Power Swing Detection and Generator Out-of-Step Protection Under Renewable Power Source Integration 9811995451, 9789811995453

Citation preview

Energy Systems in Electrical Engineering Series Editor Muhammad H. Rashid, Florida Polytechnic University, Lakeland, USA

Energy Systems in Electrical Engineering is a unique series that aims to capture advances in electrical energy technology as well as advances electronic devices and systems used to control and capture other sources of energy. Electric power generated from alternate energy sources is getting increasing attention and supports for new initiatives and developments in order to meet the increased energy demands around the world. The availability of computer–based advanced control techniques along with the advancement in the high-power processing capabilities is opening new doors of opportunity for the development, applications and management of energy and electric power. This series aims to serve as a conduit for dissemination of knowledge based on advances in theory, techniques, and applications in electric energy systems. The Series accepts research monographs, introductory and advanced textbooks, professional books, reference works, and select conference proceedings. Areas of interest include, electrical and electronic aspects, applications, and needs of the following key areas: • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •

Biomass and Wastes Energy Carbon Management Costs and Marketing Diagnostics and Protections Distributed Energy Systems Distribution System Control and Communication Electric Vehicles and Tractions Applications Electromechanical Energy Conversion Energy Conversion Systems Energy Costs and Monitoring Energy Economics Energy Efficiency Energy and Environment Energy Management, and Monitoring Energy Policy Energy Security Energy Storage and Transportation Energy Sustainability Fuel Cells Geothermal Energy Hydrogen, Methanol and Ethanol Energy Hydropower and Technology Intelligent Control of Power and Energy Systems Nuclear Energy and Technology Ocean Energy Power and Energy Conversions and Processing Power Electronics and Power Systems Renewable Energy Technologies Simulation and Modeling for Energy Systems Superconducting for Energy Applications Tidal Energy Transport Energy

The books of this series are reviewed in a single blind peer review process by the main editor and/or the advisory board.

Jignesh Kumar P. Desai · Vijay Makwana

Power Swing Detection and Generator Out-of-Step Protection Under Renewable Power Source Integration

Jignesh Kumar P. Desai Department of Electrical Engineering Ganpat University Mehsana, India

Vijay Makwana Department of Electrical Engineering G H Patel College of Engineering & Technology Anand, India

ISSN 2199-8582 ISSN 2199-8590 (electronic) Energy Systems in Electrical Engineering ISBN 978-981-19-9545-3 ISBN 978-981-19-9546-0 (eBook) https://doi.org/10.1007/978-981-19-9546-0 © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 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

Dedicated to my beloved parents, my brother Ankur Desai, my wife Mruga Desai and my son Darsh Desai for their love and constant encouragement. —Jigneshkumar P. Desai Dedicated to my wife Manisha and son Tirth for their patience and encouragement. —Vijay H. Makwana

Foreword

Power swing detection and out-of-step protection are critical components in preventing significant power system disruptions. With the right generator out-ofstep relaying decision, power system stability can be improved. The existing power system has a higher percentage of renewable energy sources, which complicates power system operation and protection. The use of artificial intelligence techniques in power system protection necessitated a growing amount of exploration in order to make power system protection adaptive. Furthermore, the wide-area measuring technique distinguishes the protective relay application from others. Dr. Jigneshkumar P. Desai and Dr. Vijay H. Makwana’s book goes into great detail about blackouts, adaptive generator relaying schemes for blackout prevention, advanced transmission line protection schemes, and artificial intelligence-based protection of major power system components with renewable integration. This book, in my opinion, is a fantastic resource for electrical engineering postgraduates, graduate students, and research scientists. Professional protection engineers may also find it beneficial. Out-of-step relaying and distance relaying are all discussed. It also includes a wavelet transform-based relaying mechanism as well as deep learning capabilities. I commend the authors on authoring Power Swing Detection and Generator Outof-Step Protection Under Renewable Power Source Integration, and I really hope that students, researchers, teachers, and practising engineers will profit from it. June 2022

Dr. Hitesh Jariwala Professor of Electrical Department Sardar Vallabhbhai National Institute of Technology Surat, Gujarat, India

vii

Preface

Due to the integration of renewable energy sources, the modern power system is heavily reliant on nature. The countries affected by the widespread blackout confront economic, social, and political difficulties. If blackouts occur more frequently, the economic loss inhibits the nation’s progress. The monitoring and protection of the power system are crucial for the system’s reliable operation. This role is handled by a modern protective relay, also known as an intelligent electronic device. Small stable power oscillations happen when the generator rotor speeds up or down when re-equalizing electric output power to mechanical input power for individual units in response to changes in load, generation, or system state. Oscillations in power systems occur as a result of fast variations in load, transmission line switching, generator tripping or removal from operation, and the occurrence of faults in a system, among other things. Given the aforementioned issue, it is critical to provide extensive information on the causes of blackout and how to prevent it utilizing adaptive protection principles. The book’s information can aid research students, graduate students, and undergraduate students in the testing, development, and modeling of protective relays for generators, transformers, and transmission lines. The reasoning and literature survey in Chap. 1 include contemporary breakthroughs, approaches, and procedures. Finally, it demonstrates a research gap. Background theory and a constructive literature assessment are presented in Chap. 2 to prepare the reader for a deep dive into the suggested implementation. Chapter 3 covers fundamentals of generator protection. PMU with a polygon-shaped graphical method for out-of-step protection of the synchronous generator was proposed in Chap. 4. Chapter 5 discussed predictive out-of-step protection, which forecasts the power swing based on some of its unique qualities before it occurs. Chapter 6 offered a final problem solution based on Wavelet transform and deep learning machine model. Finally, Chap. 7 summarized all of the strategies and their practical usefulness, balancing superiority, and limitation. Surat, India May 2022

Dr. Jignesh Kumar P. Desai Dr. Vijay Makwana

ix

Acknowledgements

We are grateful to God and our parents for their blessings, without which this project would not have been possible. We are grateful to Gujarat Technological University as the content of the published work is part of the thesis submitted in partial fulfillment for award of the degree of Ph.D. in Electrical Engineering of Gujarat Technological University. This endeavor would not have been possible without the help of family and friends. We would want to express our gratitude to family members and friends for their support and patience. Dr. Hitesh R. Jariwala (SVNIT, Surat) and Dr. (Prof.) Mahmadasraf A. Mulla (SVNIT, Surat) deserve special gratitude for their direction, monitoring, ongoing encouragement, and necessary review at various stages of the project. The blessings, assistance, and direction they provide from time to time will go with us a long way in life.

xi

Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2.1 Literature Review on Most Recent Work . . . . . . . . . . . . . . . . 1.2.2 Literature Review on Different Methods of Out-of-Step Protection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2.3 Literature Review on Impact of Renewable Power Sources on Protective Relays . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1 1 4 4 4 9 10 11

2 Power Swings with Distance Relay and PMU . . . . . . . . . . . . . . . . . . . . . 2.1 Mathematical Analysis of the Research Problem . . . . . . . . . . . . . . . . 2.2 Phasor Measurement Unit and WAMS . . . . . . . . . . . . . . . . . . . . . . . . . 2.3 Power Swing Detection and Out-of-Step Detection Technique . . . . 2.3.1 Double Blinder Based Technique . . . . . . . . . . . . . . . . . . . . . . . 2.3.2 Distance Relay and Out-of-Step Relay for Power Swing Blocking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

15 15 18 20 20 20 22 22

3 Protection of Synchronous Generator . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1 Class of Generator Protection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.1 Class A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.2 Class B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.3 Class C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Faults and Abnormal Conditions with Its Protection . . . . . . . . . . . . . 3.2.1 Insulation Failure of the Stator Winding . . . . . . . . . . . . . . . . . 3.2.2 Insulation Failure of the Rotor Winding . . . . . . . . . . . . . . . . . 3.2.3 Uneven Loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.4 Field Failure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.5 Prime Mover Failure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

25 25 26 26 26 26 27 35 36 38 40 xiii

xiv

Contents

3.2.6 Over Load Protection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.7 Over Voltage Protection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.8 Over Speed Protection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

42 42 43 44 44

4 Adaptive Out-of-Step Protection of Synchronous Generator . . . . . . . . 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 Mathematical Analysis of Proposed Out-of-Step Protection Scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3 Proposed PMU-Incorporated Polygon-Shaped Out-of-Step Protection Scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4 Implementation and Testing of the Proposed Algorithm . . . . . . . . . . 4.4.1 Event-1: Fault Remove Before CCT . . . . . . . . . . . . . . . . . . . . 4.4.2 Event-2: Fault Remove After the CCT . . . . . . . . . . . . . . . . . . 4.5 Testing of Proposed Technique for LG, LL Faults and Swing Center Arise Inside Generator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.6 Performance Test of the Proposed Algorithm . . . . . . . . . . . . . . . . . . . 4.6.1 Test Under Different Level of Renewable Power Penetration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.6.2 Test Using Real Physical System . . . . . . . . . . . . . . . . . . . . . . . 4.7 Comparison of the Proposed Relaying Algorithm with Existing Ones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.8 Limitation of the Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.9 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

45 45

5 Predictive Out of Step Protection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 Modeling of Synchronous Generator with the Turbine, Exciter, AVR, Governor and PSS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3 Power Swing Detection Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.1 Repetitiveness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.2 Continuation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.3 Regularity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.4 Impedance Travel Through Outer to Inner Layer . . . . . . . . . . 5.4 Out of Step Protection of Synchronous Generator . . . . . . . . . . . . . . . 5.4.1 Swing Center Appear in the Middle of the Transmission Line Section . . . . . . . . . . . . . . . . . . . . . . . 5.4.2 Swing Center Arise Near the Generator Transformer . . . . . . 5.5 Comparison with Existing Methodology . . . . . . . . . . . . . . . . . . . . . . . 5.6 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

48 51 53 53 55 55 57 59 59 61 62 62 68 71 71 73 74 75 76 77 77 77 78 79 80 81 81 82

Contents

6 Wavelet Transform and Deep Learning Machine Model-Based Out-of-Step Relay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2 The System and Power Swing Conditions . . . . . . . . . . . . . . . . . . . . . . 6.3 Development of the Proposed Algorithm . . . . . . . . . . . . . . . . . . . . . . . 6.3.1 Training Data Selection and Pre-processing . . . . . . . . . . . . . . 6.3.2 Design of the Mathematical Structure of Deep Learning Neural Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3.3 The Design of the Proposed Relaying Algorithm . . . . . . . . . 6.3.4 Training, Validation, and Testing . . . . . . . . . . . . . . . . . . . . . . . 6.4 Performance Validation by Unknown Events . . . . . . . . . . . . . . . . . . . 6.4.1 Test Using Unknown Data of Different Fault Location . . . . . 6.5 Development and Performance on a 29-Bus System . . . . . . . . . . . . . 6.6 Comparison of the Proposed Algorithm . . . . . . . . . . . . . . . . . . . . . . . . 6.7 Discussions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

xv

85 85 88 89 89 93 95 96 98 98 99 101 103 103 105

7 Out-of-Step Protection Schemes Summary and Future Scope . . . . . . . 107 7.1 Out-of-Step Protection Schemes Summary . . . . . . . . . . . . . . . . . . . . . 107 7.2 Future Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 Appendix A: Kundur Two-Area System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 Appendix B: Type-IV Wind Turbine System Parameter for Modified Kundur Two-Area System . . . . . . . . . . . . . . . . . 113 Appendix C: Indian Grid System’s Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117 Appendix D: A 29-Bus Hydro-Quebec System . . . . . . . . . . . . . . . . . . . . . . . . 119

About the Authors

Dr. Jignesh Kumar P. Desai received his Ph.D. in Electrical Engineering from Gujarat Technological University, India, in 2022. He is working as an Assistant Professor in the Electrical Department at the U.V. Patel College of Engineering at Ganpat University, India. He has 10 years of experience as an Assistant Professor. He worked in the industry for one year and six months. He has so far had six research articles published in peer-reviewed journals that are indexed by SCI/Scopus. More than four research publications were presented and published at international IEEE conferences. He is invited to be a reviewer for the IEEE Transaction on Power Systems. He has published two books and two book chapters with an international publisher. He has filed and published one patent and one design with the Indian Patent Office (IPO). Power system protection and AI for power system protection are some of his research interests. Dr. Vijay Makwana is a Professor in the Electrical Department of G H Patel College of Engineering & Technology, Vallabh Vidyanagar. He completed BE (Electrical Engineering) and ME (Electrical Power Systems) from BVM Engineering College, Sardar Patel University, in 1999 and 2002. He obtained Ph.D. in Electrical Engineering from Sardar Patel University in 2013. He has 20 years of teaching experience. His research interests include power system protection, reactive power compensation and FACTS. During his academic career, he guided 70 BE, 7 ME, and 2 Ph.D. students. Presently, he is guiding 3 Ph.D. students studying at Gujarat Technological University. He has published 18 papers in international peer-reviewed SCI and Scopus-indexed journals and 10 papers in international conferences. He has written a book “Power System Protection and Switchgear” published by Tata McGraw Hill in 2010 and a monograph “Transmission Line Protection Using Digital Technology” published by Springer in 2016.

xvii

Abbreviations

AFC ANN AVR CCT CPU CT CTR DB DFIG DL DSC DWT ER FFT GB GOV GTU HIL HV HVDC IEEE IIT MATLAB MB MTL NER NR OOS PMU PSB PSCAD

Automatic Frequency Control Artificial Neural Network Automatic Voltage Regulator Critical Clearing Time Central Processing Unit Current Transformer Current Transformer Ratio Double Blinder Doubly-Fed Induction Generator Deep Learning Digital Signal Controller Discrete Wavelet Transforms Eastern Region Fast Fourier Transform Generator Breaker Governor Generator Transformer Unit Hardware In Loop High Voltage High-Voltage Direct Current Institute of Electrical and Electronics Engineering Indian Institute of Technology Matrix Laboratory Multi-band Montreal Northeast Region Northern Region Out Of Step Phasor Measurement Unit Power Swing Blocking Power System Computer-Aided Design xix

xx

PSP PSS PT PV RAM RPES SMIB SR SSD SVC SVM TFC TMS USA WAFMS WAMS WR

Abbreviations

Pole Slip Protection Power System Stabilizer Potential Transformer Photovoltaic Random Access Memory Renewable Power Energy Sources Single Machine Infinite Bus Southern Region Solid-State Drive Swing Voltage Center Support Vector Machine Through Fault Current Transient Monitoring Function United States of America Wide-Area Frequency Measurement System Wide-Area Monitoring System Western Region

Symbols

A ai aii B bi C Cj C ptr D1 D2 dii Ef Eg ER Es ES fg fs H HG I I0 I1 I1 I2 I2 Idi f f Ie If I f max IR

Total area Tan-sigmoid transfer function output ith Approximate co-efficient of wavelet transform Line susceptance Bias co-efficient Continuation threshold Criteria of stability Stability margin Entering angle at right slope 1 Entering angle at right slope 2 ith Detailed co-efficient Desired field voltage Voltage behind transient reactance of generator G 1 Voltage magnitude of source B or grid System voltage Voltage magnitude of source A Power generation frequency Rated system frequency of the system in Hz Inertia constant Inertia constant of generating plant MW-s/MVA Positive sequence current Zero sequence current Positive sequence current Secondary-side current of primary-side CT of transformer Negative sequence current Secondary-side current of secondary-side CT of transformer Differential current CT magnetizing current Field current Maximum earth-fault current Relay current xxi

xxii

Ir elay Is Istab IT JQ K K0 l N ni PG Pg1 Pg2 PL Pr e f Q g1 Q g2 R RR R ST S SG T1 Ts V V0 Vcr Vm Vr VR Vr Vr elay Vs VT wi Wr e f X ( ji) X d’ Xg Xl XT X t’ Xt yj Z

Symbols

Transmission line current measured by the relay Relay spill current setting Stabilizing current Terminal current Threshold value of transient voltage magnitude of the line Biasing constant of bias differential relay Spring constant of bias differential relay Leakage reactance No. of turns First hidden layer output Generation output power in MW Active power (PV bus) Active power (slack bus) Loading of the system in MW Machine reference active power Reactive power (PV bus) Reactive power (slack bus) Line resistance Burden resistance of CT Stabilizing resistor Regularity threshold of wavelet transform Generator rating in MVA Time measured between right slopes 2 and 1 Set delay time Positive sequence voltage Nominal voltage at the bus Critical voltage at bus Maximum value of voltage Machine reference voltage Voltage across the relay Voltage at receiving bus Busbar voltage measured by the relay Voltage at sending bus Terminal voltage Weight co-efficient Reference angular frequency Input vector Transient reactance for generator (direct axis) The equivalent reactance of the generator Line reactance Reactance of the transformer Transient reactance for transformer Transmission line reactance Output of Deep Learning network at jth sample Positive sequence impedance

Symbols

Z0 Z1 Z1 Z2 Z2 Z3 Zg Zn ZS Zs Z seen ZT δ δA Z θ θ1 θ2 θmax σ (ni) φ φm φR

xxiii

Zero sequence impedance Positive sequence impedance Zone 1 Negative sequence impedance Zone 2 Zone 3 Grid impedance Neutral impedance Source internal impedance System impedance Impedance measured by the relay Total impedance from source to grid Power angle Change in area Change in impedance Relative angle between positive sequence V and I Absolute angle of voltage Absolute angle of current Maximum power transfer angle SoftMax transfer function output Initial flux Maximum flux Residual flux

Chapter 1

Introduction

Abstract The major blackout events with the process of generator oscillations are detailed in this chapter, along with their causes, at the outset. In the course of a complete literature evaluation, various forms of power swing detection and generator out-of-step protection systems have been detailed, together with their benefits and drawbacks. Based on the comprehensive review, the research aperture was emphasized at the conclusion. The research gap highlights the need for better power swing detection and generator out-of-step protection while using renewable energy sources in existing power systems. Keywords Blackout · Intigration · Power system protection · Renewable · Research gap

1.1 Motivation Power systems are complicated, non-linear, and highly interconnected networks that supply customers with high-quality, reliable electrical power at affordable prices. Power systems’ typical physical and functional components include generation, transmission, distribution, operation, electricity market, and utilization. The power system’s function is to provide power generation, delivery, and utilization to the endusers. Generations comprise a variety of resources, including traditional fossil fuel and renewable energy. Power delivery contains an energy route to distribute electrical energy from sources to buyers, including transmission and distribution lines and their functions. The power market handles all economic and energy trades. Power utilization is the final terminus of the power system energy circle, which involves customers. The most common and inevitable disturbances of the power system are usually faults and sudden load changes. These disturbances lead to transient instability, depending upon its severity (Gunasegaran et al. 2015). Stable power oscillations are not harmful and regularly occur in the electrical power system. Sudden changes in load, power generation, and system topology create a transient perturbation. The generator rotor accelerated and decelerated to balance output power to mechanical input power to counteract changes in load, power generation, and system topology. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 J. K. P. Desai and V. Makwana, Power Swing Detection and Generator Out-of-Step Protection Under Renewable Power Source Integration, Energy Systems in Electrical Engineering, https://doi.org/10.1007/978-981-19-9546-0_1

1

2

1 Introduction

Fig. 1.1 The process of generator oscillation state (Desai and Makwana 2022)

The stability of the rotor angle is established with the steady power oscillation. In case of an unstable power oscillation, they move backward and forward beyond the system once it is initiated. Most of the time, power oscillations occur during rapid changes in the load, switching of power lines, activation/exclusion of generators from service, and degree of failure of a system (Thorp 2012). A generator oscillation state can transpire in the course of states, as shown in Fig. 1.1. Every event has its own corrective action time, which depends on the type of disturbance and coordination among the protection schemes. The failure of protection coordination results in a widespread blackout. So, there is a need to provide financial and deployment of the correct protection coordination design in the modern power system. No single solution exists for stopping power outages, but many things can be done to reduce disturbance, including investigations, defensive measures, corrective measures, intensified monitoring, diagnostics, and control center performance improvements. Additionally, guaranteeing real-time operating limits daily, improvement in protection coordination, the analysis of relays, examination of dynamic voltage and transient stability, appraisal of the condition of aging infrastructure, and safe control and protection system can limit the large spread blackouts (Phadke and Thorp 2009).

1.1 Motivation

3

The unintended tripping of transmission lines requires a power swing blocking function with the distance relay (Rao and Pradhan 2017). The distance relay, close to the electrical center, may pick-up even during stable power swings under a renewable power integrated system. Power swing is a transient stability problem of the power system. The normalizing positive-sequence voltage and its phasor retain a committed relationship during transient instability. The supervisory scheme can be designed to predict the zone encroachment using synchronous measurements from both the ends comprising standard voltage and current (Lavand and Soman 2022). A power swing is a phenomenon that usually occurs due to suddenly cutting off the heavy loads and tripping the transmission lines due to the faults in the system. The unstable power swing required quick detection; otherwise, there may be damage to electrical equipment and untrue operations of distance relay (Shimpi et al. 2017). The substantial disturbance that occurred in the northern grid of India on 30th July 2012 caused a blackout in eight states. The leading causes were the malfunctioning of the distance relay due to a power swing, weak inter-regional corridors, and multiple line outages (Warathe and Patel 2015). Investigation of distance relays located on transmission lines originated from renewable power generators are at high risk of malfunctioning (Fang et al. 2019a, b). A significant amount of renewable power injection in an existing power system network changes the topology, power flow, and especially the fault characteristic (Ren et al. 2017). The fault characteristics influenced by renewable power energy sources (RPES) exhibit an unstable internal impedance, limited current amplitude, current frequency offset, and controlled current phase angle, which is different that of the synchronous generator (Hooshyar and Baran 2013; Nimpitiwan et al. 2007). In Rampurkar et al. (2016), a detailed sequence of events is provided that led to a severe blackout that occurred in the northern grid of India on 30th July 2012. This blackout occurred on two consecutive days. The Indian power grid is divided into five regions. The regions are the northern region (NR), southern region (SR), eastern region (ER), western region (WR), and northeast region (NER). On July 30, 2012, the NR was heavily loaded, and the inter-regional links from the WR–NR, and WR–ER–NR could not handle the power flow. In the Rampurkar et al. (2016), it shows a comprehensive analysis of the blackout using four PMU (Phasor measurement units) placed in the WR and NR. These measurements provide sufficient evidence of small-signal oscillations occurring on both days. Keeping the grid in the steady-state condition is prioritized since the power system is highly interconnected and apt to stability issues. Based on the above discussion, out-of-step protection is essential in power systems. Also, the PMUs are becoming key equipment for operation and protection. Further, Artificial intelligence-based outof-step relay is the need of the hour. The conventional out-of-step protection relied on local measurements and had various issues, including reliance on the network topology and system parameters. Thus, this work develops a novel power swing detection and out-of-step protection algorithm to prevent an interconnected power system from blackout. The part content of Chap. 1 is adapted from Desai (2021).

4

1 Introduction

1.2 Literature Review 1.2.1 Literature Review on Most Recent Work OOS protection based on frequency has been presented by the researcher in Zhang and Zhang (2019) recently. The frequency variation of voltage is analyzed using equivalent two-terminal system. A fast frequency calculation algorithm is used with WAMS in Zhang and Zhang (2019). The work of Zhang and Zhang (2019) prove that when δ < 70◦ , the frequency of voltage at all point changes between f 1 to f 2 which is smaller than f = f 2 − f 1 and δ > 70◦ , frequency difference between twoterminal of equipment called p and q which is  f pq > f. Here, δ is expressed as 2π ( f 2 − f 1 ) θo2 , where, θo2 is initial angle of E2. Further, E1 and E2 are the source 1 and source 2 voltages in two-terminal systems considered in Zhang and Zhang (2019). Zubov’s approximation boundary concept has been extended in Yellajosula et al. (2020). The PMU has been incorporated in the algorithm proposed in Yellajosula et al. (2020). The work doesn’t define the refreshing rate required for boundary updates. The OOS detection method in Hashemi and Sanaye-Pasand (2019) is not functional for the dynamic power system as the method detects OOS after its occurrence. The Lissajous figure-based power swing detection from fault has been presented in Patel and Bera (2018). The technique in Patel and Bera (2018) demands a higher sampling rate of frequency and knowledge of system frequency and swing frequency. The paper Hashemi and Sanaye-Pasand (2018) shows two distinct techniques, such as the rate of change of zero sequence current and magnitude of voltage and current of each phase for asymmetrical power swing arising due to single-pole tripping. The work in Hashemi and Sanaye-Pasand (2018) presents the effect of symmetrical and unsymmetrical power swing on distance relaying. Table 1.1 shows the summary of all the papers discussed above with their merits and demerits.

