Therefore, this chapter proposes a bi-level optimization approach for optimally coordinating the grid-interactive buildings and the MEG and further satisfy the interests of both the buildings and the MEG operator.

10 Optimization in Grid-Interactive Buildings

233

2 Optimization Framework for Grid-Interactive Buildings in the MEG 2.1 Framework for Interactions Between Buildings and Energy Grids This chapter proposes a B2EG framework, where the MEG operator can trigger the HDR from the grid-interactive buildings, while the grid-interactive buildings can enjoy cost saving for heating. In addition, the HSP is optimized by the MEG operator and the grid-interactive buildings together with the B2EG framework to optimally coordinate the MEG operator and the buildings. The proposed B2EG framework in the multi-energy grids is described in Fig. 1. As can be observed, the CHP unit and the heat pump couple the heating system, the natural gas system, and the electricity system together. Therefore, the MEG operator can optimally determine the schedules of the energy purchases and the schedules of the CHP unit and the heat pump. To trigger the HDR from buildings, the MEG operator optimizes the HSPs. Then, consumers can optimally adjust the flow rates flowing in the indoor radiators based on the HSPs to reduce their heating costs. In this way, the buildings can enjoy economic and efficient energy utilization solutions provided by the MEG operator. In this context, buildings to energy grid (B2EG) integration is beneficial for both the consumers and the MEG operator.

Upper energy grids Transmission grids

Multi-energy grids MEG operator Distribution grids

Buildings

Transformer

CHP

Heat pump

Heat exchanger

Gas grid

Electricity grid

Fig. 1 Framework of the B2EG in the multi-energy grid

Heating grid

234

X. Jin et al.

Upper energy gridss

Energy markets

Natural gas Heating Electricity

Cetbuy _ Oper Cg tbuy _ Oper Chtbuy _ Oper

Pg

Ph

Pe

Energy prices

Leader: MEG operator Php

chp

P , ηe,ηh Hchp CHP Sale heating price: Chsale

ηhp

Hhp

Heat pump

Follower: Grid-interactive buildings Building

Building Building

Building heating loads: Ph,L

Building

Fig. 2 Bi-level optimization for B2EG

2.2 Optimal Operations of Energy Grid-Interactive Buildings Figure 2 describes the bi-level optimization model for B2EG. As can be observed, there are interactions between the MEG operator and the buildings. The MEG operator optimizes the HSP (Chsale ) to the buildings, the energy purchase schedules (Pe , Ph , Pg ), and the energy schedules of the heat pump (Php ) to maximize its profit. Consumers optimally adjust the flow rates flowing in the indoor radiators based on the heating prices to reduce their heating costs.

3 Optimization Model for Grid-Interactive Buildings in the MEG 3.1 MEG Operator Upper-level problem’s objective is the maximum profit of the MEG operator. max T P(x1 , u 1 ) =

T  N  e,L sale h,L [ (Cesale t Pn,t t + Cht Pn,t t) t=1 n=1

10 Optimization in Grid-Interactive Buildings buy_Oper e Pt t

−(Cet

235 buy_Oper g Pt t

+ Cgt

buy_Oper h Pt t)]

+ Cht

(1)

where Cesale is the electricity sale price of MEG operator; Chsale is the MEG operator’s t t e,L h,L is consumer n’s heating HSP; Pn,t is the consumer n’s electricity load at time t; Pn,t buy_oper buy_oper buy_oper , Cgt , Cht are electricity/gas/heating’s purchasing prices load; Cet g for the MEG operator at time t; Pte , Pt , Pth is the purchased electricity/gas/heating by the MEG operator at time t, respectively. The constraints include the electricity balance, the heating balance, the constraints of energy devices, the HSPs, the heating distribution grid, and of the electricity distribution grid. • Electricity balance

chp

Pte + Pt

N 

hp

= Pt +

e,L Pn,t

(2)

n=1 chp

Pt

g

= ηe Pt

(3)

chp

hp

where Pt is the CHP unit’s power generation; Pt represents heat pump’s power consumption; ηe represents CHP unit’s efficiency for electricity generation, and it is set as 0.3 [3]. • Heating balance

chp

Ht

N 

hp

+ Ht + Pth =

h,L Pn,t

(4)

n=1 chp

Ht

hp

g

(5)

hp

(6)

= ηh Pt

Ht = ηhp Pt chp

hp

where Ht is the CHP unit’s heating generation at time t; Ht is the heat pump’s heating generation at time t; ηh is the CHP unit’s efficiency for heating generation; ηhp is heat pump’s efficiency of heating generation. • Constraints of energy devices There are limits for the energy purchases from the upper energy grids due to the CHP unit’s limited capacity and the heat pump’s limited capacity, as shown in Eqs. (7)–(10).

236

X. Jin et al.

0 ≤ Pte ≤ Pte,max g

(7)

0 ≤ Pt ≤ Pt

g,max

(8)

0 ≤ Pth ≤ Pth,max

(9)

hp

hp,max

0 ≤ Pt ≤ Pt

(10)

• Constraints of HSPs The optimal HSPs are constrained according to consumers’ purchase price for heating to guarantee the MEG operator’s profit, as shown in (1) [12]. The average value of the optimal HSPs is also limited to guarantee consumers’ willingness to provide HDR, as shown in Eq. (12). buy_USER

α1 Cht

T 

Chsale t /T ≤

t=1

buy_USER

≤ Chsale ≤ α2 Cht t T 

buy_USER

Cht

/T

(11)

(12)

t=1

buy_USER

is the consumers’ purchasing price for heating at time t; α 1 is set where Cht as 0.9 and α 2 is set as 1.1 [12]. • The heating distribution grid’s constraints The heating distribution grid enabling B2EG is shown in Fig. 3 [13]. The mass flow rate flowing into each heating node is equal to the mass flow rate flowing out of each heating node, as shown below:

Buildings Operator Heating transmission grid Supply pipe Return pipe

Fig. 3 Heating distribution grid enabling B2EG

Heating distribution grid

10 Optimization in Grid-Interactive Buildings

237

˙ pipe = m ˙ node YHDN m

(13)

Water flow rates are limited, as shown below: pipe,min

ml

pipe

pipe,max

≤ m l,t ≤ m l

∀l ∈ N pipe , t ∈ T

(14)

The drop of the node pressure along the pipe is shown in (15): pipe

pn,t − pn+1,t = ξl · (m l,t )2 ∀l ∈ N pipe , n ∈ N node ξl =

8κl L l , ∀l ∈ N pipe dl5 π 2 ρ

(15) (16)

where pn,t is heating node n’s node pressure; κ l is friction factor; L l is the length of pipe l; d is inner pipe’s diameter; ρ is water density. Constraints of the node pressure are described in Eq. (17) [14]. pnmin ≤ pn,t ≤ pnmax

(17)

• Constraints of the electricity distribution grid The power flows equations for an electricity distribution feeder in Fig. 4 are shown as follows [15]: Pn+1 = Pn − r f

Pn2 + Q 2n load − Pn+1 Vn2

(18)

Q n+1 = Q n − x f

Pn2 + Q 2n − Q load n+1 Vn2

(19)

2 Vn+1 = Vn2 − 2(r f Pn + x f Q n ) + (r 2f + x 2f )

Fig. 4 Example of a distribution feeder

Pn2 + Q 2n Vn2

(20)

238

X. Jin et al.

where V n represents the voltage of node n; Pn + iQn represents apparent power flow; load + iQload Pn+1 n+1 represents the electricity load; r f + ix f represents the impedance of the line linking node n and n + 1. Voltage is constrained, which is shown in (21) [15]. 1 − ε ≤ Vn ≤ 1 + ε

(21)

3.2 Buildings The lower level problem’s objective is the minimum cost for each consumer in the building. min OC(x2 , u 2 ) =

T 

e,L sale h,L (Cesale t P1,t t + Cht P1,t t)

(22)

t=1

• Constraints of buildings’ thermal dynamics We use the Resistor–Capacitor (RC) thermal network method [16] to model buildings’ thermal dynamics, as shown in Fig. 5. As can be seen, we use the wall node and heating zone node in the RC model to represent the wall temperature and the indoor temperature. All the nodes are connected together by the thermal resistance. The thermal resistance represents thermal mass’s ability to transmit heat. Meanwhile, all the nodes are grounded by thermal capacity. The thermal capacity represents thermal mass’s ability to preserve heat. Walls’ thermal balance according to the RC is formulated in (23). The indoor air’s thermal balance according to the RC is formulated in (24). ⎧ 1,2 ⎪ 1,2 dTw ⎪ Cw ⎪ ⎪ ⎪ dt ⎪ ⎪ ⎪ 1,3 ⎪ dTw ⎪ ⎪ ⎨ Cw1,3 dt 1,4 ⎪ dTw ⎪ 1,4 ⎪ Cw ⎪ ⎪ ⎪ dt ⎪ ⎪ ⎪ 1,5 ⎪ ⎪ ⎩ Cw1,5 dTw dt

