Futuristic Trends in Numerical Relaying for Transmission Line Protections 9789811584640, 9789811584657

This book presents the state-of-the-art approach for transmission line protection schemes for smart power grid. It provi

189 43 7MB

English Pages 146 [163] Year 2020

Report DMCA / Copyright

DOWNLOAD PDF FILE

Table of contents :
Preface
Acknowledgements
Contents
About the Authors
Abbreviations
List of Figures
List of Tables
1 Transmission Line Protection Philosophy
1.1 Introduction
1.1.1 Transmission Network Protection Schemes
1.1.2 Necessity of Advanced Power System Protection
1.2 Traditional Transmission Line Protection
1.2.1 Time-Graded Protection
1.2.2 Differential Protection
1.2.3 Distance Protection
1.2.4 Carrier-Aided Protection
1.3 Literature Review
1.4 Review on Phasor Estimation Techniques
1.5 Numerical Distance Protection for Fault Context Identification
1.5.1 Review on Microprocessor-Based Protection Schemes
1.5.2 Review on Neural Network-Based Protection Schemes
1.5.3 Review on Traveling-Wave-Based Protection Schemes
1.5.4 Review on Wavelet Transform-Based Protection Schemes
1.6 Power Swing Detection Methods
1.6.1 Review on Power Swing Detection for Uncompensated Transmission Lines
1.6.2 Review on Power Swing Detection for SCTL
1.7 Auto-Reclosure Technology Review
2 Transmission Line Protection: Issues and Research Needs
2.1 Issues in Numerical Distance Relays
2.1.1 Effect of DC Component
2.1.2 Fault Inception Angle and Power Flow Angle
2.1.3 Close-In Fault
2.1.4 Influence of Fault Resistance
2.1.5 Load Encroachment and Evaluation of Zone 3 Relay Settings
2.1.6 Transient Condition and Implementation of Auto-Reclosure
2.1.7 Effect of Power Swing
2.1.8 Series Compensation in Transmission Line
2.2 Objectives of Research
2.3 Research Plan
3 Adaptive Numerical Distance Relaying Scheme
3.1 Introduction
3.2 Phasor Estimation Techniques
3.2.1 Discrete Fourier Transform
3.2.2 Modified Full-Cycle Discrete Fourier Transform
3.3 System Modeling for Proposed Relaying Scheme
3.4 Proposed Methodology for Transmission Line Protection
3.4.1 Phasor Estimation Using MFCDFT
3.4.2 Impedance Reach Determination
3.4.3 Relay Settings for Protection Zones
3.4.4 Adaptive Slope Tracking Method
3.5 Validation of Proposed Technique
3.5.1 Results of Phasor Estimation
3.5.2 Performance Evaluation of the Proposed Algorithm
3.5.3 Fault Classification
3.5.4 Fault Location Estimation
3.5.5 Fault Cases with CT Saturation
3.6 Advantages of Proposed Algorithm
3.7 Outcome of Proposed Technique
4 Discrimination Between Power Swing and Line Fault Based on Voltage and Reactive Power Sensitivity
4.1 Introduction
4.2 System Modeling
4.3 Power Swing Detection: Problems and Remedies
4.3.1 Problem Description
4.3.2 Proposed Method for Power Swing Detection
4.4 Simulation Results and Discussions
4.4.1 Power Swing Cases Due to Electrical Load Switching
4.4.2 Power Swing Cases Due to Mechanical Disturbances
4.4.3 Fault Simulation on Protected Line (L2)
4.4.4 Fault Cases During Power Swing
4.4.5 Power Swing Cases Due to Post-Fault Isolation on Line L1
4.4.6 Relay Backup for the Fault on Parallel Transmission Line
4.5 Research Outcome of Proposed Technique
5 Sequence-Space-Aided Disturbance Classifier Scheme Based on Support Vector Machine
5.1 Introduction
5.2 System Modeling
5.3 Sequence-Space-Based SVM Classifier Scheme
5.4 Tenfold Cross-Validation
5.5 Data Mining Using SVM Classifier for Power Swing Cases
5.5.1 Power Swing Due to Load Switching
5.5.2 Power Swing Due to Mechanical Disturbances
5.5.3 Power Swing Due to Fault Isolation on Adjacent Line
5.5.4 Power Swing Due to Adjacent Line Switching
5.6 Data Mining Using SVM Classifier for Fault Cases
5.6.1 Solid Faults on Transmission Line
5.6.2 Faults During Power Swing
5.7 Result Discussions
5.7.1 Power Swing Detection
5.7.2 Validation for Fault Cases
5.7.3 Validation for Fault During Power Swing
5.8 Comparative Analysis
5.9 Effective Outcome
6 Auto-Reclosing Scheme with Adaptive Dead Time Control Based on Synchro-Check Principle
6.1 Introduction
6.2 System Modeling
6.3 Proposed Fault Detection and Auto-Reclosing Technique
6.4 Hardware Implementation
6.5 Validation of Proposed Algorithm Using Simulation
6.5.1 Validation for Transient Faults
6.5.2 Validation for Permanent Fault
6.6 Emulation of Proposed Algorithm
6.7 Research Outcome of Proposed Auto-Reclosing Scheme
Appendix: System Parameters During Modeling in PSCAD
7 Summary of Proposed Work
7.1 General
7.2 Summary of Research Work
7.3 Scope of Future Research
References
Recommend Papers

Futuristic Trends in Numerical Relaying for Transmission Line Protections
 9789811584640, 9789811584657

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

Energy Systems in Electrical Engineering

Ujjaval Patel Praghnesh Bhatt Nilesh Chothani

Futuristic Trends in Numerical Relaying for Transmission Line Protections

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

More information about this series at http://www.springer.com/series/13509

Ujjaval Patel Praghnesh Bhatt Nilesh Chothani •



Futuristic Trends in Numerical Relaying for Transmission Line Protections

123

Ujjaval Patel Department of Electrical Engineering Adani Institute of Infrastructure Engineering Ahmedabad, Gujarat, India

Praghnesh Bhatt Department of Electrical Engineering Pandit Deendayal Petroleum University Gandhinagar, Gujarat, India

Nilesh Chothani Department of Electrical Engineering Adani Institute of Infrastructure Engineering Ahmedabad, Gujarat, India

ISSN 2199-8582 ISSN 2199-8590 (electronic) Energy Systems in Electrical Engineering ISBN 978-981-15-8464-0 ISBN 978-981-15-8465-7 (eBook) https://doi.org/10.1007/978-981-15-8465-7 © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd. The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721, Singapore

Preface

This book has been organized into seven chapters as following: Chapter 1 outlines the introduction and traditional protection philosophy adopted to shelter the power system under consideration along with state-of-the-art reviews on the existing methods. The literature review starts with the technological developments in the field of phasor estimation of analog input signal applied to numerical relays to initiate the relaying actions. It also covers the reviews on numerical distance protection schemes along with separate review on widely used methods based on traveling wave, artificial neural network, and wavelet transform techniques. This chapter also covers exhaustive literature survey on power swing detection methods for uncompensated and series compensated transmission lines. At the end, the literature survey is also provided on auto-reclosure technology used to discriminate between transient fault and permanent faults. Chapter 2 presents critical issues that influence the performance of numerical distance relays along with appropriate mathematical fundamentals. The main research objectives are narrated in order to clarify the scope of research in the area of smart grid protections. This chapter also outlines the plan of work carried out during the research. Chapter 3 demonstrates the development of adaptive quadrilateral numerical relaying distance scheme for fault impedance compensation. It proposes an innovative solution over conventionally used distance relaying schemes. The proposed scheme adaptively modifies the characteristic of numerical relay adaptively depending on the magnitude of fault impedance. It also identifies the fault type and estimates fault location along with fault instance which are the key attributes of multi-functional numerical relays. Chapter 4 outlines a simple and effective method for discrimination between power swing and fault for uncompensated transmission line. The developed scheme is validated for various kinds of fault and power swing conditions, including symmetrical fault occurring during the power swing.

v

vi

Preface

Chapter 5 reveals development of sequence-space-aided disturbance classifier technique for series compensated transmission line. It signifies usage of support vector machine (SVM) to detect and classify the fault occurring during the power swing. It proves to be very advance and effective discrimination method in comparison with schemes developed by many researchers. Chapter 6 shows emulation of synchro-phasor-aided auto-reclosing scheme to discriminate between transient fault and permanent fault occurring in the transmission line. An extensive simulation and hardware emulation have been carried for detection of transient fault with adaptive dead time control. It proves to be an effective auto-reclosing technology for both uncompensated and compensated EHV and UHV transmission lines. The developed scheme indicates remarkable improvement in the stability and reliability of power grid network. Chapter 7 concludes the summary of this book for the protection of transmission line along with suggested modifications for smart protective schemes. The scope of future research has been also explored for the technocrats working in this field. Finally, literatures used during the research are outlined in references as supporting fundamentals behind this research. Ahmedabad, India Gandhinagar, India Ahmedabad, India

Ujjaval Patel Praghnesh Bhatt Nilesh Chothani

Acknowledgements

As we complete this book, we acknowledge the blessings of Almighty for enabling us to complete our research successfully. We are very much thankful to the management of Adani Institute of Infrastructure Engineering and Pandit Deendayal Petroleum University for providing great support during our research and providing the essence of research in terms of time, support, and freedom. We acknowledge the constant motivation provided to us during our crucial time. Obviously, this would not be possible without active cooperation of technical and nontechnical staff members of Electrical Engineering Department of both institutions. We obliged them for their direct and indirect support in carrying out various tasks associated with our research work. We are grateful to the following journals for permission to reprint essays: Chap. 3 was published as ‘Adaptive Quadrilateral Distance Relaying Scheme for Fault Impedance Compensation’, International Journal of Electrical Control and Communication Engineering, 14(1), (2018): 58–70; Chap. 4 was published as ‘Distance Relaying with Power Swing Detection based on Voltage and Reactive Power Sensitivity’, International Journal of Emerging Electric Power Systems, 17 (1), (2016): 27–38; Chap. 5 was published as ‘A Novel Sequence-Space Aided SVM Classifier for Disturbance Detection in Series Compensated Transmission Line’, IET Science Measurement & Technology, 12(8), (2018): 983– 993; Chap. 6 was publicized in ‘Novel Auto-Reclosing Scheme with Adaptive Dead Time Control for EHV Transmission Line’, IET Science Measurement & Technology, 12(8), (2018): 1001–1008 for uncompensated transmission line. It was further extended for compensated transmission line and was published as ‘Emulation of Auto-Reclosing Scheme with Adaptive Dead Time Control for Protection of Series Compensated Transmission Line’, Electric Power Component & Systems, 47(1), (2019): 77–89. This acknowledgement will be inconclusive without the mention of our dear family members for their support and understanding despite the hardships they

vii

viii

Acknowledgements

endured. Besides this, we would like to extend our regards to all dear friends and well-wishers for their constant greetings and support. Special thanks to the Springer Nature publication and associated press for the care they have given during the preparation and production of this book. Ujjaval Patel Praghnesh Bhatt Nilesh Chothani

Contents

. . . . . . . . . . . . . . .

1 1 2 3 4 4 5 6 8 9 9 10 11 11 12

.. ..

13 14

.. .. ..

15 17 18

. . . . . .

. . . . . .

21 21 21 22 23 23

....

24

1 Transmission Line Protection Philosophy . . . . . . . . . . . . . . . . . . . 1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1.1 Transmission Network Protection Schemes . . . . . . . . . . 1.1.2 Necessity of Advanced Power System Protection . . . . . . 1.2 Traditional Transmission Line Protection . . . . . . . . . . . . . . . . . 1.2.1 Time-Graded Protection . . . . . . . . . . . . . . . . . . . . . . . . 1.2.2 Differential Protection . . . . . . . . . . . . . . . . . . . . . . . . . 1.2.3 Distance Protection . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2.4 Carrier-Aided Protection . . . . . . . . . . . . . . . . . . . . . . . 1.3 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.4 Review on Phasor Estimation Techniques . . . . . . . . . . . . . . . . 1.5 Numerical Distance Protection for Fault Context Identification . 1.5.1 Review on Microprocessor-Based Protection Schemes . . 1.5.2 Review on Neural Network-Based Protection Schemes . 1.5.3 Review on Traveling-Wave-Based Protection Schemes . 1.5.4 Review on Wavelet Transform-Based Protection Schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.6 Power Swing Detection Methods . . . . . . . . . . . . . . . . . . . . . . . 1.6.1 Review on Power Swing Detection for Uncompensated Transmission Lines . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.6.2 Review on Power Swing Detection for SCTL . . . . . . . . 1.7 Auto-Reclosure Technology Review . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . .

2 Transmission Line Protection: Issues and Research Needs . . . . 2.1 Issues in Numerical Distance Relays . . . . . . . . . . . . . . . . . . 2.1.1 Effect of DC Component . . . . . . . . . . . . . . . . . . . . . 2.1.2 Fault Inception Angle and Power Flow Angle . . . . . . 2.1.3 Close-In Fault . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1.4 Influence of Fault Resistance . . . . . . . . . . . . . . . . . . 2.1.5 Load Encroachment and Evaluation of Zone 3 Relay Settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . .

. . . . . .

ix

x

Contents

2.1.6 Transient Condition and Implementation of Auto-Reclosure . . . . . . . . . . . . . . . . . . 2.1.7 Effect of Power Swing . . . . . . . . . . . . . . . 2.1.8 Series Compensation in Transmission Line 2.2 Objectives of Research . . . . . . . . . . . . . . . . . . . . 2.3 Research Plan . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

26 26 28 30 32

3 Adaptive Numerical Distance Relaying Scheme . . . . . . . . . . . 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Phasor Estimation Techniques . . . . . . . . . . . . . . . . . . . . . . 3.2.1 Discrete Fourier Transform . . . . . . . . . . . . . . . . . . . 3.2.2 Modified Full-Cycle Discrete Fourier Transform . . . 3.3 System Modeling for Proposed Relaying Scheme . . . . . . . . 3.4 Proposed Methodology for Transmission Line Protection . . 3.4.1 Phasor Estimation Using MFCDFT . . . . . . . . . . . . . 3.4.2 Impedance Reach Determination . . . . . . . . . . . . . . . 3.4.3 Relay Settings for Protection Zones . . . . . . . . . . . . 3.4.4 Adaptive Slope Tracking Method . . . . . . . . . . . . . . 3.5 Validation of Proposed Technique . . . . . . . . . . . . . . . . . . . 3.5.1 Results of Phasor Estimation . . . . . . . . . . . . . . . . . 3.5.2 Performance Evaluation of the Proposed Algorithm . 3.5.3 Fault Classification . . . . . . . . . . . . . . . . . . . . . . . . 3.5.4 Fault Location Estimation . . . . . . . . . . . . . . . . . . . 3.5.5 Fault Cases with CT Saturation . . . . . . . . . . . . . . . 3.6 Advantages of Proposed Algorithm . . . . . . . . . . . . . . . . . . 3.7 Outcome of Proposed Technique . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . .

35 35 36 36 38 42 43 43 45 46 49 53 53 53 59 60 60 63 64

. . . . . . . . . . .

. . . . . . . . . . .

65 65 67 68 68 70 72 74 75 75 76

..

79

.. ..

79 81

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

4 Discrimination Between Power Swing and Line Fault Based on Voltage and Reactive Power Sensitivity . . . . . . . . . . . . . . . . . . 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 System Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3 Power Swing Detection: Problems and Remedies . . . . . . . . . . . 4.3.1 Problem Description . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.2 Proposed Method for Power Swing Detection . . . . . . . . 4.4 Simulation Results and Discussions . . . . . . . . . . . . . . . . . . . . . 4.4.1 Power Swing Cases Due to Electrical Load Switching . . 4.4.2 Power Swing Cases Due to Mechanical Disturbances . . 4.4.3 Fault Simulation on Protected Line (L2) . . . . . . . . . . . . 4.4.4 Fault Cases During Power Swing . . . . . . . . . . . . . . . . . 4.4.5 Power Swing Cases Due to Post-Fault Isolation on Line L1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.6 Relay Backup for the Fault on Parallel Transmission Line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.5 Research Outcome of Proposed Technique . . . . . . . . . . . . . . . .

Contents

xi

5 Sequence-Space-Aided Disturbance Classifier Scheme Based on Support Vector Machine . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 System Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3 Sequence-Space-Based SVM Classifier Scheme . . . . . . . . . . . 5.4 Tenfold Cross-Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.5 Data Mining Using SVM Classifier for Power Swing Cases . . 5.5.1 Power Swing Due to Load Switching . . . . . . . . . . . . . 5.5.2 Power Swing Due to Mechanical Disturbances . . . . . . 5.5.3 Power Swing Due to Fault Isolation on Adjacent Line . 5.5.4 Power Swing Due to Adjacent Line Switching . . . . . . 5.6 Data Mining Using SVM Classifier for Fault Cases . . . . . . . . 5.6.1 Solid Faults on Transmission Line . . . . . . . . . . . . . . . 5.6.2 Faults During Power Swing . . . . . . . . . . . . . . . . . . . . 5.7 Result Discussions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.7.1 Power Swing Detection . . . . . . . . . . . . . . . . . . . . . . . 5.7.2 Validation for Fault Cases . . . . . . . . . . . . . . . . . . . . . 5.7.3 Validation for Fault During Power Swing . . . . . . . . . . 5.8 Comparative Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.9 Effective Outcome . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . .

. 83 . 83 . 84 . 86 . 91 . 92 . 92 . 93 . 93 . 94 . 95 . 95 . 95 . 97 . 97 . 99 . 101 . 102 . 104

6 Auto-Reclosing Scheme with Adaptive Dead Time Control Based on Synchro-Check Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2 System Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3 Proposed Fault Detection and Auto-Reclosing Technique . . . . . 6.4 Hardware Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.5 Validation of Proposed Algorithm Using Simulation . . . . . . . . . 6.5.1 Validation for Transient Faults . . . . . . . . . . . . . . . . . . . 6.5.2 Validation for Permanent Fault . . . . . . . . . . . . . . . . . . . 6.6 Emulation of Proposed Algorithm . . . . . . . . . . . . . . . . . . . . . . 6.7 Research Outcome of Proposed Auto-Reclosing Scheme . . . . . . Appendix: System Parameters During Modeling in PSCAD . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

107 107 109 110 116 120 120 121 127 130 131

7 Summary of Proposed Work . . 7.1 General . . . . . . . . . . . . . . . 7.2 Summary of Research Work 7.3 Scope of Future Research . .

. . . .

. . . .

133 133 134 136

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139

About the Authors

Dr. Ujjaval Patel is working as Assistant Professor in the Department of Electrical Engineering at Adani Institute of Infrastructure Engineering since January 2019. He completed his B.E. in Electrical Engineering from L D College of Engineering, Ahmedabad, and M.E. in Electrical Engineering from Maharaja Sayajirao University, Vadodara, with specialization in Microprocessor Systems and applications. He received his PhD from Charusat University, Anand. He has 16 years’ experience in academia and 3 years’ experience in industry. He is Life member of Indian Society of Technical Education (ISTE), Member of Institution of Engineers (India) and Institute For Engineering Research Publications (IFERP). Dr. Ujjaval Patel has presented several research papers in IEEE sponsored international conferences. He has also published many research papers in many peer reviewed international journals. He is also working as reviewer of several peer reviewed national and international journals like IEEE transaction on power delivery, Electric Power Components & Systems, IET Science Measurement & Technology and many more. Currently, he is also guiding many PG and PhD research scholars under his supervision. In pace with ongoing developing technologies, his research interest includes digital protections of power systems along with applications of Internet of Things (IoT) for Home and Industrial automations for growth of society. Dr. Praghnesh Bhatt is presently working as Associate Professor and Head in Department of Electrical Engineering, School of Technology of Pandit Deendayal Petroleum University (PDPU), Gandhinagar. He completed his B.E. in Electrical Engineering from L D College of Engineering, Ahmedabad, and M.E. in Electrical Engineering from BVM Engineering College, Vallabh Vidhyanagar, with specialization in Electrical Power Systems. He received his PhD from S V National Institute of Technology (SVNIT), Surat. He has about 15 years of teaching experience at UG and PG level. He successfully guided 4 PhD research scholars and other 5 are pursuing their research under his supervision. He has published more than 50 research papers in globally reputed international journals and conferences. He is a member of IEEE and IEEE Power & Energy Society. He is also a life member of Indian Society of Technical Education (ISTE). He delivered many expert talks at xiii

xiv

About the Authors

international/national conferences/workshops/faculty training programs and also acted as session chair in national/international conferences. He has been awarded research funding projects from Royal Academy of Engineering (UK), AICTE and GUJCOST. He travelled to UK, Denmark and Hong Kong for academic and research works. He is a reviewer of international journals such as IEEE Transactions on Power System, IET GTD, IJEPES, Applied Energy, Electrical Power Components & Systems and Canadian Journal of Electric and Computer Engineering and more than 15 international conferences. Dr. Nilesh Chothani is an Associate Professor in the Department of Electrical Engineering at Adani Institute of Infrastructure Engineering, Ahmedabad, Gujarat. He received B.E. degree from Saurashtra University, Rajkot, Gujarat, in 2001. He received Master degree in power system and the PhD degree in electric engineering from the Sardar Patel University, Vallabh Vidyanagar Gujarat, India, in 2004 and 2013, respectively. He has more than two decades of teaching experience. Dr Chothani has published several papers in reputed international journals and conferences. Three of his papers are awarded with work of excellence in IEEE conference. He is a life member of IE(India) and ISTE. His areas of interest include digital protection, power system modeling & simulation, and artificial intelligence techniques. He has developed state of the art power system protection laboratory including real time operation of digital/numerical relaying scheme. He also received a research grant funded by Department of Science and Technology, SERB, New Delhi, Government of India.