1.2.2 Literature Review on Different Methods of Out-of-Step Protection Transient instability is one of the major security threats to the system, resulting in a pole slip. The generators are subject to mechanical and thermal damage due to a pole slip. Therefore, in the case of out-of-step, early detection and disconnection of the generator from the grid is essential (Sobbouhi and Aghamohammadi 2015). The literature review on Power swing detection and generator out-of-step protection has been divided based on the concept used by different researchers around the world. Hence, the literature survey has been divided into the following subcategories: (1) Literature survey on rate of change of impedance-based technique (2) Literature survey on equal Area Criterion (EAC) based technique (3) Literature survey on

1.2 Literature Review

5

Table 1.1 Summery of most recent reported work, their merits and demerits Reference and Technique Main feature Merits Demerits Year Zhang and Zhang Frequency-based (2019) OOS

Yellajosula et al. (2020)

Extended Zubov’s approximate boundary

Hashemi and Sanaye-Pasand (2019)

Current-based OOS detection method

Patel and Bera (2018)

Lissajous figure-based power swing and fault detection

Hashemi and Sanaye-Pasand (2018)

Rate of change zero sequence current and voltage and current magnitude of each phase

Detect OOS Easy to condition in first implement oscillation cycle and determine the swing center location

Setting process involved, few assumptions like first OOS oscillation period, higher rate of sampling of PMU Detect OOS and Offline system Boundary update Power swing studies not refreshing rate is condition needed not defined. Higher refreshing rate for each boundary, not suitable for faster relaying action Detect OOS Enhance line Possibility of detection for differential misstransmission line protection identification differential between fault and protection OOS condition, not suitable for dynamic system Useful for OOS Identify power High sampling detection of both swing from fault frequency overhead line and required, under ground knowledge of cable swing and system frequency needed Useful in case Detect earth-fault Required PSB for asymmetrical from Earth relay and power swing asymmetrical each phase during power swing independently single-pole switching

Lyapunov Theory-based Approaches (4) Literature survey on adaptive relayingbased approach (5) Literature survey on PMU-based Approaches (6) Literature survey on AI-based technique. Table 1.2 shows the brief summary of various methods used in literature for OOS protection with its limitation. Nowadays, the power system integrated with PV (Photovoltaic) generators leads to different transient events. The power swing characteristics of such a modern power system are complex as compared to the power system in which only a synchronous

6

1 Introduction

Table 1.2 Summary of different power swing detection and OOS protection technique with its limitation Reference Technique Main feature Limitation and research gaps Hashemi and Sanaye-Pasand (2018); Sobbouhi and Aghamohammadi (2015) Fitzgerald (2003); IEEE PSRC WG D6 (2005); Krause (2002); Tziouvaras and Hou (2004)

Rate of change of parameter based approach

Easy to implement

Centeno (1997); Eugene (1965); Margolis (1962); Margolis and Vogt (1962); Pai (1981); Paudyal et al. (2010); Saaty (1983); Wu (2009); Yannan and Vongsuriya (1967) Evangelos et al. (2011); Pai (1981); Zubov (1955)

Lyapunov Theory Based Approaches

Does not required numerical solution

Adaptive relaying approach

Work under changed topological condition

Franco et al. (2013); Phadke and Thorp (2017); Zhang and Zhang (2017)

PMU Based Approaches

Makes the relaying adaptive

Equal Area Criterion Useful for transient (EAC) based approach stability studies

Ariff and Pal (2016); AI-based approach Donald (2006); Lavand and Soman (2016); Regulski et al. (2018); Zare et al. (2018)

No extensive study of system required, Automatic settings, faster response

Stability boundary constantly changes due to different power system operating conditions It can’t be applied directly for Multi-machine system, required additional communication devices Difficulty of finding distribution of eigen values

Complex methods and not verified under renewable integration for OOS protection PMU-based solution add extra cost and higher sampling of phasor, not verified under renewable integration for OOS protection Proper data selection is must, ANN methods are selected according to the problem so one ANN method cannot be suitable for each problem. Not used for unstable power swing detection so far yet, not verified under renewable integration so far yet for OOS protection

1.2 Literature Review

7

X (Ω) Outer Zone

X (Ω)

Innerr Zone

Outer Zone

Inner Zone R (Ω)

R (Ω)

Fig. 1.2 Different OOS relay characteristics

generator is the primary source of power generation (Ariff and Pal 2016). So, this work has carried out an exclusive literature survey on the impact of renewable on the out-of-step relay. Further, an exhaustive literature survey on worldwide blackout has been considered to identify the effect of mal-operation of protective gears on the power grid.

1.2.2.1

Rate of Change of Parameter-Based Approach

To detect the out-of-step condition, OOS relays equipped with special characteristics are used. Some characteristic elements have been used in practice, including concentric polygons and concentric circles as shown in Fig. 1.2 (Tziouvaras and Hou 2004). Impedance-based OOS relays differentiate between power oscillations and failures because power oscillations are electromechanical and slower than failures that are electromagnetic transients. A timer is used to record the time spent by the impedance path moving from the outside to the inside of the relay characteristics. The relay declares a power swing if the dwelling time is longer than a threshold. In addition to the relay parameters, it is important to place the OOS relays in optimal places to facilitate out-of-step detection. The electrical center of the system is considered the best location for establishing OOS relays (IEEE PSRC WG D6 2005). However, stability studies must be carried out to choose the optimal location because the electrical center constantly changes under different operating conditions.

1.2.2.2

Equal Area Criterion (EAC) Based Approach

The Equal Area (EA) criterion is a well-known tool for transient stability studies in power systems. EAC describes the system’s stability based on the surface beneath the P-δ curve. This approach applies directly to single machine infinite bus systems (SMIB) (Centeno 1997; Fitzgerald 2003; Krause 2002; Wu 2009). The EAC algorithm in the P-δ domain uses the out-of-step detection method to compare the power

8

1 Introduction

angle to the critical angle. In Wu (2009), based on EAC, critical clearing angle is computed and compared with the power angle of the SMIB system to decide on out-of-step conditions. The main disadvantage of the EAC method is that it cannot be applied directly to multiple-machine systems. In addition, the EAC-based method in the P-δ domain requires information about the power angle of two areas, which means the requirement for additional communication devices.

1.2.2.3

Lyapunov Theory Based Approaches

From the standpoint of control theory, the Out-of-step decision is to find an analytical solution to the equations of the swing after perturbations. Instead of directly solving a system of partial differential equations (PDE), Lyapunov proposed a method that does not require numerical or analytical solutions; rather, it analyzes the eigenvalue distributions of the system to find out the stability region (Centeno 1997; Eugene 1965; Evangelos et al. 2011; Margolis 1962; Margolis and Vogt 1962; Pai 1981; Paudyal et al. 2010; Saaty 1983; Yannan and Vongsuriya 1967). Though Lyapunov’s theory is excellent for stability analysis, however, in some critical cases, even distributions of eigenvalues are not easily calculated.

1.2.2.4

Adaptive Relaying Approach

The Kumar et al. (2022) provides a step-by-step adaptive relay logic. The angular separation is measured and compared with minimal and maximum angles. It simultaneously monitors the energy of the rotor. Using all the information, it finds a stable or volatile power oscillation. The Wide Area Monitoring System (WAMS) on the network gives the real-time angular measurement to determine if the system will collapse or stay stable (Phadke and Thorp 2017). However, the method used in Phadke and Thorp (2017) results in delays in triggering large generators, which can result in damage if the device malfunctions. The WAMS used in Zhang and Zhang (2017) introduces the stepwise separation protection based on the voltage phase angle information for each bus. The paper discussed the technique that is not complicated in mathematics but relatively simple, which encourages the use of PMU for out-of-step protection. In addition, the extended protection plan for the prevention of generalized power outages is described in Franco et al. (2013) local signals used for distance relay and global signals using synchrophatic measurements. It used to be islanding controlled. Local and global signals can also be used to make a good decision from out-of-step relays after a sudden disturbance. The authors of Zare et al. (2018) proposed a large-scale out-of-step prediction with the help of an adaptive control island concept. I note a new adaptive step-based protection approach in Zare et al. (2018) and Ariff and Pal (2016). However, the methods used for out-of-step protection in Zare et al. (2018) and Ariff and Pal (2016) are complex and need to be verified with various levels of renewable penetration.

1.2 Literature Review

1.2.2.5

9

PMU Based Approaches

The document (Regulski et al. 2018) describes the PMU application in the out-ofstep protection. It has used direct phase angle comparison between different locations to identify out-of-step condition of the generator. However, the document does not compare the proposed method with conventional out-of-step protection schemes such as the double-blind scheme. Distance relays are most likely to be picked up during the condition when the oscillation positioned sits near them on the transmission line (Lavand and Soman 2016). PMU signals, as well as line entries, are used to generate PSB (power swing blocking) signals under such conditions. The approach is described using the linear normalized positive-sequence minimum voltage as coordinate y and the positive-sequence voltage angle as x-co-ordinate in Lavand and Soman (2016). The paper has shown the method for a line out-of-step protection and not for generator out-of-step protection. The generator out-of-step protection is the most important during power swing because recent advancement in the power system moves the swing positioned away from the transmission line toward the generator and generator transformer (Donald 2006).

1.2.2.6

AI Based Approach

ANN delivers faster responses and requires a quarter of the error signal cycle to identify the listed faults (Martin and Aguado 2003). For example, an ANN-based remote relay can provide fast, and accurate operation (Coury and Jorge 1998). Nevertheless, Coury and Jorge (1998) does not address the other distance relay problems, such as mal-operation of a distance relay during a power swing and for a system with large renewable integration. In Brahma (2007), Avdakovic et al. (2012), Choudhary and Sharma (2015), Koley et al. (2017), Martin and Aguado (2003), Coury and Jorge (1998), once the power oscillation is initially identified based on the fault, further identification is missing. But that’s the most important part of a modern power system. Advanced power system controllers can cushion stable power oscillations, and there is no need for quick disconnection of affected elements. However, advanced power system controllers cannot dampen unstable power oscillations, so prompt and proper detection is required.

1.2.3 Literature Review on Impact of Renewable Power Sources on Protective Relays The paper (Eftekharnejad et al. 2013) concluded that In almost all the case studies, bus voltage magnitudes are the most adversely affected system parameters during the transients. It is observed that systems with high PV penetration levels achieve

10

1 Introduction

greater voltage dips following most disturbances. The reduction in voltage bus leads the system toward instability. The penetration of PV and DFIG (doubly-fed induction generator) improves the system’s damping, but up to a certain level. If the penetration of solar PV is more than 50%, then it creates a negative effect on the damping of the system (Choudhary and Sharma 2015). The energy reduction caused by wind turbines is proportionate to its natural time constant Tang et al. (2008). The inertia of the wind generator makes the power swing, not as it has been considered so far. The extensive penetration of inverter generation, such as solar photovoltaic generators, decreases the critical compensation time of the system (Kumaran Gunasegaran 2016). Large-scale integration of wind generation changes the power swing characteristics Haddadi et al. (2019). The power swing blocking time delay needs frequent revision with increased wind power penetration in the system. The paper Haddadi et al. (2019) addresses the concern of protection improvements due to the negative impact of wind power generation but does not reveal any proposed solution to resolve the problem.

1.3 Conclusion Based on the above survey on different methods developed by the researcher and recent work presented by the researchers, the research gap has been found that most of them did not have implemented their technique under the impact of renewable power integration. Today’s power system is highly integrated with renewable; filling out this research gap is very important. Further, It is found that the impact of renewable power integration results in the mal-operation of protective gear. There is still no proper solution found, especially in the area of power swing detection and out-ofstep protection. Indeed, huge research is required to determine the impact of large renewable power integration effects on out-of-step relay and coordination. In a traditional power system, the protective function was easy to implement due to the vertically integrated structure (Hyro 2018). Modern power systems’ increasing demands made it complex and worked near the steady-state stability limit. During the cascading failure, the power system does not have sufficient reserve time to take preventative measures. Many countries have increased the penetration of renewable to compensate for the increasing demand. The integration of renewable power creates many problems with traditional protection schemes and control techniques (Laghari et al. 2013). The integration of a large amount of renewable energy into the grid reduced the inertia of the grid. The weak inertia creates frequency instability and power swing problems. The main challenge in the modern power system is the existence of old protective functions that cannot work accurately under the deregulated structure and with a new type of power generation such as wind, solar and geothermal, etc.

References

11

In most of the recent blackout cases, it is noticed that the malfunction of protective gear is the reason behind the blackouts (Kuria 2022; Liu et al. 2018). The other major challenge is the reliability of the system, which depends on the reliability of the system component (Comparison of three Under-Frequency Load Shedding Schemes referring to the Power System of Sri Lanka 2022; Hyro 2018). A literature survey of the most recent paper suggests the use of ANN. However, it is found that identification of complex, unstable power swing from stable power swing has not been addressed by most of them. Most of the work doesn’t consider recent topological changes like reactive power compensation, PSS, AVR, AFC, and renewable integration. Most of all, the method survey required either system studies or some threshold estimation. The major research gap between indirect measurement approach to direct measurement approach till setting free approach for power swing detection and out-of-step protection has been required to fill out. Also, the researcher has not explored blackout prevention using power swing detection and out-of-step protection. The details of conventional OOS methods, PMU and WAMS functioning, and mathematical analysis of the problem with illustration have been presented in Chap. 2.

References Ariff MAM, Pal BC (2016) Adaptive protection and control in the power system for wide-area blackout prevention. IEEE Trans Power Delivery 31(4):1815–1825. https://doi.org/10.1109/TPWRD. 2016.2518080 Avdakovic S, Nuhanovic A, Kusljugic M (2012) Wavelet transform applications in power system dynamics. Lancet 83. https://doi.org/10.1016/j.epsr.2010.11.031 Brahma SM (2007) Distance Relay With Out-of-Step Blocking Function Using Wavelet Transform. IEEE Trans Power Delivery 22(3):1360–1366. https://doi.org/10.1109/TPWRD.2006.886773 Centeno V (1997) An adaptive out-of-step relay for power system protection. IEEE Trans Power Delivery 12(1):61–71 Choudhary S, Sharma FB (2015) Small-signal stability analysis of renewable source connected power system and identification of oscillatory modes using wavelet transform. In: 2015 international conference on smart grid and clean energy technologies (ICSGCE), Offenburg, pp 23–29. https://doi.org/10.1109/ICSGCE.2015.7454264 Comparison of three Under-Frequency Load Shedding Schemes referring to the Power System of Sri Lanka Coury DV, Jorge DC (1998) Artificial neural network approach to distance protection of transmission lines. IEEE Trans Power Delivery 13(1):102–108. https://doi.org/10.1109/61.660861 Desai JP (2021) Analysis of power swings and blackouts. IEEE Congreso Estudiantil de Electrónica y Electricidad (INGELECTRA) 2021:1–6 Desai JP, Makwana VH (2022) Modeling and implementation of power swing detection and outof-step protection. J Inst Eng India Ser B 103:541–548 Donald R (2006) Protective relaying for power generation systems. CRC/Taylor & Francis 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. https://doi. org/10.1109/TPWRS.2012.2216294 Eugene LJ (1965) A guide of the application of the Liapunov direct method to flight control systems. NASA Contractor Report: NASA CR-209

12

1 Introduction

Evangelos F, Renke H, George JC, . Meliopoulos AP (2011) A predictive out of step protection scheme based on PMU enabled dynamic state estimation. In: 2011 IEEE Power Energy Society General Meeting, pp 1–8 Fang Y, Jia K, Zhu Z, Feng T, Yang Z, Liu B (2019a) Adaptability analysis of distance protection on the transmission lines emanating from renewable power generators. In: 2019 IEEE innovative smart grid technologies - Asia (ISGT Asia), pp 275–279. https://doi.org/10.1109/ISGT-Asia. 2019.8881096 Fang Y, Jia K, Yang Z, Li Y, Bi T (2019b) Impact of inverter-interfaced renewable energy generators on distance protection and an improved scheme. IEEE Trans Ind Electron 66(9):7078–7088. https://doi.org/10.1109/TIE.2018.2873521 Fitzgerald AE (2003) Electric machinery, 6th edn. McGraw-Hill, New York Franco R, Sena C, Taranto GN, Giusto A (2013) Using synchrophasors for controlled islanding a prospective application for the Uruguayan power system. IEEE Trans Power Syst 28(2):2016– 2024. https://doi.org/10.1109/TPWRS.2012.2224142 Gunasegaran MK, Tan C, Bakar AHA, Mokhlis H, Illias HA (2015) Progress on power swing blocking schemes and the impact of renewable energy on power swing characteristics: a review. Renew Sustain Energy Rev 52:280–288. https://doi.org/10.1016/j.rser.2015.07.066 Haddadi A, Kocar I, Karaagac U, Gras H, Farantatos E (2019) Impact of wind generation on power swing protection. IEEE Trans Power Delivery 34(3):1118–1128. https://doi.org/10.1109/ TPWRD.2019.2896135 Hashemi SM, Sanaye-Pasand M (2018) Distance protection during asymmetrical power swings: challenges and solutions. IEEE Trans Power Delivery 33(6):2736–2745. https://doi.org/10.1109/ TPWRD.2018.2816304 Hashemi SM, Sanaye-Pasand M (2019) Current-based out-of-step detection method to enhance line differential protection. IEEE Trans Power Delivery 34(2):448–456. https://doi.org/10.1109/ TPWRD.2018.2873698 Hooshyar H, Baran ME (2013) Fault analysis on distribution feeders with high penetration of PV systems. IEEE Trans Power Syst 28(3):2890–2896. https://doi.org/10.1109/TPWRS.2012. 2227842 Hyro BC (2018) BC hydro crews continue to repair damage after historic windstorm, Extended Outages Expected; Technical Report; BC Hyro: Vancouver. BC, Canada IEEE PSRC WG D6: Power swing and out-of-step considerations on transmission lines (2005) Koley E, Shukla SK, Ghosh S, Mohanta DK (2017) Protection scheme for power transmission lines based on SVM and ANN considering the presence of non-linear loads. IET Gener Transm Distrib 11(9):2333–2341. https://doi.org/10.1049/iet-gtd.2016.1802 Krause PC (2002) Analysis of electric machinery and drive systems, 2nd edn. Wiley, New York Kumaran Gunasegaran M (2016) Investigation of power swing phenomenon in a network integrated with photovoltaic generation. http://studentsrepo.um.edu.my/id/eprint/6398 Kumar N, Nagaraja DR, Khincha HP (2022) A smart and adaptive scheme for generator out of step protection Kuria M, Operational challenges facing performance of thermal power plants in Kenya, p 60 Laghari JA, Mokhlis H, Bakar AHA, Mohamad H (2013) Application of computational intelligence techniques for load shedding in power systems: a review. Energy Convers Manag 75:130–140 Lavand SA, Soman SA (2016) Predictive analytic to supervise zone 1 of distance relay using synchrophasors. IEEE Trans Power Delivery 31(4):1844–1854. https://doi.org/10.1109/TPWRD. 2016.2521784 Lavand SA, Soman SA (2016) Predictive analytic to supervise zone 1 of distance relay using synchrophasors. IEEE Trans Power Deliv 31(4):1844–1854. https://doi.org/10.1109/TPWRD. 2016.2521784 Liu B, Zhou B, Jiang D, Yu Z, Yang X, Ma X (2018) Distributed accommodation for distributed generation – from the view of power system blackouts. In: Advances in Green Energy Systems and Smart Grid, Singapore, pp 236–246

References

13

Margolis SG (1962) Computation of regions of asymptotic stability for closed-loop systems by Liapunov’s second method. Ph.D. Dissertation, University of Pittsburgh Margolis SG, Vogt WG (1962) Control engineering applications of V.I.Zubov’s construction procedure for Lyapunov functions. IEEE Trans Automat Contr 8(2)104–113 Martin F, Aguado JA (2003) Wavelet-based ANN approach for transmission line protection. IEEE Trans Power Delivery 18(4):1572–1574. https://doi.org/10.1109/TPWRD.2003.817523 Nimpitiwan N, Heydt GT, Ayyanar R, Suryanarayanan S (2007) Fault current contribution from synchronous machine and inverter based distributed generators. IEEE Trans Power Deliv 22(1):634– 641. https://doi.org/10.1109/TPWRD.2006.881440 Pai MA (1981) Power system stability analysis by the direct method of Lyapunov. North-Holland, New York Patel B, Bera P (2018) Detection of power swing and fault during power swing using Lissajous figure. IEEE Trans Power Delivery 33(6):3019–3027. https://doi.org/10.1109/TPWRD.2018.2850958 Paudyal S, Ramakrishna G, Sachdev MS (2010) Application of equal area criterion conditions in the time domain for out-of-step protection. IEEE Trans Power Delivery 25(2):600–609 Phadke AAAAAAG, Thorp JS (2009) Wide area measurement applications. Computer relaying for power systems. Wiley, pp 285–307. https://doi.org/10.1002/9780470749722.ch10 Phadke AG, Thorp JS (2017) Synchronized phasor measurements and their applications, 2nd ed. Springer International Publishing Rampurkar V, Pentayya P, Mangalvedekar HA, Kazi F (2016) Cascading failure analysis for Indian power grid. IEEE Trans Smart Grid 7(4):1951–1960. https://doi.org/10.1109/TSG.2016.2530679 Rao JG, Pradhan AK (2017) Supervising distance relay during power swing using synchrophasor measurements. IET Gener Transm Distrib 11(17):4136–4145. https://doi.org/10.1049/iet-gtd. 2016.1110 Regulski P, Rebizant W, Kereit M, Herrmann H-J (2018) PMU-based generator out-of-step protection. IFAC-PapersOnLine 51(28):79–84. https://doi.org/10.1016/j.ifacol.2018.11.681 Ren G, Liu J, Wan J, Guo Y, Yu D (2017) Overview of wind power intermittency: impacts, measurements, and mitigation solutions. Appl Energy 204:47–65. https://doi.org/10.1016/j.apenergy. 2017.06.098 Saaty TL (1983) Modern nonlinear equations. Dover, New York Shimpi KS, Student ME, Jain AM (2017) Power swing detection, blocking and unblocking logic in distance relay. In: 2017 international conference on computing methodologies and communication (ICCMC), pp 904–908. https://doi.org/10.1109/ICCMC.2017.8282597 Sobbouhi AR, Aghamohammadi MR (2015) A new algorithm for predicting out-of-step using rotor speed-acceleration based on phasor measurement units (PMU) data. Electric Power Compon Syst 43(13):1478–1486. https://doi.org/10.1080/15325008.2014.1001089 Tang C, Pathmanathan M, Soong WL, Ertugrul N (2008) (PDF) Effects of inertia on dynamic performance of wind turbines. Australasian Universities Power Engineering Conference (AUPEC’08) The impact of social networks on behavioral change: a conceptual framework Thorp J (2012) Disturbance analysis for power systems [Book Reviews]. IEEE Power Energy Mag 10(3):89–90 Tziouvaras DA, Hou D (2004) Out-of-step protection fundamentals and advancements. In: Proceedings of the 57th annual conference protective relay engineering, pp 282–307 Warathe S, Patel RN (2015) Algorithm for power swing protection using DSP based numerical relay. In: 2015 2nd international conference on signal processing and integrated networks (SPIN), pp 762–767. https://doi.org/10.1109/SPIN.2015.7095411 Wu QH (2009) Protective relaying of power systems using mathematical morphology. Springer, New York Yannan Y, Vongsuriya K (1967) Nonlinear power system stability study by Liapunov function and Zubov’s method. IEEE Trans Power App Syst PAS-86(12):1480–1485 Yellajosula JRAK, Wei Y, Grebla M, Paudyal S, Mork BA (2020) Online detection of power swing using approximate stability boundaries. IEEE Trans Power Delivery 35(3):1220–1229. https:// doi.org/10.1109/TPWRD.2019.2941522

14

1 Introduction

Zare H, Yaghobi H, Alinejad-Beromi Y (2018) Adaptive concept of controlled islanding in power systems for wide-area out-of-step prediction of synchronous generators based on adaptive tripping index. Transm Distrib IET Generat 12(16):3829–3836. https://doi.org/10.1049/iet-gtd.2018.0319 Zhang S, Zhang Y (2017) A novel out-of-step splitting protection based on the wide area information. IEEE Trans Smart Grid 8(1):41–51. https://doi.org/10.1109/TSG.2016.2593908 Zhang S, Zhang Y (2019) Characteristic analysis and calculation of frequencies of voltages in out-of-step oscillation power system and a frequency-based out-of-step protection. IEEE Trans Power Systems 34(1):205–214. https://doi.org/10.1109/TPWRS.2018.2866022 Zubov VI (1955) Voprosy teorii vtorogo metoda Lyapunova, postroenie obshchego resheniya v oblasti asimptoticheskoi ustoichivosti,(Problems in the theory of the second method of Lyapunov, construction of the general solution in the domain of asymptotic stability). Prikladnaya Matematika i Mekhanika 19:179–210

Chapter 2

Power Swings with Distance Relay and PMU

Abstract The background of PMU and WAMS, as well as their applications, are discussed in this chapter. Beginning with a doubly-fed source system, the notion of out-of-step circumstances was explained. Later in the chapter, the theory and limits of double blinder-based and distance relays with power swing blocking functions are discussed. This chapter covers the principles of PMU and WAMS, which will be used in later chapters to implement and understand the proposed solution. Keywords Blinders · Distance relay · Impedance Trajectory

2.1 Mathematical Analysis of the Research Problem The doubly-fed source system is shown in Fig. 2.1. Source A and source B are the generator and grid, respectively, in Fig. 2.1. The relay R1 is situated near busbar-A, which measures transmission line current Ir elay and Busbar voltage Vr elay to take relaying decision. Source A has its internal source impedance which is denoted as Z S , and Z g is indicated for the grid equivalent impedance in Fig. 2.1. the voltage of Source A is at angle δ ◦ and voltage of source B is at angle 0◦ . E S and E R are the voltage magnitude of source A and source B, respectively. Now, the current measured by the relay R1 is mathematically given as IRelay =

Es ∠δ − ER ZT

(2.1)

where, Z T = Z S + Z L + Z g . From Eq. 2.1, the impedance seen by the relay is given as Zseen =

Vrelay Es ∠δ − IRelay ZS = Irelay Irelay

(2.2)

Simplifying the Eq. 2.2,

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 J. K. P. Desai and V. Makwana, Power Swing Detection and Generator Out-of-Step Protection Under Renewable Power Source Integration, Energy Systems in Electrical Engineering, https://doi.org/10.1007/978-981-19-9546-0_2

15

16

2 Power Swings with Distance Relay and PMU

Fig. 2.1 The doubly-fed source system (Desai 2022)

 = −Zs +

Es ∠δ Es ∠δ − ER 

ER ES

ZT

1 ER 1 − ES ∠ − δ

= −Zs + ZT For simplicity assumed

 (2.3)

 (2.4)

= 1 then Eq. 2.4 is written as 

Zseen = −Zs + ZT

1 1 − cos δ + j sin δ

 (2.5)

Now, Eq. 2.5 is further simplified as ZT = −Zs + 2

 sin2

ZT = −Zs + 2 sin

δ 2

ZT = −Z s + 2

δ 2

1 + j sin 2δ cos 2δ



  δ δ sin − j cos 2 2 

δ 1 − j cot 2

(2.6)

(2.7)

 (2.8)

Finally, Impedance seen (Z seen ) by the relay is achieved as = −Zs +

ZT ZT δ −j cot 2 2 2

(2.9)

From Eq. 2.9, impedance swing has been plotted on the R-X plane which is shown in Fig. 2.2. In Fig. 2.2, line section OA indicate Z S , line section OD indicate Z T , line section BD represent Z g , line section AC represents −Z S + Z2T and PC shows Z2T cot 2δ . The line section PC is an imaginary line of impedance swing which indirectly

2.1 Mathematical Analysis of the Research Problem

17

Fig. 2.2 Impedance trajectory during power swing on R-X plane (Desai 2022)

represents power swing which is shown as black colored dotted line interacting line section AB at the center point C. The center point C is said as swing center location for the condition of EE RS = 1. The line connecting point A to B is said as system slope, and at δ = 180, impedance trajectory intersects the line section AB. At that moment, sources A and B are subjected to out-of-step conditions. In the above analysis, the assumption of EE RS = 1 is not possible under transient conditions that arise due to fault and sudden load changes. If EE RS > 0 then the plot is above the line PC and If ES < 0 then the plot is below the line PC as shown in Fig. 2.2. ER In practical, EE RS is continuously changing because of the following reasons: (1) EE RS depends on excitation control, governor control, and PSS control of the generator. (2) EE RS changes with the system topology, condition, frequency, phase, and voltage. The power system is a non-linear system that depends on more than one control settings. So, at one point of time, EE RS is greater than 1 at one control setting of the power system under a transient state. At the second point of time, the EE RS becomes less than 1 at the next control settings of the power system. These variable conditions occurred during a transient disturbance, such as sudden load changes, faults, and system parameters change. These transient conditions in modern power systems create very complex power swing characteristics on the R-X plane. The consequences of unstable power swings are mostly mal-operation of the transmission line’s distance relay, damage to the generator and turbine-generator unit. The results of dangerous power swings cause the cascade failure of numerous transmission lines, transformers, and generators. Ultimately, the effect in power system blackouts.