Troom − Tw1,2 Rw1,2 room T − Tw1,3 = Rw1,3 room T − Tw1,4 = Rw1,4 room T − Tw1,5 = Rw1,4 =

T2 − Tw1,2 Rw1,2 3 T − Tw1,3 + Rw1,3 4 T − Tw1,4 + Rw1,4 5 T − Tw1,5 + Rw1,4 +

+ r 1,2 α 1,2 Aw1,2 Qrad1,2 : λ1,t + r 1,3 α 1,3 Aw1,3 Qrad1,3 : λ2,t + r 1,4 α 1,4 Aw1,4 Qrad1,4 : λ3,t + r 1,5 α 1,5 Aw1,5 Qrad1,5 : λ4,t (23)

Cr1

4 dTroom  Tw1,j − T room T out − T room = + Q load + + Q int 1 1,j dt Rwin Rw j =1

+ τ win Awim Qrad1,5 : λ5,t

(24)

10 Optimization in Grid-Interactive Buildings

239

T3

Cw1,3

T

Qr1

4

Cw1,4

Cr

T2

T Qint 1

1,4 rad

Q

Cw1,2 Cw1,5

1,5 Qrad

Rwin

T

5

Window Zone 4

Window Zone 3

Radiator

Radiator

Radiator

Radiator

Zone 1 Window

Zone 2 Window

Tout Tr Rwin T

out

C

R1

R2 2

Tw Qrad

Tw

Cw

R2 2

Tr

R3

Cr

Tw Qr

Qint

Fig. 5 RC thermal network model for one heating zone

where Twi,j represents the wall temperature; Ttroom represents heating zone’s indoor air temperature at time t; Cwi,j represents wall’s thermal capacity; Rwi,j represents wall’s thermal resistance; Awi,j represents wall’s area; α i,j represents wall’s coefficient of the radiative heat absorption; Qradi,j represents the density of the radiative heat flux on the wall; Cri represents the zone’s heat capacity; Rwin represents the window’s thermal resistance; Qint is the zone’s internal heat gain; τ win is the window’s transmittance; Awin is the zone’s window area; Ttout is the outdoor temperature at time t; λ1,t , λ2,t , λ3,t , λ4,t , λ5,t are dual variables.

240

X. Jin et al.

The zone i’s heating load (Q iload ) is calculated by (25). load = c p ∗ m rt ∗ (Ts − Tr ) Q i,t

(25)

where cp is the water’s specific heat capacity; T s , T r are the temperatures of the supply and return water. One consumer’s heating load containing I zones can be calculated by (26). h,L P1,t =

I 

load Q i,t

(26)

i=1

• Constraints of the indoor air temperatures The indoor temperatures are limited to ensure consumers’ comfort level: L U T room,min ≤ Ttroom ≤ T room,max : β1,t , β1,t

(27)

• Constraints of the water flow rates

L U m r,min ≤ m rt ≤ m r,max : β2,t , β2,t

(28)

U U L L where βl,t , βl,t , β2,t , β2,t are dual variables of inequality constraints.

4 Optimization Solution for Grid-Interactive Buildings 4.1 Transformation of Bi-Level Optimization Model The HSPs are optimized in the upper-level optimization problem, thus they are constant inputs for the lower-level optimization problem. Therefore, the lowerlevel optimization problem is equivalent to its KKT conditions since it is convex. Therefore, we replace the lower-level optimization problem using its KKT conditions. Consequently, bi-level optimization problem is reformulated as a mathematical problem with complementarity constraints (MPEC) [17].

10 Optimization in Grid-Interactive Buildings

241

KKT conditions are shown in (32)–(34). • Lagrange function

L(Ttroom , T wt1,2 , T wt1,3 , T wt1,4 , T wt1,5 , m t ) = −



βjh j +

j∈ inequal

T 

e,L sale h,L (Cesale t P1,t t + Cht P1,t t)

t=1



λk gk

k∈ equal

(29) • Stationarity conditions

∂L =λ1,t + λ2,t + λ3,t + λ4,t ∂ Ttroom ⎛ ⎛ ⎛ ⎞ ⎞⎞ 5 5   1 1 4 ⎠− ⎠⎠λ5,t −⎝ +⎝Cr*⎝ 1,j 1,j Rwin Rw Rw j=2 j=2 ⎛ ⎛ ⎞⎞ 5  1 1 ⎠⎠ L U ∗ Cr ∗ ⎝ −λ5,t−1 ⎝ + β1,t =0 − β1,t 1,j Rwin Rw j=2 ⎧ ∂L ⎪ ⎪ ⎪ ⎪ ⎪ ∂ T wt1,2 ⎪ ⎪ ⎪ ⎪ ∂L ⎪ ⎪ ⎪ ⎨ ∂ T w 1,3 t ∂L ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ∂ T wt1,4 ⎪ ⎪ ⎪ ⎪ ∂L ⎪ ⎪ ⎩ ∂ T wt1,5

2 )λ1,t + Rw1,2 2 = (Cw1,3 − )λ2,t + Rw1,3 2 = (Cw1,4 − )λ3,t + Rw1,4 2 = (Cw1,5 − )λ4,t + Rw1,5 = (Cw1,2 −

1 λ5,t − Cw1,2 λ1,t−1 Rwin 1 λ5,t − Cw1,3 λ2,t−1 Rwin 1 λ5,t − Cw1,4 λ3,t−1 Rwin 1 λ5,t − Cw1,5 λ4,t−1 Rwin

(30)

=0 =0 (31) =0 =0

∂L L U = Chsale ∗ (cp ∗ (Ts − Tr ) ∗ I) + λ5,t ∗ (cp ∗ (Ts − Tr )) − β2,t + β2,t =0 t ∂m rt (32) • Complementary slackness conditions 

L 0 ≤ β1,t ⊥(Ttroom − T room,min ) ≥ 0 U ⊥(T room,max − Ttroom ) ≥ 0 0 ≤ β1,t

(33)

242

X. Jin et al.



L 0 ≤ β2,t ⊥(m rt − m min ) ≥ 0 U ⊥(m max − m rt ) ≥ 0 0 ≤ β2,t

(34)

4.2 Linearization of the MPEC Model h,L There is a bilinear term Chsale t Pn,t in (1), (33) and (34) are nonlinear. Therefore, the MPEC model is also nonlinear. If the problem is convex, then the primal problem’s objective is equal to the corresponding dual problem’s objective according to the strong duality theorem [17–19]. Therefore, the lower-level problem’s objective (i.e., Eq. (22)) equals its dual problem’s objective, as shown in (35). Then, we can linearize h,L the bilinear term Chsale t Pn,t .

Dual_inner1,t =

T 

h,L sale e,L (Chsale t P1,t t + Cet P1,t t)

t=1

=

T 

e,L Cesale t P1,t t

−

t =1

+

T 

+

L β1,t Ttroom,min −

T 

T 

T  Tt2 λ ) + (−r 1,2 α 1,2 Aw1,2 Qrad1,2 1,t t λ1,t ) Rw1,2 t =1

(−

T  Tt3 λ ) + (−r 1,3 α 1,3 Aw1,3 Qrad1,3 λ2,t ) 2,t Rw1,3 t =1

(−

T  Tt4 λ ) + (−r 1,4 α 1,4 Aw1,4 Qrad1,4 λ3,t ) 3,t Rw1,4 t =1

(−

T  Tt5 λ ) + (−r 1,5 α 1,5 Aw1,5 Qrad1,5 λ4,t ) 4,t Rw1,5 t =1

(−

T T   Ttout λ5,t ) + (−Q int (−τ win Awin Q win t λ5,t ) + t λ5,t ) Rwin t =1 t =1

t =1

+

T  t =1

+

T  t =1

U β1,t Ttroom,max

(−

t =1

+

T  t =1

t =1

+

U β2,t m r,max t

t =1

t =1 T 

T 

(35) According to Eq. (35), the objective of the MPEC model can be updated, as shown in (36).