Abbreviations

ADC AG ANFIS ANN CB CERC CT CVT DA DFT DIF DIT DSC DSO DWT EHV EPRI ETPTL FACT FFT FIA FL FST FT GA GDF GNN GT HIF HVDC

Analog-to-digital converter Attribute grammar Artificial neuro-fuzzy inference systems Artificial neural network Circuit breaker Central Electricity Regulatory Commission Current transformer Capacitive voltage transformer Directional angle Discrete Fourier transform Decimation in frequency Decimation in time Digital signal controller Digital storage oscilloscope Discrete wavelet transform Extra high voltage Electrical Power Research Institute Equal transfer process of transmission lines Flexible AC transmission Fast Fourier transform Fault inception angle Fault location Fast S-transform Fault type Genetic algorithm Ground distance function Generalized neural network Generator transformer High impedance faults High-voltage direct current

xv

xvi

IMSUM IRPCs ISTS IT LDA LSE MEB MFCDFT MIPS MM MMG NFV NRA OOS/OST PFA PMU PSB PSCAD PSD PSRC PT RBF REB RFFD RMS SCC SCTL SCV SEB SES SERC SPS SVM TBP TCSC UPFC USB WAMS WT

Abbreviations

Integrated moving sum Incremental reactive power coefficients Interstate transmission system Instrument transformer Linear discriminant analysis Least square error Middle-end bus Modified full-cycle discrete Fourier transform Million instructions per second Mathematical morphology Mathematical morphological gradient Normalized feature value Negative restraint angle Out of step Power flow angle Phasor measurement unit Power swing blocking Power system computer-aided design Power swing detection Power System Relaying Committee Potential transformer Radial basis function Receiving-end bus Robust fault detection and discrimination Root mean square Signal conditioning circuit Series compensated transmission line Swing center voltage Sending-end bus State Electricity Supply State Electricity Regulatory Commissions Special Protection Scheme Support vector machine Transient-based protection Thyristor-controlled series capacitor Unified power flow controller Universal Serial Bus Wide area measurement systems Wavelet transform

List of Figures

Fig. 1.1 Fig. 1.2 Fig. Fig. Fig. Fig. Fig. Fig.

1.3 1.4 1.5 1.6 2.1 2.2

Fig. 2.3 Fig. 2.4 Fig. 2.5 Fig. Fig. Fig. Fig. Fig.

2.6 2.7 3.1 3.2 3.3

Fig. 3.4

Fig. 3.5

Fig. 3.6

Time-graded—ring main system . . . . . . . . . . . . . . . . . . . . . . . Merz-price voltage balance differential protection scheme [5] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Basic principle of operation of distance relays . . . . . . . . . . . . Impedance calculation in distance relays . . . . . . . . . . . . . . . . . Operating zones of distance protection scheme . . . . . . . . . . . . Three zone distance protection scheme . . . . . . . . . . . . . . . . . . High resistance fault in transmission line . . . . . . . . . . . . . . . . Comparison between mho and quadrilateral settings for load encroachment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Fault and swing impedance trajectory . . . . . . . . . . . . . . . . . . . Series compensation in transmission line . . . . . . . . . . . . . . . . Influence of dynamic situations on conventional distance relays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Effect of fault impedance . . . . . . . . . . . . . . . . . . . . . . . . . . . . Proposed scheme for numerical distance protection . . . . . . . . Single line diagram of power system . . . . . . . . . . . . . . . . . . . Proposed methodology for transmission line protection . . . . . a Stepped distance quadrilateral characteristic for all zones, b control circuit diagram. . . . . . . . . . . . . . . . . . . . . . . . . . . . . Impedance trajectory ZT during variable fault resistance RF for different power flow conditions between SEB and MEB such as a no power transfer; b exporting power flow; c importing power flow; and d adaptive characteristic during variable power flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Phasor estimation for fault current using DFT and MFCDFT during L-G fault applied at 0.1 s (400 samples) with FIA = 0° . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . a Bus voltages, b line currents and estimated value of fault current, c fault signal, and d trip signal during L-G fault applied at 50 km at 0.1 s with RF = 0.01 Ω and FIA = 0° . . .

..

4

. . . . . .

. . . . . .

5 6 6 7 8 24

.. .. ..

25 27 29

. . . . .

. . . . .

30 31 32 42 44

..

47

..

49

..

53

..

56 xvii

xviii

Fig. 3.7

Fig. 3.8 Fig. 3.9 Fig. 3.10 Fig. 3.11

Fig. 4.1 Fig. 4.2 Fig. 4.3 Fig. 4.4 Fig. 4.5 Fig. 4.6 Fig. 4.7 Fig. 4.8 Fig. 4.9 Fig. 4.10 Fig. 4.11 Fig. 5.1 Fig. 5.2 Fig. 5.3 Fig. 5.4 Fig. 5.5 Fig. 5.6 Fig. 5.7 Fig. 5.8 Fig. 5.9

List of Figures

a Bus voltages, b line currents and estimated value of fault current, c fault signal, and d trip signal during L-G fault applied at 50 km at 0.1 s with RF = 20 Ω and FIA = 0° . . . . Fault impedance trajectory during a Case 1, b Case 3, c Case 17, d Case 18, e Case 26, f Case 27 of Table 3.1 . . . . Output of fault classifier module for AB-G fault . . . . . . . . . . . Percentage of fault location error versus fault distance for L-G fault with FIA = 0° . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Waveforms of a bus voltages, b line currents, c fault signal, d trip signal during close-in fault at 5 km with FIA = 0° with CT secondary burden of 15 Ω . . . . . . . . . . . . . . . . . . . . Single-line diagram of power system network . . . . . . . . . . . . Impedance locus of line L2 relay at SEB during 25% load increase at REB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Distance relay operation during heavy load fall on system . . . Proposed algorithm to discriminate power swing and fault condition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Real and reactive power flow in transmission line during change in d . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Variations in voltage and reactive power during 25% overloading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Power swing due to change in mechanical power input to the turbine of generator G1. . . . . . . . . . . . . . . . . . . . . . . . . Relay response during fault on line to be protected (L2) . . . . Performance of proposed method for fault discrimination during power swing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . System condition during fault isolation on adjacent line . . . . . Performance of proposed method for fault on parallel transmission line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Single line diagram of power system . . . . . . . . . . . . . . . . . . . Proposed Methodology of SVM-ased Classifier Scheme . . . . . Feature vectors during faulty conditions for training and testing of SVM. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Feature vectors during power swing conditions for training and testing of SVM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mapping of support vectors in the sequence-space . . . . . . . . . Validation for 50% load switching at MEB with 8% compensation and PFA = 12° at 0.5 s . . . . . . . . . . . . . . . . . . Validation for 50% change in mechanical input to generator at 0.5 s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Validation for fault isolation of adjacent line . . . . . . . . . . . . . Waveform during switching of line L2 . . . . . . . . . . . . . . . . . .

..

57

.. ..

58 61

..

62

.. ..

63 67

.. ..

69 70

..

72

..

73

..

74

.. ..

76 77

.. ..

78 80

.. .. ..

81 84 86

..

89

.. ..

90 90

..

98

.. 99 . . 100 . . 101

List of Figures

Fig. 5.10

Fig. Fig. Fig. Fig.

5.11 6.1 6.2 6.3

Fig. 6.4 Fig. 6.5 Fig. 6.6 Fig. 6.7 Fig. 6.8 Fig. 6.9

Fig. 6.10

Fig. 6.11

Fig. 6.12

Fig. 6.13

Validation for LL-G fault applied at 60 km with RF = 8Ω, FIA = 0°, PFA = 12° with 30% compensation level applied at 0.5 s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Validation for fault during power swing scenario . . . . . . . . . . System modeling in PSCAD . . . . . . . . . . . . . . . . . . . . . . . . . . Proposed technique for auto-reclosing. . . . . . . . . . . . . . . . . . . Waveforms during transient L-G fault a differential voltage across contacts of circuit breaker Vx(k), b phasor estimation of Q(k) and A(k), c frequency estimation F(k), and d Reclosing signal R(k) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Voltage magnitude and angle difference check characteristic for high-speed auto-reclosing supervision . . . . . . . . . . . . . . . . Waveforms of line side CVT voltage during L-G transient fault applied at 0.2 s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Block diagram for emulation of the proposed algorithm . . . . . Hardware setup of the proposed scheme in laboratory environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Proposed control circuit for auto-reclosing scheme . . . . . . . . . Waveforms during transient fault a line current, b phasor estimation of differential voltage (magnitude and angle) Q(k) and A(k), c frequency estimation F(k), and d reclosing signal R(k) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Waveforms during permanent LL-G fault a line current, b phasor estimated differential voltage magnitude and phase angle, c frequency response, and d response of relay and reclosure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Waveforms during transient fault with high value of fault resistance a line current, b phasor estimation of differential voltage (magnitude and angle) Q(k) and A(k), c frequency estimation F(k), and d reclosing signal R(k) . . . . . . . . . . . . . . Waveform during emulation of transient fault a voltage (Vx) across circuit breaker, b line current during L-G fault, and c generated fault signal waveform . . . . . . . . . . . . . . . . . . . . . . . Waveform during emulation of permanent fault a voltage (Vx) across circuit breaker, b line current during L-G fault, and c relay response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

xix

. . . .

. . . .

102 103 109 110

. . 112 . . 113 . . 114 . . 117 . . 118 . . 119

. . 121

. . 122

. . 128

. . 129

. . 129

List of Tables

Table 3.1 Table 3.2 Table 3.3 Table 4.1 Table 5.1 Table 5.2 Table Table Table Table Table Table Table

5.3 5.4 5.5 5.6 5.7 5.8 5.9

Table 6.1 Table Table Table Table

6.2 6.3 6.4 7.1

Comparison of the proposed technique with existing schemes for variation in fault context . . . . . . . . . . . . . . . . . . . . . . . . . Output of fault classifier module for each type of fault for various simulated cases . . . . . . . . . . . . . . . . . . . . . . . . . . Fault location estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . Comparison of proposed and existing scheme for symmetrical and asymmetrical faults during power swing situation . . . . . . Accuracy variation during SVM parameter settings . . . . . . . . Power Swing cases due to switching of electrical load connected at MEB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Power swing cases due to variation in mechanical input . . . . Power swing cases due to fault isolation on adjacent line . . . Power swing cases due to adjacent line switching . . . . . . . . . Fault cases for fault on transmission line L1 . . . . . . . . . . . . . Evaluation of fault during power swing on line L1 . . . . . . . . Overall accuracy for fault and swing conditions . . . . . . . . . . Performance comparison of proposed algorithm with existing schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Performance of the proposed auto-reclosing algorithm during for uncompensated transmission line . . . . . . . . . . . . . . . . . . . Performance for 10% compensation level . . . . . . . . . . . . . . . Performance for 30% compensation level . . . . . . . . . . . . . . . Performance for 60% compensation level . . . . . . . . . . . . . . . Comparative analysis of existing and proposed method . . . . .

..

54

.. ..

60 62

.. ..

79 91

. . . . . . .

. 92 . 93 . 94 . 95 . 96 . 97 . 104

. . 105 . . . . .

. . . . .

123 125 126 127 137

xxi

Chapter 1

Transmission Line Protection Philosophy

1.1 Introduction According to Ministry of Power, India, energy sector is growing at rapid pace. During the financial year 2017–18, the peak demand of India was 164.1 GW with the installed capacity of 330.8 GW with generation mix of thermal (66.2%), hydro (13.6%), renewable (18.2%), and nuclear (2.0%). Powergrid Corporation of India Limited governs planning and distribution of power through interstate transmission system (ISTS). State Electricity Boards (SEBs) are mainly responsible for the successful operation of ISTS. Smart power grids have been developed for evacuating power produced by numerous generating plants and distributing the power to the end users. The transmission lines are constructed with different voltage levels by taking into consideration of quantum of power and the distance involved. The extra high voltage (EHV) lines in trend are ±800 kV high-voltage direct current (HVDC) along with 66, 110, 220, 400, and 765 kV AC lines. The length and capacity of transmission system of 220 kV and above voltage levels, in the country as on November 30, 2017, were 3,81,671 ckm of transmission lines and 7,91,570 MVA, respectively. The transmission grids are in operation following regulations directed by Central Electricity Regulatory Commission (CERC) and State Electricity Regulatory Commissions (SERC). The loading on transmission grids must be controlled by considering voltage magnitude and angle stability, power flow conditions, and security aspects. With this huge power transmission infrastructure, the secure uninterrupted power transmission and distribution network must have smart protective devices to avert unwanted faults and widespread blackouts. The USA–Canada task force on the August 14, 2003, blackout and subcommittee on grid disturbance in India that affected 250 million people, concluded that inappropriate and uncoordinated relay protection settings were one of the principle reasons of cascade outage and widespread blackout[1]. This research work presents the challenges and proposed remedies for the future electric power industry to implement new adaptive stateof-the-art techniques in the field of transmission grid protection. The main issues © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 U. Patel et al., Futuristic Trends in Numerical Relaying for Transmission Line Protections, Energy Systems in Electrical Engineering, https://doi.org/10.1007/978-981-15-8465-7_1

1

2

1 Transmission Line Protection Philosophy

for power transmission utilities are to sustain stability and reliability of the network with implementation of smart protection schemes. The power grid must be sheltered from variety of disturbances developing in the system. A small breakdown of protection scheme can result in huge amount of power loss which ultimately reflects large wastage of fossil fuels. The current situation urges development of dedicated adaptive numerical relaying scheme for implementation of new ideas in the field of power system transmission line protection technologies. This section outlines protection schemes of power transmission network along with the necessity of advanced protection technologies in brief.

1.1.1 Transmission Network Protection Schemes Power system control schemes consist of mainly two types of systems: 1. Dynamic local equipment protection schemes 2. Synchronized wide area system protection schemes Both kinds of protections work independently of one another in most cases. Dynamic local protection is normally applied to individual piece of equipment such as transformers, generators, transmission lines, etc. These protective schemes are intended to shelter individual section affected by faults or disturbances which can protect the system within milliseconds in most cases. Local protection acts dynamically in coordination with other protection schemes or network management systems. System-wide control schemes protect the interconnected power system rather than individual pieces of equipment. Examples include undervoltage or underfrequency load shedding, generation rejection, and controlled separation (islanding) to mitigate the severity of disturbances. Electric Power Research Institute (EPRI) has investigated cascade tripping and wide-spread blackout in smart power grid network of USA. It has been suggested that protection schemes and control strategies must stop system degradation, minimize impacts, and facilitate the system restoration. EPRI and IEEE Power Society, USA, had prepared report on power system protection and identified three major areas of concern [1]: 1. Prevention of blackouts through undervoltage load shedding: During major power system disturbances, the prime line of defense is accurate and timely determination of load shedding and its proper implementation. It can be achieved through technological advancement in various areas. 2. Avoiding cascading failure through improved Special Protection Schemes (SPSs): This can be achieved by the application of islanding which is one of the contemporary problems of research. Conventionally used ’fixed’ islanding in which

1.1 Introduction

3

islanding points are predetermined, it can be replaced by an ’intelligent’ separation system in which islanding points can be predicted based on dynamic real-time power grid scenario. As with load shedding, intelligent islanding can benefit substantially from the integration of information gathered and processed by wide area measurement systems (WAMS). 3. Transmission corridor protection through use of numerical relaying schemes: An overarching concern is the need to address and remedy the lack of relay protection research for students and practicing engineers. As stated in different wellknown literatures, out of all the faults, occurring in the power system, around 50% of the faults are occurring on transmission lines. The chances of fault occurring are due to touching of trees on the lines, heavy storms, flashover of insulators resulting from dirt deposits and improper relay coordination, etc. Thus, in order to maintain the uninterruptable power supply to the potential consumers and critical loads, the accurate and reliable transmission line protection schemes are highly required. The protection schemes must be adaptive in nature to mitigate the disturbances developing in the system. The protection scheme should be able to cope up with the power swing developed in the system due to sudden application/removal of the load. The service continuity should be regained as soon as fault is cleared in the system which maintains system reliability. The next section highlights the requirement of advanced power system protection schemes.

1.1.2 Necessity of Advanced Power System Protection Taskforce on grid disturbance in India has insisted that numerical relays used for protection of EHV transmission line must remain stable during power swing [2]. In the event of a major power system disturbance, control and security actions must shelter the power system degradation, minimize impacts, and enable system restoration. However, today’s protection and control actions were not well designed and coordinated for a fast-developing disturbance and may be too slow to minimize the impact [2]. Protection engineers normally analyze dynamic situations through simulation software using off-line analysis. As a result, technocrats are sometimes forced to address complex issues by relying on static company policies and past experience. A large percentage of major system disturbances can be traced to ‘hidden failures’ in protective elements which remain undetected until the system is exposed to certain system disturbances [3]. The ability of a numerical relay to detect type of disturbance occurring in smart power grid network before an incorrect operation occurs is prime requirement of digital protection. In order to minimize impacts, protective relaying should be able to monitor the real-time system disturbances and execute corrective actions. To justify these, advanced numerical relay-based schemes, which incorporates adaptive techniques in pace with disturbances occurring externally, can

4

1 Transmission Line Protection Philosophy

be implemented. In the following topics, the traditional transmission line protections are outlined along with in-depth literature review.

1.2 Traditional Transmission Line Protection Transmission lines are generally provided with following protection schemes [4]: 1. 2. 3. 4.

Time-graded protection Differential protection Distance protection Carrier-aided protection

In order to justify the area of research, basic operation and setback of each protection schemes are described here in brief.

1.2.1 Time-Graded Protection In this scheme of protection, discrimination of time is incorporated. In other words, time setting of relays is so graded that in the event of fault, the smallest possible part of the system is isolated by operating the relay nearer to the fault location. In time-graded protection system, the time of operation of the each relay is kept fixed as shown in Fig. 1.1. Whenever fault occurs at E, the relay located at point D will operate within 0.5 s. In case it fails, then the successive relays will provide backup protection and operate after the definite time lag as shown in Fig. 1.1 for the fault at point E. It is independent of the magnitude of operating current. The major drawback is that continuity of the supply cannot be maintained at the receiving end in case if the fault is transient in nature. So, it is mainly used in the system where discontinuity of the supply does not matter. Fig. 1.1 Time-graded—ring main system

A

Operang Time

G

2 Sec Line-1

B

1.5 Sec Line-2

C

1 Sec

D 0.5 Sec E

Line-3

Line-4

2 Sec 1.5 Sec 1 Sec 0.5 Sec Distance

1.2 Traditional Transmission Line Protection

5

1.2.2 Differential Protection It is further classified into current differential protection scheme and Merz-Price voltage balance scheme [5]. The principle of operation of both the schemes is based on differential circulating current protection. The basic operation along with its main setbacks is outlined herein brief. In Merz-Price voltage balance scheme as shown in Fig. 1.2, current transformers (CTs) are connected in series with a relay in such a way that under normal conditions, their secondary voltages are equal and in opposite direction, i.e., they balance each other. Therefore, there is no driving force for circulation of current through relay coils. When a fault occurs in the protected zone of transmission line then more current will flow through CT of supply side than through load-side CT. Therefore, their secondary voltages become unequal and circulating current flows through the pilot wires and relays. The circuit breaker at both the ends of the line will trip out, and the faulty line will be isolated. Though it provides very fast protection against ground faults and it can be used for ring mains/parallel feeders and reduces the damage, still it has some major drawbacks: 1. Current transformer at both the ends should be exactly identical, which is very difficult. 2. If there is a break in pilot wire circuit, the system will not operate. 3. This system is very expensive due to greater length of pilot wire. 4. Due to charging current of pilot wire capacitance, relay may operate even during normal operating conditions. CB R

CT1

Transmission Line

CT2

CB R

Y

Y

B

B

Relays To Trip Circuit

Pilot Wires To Trip Circuit

Fig. 1.2 Merz-price voltage balance differential protection scheme [5]

Relays

6

1 Transmission Line Protection Philosophy

1.2.3 Distance Protection As time-graded system gives long time delay in fault clearance at the generating station and differential protection becomes too expensive owing to its greater length of pilot wires, these protection schemes are not suitable for very long and high-voltage transmission lines. Therefore, there is need of distance protection in which the action of relay depends upon the distance (or impedance) between the point where relay is installed and the point of fault. At present, majority of all the transmission lines are protected by different kinds of distance relays for primary and backup protections; hence, its basic fundamentals along with research issues are described here in detail. Figure 1.3 explains basic principle of operation for distance relay. It is double-actuating quantity relay with one coil energized by voltage, and the other coil is energized by current. The relay operates when the ratio V /I is below the set/threshold value [6]. Distance relays measure impedance using bus voltages and line currents as shown in Fig. 1.4 at the relay location. When the fault occurs, it calculates the impedance (Z F ) between relay location and fault. If calculated impedance is less than the set value (Z set ) then relay will generate the trip signal to operate the circuit breaker and relay deems this fault F 1 as internal fault. On the other side, if calculated value of impedance is more than the preset threshold value then the trip signal will not be issued instantaneously and this fault is considered as external fault. To Trip Circuit

F

Armature From P.T.

Armature

V

I

From C.T.

Fig. 1.3 Basic principle of operation of distance relays

Z ZF G

A

CB

B

CT

CB

CB

Impedance Relay

F1 PT

Fig. 1.4 Impedance calculation in distance relays

F2

1.2 Traditional Transmission Line Protection

7

When fault occurs on the transmission line, the fault current increases and the voltage at fault point reduces. The ratio is measured at the location of current transformer (CT) and potential transformer (PT). The current gives operating torque and voltage gives restraining torque. The voltage at PT depends on the distance between the PT and the location of fault. Farther is the distance of fault from PT, more will be the voltage available at PT and smaller is the distance, less will be the voltage. The ratio of voltage and current reflects impedance seen by the relay. It is proportional to the distance between the relaying point and the fault along the line. Hence, such relay provides protection only up to certain length of the line equivalent to its impedance setting, and hence, it is called distance relay or impedance relay. Figure 1.5 shows a simple power system consisting of relays at station A, B and C which are set to operate for impedances less than Z 1 , Z 2 , and Z 3 , respectively. The first section covers 80% portion of line AB falls under zone 1 protection of relay at station A. Zone 1 provides instantaneous protection against the fault. Zone 2 and Zone 3 provide backup protection to the next section of lines and operates with some definite time lag. Zone 2 normally covers whole primary line section (AB) and additional 50% portion of the next line section. Zone 3 provides protection to remaining portion of the transmission line. Suppose a fault occurs between substation B and C, the fault impedance at A and B will be Z 1 + Z and Z respectively. The impedance value for relay B is less than its set value Z 2 , and hence, relay B will operate instantaneously. If relay B fails to operate, then relay A will provide zone 2 backup protection, and it will operate with a time lag of T 1 . In this manner, primary and backup protection can be obtained for all zones of power transmission network. In actual practice, it is not possible to obtain instantaneous protection for complete length of the line due to inaccuracies in the relay elements and instrument transformers. Thus, the relay at A would not be very reliable in distinguishing between a fault at 99% of the distance AB and the one at 101% of distance AB. This difficulty can be overcome by using ‘zone 3’ distance protection scheme as shown in Fig. 1.6 below. Z1

Z2

Z3

FD

Z1 A G

CB

Z2

Z3

T1

CT

R

Instantaneous

PT

Fig. 1.5 Operating zones of distance protection scheme

B

T2 C

Adjoining line

8

1 Transmission Line Protection Philosophy Zone 3

Relay A Time

Fig. 1.6 Three zone distance protection scheme

Zone 2 Zone 1

Zone 1

Relay B Zone 2 Relay C

G A

B

C

In this scheme of protection, three distance elements are used at each terminal of the line. The zone 1 element covers first 80–90% of the line and is arranged to trip instantaneously for faults in this portion. The zone 2 element trips for faults in the remaining 10–20% of the first section and for the fault in 50% area of next line section with definite time lag. The zone 3 element provides back-up protection in the event a fault in the subsequent section which is not cleared by its breaker. Many types of distance relays have been developed and applied for transmission line protections [7] like: 1. 2. 3. 4. 5. 6.