18

2 Power Swings with Distance Relay and PMU

2.2 Phasor Measurement Unit and WAMS A Phasor Measurement Unit (PMU) is an apparatus for capturing phasors of Alternating Current (AC) signals. Phasors of AC waveform consist of RMS magnitude values and phasor angles. The following equation can mathematically characterize an AC waveform: (2.10) x(t) = X m cos(2π f t + φ) where the X m shows the magnitude of the sinusoidal waveform, f is a power system synchronous frequency, which is generally 50 or 60 Hz, and φ represents the phase shift of the waveform. The phasor of the AC waveform in Eq. (2.10) is given by Xm X = √ ∠φ 2

(2.11)

PMU is synchronized with GPS time signals which are referred to as synchrophasors. Synchrophasors estimate AC signals from a frequency of either 50 or 60 Hz power grid at a standard sampling rate of 48 phasors per second, high up to 60 phasors per second, in the reference waveform. As shown in Fig. 2.3, the phasor quantities are measured using a reference cosine function. A Coordinated Universal Time (UTC) reference signal is then used to time-stamp the phasors measured throughout the power system for the synchronous intention. The UTC signal reaches from a GPS clock, ensuring the same time reference for all phasor measurements and guarantees the clock’s accuracy. A typical PMU functional block diagram is shown in Fig. 2.4, consisting of five main components: Filter, Analog to Digital Converter, GPS Clock Receiver, Phasor-locked Oscillator, Phasor Calculation, and Output Interface. Power system AC waveforms are captured using current or voltage transformers, then processed through anti-aliasing filters. The embedded Analog then digitizes analog signals to Digital Converter. A phase-locked oscillator uses a reference time signal from the GPS receiver required for synchro-

Fig. 2.3 A sinusoidal waveform explain synchrophasors

2.2 Phasor Measurement Unit and WAMS

19

Fig. 2.4 Simple block diagram of PMU

nized samples, typically having an accuracy of 1 ms. The embedded microprocessor contains frequency estimation and phasor estimation algorithms (Fast Fourier Transform or Discrete Fourier Transform), which could send information to power system operators with a typical reporting rate of 1–2 phasor per second through Interface. A general phasor measurement unit data acquisition and transmission network is shown in Fig. 2.5. The first communication level is between PMU devices and its closest phasor data concentrator (PDC). The phasor data available at this level is very high rate typically 10–30 phasors per second. These PMU measurements can be used. The real-time protection algorithm needs fast and large data from the system. The next level is the communication between different local PDC layers. The

Fig. 2.5 Power system synchrophasor network

20

2 Power Swings with Distance Relay and PMU

third communication level eventually connects all local PDC layers to a master level. The operating center could quickly draw an entire detailed grid status picture based on these transmission procedures. Hence, this network could be suitable for widearea controlling or protecting. However, due to the low transmitting rate of phasor data between the second communication level, which are typically not more than ten phasors per second, and the third level, which is even lower, down to 1 phasor per second, the PMU network is now mainly used for system-wide monitoring and after-fault event analysis.

2.3 Power Swing Detection and Out-of-Step Detection Technique 2.3.1 Double Blinder Based Technique The logic of the OOS tripping scheme is initiated when the swing locus crosses the outer blinder-R1, on the right at the separation angle α. The system will only undertake measurements when a swing is detected while the swing locus passes through the internal blinder-R2 (Hasnain et al. 2020). Thus, adjusting the internal blinder is essential to prevent a trip on a stable power swing. At this stage, the schematic logic makes and seals the offstage trip decision at the separation angle β. Tripping generally argues that the impedance locus leaves the characteristic schema at the separation angle δ. The essential feature of ‘exit’ depends on the particular scheme of the manufacturer but could use one of the inner or outer blinders or a separate mho circle. The manufacturer may also include a separate user-defined delay timer to further travel after leaving the characteristic. This sequence controls tripping so that the separation angle when opening the breaker does not exceed the breaker’s rated capability, typically 90◦ unless the breaker is specifically rated for 180◦ opening. When a swing is detected, and the swing locus crosses the inner blinder-R2, the trip decision is made. The swing can leave the inner and outer blinkers in both directions, and the trigger will occur. Therefore, the inner shield must be adjusted such that the separation angle, β, is greater than the critical clearance angle, δC . Transitional stability studies are required to determine an appropriate adjustment of the internal blinder. In this regard, the dual shielding system is quite similar to dual or triple-lens systems and has numerous transmission OOS relay features (Fig. 2.6).

2.3.2 Distance Relay and Out-of-Step Relay for Power Swing Blocking A system may experience a power swing following a disturbance such as a fault, generator disconnection, or switching on/off a large load. Due to significant changes

2.3 Power Swing Detection and Out-of-Step Detection Technique

21

Fig. 2.6 Double blinder relay settings method on R-X plane

in system current (I) and voltage (V) magnitudes, the impedance trajectory observed by a distance relay may enter into the protective zones and can cause unintended trip during a power swing. The impedance trajectory of a distance relay during stable and unstable power swings cases is shown in Fig. 2.7 (Koteswara and Ahmad 2017). The stable and unstable power swings cause unwanted initiation of tripping of distance relay (Makwana and Bhalja 2011). This may further lead to cascade outage of transmission lines and power system blackout. To ensure high security in protection decisions, PSB function is integrated with the protective relay. Traditional methods initiate the blocking function without observing the impedance trajectory concerning relay operating characteristics, resulting in unwanted blocking of PSB function. The blocking can be achieved much before the impedance trajectory encroaches to the zone-1 elements, averting unintended outages. In addition, whenever a fault appears in the transmission system during a power swing, the distance relay must identify the fault and trip the required circuit breaker immediately. Due to significant oscillations in pre-fault voltage and current magnitudes, the conventional fault detection algorithms good at steady-state conditions may not qualify to judge the fault during power swing.

22

2 Power Swings with Distance Relay and PMU

Fig. 2.7 Power swing blocking function with distance relay settings

2.4 Conclusion The underlying theory of power swing detection and out-of-step protection, as well as the severity of the problem as a result of power system modernization, is described in this chapter. Furthermore, for the reader’s convenience, the PMU and WAMS have been briefly discussed. The chapter concludes with a discussion of the most extensively utilized concepts, such as power swing detection, OOS protection, and PSB, as well as their limitations. The chapter as a whole establishes the context for the book’s later discussion of the proposed protection technique and tools.

References Desai JP (2021) Analysis of power swings and blackouts. IEEE Congreso Estudiantil de Electrónica y Electricidad (INGELECTRA) 2021:1–6 Hasnain A, Sukumar B, Dale F, Normann F, Dale F, Juan G, Ramakrishna G, Gene H, Prem K, Mukesh N, Eli P, Robert P, Michael R, Chris R, Pragnesh S, Phil T, Steve T, Demetrios T, Joe U, Jun V, Murty Y (2020) Application of out-of-step protection schemes for generators. In: Power System Relaying and Control Committee (PSRC), Rotating Machinery Subcommittee, Working Group J5

References

23

Koteswara Rao AV, Ahmad A (2017) Power swing blocking (PSB) function for distance relay using prediction technique. Int J Syst Assur Eng Manag 8:301–307. https://doi.org/10.1007/s13198016-0434-2 Makwana VH, Bhalja BR (2011) A new adaptive distance relaying scheme for mutually coupled series-compensated parallel transmission lines during intercircuit faults. IEEE Trans Power Delivery 26(4):2726–2734. https://doi.org/10.1109/TPWRD.2011.2159248

Chapter 3

Protection of Synchronous Generator

Abstract A synchronous generator’s protection system must be carefully designed since an unintended operation of the relay is almost as dangerous as a loss of operation. This is due to the fact that disconnecting a large generator may overwhelm the remainder of the system, causing power oscillations and an unstable power supply. Failure to rapidly remove a problem, on the other hand, may cause substantial damage to the generator and, as a result, disturb the entire system. In this chapter, major faults and abnormal conditions of a synchronous generator with its traditional protection system has been covered. The chapter fills up a bridge between the generator protection to the generator out-of-step protection. Keywords Generator · Loss of Excitation · Stator Earth Fault · Prime mover

3.1 Class of Generator Protection If a fault is of a severe character then the protective system used is known as Class A protection. Certain failures have the effect of not instantly disconnecting the generator from the infinite bus, but tripping the prime mover and boiler. The generator will lose input as a result of this tripping, and hence the power output will steadily decrease. The generator does not speed up as a result of this activity, and the stored kinetic energy is used. The protective system that commences the cycle described above is known as Class B protection. Certain errors have such severe implications that the generator just has to be disconnected from the infinite bus. As a result, the generator will only feed its auxiliary. Once the source of the issue has been identified and the fault has been cleared by an appropriate breaker, the generator may be re-synchronized with the system. The synchronization procedure does not take long. Class C protection refers to the protective strategy that simply trips the generator breaker (Oza 2010).

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 J. K. P. Desai and V. Makwana, Power Swing Detection and Generator Out-of-Step Protection Under Renewable Power Source Integration, Energy Systems in Electrical Engineering, https://doi.org/10.1007/978-981-19-9546-0_3

25

26

3 Protection of Synchronous Generator

3.1.1 Class A Actions taken when Class A protection is activated: (i) Tripping of Generator breaker (ii) Tripping of Generator field breaker (iii) Tripping of Unit auxiliary transformer incoming breakers (iv) closing of Tie breakers between the auxiliary station bus and the auxiliary unit bus (v) Tripping of Boiler (vi) Tripping of Prime-mover (vii) Tripping of All unit auxiliaries (viii) ‘Class A Trip’ annunciation appears.

3.1.2 Class B Actions are taken when Class B protection is activated: (i) Tripping of The boiler (ii) Tripping of the turbine (iii) the ‘Class B Trip’ annunciation appears (iv) Class A protection will be provided by a low forward power relay. A time-delayed relay is one with low forward power.

3.1.3 Class C Actions are taken when Class C protection is activated: (i) Tripping of generator breaker (ii) Generator feeds its house load until the system fault is cleared.

3.2 Faults and Abnormal Conditions with Its Protection Various faults and abnormal conditions occur in synchronous generators, allowing the protective system to be designed. The significant faults and aberrant conditions have been covered in this section.

3.2 Faults and Abnormal Conditions with Its Protection

27

3.2.1 Insulation Failure of the Stator Winding Insulation failure can cause a fault between conductors and between the conductor and the iron core. Overvoltage or overheating can cause the breakdown, which can be caused by overloads, unbalanced currents, ventilation issues, or cooling system failure. It can also be caused by insulation damage induced by conductor movement produced by pressures generated by short circuits or out-of-step situations (Oza 2010). The protection that suggested for the above condition is differential protection. However, for small ratings, one can use the voltage-monitored time overcurrent relays.

3.2.1.1

Generator Differential Protection

(1) Current/Mertz-Price Differential Protection: Figure 3.1 shows Mertz-price differential protection. The current differential protection protects against all kinds of phase-phase and phase-to-ground faults. A high speed differential relay covering the three phases independently is generally used to protect against stator phase defects. Applied to generators with capacities greater than 1 MVA. If the CTs are equal in nature, the differential relay will operate correctly. In reality, however, CTs with comparable saturation characteristics are impossible to obtain. As a result, even if the main currents are the same, the secondary currents of CTs are asymmetrical. This current is commonly referred to as spill current. This spill current flows through the relay and may cause it to malfunction if its value exceeds the relay’s preset. Furthermore, if the length of the connecting wires (also known as pilot wires) is uneven, the spill current rises. One strategy is utilized to avoid differential relay malfunction in these instances. One can use a stabilizing resistance in the spill current

Fig. 3.1 Circulating current/Mertz-Price differential protection

28

3 Protection of Synchronous Generator

Fig. 3.2 High impedance differential protection

path. The inclusion of stabilizing resistance, on the other hand, lessens the sensitivity of the relay during an internal malfunction (Bhalja et al. 2011). (2) High Impedance Differential Protection: Consider Fig. 3.2 for high impedance differential protection. The primary operating current is given by I R = C T R × (Is + (n Ie )) (3.1) where I R : Primary operating current CTR: CT ratio n: Number of CTs in parallel with the relay element Ie : CT magnetizing current Is : Relay setting of spill current Voltage across the relay circuit is given by VR = I f × (RC T + R L )

(3.2)

VR = I R × (R R + R ST )

(3.3)

I R = I f × (RC T + R L )/(R R + R ST )

(3.4)

The knee-point voltage (KPV) of a CT decides the working range of the CT, and it should be high for higher saturation flux density. The KPV of the CT is given by VK ≥ 2 × Vr

(3.5)

3.2 Faults and Abnormal Conditions with Its Protection

29

Stabilizing resistor R ST limits the spill current below the relay setting. R ST =

Vr RR Is

where, R R is the relay burden resistance given by R R =

(3.6) Relaybur den . (Relaysetting)2

When the computed value of the stabilizing resistance (RST) is introduced to the circuit, the sensitivity of the relay to internal faults is lowered. Furthermore, when severe external failures, it generates a large voltage across the CT. To avoid this difficulty, in reality, the value of RST might be taken to be around one-third of the computed value. Problem 3.1 3.1 A 200 MW, 13.8 kV, 0.9 power factor (PF), 50 Hz, 3-phase, Y-connected generator is protected by differential protection with CTs with ratio 10000:5 A, CT secondary resistance of 1.5 , and lead resistance of 0.30 . The rated current of the differential relay is 5 A, and its setting range is 5–20% of the relay rated current. The relay burden is 1 VA. If a through-fault occurs, with the fault current 12 times the full load current of generator, then determine the value of the stabilizing resistance. In addition, suggest the suitable value of KPV of the CT. The primary disadvantage of Mertz-Price differential protection is that the relay’s sensitivity is reduced due to the inclusion of the stabilizing resistance. As a result, biased differential protection is employed to reduce this impact while simultaneously increasing the sensitivity of the differential relay. Solution: Rated full load current (I f l ) of the generator, M W × 106 = 9297A Ifl = √ 3 × kV × p · f. × 103 Now, the fault current is 12 times the full load current of the generator Hence, I F = 12 × 9297 = 11565 A The CT secondary current for this fault current is given by if =

11565 × 5 IF = = 55.782 A CT R 10000

A generator differential relay is set to pick-up at 5–20% of the CT secondary current. Therefore, we select the relay setting (Is) = 10% of the relay rated current, that is, 0.5 A. Relay resistance (RR ) =

1VA 1VA Relay burden = 2 = = 4 ( Pick-up )2 Is (0.5)2

30

3 Protection of Synchronous Generator

Voltage across the relay circuit is given by VR = i f (RCT + R L ) = 55.782(1.5 + 0.3) = 100.4076 V Stabilizing resistance is given by, R ST =

VR 100.4076 − 4 = 196.8152 − RR = IS 0.5

The actual value of RST is one-third of the calculated value. Hence, RST = 65.6 = 66  can be selected. This resistance is connected in series with relay. The KPV of CT is given by VK > 2 × VR = 200.8 V Hence, considering some safety margin, the KPV of the CT used for differential protection should not be less than 250 V. (3) Dual Slope Bias Differential Protection: For differential protection, the current entering the generator (current at the neutral end) and the current exiting the generator (current at line end) can be monitored at regular intervals (sampling rate). The following is an explanation of the mathematical process: Idi f f = I1 − I2 = Differential current; Istab = I1 + I2 = Stabilizing current; Class A protection is activated if Idi f f > K × Istab . As seen in Fig. 3.3, the characteristic might have two slopes (K). In Fig. 3.3, S is the sensitivity threshold which is set according to how much minimum current is required to be sensed. S-1 is slope-1 which avoids malfunction due to CT errors. S-2 is slope-2 which provides through fault stability and avoids problems of CT mismatches (Desai and Makwana 2021). Problem 3.2 3.2 Figure 3.4 shows the single line diagram of a generator winding which is protected by a percentage differential protection scheme. The relay settings for the said scheme are as under. Pick-up current = 0.05 A Slope = 10 % A high resistance single line-to-ground fault occurs when the generator is supplying the power to the load. The magnitude of current through CT1 and CT2 is 400 and 375 A, respectively. Determine whether the CB will trip by the relay in case of the given situation. Also, find out whether the relay will operate at the given value of fault current if the generator were carrying no load with the CB open. Solution: Case 1: CB (52) is in closed condition: When CB is in closed condition, the operating and restraining current is calculated as under.

3.2 Faults and Abnormal Conditions with Its Protection

31

Fig. 3.3 Dual slope characteristic of synchronous generator differential protection

Fig. 3.4 A single line diagram of a generator winding which is protected by a percentage differential protection scheme

i 1 = I1 /CT ratio = 400/400 = 1 A, i 2 = I2 /CT ratio = 375/400 = 0.9375 A Operating current = i 1 − i 2 = 1 − 0.9375 = 0.0625 A. Restraining current = (i 1 + i 2 )/2 = 0.96875 A. Operating current/Restraining current is less than the bias percentage 0.0625/0.96875 0.1 and the relay current (i 1 − i 2 ) is also more than the minimum pick-up current of 0.05 A, the relay will operate. (4) 100% stator earth-fault protection A restricted earth-fault protection can be used for small generators and when the neutral of the generator stator windings is earthed through a modest impedance. However, It cannot protect 100% stator windings. It is evident that an earth-fault at the generator’s neutral is not dangerous since no-fault current flows. However, if an insulation failure at the neutral goes undetected, the machine is directly earthed, and a second earth-fault is likely to cause entire machine damage due to extraordinarily high fault currents and concomitant mechanical pressures. The most effective technique to safeguard 100% of the winding is to employ a third harmonic (zero sequence) line-to-neutral voltage generated by most machines under normal conditions. Figure 3.5 depicts a relaying scheme based on this approach. Even under normal operating conditions, the generators produce a third harmonic voltage of 1–3%. The relay 2 in Fig. 3.6 contains a blocking filter that renders it relatively insensitive to fundamental frequency voltage. The third harmonic voltage setting on this relay is quite sensitive. It is tuned at between 0.3 and 0.6 V. (neutral PT secondary voltage is 110 V). As a result, in normal operation, the relay 2 is picked up and its contact is open. Relay 3 is configured to operate at the rated generating voltage and is sensitive to fundamental frequency voltage. As a result, the relay 3 remains energized and its contact is closed under normal conditions. If a fault develops near the generator neutral, the third harmonic voltage Vn becomes very low (Vn = 0 for the neutral fault), causing the relay 2 to de-energize and its contact to shut. As a

3.2 Faults and Abnormal Conditions with Its Protection

33

Fig. 3.6 Operation of 100% stator earth-fault protection

result, the alarm or tripping will be set as intended. 95% relay 1 with a blocking filter (blocking the fundamental frequency voltage) is configured to create a greater third harmonic voltage than a machine working normally. Relay 1 is still turned off. If a fault occurs within 95% of the winding from the terminal, the relay 1 functions, closes its contacts, and the circuit breaker trips. The interlocking contact of the relay 3 is desirable because it will open out while manually halting the machine, avoiding unwanted tripping of the 100% stator earth-fault system, which would otherwise be established by closing the contact of the relay 2 while stopping the machine. When starting the machine in no-fault mode, the relay 2 must always be activated before the relay 3. This is how a 100% stator earth-fault protection system is created to work. Class A protection is provided. The relays utilized in the scheme feature a time delay that may be adjusted. Problem 3.3 3.3 An earth-fault relay with a 10% setting protects an 11 kV, three-phase, 30 MVA star-connected alternator. Determine the amount of the resistor and the percentage of the winding protected if the neutral resistance restricts the maximum earth-fault current to 40% of full load value. Determine the value of the earth resistor required to leave only 9.5% of the winding exposed. The CT/A ratio is 2000/1.

34

3 Protection of Synchronous Generator

Full load current of the generator, 30 × 103 IF L = √ = 1574.6 A 3 × 11 Maximum earth-fault current, i.e., the current for the earth-fault at the terminal of the generator, I f m max = 1574.6 × 0.4 = 629.8 A The value of this earth-fault current can also be found in the equation I f max =

3E ph Z 1 + Z 2 + Z 0 + 3Z n

where Eph is phase-to-neutral voltage of the generator; Z 1 , Z 2 , Z 0 are positive, negative and zero sequence impedances, respectively, and Z n is the neutral circuit impedance. Normally, Z n is much large compared to Z1, Z2 and Z 0 and hence the earth-fault current can be approximated as I f max =

E ph Zn

For finding out the value of neutral resistor, E ph 11 × 103 =√ = 629.8 Zn 3 × Rn Rn = 10.08 The earth-fault current for the earth-fault at p% of the winding from the neutral end, 11 × p × 103 = 6.3pA If = √ 3 × 100 × Rn This when equated to sensitivity of the relay, 6.3p = 200 p = 31.74 This means that 31.74% of winding is unprotected or 68.26% of the winding is protected. For allowing only 9.5% of the winding unprotected, the earth-fault current will be 11 × 9.5 × 103 If = √ 3 × 100 × Rn 603.33 = 200 Rn

3.2 Faults and Abnormal Conditions with Its Protection

35

Rn = 3.016 This will certainly increase the maximum earth-fault current 11 × 103 I f max = √ = 2105.72 = 133.73% A 3 × 3.016 This calculation clearly indicates that any effort to increase the portion of the winding protected will increase the maximum earth-fault current if the relay sensitivity is kept constant.

3.2.2 Insulation Failure of the Rotor Winding A fault to earth has little impact if the rotor winding is ungrounded, as is common practise, but a second fault to earth will increase the current in a portion of the winding and may also imbalance the air-gap fluxes, causing strong vibrations that may cause serious damage. A rotor’s second earth-fault may also generate local heating, which can gradually bend the rotor and cause severe eccentricity; this can also cause vibrations and significant damage.

3.2.2.1

Rotor Earth Fault Protection Scheme

Figure 3.7 depicts one way for identifying rotor earth-faults. A dc voltage biases the field circuit, as shown in Fig. 3.7. If there is a ground fault, electricity will flow via an extremely sensitive dc relay, which can trigger an alarm or a class A trip as needed. The relay is a polarization sensitive moving iron relay. Because ac relays cannot be made exceedingly sensitive. This is owing to the current that ordinarily flows via the capacitance of the rotor winding to its core, and then through the bearings to the ground. When the first ground fault occurs at an unattended station, the protective relaying equipment must be configured to trip the main and field breakers of the generator. The standard procedure in attended stations is to trigger an alert when the rotor’s first earth-fault occurs. If a second earth-fault occurs, the generator’s main and field breakers must be tripped immediately. This approach, however, carries some danger since the vibration induced by a second earth-fault cannot be halted instantaneously, and the two ground faults may occur simultaneously or in fast succession.

36

3 Protection of Synchronous Generator

Fig. 3.7 Earth fault protection to rotor

3.2.3 Uneven Loading During normal synchronous rotation, the negative sequence component of unbalanced stator currents creates double frequency currents in the rotor. If the degree of imbalance is high, significant overheating can occur in the structural sections of the rotor, softening and weakening slot wedges and retaining rings, which are generally already under tremendous stress in big turbo-alternators. The system circumstances that would induce these dangerous unbalanced situations include, (i) Phase open-circuiting or failure of one circuit breaker contact (ii) An unsymmetrical fault near the power station that is not rapidly cleared. (iii) A fault in the stator winding. The time required for the rotor to tolerate this situation changes inversely as the square of the negative sequence current, i.e., I22 tK, where K is a constant ranging from 7 for a big steam turbo-alternator to roughly 60 for a salient pole hydro machine (see Fig. 3.8).