10 Optimization in Grid-Interactive Buildings

⎡ max TP =

T  t=1

243

⎤

N 

(Dual_inner n,t ) ⎢ ⎥ ⎢ ⎥ (36) ⎣ n=1 ⎦ buy_Oper e buy_Oper g buy_Oper h Pt t + Cgt Pt t + Cht Pt t) −(Cet

Then, we linearize the complementary slackness conditions using the Big-M method [20]. The linearized constraints are shown below: L L 0 ≤ β1,t ≤ Mθ1,t

(37)

L 0 ≤ Ttroom − T room,min ≤ M(1 − θ1,t )

(38)

U U 0 ≤ β1,t ≤ Mθ1,t

(39)

U 0 ≤ T room,max − Ttroom ≤ M(1 − θ1,t )

(40)

L L 0 ≤ β2,t ≤ Mθ2,t

(41)

L 0 ≤ m rt − m min ≤ M(1 − θ2,t )

(42)

U U 0 ≤ β2,t ≤ Mθ2,t

(43)

U 0 ≤ m max − m rt ≤ M(1 − θ2,t )

(44)

It is worth noting that Eq. (15) is nonlinear. We use piecewise linear function [21] to linearize (15). The water flow rate in pipeline l in (15) can be decomposed into Q pipe,min pipe,max , ml ] as abscissas. The ordinates and slopes are shown segments in [m l in Eqs. (45)–(46). q

q

yl = 0, xl = 0, q = 1 pipe,max

q

q

q

yl = (xl )2 , xl =

ml

Q

·q

q

, kl =

(45)

q

q−1

q

q−1

yl − yl xl − xl

q≥2

(46)

As can be observed from (47), by dividing the function’s curve in (15) into many segments, it is replaced by many linear functions. The piecewise linear segment’s constraints are described in (48)–(49). pn,t − pn+1,t = ξl ·

Q  q=1

pipe

q q

q

q q

[(Al,t − xl δl,t )kl + yl δl,t ]

(47)

244

X. Jin et al. q

pipe

q

q

q+1

δl,t xl ≤ Al,t ≤ δl,t xl Q 

(48)

q

δl,t = 1

(49)

q=1

pipe

pipe

q

where Al,t is numerically equal to m l,t , and δl,t are auxiliary binary variables. They pipe represent whether the value of Al,t is in the abscissas of q. We linearize (18)–(20) according to two commonly used assumptions in low P 2 +Q 2

voltage electricity distribution network [15]: ➀ The nonlinear terms j V 2 j reprej senting power losses are much lower than the value of the branch’s power terms [22]; ➁ (Vn − V0 )2 ≈ 0 which leads to Vn2 ≈ V02 + 2V0 (Vn − V0 ) [15]. Then, the linear constraints of the electricity distribution network are described in (50)–(52). load Pn+1 = Pn − Pn+1

(50)

Q n+1 = Q n − Q load n+1

(51)

Vn+1 = Vn −

r f Pn + x f Q n V0

(52)

After the linearizations, the whole MPEC problem is a typical MILP model.

5 Case Studies for Grid-Interactive Buildings 5.1 Case Setting The MEG with multiple grid-interactive buildings is utilized to test the proposed method. The schematic of the MEG with buildings is shown in Fig. 6. We assume that one building contains 20 floors and each floor contains two consumers. Each consumer contains four heating zones. Each zone is 6 m long, 6 m wide, and 3 m high. In this chapter, heating nodes H4, H6, H8, and H9 and electricity nodes E1, E2, E3, and E4 connect 10, 8, 6, and 6 buildings. One heating zone’s thermal parameters are listed in Table 1. The data of the solar radiation and the outdoor temperature are taken from [16]. The MEG operator’s energy purchase prices and the consumers’ energy purchase prices are shown in Fig. 7.

10 Optimization in Grid-Interactive Buildings

245

Fig. 6 Schematic of the MEG with buildings

Table 1 Thermal parameters of the buildings Rw1,j (K/W)

Rwin (K/W)

Cr (J/K)

Cw1j (J/K)

Awin (m2 )

0.06

0.02

2.5e + 5

7.9e + 5

4

5.2 Results

(1)

MEG operator

The MEG operator’s optimal energy purchase schedules are shown in Fig. 8. As can be seen, the MEG operator is able to optimally decide its schedules of energy purchases according to the energy prices. MEG operator’s energy supply schedules are shown in Fig. 9. As shown in Fig. 9a, the MEG operator tends to commit the heat pump to supply heating for buildings because the heat pump’s heating generation efficiency is higher than the CHP unit’s heating generation efficiency. Due to the heat pump’s limited capacity, the MEG operator has to commit CHP (e.g., at 21:00-22:00) or directly import heating energy from the upper heating grid (e.g., at 01:00-04:00) when the heating loads peak. The optimal HSPs are depicted in Fig. 10. As can be seen, the MEG operator optimizes the HSPs within the optimal range. (2)

Grid-interactive buildings

The optimal scheduling results for the grid-interactive buildings (taking one heating zone as an example) are depicted in Figs. 11 and 12. As can be observed from Fig. 11, the HSPs’ valleys lead the water flow rates’ peaks. On the contrary, HSPs’ peaks lead to the water flow rates’ valleys. Therefore, the HSPs can trigger the HDR from the

246

X. Jin et al.

Fig. 7 Energy purchase prices

Electricity

0.2

Heating

Gas

Price ($/kWh)

0.16 0.12 0.08 0.04 0

1 3 5 7 9 11 13 15 17 19 21 23 Hour (h)

(a) Energy purchase prices for the MEG operator

Electricity

0.2

Heating

Price ($/kWh)

0.16 0.12 0.08 0.04 0

1 3 5 7 9 11 13 15 17 19 21 23 Hour (h)

(b) Energy purchase prices for consumers Electricity

16000

Energy purchase (kWh)

Fig. 8 MEG operator’s energy purchase schedules. a Heating. b Electricity

Heating

Gas

12000 8000 4000 0

1

3

5

7

9

11 13 15 17 19 21 23 Hour (h)

buildings. The indoor temperatures are adjusted in the consumer’s comfort range, as shown in Fig. 12. (3)

Comparative studies

We conduct three scenarios to analyze the benefits of the proposed method for B2EG, as shown in Table 2.

10 Optimization in Grid-Interactive Buildings

Heating load Purchased heating

Heating generated by HP Heating generated by CHP

Heating (kWh)

10000 7500 5000 2500 0 250050007500- 10000 1

247

3

5

7

9

11 13 15 17 19 21 23 Hour (h)

(a) Heating

Power (kWh)

8000 6000 4000 2000 0 - 2000 - 4000 - 6000 - 8000 1

Electricity consumption of HP Electricity generated by CHP

Electric load Purchased electricity

3

5

7

9

11 13 15 17 19 Hour (h)

21 23

(b) Electricity Fig. 9 MEG operator’s energy supply schedules Fig. 10 Optimal heating prices

Sale heating price

Upper limits Lower limits

0.12 Heating price ($/kWh)

0.08 0.04 0

1

3

5

7

9

11 13 15 17 19 21 23 Hour (h)

248

X. Jin et al. Flow rate 0.030 Flow rate (kg/s)

0.08

0.020

0.04 0

0.010

1

3

5

7

9

5

7

11 13 15 17 19 21 23 Hour (h)

0

24.0 Indoor temperature (

Fig. 12 Optimal indoor temperature of one heating zone

Sale heating price

0.12 Heating price ($/kWh)

Fig. 11 Optimal water flow rate of the radiator in one heating zone

23.0 22.0

C) 21.0

1

3

9

11 13 15 17 19 Hour (h)

21 23

Table 2 Three scenarios Scenarios

HSP is optimized?

Consumers are involved in heating pricing?

Consumers provide HDR?

I

Yes

Yes

Yes

II

No

No

Yes

III

No

No

No

Scenario I is the B2EG solution, where the HSP is optimized and consumers in buildings provide HDR to the MEG operator. In scenario II, the MEG operator sells heating to the consumers in a traditional way, buy_ USER i.e., the HSPs are fixed values. In this scenario, 40 different HSPs from α1 Cht buy_ USER buy_ USER to α2 Cht with a step size of 0.05Cht are simulated. In scenario III, the buy_ USER , α 3 = 0.905), and accordingly the MEG operator also sets fixed HSP (α3 Cht consumers’ heating loads are not adjustable. Therefore, consumers do not provide HDR in this scenario. The consumers’ energy costs and the MEG operator’s profits in scenario I and scenario II are compared, as demonstrated in Fig. 13. As can be observed from the results, in scenario II, the higher the profit of the MEG operator, the more the consumer’s energy cost, and vice versa. Therefore, the benefits of the MEG operator and the consumers in buildings cannot be balanced in this scenario. In contrast, scenario I optimizes the HSP and thus is able to benefit the MEG operator and the consumers in buildings at the same time.