Impedance Relays Reactance relay Mho relay Ohm or angle impedance relay Offset Mho relay Quadrilateral relay and other special characteristics

Because of inherent directional feature, Mho relay is used for primary protection and backup protection in zone 2. However, it incorporates limited fault impedance, and hence, a quadrilateral relay is also widely used for all zone protection [8]. However, protection engineers can use any preferred characteristics depending on system and line parameters. Traditional transmission lines are also protected by carrier-aided scheme as outlined in the next section.

1.2.4 Carrier-Aided Protection In carrier-aided protection of transmission line, carrier signal is transmitted through cable or telephone lines or fiber optic cable, etc. However, there is always a chance of loss of carrier signal and delay in transmission of this signal. It can subsequently result in subsequent mal-operation of the relay, which is major drawback of carrier-aided protection systems [8]. It can be narrated that because of significant advantages over the other protection schemes, distance protections are wide used for transmission line protection. However, it can sometimes fail to execute the protective action because of abnormal system and fault parameters which is a contemporary problem of research. Many

1.2 Traditional Transmission Line Protection

9

researchers have evolved numerous techniques to shelter transmission grid network. An exhaustive literature review on the same is outlined in next section.

1.3 Literature Review During the last three decades, remarkable success has been achieved in the field of distance protection schemes using digital/numerical relaying schemes. Many digital relaying schemes have been developed by the designers and researchers using microprocessors, microcontrollers, and digital signal processors. Majority of all these schemes employ various artificial intelligence/wavelet transform techniques for adaptive innovative protections of transmission lines. In this section, a detailed literature review is presented on the same along with scope for further improvements. One of the fundamental requirements for any numerical relays is fast and accurate phasor estimations of voltage and current phasors which can accurately remove harmonics and DC component from the fundamental frequency component. By using the estimated phasor values, faults can be classified, trip signal can be generated for varying system conditions or prohibited in case of stable power swing and automatic reclosing action can be taken to regain service continuity. Hence, literature survey has been devised into four major areas which directly concerns with the advanced protection systems. These areas are: 1. 2. 3. 4.

Phasor estimation techniques Numerical distance protection for fault context identifications Power swing detection methods Automatic reclosing technologies

1.4 Review on Phasor Estimation Techniques Recently, numerous filter algorithms are available for accurate and fast phasor estimation [9, 10]. Some algorithms implement filters like modified notch [11], mimic [12, 13], Walsh [14], Kalman [15, 16] and least square error (LSE) techniques [17], but their slower convergence speed is the critical issue. In majority of all the above methods, phasor estimation is performed by sampling the voltage and current waveforms for at least one cycle or more. In order to detect the fault and taking the adaptive corrective action, instantaneous phasor estimation is required for fast and accurate protection trip, use of which is limited to permanent faults. The accuracy and response time of any numerical distance relay depend on algorithm used for phasor estimation of input signals used in computer numerical control. The main issues with these algorithms are error in estimation and delayed operation due to influence of decaying DC component. Apart from this, the time constant of the decaying DC

10

1 Transmission Line Protection Philosophy

component is unknown; if it matches with the assumed value of the time constant, then only accurate phasor estimation will be performed. An application of distance relaying scheme to compensate fault location errors due to fault resistance is presented in [18] based on discrete Fourier transform (DFT). It is found that the error in estimation increases up to 6.5%, which is a major indicator of inappropriate phasor estimation. Linˇciks et al. [19] proposed distance relaying algorithm for L-G fault on mediumvoltage network based on digital computation for fault location estimation. Two methods are proposed, namely improved fundamental frequency method and admittance method. Former provides an accuracy of 10%, and latter provides an accuracy of 5% for locating the faults which are not sufficient during practical implementation. Wen et al. [20] presented fast distance relaying scheme based on equal transfer process of transmission lines (ETPTL) which implements a low-pass filter for the elimination of high-frequency components. However, the error in fault location estimation up to 5% is achieved which is not enough for accurate and fast relaying applications. Hence, fast and accurate phasor estimation which can eliminate the harmonic and DC component within subcycle time and generate trip signal is the need for the protection engineers. In the present era, the more efficient distance relaying algorithms are evolved by many researchers. In majority of these algorithms, phasor estimation has been performed by Fourier or wavelet transforms for generation of feature vector which is then applied to soft-computing-based classifier to estimate fault context. Out of twelve discrete relaying algorithms referred in [21], only distance protection algorithm outlined in [22, 23] and modified Fourier filter algorithm [24] respond satisfactorily and robustly to wide range of input signals.

1.5 Numerical Distance Protection for Fault Context Identification Digital/Numerical relays provide significant benefits for industrial as well as commercial systems where an economical but effective protection scheme is required. These techniques are increasingly being adopted by major power system networks for improving power system operation [25–27]. Further, they provide many advantages such as reduction in lifetime management costs, maintenance costs and also have the ability to provide primary and backup protection with great deal of reliability. Nowadays, due to technical advancement in the field of microprocessors and microcontrollers, the numerical relays are developed by many industries. Numerous techniques are available that can provide advanced protective features in the numerical relays. Some of the widely used techniques are based on artificial neural network (ANN), traveling-wave-based and wavelet transform-based techniques. However, due to improper software development in digital relays, it can mal-operate during

1.5 Numerical Distance Protection for Fault Context Identification

11

various unexpected disturbances. The detailed literature review based on these protection philosophies has been outlined in this section along with the possible scope of improvement.

1.5.1 Review on Microprocessor-Based Protection Schemes Brewis et al. [28] described a scheme which reports on in-service performance. The scheme is based on instantaneous elements contained with numeric over current relays and discussed design issues for transmission line and busbar protection. But the failure of instantaneous blocking element in case of power swing, close-in fault, and transient fault is the prime limitation of the proposed scheme. This can be overcome by innovative and adaptive distance relaying schemes [29–32], but the outlined schemes are not validated for power swing scenario. Kennedy and his colleague [33] proposed a differential relay using microprocessors for transmission line and busbar protection based on harmonic current restraint principle. However, the prime limitation of this scheme is inhibition of relay operation if the content of harmonics in the differential relay exceeds the threshold value. Many researchers have represented realization of distance protection using microprocessors and controllers, but protection against extreme fault conditions are realized only in off line simulations [34, 35], their hardware implementation is the need for future implementations. However, microprocessor-based relays are suffered from the disadvantages of interfacing of analog-to-digital converter (ADC) and its digital signal processing time and adjusting sample and hold time in it, because of which distance relay response time will be limited [36]. These disadvantages can be overcome by advanced microprocessor and digital signal processors having on-chip ADC.

1.5.2 Review on Neural Network-Based Protection Schemes Nowadays, machine learning-based approaches are widely used in various nonlinear applications for classification and regression problems. One of the most widely used approaches is artificial neural network (ANN). Jamil et al. [37] have introduced generalized neural network (GNN) and wavelet transform (WT)-based approach for fault location estimation of a transmission line considering the effect of system loading level, fault inception angle, fault resistance, DC offset and harmonics contents in the transient signals of faulted transmission line. Authors have claimed fault location estimation accuracy of the order of 3%. The effect of power swing and operation of relay in zone 3 of transmission line with high accuracy and low tripping time are not considered. In literature [38, 39], a new adaptive digital distance relaying scheme is proposed which takes care of all such abnormalities of the conventional ground distance relays

12

1 Transmission Line Protection Philosophy

and measures the correct value of impedance during phase-to-phase and phase-tophase-to-ground intercircuit faults. MATLAB/Simulation results demonstrate the effectiveness of the proposed scheme since the maximum percentage error is within ±2%. The algorithm based on artificial neural network has also been developed in [40], but the algorithm with artificial neural network (ANN) configuration requires sufficient training data for fast fault detection and hence requires large memory. Yadav has presented a transmission line relaying scheme for fault detection and classification using ANN based [41] in thyristor-controlled series capacitor (TCSC) compensated transmission line. However, the effect of decaying DC component and about to reach operation of the relay is not considered. In the present era, more sophisticated soft computing-technique-based algorithms are evolved by many researchers. In majority of these algorithms, the phasor estimation of analog input signals is performed by various techniques outlined in Sect. 1.4. The estimated phasors are used to derive the feature vectors which are further applied to soft-computing-based classifier to identify the fault context. The feature vectors extracted by discrete wavelet transform (DWT) are utilized by genetic algorithm (GA)-based fault classifier in [42] and support vector machine (SVM)-based classifier in [43]. The main setback of above algorithms as applied to distance protection is training and testing accuracy, determination of unknown regulation parameters, and complexity for implementation. Moreover, aforesaid algorithms can be applied to the configuration of transmission for which it has been trained. It cannot be applied to other standalone power system network configuration without accurate training and testing. In the last two decades, due to the development in memory technologies, the use of soft-computing-based techniques for optimizing relaying performances in the power system is rapidly increasing. The ANN consists of subneural networks which can provide protection against power swing detection, fault detection, fault phase identification, and fault location estimation [44–46], but its hardware implementation is not practically feasible because of amount of data needed for training and testing. Inception of various types of intercircuit faults on a series compensated parallel transmission line involving fault resistance makes the protection system quite complicated.

1.5.3 Review on Traveling-Wave-Based Protection Schemes Wang et al.[47] presented a transmission line and busbar protection scheme based on polarities of transient current waves for discrimination of faults and abnormalities. Jiang et al. [48] proposed transient-based protection (TBP) technique based on energy spectrum of fault transient. Subsequently, Chen and his co-workers [49] described busbar protection scheme based on wavelet analysis which works on fault-generated transient current signals. However, when a fault occurs at a voltage inception angle close to zero degrees, it does not generate significant traveling wave components. In addition, the bandwidth limitation of the transducers, particularly the capacitive

1.5 Numerical Distance Protection for Fault Context Identification

13

voltage transformer (CVT), limits the applicability of such techniques, particularly for close-in faults, which generates very high-frequency traveling waves that are outside the bandwidth of perceptibility of relays [50]. Many researchers have also realized traveling-based differential protection scheme in power system computeraided design (PSCAD) software [51]. Due to implementation of adaptive numerical local protection scheme for fault context identification and soft-computing-based optimization techniques for power swing detection; nowadays, traveling-wave-based transmission line protection schemes are not implemented in commercial relays.

1.5.4 Review on Wavelet Transform-Based Protection Schemes Nowadays, wavelet transform finds many applications in the field of power system protection as compared to conventional Fourier transform technique. In wavelet transform, time domain signal is converted into frequency domain signal using mother wavelet. In frequency domain, it can be further decomposed into various wavelet components by selecting an appropriate level decomposition method to analyze the discontinuity and spikes in the signal. The extracted information can be used for further digital signal processing and analysis for conducting protective actions. Recently, many filter algorithms are available for accurate and fast phasor estimation using wavelet analysis. Digital protective relaying of transmission lines has been greatly benefited from the development of artificial intelligence and signal processing techniques. Yadav [52] has presented a novel transmission line relaying scheme for fault detection and classification using wavelet transform and linear discriminant analysis, which indicates that the use of wavelet transform in power system protection is increasing day by day. Authors have considered the parameter variation such as faulty type, fault location, fault resistance, fault inception angle, power flow angle for single-circuit and double-circuit transmission line and claimed an accuracy of 100%, but the effect of power swing, degree of compensation and overloading issues are not considered in the same. Samantaray et al. [53, 54] have proposed fast S-transform (FST)-based distance relaying in transmission line. It is concluded that S-transform combines elements of wavelet transform and short-time Fourier transform. However, it imposes higher computational burden which can be reduced by down sampling. The results show that FST can suitably replace the existing DFT-based algorithm for digital distance relaying task in power system transmission line network. The prime limitation of the same is proper estimation of fault clearing time is not mentioned. Secondly, the effects of transmission line parameters like high resistance fault, fault inception, type of fault, multi-location fault, power flow angle, power swing, and overloading are not considered. Jamil M. and his colleagues [55] propose new power swing blocking function based on wavelet transform (WT). Although numbers of transmission line protection

14

1 Transmission Line Protection Philosophy

schemes have been proposed using wavelet transform [56–62], it is suitable for discrimination of DC component from the extracted waveform because of which overreaching problems can be eliminated. There exists a lot of scope for further improvement especially for the effective discrimination between overloading and power swing which can eliminate overreach and under reach problems effectively. Sudha et al. [63] have investigated a comparison between different approaches for fault classification in transmission lines. Three fault classification techniques are presented for comparison in terms of complexity and time. Firstly, based on evaluation of symmetrical components of fault currents, secondly by applying wavelet transform on fault voltages and currents subsequently followed by energy computation and thirdly after modal transformation, signal decomposition is performed to determine the energy densities for fault classification. The fault data for various fault locations and fault inception angles is obtained by simulating the power system in MATLAB/Simulink. The results show that the last algorithm provides a fast fault classification with adaptive window programming feature. The main issues with these algorithms are error in estimation and delayed operation due to influence of decaying DC component. Compensation of ground distance function (GDF) and resistive reach assessment in quadrilateral characteristics is proposed by Sorrentino [64]. This method of compensation gives the suitable results during phase impedance calculations but still needs improvement in adaptive calculation of resistive reach of relay. Pasand [65] and Yadav [66] described adaptive decision logic to enhance distance protection of transmission line. However, during occurrence of high resistance faults, delayed tripping of numerical relay cannot be avoided by the said algorithms. Majority of above literatures, incorporate limited fault and system dynamics like fault resistance, fault inception, type of fault, multi-location fault, fault inception angle, power flow angle, power swing and overloading. However, the distance relays must be capable of detecting the fault with all above possible issues with highest accuracy and lowest tripping time for increasing reliability of transmission line protection scheme, although it’s a matter of contemporary research. The next section demonstrates the detailed literature reviews on power swing detection methods.

1.6 Power Swing Detection Methods Distance relay may mal-operate during power swing condition which can arise due to variation of mechanical input to the generators, load encroachments and switching of adjacent loaded line during fault or maintenance. Hence, it is necessary to discriminate between fault and power swing condition in case of line protection. The distance relays used for the protection of transmission lines must generate trip signals at the time of fault. On the other hand, it must block the tripping action during the power swing generated in the grid due to aforesaid disturbances. Detection of faults occurring during the power swing is the most complicated task for the protection engineers.

1.6 Power Swing Detection Methods

15

Hence in this section, literature review has been presented for power swing detection for both uncompensated and compensated transmission lines.

1.6.1 Review on Power Swing Detection for Uncompensated Transmission Lines In worst case (during large disturbances), the magnitude and phase of both current and voltage change abruptly which may result into loss of synchronism and/or out of step situation [67]. During power swing phenomenon, the apparent impedance seen by a distance relay may reach to a value lower than the line zone setting and hence can cause unwanted tripping of the line protected by the distance relay [68]. This mal-operation of distance relay has to be prevented using power swing blocking (PSB) function. On the other hand, to ensure reliability and system stability, it must be operated without any undesired time delay for different type of short-circuit faults to generate out-of-step (OOS) signal for the relay on the transmission line [69, 70]. Many techniques have been proposed by researchers to distinguish between power swing and fault condition. Mechraoui [71] has presented power swing discrimination method based on calculation of phase angle difference of voltages at local and remote ends. The same authors have proposed a scheme for detection of high resistance earth faults which may occur during a power swing [72]. However, the validation of the proposed technique does not include symmetrical fault cases. Lin et al. [73] proposed a cross-blocking scheme based on rate of change of active and reactive power to detect symmetrical faults during power swing condition. Furthermore, Su et al. [74] introduced an improved fast fault detection method based on the swing center voltage (SCV). However, the said schemes result into delayed operation in case of fault occurrence during power swing. This cannot be acceptable for EHV transmission lines of 220 kV and above voltage levels. Later on, Brahma [75] has proposed wavelet-transform-based technique which detects power swing as well as symmetrical fault during power swings. However, the prime limitation of this scheme is that it requires a high sampling rate of 40.96 kHz. Reddy et al. [76] have demonstrated modified wavelet and S-transformer-based technique for power quality disturbance in power system. Moreover, Bhalja et al. [77] and Mohamad et al. [78] have presented an approach to detect power swing based on S-transform signal processing tool. Conversely, high filtering followed by decomposition in wavelet increases the computation time of algorithm and demands more hardware requirements. Abidin et al. [79] offered a blocking scheme for distance protection during power swing based on derivative of the reactive power as seen by the relay. However, dQ/dt criterion alone fails to discriminate the fault and swing condition during severe system disturbances. Pang and his colleague [80] presented an algorithm for detection of symmetrical faults during power swing based on extraction of high-frequency energy component of forward and backward traveling waves induced by faults using wavelet

16

1 Transmission Line Protection Philosophy

transform. However, traveling-wave-based technique requires high sampling rate and also has difficulties in distinguishing traveling waves reflected from the fault point and those from remote end of the line. Lotfifard et al. [81] have proposed a method based on extraction of current components to detect symmetrical faults during power swing using Prony’s algorithm. However, this method depends on the relationship between faults and decaying DC component of fault current and hence, cannot be used as a standalone scheme. Afterward, Gautam et al. [82] proposed out of step blocking function for distance relays using mathematical morphology (MM). However, the process of design and selection of structuring element in MM is very complex and time consuming. Another method is proposed in [83] based on detection of the fundamental frequency component created on the instantaneous three-phase active power after inception of a symmetrical fault. However, the fault resistance is not considered in his technique. Sharifzadeh et al. [84] have presented a method based on rates of change of voltage and current magnitudes to block the distance relay during voltage-degraded condition. Jafari et al. [85] demonstrated a scheme based on the circular locus of the admittance trajectory and its center behavior. Zadeh et al. [86, 87] demonstrated an artificial neuro-fuzzy inference system (ANFIS)-based scheme for power swing blocking. However, the main disadvantage of fuzzy-based schemes is that it uses explicit knowledge base fuzzy rules and hence gives insufficient information about the fault symptom relationship. Moreover, it needs a large number of training patterns in achieving a consistent relay operation. ANFIS-based scheme claims training accuracy of 86.02% and testing accuracy of 83.84%. Later on, Seethalekshmi et. al. [88] evolved SVM-based power swing classifier scheme for discrimination between fault and power swing under the existence of unified power flow controller (UPFC) and claims training accuracy of 99.83% (against 86.02% by ANFIS-based scheme) and testing accuracy of 99.30% (against 83.84% by ANFIS-based scheme) after 3427 iterations and using 1196 support vectors. But the response time of the same will be very high due to larger iterations and amount of memory required. Simultaneously, it does not consider the asymmetrical faults occurring during the power swing. Sauhats et al. [89] presented out-of-step relay testing procedure for distance relays, but the same is not validated for symmetrical fault detection during power swing involving high electrical resistance. It is always desired for protection schemes to be fast, reliable, and secure which can efficiently detect any type of power system disturbance and can also discriminate between fault and power swing condition. Hence, adaptive and fast numerical relaying scheme is required which can improve the training and testing accuracy to a considerable level. Moreover, detection of symmetrical fault during power swing is challenging task for protection engineers especially in series compensated transmission line (SCTL). Compensation plays important role in EHV systems for improving stability and power transfer capacity. Hence, in the next section, the literature review has been presented on power swing detection for transmission lines equipped with compensation.

1.6 Power Swing Detection Methods

17

1.6.2 Review on Power Swing Detection for SCTL The critical requirements for digital protection of SCTL are reliable and fast phasor estimation of input signals to estimate correct impedance reach. Recently, many filter algorithms are available for accurate and fast phasor estimation using wavelet analysis. Kale et al. [90] presented an excellent comparison between discrete wavelet transform (DWT) and discrete Fourier transform (DFT). It is observed that the wavelet transform has wide application in power system protection, but it has limited role in phasor estimation. Later, Garcia et al. [91] evolved fault detection and classification scheme based on model transformation and wavelet analysis but the same is not applied for discrimination between power swing and faulty condition. In majority of reported algorithms, phasor estimation has been performed by Fourier or wavelet transforms for generation of feature vector which is further applied to soft-computing-based classifier. Soft-computing-based power swing detection techniques are extensively used by many researchers. In majority of the schemes proposed by researchers, phasor estimated values are used as feature vectors. The extracted vectors are then applied soft computing algorithm to discriminate between the fault and power swing scenario. The features extracted by discrete wavelet transform (DWT) are utilized by genetic algorithm (GA)-based fault classifier in [92], and accuracy of the order of 99% has been achieved but the same is not validated for fault during power swing conditions. Further, application of DWT-based vectors to artificial neural network (ANN)-based classifier is presented, and accuracy of the order of 96% has been achieved in [93] and 97.5% in [94]. The main issues with these algorithms as applied to distance protection are less training and testing accuracy due to improper settings of regulation parameters. Chothani et al. [95] have applied DFT-based feature extraction technique to SVM for discrimination between fault and power swing conditions. Extensive analytical approach indicates that SVM provides very fast, reliable, and secure response from protection view point. Seethalekshmi et al. [96] have evolved SVM-based fault classification algorithm to prevent mal-operation of distance relays under power swing and voltage instability using two-stage SVM. However, the outlined algorithms [95, 96] are not validated for the lines with compensation in which nonlinearity will be profound as compared to uncompensated transmission line. Recently, Vyas et al. [97] have used SVM as pattern recognition tool for the lines with TCSC, and its performance has been compared with ANN-based scheme. It demonstrates efficacy of SVM to solve data mining problems over ANN-based techniques but the same is not validated for variety of power swing scenario as discussed the Chap. 5. Afterward, phase-space (PS)-based symmetrical fault detection technique for SCTL is developed by Dubey [98] using embedding theorem. It demonstrates that time series can be mapped to a higher-dimensional space called phase-space through embedding. However, the presented scheme is not investigated for line with compensation subjected to wide variation in fault contexts as well as power swing scenario. An innovative traveling-wave-based fast fault detection scheme for SCTL during

18

1 Transmission Line Protection Philosophy

power swing has been reported in [99]. The voltage and current samples are applied for model transformation to calculate mathematical morphological gradient (MMG) applied to SVM. It has been reported in many literatures that mathematical morphological algorithms impose high computational burden and result in delayed response from the protection view point. Ebha et al. [100] evolved a transmission line protection scheme based on SVM and ANN in presence of nonlinear loads. The post-fault voltage signals processed by Kalman filter serve as feature vectors for disturbance detection using SVM and fault zone identification using ANN. However, validation of the same is not carried out on the line equipped with compensation. Also, the developed scheme is not tested for variety of fault during power swing scenario. Pavlatos et al. [101] have proposed linguistic representation of power system signals. The presented attribute grammar (AG) is further used for syntactic pattern recognition of power system waveforms to detect the faults in transmission lines in [102]. The developed pattern recognition technique can constitute a promising alternative approach for detecting disturbance occurring in power system. A universal pilot relaying scheme for series and shunt compensated transmission line is presented in [103] based on incremental reactive power coefficients (IRPCs). If IRPC of any phase is found less than −0.5, the internal fault is detected else fault is discriminated as the external. However, pilot relaying scheme requires high-speed communication channel for transferring traveling waves. Subsequently, a robust fault detection and discrimination (RFFD) technique is proposed in [104], which utilizes Hilbert transform for phasor estimation. The estimated phasors are used to derive normalized feature value (NFV) and global trip (gTRIP) parameters to increase sensitivity and selectivity of the relay for fault zone identification. However, the performance of the same is not validated for the line with compensation immune to power swing scenario. It can be narrated that protection task becomes more complicated when the lines are equipped with compensation to improve the power transmission capacity. The series compensation adds more nonlinearity in the system. Hence, power detection in series compensated transmission line is the most complicated task for the protection engineers. It is also narrated in [2] that distance relays mal-operate due to power swing which ultimately leads to wide spread blackout problem. Further, symmetrical fault detection during power swing scenario is the challenging task for the researchers in the era of smart grid technologies. The next section demonstrates an extensive literature reviews on the yet another issue concerning identification of transient fault occurring in the transmission line in order to improve the stability of the system under consideration.