3.2 Faults and Abnormal Conditions with Its Protection

37

Fig. 3.8 Protection scheme against uneven loading

3.2.3.1

Negative Phase Sequence Relay

Figure 3.9 shows the negative phase sequence relaying algorithm which used to protect generator against any uneven loading condition. If the stator is carrying imbalanced currents, the rotor gets overheated. How long the generator can run under imbalanced loads is determined by the machine’s thermal withstand capability, which is determined by the type of cooling system used. The rate of heat generation is proportional to I22 Rt, where t is the time and I2 is the negative sequence current. Because a machine’s capacity to properly disperse energy is restricted to a fixed value k, we may write I22 Rt = k Assuming R to be a constant, and K = k/R, we get the thermal characteristics of the machine as I22 t = K So, IDMT with the below setting can detect the fault, t≤

K I22

38

3 Protection of Synchronous Generator

Fig. 3.9 Negative phase sequence relaying algorithm

The foregoing reasoning shows that if we could extract the negative sequence component of the stator current, we could construct protection against imbalanced loading by using the algorithm depicted in Fig. 3.9.

3.2.4 Field Failure Field excitation can be lost owing to a variety of factors, including: (1) Field loss to the primary exciter. (2) The field breaker was accidentally tripped. (3) A short circuit has occurred in the field winding. (4) Inadequate brush contact in the exciter. (5) Failure of the field circuit breaker latch. (6) Loss of alternating current supply to the excitation system. The protection against field failure popularly knows as loss of excitation protection of the generator. Consider the situation when the field excitation is lost but the mechanical input is not. Because the generator is already synced with the grid, it would try to maintain that synchronization by operating as an induction generator. As an induction generator, the machine runs slightly faster than synchronous speed

3.2 Faults and Abnormal Conditions with Its Protection

39

and is excitation powered by the grid. As a result, the flow of slip frequency current in the rotor, the current flowing in the damper winding, as well as the slot wedges and the surface of the solid rotor body. There are now two alternatives. The grid can either fully or partially supply this reactive power demand. If the grid can completely provide this need for reactive power, the machine continues to supply active power of P, MW but pulls reactive power of Q g MVA with no danger of instability. However, because the generator is not intended to be an induction machine, anomalous heating of the rotor and overloading of the stator winding will occur. If the grid was only partially capable of meeting the reactive power demand, the generator terminal voltage would decline. The generator would be overworked. There are several restrictions on how far a generator may be operated in the under-excited state. As a result, in the event of a loss of excitation, the operation must be rapidly noticed and verified in order to avoid a generator shutdown.

3.2.4.1

Protection Against Field Failure

The easiest way to identify loss of excitation is to monitor the generator’s field current. A loss of field signal can be raised if the field current falls below a certain threshold. The slip frequency current created in the case of loss of excitation and operation as an induction generator complicates this protection. When a generator loses field excitation, the impedance measured at the stator terminals varies the greatest. When excitation is lost, the terminal voltage decreases, and the current increases, resulting in a drop in impedance and a change in power factor. As a result, a Mho relay situated at the generator terminals can detect loss of excitation unambiguously. Figure 3.10 shows the offset Mho relay located at generator terminal to measure the three-phase voltages and current in the event of loss of excitation. In order to keep the generator running as long as it is safe, the generator may not be tripped instantly in the event of a loss of excitation. When the relay detects a lack of excitation, an alert is raised and an attempt is made to see whether excitation can be restored. To prevent against loss of excitation, a mho type of distance relay with an offset characteristic can be utilized. The offset is calculated as Xd’/2. The relay’s impedance setting is Xd at a −90” angle, as illustrated in Fig. 3.11. The impedance trajectory can be traced at low initial output condition with loss of excitation which moves from first to third

Fig. 3.10 Off-set Mho Relay

40

3 Protection of Synchronous Generator

Fig. 3.11 Off-set Mho Relay settings with loss of excitation

quadrant and enter settings as depicted in Fig. 3.11. The relay action can be delayed by around 0.5–3 s to allow time for the control circuitry to switch over to the standby exciter.

3.2.5 Prime Mover Failure In the event of a primary mover failure, i.e., lack of mechanical input, the machine remains synced with the grid and operates as a synchronous motor. The machine now needs a tiny amount of active power (in comparison to its rating) from the grid to run the turbine and meet the machine’s losses. At the same time, because its excitation remains intact, the machine provides reactive power to the grid. Running in this mode is not detrimental to the generator, however, it is detrimental to a primary mover such as a steam turbine. In most cases, a loss of steam supply to the turbine results in the loss of the primary mover. When the machine is used as a motor, trapped steam in the turbine causes an unpleasant temperature rise and damage to the blades. As a result, if the prime mover fails, the generator must be tripped immediately. When the primary mover fails, the generator begins pulling actual power from the grid while continuing to deliver reactive power to the grid.

3.2 Faults and Abnormal Conditions with Its Protection

41

Fig. 3.12 Direction relay with MTPA settings for loss of prime mover

In comparison to the generator rating, the actual power consumed from the grid is fairly little. The generator pulls actual power, which is just enough to cover the turbine’s losses and load. As a result, the magnitude of the stator current is less than it was when it was producing, yet the stator current suffers a 180◦ phase shift (Donald 2006).

3.2.5.1

Protection Against Prime Mover Failure

From the previous section it implies that if we employ a directional relay with an MTA of 180◦ , it will detect the loss of the prime mover when the current phasor reverses. Figure 3.12 shows the location of directional over the current relay with required settings. The amplitude of this reversed current phasor, however, is fairly tiny in comparison to the forward current since the generator only pulls enough actual power to satisfy the losses and operate the turbine. As a result, the directional relay used to detect prime mover failure must be more sensitive than directional relays used for continuous-current applications. A reverse power relay detects this power reversal. As the alternator operates as an overexcited synchronous motor, it is essentially a directed relay with a leading maximum torque angle. It is a timed and class A protection system. Overheating of turbine blades does not occur instantly once the generator begins to function as a motor. In the event of an internal problem, differential protection kicks in immediately. At that moment, the bus feeds the internal fault, and if reverse power also trips instantly, operators will be confused about the cause of the trip. To minimize unwanted tripping during sudden power reversals, a sufficient time delay should be given. As power reverse, it is popularly known as reverse power protection. As depicted in Fig. 3.13, Low forward power relays 32GR, 32GY, and remain operational during normal alternator operation. These relays are programmed to trip when the forward power falls below 0.53% of the alternator’s rated power.

42

3 Protection of Synchronous Generator

Fig. 3.13 Circuit diagram of reverse power protection

3.2.6 Over Load Protection When the generator is overloaded, the winding insulation overheats in proportion to the overload. If the insulation’s allowed temperature limit is exceeded, the insulation will rupture, resulting in a stator problem. As a result, the stator winding temperature must be kept within safe limits. If a generator is loaded over its rated capacity, an alert is generally sounded so that the operator can throttle the turbine’s steam control valve to relieve the generator of overload. For overload protection, a fixed-time overcurrent relay might be employed.

3.2.7 Over Voltage Protection Aside from transient overvoltage induced by lightning, etc., overvoltage can be produced by overspeed or by a faulty voltage regulator. Lightning arrestors can provide protection against transient high frequency or impulsive overvoltage caused by lightning and switching surges. However, an overvoltage relay is necessary to safeguard the generator’s stator conductor insulation against power-frequency overvoltages.

3.2 Faults and Abnormal Conditions with Its Protection

43

On contemporary steam-driven generators, the voltage regulators react quickly enough to prevent major overvoltage from occurring when the generator loses load and terminal voltage rises due to acceleration or line charging current. The best overvoltage relay will have two units: an instantaneous unit that trips at 25% (steam) or 40% (hydro) overvoltage, and an inverse time unit that starts at 10% overvoltage. The relay will be powered by a PT secondary.

3.2.8 Over Speed Protection Consider a turbo-alternator producing its rated actual electrical power P to the grid. Its mechanical input Pm is virtually identical to P (save for losses), and the machine works at a constant synchronous speed Ns. Consider that the generator has tripped and has been disconnected from the grid owing to some defect. As a result, P equals zero. However, the mechanical power input Pm cannot be rapidly lowered to zero. As a result, the generator has full input mechanical power but no output electrical power. If the mechanical input is not rapidly decreased by the speed-governing system, the machine will accelerate to dangerously high rates. Protection against such an occurrence can be given by monitoring overspeeding and taking actions such as closing the steam valve to cease steam input to the turbine. The turbine’s speed-governing device, or speed governor, is primarily responsible for recognizing this situation. An over-frequency relay or monitoring the output of the tachogenerator installed on the generator shaft can also identify overspeeding. Figure 3.14 depicts the rationale of anti-over-speeding protection.

Fig. 3.14 Protection algorithm against over speed or frequency

44

3 Protection of Synchronous Generator

3.3 Conclusion This chapter has addressed generator protection from stator defects through practical instances. With the slope-1 and slope-2 settings described in this chapter, a dualslope differential relay provides dependable protection. However, a 100% stator fault prevention technique that overcomes the limitations of current differential schemes is disclosed. This chapter also discusses synchronous generator abnormalities and its remedy such as loss of excitation, loss of prime mover, over speed, etc. A specific OOS problem is described in Chap. 4. This chapter provides the most professional protection practise for synchronous generators against faults and anomalies.

References Bhalja B, Maheshwari, Chothani N (2011) Protection and Switchgear. OUP India Desai JP, Makwana VH (2021) Modeling and implementation of percentage bias differential relay with dual-slope characteristic. IEEE Texas Power and Energy Conference (TPEC) 2021:1–6 Donald R (2006) Protective relaying for power generation systems. CRC/Taylor & Francis Oza BA (2010) Power system protection, and switchgear. Tata McGraw-Hill Education Private Ltd

Chapter 4

Adaptive Out-of-Step Protection of Synchronous Generator

Abstract The existing out-of-step protection schemes have proven deficient in preventing significant outages (Liu et al. 2018). Out-of-step protection schemes must not operate in stable power swing and rapidly isolate an asynchronous generator or group of generators from the rest of the power system in case of unstable power swing. The chapter proposes a novel phasor measurement unit (PMU) which incorporated a polygon-shaped graphical algorithm for out-of-step protection of the synchronous generator. The unique PMU-based logic further calculates the type of swing once the graphic scheme detects it. The novel graphical logic scheme design in this work can identify the complex power swing produced in the modern power system. The proposed algorithm can take the correct relaying decision in power swing due to renewable integration, load encroachment, and transient faults. This work uses the original and modified Kundur two-area system with a power system stabilizer (PSS) to test the proposed algorithm using MATLAB software environment. It provides assessment results of the proposed relay on the Indian grid system during the July2012 blackout. The results declared that the proposed algorithm is fast, accurate, and adaptive in the modern power system and has better performance than the existing out-of-step protection schemes. Keywords Blackout · Phasor Measurment Unit · Power swing · Out of step protection

4.1 Introduction The most common and inevitable disturbances of the power system are faults and sudden load changes. These disturbances lead to transient instability, depending upon its severity (Gunasegaran et al. 2015). Power system stability plays a vital role in the reliable operation of the power system. The transient disturbance causes power oscillations, which is known as a power swing (Ayer and Gokaraju 2019). Power swings can be synchronous (synchronous generators do not lose synchronism) or asynchronous (synchronous generators lose synchronism) depending upon the severity of the disturbance (Machowski 2019). On the generation side, when the genera© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 J. K. P. Desai and V. Makwana, Power Swing Detection and Generator Out-of-Step Protection Under Renewable Power Source Integration, Energy Systems in Electrical Engineering, https://doi.org/10.1007/978-981-19-9546-0_4

45

46

4 Adaptive Out-of-Step Protection of Synchronous Generator

tor or group of generators is exposed to the asynchronous power swing, instability may occur, and the generator experiences a pole slip. The pole slipping can cause stress on the system and pole (coil) (Gunasegaran et al. 2015). Deep synchronous or asynchronous power swings are accompanied by massive changes in voltages and currents, posing a serious threat to power system operation. An unstable power swing may cause mal-operation of some protection systems, especially distance protection and under impedance protection which in turn may lead to cascaded outages and blackouts (Machowski et al. 2020). For reliable operation, out-of-step protection scheme is provided, which has the following main elements (Machowski et al. 2020): • • • •

Special protection and supplementary control. Power swing blocking (PSB) of distance protection. Pole-slip protection (PSP) of synchronous generators. Out-of-step tripping (OOS) in transmission networks.

Nowadays, the power system integrated with PV (photovoltaic) generators leads to different transient events. The power swing characteristics of such a modern power system are complex as compared to the power system in which only a synchronous generator is the primary source of power generation (Gunasegaran et al. 2015). The penetration of PV and DFIG (doubly-fed induction generator) improves the system’s damping, but up to a certain level. If the penetration of solar PV is more than 50%, then it creates a negative effect on the damping of the system (Choudhary and Sharma 2015). The power system is protected against the power swing using the blinderbased out-of-step protection scheme. The method measures the Z to detect the outof-step impedance trajectories (C. B. S. Publishers 2021). The blinder-based relay may mal-operate for stable swing unless the user makes exact threshold estimation (Ambekar and Dambhare 2012). Furthermore, the OST (out of step) relays based on blinder must improve under different renewable penetration levels. Renewable power penetration increment into the existing power system can alter the coherent group of generators, swing frequency, and impedance trajectory. Furthermore, the increased penetrating leads to timely changes in the OOS scheme (Liu et al. 2018). Large-scale integration of wind generation changes the power swing characteristics (Haddadi et al. 2019). The power swing blocking time delay needs frequent revision with increased wind power penetration in the system. The paper (Haddadi et al. 2019) addresses the concern of protection improvements due to the negative impact of wind power generation but does not reveal any proposed solution to resolve the problem. The paper (Regulski et al. 2018) explains PMU application in the out-of-step protection. It has used direct phase angle comparison between different locations to identify out-of-step condition of the generator. However, the paper does not compare the proposed method with conventional out-of-step protection schemes like the double blinder. Distance relays are most likely to be picked up during the condition when the swing positioned arises near them on the transmission line (Lavand and Soman 2016). The PMU signals and line inputs are used to produce PSB (power swing blocking) signals during such conditions. The approach is described using normalized minimum positive sequence voltage on line as y-coordinate and positive sequence voltage angle

4.1 Introduction

47

as x-coordinate in Lavand and Soman (2016). The paper has shown the method for a line out-of-step protection and not for generator out-of-step protection. The generator out-of-step protection is the most important during power swing because recent advancements in the power system move the swing positioned away from the transmission line toward the generator and generator transformer (Donald 2006). The paper Kumar et al. (2022) presents adaptive out-of-step relay logic. The angular separation is calculated and compared with minimum and maximum angles. It simultaneously checks the energy of the rotor. Using all the information, it finds a stable or unstable power swing. Wide area monitoring system (WAMS) at network gives the real-time angular measurement to determine whether the system is going to collapse or remain stable (Phadke and Thorp 2017). However, the method used in Phadke and Thorp (2017) gives delays in the tripping of large generators, which may lead to damage in case of mal-operation of the scheme. WAMS, used in Zhang and Zhang (2017), introduces the out-of-step splitting protection based on the information of voltage phase angle at each bus. The paper discussed the technique, which is not complex mathematical but relatively simple, which encourages the use of PMU for the out-of-step protection. Adding to it, the wide-area protection scheme for prevention of widespread blackout is described in Franco et al. (2013) which used local signals available for distance relaying and global signals using synchrophasor measurements. It is used for controlled islanding. The local and global signals also can be utilized to take the correct decision from out-of-step relays after sudden disturbance. The authors of Zare et al. (2018) proposed wide-area out-of-step prediction using an adaptively controlled islanding concept. We note a new adaptive out-of-step protection approach in Zare et al. (2018) and Ariff and Pal (2016). But, the methods used for out-of-step protection in Zare et al. (2018) and Ariff and Pal (2016) are complex and must verify with different levels of renewable penetration. We proposed the novel solution, which overcomes issues of out-of-step relaying due to the impact of renewable integration, the usage of PSS, the fault during power swing, the load encroachments, and the symmetrical faults. The proposed algorithm identifies the out-of-step generator or group of generators during asynchronous power swing events. In this chapter, we found some underlying facts during power swing using PMU measurements, which can detect unstable swings early regardless of system changes and configuration. The settings and threshold of the proposed algorithm need not require frequent revision like a blinder-based relay. So, the proposed work significantly improves the out-of-step relaying by providing pole slipping protection to a generator in the present power system. Section 4.2 explains underlying facts during power swing using the mathematical analysis of PMU (phasor measurement unit) incorporated graphical OOS scheme. Section 4.3 explains the new algorithm steps and test system. Section 4.4 shows the test cases for different stable and unstable power swings due to the symmetrical faults with PSS. In Sect. 4.5, we describe the performance of the proposed PMU-based polygon shape OOS scheme for the line to line (LL), the line to ground (LG), and significant generator fault with the stable and unstable swing. A modified Kundur two-area system is used in Sect. 4.6 to see the performance of the proposed out-of-step relay under the impact of renewable integration. Section 4.6 also provides assessment results of the proposed relay on the

48

4 Adaptive Out-of-Step Protection of Synchronous Generator

Indian grid system during the July-2012 blackout. The comparison of the proposed relay with impedance-based relay and double blinder-based relay is explained in Sect. 4.7. Finally, the chapter presents the result and discussion in the end.

4.2 Mathematical Analysis of Proposed Out-of-Step Protection Scheme PMU at the generator bus can measure the magnitude and phase of positive sequence voltages and currents, local frequency, the local rate of change of frequency, circuit breaker, and switch status synchronously. Figure 4.1 shows the schematic diagram of the PMU-based generator out-of-step system. The advantage of PMU over the local measurement is that it can measure an absolute angle regarding the global timestamp or pure cosine wave. A positive sequence impedance can be measured using PMU data which is expressed below: Z ∠θ =

|V | (θ1 − θ2 ) |I |

(4.1)

where Z : positive sequence impedance, V : positive sequence voltage, I : positive sequence current, θ : relative angle between positive sequence V and I, θ1 : absolute angle of V and θ2 : absolute angle of I, f g : power generation frequency. The expression of a positive sequence current and a positive sequence voltage at the generator terminal is given as  I ∠θ2 =

E g ∠δ − E s X g + X T + Zs



V ∠θ1 = E g ∠δ − (I ∠θ2 ) × (X g )

(4.2) (4.3)

where E g = voltage behind transient reactance of generator G 1 , E s = system voltage, δ : the changing angle between the generator and system voltage, X g = the equivalent reactance of the generator, X T = reactance of the transformer, and Z s = system impedance.

Fig. 4.1 PMU-based generator out-of-step detection system

4.2 Mathematical Analysis of Proposed Out-of-Step Protection Scheme

49

Fig. 4.2 θ1 (degree), θ2 (degree), and θ (degree) versus time (s) for in event of unstable power swing

Substituting the value of I ∠θ2 in Eq. (4.3)  V ∠θ1 = E g ∠δ −

E g ∠δ − E s Xg + XT + Xs

 × (X g )

(4.4)

Assuming special condition, where n=

Eg = 1&1∠δ = cosδ + jsinδ Es

(4.5)

From (4.5), impedance observed by PMU data in real time is given by  Z ∠θ = (X g + X T + Z s ) × n ×

(n − cosδ) − jsinδ) (n − cosδ)2 + sin 2 δ)

 − Xg

(4.6)

From Eq. (4.6) it is clear that θ is propositional to δ. Where δ is angle between E s and E g and θ = θ1 − θ2 . Now if δ = 180◦ during pole slipping then θ = 180◦ . So, during pole slipping the θ is given by (4.7) θ1 − θ2 = 180◦ using Eq. 4.7, during the pole slipping event if an absolute angle of voltage θ1 is at 180◦ then an absolute angle of current θ2 approaches to 0◦ . After θ2 approaches to 0◦ , it crossed zero on time axis and reaches to −180◦ as shown in Fig. 4.2. At the time when the swing center arises near the GTU (generator transformer unit), then there is a sudden change of frequency (f ). If the rate of change of frequency ( ddtf ) exceeds the threshold value, which is 1.5 Hz/s in this work, then the generator may fall apart from the significant loads or completely (Elansari et al. 2012). The threshold value of ( ddtf ) is dependent on maximum permissible unbalance power through the synchronous generator and according to grid code for the rate of change frequency relay settings. The value of ( ddtf ) is calculated in this work using Eq. (4.8) at terminal G 1 by allowing maximum possible unbalance power which must produce the frequency variation less than the permissible range for grid frequency (Gupta et al. 2017). The permissible limit considered in the setting is ±3%, referring to the Indian grid code.

50

4 Adaptive Out-of-Step Protection of Synchronous Generator

Fig. 4.3 Polygon-shaped graphical characteristic of the proposed scheme

df (PG − PL ) × fs = dt 2 × SG × HG

(4.8)

where PG : generation output power in MW, PL : loading of the system in MW, f s : rated system frequency of the system in Hz, SG : generator rating in MVA, and HG : inertia constant of generating plant MW-s/MVA. Power swing impedance trajectory which passed 120◦ is generally not recoverable (Donald 2006). An associate generator breaker (GB) gets much stress when OOS trips after 120◦ to 180◦ . The switching of breaker after 270◦ produces lower stress on GB. We have considered Eg = Es (n = 1) in Eq. (4.6) for visualization of impedance trajectory. The visualization of impedance trajectory is shown in Fig. 4.3. Rslope1 and Rslope2 settings are based on the point A and point B, respectively, as shown in Fig. 4.1. Now, point A is set to 2  which is inside the maximum loading condition preferably near to zone-6. Now Point B is set to 0.79911  which is inside 35◦ from Point A. The parameter setting procedure is shown in Table 4.1. In Fig. 4.3, point A is the intersection point between the right slope 2 (Rslope2) with horizontal line of resistance axis (X = 0); point B is the intersection point between the right slope 1 (Rslope1) with horizontal line of resistance axis (X = 0); point C is any point on the maximum reactance reach of the inner polygon; point D is any point on the maximum reactance reach of the outer polygon; D1 and D2 are the power swing angles at points A and B, respectively; and Lslope1 and Lslope2 are the left slopes 1 and 2, respectively. The actual swing characteristic is not as same as considering using special conditions. So, to detect the sophisticated and unusual swing in the case of PSS and increasing renewable penetration, a new polygon-shaped graphical characteristic is designed. The kundur two-area system is considered as a test system in this work, and its details are in appendix (Kundur

4.3 Proposed PMU-Incorporated Polygon-Shaped Out-of-Step Protection Scheme Table 4.1 Settings of polygon-shaped characteristic Settings Value

51

Calculation procedure

Radius of Z 1 Radius of Z 2 Radius of Z 3 Distance of point A Distance of point B Point D Point C Slope of line A and B X d XT Zs

0.4253  0.6380  1.0633  2 0.79911  1.46665  1.333  tan60◦ 0.73315  0.3  2.81 

80%Z G S 120%Z G S 200%Z G S Inside the maximum load From A, inside 35◦ 2 × X d From D, inside 30◦ At system line slope angle Transient reactance From ratting Assuming healthy system

Ts

0.025 s

(D2◦ −D1◦ )× f n Sli p f r equency×360◦

Fig. 4.4 A kundur two-area system (test system)

2021). MB (multi-band) PSS with simplified settings (IEEE type PSS4B, according to IEEE Std 421.5) is working as PSS in the test system shown in Fig. 4.4.

4.3 Proposed PMU-Incorporated Polygon-Shaped Out-of-Step Protection Scheme Figure 4.5 shows the flowchart of the proposed out-of-step protection scheme, and the algorithm of the proposed scheme is as follows: Step 1: PMU continuously measures the impedance Z ∠ θ . If the impedance trajectory crosses from right slope 2 to 1, the timer starts and calculates T1 . For visualization, refer Fig. 4.3.

Fig. 4.5 Flowchart of the proposed out-of-step protection scheme

52 4 Adaptive Out-of-Step Protection of Synchronous Generator

4.4 Implementation and Testing of the Proposed Algorithm

53

Step 2: If T1 ≥ Ts (Delay time) indicates slow power swing due to sudden load change, load encroachment, or other system power flow changes. If T1 < Ts then it means fast power swing due to the fault. Step 3: Now, the graphical algorithm checks that Z∠ θ passed through system slope or passed the left slope 2 from the right slope 2 and also simultaneously from PMU, it checks the condition that whether the θ > 270◦ in real time. To visualization this condition, refer Fig. 4.3. If both the requirements explained in step 3 are satisfied in real time, the out-of-step pick-up condition is detected. The OST is declared after step 4. Step 4: Using PMU measurements at generator bus, the algorithm checks the following conditions: (1) ddtf > j (where j = threshold value) (For the test system, j = 1.5 Hz/s); (2) V ∠ θ1 crossing zero; (3) I∠ θ2 crossing zero; (4) θ passed left slope 1. Finally, the OST declares if step 3 and step 4 are correct. Table 4.1 gives the setting and its calculation procedure for polygon-shaped characteristics using G 1 data of Kundur two-area system.

4.4 Implementation and Testing of the Proposed Algorithm The three-phase short circuit fault has been created just before GTU, as shown in Fig. 4.4. In this first event, the fault removes before the critical clearing time (CCT) of the system. The results of the impedance trajectory of this case are shown in Fig. 4.6. CCT has been found by repetitive simulation at given fault location by varying the fault duration till system lost its synchronism. Also, Liu et al. (2018) has been used for CCT for given location. In the next event, the fault is removed after the CCT and so the system becomes unstable which is shown in Fig. 4.7.