10 Optimization in Grid-Interactive Buildings Scenario II

8500

1.1Chtbuy_User

8000 7500

Scenario I

7000 6500

16.26

16.39

16.13

15.88

16

15.75

15.62

15.54

15.41

15.28

15.16

15.03

14.9

6000

0.9Chtbuy_User

14.77

Profits of the MEG operator ($)

Fig. 13 The consumers’ energy cost and the MEG operator’s profits in scenarios I and II

249

Energy cost of one consumer ($)

Table 3 The MEG operator’s profits and consumers’ energy cost Scenario

The MEG operator’s profits ($)

One building consumer’s energy costs ($)

All building consumers’ energy costs ($)

I

7334.45

15.60

18,720

II

[6371.32, 8307.64]

[14.77, 16.39]

[17724, 19668]

III

7337.72

16.14

19,368

The MEG operator’s profits and the consumers’ energy costs in the three scenarios are shown in Table 3. As can be observed from the results, in scenarios I and III, the MEG operator’s profits are almost the same in the two scenarios. However, one consumer’s energy cost in scenario III is about 4% larger than one consumer’s energy cost in scenario II because consumers cannot adjust the water flow rates in their radiators in scenario III.

6 Conclusions This chapter developed a bi-level optimization method for grid-interactive buildings integrated into the MEG. The MEG operator optimizes the HSP to the buildings and the energy schedules by dispatching the multi-energy devices. Thanks to buildings’ thermal inertia, consumers in buildings can optimally adjust the water flow rates in the indoor radiators to reduce the heating costs and provide HDR to the MEG operator. The following conclusions can be achieved: (1) The proposed bi-level optimization method is able to benefit the MEG operator and the consumers in buildings at the same time. Thus, the grid-interactive buildings and the MEG operator can be coordinated optimally. (2) The flexibility of the grid-interactive buildings can be fully used in the optimization of the MEG. Moreover, the heating costs of the buildings can be significantly reduced with the proposed bi-level optimization method.

250

X. Jin et al.

References 1. Jin X, Wu Q, Jia H (2020) Local flexibility markets : literature review on concepts, models and clearing methods. Appl Energy 261:114387 2. He Z, Hou K, Wang Y et al (2019) Reliability modeling for integrated community energy system considering dynamic process of thermal loads. IET Energy Syst Integr 1(3):173–183 3. Lin W, Jin X, Mu Y et al (2018) A two-stage multi-objective scheduling method for integrated community energy system. Appl Energy 216:428–441 4. Wang D, Hu Q, Jia H et al (2019) Integrated demand response in district electricity-heating network considering double auction retail energy market based on demand-side energy stations. Appl Energy 248:656–678 5. Finck C, Li R, Zeiler W (2020) Optimal control of demand flexibility under real-time pricing for heating systems in buildings: a real-life demonstration. Appl Energy 263:114671 6. Zhang R, Jiang T, Li W et al (2019) Day-ahead scheduling of integrated electricity and district heating system with an aggregated model of buildings for wind power accommodation. IET Renew Power Gener 13:982–989 7. Dai Y et al (2019) Dispatch model for CHP with pipeline and building thermal energy storage considering heat transfer process. IEEE Trans Sustain Energy 10(1):192–203 8. Li X, Li W, Zhang R (2020) Collaborative scheduling and flexibility assessment of integrated electricity and district heating systems utilizing thermal inertia of district heating network and aggregated buildings. Appl Energy 258:114021 9. Li Z, Su S, Jin X (2021) A hierarchical scheduling method of active distribution network considering flexible loads in office buildings. Int J Electr Power Energy Syst 131:106768 10. Cai H, Ziras C, You S et al (2018) Demand side management in urban district heating networks. Appl Energy 230:506–518 11. Li Z, Su S, Jin X et al (2021) Distributed energy management for active distribution network considering aggregated office buildings. Int Renew Energy 180:1073–1087 12. Li Y, Feng C, Wen F et al (2018) Energy pricing and management of park energy internet including electric vehicles and electricity to gas. Autom Electr Power Syst 42(16):192–196 13. Lu Y, Yu X, Jin X, Jia H, Mu Y (2021) Bi-level optimization framework for buildings to heating grid integration in integrated community energy systems. IEEE Trans Sustain Energy 12(2):860–873 14. Li J, Lin J, Song Y et al (2019) Operation optimization of power to hydrogen and heat (P2HH) in ADN coordinated with the district heating network. IEEE Trans Sustain Energy 10(4):1672– 1683 15. Yeh H, Gayme DF, Low SH (2012) Adaptive VAR control for distribution circuits with photovoltaic generators. IEEE Trans Power Syst 26(3):1656–1663 16. Jin X, Wu Q, Jia H et al (2021) Optimal integration of building heating loads in integrated heating/electricity community energy systems: a bi-level MPC approach. IEEE Trans Sustain Energy 12(3):1741–1754 17. Wang Y, Dvorkin Y, Fernández-Blanco R et al (2017) Look-ahead bidding strategy for energy storage. IEEE Trans Sustain Energy 8(3):1106–1117 18. Pourakbari-Kasmaei M, Asensio M, Lehtonen M, Contreras J (2020) Trilateral planning model for integrated community energy systems and PV-Based prosumers—A bi-level stochastic programming approach. IEEE Trans Power Syst 35(1):346–361 19. Fernández-Blanco R, Arroyo JM, Alguacil N (2014) Network-constrained day-ahead auction for consumer payment minimization. IEEE Trans Power Syst 29(2):526–536 20. Jia Y, Mi Z, Yu Y et al (2018) A bilevel model for optimal bidding and offering of flexible load aggregator in day-ahead energy and reserve markets. IEEE Access 6:67799–67808 21. Carrion M, Arroyo JM (2006) A computationally efficient mixed-integer linear formulation for the thermal unit commitment problem. IEEE Trans Power Syst 21(3):1371–1378 22. Baran ME, Wu FF (1989) Network reconfiguration in distribution systems for loss reduction and load balancing. IEEE Trans Power Del 4(2):1401–1407

Chapter 11

Cost-Benefit and Short-Term Power Flow Analysis of Grid Integrated Residential Photovoltaic-Battery Energy System Mohamed J. M. A. Rasul, Naleen de Alwis, and Mohan Lal Kolhe

Abstract A recent progressing trend on grid-connected residential photovoltaic (PV) system has been reported all around the world with the focuses on various aspects, mainly on the reduction of grid consumption and demand-side management with peak shaving and peak shifting. Subsequently, a huge impact on the distribution network is caused due to the power quality problems arising from grid-connected PV systems. As the most promising solution, the integration of battery energy storage with the PV system is going to improve the operation of distribution network by enhancing self-consumption as well as in demand side management. The purpose of this study is to investigate the technical performance of a battery-coupled gridconnected PV system and to assess the economic viability of such integrated renewable energy systems in Sri Lanka. In this work, a residential PV-battery energy system is designed and developed considering a control algorithm for energy efficient system operation at conditions to maximize the PV self-consumption through battery energy throughput and to reduce the utility grid consumption accordingly. The system is analysed for dynamic load profile under variable PV conditions with dynamic battery state of charge (SOC) conditions. The results are analysed for system performance and stability conditions. Overall system performance indicated a stable system operation for each investigated condition by effectively managing the PV, battery and grid systems to meet the real-time residential demand. The economic viability of the proposed PV system is analysed using the discounted cash flow (DCF) approach using the indices of benefit-cost ratio (B/C), net present value (NPV), internal rate of return (IRR), and payback period, which revealed to be viable in regular Sri Lankan grid connected PV-battery energy residential system. M. J. M. A. Rasul (B) · N. de Alwis Faculty of Engineering and Management, Ocean University of Sri Lanka, Crow Island, Mattakuliya, 15, Colombo, Sri Lanka N. de Alwis e-mail: [email protected] M. L. Kolhe Faculty of Engineering and Science, University of Agder, PO Box 422, NO 4604 Kristiansand, Norway e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 A. Tomar et al. (eds.), Control of Smart Buildings, Studies in Infrastructure and Control, https://doi.org/10.1007/978-981-19-0375-5_11