1.7 Auto-Reclosure Technology Review The critical requirements for numerical distance protection used for transmission line are reliable and fast relay responses in order to maintain the power system stability.

1.7 Auto-Reclosure Technology Review

19

Faults occurring in transmission line are mainly transient in nature which can result in loss of continuity of power. Auto-reclosure is one of foremost ways out for the discrimination between transient fault and permanent fault which can regain service continuity. Transient faults are caused due to flashover of the insulator as a result of lightening surge or due to momentary tree contact. For such faults, auto-reclosing improves transient stability and reduces adverse effects on alternator shafts [105]. The conventional auto-reclosing schemes applied to EHV power transmission lines are based on fixed time interval reclosure, i.e., the circuit breaker recloses after fix time known as dead time following trip signal generated by the relay. However, unsuccessful reclosing with fixed dead time or reclosing on permanent fault threatens stability of power system and aggravate adverse effects on protective system. An adaptive successful single-phase auto-reclosing scheme can be the optimum outcome for improving transient stability of the power system. In the present era, many researchers have presented various discrimination algorithms to implement auto reclosing schemes for transmission line protection [105, 07]. Artificial neural network (ANN)-based schemes have been demonstrated in [108, 110], but ANN-based scheme must be trained for each individual transmission system topology. Moreover, the behavior of the ANN cannot be predicted when exposed to some patterns outside the training classes. Wavelet analysis and neural network-based auto-reclosing scheme have been developed to recognize certain situations in order to deduce whether and when to reclose the circuit breakers [111]. Stability margin-based variable dead time auto-reclosing scheme has been presented [112] which is only suitable when the relative degree of power system transient stability is sufficiently large. Techniques based on mathematical morphology [113] require high sampling frequency to obtain the reliable results. This increases computational burden and time of operation of reclosure. Recently, many filter algorithms are available for accurate and fast phasor estimation using wavelet analysis [114–116] and conventional DFT [117] which issues reclosing command after some fixed dead time. Algorithm outlined in [118] implements mimic filter for phasor calculation of analog input signals. With this, the time constant of the decaying DC is unknown, and if it is matching with the assumed value of the time constant, then only accurate phasor estimation will be performed. Biswal [119] has presented transmission line fault classification using integrated moving sum (IMSUM) technique, but the same is not applied for discrimination between transient and permanent fault condition. Although numerous techniques are evolved by many researchers for discrimination between transient fault and permanent faults, still there is huge scope of improvement in auto-reclosing technologies which can reduce the reclosing time adaptively as and when transient fault is cleared. The developed scheme presented in this book remains stable for transient fault detection on series compensated transmission lines. Based on the exhaustive literature reviews, the critical issues influencing the performance of numerical distance relays used for security of smart power grid network are outlined in next chapter along with the objectives and systematic plan of work to execute protective actions.

Chapter 2

Transmission Line Protection: Issues and Research Needs

2.1 Issues in Numerical Distance Relays There are many factors which can be responsible for incorrect implementation and operation of distance protection schemes. Out of the many issues, the potential issues are presented in this section along with mathematical and technical fundamentals for each of them. The main influencing parameters are: 1. 2. 3. 4. 5. 6. 7. 8.

Effect of DC component. Close-in fault. High resistance fault. Fault inception angle (FIA) and power flow angle (PFA). Load encroachment and uncoordinated zone 3 relay settings. Transient faults and auto-reclosing schemes. Power swings. Series compensation in transmission lines.

Due to the effect of above-listed factors, the distance relays can mal-operate. The main reason behind it is accuracy of phasor estimation of applied analog signals which can result in either delayed or false tripping. Here, the protective issues arise due to the above factors are outlined in subsequent sections.

2.1.1 Effect of DC Component The accuracy of any distance relaying algorithm depends on proper extraction of fundamental complex phasor. However, when the fault occurs, the DC component and harmonics will be superimposed on the original phasor resulting incorrect estimation. The most widely used phasor estimation technique is discrete Fourier transform (DFT) for proper extraction of fundamental component from the analog signal by © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 U. Patel et al., Futuristic Trends in Numerical Relaying for Transmission Line Protections, Energy Systems in Electrical Engineering, https://doi.org/10.1007/978-981-15-8465-7_2

21

22

2 Transmission Line Protection: Issues and Research Needs

separating DC component and harmonics. The detailed discussion on the same is as follows: Let, x(t) be the sinusoidal time varying signal which is to be analyzed. X(k) is the discrete time signal derived from x(t) by sampling it with sampling frequency F s . Assuming, T = fundamental cycle time; N = number of samples in window of a full cycle; T = sampling time = T /N; n indicates frequency component contained in complex phasor x(t). n = 1 indicates fundamental frequency component of the complex phasor. Assuming that the continuous time signal containing DC offset Ao , it can be represented by the equation: x(t) = Ao +

N 

An cos(nωt + n)

(2.1)

n=1

Its discrete time signal can be obtained by putting t = kT. X (k) = Ao +

N 

An cos(nωkT + n)

(2.2)

n=1

where k is the sample number and X(k) indicates complex phasor of the kth sample. Now, ω = 2π f = 2π = N2π . T T Putting this value of ω in Eq. 2.2, we get X (k) = Ao +

N  n=1

    2π kT + n An cos n N T

X (k) = Ao +

N  n=1

 An cos

2π kn + n N

(2.3)

 (2.4)

Equation 2.4 represents DFT of an input analog signal x(t). The effect of DC component and error in phasor estimation due to variation in fault inception angle (FIA) are discussed in the next section.

2.1.2 Fault Inception Angle and Power Flow Angle As EHV and UHV transmission lines are protected by distance relays, and majority of all distance relays are operated based on calculated impedance at the time of fault. The impedance depends not only on values of fault voltage or current, but it also depends on fault inception angle. The fault inception angle is the angle between fault voltage and fault current which mainly governs the value of impedance at the time of fault. Hence, for a fault at particular distance, if impedance has been set by considering only magnitudes of voltage and current, then at the time of fault due to

2.1 Issues in Numerical Distance Relays

23

fault inception angle, calculated impedance can be much higher than the impedance setting which results in under reaching problem. The magnitude of DC component depends on the fault inception angle α as given by the equation; −t

Ao = An sin( − α)e λ

(2.5)

where An indicates maximum magnitude of the signal,  = phase angle of the signal, α = fault inception angle and λ is the time constant of the system. The DC component will be maximal when the FIA is 0° and minimal when the FIA is 90°. During FIA = 0°, due to the error in the phasor estimation, relay can mal-operate. If the impedance locus is lying near the boundary of the characteristic of the relay, then it can result in under reaching problem. In the exactly same way, during the power swing and/or overloading condition, due to the angle separation between the generator and infinite busbar, the power flow angle could cause the locus of the impedance to enter in the zone of the relay, which results in mal-operation of the relay and cause overreaching problem of the relay.

2.1.3 Close-In Fault One of the serious problems in case of distance relaying is close-in fault due to which relay will not get enough polarizing voltage as specified in the datasheet of the relay, and it will fail to operate because of its lower sensitivity. It is frequent in Mho relay which is widely used for primary and backup protection. When the fault occurs on heavily loaded line, especially when PT is on the secondary side of the busbar, the origin may get excluded from the relay operating zone and relay may fail to operate. There are various techniques to avoid close-in fault. One technique is to use offset Mho characteristic to bring the origin inside its operating zone which can make the relay directionally insensitive. Another technique is cross-polarization in which polarization signal is derived from the healthy phase, but this technique can fail during three-phase fault.

2.1.4 Influence of Fault Resistance When fault involving ground occurs, the fault resistance (RF ) is given by R F = Rarc + RT + Rg . where Rarc Arc resistance which depends on fault MVA.

(2.6)

24

2 Transmission Line Protection: Issues and Research Needs Bus B

Bus A A

IA

B

IF RF

IB

Fig. 2.1 High resistance fault in transmission line

RT Rg

Tower footing resistance which depends on resistivity of the soil, its typical values are 0.5–50 . Ground resistance which depends on ground surface.

Figure 2.1 shows simple two-terminal transmission line between two buses. If fault occurs in the middle of transmission line (between bus A and B), the impedance seen by the relay can be given by   IB Z seen = d · Z L + R F · 1 + IA

(2.7)

where Z L is the unit line impedance and d indicates fault location. Also, I F = fault current = I A + I B . When HIF occurs, not only RF changes but magnitude and phase of fault current will also change depending on pre-fault power flow conditions. Thus due to pre-fault condition for L-G fault, if Z seen is higher than the set value in the distance relay, then under reach can happen. Hence, there is a need of adaptive distance relay which can change its characteristics according to magnitude of fault resistance dynamically during real-time operations.

2.1.5 Load Encroachment and Evaluation of Zone 3 Relay Settings The zone 3 relay, often considered a remote relay or breaker backup, is slower acting and monitors for faults outside the length of the line. This relay should not trip the breakers under typical overloading conditions. If zone 3 relay senses a fault in the immediate reach of the line and its zone 1 and zone 2 settings, it waits for 1–2 s to allow the primary line protection to act first. The length and configuration of some lines require higher apparent impedance setting. Hence, the zone 3 settings are designed with overload margins close to the long-term loadability limit of the line. In such situations, zone 3 relay could cause a breaker to trip in an extreme overload situation even though a fault does not exist. The final reports on blackout [2, 3] concluded that if the relay setting is sensitive to the load encroachment on the transmission system then it can result in cascade tripping and widespread blackout. While all of the relays operated according to their

2.1 Issues in Numerical Distance Relays

25

settings, some zone 3 relays (and few zone 2 relays acting like zone 3 relays) acted so quickly in response to line overload that they did not provide time for the electric system transients to settle. This is so-called overreach of the distance relay. There are two possible solutions to avoid this overreaching problem. Firstly, by proper coordination between primary and backup protection. Secondly, use of relay with quadrilateral characteristic which has better loadability limit. Hence, there is a need of adaptive distance relay which can change its characteristics according to calculations of fault condition dynamically during real-time operations. Heavy loads in the system can be mistaken as faults occurring in zone 3 as shown in Fig. 2.2. Such conditions lead to overreaching problems. The incorrect operation of distance relays due to large load encroachments was the root cause of several historical blackouts in India and the USA. In the past when the electromechanical relays were used, circular Mho characteristic was the only possible choice. Nowadays, variety of relay characteristics can be realized with the numerical relays. Figure 2.2 shows quadrilateral relay with the same reach settings as that of relay with Mho characteristics. It can be observed that Mho settings are more immune to heavy loads which ultimately results in mal-operation. However, with the use of quadrilateral settings, trip signal can be blocked during heavy load conditions. Hence, in order to make the relay more sensitive to high resistance faults and less sensitive to heavy loads, quadrilateral characteristic is employed in the present work of developed scheme as discussed in Chap. 3. Moreover, it was observed that for very heavy loading conditions the load impedance can encroach even quadrilateral settings as well in zone 3. Hence, the relays with quadrilateral characteristics can also mal-operate during load encroachments. The only possible solution is to develop a state-of-the-art system which must block the trip signal during such scenario. The more discussion on development of such scheme is detailed in Chap. 3.

Fig. 2.2 Comparison between mho and quadrilateral settings for load encroachment

26

2 Transmission Line Protection: Issues and Research Needs

2.1.6 Transient Condition and Implementation of Auto-Reclosure Almost 80% of the faults involving the ground are transient in nature. The transient faults are normally cleared in 1–2 s. Once the fault is occurred, the protective element will generate the trip signal to isolate the faulty system from healthy section. The generation of trip signal subsequently results in stoppage of power supply continuity. In order to provide uninterruptable power supply to the potential consumers, service continuity must be regained automatically as soon as transient fault is cleared. Thus, there should be proper discrimination between transient fault and permanent fault. However, transient conditions can exist in two cases. (i) Switching transient due to system reactance. (ii) Transient fault. Whenever fault or dynamic switching occurs on the transmission line, the magnitude of the current is very high and asymmetrical in nature. The time response of the fault current consists of transient part and steady-state component. The transient part consists of DC component and subharmonic components exponentially decaying in nature. The time response of the system will be oscillatory and unstable if the system is un-damped. The damping factor mainly depends on system reactance and series capacitance of the series compensated transmission line. Due to the presence of DC component superimposed on fundamental frequency component, there can be overreaching problem for zone 1 of distance relaying. Similarly, in case of transient fault which mainly die out in few cycles, there should be proper discrimination of transient fault and permanent fault in order to automatically reclose the circuit breaker. To mitigate these problems, digital distance relays and auto-reclosures use discrete Fourier transform (DFT). However, it provides more time delay and also suffers from sensitivity problems for processing the algorithm. Hence, proposed solution is to implement least square error (LSE) technique and fast Fourier transform (FFT) using decimation in time (DIT) or decimation in frequency (DIF) algorithm using which mathematical computations can be greatly reduced. It can be also analyzed using modified Fourier transform, wavelet transform, support vector machine, or artificial neural network to improve its performance.

2.1.7 Effect of Power Swing The fluctuations of voltage and current because of abrupt change in mechanical input to generators, switching of transmission line due to fault or sudden application/removal of load is called power swing. Due to power swing condition, the impedance locus falls under the zone 3 characteristics of the relay momentarily. It complicates the operation of the relay to discriminate between three-phase faults and power swings. The conventionally used load blinder technique works on the rate of

2.1 Issues in Numerical Distance Relays

27

change of impedance with respect to time as shown in Fig. 2.3 between two blinder A and blinder B. During fault, its magnitude will be more, whereas during power swing, its value will be less. However, proper determination of threshold limits for the same is the challenging task for the protection engineers. Thus, the swing impedance trajectory may enter the fault detection zone of distance relays as shown in Fig. 2.3 Fault and swing impedance trajectory. With stable swings, the swing impedance trajectory returns to the actual load impedance locus. Thus, all distance relays in the power system subjected to swings need to be securely blocked for the time the swing impedance remains within the distance relay characteristic in order not to disrupt the power system integrity. The swing impedance moving along its trajectory needs some time to travel through the two blinders as shown in Fig. 2.3. Its traveling speed is slow compared to the sudden impedance jump into the characteristic during faults occur. Traditional power swing detection is based on the measurement of time t that elapses as the traveling swing impedance trajectory enters and leaves two thresholds (circles or blinders). If the time, the swing impedance requires to pass through the two impedance set points is longer than set time t, the swing detector will block the distance relays. The common predictive method to determine loss of synchronism is the equal area criterion. It assumes that the power system behaves like a two-machine model where one area oscillates against the rest of the system. In reality, a power system is more complex and changes its parameters over time. The time t as criterion does not fully cover all possible situations. The set points are fixed and do not adapt to power system changes. Finding proper settings for traditional swing detectors is not simple and often requires comprehensive grid studies. If the study does not consider worstcase conditions, then the relays may lack security. When single pole auto-reclosure is applied, the swing detector may not assume symmetrical conditions during the auto-reclose dead time. A more comprehensive logic is required to cope with open pole conditions in the presence of power swings. Another challenge is the clearance of line faults during power swings. The swing detector blocks the ordinary distance processing under those conditions. Thus, the Blinder B

Fig. 2.3 Fault and swing impedance trajectory

X

Line

Blinder A ΔZ

Δt

Instant jump from load to Fault impedance Power swing impedance trajectory

R

Load prior to swing

28

2 Transmission Line Protection: Issues and Research Needs

distance relay itself requires extra provisions to cover faults during power swings. For all asymmetrical faults, the negative or zero-sequence current or voltage which are not affected by power swings, may be used for line fault calculation. In conclusion, new swing detectors and their associated distance relays must. • • • • •

Cover extremely fast swings frequencies. Block the trip signal during power swing. Generate the trip signal during potential faults. Clear all kinds of internal fault during power swings. Should be virtually setting free.

Moreover, the power swing detection scheme must remain stable for the lines equipped with compensation. The detained discussion on the same is carried out in Chaps. 4 and 5. However, the fundamentals of series compensation are presented in the next section.

2.1.8 Series Compensation in Transmission Line Due to the rapid development of industrial and commercial sectors, overall power demand increases at very rapid rate. Simultaneously, the installed capacity is very less as compared to demand which ultimately results in shortage of energy. In order to cope with this situation, one of the possible solutions is to introduce a flexible AC transmission (FACTs) compensating device in series with the transmission line. The compensating devices are of two types: fixed series compensation and variable compensation. The compensator adds the appropriate value of capacitor in series with the transmission line such that voltage profile will be improved. It also increases the power transfer capacity of the line under consideration. Moreover, it enhances stability of the system by ensuring proper balance of load in the system. It is frequently used in extra high voltage (EHV) and ultra-high voltage (UHV) lines. The power transmitted through the line between sending end and receiving end is given by, P1 =

VS · V R sin δ XL

(2.8)

where P1 power/phase, transmitted by the uncompensated transmission line from sending end to receiving end. V s sending-end voltage/phase. V R receiving-end voltage/phase. X L total reactance of transmission line. δ angle difference between V s and V R .

2.1 Issues in Numerical Distance Relays

29

Fig. 2.4 Series compensation in transmission line

In order to improve the power transfer capacity, if a series capacitor of X c is introduced in series with the line to be protected as shown in Fig. 2.4, then power transferred will be given by, P2 =

VS .VR sin δ X L − XC

(2.9)

Equation 2.9 gives the power transferred by the same transmission line after incorporation of series compensation in the transmission line. By taking a ratio of Eqs. 2.8 and 2.9, we get, XL 1 1 P2 = = = XC P1 X L − XC 1 − k 1 − XL

(2.10)

where k = XX CL indicates degree of compensation/compensation level of transmission line. In actual practice, k lies in the range of 0.1–0.7. If the value of k = 0.5 then, P2 1 =2 = P1 1 − 0.5

(2.11)

It reflects that by increasing compensation level by 50%, the power transfer capacity of the transmission line will get doubled. Its maximum value depends on long-term loadability limit of the line to be protected. The compensation level of the line can be varied between 10–70% in order to improve the power transfer capacity. However, by involving the series compensation to improve the power transfer capacity makes the protection of line more complicated. The distance relays may maloperate because of the reactance of compensated device at the time of fault. It also affects the tripping characteristics of the distance relays. During power swing and load encroachment problems, it also inserts nonlinearity in the waveforms of voltage and current. Thus, conventional protection philosophy may not operate properly which is a major challenging task for the protection engineers. All these requirements cannot be met by one single measurement approach. Only an adaptive machine learningbased protective approach may cope with these numerous and changing measurement environment.

30

2 Transmission Line Protection: Issues and Research Needs

2.2 Objectives of Research Undesirable operation of numerical relays may lead to catastrophic breakdown in the power grid. The digital numerical relays used for in-system operation nowadays must be adaptive in order to avoid the undesirable effects of conventional protective system as described in Sect. 2.1. As outlined in previous sections, there are numerous factors influencing the performance of numerical distance relays. As shown in Fig. 2.5, the power swing, load encroachment, and fault occurring in the transmission lines are the three main dynamic situations occurring in power grid. During power swing and load encroachments, impedance trajectory enters in the trip characteristic of impedance relays even without occurrence of fault, and hence, the relays will generate false tripping on such events. On the contrary, there are some fault situations during which the conventionally used numerical relay fails to operate. Majority of distance relay incorporates certain amount of arc resistance and reactance at the time of fault. However during the fault, if the resistance and reactance seen by the relay are more than the reach setting of the distance relays, then impedance trajectory will fall outside the trip region as shown in Fig. 2.6. It will ultimately lead to under reach operation of the relay. Hence, main objective of this research is to develop the adaptive quadrilateral distance relay which modifies its characteristics adaptively based on the variation in fault and system parameters. The developed relay must be fast and accurate as compared to existing schemes of protection to maintain power system stability and reliability. The protective scheme must be sensitive enough to critical parameters because of which currently adopted relaying scheme mal-operates. It makes modern relay capable enough to effectively settle down disturbances and communicate in better way with phasor measurement units (PMUs). The major factors that need to consider are high resistance fault, load encroachment problem, power swing, closein faults, saturation of instrument transformer and transient faults occurring in the transmission line. Fig. 2.5 Influence of dynamic situations on conventional distance relays

2.2 Objectives of Research

31

Fig. 2.6 Effect of fault impedance

Secondly, majority of blackouts reported are due to power swing scenario [1]. Conventionally used distance relays adopt power swing blocking (PSB) feature. These relays mainly operate on the rate of change of single operating quantity. However, these developed schemes fail to operate for the faults occurring during power swing. Recently, soft computing-based disturbance classifiers found many applications in the field of smart grid protections against power swing. Hence, there is always a scope of improvement for the development of powerful classifier algorithm which can properly discriminate between fault and swing scenario. The developed scheme must be sensitive for the symmetrical and asymmetrical faults for both compensated and uncompensated transmission lines. Apart from these, the transient fault occurring in the transmission line also needs critical attention as it severely affects the stability of the system. Presently used autoreclosure scheme provides fixed dead time for detection of fault. It also provides multi-shot operation in order to monitor the clearance of transient fault. The multishot operations are practically not feasible on heavily loaded lines. Additionally, the detection of transient fault is very much complicated task in series compensated transmission lines. Hence, auto-reclosing schemes are mainly applied to short transmission lines. With the advancement of synchro-phasor technology, the phasors can be monitored continuously in order to set the dead time of auto-reclosing system adaptively. It can also improve the dynamic stability of the system. Considering all above critical aspects, the main attributes of numerical relaying can be outlined as follows:

32

2 Transmission Line Protection: Issues and Research Needs

• It must be adaptive for fault impedance during wide variation in fault resistance, fault type, fault inception angle, power flow angle, and series compensation in transmission line. • The protective scheme should be sensitive to close-in fault, load encroachment, and saturation of instrument transformer. • It must be able to discriminate between fault and power swing conditions and detect the symmetrical faults during power swing for series compensated transmission line. • It should be also able to discriminate between transient and permanent faults with adaptive dead time control.