4.4.1 Event-1: Fault Remove Before CCT The CCT for the three-phase fault of the system at fault located in Fig. 4.4 is 0.225 s, which is found by a simulation study. At t = 0 s, the impedance point is outside the right slope 2. After 0.004 s, the impedance trajectory crossed the right slope 2, and in the next 0.003 s, it crossed the right slope 1. So, Step 1 of the proposed algorithm gets satisfied, and the timer calculates time (T). The calculated time T = 0.0036 s ≤ Ts (delay time), which indicates a fast power swing from step 2. After the fault clearance, the point comes right back near the previous load point after 0.25 s. The algorithm verifies that θ is crossing 270◦ after fault is cleared, so OOS picks up is announced. Figure 4.6 shows the plot of the impedance trajectory of event-1.

54

4 Adaptive Out-of-Step Protection of Synchronous Generator

Fig. 4.6 Stable power swing due to swing center arises near GTU of G 1

Fig. 4.7 Impedance trajectory of unstable power swing

Step 4 decides whether this complex swing is stable or not. In this event, conditions (2), (3), and (4) of step 4 are satisfied, but (1) is not satisfied. Hence, OOS blocking was detected, and no tripping was announced. Step 4 identified it as a stable power swing. The decision taken by step 4 is verified as the system gains its equilibrium as shown in Figs. 4.6 and 4.8.

4.5 Testing of Proposed Technique for LG, LL Faults and Swing …

55

Fig. 4.8 Pole a slip occurrence and system remains stable

Fig. 4.9 δ f /δt ≥ 1.5 (Hz/s) versus time (s)

4.4.2 Event-2: Fault Remove After the CCT In this event, within 0.02 s after the fault, the swing passed the right slope 2 to right slope 1 via the system slope and moved back with crossing left slope 1. The timer calculated total time T = 0.0036 s ≤ Ts (delay time) and so step 2 indicates as fast power swing due to fault. The proposed algorithm’s step 3 is satisfied very quickly in this event, and OOS picks up announced. Now, step 4 decides the type of power swing. From Figs. 4.7 and 4.9, it can be seen that conditions (1), (2), (3), and (4) are satisfied. Finally, the OST command is sent to the generator breaker for rapid disconnection. Interestingly, the actual loss of synchronism due to angle separation takes 3.92 s after the fault, but the proposed algorithm detects it before 1 s of it. Early detection is beneficial for decision-making to other relays in coordination with the out-of-step relay. The decision is verified correctly as the diameter of the swing trajectory keeps on increasing, as shown in Fig. 4.7.

4.5 Testing of Proposed Technique for LG, LL Faults and Swing Center Arise Inside Generator The different faults at a given location in Fig. 4.4 have different critical clearing times (CCT). The proposed algorithm tests for LG, LL, and LLL faults at a given location. CCT found 0.7 s for LG fault, 0.4 s for LL fault, and 0.225 s for LLL fault.

56

4 Adaptive Out-of-Step Protection of Synchronous Generator

Fig. 4.10 Rotor angle deviation during power swing due to LG, LL, and LLL faults

Table 4.2 Comparison of proposed algorithm and DB scheme in different transient events with PSS enable Event DB decision Proposed scheme LLL fault (stable swing) LLL fault (unstable swing) LG fault (stable swing) LG fault (unstable swing) LL fault (unstable swing) LL fault (stable swing) Swing center in G 1 (unstable swing)

Incorrect Correct Incorrect Correct Correct Incorrect Correct

Correct Correct Correct Correct Correct Correct Correct

We simulated the swing center inside the generator by creating a three-phase fault at the terminal of the generator. Figure 4.10 shows the rotor angle deviation (δ) for different power swing conditions due to the LG, LL, and LLL faults. The rotor angle deviation increases as the fault clearing time increases. For the fault removed before the CCT of the system, the deviation is less than 2.5 rad, which finally achieves rotor angle stability under the action of PSS. For the LG, LL, and LLL faults, which are removed after the CCT of the system, the rotor angle deviation is too high which results in first swing instability under the action of PSS. From Fig. 4.10, it is clear that LLL fault near the GTU of G 1 significantly affects the rotor angle stability and provides a very less margin of control as compared to LL and LG fault near GTU of G 1 . Thus, the design of out-of-step protection required carefully considering the three-phase fault near the GTU of G 1 . Furthermore, analysis of different power swings due to LL, LG, and LLL faults also shows that the speed variation of the generator has less margin in the case of power swings than the transient faults. Table 4.2 shows the results of all the cases of faults. Table 4.3 presents the settings calculated for DB (double blinder) scheme for comparison with the proposed scheme.

4.6 Performance Test of the Proposed Algorithm Table 4.3 Settings used in DB scheme Type of settings Radius of mho element at origin Outer right blinder Inner right blinder Outer left blinder Inner left blinder Td

57

DB scheme 1.1 2 0.79 −2 −0.79 0.004 s

4.6 Performance Test of the Proposed Algorithm The performance test is divided into two parts: (1) Test under different levels of wind power penetration. (2) Test using the real physical system. I have designed the modified kundur two-area system for the simulation system to find the effect of different levels of renewable penetration on the proposed out-ofstep relay. Four identical type-4 DFIGs (doubly-fed induction generator) have been connected at buses BG 1 , BG 2 , BG 3 , and BG 4 such that total power flow remains the same from area 1 to area 2 (Liu et al. 2018). Figure 4.11 shows the modification in the Kundur two-area system to inject power from wind power sources by making the power transfer as same as the original kundur two-area system. I have provided results by assessing the proposed algorithm on the real Indian grid, which is shown in Fig. 4.13 as a simplified system just before the blackout. The Indian grid experienced the blackout on 30 (day-1) and 31 (day-2) July 2012. f21 Grid Disturbance (2012) and Report of the Enquiry Committee (2012), which are shown in Fig. 4.14. We have also used the real PMU data collected from data logged at a wide-area frequency measurement system (WAFMS) at the Indian Institute of Technology-Mumbai (IIT-Mumbai) to the proposed relay.

Fig. 4.11 A modified kundur two-area system

58

4 Adaptive Out-of-Step Protection of Synchronous Generator

Fig. 4.12 Impedance trajectory at different levels of renewable power penetration

Fig. 4.13 Antecedent condition of Indian grid on 30 July 2012 Fig. 4.14 Real PMU measurement on 30 July 2012 (Report of the Enquiry Committee 2012)

4.6 Performance Test of the Proposed Algorithm

59

4.6.1 Test Under Different Level of Renewable Power Penetration The CCT of the system at the given location for three-phase fault is keeping decreasing as the penetration of wind power increases. Table 4.4 shows the tripping time with different penetration. The impedance trajectory for different penetration levels is shown in Fig. 4.12. The detail of DFIG-IV is available in Trevisan et al. (2018). The proposed algorithm performs perfectly fine under renewable integration and gives better performance when the system keeps on weaker. The changed path of impedance trajectory with every increment of renewable power does not require to change the threshold and slopes in the proposed algorithm. The power swing characteristics are influenced by renewable power energy sources (RPES) and exhibit an unstable internal impedance. It also altered the reactance and resistance reach as shown in Fig. 4.12. This is due to limited current amplitude, current frequency offset, and controlled current phase angle present due to inverter-driven sources which is different than the synchronous-generator-based sources.

4.6.2 Test Using Real Physical System The different successive disturbing events on July 30, 2012 to the Indian grid result in a blackout. It is observed that before the blackout, number of transmission lines are not available or in force outage or planned outage or kept out to control high voltages. The lines which are out of service are presented with a dotted line in Fig. 4.13. The number of out-of-service lines results in a weak inter-region transmission network with high demand in the northern region (NR). Figure 4.13 shows 220 kV line with green color, 400 kV with purple color, and sequence of events with numbers. Also, Fig. 4.13 shows the condition before the disturbance with schedule power flow and actual power flow between different regions of the Indian power grid. It can be observed from Fig. 4.13 that grid is in stress condition before the blackout. The timing format used is HR:MIN:SEC:MILLISECONDS. The events before complete area separation between WR (western region)–ER (eastern region)–NR (northern

Table 4.4 Tripping time of proposed algorithm in different level renewable power penetration Penetration level (%) Swing due to disturbance Tripping time 7.6 14.1 21.2 28.3 35.4

Stable swing Stable swing Unstable swing Unstable swing Unstable swing

No trip No trip 0.16 s 0.06 s 0.025 s

60

4 Adaptive Out-of-Step Protection of Synchronous Generator

region) to NR are described as follows (Report of the Enquiry Committee 2012): • Event-1,02:33:11:907AM: 400 kV-Bina-Gwalior-1 line tripped because of zone3 tripping and 220 kV-Gwalior-Malanpur-1 transmission line tripping causing Malanpur and Mehgaon loads to be fed from the NR system. • Event-2,02:33:13:438AM: 220 kV-Bhinmal-Sanchor line trip on zone-1 due to power swing. • Event-3,02:33:13:927AM: 400 kV-Jamshedpur-Rourkela line-2 tripped on Zone-3 distance relay. • Event-4,02:33:13:996AM: 400 kV-Jamshedpur-Rourkela line-1 tripped on Zone-3 distance relay. • Event-5,02:33:15:400AM: 400 kV-Gorakhpur-Muzaffarpur-2 tripped on power swing. The above sequence is shown in Fig. 4.13 with its location in the Indian grid system. The assessment of the proposed algorithm for the above events is described as follows: • Step 1: Considering the NR end at 02:33:15:400HR, on 400 kV-GorakhpurMuzaffarpur-circuit-2, the measured phase voltages are 123, 116, and 115 kV and currents are 2.42, 2.48, and 2.45 kA for RYB phase, respectively. The calculated impedance for the control area at NR end during this time is in order of 0.65 . Hence, the impedance crosses right slope 2 to 1 in the proposed relay. • Step 2: It detects the power swing due to load change at 02:33:15:400HR due to slow trajectory reach to system slope. • Step 3: The θ > 270◦ from PMU measurement is not satisfied at 02:33:13:996HR as the angle difference of WR-ER-NR to NR does not exceed 270 ◦ C as shown in Fig. 4.14. The phase difference less than 270◦ implies that the relative angle (θ ) also not exceed the set value, which means that till at 02:33:13:996HR, the proposed algorithm is immune to trip the control area of NR at 2:33:13:996HR which is the desired operation. At 2:33:15:996HR, θ > 270◦ is satisfied and step 3 announces out-of-step pick-up as per the proposed relay logic. • Step 4: It is explained with each condition as follows: (1) ddtf > j (considering j = 1.5 Hz/s) is true at time of 2:33:11:8HR because as per our test using real measurements data logged at WAFMS-IIT Mumbai shows df > 1.5 Hz/s triggered at 2:33:11:8HR on NR end. dt (3) I∠θ2 also crosses zero at the time of 02:33:15:400HR on NR end. (4) θ already crossed 270◦ at time of 02:33:15:400HR on NR end. As per the proposed algorithm, out-of-step condition is identified at 02:33:15: 400HR for NR. The inquiry report also found that the actual separation of NR from WR-ER-NR occurred around at 02:33:15:542HR (Report of the Enquiry Committee 2012).

4.7 Comparison of the Proposed Relaying Algorithm with Existing Ones

61

4.7 Comparison of the Proposed Relaying Algorithm with Existing Ones The reliability of the proposed scheme is excellent as it used a balanced combination of direct and indirect measurements for power swing detection as compared to existing schemes like the DB scheme and impedance-based scheme. Table 4.5 shows the contribution of the proposed algorithm in terms of performance compared to existing methods like the impedance-based scheme and the DB scheme. Impedance-based scheme alone is highly unreliable in present power swing conditions as its relay reach if set high, then it may trip for stable swing and if relay reach is too small, then it is mal-operated for unstable swings (Kundur 2021). In the modern system, the reach is subjected to variation with renewable integration, as found using the analysis of Fig. 4.12. DB scheme is average reliable as it is based on a blinder, which measures the rate of change of impedance. However, the rate of change of impedance is subjected to change when the system used PSS. The results of Table 4.2 indicate that the DB scheme is not much reliable. Also, both schemes which are the existing ones are not adaptable for the same reason discussed just before. For the comparison of the accuracy of the proposed relay and existing ones, we have considered three criteria: (1) detection of swing, (2) time of out-of-step detection, and (3) detection at other places of generator location. According to the above three aspects, impedance-based relay performance is inferior and mal-operated frequently without changed settings (Kundur 2021) where the DB scheme is reasonable at detection of unstable power swing but mal-operated for a stable power swing as per results of Table 4.2. The result of the Tables 4.2 and 4.4 shows that the proposed relay is accurately operated at just before the actual loss of stability. Impedance-based method has no provision to identify symmetrical fault from power swing (Kundur 2021) where DB and proposed scheme have the ability to identify three-phase fault from swing but results shown in Table 4.2 that the DB scheme is not much accurate due to its existing methodology. Further, the effect of PSS does not affect the performance of the proposed method, which affects more and needs change in settings for existing methods.

Table 4.5 Performance comparison of the proposed scheme with existing schemes Parameter Impedance based DB scheme Proposed Reliability Adaptability Accuracy Detection of symmetrical fault Detection of fault during swing Effect of PSS

Less Not adaptable Less Not accurate

Average Not adaptable Average Not accurate

Excellent Adaptable Excellent Accurate

Not possible

Not possible

Possible

More

More

Not affected

62

4 Adaptive Out-of-Step Protection of Synchronous Generator

4.8 Limitation of the Method The method has high accuracy because of PMU Data. However, there are some limitations which have limited its usage. The limitation of the proposed method is as follows: (1) The PMU imposed additional cost and communication channel. (2) The local relay can take out-of-step decision after PMU agreement. (3) The system slope needs revision after long period of time.

4.9 Results and Discussion The effect on CCT of the LG fault with different level of penetration is shown in Table 4.6. The CCT of the LG fault is significantly reduced after 50% renewable power penetration. The proposed algorithm gives correct OOS tripping when the system becomes unstable due to the reduction of CCT and remains immune when the swing is stable at the same fault duration with different level of penetration. The proposed out-of-step algorithm adapts the renewable power environment and detects unstable swing faster as compared to without a renewable penetrated the system. For LG fault applied at a given location in Fig. 4.4 for a time more than CCT. The proposed algorithm gives a decision after 0.32 s of fault clearance when the actual loss of synchronism occurs after 1.82 s of fault clearance. DB scheme detects it faster than the proposed algorithm, less than 0.1 s of fault clearance. However, very fast tripping is not good when the large alternator is to be protected. The algorithm needs to take enough time to decide the swing nature and at the same time have to give a trip at least sometimes before the actual loss of synchronism occurs to left time for island action. The DB scheme gives false tripping for stable power swing produced due to transient LG fault. The proposed scheme proves very reliable here and does not produce an OST trip in a stable power swing event. Figure 4.15 shows impedance trajectory in the proposed scheme and DB scheme, respectively, for an unstable power swing due to LG fault. The system which is stable at 0–14% penetration has become unstable after 14–35% for the same disturbance at the same location in renewable power integrated power system. The proposed relay gives the correct OOS trip under

Table 4.6 Effect of renewable power penetration on system CCT for LG fault Penetration level (%) CCT of system (s) 0 8 36 52 58 62

0.6 0.55 0.41 0.2 0.15 0.15

4.9 Results and Discussion

63

Fig. 4.15 Impedance trajectory for LG fault in the proposed and DB scheme of unstable swing Fig. 4.16 Impedance trajectory for LL fault of unstable swing

unstable power swing after 14% penetration and remains immune for stable power swing up to 0–14% penetration. The swing center that arises inside the generator is simulated by applying a threephase fault at the terminal of generator 1. The swing trajectory is somewhat more dangerous when arising inside the generator and so the proposed algorithm detects it faster. The fault is applied at 2 s of run time and the proposed scheme gives the trip command after 0.106 s of fault clearance and the actual loss of synchronism is after 0.525 s. Figure 4.18 shows the OOS trip time comparison of DB and the proposed scheme. The DB scheme is not able to handle the severity of the fault and takes more time in this case and gives tripping after 0.558 s. The assessment of the proposed algorithm using real field data shows that the proposed algorithm has the capability to detect loss of synchronism at the correct time, which prevents the generators from damage due to power swings (see Figs. 4.16 and 4.17).

64

4 Adaptive Out-of-Step Protection of Synchronous Generator

Fig. 4.17 Untrue operation of DB scheme in stable power swing due to LL fault

Fig. 4.18 Tripping time comparison with DB scheme when swing center arises inside generator

Conclusion The transient events of modern power systems with PSS in use can be able to produce well-damped power oscillations. It produces power swings, which are much more challenging to detect. The PMU-based polygon-shaped graphical logic proposed in the algorithm accurately performs the task and detects complex power swing. The system is able to attain stability after one pole slipped in some events where the DB scheme mal-operated. The DB scheme is not reliable with PSS-enabled power systems as well as under renewable integration. Further, the DB scheme requires

4.9 Results and Discussion

65

modifications of blinders and time delays for correct operation in different power swing conditions. The proposed scheme is advantageous when a synchronous generator or group of generators is required to protect from unstable power swing. The proposed algorithm gives faster tripping when different levels of renewable power penetration govern unstable power swing. Impedance trajectory travels more distance from right to left while increasing the wind power penetration for the same fault, fault duration, and location. The result clarifies that the proposed scheme is speedy and correct when the swing center comes inside the large generator. The proposed out-of-step relay did not require revision of settings when system conditions changed. PMU data incorporated with a novel graphical scheme makes the algorithm adaptive in modern power system conditions. This chapter provides solution to the out-of-step protection of synchronous generator by combination of direct and indirect measurements. The next chapter focuses on an objective to prevent blackout by correct power swing detection thereby reducing unwanted tripping of out of step and its associate relay such as distance relay. In the next chapter, reader will understand OOS and PSB functions and its coordination with proposed direct method of OOS protection.

Application and Design Problem with Guideline 4.1 Implement the single-blinder-based out-of-step protection scheme for simple system shown in Fig. 4.19. The generator transient reactance is 0.2 p.u., transformer reactance is 0.07 p.u., and equivalent grid impedance is 0.14 p.u. at 85◦ with normal loading and 0.3 p.u. at 75◦ with line tripped. All the p.u. data is considered on generator base. The OOS relay is located as shown in Fig. 4.19. Assume generator base KV is 11 KV and base MVA = 100 MVA. Solution: Here, system impedance is given in p.u. which needs to be converted in ohms. As the OOS relay is located at generator terminal, full load current through generator is given as

Fig. 4.19 Simple system of generator connected to grid

66

4 Adaptive Out-of-Step Protection of Synchronous Generator

IF =

100000 √ 3×11

= 5248.52 A

11000 = 1.21  Now, Z base = √3×5248.52 Using Z base , circuit impedance can be calculated as given below: 

X d = 0.2j × 1.21 = 0.242j  X T R = 0.07j × 1.21 = 0.0847  Z s = 0.14 ∠85◦ × 1.21 = 0.1694  ∠85◦ The reliable settings are minimum settings for the scheme. It is fact that lower system impedance have minimum diameters swing locus. The values calculated above are primary site of CT. Those all values are converted into secondary side with CT ratio = 1000/1 A and 11000/110 V. 

= 2.42j  (s) xd = 0.242j × 1000 100 x T R = 0.0847j × × 1000 = 0.847  (s) 100 z s = 0.1694 ∠85◦ × 1000 = 1.694  ∠85◦ (s) 100 The values are plotted as shown below on R-X-plane. Basically the single blinder is set on 120◦ and 240◦ for outer and inner blinders, respectively. For the simple system, n = EEGS = 1. These is minimum settings which ensure that unstable power swing won’t travel the blinders before minimum set time which is generally four cycles. The points O and O’ have been calculated by line PA with QB and PC with QD intersection, respectively. Now blinder angle is calculated by X T ,  X T = xd + x T R + = 0.1476 + 4.9545j = 4.956  ∠88.29◦ . Finally, blinder 1 and blinder 2 are set at 88.29◦ with parallel to line PQ and passing from O and O’, respectively. Now D1 is the time set between blinder 1 and PQ line and D2 is time set between line PQ and blinder 2 which is proposal to x-axis distance between them. Here D1 = 2.1 s and D2 = 1.8 s approximately. Figure 4.20 shows the final design of single blinder scheme for protection of G1 of simple system as shown in Fig. 4.19. 4.2 A three section line feeder is shown in Fig. 4.21 which is to be protected by mho relay. The data of the system has been given as follows: (1) Section 1 impedance: 5 + 20j ohms (primary);m (2) Section 2 impedance : 4 + 15j (primary); (3) Section 3 impedance 3 + 10j (primary); (4) Rated load current of feeder is 1000 A at 0.8 power factor lag; (5) possible overloading = 200% of the rated current; and (6) possible voltage dip = 10%. Consider CT ratio = 1000/1 A and PT ratio = 132 kV/110 V. For Relay R1: (1) transient over reach = 10%; (2) characteristic angle = 60% of mho relay. Find the settings of zones 1, 2, and 3 of the distance relay R1. Solution: For the system shown in Fig. 4.21, Zone-3 reach is decided based on how much overloading and voltage dip is considered. The load impedance Z L for 10% voltage dip is given as

4.9 Results and Discussion

67

Fig. 4.20 Blinder settings

Fig. 4.21 Transmission network of Example 2

√ [(132000/ 3) × 0.9] ohms 2000∠ − 36.87 = 34.29∠36.87ohms

ZL =

(4.9)

Now 5% margin for relay errors, Z L = 0.95 × 34.29∠36.87ohms = 32.57∠36.87ohms 1000/1 × 32.57∠36.87ohms Z L ( secondary) = 132000/110 = 27.14∠36.87 ohms therefore

Z L = 29.51ohms cos(60 − 36.87) Z 3 = K 3 cos(φ − θ )

(4.10)

T3 =

= 29.51 cos(73.30 − 60) = 28.71ohms Impedance of Section 1 is Z 1 = 5+20j = 20.61 ∠75.96◦ . With 10% over reach = 1.1 × 20.61 ∠75.96◦ = 18.74∠75.96◦ .

(4.11)

68

4 Adaptive Out-of-Step Protection of Synchronous Generator

Fig. 4.22 Zone settings plots

At secondary side, Z 2 = 15.62∠75.96◦ . T1 =

Z1 = 16.27ohms cos(φ − θ )

(4.12)

Now, impedance of Sections 1 and 2 = 9 + 35j = 36.13 ∠75.57. At secondary side, the value becomes 30.11 ∠75.57. Here, it can be seen that Z 3 is less than above figure which indicates that Z 3 covers Section 1 and 98% Section 2 which is acceptable. Now, for Z 2 , considering Section 1 and 50% of Section 2 = 20.61 ∠75.96◦ + 50% (4+15j) = 7 + 27.5j = 28.37 ∠75.71◦ . At secondary side Z 2 = 23.54 ∠75.71◦ T2 = 24.37 ohms. Final settings Z 3 , Z 2 , Z 3 , T1 , T2 , T3 , and Z L  are shown in Fig. 4.22.