251

252

M. J. M. A. Rasul et al.

1 Introduction In the last decades, power generation from renewable energies such as wind and solar has been significant growth around the world to meet the rising energy demand day by day with the focus of moving towards a sustainable and green energy future. Most importantly, the global grid-connected solar power capacity reached 586.4 GW at the end of 2019 and it has been continuing to grow with the advancement of power electronics technologies [1]. Thus, the capacities of implemented PV systems are ranging from few kilowatts to hundreds of megawatts, which can be categorized as small scale (residential level), medium-scale (commercial level) and large scale (utility-scale). According to the configuration of the PV system, there are four common types such as PV-direct, off-grid or stand-alone, battery-based grid-tied and battery-less gridtied. Moreover, the utility-scale systems are connected to the medium voltage (MV) network while the small scale to medium-scale systems are connected to low voltage (LV) network [2]. Notably, grid-connected PV systems in medium and small scales have been progressing rapidly all around the world in recent years as the electricity consumers have been allowed to set up grid-connected PV systems as distributed energy resources (DERs) by the grid operators. This rapidly growing trend is due to the high electricity tariffs, which varies from time to time, and it is providing an opportunity for future investment in PV systems. Although this is an effective solution for grid operators to ease the burden on the national grid owing to rapid energy demand growth, these causes severe technical impacts on the entire distribution network as these PV systems are intermittent and inject fluctuating real power to the grid at many points in the LV distribution network. The several impacts caused by PV systems due to high penetration conditions, critical issues as voltage rise, voltage unbalance and reverse power flow have been addressed in previous literature and a comprehensive review on the impact of PV systems in LV distribution networks is presented in [3]. As mitigation solutions for high PV penetration effects in the LV distribution network, distribution network reinforcements or expensive grid infrastructures such as integration of on-load tap changers (OLTC), step voltage regulators (SVR) [4–6], static VAR compensators (SVC) [7] or fixed and switchable capacitors (FC and SC) can be possibly deployed by the grid operators. But, it is a practically questionable fact that how far these costly investments of grid oriented approaches can perform effectively to mitigate the cumulative impact of PV systems at different locations in the distribution network [8]. Also, active power curtailment of PV systems [9], reactive power consumption by PV inverters [10, 11] or integration of storage devices with PV systems [12] are feasible approaches implemented specifically on individual PV systems. Notably, the active power curtailment method may not be an economically beneficial solution for the PV system owners as it limits the PV production feed of the system while the reactive power compensation option may raise reactive power flow over the network to a higher level and also cause extra losses in the LV feeder network. In particular, the integration of electrical energy storage (EES) system with DERs is one of the popular trends in order to overcome the intermittency

11 Cost-Benefit and Short-Term Power Flow Analysis …

253

of DERs and simultaneously smooth their energy output throughout the operation. It is reported that batteries, being the most proposed EES system [13], have been able to address the high PV penetration issues as the batteries enable to store surplus PV production without disturbing the PV production and consequently, limit the high penetration into the distribution network. Different approaches and strategies implemented on residential PV-battery energy system operation have been proposed in several works of literature where the main objective thereof is to eliminate the voltage violations due to the high PV system penetrations in the distribution network [14, 15]. In the meantime, particularly, the self-consumption in residential applications is enhanced to a maximum level with the use of a battery energy system (BES) meanwhile properly maintaining the battery performance and condition at an optimum level considering battery lifetime [16]. This is due to the facts of battery cost and lifetime, which are the main constraints of PV-battery system implementations. Since the implementation cost of a PV-battery system is indeed significantly higher than that of a grid-connected PV system with the same PV capacity, the economic benefits and feasibility, in addition to technical benefits, are also critical factors in such system implementations. With respect to economic feasibility, several investigations were conducted considering different scenarios to analyse the economic benefits of the PV-battery energy system in comparison to the PV systems [17–22]. With the initiation of ‘Soorya Bala Sangramaya’, the national level power generation project in Sri Lanka which has enabled electricity consumers to set up small solar power plants on the rooftops of households, religious places, hotels, commercial establishments and industries, installation of rooftop solar power plants has been significantly escalated. It is reported that the total installed capacity of rooftop solar plants was 70.442 MW at the end of 2017 and 217.263 MW at the end of 2019 [23, 24]. Notably, rooftop solar plants have been rising annually at a rapid rate in Sri Lanka, and the issue of high penetration residential PV system effect has already been experienced [25]. Ceylon Electricity Board (CEB), the electrical power generation, trans-mission and distribution authority of Sri Lanka, has taken necessary action to regulate the PV installation capacity under the distribution substation as follows. Furthermore, the determination of PV capacity to be installed in a residential place is based on the average monthly demand of that particular household. Consequently, most of the consumers might have lost the opportunity to implement rooftop solar plants, although they are ready to accommodate. Hence, with this approach, the transition from consumer to prosumer in the country might be restricted in the next few years. As the main grid operator, CEB is always responsible to fulfil the maximum daily peak demand (night peak demand) of the country while continuously serving the daily demand of the country irrespective of any conditions. The energy demand has been continuously growing, and it is reported that the annual maximum night peak demand of Sri Lanka has been stepping up gradually each year [26]. Although the LV distribution feeders, which are radial type and the power flow from upstream MV distribution network, are continuously monitored at the respective distribution substations, the implementation of residential PV-battery energy systems

254

M. J. M. A. Rasul et al.

Fig. 1 Residential PV-battery energy system configuration

has not yet permitted in Sri Lanka. As a result, this research mainly focuses on the technical performance evaluation of a battery-coupled residential PV system in which the utility grid is maintained as backup power, and economic feasibility assessment of such energy system implementation at the consumer level. The general configuration of the grid-connected residential PV-battery energy system is illustrated in Fig. 1. Since the PV power generation is dependent upon weather condition, BES is necessary to get stable and reliable power from the PV system for both loads and the utility grid. The maximum power point tracking (MPPT) ensures the PV system to provide its instantaneous maximum power. As BESs, lithium-ion batteries with high power and energy density are more dominant. Bi-directional converters are deployed to charge and discharge the BES. Power exchange with the grid is limited by managing residential load demand with the PV system.

2 PV-Battery Residential Energy System Implementation The residential PV-battery energy system configuration shown in Fig. 1 is modelled with the relevant control systems for the PV system, battery system and inverter system as illustrated in Fig. 2. Hence, the PV module, BES, power electronic converters, residential load and utility grid are the main components. The system components are connected by the solid lines with the representation of the power flow directions, and the signal flows of control systems are represented by the dotted lines. For this study, a system of 3 kW PV capacity and 2.5 kWh battery capacity residential energy system is exploited for system performance analysis.

11 Cost-Benefit and Short-Term Power Flow Analysis …

255

Fig. 2 System configuration diagram of the residential PV-battery energy system

2.1 PV System Model The PV system model comprises the PV module and DC-DC boost converter with the maximum power point tracking (MPPT) controller as shown in Fig. 3. The PV module is modelled with the PV array, and the model specifications are presented in Table 1.

Fig. 3 PV system model

Table 1 PV array specifications

PV array Maximum power (kW)

3.052

Open circuit voltage (V)

321

Short circuit current (A)

11.92

Voltage at MPP (V)

273.5

Current at MPP (A)

11.16

256

M. J. M. A. Rasul et al.

As the PV output power varies with the intensity of solar radiations and the cell temperature, it is significant to continuously maintain the maximum power point (MPP) operating condition of the PV module in all meteorological conditions for an effective PV system operation. Therefore, the DC-DC boost converter is equipped with an MPPT controller which efficiently tracks the MPP of the PV module by varying the duty cycle and generates the pulses for the boost converter operation simultaneously. The incremental conductance algorithm (the MPPT algorithm) is utilized in the MPPT controller to provide maximum continuous power from the PV module under the variation of irradiation and temperature.

2.2 Battery System Model Figure 4 depicts the battery system model which includes the battery module and the bi-directional converter with the battery controller. The battery module is a vital component in the system where it is responsible to optimize the produced solar power as required by the residential load conditions and to charge the battery when needed. Proper controlling of both charging and discharging modes of battery system permits the overall system to operate in all PV power conditions of surplus, shortage or out-age. The parameters of battery model is used from the Simulink library. The lithium-ion battery type is chosen, and the relevant model specifications are given in Table 2.

Fig. 4 Battery system module

Table 2 Battery module specifications

Battery module Nominal voltage (V)

48

Rated capacity (Ah)

53

Initial state-of-charge (%)

80

Nominal discharge current (A)

23.0435

Maximum charged voltage (V)

55.8714

11 Cost-Benefit and Short-Term Power Flow Analysis …

257

Fig. 5 a Buck converter. b Boost converter. c Bidirectional converter

The battery controller system ensures the proper operation of charging and discharging mode of the battery system while maintaining the DC bus voltage at 400 V. Proper management of a battery system leads to a long life of the battery system by ensuring the operation of the battery system in charging and discharging modes with specified regions. Conventional DC-DC buck and boost converters allow only unidirectional power flow either high voltage (HV) to the low voltage (LV) side or vice versa. A combination of both buck and boost converters by replacing the diode with a controllable switch can be implemented as a bidirectional converter as shown in Fig. 5. Hence, the bidirectional converter facilitates power flow in both directions. Therefore, it is used to charge and discharge the BES with the charge controller.