2.3 Research Plan Keeping in view the research gaps identified, whole research work can be divided into different sections as outlined in Fig. 2.7: Section 1

Fig. 2.7 Proposed scheme for numerical distance protection

2.3 Research Plan

33

• One of the fundamental requirements for the distance protection is fast and accurate phasor estimation of voltage and current phasors which can accurately remove harmonics and DC component from the fundamental frequency component. For detecting the fault and taking the adaptive corrective action, instantaneous phasor estimation is evolved by many researchers for fast and accurate protection trip, use of which is limited to transient fault analysis. Permanent fault and load encroachments are also critical issues to be considered for phasor estimation. Hence, fast and accurate phasor estimation which can eliminate the harmonic and DC component within subcycle time (within 20 ms) is the need for the protection engineers. • An innovative transmission line protection scheme which can detect close-in fault and high resistance fault can be designed. The developed scheme must be fast and accurate against variation in fault and system parameters. • The developed scheme must remain inoperative during load encroachment problems to maintain stability of power system. Section 2 • An innovative power swing detecting algorithm can be evolved for EHV/UHV long transmission line which can detect the symmetrical fault during power swing. • It should generate power swing blocking (PSB) signal during fast and slow stable power swing and trip signal for unstable power swing operation for the numerical relay with fast and accurate response. • The developed scheme should also remain stable for operation during series compensated transmission line. It must respond with highest accuracy and fast enough as compared to other existing algorithms. Section 3 • As 80% faults occurring in the transmission lines are transient in nature, the main motivation is to design fast and reliable adaptive, single pole auto-reclosure with optimization of reclose time depending on the power system disturbances. • The developed auto-reclosing scheme which can discriminate between transient fault and permanent fault simultaneously eliminating the research gaps outlined in the literature review for both compensated and uncompensated transmission lines. The existing gap provides tremendous opportunities for further development of intelligent and adaptive digital protection for transmission line, many of which are subjects of contemporary research. The detailed discussion on the work carried out in each phase has been outlined in subsequent chapters.

Chapter 3

Adaptive Numerical Distance Relaying Scheme

3.1 Introduction Impedance reach of numerical distance relay is severely affected by fault resistance (RF ), fault inception angle (FIA), fault type (FT), fault location (FL), power flow angle (PFA) and series compensation in transmission line as discussed in Chap. 2. This chapter presents a novel standalone adaptive distance protection algorithm for detection, classification and location of fault in presence of variable fault resistance. It is based on adaptive slope tracking method to detect and classify the fault in combination with modified Fourier filter algorithm for locating the fault. To realize the effectiveness of the proposed technique, simulations are performed in PSCAD [121] using multiple run facility and validation is carried out in MATLAB® [122] considering wide variation in power system disturbances. Due to adaptive settings of quadrilateral characteristics in accordance with variation in fault impedance, the proposed technique is 100% accurate for detection and classification of faults with error in fault location estimation to be within 1%. Moreover, the proposed technique provides significant improvement in response time and estimation of fault location as compared to existing distance relaying algorithms, which are the key attributes of multi-functional numerical relay. Many numerical distance relaying algorithms were suggested by researchers in the past for detection of high impedance faults (HIF). Quadrilateral relay characteristic is widely used for the same, which incorporates more fault impedance depending on its setting compared to Mho-type distance relay. If zone 1 fault involves resistance higher than the setting of quadrilateral characteristic, the relay deems this fault as of zone 2 or zone 3 and will issue delayed trip signal. A delayed tripping of circuit breaker during fault can increase stress on power system for longer time duration. The distance relay is more susceptible to under reach when HIF occurs in the zone 3. This research is mainly intended to completely eliminate the effect of any value of fault resistance by modifying the quadrilateral relay characteristic settings adaptively. Adaptive protection modifies the preferred protective response in order to © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 U. Patel et al., Futuristic Trends in Numerical Relaying for Transmission Line Protections, Energy Systems in Electrical Engineering, https://doi.org/10.1007/978-981-15-8465-7_3

35

36

3 Adaptive Numerical Distance Relaying Scheme

make it more attuned to prevailing power system conditions. In the proposed work, quadrilateral characteristic is adaptively set to improve the response time by mitigating the adverse effect of fault resistance, DC component, and pre-fault power flow condition. The prime reason for mal-operation of any numerical distance relaying algorithm is incorrect phasor estimation. The conventionally used discrete Fourier transform (DFT) fails during some dynamic situations. In order to avoid the same, modified fullcycle discrete Fourier transform (MFCDFT) [24] is used in the proposed work. The extensive literature survey on phasor estimation technique is presented in Sect. 1.4. The detailed discussion on the phasor estimation technique is outlined in Sect. 3.2. The system modeling for validation of the proposed algorithm has been outlined in Sect. 3.3. The proposed methodology for fault impedance compensation has been outlined in Sect. 3.4. The phasors of analog input signals are calculated by MFCDFT which is further used to estimate fault phase and line impedance using ground distance function (GDF). For the detection of fault, the magnitude of the ground distance function (K 0 ) is adaptively calculated based on real-time sequence components of the faulty phase voltage and currents. Nevertheless, the same is failing only during the high resistance faults. The high resistance faults are detected by using slope tracking method of the fault impedance trajectory and the comparison of the same with transmission line impedance trajectory which is to be protected using the knowledge of X/R ratio of transmission line parameters. In order to make the proposed methodology immune to load encroachment problems, zero-sequence current of line currents is used. The proposed method is highly accurate for detecting severe power system disturbances like high resistance fault and yields accurate phasor estimation even during instance of voltage zero crossing, close-in fault, and variation of other parameters like FT, FL, and FIA as compared to existing numerical distance relaying algorithms. The detailed discussion on applied phasor estimation technique is presented in the next section.

3.2 Phasor Estimation Techniques 3.2.1 Discrete Fourier Transform The most widely used phasor estimation technique is DFT which is used for extracting fundamental complex phasor from the original phasor and separates DC component and higher order harmonics. It can be performing the following operation. Let, x(t) is the sinusoidal time varying signal which is to be analyzed. Also, X(k) is the discrete time signal derived from x(t) by sampling it with sampling frequency F s with fundamental cycle time of T sec. Where, k indicates sample number. If N is number of sample in window of a full cycle then the sampling time can be given by T = T /N.

3.2 Phasor Estimation Techniques

37

If n indicates frequency component contained in complex phasor x(t) then n = 0 indicates fundamental frequency component of the complex phasor. If a faulted signal can contain up to N − 2 order harmonic, the continuous time signal containing DC offset Ao can be represented by the equation: x(t) = Ao +

(N −2) 

An cos(nwt + θ n)

(3.1)

(n=1)

Its discrete time signal can be obtained by putting t = kT. N −2 

X (k) = Ao +

An cos(nwkT + θ n)

(3.2)

n=1

Here, X(k) indicates complex phasor of the kth sample. By substituting w = 2π f = 2π = N2π in the Eq. 3.2, we get T T X (k) = Ao +

N −2  n=1

    2π kT + θ n An cos n N T

X (k) = Ao +

N −2 

 An cos

n=1

2π kn + θn N

(3.3)

 (3.4)

Equation 3.4 represents DFT of an input analog signal x(t). The fundamental complex phasor containing real part X r (k) and imaginary part X i (k) can be obtained by   k+N −1 2π n 2  X r1 (k) = X (n) cos N n=k N

(3.5)

For considering previous (N − 1) samples, both lower and upper limit in above equation can be subtracted by N − 1. X r1 (k) =

2 N

2 X r1 (k) = N

k+N −1−(N  −1) n=k−(N −1) k 

 X (n) cos 

2π n N

2π n X (n) cos N n=k−N +1

Similarly, the imaginary component can be given by



 (3.6)

38

3 Adaptive Numerical Distance Relaying Scheme

X i1 (k) =

  k 2π n −2  X (n) sin N n=k−N +1 N

(3.7)

Equations 3.6 and 3.7 represent real and imaginary components of fundamental complex phasor, respectively, using full cycle DFT. By using above basic equations of full cycle DFT, phasor estimation of voltage and current waveform can be performed for the fault analysis and impedance can be calculated. The main limitation of this algorithm is removal of decaying DC component during the zero crossing of the faulty signal. The DC component will be maximal when the FIA is 0° and minimal with FIA of 90°. During FIA = 0°, due to the error in the phasor estimation, relay can mal-operate, if the impedance locus is lying near the boundary of the characteristics of the relay. So, these limitations can be easily overcome by modified full-cycle DFT (MFCDFT) [24] by performing accurate phasor estimation during extreme fault conditions.

3.2.2 Modified Full-Cycle Discrete Fourier Transform The main limitation of conventional DFT is improper phasor estimation during extreme fault conditions due to unknown time constant of the exponentially decaying DC component in the faulty signal which can be easily identified by using MFCDFT. In modified algorithm, the faulty signal is filtered by using low pass butter worth filter for suppressing higher order harmonics. The output of low pass filter contains additional DC component. Fortunately, the time constant of the same depends on the filter parameters and hence it is known and can be removed easily from the original signal which results in accurate phasor estimation. Assuming that during the fault or transient condition, exponentially DC component will be superimposed on the fundamental component having a magnitude D and time constant τ. The magnitude of exponentially decaying component depends on transmission line parameters. If we consider a lumped transmission line having resistance R and inductance L, the time constant will be given by τ = LR . −t

X dc = D · e τ Hence, X (t) can be modified as X (t) =

N −2 

−t

An cos(nwt + θ n) + D · e τ

(3.8)

n=1

This time varying signal can be filtered by a first-order analog low pass filter having characteristic equation S + τ11 = 0 where τ1 > 0. Normally, the time constant of analog filter (say D1 ) will be known and hence X (t) can be modified as

3.2 Phasor Estimation Techniques

X (t) =

N −2 

39 −t

−t

An cos(nwt + θ n) + D · e τ + D1 · e τ1

(3.9)

n=1

Its discrete time signal can be obtained by putting t = kT. X (k) =

N −2 

An cos(nwkT + θ n) + D · e

−kT τ

+ D1 · e

−kT τ1

(3.10)

n=1

The real part of X (k) can be given by Eq. 3.6 as X r (k) = X r1 (k) +

2 N

k  

D·e

−kT τ

+ D1 · e

−kT τ1



 cos

n=k−N +1

2π n N

 (3.11)

Similarly, the imaginary part X i (k) at any kth instance will be given by, X i (k) = X i1 (k) +

2 N

k  

D·e

−kT τ

+ D1 · e

−kT τ1



 sin

n=k−N +1

2π n N

 (3.12)

For implementing phasor estimation of fundamental complex phasors, Eqs. 3.11 and 3.12 can be modified as, 2 X r1 (k) = X r (k) − N 2 X i1 (k) = X i (k) − N

k  

D·e

−kT τ

D·e

−kT τ

+ D1 · e

−kT τ1



n=k−N +1 k   n=k−N +1

+ D1 · e

−kT τ1





2π n cos N 

2π n sin N

 (3.13)  (3.14)

So, for estimating the instantaneous value of a signal at any kth instance, previous N samples are required. The value of D and D1 can be found using above Eq. 3.11 by calculating the next successive three samples,X r (k + 1), X r (k + 2) and X r (k + 3). Assuming unknown time constant of the decaying DC component as X, and known −T −T time constant of the analog filter as K 1 , where X = e τ and K 1 = e τ 1 

 N  N

X r (k + 1) − X r (k)  2 cos 2π N



  = D X X N − 1 + D1 K 1 K 1N − 1 = A1

X r (k + 2) − X E (k + 1)  2 cos 4π N X r (k + 3) − X r (k + 2)  2 cos 6π N



(3.15)

  = D X 2 X N − 1 + D1 K 12 K 1N − 1 = A2 (3.16)   = D X 3 X N − 1 + D1 K 13 K 1N − 1 = A3 (3.17)

40

3 Adaptive Numerical Distance Relaying Scheme

By using Eq. 3.15, 3.16, and 3.17  (A2−A1)K 1 = D X X N − 1 (X − K 1 )

(3.18)

 (A3−A2)K 1 = D X 2 X N − 1 (X − K 1 )

(3.19)

The value of X can be obtained by dividing Eqs. 3.19 and 3.18, X=

(A3−A2) (A2−A1)

(3.20)

After finding the value of X from the above equation, D and D1 can be obtained from Eqs. 3.18 and 3.15, respectively. So, the time constant of the exponentially decaying DC component can be calculated from the equation X =e ∴ ∴

−T τ

−T τ −T τ= log X

log X =

The negative sign indicates the decaying nature of the DC component. So, it can be ignored.X r 1 (k)andX i1 (k) can be calculated from Eqs. 3.13 and 3.14 by eliminating the effect of DC component. During phasor estimation using MFCDFT, it has been found that it is unable to operate during certain extreme power system disturbances like variation in fault inception and power flow angle. Therefore, for monitoring these abnormalities, some further modifications are required in the above algorithm. Apart from this, the MFCDFT algorithm outlined in the previous section can be used for phasor estimation of only single quantity, but during the relaying decision phasor estimation of all the quantities should be performed simultaneously. Here, in this investigation, the magnitude and phasor angle of fundamental complex phasors like three-phase voltages V a , V b , and V c and three-phase currents I a , I b , and I c are estimated simultaneously by incorporating few modifications in the above algorithm. Also, the effect of DC component is profound in first cycle after fault inception and plays a key role in accurately estimating magnitude and phase angle and it must be given enough importance during relaying decision in order to accurately estimating the parameters of any signal. Thus, in the proposed algorithm, in order to consider the DC component present in the fundamental complex phasor after the inception of the fault for estimating the phasor, modified full-cycle algorithm is used for all the quantity. In order to estimate all the voltages and currents simultaneously at any instance, the phasor estimation (real and imaginary part) for jth signal at any kth instance can be obtained by the following equation.

3.2 Phasor Estimation Techniques

X r1 ( j, k) = X r ( j, k) −

2 N

41 k  

D( j) · e

−kT τ ( j)

+ D1 ( j) · e

−kT τ1



 cos

n=k−N +1

2π n N



(3.21) 2 X i1 ( j, k) = X i ( j, k) − N

k  

D( j) · e

−kT τ ( j)

+ D1 ( j) · e

−kT τ1

n=k−N +1





2π n sin N



(3.22) The magnitude and phase angle can be calculated by using the equation, |X ( j, k)| =



(X r1 ( j, k))2 + X i1 ( j, k)2

∠(X ( j, k)) = tan

−1



X i1 ( j, k) X r 1 ( j, k)

(3.23)

 (3.24)

Above equation gives magnitude and phase angle of jth signal at Kth instance. It is to be noted that Eq. 3.23 will result into the maximum magnitude of the fundamental complex phasor. Normally, all the relaying decisions are taken based on root mean square (RMS) value and hence to find out the RMS value of the fundamental complex phasor, M( j, k) =

|X ( j, k)| √ 2

(3.25)

Here, the cosine Fourier transform has been performed for estimating the phasor magnitude and hence there will be phase displacement of 90° and delay of three samples is also to be taken into account. Hence, it is mandatory to compensate the said error in estimation of phase angle of any quantity. If we sample a 50 Hz fundamental signal with a sampling frequency of 12 kHz, then total error in the phase will be equal to 90° plus phase angle corresponding to three samples (4.5°). To compensate for this error in estimation, the actual phase angle of the fundamental complex phasor can be given by,    360◦ ◦ θ ( j, k) = ∠(X ( j, k)) − 90 + 3 N

(3.26)

Equations 3.25 and 3.26 result in RMS magnitude and phase angle of fundamental complex phasor after removal of all the harmonics and decaying DC component and hence it can be used for relaying applications. All phase voltages and currents can be found by using these equations which are very much helpful for the further digital signal processing for adaptively taking the relaying decision for correct operation of the relay. The detailed comparative analysis has been performed in Sect. 3.5.2. The

42

3 Adaptive Numerical Distance Relaying Scheme

system modeling for validation of developed algorithm is demonstrated in the next section followed by the proposed methodology and comparative result analysis.

3.3 System Modeling for Proposed Relaying Scheme The single line diagram of a 220 kV power system network is shown in Fig. 3.1. There are two generators (G1 and G2) connected by the first section of 120 km parallel transmission lines (L1 and L2) between sending-end bus (SEB) and middleend bus (MEB) followed by the second section of 100 km single transmission line (L3) between MEB and receiving-end bus (REB). Generators (G1 and G2) are modeled as an equivalent dynamic source representing a multi-machine system. The voltage and current signals of bus PTs at SEB and line-1 CTs are sampled at a frequency of 4 kHz (80 samples/cycle) and applied to the numerical distance relay R (Fig. 3.1). Bergeron line model with distributed parameters is used for modeling of transmission line. The performance of numerical distance relay R is tested for all protective zones of power system network. Zone 1 covers 100 km of transmission line L1. Zone 2 covers remaining part of L1 plus 50% length of L2 or L3, and zone 3 covers the remaining portion of L2 or L3. The proposed technique is validated for different faults such as L-G, LL, LL-G, LLL-G (10 types) at several locations in the modeled power system network to cover the effect of close-in fault, in-zone fault, and out of zone fault. Moreover, faults are simulated with different fault inception angle (0°−315°), power flow angle (5°, 10°, 15°, 25°, 35°, 45°), and varying fault resistance (0.01 to 200 ). The load in power system is varied by adjusting the variable load angle between the two generators. Large numbers of simulation cases are generated using multi-run facility available in PSCAD/EMTDC. The phasor values of voltage and current signals are successfully estimated using modified full-cycle discrete Fourier transform (MFCDFT) and

Fig. 3.1 Single line diagram of power system

3.3 System Modeling for Proposed Relaying Scheme

43

applied to adaptive slope tracking algorithm for HIF detection. The parameters of generators, transformers, and transmission lines are as mentioned below. The detailed discussion on the proposed methodology is outlined in the next section. Generator data (G1 and G2) 615 MVA, 13.8 kV, 50 Hz Inertia constant (H) = 4 MWs/MVA Positive-sequence impedance = 0.871 + j9.96  Phase angle = 0° (G1) and phase angle = variable (G2) X d = 1.81 pu, X d  = 0.3pu, X d  = 0.23 pu, T do  = 8 s, T do  = 0.03 s, X q = 1.76 pu, X q  = 0.25 pu, T  = 0.03 s, Ra = 0.003 pu, X p (Potier reactance) = 0.15 pu. Transmission line data Line length: L1 & L2 = 120 km, system voltage = 220 kV Positive-sequence impedance = 0.0297 + j0.332 /km Zero-sequence impedance = 0.162 + j1.24 /km Positive-sequence capacitance = 12.99 nF/km Zero-sequence capacitance = 8.5 nF/km. Transformer data 650 MVA, 13.8 kV/220 kV, DYng, 50 Hz three-phase transformer with leakage reactance of 12%.

3.4 Proposed Methodology for Transmission Line Protection Figure 3.2 represents the proposed algorithm which consists of three main steps: (i) Phasor estimation using MFCDFT. (ii) Impedance reach calculation using slope tracking method. (iii) Adaptive setting of quadrilateral characteristic for determination of relay decisions.

3.4.1 Phasor Estimation Using MFCDFT In conventional numerical relay, DFT is used for phasor estimation to extract fundamental component from the original complex phasor. The main limitation of DFT algorithm is inaccurate removal of decaying DC component and higher order harmonics from the original complex phasors. Hence, to overcome the said error for accurate phasor estimation during extreme fault conditions, MFCDFT is realized as outlined in Sect. 3.2. The fault and abnormal conditions are discriminated

44

3 Adaptive Numerical Distance Relaying Scheme

Fig. 3.2 Proposed methodology for transmission line protection

with the help of fault detection algorithm based on MFCDFT followed by low pass Butterworth filter to remove harmonic and aliasing effects. A signal compensation for accurate phasor estimation is employed in account of three more samples. Figure 3.2 shows the proposed methodology for fault detection and classification algorithm. The system modeling outlined in Sect. 3.3 is simulated using PSCAD software package in order to generate various faults. The three-phase voltages and currents are captured with a sampling frequency of 12 kHz. The phasor estimation of all analog signals is performed simultaneously using the proposed MFCDFT

3.4 Proposed Methodology for Transmission Line Protection

45

algorithm. The estimated phasors are used to calculate fault phase and line impedance as outlined in the next section.