References Ambekar V, Dambhare S (2012) Comparative evaluation of out of step detection schemes for distance relays, p 6 Ariff MAM, Pal BC (2016) Adaptive protection and control in the power system for wide-area blackout prevention. IEEE Trans Power Delivery 31(4):1815–1825. https://doi.org/10.1109/TPWRD. 2016.2518080 Ayer N, Gokaraju R (2019) Online application of local OOS protection and graph theory for controlled islanding. IEEE Trans Smart Grid 1-1. https://doi.org/10.1109/TSG.2019.2943525

References

69

CBS Publishers, DP. Ltd204, PIA Delhi 110 092 (2021) Protective relaying: principles and applications, 4th edn. CRC Press, Routledge. https://www.routledge.com/Protective-RelayingPrinciples-and-Applications-Fourth-Edition/Blackburn-Domin/p/book/9781439888117. Accessed 18 Jan 2021 Choudhary S, Sharma FB (2015) Small-signal stability analysis of renewable source connected power system and identification of oscillatory modes using wavelet transform. In: 2015 international conference on smart grid and clean energy technologies (ICSGCE), Oct 2015, pp. 23–29. https://doi.org/10.1109/ICSGCE.2015.7454264 Donald R (2006) Protective relaying for power generation systems. CRC/Taylor & Francis Elansari AS, Edrah MF, Khaled SM (2012) Improve transient frequency response by adjusting generators’ over frequency relays. In: 2012 IEEE international conference on power and energy (PECon), pp 458–463. https://doi.org/10.1109/PECon.2012.6450257 Franco R, Sena C, Taranto GN, Giusto A (2013) Using synchrophasors for controlled islanding A prospective application for the Uruguayan power system. IEEE Trans Power Syst 28(2):2016– 2024. https://doi.org/10.1109/TPWRS.2012.2224142 Grid disturbance on 30th July 2012 and grid disturbance on 31st July 2012, Compliance, Aug 2012 [Online] Gunasegaran MK, Tan C, Bakar AHA, Mokhlis H, Illias HA (2015) Progress on power swing blocking schemes and the impact of renewable energy on power swing characteristics: A review. Renew Sustain Energy Rev 52:280–288. https://doi.org/10.1016/j.rser.2015.07.066 Gupta P, Bhatia RS, Jain DK (2017) Active ROCOF relay for islanding detection. IEEE Trans Power Delivery 32(1):420–429. https://doi.org/10.1109/TPWRD.2016.2540723 Haddadi A, Kocar I, Karaagac U, Gras H, Farantatos E (2019) Impact of wind generation on power swing protection. IEEE Trans Power Delivery 34(3):1118–1128. https://doi.org/10.1109/ TPWRD.2019.2896135 Kumar N, Nagaraja DR, Khincha HP (2022) A smart and adaptive scheme for generator out of step protection Kundur P (2021) Power system stability and control. Accessed 18 Jan 2021 Lavand SA, Soman SA (2016) Predictive analytic to supervise zone 1 of distance relay using synchrophasors. IEEE Trans Power Delivery 31(4):1844–1854. https://doi.org/10.1109/TPWRD. 2016.2521784 Liu G, Azizi S, Sun M, Popov M, Terzija V (2018) Performance of out-of-step tripping protection under renewable integration. J Eng 2018(15):1216–1222. https://doi.org/10.1049/joe.2018.0180 Machowski J (2012) Selectivity of power system protections at power swings in power system. Acta Energetica, Dec. 31. http://actaenergetica.org/article/en/selectivity-of-power-systemprotections-at-power-swings-in-power-system.html. Accessed 23 Nov 2019 Machowski J, Lubosny Z, Bialek JW, Bumby JR (2020) Power system dynamics: stability and control. Wiley Phadke, AG, Thorp JS (2017) Synchronized phasor measurements and their applications, 2nd ed. Springer International Publishing Regulski P, Rebizant W, Kereit M, Herrmann H-J (2018) PMU-based generator out-of-step protection. IFAC-PapersOnLine 51(28):79–84. https://doi.org/10.1016/j.ifacol.2018.11.681 Report of the Enquiry Committee on grid disturbance in Northern Region on 30 July 2012 and in Northern, Eastern and North-Eastern region on 31 July 2012 New Delhi, India, 2012 [Online]. http://powermin.nic.in/upload/pdf/GRIDENQREP16812.pdf Trevisan AS, El-Deib AA, Gagnon R, Mahseredjian J, Fecteau M (2018) Field validated generic EMT-Type model of a full converter wind turbine based on a gearless externally excited synchronous generator. IEEE Trans Power Delivery 33(5):2284–2293. https://doi.org/10.1109/ TPWRD.2018.2850848

70

4 Adaptive Out-of-Step Protection of Synchronous Generator

Zare H, Yaghobi H, Alinejad-Beromi Y (2018) Adaptive concept of controlled islanding in power systems for wide-area out-of-step prediction of synchronous generators based on adaptive tripping index. Transm Distrib IET Gener 12(16):3829–3836. https://doi.org/10.1049/iet-gtd.2018.0319 Zhang S, Zhang Y (2017) A novel out-of-step splitting protection based on the wide area information. IEEE Trans Smart Grid 8(1):41–51. https://doi.org/10.1109/TSG.2016.2593908

Chapter 5

Predictive Out of Step Protection

Abstract To avoid a blackout, power swing detection and an out-of-step protection method are essential. With the correct generator out-of-step relaying decision, power system stability improves. The chapter begins by presenting an accurate model of a synchronous generator. The predictive power swing detection and out-of-step tripping method are then presented in this chapter. In this chapter, the proposed out-of-step protection mechanism is implemented using the single machine infinite bus (SMIB). For the examination of the proposed out-of-step protection method for the transmission line, the contrasting situations with variations in fault resistance, position, and duration are considered. In the end, the chapter discusses the impact of small-signal disturbance and fault resistance on out-of-step protection. The proposed technique provides appropriate power swing detection and measurement performance. This chapter is based on Desai and Makwana (2022). Keywords Predictive · Power Swing blocking · Single machine infinite bus

5.1 Introduction The article Kosterev et al. (1996) provides an answer to the problem of how to assess remote protection relay tripping behavior when direct methods are used. Comprehensive tutorials on blocking the power oscillation and triggering out-of-step (OOS) relay collectively with advanced unconventional methods are shown in the article McDonald et al. (2005). Conventional Power swing Blocking (PSB) systems use the diversity between the impedance change rate during a fault and during a power oscillation to distinguish between a fault and an oscillation. If the measured impedance crisscrosses the concentric characteristics earlier than the timer lapses, the relay reports the incident as a system fault otherwise as a power oscillation Alinezhad and Karegar (2017). The blinder unit cannot distinguish between a fault and an OOS condition until the swing is passed through the second blinder within a given period. As such, the blinder-based scheme cannot be used to obstruct phase distance relays from tripping for unstable power swings because the relays will have tripped before the scheme declaring an unstable condition. The scheme can be used to stop the © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 J. K. P. Desai and V. Makwana, Power Swing Detection and Generator Out-of-Step Protection Under Renewable Power Source Integration, Energy Systems in Electrical Engineering, https://doi.org/10.1007/978-981-19-9546-0_5

71

72

5 Predictive Out of Step Protection

automated re-closure for a detected unstable power oscillation Oza (2010). Zubov’s stability boundary has been used in the paper Yellajosula et al. (2020) using the PMU measurements. Nevertheless, referring to IEEE C37.118 standard, the most prevalent reporting rate of industrial PMU is equal to one power system cycle. This means that synchronized measurements from the remote end are not available to the relay at the same sampling rate as local signals are sampled. the weakness of PMU necessitates the out-of-step detection based on local measurements. Further, the autonomous use of PMU is observed as unsafe when the swing center is far from the generator. The paper Desai and Makwana (2020) presents a phasor measurement unit that incorporates an adaptive out-of-step scheme that can distinguish the complex power swing. The mal-operation of transformer protection scheme in case of power swing due to outside high current faults can be restricted using percentage bias differential relay with dual-slope characteristic in the paper Desai and Makwana (2021). The variance in the frequency of voltage is determined during out-of-step oscillation and investigated in the research paper Zhang and Zhang (2019). The frequency differencebased algorithm for out-of-step protection of multi-terminal line was proposed by the researcher in the paper Krata et al. (2014). Using the PMU, the paper Zhang and Zhang (2017) presents out-of-step splitting protection. Still, the PMU application raises the cost of out-of-step protection scheme in Zhang and Zhang (2017). The use of artificial intelligence techniques for power system planning and protection is given in the paper Desai and Makwana (2017). The power swing influences the distance relay. Out-of-step relay required to produce power swing blocking command in stable power swing condition. The method using geometrical understanding using Clarke’s transformation is defined in the paper Camarillo-Peñaranda et al. (2020). The method is used to detect power swing from the three-phase fault with help of a swing voltage center (SVC). The counting method in paper Camarillo-Peñaranda et al. (2020) calculates the αβ voltage vector entry in the unstable oscillation zone. The power swing has complex nature under renewable power penetration conditions Desai and Makwana (2022). Under the before-mentioned condition, the method of paper Camarillo-Peñaranda et al. (2020) merely gives accurate detection. The predictive calculation-based power swing detection is described in the paper Khodaparast and Khederzadeh (2016). The method has some uncertainties of malfunction when the impedance trajectory is quicker than the computational time used. The paper Desai and Makwana (2020) presents the power swing blocking algorithm which measures the relative frequency between sending end bus and system frequency along with the voltage decay of the transmission line. The algorithm required communication between both ends of the transmission line. I have used local measurement alone which eliminates the use of PMU and communication facility in this paper. The approach proposed in this work can predict the power swing before it occurred which gives an edge over the method used by the researchers in papers Krata et al. (2014); Desai and Makwana (2020). Also, the method of this paper is based on local measurement which eliminates the weaknesses of PMU-based OOS scheme.

5.2 Modeling of Synchronous Generator with the Turbine, Exciter …

73

In this chapter, Sect. 5.2 starts with explicit modeling of the synchronous generator in simulation environments. Section 5.3 presents the proposed power swing detection method. Section 5.4 describes the proposed out-of-step (OOS) protection algorithm. Finally, the proposed OOS scheme is performed on the SMIB system and experimented with different power swing conditions in Sect. 5.5. The outcome and discussion are described in Sect. 5.6.

5.2 Modeling of Synchronous Generator with the Turbine, Exciter, AVR, Governor and PSS For a more practical simulation, the exciters, governors, and turbine models have to be incorporated in the simulation. Figure.5.2 shows a complete model of a synchronous machine with the turbine, the exciter, the automatic voltage regulator (AVR), governor, and power system stabilizer model (PSS). The generation model is developed in a PSCAD environment where it is feasible to configure a specific model. The type of model used in the system are as follows: (1) Generator: Hydro type is employed. (2) Governor: IEEE GOV1 type is employed (Mechanical-Hydraulic Controls). (3) Turbine: 4 number of the turbine with 3600 revolution/min. (4) Exciter: IEEE ACIA Type is employed. (5) Power system Stabilizer: IEEE PSS1A type is employed. The exciter is initialized at the instant when the source to machine signal is given. The same signal is applied inside the exciter model to define the instant that it should initialize its internal parameters and output the desired field voltage (E f ). In summing up to the aspired output field voltage to the machine reference voltage (Vr ), the field current (If), terminal voltage, and current (VT / IT ) are given as inputs to the AVR model. The program then determines the initialized value of the reference voltage (Vr ) required to keep the steady-state operating condition. Vr is the initialized voltage set point to keep up the stipulated steady-state terminal setting. After the initial transients have ended, the machine mode is initiated by switching signals. At this instant, the rotor will be rotating at a constant speed as the machine is still in the ‘locked rotor’ state. The governors and turbines may be initialized at the time instant when the rotor is unlocked, i.e., when the lock rotor signal goes high. Once this occurs, the mechanical dynamics are active. The simulation model with a governor or turbine model initialized in the same way as that used to initialize the exciter. Pr e f and ωr e f is the control signal that activates the initialization of the governor (Ambekar and Dambhare 2012). After the detail modeling and initialization, the machine reaches its steady-state output frequency at 15 s as shown in Fig. 5.1. The parameter of SMIB system shown in Fig. 5.2 is described in Table 5.1.

74

5 Predictive Out of Step Protection 377.35 377.3

ω (Rad)

377.25 377.2 377.15 377.1 377.05 377 0

2

4

6

8

10

12

14

16

Time(s) Fig. 5.1 Rated output speed waveform of synchronous generator model

Fig. 5.2 Modeling of synchronous machine in PSCAD

5.3 Power Swing Detection Method The power swing detection is identified using the below criteria: (1) Repetitiveness, (2) Continuation, (3) Regularity, (4) Impedance travel through the outer to the inner layer. The algorithm of power swing detection is shown in Fig. 5.4. ‘AND’ Gate is used at the end of both conditions in Fig. 5.4. Because it waits for the second part to be processed, and if one condition is satisfied early, the proposed algorithm does not declare a decision and waits for the second to be complete. That’s why the gate is used.

5.3 Power Swing Detection Method

75

Table 5.1 Test system components and its ratting Model name Parameter values Generator Transformer Transmission line Potential Transformer (PT) Current transformer (CT) Grid

V = 11 kV, I = 5kA, f = 50 Hz, H = 3.1 s 11 kV/230 kV, l = 0.12 pu, 50 Hz, 25 MVA 70 km, R = 0.0032 pu, Xl = 0.03 pu, B = 0.07 pu 230 kV/110 V, burden= 301  160/1, burden = 2.5 p.u 50 Hz, 230 kV

5.3.1 Repetitiveness The resistance (R) and reactance (X) rate is calculated and swing is considered to detect the change which verifies repetitiveness (Liu et al. 2018). It is important for power swing detection that at a time direction of R or X is fixed till the swing crosses the swing detection area to ensure repetitiveness. The zone-1, zone-2, zone-3, outer layer using zone-6, and inner layer using zone-5 are shown in Fig. 5.3. The repetitiveness is calculated using the following: If X 1 and Y1 is measure at ith sample and X 2 and Y2 is measure at (i + 1) sample then repetitiveness can be found using, X (i+1) = X 2(i+1) − X 1(i)

(5.1)

Y (i+1) = Y2(i+1) − Y1(i)

(5.2)

Reactance (X) ohm

20 10

0 -10 -20

Zone-1

-20

Zone-2

-10

Zone-5 (Inner layer)

Zone-6 (Outer layer)

0

10

Zone-3

20

Resistance (R) ohm Fig. 5.3 Different zones of protection, outer layer and inner layer settings in R-X plane

76

5 Predictive Out of Step Protection

Fig. 5.4 Power swing detection algorithm

From Eqs. (5.1) and (5.2), either X (i+1) or Y (i+1) has to be near to 0 or very small for impedance trajectory to be considered as repetitive.

5.3.2 Continuation To ensure continuation, the impedance trajectory needs to exceed the minimum C value, where C = continuation threshold. The continuation is measured as below Firstly, it measured, X (i+1) and Y (i+1) . secondly, at (i+1) sample, Z Ci+1 = X (i+1) + Y (i+1)

(5.3)

Finally, If Z c calculated using Eq. (5.3) is higher than C (continuation threshold) then swing consider as perpetual and not steady at a point. The continuation threshold depends on the sampling rate of the relay and for a greater sampling rate, the Z c is calculated as a summation of 1 to nth samples.

5.4 Out of Step Protection of Synchronous Generator

77

5.3.3 Regularity The regularity is measured after the impedance path crosses the inner layer (Zone5). Regularity will be ensured that Z is slow and approaching zone-2. To ensure the regularity of the impedance trajectory, the difference between two consecutive measured Z needs to be below the S (Regularity threshold) value. The two consecutive measures are taken at a time when impedance passed Zone-5. The value of S is equal to or less than the difference between Z (Impedance) measured at any point on Zone-5 and Zone-2.

5.3.4 Impedance Travel Through Outer to Inner Layer There are two boundary layers are outlined to detect the power swing. The outer layer is on the zone-6 boundary of the step distance protection scheme and the inner layer is on zone-5. The step distance scheme of Oza (2010) is employed to design the zones. The three-step mho type distance relay recognize for purpose as it covers the area on the R-X plane which reduced the power swing effect Paithankar and Bhide (2011). The time delays are used between Zone-6 to Zone-5, for Zone-3 and Zone2. Outer layer (zone-6) and inner layer(Zone-5) are used to detect slow impedance trajectory approaching to the zones of protection.

5.4 Out of Step Protection of Synchronous Generator The out-of-step (OOS) protection is performed for synchronous generators using the power swing detection algorithm explained in Fig. 5.4. The SMIB system as shown in Fig. 5.2 is considered for implementation. Once the power swing is recognized then the OOS scheme holds the following condition to declare the OOS tripping. (1) Condition-1: Swing must enter into zone-2. (2) condition-2: Once the swing is entered into zone-2 then it must not alter its direction X after leaving the zone. The condition-1 is tested by the steps distance relay unit of the transmission line which ensures that the swing has lain on the transmission line. The condition-2 ensures that the swing is not reversible once it moves toward the transmission line which means that it is an uncontrollable power swing. The flowchart of OOS tripping and power swing blocking (PSB) is shown in Fig. 5.5. If the OOS conditions are not fulfilled then PSB is transferred to the distance relay and the final decision is executed through the distance relay. The proposed power swing detection and out-of-step protection scheme are performed on the SMIB system and verified using different power swing conditions.

78

5 Predictive Out of Step Protection

Fig. 5.5 Power swing detection using outer and inner layer

In the case when the PSB is transferred to distance relay then distance relay verify zone entry. It holds the following settings for Z R (Relay measured impedance): (1) Z R < Zone-3 settings; (2) Z R < Zone-2 settings. In most of the cases, the calculated impedance is less than Zone-2 settings but in many circumstances, it may be the case that impedance is located in Zone-3. According to the settings of zones, the delayed tripping is sent to the transmission line circuit breaker. For Zone-3, the Td (delay time) is usually 0.5s or less and for Zone-2, the Td (delay time) is usually 0.3s or less. If the fault is cleared before the delay timer lapsed then no tripping has been sent.

5.4.1 Swing Center Appear in the Middle of the Transmission Line Section The three-phase fault has been applied at the middle of the transmission line to create a swing center. The fault is applied for varying duration with various fault resistance. Figure 5.6 shows impedance trajectory for the condition in which fault is applied for 0.2s duration with fault resistance of 0.01 in the SMIB system. Figure 5.6 points that the continuation, repetitiveness, and regularity are recognized. Also, the swing enters

5.4 Out of Step Protection of Synchronous Generator

Reactance (X) ohm

20

79

Reverse Direction Condition 1: False X = -0.8 (reverse)

10

0

-10

X: 4.739 Y: -2.049

Enter Zone-2 Condition 1: True Regularity Z = 0.2 < S

Repetitiveness X = 2.72 Y = 0.2

Continuation

-20

i+1 ZC

Zone-1

-2 0

Zone-2

=1.93 >C

Zone-5

Zone-6

0

10

-1 0

Z

path

20

Resistance (R) ohm Fig. 5.6 Power swing arise on the transmission line

the outer layer (Z6) to the inner layer (Z5) which identifies the swing perfectly. As per the proposed algorithm, the swing has to be valid for condition-1 and condition-2. From Fig. 5.6, it is obvious that condition-1 is true but condition-2 false as the swing comes opposite. Hence, the OOS was not detected which is the true decision.

5.4.2 Swing Center Arise Near the Generator Transformer The three-phase fault has been applied near the transformer to produce a swing center. Once again, the fault is applied for varying duration with various fault resistance. Figure 5.7 shows the impedance trajectory for the condition in which fault is applied for 0.04 s duration with fault resistance of 1 in the SMIB system. Figure 5.7 points that the continuation, repetitiveness, and regularity are detected. Also, the swing enters the outer layer (Z6) to the inner layer (Z5) which recognizes the swing correctly. As per the proposed algorithm, the swing has to be valid for condition-1 and condition-2. From Fig. 5.7, it is clear that condition-1 and 2 are true. Hence, the OOS detected is the true decision.

80

5 Predictive Out of Step Protection

Reactance (X) ohm

30 20

Same Direction Condition-2:valid

10 0 -10

Enter Zone-2 Condition-1:valid

-20

Zone-1

-20

Zone-2

-10

Zone-5

Zone-6

0

10

Repetitiveness Regularity Continuation

20

Resistance (R) ohm Fig. 5.7 Power swing arise near the transformer

5.5 Comparison with Existing Methodology The research paper Blumschein et al. (2014) uses the monotony, continuation, and smoothness of the power swing. Nevertheless, the speed of the impedance is not taken into account when entering the swing area. The shape of the zone considered is a polygon in Blumschein et al. (2014) by the researcher. The polygon shape covers an additional area on the R-X plane. The proposed power swing detection algorithm used repetitiveness where insignificant changes in R or X are seen as more reasonable. The calculation of the continuity threshold is not given in Blumschein et al. (2014) which immediately depends on the sampling rate of the relay. The proposed power swing detection technique in section-3 reveals the practical way to determine the continuation threshold which is flexible with the sampling rate and used Z rather than R and X. Instead of smoothness, the proposed algorithm employed regularity because smoothness can identify abrupt change but not the direction of abrupt change. The proposed power swing detection method in this paper used regularity which is completely different from the smoothness concept. The proposed algorithm used Z which is conventionally evidenced to be the ideal for obtaining directional abrupt changes. The paper Blumschein et al. (2014) used a record in the power swing area where the proposed detection used a record along with the rate of impedance path. Also, the shape of the boundary used in this work is round which covers the smaller area on the R-X plane. Once the power swing is detected, the out-of-step condition is determined by the proposed algorithm using Zone-2 entry and direction which is the most unique and safe method used in this work. No counter is required like Blumschein et al. (2014) in the proposed algorithm which excludes the need for a comprehensive system study.

5.7 Conclusion

81

Table 5.2 Results of power swing with different fault location, resistance, and duration Swing Fault duration Fault Swing Actual OOS location resistance detection (ohm) Transmission

Transformer

0.2 0.2 0.2 0.5 1 2 5 0.04 0.1 1 5 0.04 0.04 0.04

0.01 1 10 0.01 0.01 0.01 0.01 1 1 1 1 0.01 5 10

Yes Yes No Yes Yes Yes Yes Yes Yes Yes Yes Yes No No

No No No No No No Yes Yes No No No Yes No No

No No No No No No Yes Yes Yes No No Yes Yes Yes

5.6 Results and Discussion The completion of different power swings with the change in swing locations, fault resistance, and fault duration are listed in Table 5.2. The proposed relay’s OOS detection and actual OOS conditions are matched. The bold cases in Table 5.2 show maloperation of the proposed relay and all others are true operations. It has been noted that small-signal disturbance is more serious as compared to transient disturbance. Also, the fault resistance can alter the impedance trajectory and due to which the swing cannot be identified. The continuity, regularity, and smoothness are the most important parameter of swing which in most cases give accurate detection. Once the power swing is detected, it is very important to decide whether the swing is uncontrollable or controllable. The results of Table 5.2 show that the out-of-detection conditions give satisfying results. From Table 5.2, it is observed that the faults on the transmission line are merely outcomes of the out-of-step condition of the generator only in the case of 5s duration of fault with 0.01 in OOS of the generator.

5.7 Conclusion The investigation of the proposed power swing detection and out-of-step protection scheme reveals that the settings of the out-of-step relay necessitated revision with variations in fault resistance value. Due to the increased fault resistance value, the

82

5 Predictive Out of Step Protection

swing could not reach the outer layer to the inner layers though it has repetitiveness, continuation, and regularity. The two important points transcribed from the results of the given cases studies are as follows: (1) Small-signal disturbances often resulted in uncontrollable power swings than transient disturbances. (2) Fault duration much high than the critical clearing time of the fault point created unstable swings. So, fault clearing must be rapid. The proposed algorithm provides often a correct classification of power swings and out-of-step conditions. In some cases where the fault resistance is high then the proposed power swing detection algorithm fails to identify. The next chapter focus on the many objectives framed in Chap. 1 such as faster OOS scheme with half cycle response time, hardware compatible algorithm, setting free relaying scheme, and reduction in unwanted tripping of associate relaying.

References Alinezhad B, Karegar HK (2017) Out-of-step protection based on equal area criterion. IEEE Trans Power Syst 32(2):968–977 Ambekar V, Dambhare S (2012) Comparative evaluation of out of step detection schemes for distance relays, p 6 Blumschein J, Yilmaz Yelgin, Kereit M (2014) Blackout prevention by power swing detection and out-of-step protection. J Power Energy Eng 02:694–703. https://doi.org/10.4236/jpee.2014. 24093 Camarillo-Peñaranda JR, Celeita D, Gutierrez M, Toro M, Ramos G (2020) An approach for outof-step protection based on swing center voltage estimation and analytic geometry parameters. IEEE Trans Ind Appl 56(3):2402–2408. https://doi.org/10.1109/TIA.2020.2973590 Desai JP, Makwana VH (2017) Development of artificial intelligence based algorithm for power system planning and protection. In: 2017 International conference on telecommunication, power analysis and computing techniques (ICTPACT proceedings), Chennai, April 2017, pp 1–5. https:// bit.ly/3tSVZiG Desai JP, Makwana VH (2020) Phasor measurement unit incorporated adaptive out-of-step protection of synchronous generator. J Modern Power Systems Clean Energy Desai JP, Makwana VH (2021) Modeling and implementation of percentage bias differential relay with dual-slope characteristic. IEEE Texas Power Energy Conf (TPEC) 2021:1–6. https://doi. org/10.1109/TPEC51183.2021.9384987 Desai JP, Makwana VH, Phasor measurement unit incorporated adaptive out-of-step protection of synchronous generator. J Modern Power Systems Clean Energy. https://doi.org/10.35833/MPCE. 2020.000277 Desai JP, Makwana VH (2022) Modeling and implementation of power swing detection and outof-step protection. J Inst Eng India Ser B 103:541–548 Desai J, Makwana V (2020) Power swing blocking algorithm based on real and reactive power transient stability. Electric Power Components Syst 48(16–17):1673–1683. https://doi.org/10. 1080/15325008.2021.1906794 Haes Alhelou H, Hamedani-Golshan ME, Njenda TC, Siano P (2019) A survey on power system blackout and cascading events: research motivations and challenges. Energies 12(4): Art. no. 4 Khodaparast J, Khederzadeh M (2016) Adaptive concentric power swing blocker. Protection Control Modern Power Syst. 1(16):324–339. https://doi.org/10.1186/s41601-016-0026-9

References

83

Kosterev NV, Yanovsky VP, Kosterev DN (1996) Modeling of out-of-step conditions in power systems. IEEE Trans Power Syst 11(2):839–844 Krata J, Balcerek P, Gajic Z (2014) The new frequency difference based out of step protection for multiterminal transmission system. In: 12th IET international conference on developments in power system protection (DPSP 2014), Mar 2014, pp. 1–6. https://doi.org/10.1049/cp.2014. 0055 Liu G, Azizi S, Sun M, Popov M, Terzija V (2018) Performance of out-of-step tripping protection under renewable integration. J Eng 2018(15):1216–1222. https://doi.org/10.1049/joe.2018.0180 McDonald M, D Tziouvaras, Apostolov et al. (2005) Power swing and out-of-step considerations on transmission lines. IEEE PSRC, WG. https://bit.ly/33KdLdD Oza BA (2010) Power system protection, and switchgear. Tata McGraw-Hill Education Private Ltd Paithankar YG, Bhide SR (2011) Fundamentals of power system protection. PHI Learning Pvt. Ltd PSCAD (2022) PSCAD on-line help system. https://bit.ly/2Qn9SYR Yellajosula JRAK, Wei Y, Grebla M, Paudyal S, Mork BA (2020) Online detection of power swing using approximate stability boundaries. IEEE Trans Power Delivery 35(3):1220–1229. https:// doi.org/10.1109/TPWRD.2019.2941522 Zhang S, Zhang Y (2017) A novel out-of-step splitting protection based on the wide area information. IEEE Trans Smart Grid 8(1):41–51 Zhang S, Zhang Y (2019) Characteristic Analysis and Calculation of Frequencies of Voltages in Out-of-Step Oscillation Power System and a Frequency-Based Out-of-Step Protection. IEEE Trans Power Syst 34(1):205–214. https://doi.org/10.1109/TPWRS.2018.2866022