2.3 Inverter System Model The main function of the single-phase inverter system is to feed the residential loads by properly managing the generated DC power from the PV module and the battery system while maintaining the input DC voltage at 400 V and converting it into a single-phase AC voltage of 240 Vrms, 50 Hz. In addition, in this proposed system, the utility grid is utilised as a backup power which is to feed the residential loads when needed and to direct the surplus PV power to the grid. The inverter system

258

M. J. M. A. Rasul et al.

Fig. 6 Inverter system module

module as indicated in Fig. 6 is equipped with an inverter, an LCL filter and an inverter operation controller.

2.4 Overall System Operation The system has different operating stages that depend on the PV and battery conditions as depicted in Fig. 7. The generated PV power is utilized simultaneously to charge the battery and fulfil the residential consumptions to maximize selfconsumption (see Fig. 7a). In the meantime, the combination of PV and battery system keeps actively powering the residential loads when the PV power is not sufficient enough to power the residential consumptions. After the battery system is fully charged by the generated PV power, the residential loads are continuously served by the PV power and the surplus PV power is fed to the utility grid as in Fig. 7b. During the time when there is no PV power as in Fig. 7c, the on-time residential demand, which is within the battery system’s maximum power capacity, is driven by the battery system alone until it reaches maximum discharging conditions. The conditions where the on-time residential demand exceeds the maximum power capacity of the battery system is served by both battery system and utility grid as in Fig. 7d, till the battery system reaches to fully discharged condition. The utility grid is then active to supply power to residential loads till the PV-battery system reaches a capable condition (see Fig. 7e).

3 Economic Assessment of PV-Battery Residential Energy System With the rising concerns globally on energy scarcity, the environmental impact of energy production and the resulting climate change, renewable resources are required to produce sustainable energy systems [27–29]. Solar photovoltaic energy has gained

11 Cost-Benefit and Short-Term Power Flow Analysis … Fig. 7 System operating stages

259

260

M. J. M. A. Rasul et al.

significant market share during the past decade globally, recording a total of 627 GW capacity, both on- and off-grid, in 2019 and self-consumption PV systems with battery storage has been a key driver contributing to this capacity expansion while most countries require related policies, regulatory frameworks, and support schemes for implementing solar PV systems [30]. Electricity generation in Sri Lanka majorly depends on fossil fuels, 55% of the total energy generation, encouraging the national policy to diversify energy production to renewable sources [31]. Sri Lankan government along with the CEB introduced different rooftop solar business models, such as net metering, which allows users to offset the consumption with the electricity generated from the PV system, net accounting, which allows users to be paid in cash for surplus electricity generated, and net plus, allows users to separate the consumption and generation and manage the finances accordingly, since 2008 and further provided subsidies and tax exemptions to install solar PV systems [31]. The national energy policy of Sri Lanka emphasizes the need to provide access to affordable, reliable, and sustainable energy for all aligning with the Sustainable Development Goals (SGDs) of the United Nations (UN) [32]. However, the existing power supply for the households in Sri Lanka has many restrictions as discussed earlier in this paper, technical and otherwise, accommodating solar PV systems such as the capacity of the transformer (TF), the distance from the TF, the reluctance of the service providers to upgrade the system etc. Due to these restrictions, only a limited number of users can install PV systems, even with the TF capacity increased, which demotes social equity as well as the potential for using clean energy. The BES eliminates the above technical restrictions of installing PV systems, where all users served by the national grid are provided with the opportunity to enjoy the benefits of solar PV energy. With the existing prices of PV systems in Sri Lanka, the BES doubles the cost of a regular PV system which is a substantial investment for a regular household user. Therefore, an economic analysis based on the Discounted Cash Flow (DCF) method is conducted using the Benefit-Cost Ratio (B/C), the Net Present Value (NPV), the Internal Rate of return (IRR) and the payback period as the indices. Benefit-Cost (B/C) ratio measures the net benefits (or savings) of a project or investment relative to its net cost using the present worth of all the cash flows [33], which is expressed mathematically in Eq. 1, where n = the number of years over which the benefits and costs are analysed, Bi = benefits in the year i, C i = costs in the year i, and d = the discount rate. B/C ratio =

 n  0

  n   i Bi /(1 + d) / Ci /(1 + d) i

(1)

0

For an investment to be economically viable or worthwhile, the benefits gained must outweigh the costs incurred, where the B/C ratio will be greater than unity [34]. Net Present Value (NPV) is the total or the sum of the real cash flows from a project or an investment over time [35], which is expressed mathematically as

11 Cost-Benefit and Short-Term Power Flow Analysis …

261

follows (Eq. 2), where n = the number of years over which the benefits and costs are analysed, Bi = benefits in the year i, C i = costs in the year i, and d = the discount rate. NPV =

n i=0

(Bi − Ci ) (1 + d)i

(2)

Although, as a rule of thumb, the investments with positive NPV are economically viable, the positive NPV alone does not necessarily determine whether the investment is worthwhile [36] unless the investments have significant NPV values [35]. Internal Rate of Return (IRR) is the discount rate that makes the NPV equal to zero in a DCF analysis which is used to estimate the profitability of investments [37], which is expressed mathematically as follows (Eq. 3), where n = the number of years over which the benefits and costs are analysed, Bi = benefits in the year i, C i = costs in the year i, C 0 = total initial investment. 0 = NPV =

n i=0

(Bi − Ci ) − C0 (1 + IRR)i

(3)

Payback is the time required for a return on an investment indicating the financial profitability of a project [38], which can be expressed mathematically as follows (Eq. 4), where C 0 = total initial investment, Bavg = average annual benefits and C avg = average annual costs. Payback in years =

C0 Bavg − Cavg

(4)

The economic analysis considers a regular urban area in Sri Lanka to measure the economic viability of installing the proposed PV systems. The average household electricity consumption of Sri Lanka is approximately 248 kWh per month in 2016 [39] which amounts to a monthly electricity bill of $40.73 [40]. Considering the average monthly consumption of a household, a PV system with a capacity of 3 kW is considered with the cost of installing the regular PV system is approximately $3,068.60, while the PV system with the battery (2.5 kWh) amounted to $5,118.33 [41]. A project lifetime of a PV-battery residential energy system is taken as 10 years considering the lifetime of a battery module is limited to 10 years, and a discount rate of 10% is applied for the calculations. Since the battery lifetime is considered as 10 years, therefore no replacement costs of the system components are used in the analysis. Since the power generation in Sri Lanka depends on coal, fossil fuels and hydropower plants, 0.09 $/kWh is considered as the cost of the environmental impact for the analysis based on previous studies [42]. The average annual exchange rate of 2019, 1$ = 178.74 LKR, is used for the analysis [43]. The average urban area selected for the economic evaluation consists of 178 households which are served by a 100 kVA transformer (T/F), the most common T/F type used in urban areas. This T/F can accommodate 14.6% of the users to install PV systems while the rest

262

M. J. M. A. Rasul et al.

cannot be accommodated without upgrading the T/F. The cost of up-grade amounts approximately to $12,430.14, which is calculated from the data acquired from CEB. Although the PV system can be utilised as a revenue generator for individual users, the following economic analysis only considers the cost-saving gained by offsetting the monthly electricity bill to zero. The economic analysis is conducted in four steps. Firstly, the four economic ratios are calculated for an individual user to install the PV system with a battery and secondly for a regular urban area in Sri Lanka with the limitations of T/F capacity. Thirdly, the economic ratios are calculated for an individual user with the cost of the battery being subsidized considering the financial assistance provided by the Sri Lankan government to promote the use of PV energy. Finally, a sensitivity analysis is conducted to account for the possible variances of the discount rate used as well as the benefit and cost factors considered.

4 Results and Discussion 4.1 Technical Performance Evaluation of PV-Battery Residential Energy System The overall system response of the PV-battery energy system for a dynamic residential load profile is investigated under two specific PV conditions: variable PV conditions and zero PV conditions. It is a vital fact to understand the system power flow relation in the system investigation process as represented in Fig. 8. Hence, concerning the PV and battery systems, a positive sign is set for power production and a negative sign is for power consumption of those systems. Consequently, the opposite sign is set for power injection and consumption of the utility grid and home load.