3.4.2 Impedance Reach Determination Once the accurate phasors for bus voltages V a ,V b ,V c and line currents I a ,I b ,I c are obtained by MFCDFT, sequence components are calculated from the phasor voltages and currents using the equation ⎤ ⎡ ⎤ ⎤⎡ ⎤ ⎤⎡ ⎤ ⎡ ⎡ I1 V1 Va Ia 1 a a2 1 a a2 ⎣ V2 ⎦ = 1 ⎣ 1 a 2 a ⎦⎣ Vb ⎦ and ⎣ I2 ⎦ = 1 ⎣ 1 a 2 a ⎦⎣ Ib ⎦, 3 3 V0 Vc I0 Ic 1 1 1 1 1 1 ⎡

(3.27)

where a = −0.5 + 0.866i. Normally, phase impedances referred to ground fault can be calculated using the equation Zp =

Vp Ip + K0 I0

(3.28)

where K 0 is ground distance function. In conventional relay, K 0 is assumed between 1.5 and 3.5, i.e., a constant value, which depends on the fault resistance, tower footing resistance and soil resistivity. A comparison of five methods of compensation for the ground distance function (GDF) and assessment of their effect on the resistive reach in quadrilateral characteristics is given by Sorrentino [64] and it is found that K 0 is calculated based on the positive- and zero-sequence impedances using, K0 =

Z L0 − Z L+ , 3Z L+

(3.29)

This method of compensation gives the ideal results during impedance calculations. But during the faulty conditions, it has been found that the magnitude of positive- and zero-sequence impedances changes according to the type of disturbances available in the power system network. In the proposed work, the positiveand zero-sequence impedances are adaptively calculated from the obtained values of voltage and current at the time of fault and its sequence component calculations. ZL+ = positive-sequence impedance = VI 11 and Z L0 = zero-sequence impedance = VI 00 . The developed scheme is also examined for the load encroachment problems of proper discrimination between heavy load and fault scenarios. In order to prevent the same, zero-sequence components of fault currents have been used for proper discrimination. Whenever zero-sequence current I 0 is more than 0.1 pu of maximum

46

3 Adaptive Numerical Distance Relaying Scheme

load current (as shown in Fig. 3.2), HIF is detected. During HIF, impedance reach has been determined from estimated sequence components and GDF. The calculated fault impedance comprises line impedance and fault resistance. In case of high resistance ground fault, the estimated fault impedance trajectory can fall outside the tripping region of the distance relay and result in delayed tripping or under reach operation of the relay. This can be overcome by estimating fault resistance using slope tracking method as discussed in Sect. 3.4.4. As one cannot modify the fault impedance seen by the relay, but it is very easy to modify the characteristics of relay itself. Hence, fault resistance estimated using slope tracking method is further used to modify the relay settings for taking protective actions. During load encroachments, the zero-sequence component remains within 0.1 pu of maximum load currents, which ultimately blocks the trip signal. Hence, even though the impedance locus falls inside the quadrilateral characteristics, the proposed methodology blocks the trip signal indicating stability and reliability of the protective scheme. The value of line impedances referred to phase fault can be calculated using the following equations: Z ab =

Va − Vb ; Ia − Ib

Z bc =

Vb − Vc ; Ib − Ic

Z ca =

Vc − Va Ic − Ia

(3.30)

If any of the complex phasors of phase impedances or line impedances fall within quadrilateral characteristic, then trip signal is generated. From the knowledge of phase or line impedance locus entering into the trip region, type of fault generated in the power system can be determined. The operating time can be calculated by considering the sample number as time stamp using the following equation: Fault time =

(K · T ) , f0

(3.31)

where K is sample number, T is sampling time, and f 0 is fundamental frequency. In the next sections, setting calculations of quadrilateral relay is explained in detail followed by slope tracking method for adaptive distance protection.

3.4.3 Relay Settings for Protection Zones Proper understanding of transmission line parameters and protection issues forms the basis for arriving at the right relay settings [123]. This section contains the settings of quadrilateral relay by using the line and system parameters for detection of high resistance fault for all zones followed by adaptive slope tracking method to detect and classify the faults. Backup protection is provided in distance protection using stepped distance characteristics. Zone-1 or the high-speed zone is set to trip without any intentional time delay and provides primary protection for the line section to be protected, which can be adjusted

3.4 Proposed Methodology for Transmission Line Protection

47

to reach 80–90% of the line length. Figure 3.3a shows setting of quadrilateral relay for different zones of protection. Zone-1 relay setting is done for initial 100 km of line L 1 for the system modeling as shown in Sect. 3.2. According to the guidelines published by subcommittee on relay/protection task force for power system analysis under contingency [123] for detecting the high resistance fault, minimum 15  arc resistance should be considered during setting of relay. In order to incorporate the effects of close-in fault, negative restraint angle (NRA) is set to 115° (line AB) and directional angle (DA) for distance protection is adjusted to −15° (line AD). As line Fig. 3.3 a Stepped distance quadrilateral characteristic for all zones, b control circuit diagram

(a) (+)

D Z2

861

Z3

Z1 T1

Auxiliary 86 Relay

T2

Timer T

(-)

(b)

862

Auxiliary Relay Contact

Auxiliary Switch Tripping Coil of CB

48

3 Adaptive Numerical Distance Relaying Scheme

impedance considered is (0.0297 + j0.332) /km, hence, the total line impedance for 100 km will be (2.97 + j33.2)  for Zone-1 protection. Here, considering X/R ratio of 11 as per guidelines in [26], line impedance can be rounded to (3 + j33) . As fault current is limited by source impedance, line impedance and fault impedance; to calculate total reactance, source impedance (0.871 + j9.96)  must be considered and hence maximum reactance during the fault will be 43  and hence line BC is drawn at a height of 43 units from the origin A. To incorporate fault resistance (RF ) will be Z T = (18 + j43) , of 15  with line resistance (RL ) of 3 , total impedance ◦ = 67.28 . Hence, considering AF = 18, line which yields line angle φ = tan−1 43 18 CD is drawn at an angle of φ with x-axis. Zone-2 is used to provide high-speed protection for the remainder of the line and also serves as backup protection for 50% section of an adjoining line. The zone 2 relays have to be time delayed to coordinate with relays at the remote bus with typical time delays of around 0.1–0.5 s. Lines AB and AD are drawn in exactly the same manner as that of Zone-1. Zone-2 covers in total 170 km from the relay location with total line resistance of 5 . By considering X/R ratio of 11, the line reactance will be 55 . By incorporating fault resistance and source impedance, total line impedance for Zone-2 will become (20 + j65) . Hence, PQ is drawn at a height of 65 units from the origin A and considering A-G = 20, line QJ is drawn at the line angle of φ with x-axis. The zone 3, designated as Zone-3, is used to provide remote backup to the zone 1 and 2 of adjacent line sections when a relay or breaker fails to clear the fault locally. The usual practice is to extend its reach beyond the end of the largest adjoining line section or more than double the line section to be protected. The zone 3 operation is usually delayed by about 0.3–2.0 s. The zone 3 reach setting is a more complex problem. It has been observed that the Zone-3 unit trips due to load encroachment problems and thereby leads to the cascade tripping of the power system. The zone 3 setting must be blocked during extreme loading conditions. In the proposed work, this blocking is achieved by monitoring the zero-sequence component of fault current. The zone 3 relay setting is done for total 220 km from the relay location. By considering fault resistance (15 ) and source impedance (0.871 + j9.96) , total impedance will become (22 + j87) . Hence, line ST is drawn at a height of 87 units from the origin A. Further considering AH = 22, line TU is drawn at line angle of φ with x-axis (Fig. 3.3a). It results in complete quadrilateral characteristic considering close-in fault with directional sensitivity for all zones of protection. The control circuit of distance protection is shown in Fig. 3.3b which includes contacts of directional element (D), three zone relay units (Z 1 , Z 2 , and Z 3 ), auxiliary relay (86), and a timer (T). As shown in Fig. 3.3b whenever any fault occurs within the reach of Zone-1, closing of contact Z 1 and directional contact D will energize auxiliary relay and trip the circuit breaker (CB) immediately. The seal-in contact (861 ) of auxiliary relay provide hold-on path to trip coil of CB. Zone-2 covers 50% of next line section. Whenever, the fault impedance lies within the reach of Zone2 but outside the reach of Zone-1, both relay units of Zone-2 and Zone-3 operate because Zone-3 relay unit cover the complete region with largest reach. Thus, for zone 2 fault, contacts Z 2 and Z 3 operate immediately. Whenever, the directional

3.4 Proposed Methodology for Transmission Line Protection

49

element (D) and Z 3 operates for zone 2 fault, the timer (T 1 ) is energized. After the completion of definite time, the closing of contact T 1 energizes the trip coil of CB. If the fault impedance is measured outside Zone-2, i.e., in Zone-3 region, after a delayed time, the contact T 2 closes and subsequent tripping is issued. It is to be noted that both time delays (T 1 and T 2 ) are individually adjustable. The next section demonstrates implementation of adaptive slope tracking method using the relays settings as discussed in this section.

3.4.4 Adaptive Slope Tracking Method Figure 3.4 shows impedance trajectory Z T = (RT + jX T ) during high impedance fault

Fig. 3.4 Impedance trajectory ZT during variable fault resistance RF for different power flow conditions between SEB and MEB such as a no power transfer; b exporting power flow; c importing power flow; and d adaptive characteristic during variable power flow

50

3 Adaptive Numerical Distance Relaying Scheme

for different power flow conditions. It is observed that during solid L-G fault (RF = 0 ), the apparent impedance Z T is located along the segment described by the line impedance Z Line , while varying RF without power transfer between SEB and MEB represents a straight line that increases depending on resistive reach as shown in Fig. 3.4a. However, when both RF and pre-fault power conditions are considered, the locus of the apparent impedance Z T follows a slight curvature like downward concave when exporting power from SEB (Fig. 3.4b) and upward concave when importing power conditions (Fig. 3.4c). Exporting power from SEB results decrease of apparent impedance and importing power results increase in apparent impedance seen by the relay. It can ultimately lead to mal-coordination between distance protection zones. As shown in Fig. 3.4d, Z a indicates fault impedance without considering high resistance fault. As shown in Fig. 3.4d, Z a indicates fault impedance without considering high resistance fault. In this case, fault impedance will be consisting of only line impedance and hence using quadrilateral characteristic, relay will be able to identify the fault. Whereas during high resistance fault, total fault impedance comprises of arc resistance, tower footing resistance, and resistance due to soil resistivity. Hence, as shown in Fig. 3.4d, the fault impedance locus will move out of the quadrilateral characteristic and hence relay will fail to detect such type of abnormality and hence under reaching can happen. In order to prevent such breakdown of the relay, during the high resistance fault, line impedance locus can be traced by taking into consideration X/R ratio of the line. During high resistance fault, Total fault impedance is given by Z T = RT + jX T . Actual line reactance can be traced to XL ∼ = X T. Actual line resistance can be traced to RL = X L tan º. Where tan º = XRLL = X/R1ratio . Hence, actual line impedance during the fault can be given by Z L = RL + jX L and fault impedance can be given by, RF = RT − RL

(3.32)

This derived value of fault impedance will result in approximate value. During solid L-G fault, due to the absence of resistance in faulty path, the quadrilateral relay will easily detect the fault. Whereas during high resistance fault, the fault impedance locus will move out of the quadrilateral characteristic and result in mal-operation of the relay. In order to prevent such breakdown of the relay during the high resistance fault, line impedance locus can be traced by considering zero-sequence component of the fault current and X/R ratio of transmission line. In normal conditions, zerosequence component of the fault current will be negligible and when ground fault occurs, it will rise abruptly. Here, in this investigation, a threshold value of 0.1 pu of maximum load current is considered for zero-sequence current. If zero-sequence current is greater than this threshold value, it indicates that ground fault has occurred. The power flow will always be exported when the fault occurs in forward direction of

3.4 Proposed Methodology for Transmission Line Protection

51

the relay. During high resistance fault, the approximate value of fault resistance can be calculated using Eq. 3.32. The scheme developed using approximate value of fault resistance detects and classifies the faults accurately, but the error is fault location estimation increases. Hence, in order to completely eliminate the fault resistance, the exact value of fault resistance can be calculated. The actual phase impedance without involving fault resistance and adaptive setting of quadrilateral characteristic as shown in Fig. 3.4d are done as follows. 1. Total fault impedance is given by Z T = RT + jX T . 2. Fault resistance can be narrated from the knowledge of X/R ratio (λ) of transmission line to be protected and sequence distribution factor ξ . From Fig. 3.4d, it can be written that Ra = RT − RF and X a = X T + X F The ratio of actual reactance X a to actual resistance Ra can be given by ⇒

Xa d · Xu XT + XF = =λ= Ra d · Ru RT − RF

where d indicates fault location, X u and Ru represents unit reactance and unit resistance of the transmission line, respectively. Simplifying, ⇒

λRT − λRF = X T + X F



λRF = λRT − X T − X F λRT − X T − X F RF = λ



(3.33)

For the transmission line without shunt or series compensation, the fault reactance can be derived from total fault reactance using sequence distribution factor as outlined by Halabi [18] as X F = (1 − ξ )X T

(3.34)

where ζ is the sequence distribution factor which is given by  ξ = 2C1 + C1

Z0 + Z A0 Z1

 (3.35)

Here, Z 0 and Z 1 are zero- and positive-sequence components of line impedances, respectively. Whereas,Z A0 is the zero-sequence impedance of generator G1 and Z B0 is the zero-sequence impedance of generator G2 . Also, C 1 is the source sequence distribution factor which can be derived using source sequence components, C1 =

Z B0 + (1 − d)Z 0 Z B0 + Z A0 + Z 0

(3.36)

52

3 Adaptive Numerical Distance Relaying Scheme

Substituting, the Eq. 3.34 in Eq. 3.33, RF =

λRT − 2X T + ξ X T λ

(3.37)

From Eq. 3.37, the fault resistance RF can be derived using total resistance, total reactance, X/R ratio (λ) and sequence distribution factor (ξ ). 1. 2. 3. 4. 5. 6. 7. 8.

Actual line resistance can be traced to Ra = RT – RF . Actual line reactance can be calculated by X a = Ra · λ. Change in fault reactance can be traced to X F = X T – X a . Total actual line impedance Z a = Ra + jX a . Absolute value of actual impedance Z am = abs(Z a ). Absolute magnitude of unit impedance Z u = abs(Z pu ). Fault location can be derived as ratio of Z am to Z u . Error in estimation of fault location can be given by % Error =

Estimated Location − Actual Location · 100%. Actual Location

(3.38)

Settings of quadrilateral characteristics can be modified after calculating the magnitudes of RF and X F with the vertices of A, B1, C1, D1 as shown in Fig. 3.4d. The fault resistance RF and fault reactance X F can be applied as correction factors in coordinates of C1 and D1. In order to sense high resistance close-in faults properly, the correction factor of only X F is applied to the coordinates of B1, which also maintains directional sensitivity of the relay. Thus, adaptive settings of quadrilateral characteristics depend on fault impedance, power flow conditions, and GDF. The estimated fault impedance locus is mapped in impedance plane of relay and if it falls within the characteristics, then the proposed numerical relaying technique issues the trip signal. Moreover, the proposed technique can also calculate fault instance by considering sample number as its time stamp. Fault classifier module is also designed to identify the type of fault occurred in power system. The proposed method is highly accurate for detecting severe power system disturbances like high resistance fault and yields accurate phasor estimation even during extreme fault conditions like voltage zero crossing (maximum DC component), close-in fault (minimum polarizing voltage), and variation of other parameters like fault types, fault locations, and fault inception angle. In order to show the effectiveness of the proposed algorithm, the results are compared with existing methods, which are implemented with the same system configuration and relay characteristics as outlined in the next section.

3.5 Validation of Proposed Technique

53

Fig. 3.5 Phasor estimation for fault current using DFT and MFCDFT during L-G fault applied at 0.1 s (400 samples) with FIA = 0°

3.5 Validation of Proposed Technique 3.5.1 Results of Phasor Estimation The results of phasor estimation by realizing algorithms of DFT and the proposed MFCDFT as mentioned in Sect. 3.2 are shown in Fig. 3.5. The results are obtained by simulating the system modeling outlined in Sect. 3.3 in PSCAD software for L-G fault applied at 0.1 s (400 samples). For phasor estimation at any kth sample, pre-fault 80 samples and post-fault 3 samples are considered. Thus, a sliding window of 83 samples is used for phasor estimation of each sample. For the waveform as shown in Fig. 3.5, the fault is applied at kth sample (400) and if it is detected on k + n sample (453), then n = 53 is the number of samples required to detect the fault (fault detection time = 13.25 ms). The results indicate that the proposed technique performs phasor estimation of faulted current in a better manner as compared to conventional DFT by removing the effect of DC component and FIA. Filtering and signal compensation in magnitude and phase impedance give improved results for each type of fault and various power system disturbances.

3.5.2 Performance Evaluation of the Proposed Algorithm In order to validate the performance and reach setting of the proposed numerical relaying technique, the results are obtained by simulating different kinds of faults with varying power system operating conditions as outlined in Sect. 3.3. Table 3.1 shows the performance of the proposed numerical algorithm in terms of operating

Fault applied

A-G

A-G

A-G

B-G

C-G

A-G

A-G

A-G

A-G

A-G

A-G

2

3

4

5

6

7

8

9

10

11

12

A-G

C-G

A-G

13

14

15

Faults in zone 1

A-G

1

Close-in faults

Case No.

30

10

5

Fault location d (km)

– 5

– 10

– 5

– 15

– 5

– 15

– 5

δ (◦52)

90

0

0

45

90

0

45

90

0

45

90

0

45

90

0

FIA (◦52)

40

20

0.01

20

10

0.01

20

10

0.01

20

10

0.01

20

10

0.01

R F ,

NOP

NOP

17.1

6.01

5.94

5.82

NOP

4.00

5.18

NOP

4.70

5.40

NOP

4.50

5.20

Conv. DFT

11.90

11.50

6.20

4.49

4.82

5.53

4.49

4.83

5.51

4.21

4.54

5.31

4.21

4.52

5.30

Proposed technique

Operation time (ms)

14.4

13.0

12.4

3.4

7.6

5.8

3.2

7.6

5.6

3.4

7.6

5.8

3.2

7.6

5.6

Phase comp.[114]

20

18

9

18

16

9

18

16

8

18

15

12

18

14

13

Wavelet and LDA [52]

Table 3.1 Comparison of the proposed technique with existing schemes for variation in fault context

22.80

21.80

14.22

20.34

11.31

10.62

20.14

10.61

9.23



12.01

9.92



12.01

9.23

ANFIS [65]

38.92

19.14

0.55

19.65

9.75

0.65

19.80

9.82

0.67

19.65

9.75

0.68

19.70

9.81

0.70

Estimated RF ()

A-G

C-G

A-G

A-G

A-G

A-G

A-G

A-G

A-G

C-G

B-G

A-G

A-G

A-G

A-G

Fault classified

30.219

30.205

30.192

10.065

10.066

10.068

10.061

10.064

10.065

5.040

5.024

5.026

5.037

5.038

5.031

Estimated fault location d  (km)

(continued)

0.73

0.68

0.64

0.65

0.66

0.68

0.61

0.64

0.65

0.80

0.48

0.52

0.74

0.76

0.62

d−d  · 100 d

% Error =

54 3 Adaptive Numerical Distance Relaying Scheme

AB-G

A-G

B-G

C-G

BC-G

A-G

A-G

A-G

AB-G

ABC-G

A-G

A-G

A-G

CA-G

16

17

18

19

20

21

22

23

24

25

26

27

28

29

99

99

90

90

65

65

65

65

65

50

50

30

Fault location d (km)

A-G

B-G

C-G

A-G

B-G

30

31

32

33

34

200

180

150

150

110

Faults in zone 2 and zone 3

Fault applied

Case No.

Table 3.1 (continued)

– 5

– 5

– 5

– 5

– 5

– 5

– 15

– 10

– 5

– 5

– 5

– 15

– 10

– 5

– 5

– 15

– 10

– 5

– 5

δ (◦52)

0

0

0

0

0

0

90

0

0

0

0

90

0

0

0

90

0

90

0

FIA (◦52)

0.01

20

20

0.01

0.01

0.01

50

20

0.01

0.01

0.01

50

20

0.01

0.01

40

20

0.01

0.01

R F ,











14.5

NOP

NOP

18.0

10.7

12.6

NOP

NOP

14.6

11.6

NOP

NOP

12.8

5.8

Conv. DFT

421.10

421.20

219.20

217.30

214.50

13.10

19.60

18.50

17.30

10.00

11.20

15.80

15.50

14.30

9.00

13.90

12.60

12.50

4.12

Proposed technique

Operation time (ms)

NOP

NOP

NOP

NOP

NOP



NOP

26.0

23.0





NOP

22.0

16.4



18.0

15.2

11.6



Phase comp.[114]

NOP

NOP

NOP

NOP

NOP

13

NOP

18

17

10

12

19

18

16

11

20

18

15

10

Wavelet and LDA [52]









23.12

22.41

200

23.81

19.64

22.41

23.81

200



15.80

22.20

23.70

22.41

15.47

21.03

ANFIS [65]

2.31

23.4

21.9

1.70

1.22

0.57

50.75

20.67

0.94

0.44

0.42

50.60

20.46

0.73

0.59

40.91

20.23

0.62

0.60

Estimated RF ()

B-G

A-G

C-G

B-G

A-G

CA-G

A-G

A-G

A-G

ABC-G

AB-G

A-G

A-G

A-G

BC-G

C-G

B-G

A-G

AB-G

Fault classified

201.810

181.770

151.150

150.960

110.570

99.523

99.614

90.712

90.501

65.290

65.324

65.513

65.122

65.170

50.384

50.452

50.441

50.436

30.127

Estimated fault location d  (km)

0.90

0.98

0.76

0.64

0.52

0.52

0.61

0.79

0.56

0.45

0.50

0.79

0.19

0.26

0.77

0.90

0.89

0.87

0.42

d−d  · 100 d

% Error =

3.5 Validation of Proposed Technique 55

56

3 Adaptive Numerical Distance Relaying Scheme

time for close-in faults, boundary location faults, and high resistance faults simulated in all three zones during different power system disturbances. The obtained results are compared with the existing methods based on conventional DFT, phase comparison principle [114], wavelet transform, linear discriminant analysis (LDA) [52], and also adaptive neuro fuzzy inference system (ANFIS) [65] as shown in Table 3.1. The waveforms of phase voltages, line currents, fault applied, and trip signals during fault resistances of 0.01  and 20  are shown in Figs. 3.6 and 3.7 for Case 17 and 18 of Table 3.1. The effect of DC component will be profound during the low fault resistance and vanish during the high fault resistance, which can be observed from Figs. 3.6b and 3.7b. The adaptive setting of the numerical relay is highly effective for the protection scheme under consideration. Figure 3.8a–e illustrates estimated impedance trajectories during low resistance L-G fault and Fig. 3.8b, d, f indicates high resistance L-G fault simulated at different

Fig. 3.6 a Bus voltages, b line currents and estimated value of fault current, c fault signal, and d trip signal during L-G fault applied at 50 km at 0.1 s with RF = 0.01  and FIA = 0°

3.5 Validation of Proposed Technique

57

Fig. 3.7 a Bus voltages, b line currents and estimated value of fault current, c fault signal, and d trip signal during L-G fault applied at 50 km at 0.1 s with RF = 20  and FIA = 0°

location on line considered between SEB and MEB. It indicates the efficacy of the proposed methodology. It is observed that during the high resistance L-G fault applied at 0.1 s, the phase impedance and line impedance locus remain outside the trip region of uncompensated quadrilateral relay characteristic for conventional DFT, phase comparisonbased algorithm and ANFIS-based algorithm. Hence, the conventional relay may mal-operate. This drawback is avoided by adaptive setting of relay characteristics to ensure proactive response. Thus, in case of proposed technique, the impedance trajectory of phase impedance (Z a ) falls inside the trip region of compensated (adaptive) quadrilateral characteristics. From the comparative results demonstrated above, sound features of the proposed technique can be narrated as follows.