Chapter 6

Wavelet Transform and Deep Learning Machine Model-Based Out-of-Step Relay

Abstract The out-of-step protection of a synchronous generator or a group of synchronous generators is unreliable with significant renewable power penetration in the power system. This work presents an innovative out-of-step protection algorithm using wavelet transform and deep learning to protect synchronous generators and transmission lines. The specific patterns are generated from a stable power swing, an unstable power swing, and a three-phase fault using the wavelet transform technique. The data containing 27008 continuous samples of 48 different features trains a two-layer feed-forward network’s particular design. The proposed algorithm gives an automatic, setting free, and highly accurate classification against the three-phase fault, the stable power swing, and the unstable power swing through pattern recognition within half cycle. The solution given in Chaps. 3, 4, and this chapter has a minor setting procedure that can be fully eliminated using the ANN approach. The proposed algorithm uses the Kundur two-area system and the 29-bus electric network for testing under different swing center locations and renewable power penetrations. The hardware-in-loop (HIL) test shows a newly developed out-of-step algorithm’s hardware compatibility. The proposed algorithm is compared with a recently reported algorithm in the end. The comparison and test results on different large-scale systems clarify that the algorithm is simple, fast, accurate, and HIL tested and not affected by the changes in power system parameters. Keywords Artificial intelligance · Deep learning · Hardware relay · Wavelet transform ,

6.1 Introduction The power swing is a phenomenon that usually occurs due to sudden disconnection of the heavy loads and tripping of the transmission lines due to the faults in the system. The protective elements must accurately and quickly detect the power swing condition. The consequences of unstable power swings are mostly mal-operation of the transmission line’s distance relay, damage to the generator and turbine-generator unit. The results of unstable power swings cause the cascade failure of numerous trans© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 J. K. P. Desai and V. Makwana, Power Swing Detection and Generator Out-of-Step Protection Under Renewable Power Source Integration, Energy Systems in Electrical Engineering, https://doi.org/10.1007/978-981-19-9546-0_6

85

86

6 Wavelet Transform and Deep Learning Machine Model-Based Out-of-Step Relay

mission lines, transformers, and generators. Additional devices and functions called out-of-step protection systems are usually augmented to avoid the consequences of an unstable power swing. The essential parts of out-of-step protection relaying are the power swing blocking (PSB) function, out-of-step (OOS) tripping function of the transmission line, and OOS protection of the synchronous generator. It is difficult to detect a symmetrical fault during power swing (Brahma 2007). The low-frequency oscillation of the power system can result in a loss of stability or a blackout of the power system. The low-frequency oscillations in power swing have been detected in Avdakovic et al. (2012) by analyzing the Daubechies-4 (db4) wavelet. The method was compared with the Prony and the Eigenvalue analysis. The proposed discrete wavelet transforms (DWT)-based approaches in Avdakovic et al. (2012) can identify the onset of the initial disturbance in the power system and identify the modes present during low power frequency oscillation. The future scope of wavelets for blackout prevention by avoiding unnecessary tripping using wavelet-based protection has been described in Choudhary and Sharma (2015). The wavelet transform can decompose the signal in the time-frequency domain. Different wavelet families are taken and compared in the paper Choudhary and Sharma (2015), which shows that high accuracy can be achieved using the db4 mother wavelet. The author projected in Brahma (2007) that the transient energy could capture in level d1 to d4 of voltage waveform during the events. It has been analyzed that the detail co-efficient-9(d9) of the current waveform can track the variation in current. The proposed algorithm in Brahma (2007) uses detail co-efficient-1 (d1) to detail co-efficient-4 (d4) for the detection of the fault. It uses the d9 as an indicator of power swing. The paper used some thresholds for a final decision. However, the identification of unstable power swing using detail co-efficient lacks in Brahma (2007). Also, the calculation of such thresholds is not described in Brahma (2007). The algorithm in Brahma (2007) used some fixed settings, making it rigid and needing close attention. I have designed the optimized deep learning (DL) machine model, which works on pattern recognition, whereas Brahma (2007) uses the feature extraction and compares it with threshold settings. The difficulties of using only a few specific features and threshold settings are explained in Sect. 6.3 of this paper. The blinder-based out-of-step relay for the synchronous generator protection is typically designed based on fixed settings. The work in Choudhary and Sharma (2015) described the impact of integrating solar and wind power generation on the small-signal oscillations in the modern power system. The authors in Koley et al. (2017) proposed a protection scheme that differentiates the type of fault from load change events. The magnitude harmonic components of the voltage signal and its fundamental part provide enough information to discriminate the fault detection, location, classification, and zone identification. The artificial neural network (ANN) is used after the support vector machine (SVM) identifies the fault type. However, the two unique techniques make the model complicated and less efficient for hardware. The algorithm in Koley et al. (2017) cannot detect unstable power swings and is used only as a PSB relay. Furthermore, the algorithm in Koley et al. (2017) is not validated on an extensive scale system with significant renewable power penetration. The ANN offered more prompt responses and required a quarter of the fault signal cycle to identify the type of fault Martin and Aguado (2003). The

6.1 Introduction

87

ANN-based distance relay can provide fast and precise operation (Coury and Jorge 1998). Nevertheless, the paper Coury and Jorge (1998) does not address the other distance relay problems, such as mal-operation of distance relay during power swing, under renewable integration. Once the power swing is identified from the fault, its further identification is lacking in Brahma (2007); Avdakovic et al. (2012); Choudhary and Sharma (2015); Koley et al. (2017); Martin and Aguado (2003); Coury and Jorge (1998), which is the most important in the modern power system. The stable power swings need not require a rapid disconnection of affected elements, and the advanced power system controllers can dampen the stable swing. Advanced power system controllers cannot dampen out the unstable power swing, so it needs rapid and correct detection. In the paper Venkatesan and Balamurugan (2001), the authors suggested using voltage and current signals for feature extraction for the machine model and advised avoiding fast Fourier transform (FFT) signal to reduce hardware complexity. The ANN model in the paper Zhou et al. (1994) uses the multi-layered perception with the back-propagation algorithm. It describes the neural network structure, with 30 neurons in the input layer, 2 neurons in the hidden layer, and 1 in the output layer. The output, higher or equal to 0.8, is classified as 1. If it is less or equals 0.2, it is classified as 0. Result 1 indicates that the system is vulnerable, and 0 means not weak. The paper Zhou et al. (1994) shows that the use of ANN for online power system dynamic security and vulnerability assessment is quite realistic. ANN improves performance in terms of adaptiveness and relay coordination (Jongepier and van der Sluis 1997). The two significant areas of the wavelet transform are power system protection and power quality. The paper Zhang and Kezunovic (2007) chooses db5 as a mother wavelet for detecting the short duration and fast decaying fault-generated transient signal. Both high-frequency and low-frequency approximation can avoid confusion between fault and non-fault events (Zhang and Kezunovic 2007). The paper Zhang and Kezunovic (2007) shows some limitations, such as specific system structure; the algorithm needs more adjustment, which shows the method’s non-adaptiveness. Incorrect input data, the proposed method in Cardoso et al. (2004) produced dubious output and lacked information. In that case, user intervention is necessary. The process can be used for significant system fault section identification as it does not depend on the size of the electric network. Identifying a failure device helps restore the system quickly after the collapse. The author has used a multilayer perceptron neural network to solve the failed device’s identification problem in Negnevitsky and Pavlovsky (2005). However, the proposed network only handles 32 alarms, so the ANN needs to be developed to deal with complex system emergencies. The paper Far et al. (2012) uses multivariate analysis and data mining techniques for synchronous generator islanding protection. The phasor measurement unit (PMU)-based adaptive out-of-step protection algorithm is presented in Chap. 5. The use of PMU not only increased the accuracy and reliability of the out-of-step protection, but it also increased the overall cost of protection. The paper Khodaparast and Khederzadeh (2016) presented the adaptive concentric power swing blocker with two concentric circles. The paper Khodaparast and Khederzadeh (2016) shows that the current signal’s static phasor estimation error can be used to find the first pair of concentric circle locations. However, the impedance rate of change used in Khodaparast and

88

6 Wavelet Transform and Deep Learning Machine Model-Based Out-of-Step Relay

Khederzadeh (2016) depends on the modern power system parameters like a voltage regulator, governor, fault type, and renewable power penetration. Furthermore, the conventional power swing blocking (CPSB) used in Khodaparast and Khederzadeh (2016) may mal-operate under renewable integration. The signal is analyzed up to the 12th level using the db4 mother wavelet in the present work. I have designed a proposed algorithm that is general enough that with modification of appropriate mother wavelet, it can be used for the protective relaying with other conditions such as topological change, loading, and fault locations. The MATLAB (Matrix laboratory) environment is used for the testing and development of the proposed ANN-based algorithm with a central processing unit (CPU) having an Intel i5 processor of 8gigabyte (8 GB) of random-access memory (RAM) and a 64-bit operating system on 256-GB solid-state drive (SSD). Section 6.2 described the system and power swing conditions. Section 6.3 explains the uniqueness in the training data selections and pre-processing of the data. It also presents the deep learning machine model’s unique mathematical modeling and the final algorithm. The unknown disturbances are manually applied to test the proposed technique, described in Sect. 6.4. The development and large-scale validation results are presented in Sect. 6.5. A comparison of the proposed algorithm with the recently reported algorithm is explained and summarized in Sect. 6.6. The results are expressed into four categories: Results during training, testing, and validations. Results of testing with unknown disturbance. Hardware-in-loop (HIL) test results. Extensive scale validation.

6.2 The System and Power Swing Conditions The Kundur two-area system is considered for the wavelet analysis of the three-phase fault, stable and unstable power swing signals, as shown in Fig. 6.1. The unstable

Fig. 6.1 A Kundur two-area system for testing

6.3 Development of the Proposed Algorithm

89

Fig. 6.2 Three-phase current waveform during unstable power swing

Fig. 6.3 Three-phase voltage waveform during unstable power swing

power swing is produced by applying three-phase faults more than CCT (critical clearing time) near HV (high voltage) of GTU (generator transformer unit) of G1 (generator 1). The three-phase current and voltage waveforms produced at bus B1 during unstable power swing after fault removed are shown in Figs. 6.2 and 6.3, respectively. The process of developing the proposed relay has been explained in four steps. The steps are (1) training data selection and pre-processing; (2) design of the mathematical structure of deep learning neural network; (3) design of the proposed relaying algorithm; and (4) training, validation, and testing.

6.3 Development of the Proposed Algorithm 6.3.1 Training Data Selection and Pre-processing The fault events are transient, and the transients are reflected in voltage and current waveforms by the change in their frequency and magnitude. The wavelet transform of a signal gives information about both frequency and time of the transient event. It is challenging to detect three-phase faults during power swing as it provides a minimal reflection at the transient stage of the voltage signal. These changes depend on the

90

6 Wavelet Transform and Deep Learning Machine Model-Based Out-of-Step Relay

Table 6.1 Energy distributions in the current waveform Levels Frequency range Three-phase fault Stable swing energy in % energy in % Level 1 Level 2 Level 3 Level 4 Level 5 Level 6 Level 7 Level 8 Level 9 Level 10 Level 11 Level 12 Approx.12

5 kHz–10 kHz 2.41 kHz– 5.19 kHz 1.21 kHz– 2.58 kHz 603 Hz–1.29 kHz 302 Hz–646 Hz 151 Hz-323 Hz 75.4 Hz-162 Hz 37.9 Hz–80.7 Hz 19.2Hz–40.3Hz 9.98 Hz–19.8 Hz 4.44 Hz-10 Hz 4.4 Hz–4.61 hz 0 Hz–0.244 Hz

Unstable swing energy in %

0.00 0.00

0.00 0.00

0.00 0.00

0.00

0.00

0.00

0.01 0.02 0.26 14.95 67.80 3.15 2.80 3.75 4.14 3.11

0.00 0.00 0.17 15.57 83.69 0.21 0.35 0.00 0.00 0.00

0.00 0.00 0.23 19.05 80.45 0.09 0.17 0.00 0.00 0.00

location of the fault, type of fault, and instant of fault. The frequency of transient is much higher than the nominal frequency of the system. Further, the phase-to-ground voltage magnitude of the faulted phase is near zero during ground faults, and the current increases considerably. Hence, the wavelet scale covering the fault frequency has higher energy than the scale covering the current wave’s nominal frequency. Power swing has a relatively low frequency which ranges from 3 to 7 Hz. The energy distribution up to the 12th level during the three-phase fault, stable power swing, and unstable power swing is shown in Table 6.1 using the current waveform and in Table 6.2 using the voltage waveform. The wavelet scales must be selected such that it covers lower frequency, which detects patterns of the low energy levels, and high frequency, which see patterns of high energy levels. The db4 wavelet decomposes the current/voltage signals up to 12th levels with a sampling rate of 20 kHz, which gives enough resolution of time–frequency variation in current and voltages during events. During the three-phase faults, the minor energy lies in the current wave from 302 to 646 Hz, reflecting in detail co-efficient d5, which is absent in power swing for the same ranges. Further, the minor energy lies in voltage waveform in the range of 302–646 Hz for three-phase fault and absent for power swing in the same range. Analysis of details co-efficient at levels 3 to 5, level 9, and level 10 shows the unique patterns of an unstable power swing, stable power swing, and three-phase fault. Figure 6.4 shows the significant pattern differences between the three-phase fault and power swing in the level 5 current signal’s detailed resolution. Similarly, the time–frequency resolution of the voltage waveform at level 9 shows the significant pattern differences between stable and unstable swings, as

6.3 Development of the Proposed Algorithm

91

Table 6.2 Energy distributions in the voltage waveform Levels Frequency range Three-phase fault Stable swing energy in % energy in % Level 1 Level 2 Level 3 Level 4 Level 5 Level 6 Level 7 Level 8 Level 9 Level 10 Level 11 Level 12 Approx.12

5 kHz–10 kHz 2.41 kHz– 5.19 kHz 1.21kHz– 2.58 kHz 603Hz–1.29 kHz 302 Hz–646 Hz 151 Hz–323 Hz 75.4 Hz–162 Hz 37.9 Hz–80.7 Hz 19.2 Hz–40.3 Hz 9.98 Hz–19.8 Hz 4.44 Hz–10 Hz 4.4 Hz–4.61 hz 0 Hz–0.244 Hz

Unstable swing energy in %

0.01 0.01

0.00 0.00

0.00 0.00

0.02

0.00

0.00

0.05 0.11 0.51 18.31 78.42 1.66 0.56 0.15 0.10 0.10

0.00 0.00 0.17 15.58 83.69 0.21 0.35 0.00 0.00 0.00

0.00 0.00 0.24 19.72 79.85 0.04 0.13 0.00 0.00 0.00

Fig. 6.4 Pattern difference at level 5 in detail resolution of current during the events

shown in Fig. 6.5. The minor pattern difference at each level between fault, stable, and unstable swings also exists, as shown in Tables 6.1 and 6.2. The little pattern difference is also useful for training the machine model and can’t be ignored. Further analysis of all features verifies that the mean and median values of d7 and a12 co-efficient of the current signals can differentiate between three-phase fault and

92

6 Wavelet Transform and Deep Learning Machine Model-Based Out-of-Step Relay

Fig. 6.5 Pattern difference at level 9 in detail resolution of voltage during the events Table 6.3 d7 and a12 co-efficients during the unstable, stable, and faulty condition Current signal during Co-efficient Mean value Median Max. value Min. value Unstable swing Stable swing Three-phase fault Unstable swing Stable swing Three-phase fault

d7 d7 d7 a12 a12 a12

0.02 0.008 59.74 0.037 0.062 1.2×104

−0.01 −0.008 262 0.02 0.035 8888

6.346 2.287 4.431×104 0.1881 0.3181 5.08×104

−7.298 −2.311 4.2×104 −0.008 −0.007 392

power swing, as shown in Table 6.3. The stable and unstable swing can be classified using mean and median values a1 and d12 of the voltage waveform, as shown in Table 6.4. First, a few selected details and approximate co-efficients of the voltage and current waveforms were used to train the DL (deep learning) algorithm using Tables 6.3 and 6.4. However, it has been observed that a few selected co-efficient cannot classify all the events altogether. A few chosen feature extraction techniques have no significant difference between the stable and unstable power swings. After that, a few more co-efficient using Tables 6.1 and 6.2 are used and tried one more time with the feature extraction method. The observation shows the pattern’s improvement and some distinct differences in each pattern of fault and swing. Finally, in this work, the patterns of the detailed co-efficient from d1 to d12 and approximate co-efficient a1 to a12 were considered. The input vector gives the unique pattern for three-phase fault, stable power swing, and unstable power swing, differentiating them completely.

6.3 Development of the Proposed Algorithm

93

Table 6.4 d12 and a1 of voltage signal statistics during the unstable, stable, and faulty condition Voltage signal during Co-efficient Mean value Median Max. value Min. value Unstable swing Stable swing Three-phase fault Unstable swing Stable swing Three-phase fault

d12 d12 d12 a1 a1 a1

1.12E+04 −1594 −7577 2632 −25.4 −336.1

1.3×104 −773.8 −8803 5201 68.84 −452.2

1.9×104 629.9 −2061 2.2×105 1.8×105 1.5×105

−3108 −5065 −1.3×104 −2.3×105 −1.9×105 −8.8×104

6.3.2 Design of the Mathematical Structure of Deep Learning Neural Network The proposed deep learning machine model has a (48×1) input vector-matrix size, x1, x2, x3, and x4 to x48, as shown in Fig. 6.6. The optimization in performance and training time gives the neural network of two hidden layers, one input layer, one output layer, and each hidden layer has ten neurons. The optimization has been achieved by minimizing the cross-entropy by changing neurons at each layer. The input vector x(j) (at the jth sample) is weighted by respective weight and bias at hidden layers, as shown in Fig. 6.6. The best possible weight and bias are determined such that it minimizes the loss function. The proposed deep learning machine model uses a scale conjugate descent method for parameter estimation. The forwarding pass uses a linear combination with non-linear activation repetitively to each layer to get the prediction. Once it reaches a prediction, the next job is to find out the loss. The loss is propagated in the reverse direction to calculate the gradient concerning the direct connections to the output layer. Then it applies the chain rule of derivative successively to find out losses at intermediate levels. The proposed deep learning model used an adaptive learning rate compared to the classical machine learning algorithm. The adaptive learning rate gives convergence faster. The loss function used is the non-convex type, and hence the proposed model used a momentum-

Fig. 6.6 DL neural network for the proposed algorithm

94

6 Wavelet Transform and Deep Learning Machine Model-Based Out-of-Step Relay

based strategy. It applies techniques of early stopping. The machine model early stops when generalization stops improving during training. Each unit of neuron has two parts of activation as follows: (1) Linear combination. (2) Non-linear activation. The linear function at first hidden layer for an ith neuron and a jth sample is given by z i = bi +

i=n  

j



wi xi

(6.1)

i=1

where bi = bais at ith neuron, wi = weight at ith neuron, and x ji = input vector at ith neuron for jth sample. The non-linear activation function used at the first layer is a, bi = bais at ith neuron; wi = weight at ith neuron; x ji = input vector at ith neuron for jth sample. The non-linear activation function used at the first layer is a tan-sigmoid transfer function, which for an ith neuron and a jth sample is described as given below: ai = F(z) =

2 −l 1 + e−2zi

(6.2)

The below equation gives the linear function at the first hidden layer for an i th neuron, and a jth sample is n i = bi +

i=n 

wi ai

(6.3)

i=1

Layer 2 has a non-linear activation function as a SoftMax transfer function. The below equation describes a SoftMax transfer function for an ith neuron and jth sample. Finally, the output is given by for a jth sample, y ( j) = bi +

i=n 

wi σi

(6.4)

i=1

where y(j) = output at a jth sample. y(j) has three sets of binary outputs which are [1 0 0], [0 1 0], and [0 0 1] for unstable swing, stable swing, and fault, respectively. The highest probability in the output set [y1, y2, y3] is considered by the nearest value 1, and it is considered equal to 1, and all other values consider 0.

6.3 Development of the Proposed Algorithm

95

6.3.3 The Design of the Proposed Relaying Algorithm Equations 6.5 and 6.6 calculate the detail and approximate co-efficient. The input signals of voltage and current are given by the current and voltage transformer, respectively.  {h(2k − n)C Al−1 (n)} (6.5) C Al (k) = n

C Dl (k) =



{g(2k − n)C Dl−1 (n)}

(6.6)

n

where C Al and C Dl represent the approximation co-efficients and detail co-efficients of the signal at level l. The signals pass through an HPF (high-pass filter) and an LPF (low-pass filter). Outputs from both filters are then decimated by 2 to obtain the detail co-efficients and the approximation co-efficients at level 1 (a1 and d1). The approximation co-efficients are then sent to the second stage to repeat the procedure (Avdakovic et al. 2012). Finally, the signal decomposes at the 12th level. Once the signal has decomposed, then it starts to create the input vector. The training data have prepared using input vector and output vector which is given by T rainingdata = (x( j), y( j))

(6.7)

where x (j) is a set of wavelet co-efficients (d1 (j) to d12 (j), a1 (j) to a12 (j)) at a jth sample for voltage and current waveform at a sampling rate of 20kHz or above and y (j)  (0,1). Figure 6.7 shows the flowchart of the proposed algorithm. The previously trained pattern recognition machine model senses the input x (j). The machine model

Fig. 6.7 A flowchart of the proposed algorithm

96

6 Wavelet Transform and Deep Learning Machine Model-Based Out-of-Step Relay

Table 6.5 Performance result of the developed machine model on the Kundur two-area system and the 29-bus system System Process Samples % error Kundur two-area system

The 29-bus system

Training

18906

0

Validation Testing Training Validation Testing

4051 4051 12603 2700 2700

0 0 1.40% 1.30% 1.30%

gives three binary outputs [1 0 0], [0 1 0], and [0 0 1] for unstable swing, stable swing, and three-phase fault, respectively. If an unstable swing is classified, then the trip signal is sent to the associate breaker at the point of separation. The algorithm produced a PSB command sent to the transmission line distance relay in a stable power swing case. Once the PSB command is sent, the algorithm samples the next data and continues until any further identification. If the fault is classified, then the tripping decision of distance relay is allowed with its zone delay settings for transmission line protection (Desai 2022).

6.3.4 Training, Validation, and Testing Table 6.5 shows the performance in terms of % error during training, validation, and testing. The 0% error indicates that no sample is mis-classified. The cross-entropy needs to minimize during training, validation, and tastings. The development stops when the training, validation, and testing plot intersect at minimum cross-entropy. After 614 epochs, the best validation is achieved, which gives a cross-entropy of 0.00013, as shown in Fig. 6.8. The confusion matrix present in Fig. 6.9 shows the performance in terms of the output class matrix to the target class. The output and target classes completely match during training, validation, and testing in the confusion matrix. If the algorithm is confused, the value is shown in the off-diagonal place; otherwise, it is placed at a diagonal location in the confusion matrix.

6.3 Development of the Proposed Algorithm

Fig. 6.8 Cross-entropy reduction after each epoch

Fig. 6.9 Confusion matrix of the developed machine model

97

98

6 Wavelet Transform and Deep Learning Machine Model-Based Out-of-Step Relay

6.4 Performance Validation by Unknown Events The signal data were used for the swing center at the HV terminal of GTU of G1 during the machine model development. After developing the proposed algorithm with the required performance, many unknown signal data of stable swing, unstable swing, and a three-phase fault were considered. The modified Kundur two-area system was used to find the effect of different renewable penetration levels on the proposed outof-step relay. Four identical type-4 DFIGs (doubly-fed induction generator) have connected at buses BG1, BG2, BG3, and BG4 such that total power flow remains the same from area 1 to area 2 in Fig. 6.1.

6.4.1 Test Using Unknown Data of Different Fault Location The following test cases produced strange signals for testing: 1. Swing center arises at HV terminal of GTU of G2. 2. Swing center appears at HV terminal of GTU of G3. 3. Renewable power penetration at a different level. 4. Swing center appears at the middle of the transmission line. An unknown input vector of features from the whole bunch of extensive sampled signal data applies to the algorithm. The output of the algorithm is rounded to its closest prediction using the following condition: (1) if the binary output is fractional value < 0.9, then consider 0. (2) if the binary output is fractional value ≥ 0.9, then consider 1. If the same binary result continues for a half cycle, it indicates its final prediction. It is worth noting for the reader that the same algorithm can be expanded for fault using Simulink Real-Time (MATLAB 2020). The Simulink relay consists of an input vector (48×1) applied to the DL machine model, which generates the binary output. The hardware relay only consists of the DL machine model inside coded using Simulink Real-Time. The data transmitter and receiver have been connected at the PD8 and PD9 pins of the hardware relay. The pin PD8 is the transmitter pin and the PD9 pin is the receiver pin of the hardware relay. Sampling time used for the data transmission is 0.005 s. Figure 6.12 shows the output comparison between Simulink-based developed relay and hardware-based designed relay. The tripping command of the hardware relay has been delayed with 0.005 s. The delay in the trip command of hardware is due to the sampling time set in data transmission. If the data transmission time is subtracted from the hardware relay response time, then the software and hardware relays both have the same instantaneous response once the event has been confirmed after continuous same binary output for half cycle detected. Figure 6.10 shows the hardware setup used for real-time HIL simulation.