Fig. 8 Sign convention of system power flow relation

11 Cost-Benefit and Short-Term Power Flow Analysis …

4.1.1

263

Short-Term Power Flow Analysis Under Variable PV Conditions

The system is analysed separately under variable PV conditions for two battery state of charge (SOC) conditions where the initial battery SOC is set at 75 and 100%. The short-term (10 s) load variations with PV output are considered for analysing the power flow and battery SOC. The short-term transient system performance for power flow is depicted in Figs. 9 and 10 for each battery SOC conditions respectively. It can be observed that the system with 75% battery SOC level under dynamic PV conditions, the residential load is fed by the PV power, and surplus PV power is utilised to charge the battery in high PV power productions, while both the PV and battery system simultaneously serve on-demand residential load when the PV

Fig. 9 System power flow at 75% battery SOC

Fig. 10 System power flow at 100% battery SOC

264

M. J. M. A. Rasul et al.

power is insufficient in low PV power production situations as in Fig. 9. Also, when the battery is in a fully charged condition (battery SOC is at 100%), the battery stops charging and surplus PV power is fed into the utility grid. In the circumstances where PV power is not adequate to meet the on-demand residential load, the necessary support from the battery system is provided for the PV system to drive the home load appropriately as in Fig. 10.

4.1.2

Short Term Power Flow Analysis Under Zero PV Conditions

Similarly, the overall system performance is analysed separately under zero PV conditions for two different battery state of charge (SOC) conditions where the initial battery SOC is set at 75 and 0% and is illustrated in Fig. 11 and Fig. 12 for each battery conditions respectively. It is noticeable fact that, under zero PV conditions, the battery system with 75% SOC fulfils on-demand residential load subjected to its maximum capacity as in Fig. 11, and the utility grid acts as a backup power source that concurrently provides support to the battery system to power the residential loads when they exceed the maximum capacity of the battery system. Furthermore, when the system with fully discharged battery state (0% battery SOC) under zero PV conditions, the utility grid is active to completely feed the residential loads as in Fig. 12. According to the simulation results of the overall system response, the PV-battery energy system steadily serves the dynamic residential loading conditions under different PV and battery stages by maximizing self-consumption. Besides, the utility grid is actively available as a backup power source to supply power to meet the on-time residential demand operating simultaneously together with or without the PV-battery system whenever the PV-battery grid connected energy system is not

Fig. 11 System power flow at 75% battery SOC

11 Cost-Benefit and Short-Term Power Flow Analysis …

265

Fig. 12 System power flow at 0% battery SOC

capable to handle the residential load conditions. Overall, the battery coupled residential PV system technically performs well with its system functions while mostly reducing the interaction with the utility grid, which yields to eliminate the technical impacts coursed by the grid-connected PV systems in LV distribution networks.

4.2 Economic Viability of PV-Battery Residential Energy System in Sri Lanka According to the results of the economic analysis indicated in Table 3, the proposed PV system with the BES is not financially viable for the individual user with respect to the related costs and benefits, as the cost of the BES doubling the price of the PV system. In the case of using the proposed system for additional revenue generation, which is not considered in this analysis, can increase the benefits gained by the individual user providing the financial viability. However, at the macro level, in terms of the costs and benefits to the economy, the proposed PV system is economically viable with a B/C ratio of 1.42, an NPV of $356,924, an IRR of 22.29% and a Table 3 Results of the economic analysis Step 1 individual user B/C ratio

0.71

NPV ($)

−1486.45

Step 2 Urban area 1.42 356,924

IRR (%)

1.10

22.29

Payback period (Years)

9.6

4.8

Step 3 individual user (subsidized) 1.18 563.29 15.36 5.7

266

M. J. M. A. Rasul et al.

payback period of 4.8 years. Considering the economic gains at the macro level and to promote social equity, it can be recommended for the government to subsidize the cost of the BES providing all users in the urban area the opportunity to utilize PV systems, which leads to the third step of this analysis providing financial viability for individual users with a B/C ratio of 1.18, an NPV of $563.29, an IRR of 15.36% and a payback period of 5.7 years. Since the economic analysis assumes the discount rate at 10%, a sensitivity analysis is conducted to account for the variance of the indices used in the economic analysis [24]. Table 4 indicates the change in the B/C ratio and NPV as the discount rate moves from 7 to 13%. Table 5 indicates the results of the sensitivity analysis with the change of cost from −20 to +20% to account for the possible variance of related cost factors considered in the economic analysis. Table 6 indicates the results of the sensitivity analysis with the change of benefits from −20 to +20% to account for the possible variance of related benefits considered in the economic analysis. The results of the sensitivity analysis indicate the economic viability of the proposed PV system in the selected urban area, even in the worst-case scenarios accounting the variance of the discount rate, possible increment of costs and the reduction of benefits. However, for the individual user, the proposed PV system Table 4 Sensitivity analysis of discount rate Discount rate (%)

Step 2 Urban area

Step 3 individual user (subsidized)

B/C ratio

NPV ($)

B/C ratio

NPV ($)

7

1.57

491,434

1.32

969.63

8

1.52

443,702

1.27

825.44

9

1.47

398,942

1.22

690.22

10

1.42

356,924

1.18

563.29

11

1.37

317,439

1.14

444.02

12

1.33

280,300

1.11

331.82

13

1.29

245,330

1.07

226.18

Table 5 Sensitivity analysis of costs Cost change (%)

Step 2 Urban area

Step 3 individual user (subsidized)

B/C ratio

NPV ($)

IRR (%)

B/C ratio

NPV ($)

IRR (%)

−20

1.77

528,477

33.14

1.48

1,177.01

23.97

−10

1.57

442,700

27.03

1.31

870.15

19.18

0

1.42

356,924

22.29

1.18

563.29

15.36

10

1.29

271,147

18.48

1.08

256.43

12.23

20

1.18

185,370

15.33

0.99

−50.43

9.60

11 Cost-Benefit and Short-Term Power Flow Analysis …

267

Table 6 Sensitivity analysis of benefits Cost change (%)

Step 2 Urban area

Step 3 individual user (subsidized)

B/C ratio

NPV ($)

IRR (%)

B/C ratio

NPV ($)

IRR (%)

−20

1.13

113,985

13.94

0.95

−163.08

−10

1.27

235,454

18.10

1.07

200.10

11.92

0

1.42

356,924

22.29

1.18

563.29

15.36

10

1.56

478,393

26.55

1.30

926.48

18.79

20

1.70

599,862

30.91

1.42

1,289.67

22.24

8.42

is financially not viable when higher costs and lower benefits are accounted. In summary, the solar-battery energy system is economical viable for urban areas in Sri Lanka. Although the proposed system is not financially viable for individual consumers, due to the cost of the battery increasing the total price significantly, financial assistance from the government subsidizing the cost of the battery can improve the viability.

5 Conclusion This paper has presented a technical performance investigation and an economic viability assessment of a residential PV-battery grid connected energy system. The overall system with a 3 kW PV system and a 2.5 kWh battery system is implemented to maximize the PV self-consumption while curtailing the power injection to eliminate critical issues caused in the LV distribution network during the period of high PV penetration. The system control algorithm is designed to perform effectively by managing PV, battery and utility grid systems and their functions in different operating stages while serving the dynamic residential loading conditions (as in Fig. 8) under different PV and battery conditions. Stable system performance is observed from the obtained results for each investigated condition with the minimum grid support. Thus, the crucial issues experienced on the LV distribution network due to the grid-connected PV systems can be mitigated with the PV-battery residential energy systems at the consumer level rather than implementing expensive system upgrades by the grid operators. With the regulations initiated to install residential PV systems in Sri Lanka to control the high penetration residential PV system effect, a limited number of consumers can accommodate residential PV systems under a particular distribution substation. Moreover, consumers have not still been granted permission to install residential PV-battery residential energy systems by CEB. Therefore, through this research study, the economic feasibility of a PV-battery energy system is assessed with the approach of a real case scenario. The PV-battery energy system is economically viable for urban areas in Sri Lanka which supports the national energy policy providing access to affordable, reliable, and sustainable energy for all. Considering

268

M. J. M. A. Rasul et al.

the alignment to the policy framework, the economic viability at the macro economic environment, and the possibility of promoting social equity, it can be recommended for the government to subsidise the cost of the proposed PV system, which provides the financial viability for individual users to implement the BES. Further, using the PV system to generate income for the individual users, in addition to offsetting the monthly electricity consumption, can strengthen the financial viability of the proposed PV system, which can be considered for further analysis. Acknowledgements Authors would like to acknowledge Mr. S. M. K. Parsad, Engineer, Ceylon Electricity Board and M. N. M. Fazlan, Engineer, Sri Lanka Telecom PLC for providing necessary information to conduct the economic assessment section in this research study.