58

3 Adaptive Numerical Distance Relaying Scheme

Fig. 3.8 Fault impedance trajectory during a Case 1, b Case 3, c Case 17, d Case 18, e Case 26, f Case 27 of Table 3.1

• The proposed technique operates faster than the existing techniques during closein faults. • For the faults involving low resistance and higher magnitude of DC component, the response time of the proposed algorithm is always less than existing methods. • When fault resistance is higher than the impedance setting of relay, then the existing scheme treats such fault either as the fault of the next zone or out of zone fault. Whereas the proposed technique accurately detects low to high resistance fault in its actual zone due to adaptive setting of impedance characteristics and ensures reliable operation.

3.5 Validation of Proposed Technique

59

• Moreover, the proposed scheme is capable to discriminate between zone 1, zone 2, and zone 3 faults and operates with proper time margin provided in it. • Table 3.1 also indicates the fault classified by the modeled fault classifier modules. It shows that each kind of fault is perfectly sensed by the scheme and classification accuracy is 100%. • Apart from this, the developed scheme remains stable during load encroachment and saturation of instrument transformers. Nevertheless, it is observed that for the far end faults, operating time is slightly less in case of wavelet and LDA-based technique [52] as compared with the proposed algorithm. Also, the operating time is very less in case of conventional DFT, but with increase in DC component and fault resistance, there is degradation in their performance. The phase comparison principle-based method [114] operates faster during close-in faults, but at the same time, its response time decreases as fault resistance increases. As the reach setting in the proposed technique is 100 km for zone 1, it does not operate quickly for the out of zone 1 fault, which indicates stability and reliability of the proposed technique. However, in ANFIS-based technique [65], due to higher reach setting, the trip signal is generated (Case 30). It can be narrated that the implemented algorithm yields optimum performance parameters in terms of accuracy and response time with adaptive approach. The proposed scheme also provides perfect fault classification as outlined in the next section.

3.5.3 Fault Classification Fault classifier modules are also designed to in order to identify the type of fault occurring in the power system network. The fault detection is extracted depending on the type of impedance locus entering into the trip region of characteristic. In fault classifier, six separate impedance modules are designed, each for individual phase (A, B, and C) and for line (AB, BC, and CA). The output signals TripA, TripB, and TripC are for phase modules and TripAB, TripBC, and TripCA are for line modules, respectively. The response of fault classifier module during validation for different kinds of faults is tabulated in Table 3.2 for each type of fault. The response of the proposed methodology for fault detection scheme is also demonstrated in Fig. 3.9 for AB-G fault applied at 0.1 s. It can be observed that for AB-G fault, the TripA, TripB, and TripAB go high, whereas remaining outputs remain stable at low. The fault classification accuracy is found to be 100% for each type of faults developing in power system network.

60

3 Adaptive Numerical Distance Relaying Scheme

Table 3.2 Output of fault classifier module for each type of fault for various simulated cases Type of fault applied

TripA

TripB

TripC

TripAB

TripBC

TripCA

Fault classified

A-G

1

0

0

0

0

0

A-G

B-G

0

1

0

0

0

0

B-G

C-G

0

0

1

0

0

0

C-G

AB

0

0

0

1

0

0

AB

BC

0

0

0

0

1

0

BC

AB-G

1

1

0

1

0

0

AB-G

BC-G

0

1

1

0

1

0

BC-G

CA-G

1

0

1

0

0

1

CA-G

ABC-G

1

1

1

1

1

1

ABC-G

1 fault is detected 0 fault is not detected

3.5.4 Fault Location Estimation The outcome of variation in fault location error with respect to actual fault distance from relay location is shown in Fig. 3.10 for L-G fault applied with FIA = 0° for the different values of fault resistance. The fault locator scheme calculates the location of the fault by taking a ratio of estimated fault impedance to the unit impedance. The actual fault impedance is estimated by using the proposed phasor estimation technique and unit impedance is derived from the knowledge of X/R ratio of transmission line to be protected. Error in the estimation is calculated using Eq. 3.38. It can be observed from the result analysis as illustrated in Fig. 3.10 that the error in fault location estimation is always less than 1% for wide variation in fault resistance. Thus, compared to existing schemes [18], [20], [22], and [37] in which the optimum error achieved is up to 3%, the fault location error estimated by the proposed method is low for different fault locations, fault types, and variation in fault resistances. Table 3.3 shows the error in fault location estimation for different fault cases with variation in fault resistance. In simulation, the fault resistance has been varied from 0  to 200  to cover the effects of solid faults and variable high impedance faults. From the results presented, it can be narrated that error in fault location estimation always remains within 1%. It is mainly due to correct and accurate phasor estimation performed by the proposed MFCDFT-based technique. It shows remarkable improvement as compared to other existing algorithms.

3.5.5 Fault Cases with CT Saturation The close-in fault imposes severe system disturbances for numerical relaying. When it occurs, fault current will be very high and CTs get saturated depending on the

3.5 Validation of Proposed Technique

Fig. 3.9 Output of fault classifier module for AB-G fault

61

62

3 Adaptive Numerical Distance Relaying Scheme

Fig. 3.10 Percentage of fault location error versus fault distance for L-G fault with FIA = 0°

Table 3.3 Fault location estimation Fault location (km)

Measured fault location (km) RF = 5 

RF = 10 

RF = 30 

% Error in estimation of fault location RF = 80 

RF = 0.01 

RF = 10 

RF = 30 

RF = 80 

2

2.009

2.011

2.012

2.013

0.45

0.55

0.60

0.65

5

5.021

5.028

5.029

5.030

0.42

0.56

0.58

0.60

10

10.03

10.05

10.06

10.07

0.30

0.50

0.60

0.70

20

20.10

20.11

20.12

20.17

0.50

0.55

0.60

0.85

40

40.12

40.14

40.17

40.18

0.30

0.35

0.42

0.45

50

50.25

50.26

50.31

50.34

0.50

0.52

0.62

0.68

60

60.23

60.29

60.31

60.38

0.38

0.48

0.51

0.63

70

70.35

70.39

70.40

70.42

0.50

0.55

0.57

0.60

80

80.38

80.40

80.44

80.48

0.47

0.50

0.55

0.60

85

85.46

85.52

85.59

85.67

0.54

0.61

0.69

0.78

90

90.53

90.57

90.64

90.78

0.58

0.63

0.71

0.86

95

95.58

95.60

95.69

95.86

0.61

0.63

0.72

0.90

magnitude of current and burden resistance in the secondary side. In this work, CT secondary burden resistance is varied from 5 to 20  during close-in faults. The proposed scheme is validated for L-G close-in fault applied at 5 km with FIA = 0° and burden resistance of 15 . The result for the same is shown in Fig. 3.11. It is observed that during severe CT saturation the proposed scheme successfully detects the fault in the zone 1 and generates trip signal within one cycle. This ensures reliability and security of the protection systems.

3.6 Advantages of Proposed Algorithm

63

Fig. 3.11 Waveforms of a bus voltages, b line currents, c fault signal, d trip signal during close-in fault at 5 km with FIA = 0° with CT secondary burden of 15 

3.6 Advantages of Proposed Algorithm From the discussions and results outlined in the previous sections, unique advantages of the proposed algorithm are as follows: • Response of the proposed algorithm is faster which accelerate the Zone-1 operation of the distance relay for high resistance faults. Complete compensation of fault resistance ensures remarkable improvement in the sensitivity of distance relay. • Delayed tripping due to fault resistance can be avoided in Zone-1 and Zone-2. • Under reaching of distance relay during HIF in Zone-3 can be avoided which improves selectivity and stability of power system. • Optimal discrimination between internal and external faults for all zone of protection.

64

3 Adaptive Numerical Distance Relaying Scheme

• Secure and fast operation as compared to existing algorithms with optimum performance parameters. Moreover, the proposed algorithm ensures reliability for close-in faults, saturation of instrument transformer, load encroachment, and decaying DC component.

3.7 Outcome of Proposed Technique This research presents novel decision logic to improve the impedance reach of numerical distance relay by adaptive settings of quadrilateral characteristics. The proposed technique implements MFCDFT algorithm for fast and accurate phasor estimation of fault impedance followed by slope tracking method for adaptive settings of the numerical relaying. System modeling and simulation are performed in PSCAD software package using multi-run facility for capturing samples of faulty signals during varying power system disturbances, and the developed algorithm is validated in MATLAB. The high impedance faults (HIFs) are successfully detected by implementing GDF calculation, slope tracking method, and adaptive quadrilateral relay characteristic. The proposed technique is found to be highly precise and faster than the existing methods during close-in fault, high resistance fault, load encroachment, influence of DC component, and CT saturation. Fault classifier modules are designed for each phase and line to identify the type of fault occurs on transmission line. In order to estimate accurate fault location, a fault locator module is also designed and error in fault location estimation is found within 1% for each zone of protection. The outcome of the proposed algorithm highlights significant contribution to improve stability, sensitivity, and speed of the numerical distance relays in comparison with existing schemes of protection. Moreover, the proposed fault identification algorithm can accurately estimate fault instance, fault location, and type of fault, which are the desirable attributes of multi-functional numerical relays.

Chapter 4

Discrimination Between Power Swing and Line Fault Based on Voltage and Reactive Power Sensitivity

4.1 Introduction Sudden changes in loading or configuration of an electrical network cause power swing which may result in an unwanted tripping of the distance relays. Hence, it becomes utmost necessary to rapidly and reliably discriminate between actual fault and power swing conditions in order to prevent instability in power network due to mal-operation of distance relay. This research proposes a novel method for the discrimination between fault and power swing based on rate of change of voltage and reactive power measured at relay location. The effectiveness of the proposed algorithm is evaluated by simulating series of power swing conditions in PSCAD/EMTDC® software for different disturbances such as change in mechanical power input to synchronous generator, tripping of parallel line due to fault, and sudden application of heavy load. It is revealed that the distance relay gives successful tripping in case of different fault conditions and remains inoperative for power swing with the implementation of the proposed algorithm. The power system is subjected to electrical oscillations due to large and sudden disturbances such as change in mechanical input to the turbine, removal of faults on transmission line, switching ON/OFF of heavily loaded transmission lines, or sudden application/rejection of heavy load. These electrical oscillations are responsible for changes in bus voltages and power transfer through the transmission lines. The phenomenon that creates large variations in power flows between two areas of power system is referred as power swing, and it is classified as stable and unstable power swing. For stable power swing, the electrical oscillations die out and system can again recover stable operating mode. However, severe disturbances result in unstable power swing. Following the unstable power swing, the angle between two equivalent machines progressively increases and eventually both machines lose synchronism [67–69]. The stable or unstable power swing may cause large fluctuations in voltages, currents, and power flows. The distance relay which measures

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 U. Patel et al., Futuristic Trends in Numerical Relaying for Transmission Line Protections, Energy Systems in Electrical Engineering, https://doi.org/10.1007/978-981-15-8465-7_4

65

66

4 Discrimination Between Power Swing and Line Fault Based …

the apparent impedance based on actuating signals of voltage and current may maloperate during power swing [70]. An undesirable tripping of transmission line due to mal-operation of distance relay against power swing may aggravate the power system stability. Hence, the operation of the distance relay has to be blocked during power swing (stable/unstable) by incorporating the power swing blocking (PSB) function in distance relays. On the other hand, the distance relay must operate correctly for the actual fault condition without any undesired time delays [71–73]. Garlapati et al. [124] and Mohammed et al. [125] described power swing detection and fault classification method based on multi-agent-aided distance protection. However, aforesaid schemes can be failed to identify symmetrical faults during the power swing and high resistance faults. Also, multi-agent systems are based on traveling waves from nearby substations which can be distorted during the fault and communication link failure during the fault. Blumschein et al. [126] suggested power swing detection and out-of-step protection using geometrical parameters of impedance trajectory. Lin et al. [127] have evolved a novel self-adaptive distance protection scheme resistant to power swing based on time difference staying within the gap between two circles and the corresponding time staying within internal operating circle covering various types of fault during power swing. The prime limitation of this scheme is that it fails during high resistance fault occurring during power swing situation. Nowadays, numbers of techniques have been suggested for power swing detection in uncompensated transmission lines as outlined in the literature reviews (Sect. 1.6.1), although there exists a lot of opportunity for further development especially on the reliable discrimination between fault and power swing condition. Majority of transmission lines in India are uncompensated transmission lines. Hence, in the proposed work, two different methodologies have been proposed for the detection of power swing. First method is simple and effective which is developed for the detection of power swing in uncompensated transmission line as discussed in this chapter. This method is perfectly suited for protecting the uncompensated transmission lines. However, its performance needs to be improved when the nonlinearity increases as in series compensated lines. Hence, in the next chapter support vector machine (SVM)-based disturbance classifier scheme is proposed. Whenever a major disturbance occurs in transmission line, the voltage and power measured at any bus will be violated significantly. Hence, based on the combination of rate of change of voltage and rate of change of reactive power, a novel method is proposed in this research to distinguish between fault and power swing condition. In order to test the proposed algorithm, various fault and swing cases have been generated by modeling a 220 kV power system network in PSCAD/EMTDC software package [122]. Validation of the proposed technique has been carried out by producing around 200 simulation cases for both power swing and fault conditions. The proposed algorithm for biasing of fault and swing condition is developed in MATLAB® software and found to be more efficient. The next section demonstrates the system modeling for the validation of proposed technique.

4.2 System Modeling

67

4.2 System Modeling Figure 4.1 shows single-line diagram of a 220 kV power system network consisting of two sources connected by parallel transmission lines at sending-end bus (SEB) and receiving-end bus (REB). In Fig. 4.1, the generators (G1 and G2) are modeled as equivalent dynamic sources representing a multi-machine system. Power is transferred from SEB to REB through parallel transmission lines. Bergeron model with distributed parameters is used for modeling of both transmission lines. Generator is modeled with one damper winding in q-axis and an IEEE Type 1(AC1A) solid-state exciter. Both sources G1 and G2 are connected to the bus through generator transformer (GT) of equivalent capacity. The three-phase variable linear and nonlinear loads are connected at REB. The load variation is performed by changing the load from 25 to 125% using multiple run facility in PSCAD software package. The sudden application and removal of loads generate the power swing in the system. The distance relay ‘R’ is located in Line 2 near SEB for which discrimination between fault and power swing is required as shown in Fig. 4.1. Sampling frequency of 4 kHz (80 samples/cycle) at 50 Hz nominal frequency is used in this study. The validation of the proposed algorithm for discrimination between fault and power swing condition, several case studies are carried out in PSCAD. The fault conditions have been created for all possible ten types of faults (L-G, LL, LL-G, and LLL-G) in power system. These faults are applied either on line L1 or line L2 by varying the fault locations with different values of fault resistances and fault inception angles. The change in loading patterns at REB, the sudden application of mechanical

SEB 13.8/220 kV 650MVA GT

B1

L1 (120km)

G1 13.8 kV ∆ Y Inera: 4 MW/MVA

B2

REB

G2

Flt-1 B3

CT

PT

R

L2 (120km)

Fig. 4.1 Single-line diagram of power system network

13.8 kV 615MVA Inera: 4 MW/MVA

Flt-2

B4

Y ∆

Variable Load – 1,2& 3 (25% 50%, 100%, 125%)

68

4 Discrimination Between Power Swing and Line Fault Based …

input to turbines of generating units at SEB and energizing/de-energizing loaded parallel line (L1) have also been considered for creation of power swing. The power system network configuration is as follows: Generator Data (G1 and G2): 615 MVA, 13.8 kV, 50 Hz, Inertia constant (H) = 4 MWs/MVA. X d = 1.81 pu, X d  = 0.3 pu, X d  = 0.23 pu, T do  = 8 s, T do  = 0.03 s, X q = 1.76pu, Xq = 0.25 pu, T  = 0.03 s, Ra = 0.003 pu, Xp (Potier reactance) = 0.15 pu. Transmission line Data: Line Length: L1 & L2 = 120 km, System Voltage = 220 kV. Positive-sequence impedance = 0.0297 + j0.332 /km. Zero-sequence impedance = 0.162 + j1.24 /km. Positive-sequence capacitance = 12.99 nF/km. Zero-sequence capacitance = 8.5 nF/km. Transformer data: 650 MVA, 13.8 kV/220 kV, DYng. 50 Hz Three-Phase Transformer with Leakage Reactance of 12%. Full Load: Load-1: 200 MW, Load-2: 400 MW, and Load-3: 500 MW, at 132 kV, 0.85 power factor, 50 Hz.

4.3 Power Swing Detection: Problems and Remedies 4.3.1 Problem Description During the fault condition, the voltage at the bus reduces and current through the line increases and hence impedance seen by the distance relay decreases [128]. Distance relay operates when measured apparent impedance (Z) enters in predefined zones (may be zone 1, zone 2, or zone 3) of distance relay and stays therein for the longer duration than the set value of operating time. The possibility of mal-operation of distance relay due to tripping of zone 3 element in the event of power swing is more as zone 3 covers highest reach on R–X plane. Hence, functionality of zone 3 of distance relay is considered in this study during severe system disturbances. Figure 4.2 represents the Mho relay characteristic (zone 3), inner and outer power swing detection (PSD) zones and impedance trajectory seen by the relay installed on line L2. The system shown in Fig. 4.1 was initially loaded, and the locus of measured impedance falls away from the outer PSD zone at point-X as shown in Fig. 4.2.

4.3 Power Swing Detection: Problems and Remedies

69

Fig. 4.2 Impedance locus of line L2 relay at SEB during 25% load increase at REB

The conventional power swing detection technique is shown in Fig. 4.2 which operates on the rate of change of the impedance (dz/dt) during its travel between blinders (PSD zones) and zone 3. This rate of change of the impedance (dz/dt) is too slow and takes more time during power swing condition to cross the distance between blinder and zone 3. On the other hand, its movement is faster during fault. However, during major system disturbances, this scheme may fail due to unwanted tripping of relay in zone 3 [84]. The system disturbance is created at time t = 2 s by applying sudden increase in load of 25% at REB. This causes the power swing in a system and apparent impedance seen by the relay travels from original operating point-X toward zone 3 of relay. If it stays therein for longer duration, then the relay issues trip signal, and finally, line L2 gets disconnected unnecessarily which results into system instability as shown in Fig. 4.3. On the other hand, during zone 3 weak fault (at far end of next line), the trajectory of impedance some time remains outside the zone 3 characteristic and relay remain inoperative even after a predefine time elapse for backup protection. Hence, blinder characteristic-based power swing blocking technique is not competent to discriminate between load encroachments and short circuit in all the way. Relay mal-operation during system disturbance other than actual fault leads to redundant system separation and affect system stability. Hence, a new method is suggested in order to prevent relay mal-operation and securing power system stability.

70

4 Discrimination Between Power Swing and Line Fault Based …

Fig. 4.3 Distance relay operation during heavy load fall on system

4.3.2 Proposed Method for Power Swing Detection For the discrimination between power swing and actual fault condition, a simple but effective method has been proposed in this work. The method utilizes the rate of change of line voltage and reactive power as a decisive factor for the required discrimination. Any change in the power system network in terms of short circuit or change in load accompanies the variation of bus voltages and power transfer through the transmission lines. These variations in generator/bus voltages and power (active and reactive) depend on the nature of change in system configuration. With the measurement of rate of change of bus voltage V S and line reactive power QS at SEB, an appropriate index is determined which can be effectively used to discriminate between fault and power swing condition. It is observed that whenever short circuit occurs on transmission line, the voltage magnitude is considerably decreased in the faulted phase [129]. However, it is also observed that the magnitude of voltage get reduces during generator outage or sudden load increase and excitation failure which can result in negative rate of change of voltage (dV /dt). Thus, the technique based on rate of change of voltage alone may not be capable to discriminate the said two power system conditions of transmission line.

4.3 Power Swing Detection: Problems and Remedies

71

To overcome the limitations of decision criteria made on the basis of only rate of change of voltage magnitude, a new algorithm is proposed which perfectly distinguishes fault and power swing conditions. The proposed algorithm is based on combination of rate of change of bus voltages (dV/dt) and line reactive power (dQ/dt) measured at relaying point at SEB. The proposed fault and power swing discriminating algorithm is shown in Fig. 4.4. The occurrence of both fault and power swing brings the impedance locus seen by the relay into the operating region of zone 3 of distance relay but the dV /dt and dQ/dt-based decision logic precisely identifies the actual condition due to which the impedance locus is brought in the operating zone of the distance relay. Decision about the fault or power swing conditions depend on the following criteria: A fault is declared, when:      dQ  dQ SET  dV  dVSET >    and (4.1)  dt  > dt  dt  dt Otherwise, power swing is declared and activates power swing blocking (PSB). Once the fault condition is declared, the relay logic waits until the timer times out. The final check for the status of impedance locus to stay within the operating region of distance relay will be ascertained. The relay issues the trip signal if the impedance locus still presents within the zone 3 boundary of the distance relay for the longer duration than time out time. The reactive power measured by the distance relay can be represented by: QS =

VS2 − VS VR cos δ X

(4.2)

where QS Sending-end reactive power, V S Sending-end voltage, V R Receiving-end voltage (reference), δ angle difference between V S and V R , X Line reactance. For smaller values of angle difference (δ) and load impedance, the amount of reactive power delivered is very low. During fault condition, the value of angle difference (δ) may change drastically within a cycle. Thus, the rate of change of δ is faster in event of faults as compared to power swing condition. Moreover, during fault on line, the relay encountered only the line reactance (X) present in the system up to a fault point. The variation in active power (PE ) and reactive power (QS ) with the change in δ is depicted in Fig. 4.5. The reactive power increment is very high during fault as compared that of swing condition.

72

4 Discrimination Between Power Swing and Line Fault Based …

Start Read Current (IR, IY and IB) and Voltage (VR, VY, VB) Samples from line CTs and bus PTs Calculate apparent impedance and reacve power

Is impedance entered into Zone-3?