6.5 Development and Performance on a 29-Bus System

99

Fig. 6.10 HIL test setup to test the proposed algorithm

6.5 Development and Performance on a 29-Bus System Figure 6.11 shows the 29-bus, 7-power plant networks based on the Canadian HydroQuebec (QUE) system. It has transmission lines of 735 kV and 26200 MVA power capacity (Initializing a 29-Bus 2022). The 735 kV transmission line is a series and shunt compensated with fixed capacitors (Cs) and inductors (Ls). The network contains seven hydrotype power plants with its regulator and power system stabilizers (Initializing a 29-Bus 2022). The sampling rate of 50 ms is considered in the simulation. The three-phase fault is applied at the MTL (Montreal) bus before and after the

Fig. 6.11 29-bus, 7-power plant network system (Hydro-Quebec)

100

6 Wavelet Transform and Deep Learning Machine Model-Based Out-of-Step Relay

Fig. 6.12 Tripping signals of the hardware relay during HIL test

Fig. 6.13 Cross-entropy reduction with each epoch for 29-bus system implementation

critical clearing time to generate a set of training data. We developed the DL model on a 29-bus large-scale system to verify the accuracy of development steps. The continuous 18003 samples of three different patterns having the vector dimension of 48×1 were used for training, testing, and validation (see Fig. 6.13). The newly developed wavelet and deep-learning-based machine model is tested with unknown power swings and three-phase fault signals. The results show that the proposed algorithm development procedure is a standard one and able to use on any scale of the system. Also, the detection of power swings and three-phase fault signals is highly accurate with an unknown input vector. It is found that the overall accuracy is reduced by

6.6 Comparison of the Proposed Algorithm

101

Fig. 6.14 Confusion matrix of the developed machine model on the 29-bus system

1.4% during the training, testing, and validation processes on the 29-bus system. The details of errors in each stage are shown in Table 5. The movement, testing, and validation results on 29 bus are shown using the confusion matrix in Fig. 6.14. The cross-entropy found on the 29-bus system is 0.0127, which shows that the deep learning model is well fitted.

6.6 Comparison of the Proposed Algorithm Comparison of the proposed algorithm with the wavelet-based algorithm of paper Brahma (2007) and SVM- and ANN-based algorithm of paper Koley et al. (2017) is summarized in Table 6.7. Both the PSB and OOS trippings are most important for a synchronous generator. The proposed algorithm can identify the type of power swing once power swing is detected, which is not possible in the method reported in Brahma (2007) and Koley

102

6 Wavelet Transform and Deep Learning Machine Model-Based Out-of-Step Relay

Table 6.6 Cases simulated to test relay logic for unknown data Case no. Type of event Power swing produced due to: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Unstable power swing Stable power swing Three-phase faults Stable power swing Unstable power swing Unstable power swing Stable power swing Unstable power swing Stable power swing Three-phase faults Stable power swing Unstable power swing Unstable power swing Stable power swing Unstable power swing Stable power swing Three-phase faults Unstable power swing Stable power swing Three-phase faults

LLLG fault (Tfault > Tcr) LLLG fault (Tfault < Tcr) Not applicable Sudden load changes Sudden load changes LG fault (Tfault > Tcr) LG fault (Tfault < Tcr) LLLG fault (Tfault > Tcr) LLLG fault (Tfault < Tcr) Not applicable Sudden load changes Sudden load changes LG fault (Tfault > Tcr) LG fault (Tfault < Tcr) 50% level of integration 7% level of integration Not applicable LLLG fault (Tfault > Tcr) LLLG fault (Tfault < Tcr) Not applicable

The output of the proposed relay [1 0 0] [0 1 0] [0 0 1] [0 1 0] [1 0 0] [1 0 0] [0 1 0] [1 0 0] [0 1 0] [0 0 1] [0 1 0] [1 0 0] [1 0 0] [0 1 0] [1 0 0] [0 1 0] [0 0 1] [1 0 0] [0 1 0] [0 0 1]

et al. (2017). Hence, the recent way reported in Brahma (2007) and Avdakovic et al. (2012) is suitable for power swing blocking (PSB) and not for OOS tripping. Further, the out-of-step relay must adapt to the change in system topology, power flow, and renewable power penetration level. The proposed scheme is validated with different levels of renewable power penetration. In contrast, the method in Brahma (2007) and Koley et al. (2017) is tested on a minimal system (less than nine buses) without the integration of renewable power resources. The proposed algorithm does not require the threshold calculation as compared to the method in Brahma (2007). The way in Koley et al. (2017) needs harmonic detection of 3rd, 5th, and 7th harmonics, and after that, it used a fault classifier for detection of the type of fault. Under renewable integration, such methods have more chances of mal-operation. The method’s operating time in Koley et al. (2017) and the proposed method is the same. Relay operating time is not fixed for the technique (Brahma 2007) as it depends on the fault location. The proposed DL model development steps are standard ones irrespective of the scale of the power system. In the paper, Brahma (2007) and Koley et al. (2017), the author hasn’t tested the algorithm on a larger scale for development standardization.

6.8 Conclusion

103

Similarly, the out-of-step relay performance in series- and shunt-compensated line is required to be evaluated, which is lacking in Brahma (2007) and Koley et al. (2017). In contrast, the proposed method works correctly under this condition verified on the 29-bus system. The proposed algorithm independently works where the technique in Brahma (2007) uses the conventional scheme, and Koley et al. (2017) uses fault classifier. The proposed algorithm works on patterns, so the protection engineer does not need to study the system topology, system parameters, source nature, and fault location. The recent methods in Brahma (2007) and Koley et al. (2017) are not that much easy. In Brahma (2007), threshold settings fluctuate with fault location, and in Koley et al. (2017), the Kalman filter design changes under increased harmonic injection by non-linear loads or unknown renewable sources. HIL test results of the proposed algorithm have been provided, which shows the hardware suitability and actual operation speed where the method in Brahma (2007), and Koley et al. (2017) was not tested for HIL.

6.7 Discussions The proposed pattern recognition machine model using wavelet transform gives favorable results during development. This work decided to test it using the events not part of the algorithm’s development. Table 6.6 shows all the different unknown event cases, including stable power swing, unstable power swing, three-phase faults, and sudden load change conditions. The proposed algorithm correctly identifies each class of power swing and three-phase fault for all the unexplained events. The proposed algorithm needs half a cycle or less to decide the type of power swing. The training time of the algorithm using the high-end processor is almost a fraction of seconds. In this work, the total time required for training, validation, and testing is around 2 s. The proposed algorithm’s sampling time can handle a higher rate depending on the processor which is deployed. When the proposed algorithm is applied in a hardware relay, it gives detection and tripping within 0.01 s with high accuracy. Hardware-in-the-loop test confirms that the developed relaying algorithm is ready for the hardware production stage and provides the same response as tested in Simulinkbased model. The proposed algorithm’s accuracy on the 29-bus system is reduced slightly during training, testing, and validation.

6.8 Conclusion The details and approximate co-efficients captured using the db4 wavelet up to the 12th level of resolution during unstable power swing, stable power swing, and threephase fault provide a unique pattern about each event with an input vector of d1 to d12 and a1 to a12 of current and voltage in the given order. The deep learning

104

6 Wavelet Transform and Deep Learning Machine Model-Based Out-of-Step Relay

Table 6.7 The comparison of the proposed algorithm with existing methods Point of comparison The proposed The wavelet based SVM and ANN based algorithm (Brahma 2007) (Koley et al. 2017) PSB Out-of-step (OOS) tripping Three-phase faults during power swing Use of threshold settings Adaptive quality Accuracy Simplicity of usage Relay operating time HIL tested Standardization Tested for compensated transmission line Tested under significant renewable integration Depended on another scheme

Yes Yes

Yes No

Yes No

Yes

Yes

Yes

No

Yes

No

Yes Excellent Yes Half cycle Yes Standard Yes

No Very good p5.93em—Needs careful settings Not fixed No Not verified No

No Very good Required filter design and fault classifier Half cycle No Not verified No

Yes

No

No

No

Yes

Yes

machine model is designed to recognize the pattern and then discriminate stable swing, unstable swing, and three-phase fault event automatically with high accuracy. The proposed algorithm is not affected by strange circumstances as the pattern’s nature remains the same, not when the signals’ features were used. The proposed algorithm is based on the wavelet transform with the DL machine model, which can detect any uncommon power swing produced due to the impact of a renewable power penetration. The strange power swings can’t classify correctly every time using the blinder-based and recently reported algorithms. The HIL test confirms that the proposed relay gives the same response in the hardware system, which it provides during testing with the simulation model. The proposed out-of-step relaying algorithm has a minimal training time and rapid response time. The proposed algorithm’s accuracy is 98.6% on the 29-bus hydro-Quebec system, which is reduced by 1.4% than the kundur two-area system due to the 29-bus hydro-Quebec system’s complexity. The comparison of the proposed algorithm with the recently reported algorithms concludes that the proposed out-of-step relay is very accurate, simple, fast, HIL tested, and adaptive in modern power system conditions.

References

105

References Avdakovic S, Nuhanovic A, Kusljugic M (2012) Wavelet transform applications in power system dynamics. Lancet 83. https://doi.org/10.1016/j.epsr.2010.11.031 Brahma SM (2007) Distance relay with out-of-step blocking function using wavelet transform. IEEE Trans Power Delivery 22(3):1360–1366. https://doi.org/10.1109/TPWRD.2006.886773 Cardoso G, Rolim JG, Zurn HH (2004) Application of neural-network modules to electric power system fault section estimation. IEEE Trans Power Delivery 19(3):1034–1041. https://doi.org/ 10.1109/TPWRD.2004.829911 Choudhary S, Sharma FB (2015) Small-signal stability analysis of renewable source connected power system and identification of oscillatory modes using wavelet transform. In: 2015 international conference on smart grid and clean energy technologies (ICSGCE), Offenburg, pp 23–29. https://doi.org/10.1109/ICSGCE.2015.7454264 Coury DV, Jorge DC (1998) Artificial neural network approach to distance protection of transmission lines. IEEE Trans Power Delivery 13(1):102–108. https://doi.org/10.1109/61.660861 Far HG, Rodolakis AJ, Joos G (2012) Synchronous distributed generation islanding protection using intelligent relays. IEEE Trans Smart Grid 3(4):1695–1703. https://doi.org/10.1109/TSG.2012. 2208659 Initializing a 29-Bus, 7-Power plant network with the load flow tool of Powergui - MATLAB and Simulink. https://www.mathworks.com/help/physmod/sps/ug/initializing-a-29-bus-7power-plant-network-with-the-load-flow-tool-of-powergui.html Jongepier G, van der Sluis L (1997) Adaptive distance protection of double-circuit lines using artificial neural networks,”. IEEE Trans Power Delivery 12(1):97–105. https://doi.org/10.1109/ 61.568229 Desai JP (2022) Protective relay for out of step protection in power systems, India. 202221018194 A, 20/2022. https://search.ipindia.gov.in/IPOJournal/Journal/ViewJournal Khodaparast J, Khederzadeh M (2016) Adaptive concentric power swing blocker. Prot Control Mod Power Syst 1(1):16. https://doi.org/10.1186/s41601-016-0026-9 Koley E, Shukla SK, Ghosh S, Mohanta DK (2017) Protection scheme for power transmission lines based on SVM and ANN considering the presence of non-linear loads. IET Gener Transm Distrib 11(9):2333–2341. https://doi.org/10.1049/iet-gtd.2016.1802 Martin F, Aguado JA (2003) Wavelet-based ANN approach for transmission line protection. IEEE Trans Power Delivery 18(4):1572–1574. https://doi.org/10.1109/TPWRD.2003.817523 MATLAB 9.8 (R2020a-trail) (2020) The MathWorks Inc., Natick, Massachusetts Negnevitsky, Pavlovsky V (2005) Neural networks approach to online identification of multiple failures of protection systems. IEEE Trans Power Delivery 20(2):588–594. https://doi.org/10. 1109/TPWRD.2004.843451 Venkatesan R, Balamurugan B (2001) A real-time hardware fault detector using an artificial neural network for distance protection. IEEE Trans Power Delivery 16(1):75–82. https://doi.org/10. 1109/61.905596 Zhang N, Kezunovic M (2007) Transmission line boundary protection using wavelet transform and neural network. IEEE Trans Power Delivery 22(2):859–869. https://doi.org/10.1109/TPWRD. 2007.893596 Zhou Q, Davidson J, Fouad AA (1994) Application of artificial neural networks in power system security and vulnerability assessment. IEEE Trans Power Syst 9(1):525–532. https://doi.org/10. 1109/59.317570

Chapter 7

Out-of-Step Protection Schemes Summary and Future Scope

Abstract The book used power swing detection and generator out-of-step protection to avert blackouts. To determine the precise problem statement and research gap on the book topic, a literature survey utilizing research papers, observations using the real-time operation of the generator, and simulated case studies were conducted. According to the survey, the power system’s power swing characteristic is difficult. Due to the intricacy of the swing under the impact of renewable power penetration, control actions by AVR and AFC, and PSS action, distinguishing the stable power swing from three-phase fault, transient fault, and unstable power swing might be difficult. Keywords Computational · Memory · Pattern recognition · Sampling rate

7.1 Out-of-Step Protection Schemes Summary This research points the following effects of renewable power penetration and controllers: • The CCT of the system at the given location for the three-phase fault keeps decreasing as the penetration of wind power increases. • DB scheme mal-operates under the impact of renewable power source because the increase in penetration more than 14.1% makes the power swing complex and unstable found by implantation on Kundur two-area system. • RPES influence the impedance trajectory and alter the reactance and resistance reach due to the limited current amplitude, current frequency offset, and controlled current phase angle. • The penetration of PV and doubly-fed induction generator (DFIG) improves the system’s damping, but up to a certain level. If the penetration of solar PV is more than 50%, it negatively affects the damping of the system. The following points are concluded from the all methods of out-of-step protection described in this book: © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 J. K. P. Desai and V. Makwana, Power Swing Detection and Generator Out-of-Step Protection Under Renewable Power Source Integration, Energy Systems in Electrical Engineering, https://doi.org/10.1007/978-981-19-9546-0_7

107

108

7 Out-of-Step Protection Schemes Summary and Future Scope

• The combination of direct measurement using PMU and indirect measurement using graphical characteristics makes the OOS relay quite costly. • The predictive technique is found prime important to eliminate the few threshold settings required by the direct and indirect measurement technique. • The investigation of the proposed predictive out-of-step relay reveals that the settings of the out-of-step relay necessitated revision with variations in fault resistance value. Due to increased fault resistance value, the swing could not reach the outer layer to the inner layers through repetitiveness, continuation, and regularity. • The fast setting free OOS relay based on wavelet and deep learning neural network is developed and implemented in this research. It shows the most accurate results on a Kundur two-area system under renewable power system integration, action AVR, action AFC, and action of PSS. The accuracy is reduced by 1.4% when implemented on 29-bus Hydro-Quebec system. • Hardware-in-loop test of the wavelet and DL-based relay gives the same results as achieved during software implementation. • HIL test performed on the OOS relay is very cost-effective and reliable, which gives the initial start point of prototype development at low cost. • Amid all different approaches of power swing detection and OOS protection to prevent and reduce the blackout, the wavelet- and deep-learning-based fast setting free OOS relay is most accurate and suitable as the solution for the problem identified in this research. Table 7.1 shows that the PMU-based scheme has high measuring accuracy due to the use of PMU. However, the PMU-based method may detect pole slipping after 180◦ , which makes detection quite slow. The CPU, RAM, and EEPROM requirements are moderate. The most advantageous feature of the PMU-based scheme is that it allows for local decision-making. As PMU involves the sampling of phasor, it needs to be done at 48 samples/s and computational time of the order of 0.4 s. The predictive scheme is again quite inaccurate due to fault nature, location, and resistance changes in the impedance trajectory. This scheme can predict the power swing before 180◦ . The main advantage of this scheme is that the system operator has enough time to prevent outages or blackouts. The predictive scheme, however, needs 0.25 ms of computational time, which requires a high sampling rate of 80 samples/s. Finally, the pattern-recognition-based scheme gives the highest accuracy among all three, with high measuring accuracy. The main concern of the scheme is very high computational time, on the order of 5 µs. However, modern FPGA and highend processors can fully meet the high computational demand at affordable prices. The pattern-recognition-based scheme needs temporary storage to train the data for every changed topology.

7.2 Future Scope

109

Table 7.1 Technical comparison of all out-of-step schemes discussed in the book OOS relaying Detection Detection and Sampling, Other parameters algorithm accuracy and response time storage, and proposed measuring computational accuracy burden PMU based

Moderate as it depends on step 4 after 180◦ . So sometimes it can issue the tripping after one pole slip to give accurate detection in worst case. Measuring accuracy: High Predictive OOS Detection accuracy: medium measuring accuracy: medium Pattern Detection recognition based accuracy: highest measuring accuracy: High

0.106 s for swing at Generator (after fault) & 1 cycle or less 0.32 s for swing at Transformer (after fault) & 1 cycle or less

Conditional logic sampling of phasor: 48 sample/cycle Less burden computational time: 0.4 ms

CPU: medium RAM: order of few MB ROM: order of few MB communication: Local EEPROM: Yes

Earlier (before swing center arise) 1 cycle or less

Mathematical computing sampling: 80 sample/cycle Computational time: 0.25 ms Mathematical computing high-end controller & FPGA sampling: 400 sample/cycle high burden computational time: 5 us

Storage: less EEPROM: Yes

Half cycle or sometimes less than that

Storage: Moderate level (temp. storage) CPU: medium RAM: 500 MB or more ROM: order of few MB EEPROM: No

7.2 Future Scope Still, there are few areas where one can work on the research topic. Following points show the future scope of work: • The predictive OOS scheme can be improved by considering the angle, voltage, frequency, and current measurements, which may eliminate the malfunctioning of the relay. • PSB and predictive OOS scheme can be further verified on New England IEEE 39 Systems. • The wavelet and deep-learning-based work can be further developed using real grid disturbance data of various large-scale outages.

Appendix A

Kundur Two-Area System

The system consists of two identical areas connected with very sensitive tie line. Each area having two generator units having a rating of 900 MVA and 20 KV. The generator parameters in p.u. on their rated MVA and KV base are as follows: X d = 1.8; X q = 1.7; X l = 0.2; X d = 0.3; X q = 0.55; X d = 0.25; X q = 0.25; Ra = 0.0025; Td 0 = 8.0 s; Tq 0 = 0.4s’; Td 0 = 0.03 s; Tq 0 = 0.05 s; As at = 0.015; Bs at = 9.6; Yt 1 = 0.9; H = 6.5 (for G 1 and G 2 ); H = 6.175 (G 3 and G 4 ); K D = 0. For each step-up transformer: Z T = 0 + j0.15 pu on 900 MVA and 20/230 KV base and off-nominal ratio = 1.0; For Transmission line: r = 0.0001 pu/km; x L = 0.001 pu/km; bC = 0.00175 pu/km. The system is operating with area 1 exporting 400 W to area 2 and generating units are loaded as follows: G 1 : P1 = 700 MW; Q1 = 185 MVAr; E t1 = 1.03 (20.2◦ ), G 2 : P2 = 700 MW; Q2 = 235 MVAr; E t2 = 1.01 (10.5◦ ), G 3 : P3 = 719 MW; Q3 = 176 MVAr; E t3 = 1.03 (−6.8◦ ), G 4 : P4 = 700 MW; Q4 = 202 MVAr; E t4 = 1.01 (−17.0◦ ). The loads and reactive power by shunt capacitors, Bus 7: PL = 967 MW; Q L = 100 MVAr; Q c = 200 MVAr and Bus 9: PL = 1767 MW; Q L = 100 MVAr; Q c = 350 MVAr.

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 J. K. P. Desai and V. Makwana, Power Swing Detection and Generator Out-of-Step Protection Under Renewable Power Source Integration, Energy Systems in Electrical Engineering, https://doi.org/10.1007/978-981-19-9546-0

111

Appendix B

Type-IV Wind Turbine System Parameter for Modified Kundur Two-Area System

EMT-type detail model for specific type-IV wind turbine topology is used to integrate into the Kundur two-area system. It is mainly based on a directly coupled (gear-less) externally excited synchronous generator connected to three-stage full converter system. The converter system contains a six-pulse diode rectifier, DC-DC boost converter, and an IGBT-based two-level voltage source converter (VSC). The data of the wind turbine characteristic data is shown in Table B.1, wind turbine data is given in Table B.2, control parameter for a wind turbine is given in Table B.3, the converter data is presented in Table B.4, and rated value of wind turbine is given in Table B.6 (see Table B.5). In modified Kundur two-area system, Fig. 4.11, a total of 400 MW power has been injected with the above model at the same time maintaining power flow from area 1 to area 2 approximately as same as original under two-area system. 100 MW was injected at each bus of all four Bg1, Bg2, Bg3, and Bg4, making an overall 400 MW injection into the system. Table B.1 Wind turbine characteristic c p versus λ Data C1

C2

C3

C4

C5

C6

C7

Cs

C9

0.2065

200

0.15

0.05

2.14

13.2

13

−0.02

−0.003

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 J. K. P. Desai and V. Makwana, Power Swing Detection and Generator Out-of-Step Protection Under Renewable Power Source Integration, Energy Systems in Electrical Engineering, https://doi.org/10.1007/978-981-19-9546-0

113

114

Appendix B: Type-IV Wind Turbine System Parameter for Modified …

Table B.2 Wind turbine rated parameters Parameter Name Rr otor lambdaopt V speednom V speedin Urated Prated Q rated i max f buck f boot f chopper V dcr e f

Wind turbine rotor radius Optimal tip-speed ratio Nominal wind speed Cut-in wind speed Phase-to-phase rated voltage Rated active power Rated reactive power Maximal acceptable converter current PWM carrier-freq. of buck converter PWM carrier-freq. of boost converter PWM carrier-freq. of braking chopper Vdc2 reference voltage value

Table B.3 Wind turbine control parameter Parameter Name T au or e f V dc1kp Vdcl ki f cut zeta K p pll K i pll T au V dc2 V dc2kp V dc2ki T au Qr e f T au Qmeas Q kp Q ki Iband

Time constant of filter Vdc1 proportional gain Vdcl integrator gain Cut-off freq. of measurement filter Damping factor of measurement filter PLL proportional gain PLL integrator gain Time constant of Vdc2 meas. filter Vdc2 controller proportional gain Vdc2 controller integrator gain Time constant of Qref-filter Time constant of Qmeas-filter Q-control proportional gain Q-control integrator gain Current control tolerance band

Value

Unit

42 7.1 11 3.5 690 2 1.5 3180 2000 2000 2000 1250

m – m/s m/s V MW MVAr A Hz Hz Hz V

Value

Unit

10 0.75 0.075 10 0.7071 0.1775 21.3 33.334 10 75 33.334 33.334 0.001 0.035 0.017

S – S −1 kHz – – S −1 ms – S −1 ms ms – S −1 pu

Appendix B: Type-IV Wind Turbine System Parameter for Modified … Table B.4 Wind turbine converter parameter Parameter Name Rs Ron C1 Rboost Lboost Rchopper C2 L1 filter Rd filter Ld filter L2 f ilter

Diode and IGBT’s snubber resistance Diode and IGBT’s conducting resistance Vdcl capacitance Boost reactor resistance Boost reactor inductance Chopper resistance Vdc2 capacitance L1 output filter inductance Damping circuit output filter resistance Damping circuit output filter inductance L2 output filter inductance

Table B.5 Load data Parameters Nominal active power (MW) Nominal reactive power (MVar)

Value

Unit

0.1 0.1 50 10 100 0.15 300 150 0.025 15 5

M M mF  mH  mF uH  mH uH

Load5

Load6

Load8

125 50

90 30

100 35

Table B.6 Wind turbine rated values Parameter Name Rr otor lambdaopt V speednom V speedin U rated Prated Q rated i max f buck f boot f chopper V dcr e f

115

Wind turbine rotor radius Optimal tip-speed ratio Nominal wind speed Cut-in wind speed Phase-to-phase rated voltage Rated active power Rated reactive power Maximal acceptable converter current PWM carrier-freq. of buck converter PWM carrier-freq. of boost converter PWM carrier-freq. of braking chopper Vdc2 reference voltage value

Value

Unit

42 7.1 11 3.5 690 2 1.5 3180 2000 2000 2000 1250

m – m/s m/s V MW MVAr A Hz Hz Hz V

Appendix C

Indian Grid System’s Data

See Tables C.1 and C.2. Table C.1 Line parameter of Bina Gwalior line Parameters WR-NR Voltage (kV) Reactance (10−3 /m) Resistance (10−4 /m) Distance (Km)

400 0.19623 0.109 350

WR-ER

NR- ER

400 0.2381 0.13 420

400 0.2381 0.13 420

Table C.2 Pre-disturbance condition of NR, ER, and WR S/N Region Generation Demand Import(+)/ (MW) (MW) Export(−) (MW) 1

Northern

32636

38,322

(+) 5686

2 3 4

Eastern 12452 Western 33024 Northeastern 1367

12,213 28053 1314

(−) 239 (−) 6229 (−) 53

Remarks (MW) Import From Bhutan 1127 – –

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 J. K. P. Desai and V. Makwana, Power Swing Detection and Generator Out-of-Step Protection Under Renewable Power Source Integration, Energy Systems in Electrical Engineering, https://doi.org/10.1007/978-981-19-9546-0

117

Appendix D

A 29-Bus Hydro-Quebec System

The model shows a 735 kV transmission network with detailed modeling of seven 13.8 kV power plants (total available generation = 26200 MVA), including hydraulic turbines, speed regulation, excitation systems, and power system stabilizers. The 735 kV transmission network is both series and shunt compensated using fixed capacitors and inductors. The load is lumped at two buses (MTL7 and QUE7). The MTL load subsystem connected to the MTL7 bus consists of four load blocks connected on the 25 kV distribution system through 735/230 kV and 230/25 kV transformers. The QUE load and wind generation subsystem uses a 6000 MW load (constant Z and constant PQ) connected on the 120 kV bus. Using an asynchronous generator, a 9 MW wind farm is connected to the 120 kV bus through a 25 kV feeder and a 25 kV/120 kV transformer.

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 J. K. P. Desai and V. Makwana, Power Swing Detection and Generator Out-of-Step Protection Under Renewable Power Source Integration, Energy Systems in Electrical Engineering, https://doi.org/10.1007/978-981-19-9546-0

119