References 1. Renewable Capacity Statistics (2020) International Renewable Energy Agency. https://www. irena.org/publications/2020/Mar/Renewable-Capacity-Statistics-2020. Accessed 07 Dec 2020 2. Mirhassani S, Ong H, Chong W, Leong K (2015) Advances and challenges in grid tied photovoltaic systems. Renew Sustain Energy Rev 49:121–131 3. Alboaouh K, Mohagheghi S (2020) Impact of rooftop photovoltaics on the distribution system. J Renew Energy 2020:1–23 4. Mawarni D, Ali M, Nguyen P, Kling W, Jerele M (2015)A case study of using OLTC to mitigate overvoltage in a rural European low voltage network. In: 50th international universities power engineering conference (UPEC) 5. Navarro-Espinosa A, Ochoa L (2015) Increasing the PV hosting capacity of LV networks: OLTC-fitted transformers vs. reinforcements. In: 2015 IEEE power & energy society innovative smart grid technologies conference (ISGT) 6. Liu X, Aichhorn A, Liu L, Li H (2012) Coordinated control of distributed energy storage system with tap changer transformers for voltage rise mitigation under high photovoltaic penetration. IEEE Trans Smart Grid 3(2):897–906 7. Xu X, Xu Z, Lyu X, Li J (2018) Optimal SVC placement for maximizing photovoltaic hosting capacity in distribution network. IFAC-PapersOnLine 51(28):356–361 8. Haque MM, Wolfs P (2016) A review of high PV penetrations in LV distribution net-works: present status impacts and mitigation measures. Renew Sustain Energy Rev 62:1195–1208 9. Etherden N, Bollen M (2014) Overload and overvoltage in low-voltage and medium-voltage networks due to renewable energy—Some illustrative case studies. Electr Power Syst Res 114:39–48 10. Rasul M, Kolhe M (2019)Reactive power control of a 3-phase grid-connected building integrated photovoltaic system. IOP Conf Ser: Mater Sci Eng 605:012012 11. Kolhe M, Rasul M (2020) 3-Phase grid-connected building integrated photovoltaic system with reactive power control capability. Renew Energy 154:1065–1075 12. Alam MJE, Muttaqi KM, Sutanto D (2013) Mitigation of rooftop solar PV impacts and evening peak support by managing available capacity of distributed energy storage systems. IEEE Trans Power Syst 28(4):3874–3884 13. Nair NC, Garimella N (2010) Battery energy storage systems: assessment for small-scale renewable energy integration. Energy Build 42:2124–2130 14. Ueda Y, Kurokawa K, Tanabe T, Kitamura K, Akanuma K, Yokota M (2007)Study on the over voltage problem and battery operation for grid-connected residential PV systems. In: Proceedings of the 22nd European photovoltaic solar energy conference

11 Cost-Benefit and Short-Term Power Flow Analysis …

269

15. Alam MJE, Muttaqi KM, 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 Trans Energy Convers 29(2):507–518 16. Farinet D, Maurer M, Vacca L, Spataru S, Stroe D (2019) Battery lifetime analysis for residential PV-battery system used to optimize the self consumption—A danish scenario. In: 2019 IEEE energy conversion congress and exposition (ECCE) 17. Naumann M, Karl R, Truong C, Jossen A, Hesse H (2015) Lithium-ion battery cost analysis in PV-household application. Energy Procedia 73:37–47 18. Schopfer S, Tiefenbeck V, Staake T (2018) Economic assessment of photovoltaic battery systems based on household load profiles. Appl Energy 223:229–248 19. Bruch M, Müller M (2014) Calculation of the cost-effectiveness of a PV battery system. Energy Procedia 46:262–270 20. Kolhe M, Kolhe S, Joshi J (2002) Economic viability of stand-alone solar photovoltaic system in comparison with diesel-powered system for India. Energy Econ 24(2):155–165 21. Kolhe M, Ranaweera K, Gunawardana A (2015)Techno-economic sizing of off-grid hybrid renewable energy system for rural electrification in Sri Lanka. Sustain Energy Technol AssessS 11:53–64. https://doi.org/10.1016/j.seta.2015.03.008 22. Sharma P, Kolhe M, Sharma A (2020)Economic performance assessment of building integrated photovoltaic system with battery energy storage under grid constraints. Renew Energy 145:1901–1909. https://doi.org/10.1016/j.renene.2019.07.099 23. Statistical digest (2018) Ceylon electricity board, 2019. https://ceb.lk/front_img/img_reports/ 1567487133Statistical_Digest_2018.pdf 24. Statistical digest (2019) Ceylon electricity board, 2020. https://ceb.lk/front_img/img_reports/ 1601877736Statistical_Digest_2019_Web_Version.pdf 25. De Silva H, Jayamaha D, Lidula N (2019) Power quality issues due to high penetration of rooftop solar PV in low voltage distribution networks: a case study. In: 2019 14th conference on industrial and information systems (ICIIS) 26. Annual Report (2018) Ceylon electricity board, 2019. https://ceb.lk/front_img/img_reports/ 1605765148CEB_Annual_Report_2018.pdf 27. Cucchiella F, D’Adamo I, Gastaldi M, Stornelli V (2018) Solar photovoltaic panels combined with energy storage in a residential building: an economic analysis. Sustainability 10(9):3117 28. Sandor D, Fulton S, Engel-Cox J, Peck C, Peterson S (2018) System dynamics of polysilicon for solar photovoltaics: a framework for investigating the energy security of renewable energy supply chains. Sustainability 10(2):160 29. Zhang Y, Zhao X, Zuo Y, Ren L, Wang L (2017) The development of the renewable energy power industry under feed-in tariff and renewable portfolio standard: a case study of china’s photovoltaic power industry. Sustainability 9(4):532 30. Report R (2020) Renewables 2020 global status report—REN21, REN21. https://www.ren21. net/reports/global-status-report/. Accessed 08 Jan 2021 31. Rooftop Solar Power Generation Project: Report (2017) Asian Development Bank. https://www.adb.org/sites/default/files/linked-documents/50373-002-sd-01%20redacted. pdf. Accessed 13 Jan 2021 32. National Energy Policy and Strategies of Sri Lanka (2019) Ministry of Power, Energy and Business Development. http://powermin.gov.lk/english/wp-content/uploads/2019/11/2135-61_E. pdf. Accessed 13 Jan 2021 33. Krarti M (2018)Chapter 5—Life-cycle cost and energy productivity analyses. In: Optimal design and retrofit of energy efficient buildings, communities, and urban centers. ButterworthHeinemann, pp 247–312 34. Cheremisinoff NP, Rosenfeld P, Davletshin AR (2008) Chapter 8—Environmental economics. In: Responsible Care. Gulf Publishing Company, pp 477–513 35. Flaig J (2005) Improving project selection using expected net present value analysis. Qual Eng 17(4):535–538 36. Ross S (1995) Uses, abuses, and alternatives to the net-present-value rule. Financ Manag 24(3):96

270

M. J. M. A. Rasul et al.

37. Patrick M, French N (2016) The internal rate of return (IRR): projections, benchmarks and pitfalls. J Prop Invest Financ 34(6):664–669 38. Port and Terminal Management (2020) Institute of chartered shipbrokers 39. Household electricity consumption per capita in Sri Lanka from 2000 to 2016 (2018) Statista. https://www.statista.com/statistics/598346/household-consumption-of-electr icity-per-capita-in-sri-lanka/. Accessed 23 Dec 2020 40. HOW CAN I CALCULATE MY BILL? (2020) Ceylon Electricity Board. https://ceb.lk/billcalculation-commercial/en. Accessed 23 Dec 2020 41. Solar Life Technologies|Solar PV Company (2020). https://solarlife.lk/. Accessed 18 Dec 2020 42. Karkour S, Ichisugi Y, Abeynayaka A, Itsubo N (2020) External-cost estimation of electricity generation in G20 countries: case study using a global life-cycle impact-assessment method. Sustainability 12(5):2002 43. International Financial Statistics (2019) International Monetary Fund. https://data.imf.org/? sk=4c514d48-b6ba-49ed-8ab9-52b0c1a0179b&sId=1409151240976. Accessed 23 Dec 2020