No

Yes

No

|dV/dt|> dVSET/dt

|dQ/dt|> dQSET/dt

No

Yes

Yes Start Timer

Power Swing

Timer Times Out

Blocking Signal

No

Sll impedance present in Yes Trip

Fig. 4.4 Proposed algorithm to discriminate power swing and fault condition

4.4 Simulation Results and Discussions In order to test effectiveness of the proposed scheme under varying system conditions, a large numbers of fault cases have been studied. These cases are simulated considering the following parameter variations:

4.4 Simulation Results and Discussions

73

Real and Reacve Power (pu)

QS

PE

0

45

90

135

180

Load

Angle δ (deg) Fig. 4.5 Real and reactive power flow in transmission line during change in δ

(i) Fault inception angle (FIA) between 0° to 180°, (ii) Ten types of faults (L-g, L-L, L-L-g, and L-L-L-g) including high resistance fault, (iii) Fault locations (FL) on line L2 which also includes close-in faults, (iv) Fault resistance (RF ), and (v) Power flow angle (δ). The detection of symmetrical fault occurring during power swing is the most difficult task for the existing numerical relays. In this investigation, the validation of the proposed scheme is also carried out for the faults which may occur when power swing is prevailing. The power swing cases are simulated considering following system parameter variations: (i) (ii) (iii) (iv)

Electrical load switching. Mechanical input variation to the turbine of the generator. Fault clearance on adjacent transmission line (L2). Post-fault isolation.

Considering all these varying conditions for faults and power swing, around 200 simulations cases have been analyzed and results are reported and discussed at length for the some of the simulated test cases.

74

4 Discrimination Between Power Swing and Line Fault Based …

4.4.1 Power Swing Cases Due to Electrical Load Switching The application of the electrical load brings the impedance locus seen by the distance relay much closer to operating region of the relay, whereas the rejection of the electrical load shifts it away. Thus, the distance relay faces the problem of overreaching and may mal-operate with the application of heavy electrical load if the zone 3 characteristic covers larger portion on R–X plane. In this work, several cases for electrical load changes are considered by increasing the load at REB by 25, 50, 100, and 125% above its original loading as given in Fig. 4.1. First two windows of Fig. 4.6 show the waveforms of voltage and current measured at the relay location in Fig. 4.1. Also, the magnitude of rate of change of voltage and reactive power for the case when the load is increased suddenly by 25% above its original value at t = 3 s (1.2e4 samples) is shown in Fig. 4.6. The rate of

Fig. 4.6 Variations in voltage and reactive power during 25% overloading

4.4 Simulation Results and Discussions

75

change of voltage and reactive power is compared against their preset values to take decision about the power swing. The threshold values for the voltage and reactive power are set at 0.2 pu/s and 0.4 pu/s, respectively, based on the several simulations studies for power swing conditions. It can be observed from Fig. 4.6 that during power swing, dv/dt and dQ/dt do not cross their respective threshold values; thus, trip signal of the relay is not generated even after sudden application of load. This clearly reveals that the proposed method can effectively identify the power swing condition and prevent the mal-operation of the distance relay.

4.4.2 Power Swing Cases Due to Mechanical Disturbances In this case, power swing is generated by changing the mechanical input to the turbine generator system. The imbalance between mechanical power input to the turbine and electrical power output of generator results in acceleration or deceleration of the generator rotor. The rotor experiences the heavy oscillations till the imbalance between input and output power persists [130]. In this case, the mechanical power input to turbine is reduced by 20% below its original value at t = 3 s. The dv/dt and dQ/dt are determined for this case, and it is observed in Fig. 4.7 that the values of these parameters are well below the set threshold values. Hence, for this case also, the relay remains inoperative against power swing conditions.

4.4.3 Fault Simulation on Protected Line (L2) Various fault cases are simulated on line L2 by considering different types of faults and variations in system parameters such as fault inception angle (FIA), fault resistance (RF ) fault locations (FL), and power flow angle (δ). For all these fault cases, distance relay must identify the faulty condition and operate as per the set operating time. Depending upon the zone of the fault, the operating time may be different. Figure 4.8 shows the variations in different parameters for the single line-to-ground fault (L-G) which is created in line L2. The FIA, RF, FL, and δ are set equal to 0°, 15 , 30 km, and 5°, respectively. With the occurrence of the fault, the voltage and reactive power changes very rapidly this can be clearly noticed from Fig. 4.8. The dv/dt and dQ/dt cross the threshold values and results in trip signal which has been initiated after waiting up to the operating time of the relay. Hence, in Fig. 4.8, the trip signal for relay has been generated after certain samples once the fault condition is correctly declared by the proposed method. The substantial changes in dv/dt and dQ/dt for the fault condition as illustrated in Fig. 4.8 are defensible as compared to those shown in Figs. 4.6 and 4.7 for power swing. These significant changes have been successfully utilized for the discrimination between fault and power swing conditions.

76

4 Discrimination Between Power Swing and Line Fault Based …

Fig. 4.7 Power swing due to change in mechanical power input to the turbine of generator G1

4.4.4 Fault Cases During Power Swing Distance relays must be blocked during power swing to ensure reliability of the power system. However, if a fault occurs during a power swing, it should be detected and the blocking function should be removed to clear the fault as soon as possible. Due to the symmetric nature of signals during the power swing, symmetrical faults are difficult to be detected. To test the proposed method, several asymmetrical and symmetrical faults are simulated on the line to be protected during different swing conditions persisting in system. As shown in Fig. 4.9, a power swing case is simulated at t = 3 s by connecting extra load of 25% at REB and at the same time by disconnected parallel line (L1). However, the impedance trajectory does not enter into zone 3 characteristic, and

4.4 Simulation Results and Discussions

77

Fig. 4.8 Relay response during fault on line to be protected (L2)

hence, relay remains unresponsive. Later on, in continue power swing situation a symmetrical three-phase fault is applied at t = 4 s to validate the proposed scheme. It has been observed from Fig. 4.9 that the dV /dt and dQ/dt value during power swing (at 3 s) are well below the threshold and hence no trip signal has been issued. On the other hand, during fault on power swing, these values crosses respective thresholds (at 4 s) and hence relay sends trip signal as shown in Fig. 4.9. Thus, the proposed method successfully identifies symmetrical faults during power swings for all different power system conditions. In order to test the proposed scheme and further to compare with existing scheme, few extra fault cases are simulated during power swing condition. Table 4.1 shows the relay operation criterion and result in terms of time of operation of the proposed scheme. The proposed algorithm is validated for the symmetrical

78

4 Discrimination Between Power Swing and Line Fault Based …

Fig. 4.9 Performance of proposed method for fault discrimination during power swing

and asymmetrical faults occurring during the power swing including variation in power system parameters like fault type, load flow angle (δ), fault location, and fault resistance. A comparison between existing method [127] and proposed method is also outlined in above Table 4.1. It is to be observed from Table 4.1 that the proposed scheme operates faster than the existing scheme during various low resistance fault generated in presence of power swing. Moreover, during the fault Case 7 in Table 4.1 for high resistance fault, the existing method fails to detect it and the proposed scheme successfully operates. In addition, during the fault at far end of transmission line (95% of line length), as per Case 8 in Table 4.1, the existing method again fails to sense the fault during swing condition, whereas the proposed scheme operates and issues delayed trip signal.

4.4 Simulation Results and Discussions

79

Table 4.1 Comparison of proposed and existing scheme for symmetrical and asymmetrical faults during power swing situation Fault case

Fault type

Fault location Fault (percentage Resi. of line length) () (%)

Tripping time (ms) using existing method [127]

Tripping time (ms) using proposed method

Tripping time (ms)

Trip output

Tripping time (ms)

Trip output

24.5

Yes

21.2

Yes

1

L-G

40

10

2

L-L

40

10

24.5

Yes

21.2

Yes

3

L-L-G

40

10

24.5

Yes

21.2

Yes

4

L-L-L

40

10

24.5

Yes

21.2

Yes

5

L-G

60

10

28

Yes

23.45

Yes

6

L-G

40

80

30

Yes

24.1

Yes

7

L-G

40

200

NOP

No

28.23

Yes

8

L-G

95

10

NOP

No

27

Yes

4.4.5 Power Swing Cases Due to Post-Fault Isolation on Line L1 The proposed methodology has been also tested for the power swing generated due to fault isolation on adjacent transmission line. Numerous faults are simulated on line L1 shown in Fig. 4.1, and effect of power swing is observed by quick isolation of fault. For this simulation case, a L-G fault is simulated at 2.9 s and thereafter power swing is generated by isolation of same fault at 3 s by opening of circuit breakers B1 and B2 of line L1 as shown in Fig. 4.10. Figure 4.10 shows the value of voltage and current monitored at line L2 during fault and on isolation of fault from parallel line. The removal of fault at 3 s on adjacent line L1 produces power swing effect on the relay located at line L2. The apparent impedance enters into zone 3 setting of relay just after the instant of fault; hence, zone 3 timer is initiated. Conversely, as fault is removed at 3 s (by local protection of line L1), the magnitude of dV /dt and dQ/dt goes down to that of the set value before the timer times out. Thus, the proposed relay remains stable and does not issue any trip signal as shown in Fig. 4.10.

4.4.6 Relay Backup for the Fault on Parallel Transmission Line Since zone 3 relay mainly serves as a backup relay, the dynamic operation of the parallel or adjacent transmission lines should be considered. In case of a single lineto-ground fault on parallel line, the dynamic process, including the initiation of the

80

4 Discrimination Between Power Swing and Line Fault Based …

Fig. 4.10 System condition during fault isolation on adjacent line

fault, single pole tripping, and the correlative voltage and reactive power changing, should all be considered. The removal of fault on adjacent line L1 produces power swing effect. However, the proposed relay (located on line L2 in Fig. 4.11) remains stable and does not issue any trip signal as discussed in previous section. On the other hand, if the fault on line L1 is not isolated by relay of parallel line L1, then the proposed relay located on line L2 will issue a trip signal and serves as backup protection. To validate the algorithm, a solid L-G fault at 70 km from REB on parallel line L1 is created and parameter variations for the same are shown in Fig. 4.11. It is observed that the impedance enters into zone 3 region of relay characteristic. Moreover, the magnitude of rate of change of voltage, i.e., dV /dt is higher than that of set value of threshold at the time of fault. Also, slightly high value of rate of change of reactive power (dQ/dt) is observed during said fault condition. It is to be noted that the impedance spot still remain in zone 3 after a set time elapsed by timer during said fault in parallel line. Thus, the proposed relay issues delayed trip signal and serves as backup protection as shown in Fig. 4.11. The results presented indicate the effectiveness of the proposed methodology for discrimination between fault and power swing during diversified disturbances developing in the system. The next section demonstrates the research outcome of the developed scheme of protection.

4.5 Research Outcome of Proposed Technique

81

Fig. 4.11 Performance of proposed method for fault on parallel transmission line

4.5 Research Outcome of Proposed Technique This research presents a new method to discriminate fault condition and power swing condition in the power system. The proposed algorithm utilizes voltage and current signal to calculate rate of change of voltage and reactive power. The magnitude changes of dV /dt and dQ/dt effectively discriminate between fault and power swing conditions. The 220 kV power system network is modeled in PSCAD/EMTDC, whereas algorithm is designed and validated using MATLAB software. Feasibility of the proposed scheme has been tested with a simulation data set of around 200 cases generated with varying faults and system conditions. Power swing cases are simulated by changing load, disconnection of parallel line, isolation of fault in adjacent line, and mechanical disturbance. Various fault cases are generated by varying fault location, fault type, power flow angle, and fault resistance on line to be protected. Moreover, the proposed scheme is also validated for symmetrical and asymmetrical faults in existence of power swing condition and backup operation of relay during zone 3

82

4 Discrimination Between Power Swing and Line Fault Based …

fault on adjacent transmission line. The proposed scheme is very simple and more effective for discrimination of fault and power swing situations and found to be highly accurate for all the simulated cases in this study.

Chapter 5

Sequence-Space-Aided Disturbance Classifier Scheme Based on Support Vector Machine

5.1 Introduction Smart power grids are equipped with series compensation in order to improve the power transfer capability. Sudden changes in loading or weak constitution of series compensated transmission line network causes power swing which may aggravate miss-operation of protective elements. Consequently, it becomes utmost essential to rapidly and accurately distinguish between fault and power swing conditions to prevent instability in smart power grid equipped with compensation. This research demonstrates an effective disturbance classifier scheme for series compensated transmission line for discrimination during disparity in power grid contexts using sequence-space-based support vector machine (SVM) classifier. The test data sets are generated by performing extensive simulations in PSCAD software by varying system and fault context. SVM architecture has been trained and tested by generating feature vector using modified full-cycle discrete Fourier transform (MFCDFT) in MATLAB. After proper extraction of features of the interest at the time of disturbance, a decision about power swing or fault has been carried out using SVM classifiers. Regulation and kernel function parameters have been tuned using tenfold cross-validation applied on training set. The developed scheme is also evaluated for symmetrical fault detection during power swing and shows remarkable improvement in accuracy and speed for protection of series compensated transmission line in comparison with existing schemes. Electric Power Research Institute (EPRI) has investigated cascade tripping and widespread blackout in smart power grid network of USA. It has been suggested that protection schemes and control strategies must stop system degradation, minimize impacts and facilitate the system restoration [1]. Taskforce on grid disturbance in India has insisted that numerical relays used for protection of Extra High Voltage (EHV) transmission line must remain stable during power swing [2]. Moreover, detection of symmetrical fault during power swing is challenging task for protection

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 U. Patel et al., Futuristic Trends in Numerical Relaying for Transmission Line Protections, Energy Systems in Electrical Engineering, https://doi.org/10.1007/978-981-15-8465-7_5

83

84

5 Sequence-Space-Aided Disturbance Classifier …

engineers especially in series compensated transmission line (SCTL) [3]. Compensation plays important role in EHV systems for improving stability and power transfer capacity. Nowadays, many researchers have evolved numerous techniques for sheltering series compensated transmission line as outlined in Sect. 1.6.2. In the proposed work, in order to precisely discriminate between fault and power swing scenario, modeling of a 400 kV network topology has been carried out in PSCAD/EMTDC, an established simulator to study power system dynamics. Different types of system disturbances are applied to realize fault and power swing conditions. Whenever abrupt changes occur, the line currents measured by the Current Transformers (CTs) will be desecrated radically. These inequalities in line currents will be reflected in terms of sequence components. Hence, based on the variation in sequence components of line currents, a novel technique is projected for disturbance classification. The phasor estimations of analog signals have been performed using MFCDFT as outlined in Sect. 3.2.2 which is further applied to sequence components-based SVM classifier as presented in Sect. 5.3. In order to improve the disturbance classification accuracy of the proposed algorithm as compared to existing algorithms, tenfold cross-validation technique is used as outlined in Sect. 5.4. The data mining for different fault and power swing scenario is outlined in Sects. 5.5 and 5.6, respectively, followed by validation of proposed algorithm in Sect. 5.7. The results obtained using developed algorithm are also compared with existing schemes as outlined in Sect. 5.8. The outcome of proposed technique is presented in Sect. 5.9.

5.2 System Modeling IEEE PES Power System Relaying Committee (PSRC) has presented single line diagram of 400 kV power transmission network as shown in Fig. 5.1 for power system

Fig. 5.1 Single line diagram of power system

5.2 System Modeling

85

analysis [131]. There are two generators (G1 and G2) connected by 120 km parallel transmission lines (L1 and L2) between sending-end bus (SEB) and middle-end bus (MEB) followed by second section of 100 km line (L3) between MEB and receivingend bus (REB). Lines L1 and L2 are equipped with series compensation protected by MOV at sending end to improve power transfer capability. Multi-machine generating system is modeled as dynamic source of power with IEEE Type-1 excitation system. Bergeron line models with distributed parameters are used for modeling of transmission line. Current signals from CTs sampled at a frequency of 4 kHz (80 samples/cycle) are applied to the numerical relay ‘R’. The system specifications are summarized in follows: Generator Data (G1 and G2) 615 MVA, 13.8 kV, 50 Hz. Inertia constant (H) = 4 MWs/MVA. Positive-sequence impedance = 0.871 + j9.96 . Phase angle = 0° (G1) and Phase angle = variable (G2) X d = 1.81 pu, X d = 0.3 pu, X d = 0.23 pu,   Tdo = 8 s, Tdo = 0.03 s, X q = 1.76 pu, X q = 0.25 pu,

T  =0.03 s, Ra = 0.003 pu, X p (Potier reactance) = 0.15 pu

Transmission Line Data Line Length: L1 and L2 = 120 km. System Voltage = 220 kV. Positive-sequence impedance = 0.0297 + j0.332 /km. Zero-sequence impedance = 0.162 + j1.24 /km. Positive-sequence capacitance = 12.99 nF/km. Zero-sequence capacitance = 8.5 nF/km. Transformer Data 650 MVA, 13.8 kV/220 kV, DYng, 50 Hz. Three-phase transformer is with leakage reactance of 12%. The proposed technique has been verified for all 10 types of faults (L-G, LL, LL-G, LLL-G) at variable locations with wide variation in Fault Inception Angle (FIA) (0o −315°) and Power Flow Angle (PFA) (0°, 5°, 10°, 15°). The degree of compensation has been varied from 0% (uncompensated line) to 40%. Power swing cases are evaluated by simulating conditions such as change in mechanical input to synchronous generator, tripping of adjacent line because of fault and sudden application/outage of intense load at MEB. The proposed scheme is also evaluated for asymmetrical and symmetrical faults detection during power swing including fault resistance (0.01–50 ) and CT saturation condition. Total 32,736

86

5 Sequence-Space-Aided Disturbance Classifier …

numbers of simulation cases are created for training of SVM and validation has been performed for 2400 test cases including unseen data.

5.3 Sequence-Space-Based SVM Classifier Scheme Figure 5.2 represents the proposed algorithm which consists of two main stages: 1. Phasor estimation using MFCDFT. 2. Sequence-space-based SVM classifier scheme. In order to detect the type of disturbance occurred in power system, samples of line currents are captured from PSCAD as shown in Fig. 5.2 using multiple run facility with sampling frequency of 4 kHz (80 samples/cycle). For phasor estimation of these signals, MFCDFT is realized in the proposed work which implements a low pass Butterworth filter to remove harmonics and aliasing effects. In the proposed work, in order to estimate fundamental component of all currents simultaneously, the phasors (real and imaginary part) for any jth signal at kth instance can be obtained by, Fig. 5.2 Proposed Methodology of SVM-ased Classifier Scheme

5.3 Sequence-Space-Based SVM Classifier Scheme

87

X r 1 ( j, k) = X r ( j, k) 2 − N

k  

D( j).e

−kT τ ( j)

+ D1 ( j).e

−kT τ1



n=k−N +1



2π n cos N

X i1 ( j, k) = X i ( j, k) −

2 N

k  

D( j).e

−kT τ ( j)

+ D1 ( j).e

−kT τ1



 sin

n=k−N +1



2π n N

(5.1)

 (5.2)

where N = Number of samples/cycle, T = Sampling time, X r ( j, k) = real component of complex phasor and X i ( j, k) = imaginary component of complex phasor. The unknown parameters D and D1 are calculated for sliding window of 80 samples using Eqs. 5.1 and 5.2 with the use of next three samples and then by taking its ratio as outlined in Sect. 3.2.2. The RMS value and phase angle for one complete cycle are calculated by using the equation, 

(X r 1 ( j, k))2 + (X i1 ( j, k))2 2   X i1 ( j, k) ∠(X ( j, k)) = tan−1 X r 1 ( j, k)

|M( j, k)| =

(5.3) (5.4)

Equations 5.3 and 5.4 give fundamental complex phasors after removal of all the harmonics and decaying DC component. These are further used for sequence-spacebased SVM classifier algorithm. Once accurate phasors for line currents Ia ,Ib ,Ic are obtained by MFCDFT as shown Fig. 5.2, sequence components are calculated using the model transformation equation, ⎤ ⎡ ⎤⎡ ⎤ I1 1 a a2 Ia 1 ⎣ I2 ⎦ = ⎣ 1 a 2 a ⎦⎣ Ib ⎦ Where, a = −0.5 + 0.866i 3 I0 Ic 1 1 1 ⎡

(5.5)

The disturbance occurring in the line can be classified in two categories, symmetrical and asymmetrical disturbance. When asymmetrical change like L-G fault and LL fault occurs, zero and negative sequence component of fault current will change remarkably and it remains almost zero during symmetrical disturbance like LLLG fault and power swing conditions. The positive-sequence component reflects significant variation in symmetrical disturbance. The absolute value of the sequence components, Qj (k) over one complete cycle can be computed using,

88

5 Sequence-Space-Aided Disturbance Classifier …

1 Q j (k) = N

k  n=k−N +1

I j (n)

(5.6)

where j = sequence number (0,1,2) and k = instantaneous sample. Absolute value Qj (k) is compared with respective preset threshold I Tj to derive model signal.

M(k) =

1Q j(k) ≥ IT j 0Q j(k) < IT j

(5.7)

The rising edge in the waveform of model signal M(k) indicates remarkable disturbance in the power grid. The values of I T 0 and I T 2 can be considered from 0.1 to 0.5 pu of maximum load current and that of I T 1 can be set within 1.1–1.3 pu for avoiding erroneous judgment. In the proposed work, based on extensive simulation studies on the system modelling outlined in Sect. 5.2, I T 0 and I T 2 are set to 0.3 pu of maximum load current whereas I T 1 is set to be 1.2 pu of maximum load current for accurate results. The raw data collection for validation in SVM has been initiated at the rising edge of model signal for one fundamental cycle. SVM is a very powerful data mining artificial intelligence-based tool used for regression and classification problems. There are basically two types of classifiers, namely single layer and multilayer [132]. The single layer classifier is a binary classifier which has two possible states like +1 (fault) and −1 (swing) used for classification problems. Whereas, multilayer classifier is a discrete classifier which is mainly used for regression problems. The inputs applied to SVM have been expanded into higher dimensional space using sequence-space based hyperplane. In this investigation, two-class (fault and power swing) of data sets (Xi|Y i) N with N number of data points are considered where i = {0, 1, 2}, X i indicates feature vectors and Y i reflects its data label [+1, −1]. The equation of separating hyperplane is given by wT + b = 0

(5.8)

The ‘w’ is weight vector and ‘b’ is bias term to determine position of hyperplane. The separation distance can be given by m=

2 ||w||

(5.9)

The separation distance can be increased by considering minimum value of w. For linear separation, SVM can be realized by support vector, v(w) =

1 T w w 2

(5.10)

Above linear support vector can be transformed into higher dimensional nonlinear vector using,

5.3 Sequence-Space-Based SVM Classifier Scheme

89

  Y i w T X i + b ≥ 1 where, m  n

(5.11)

In practice, transformation of nonlinear data sets can be accomplished by an appropriate kernel function. The selection of kernel function plays an important role for gaining accuracy of classification. In many literatures, it is narrated that Radial Basis Function (RBF) kernel is ideal choice for nonlinear classification and regression problems [133] which is defined by, 1 2 K (X i, Y i) = e(−γ ||X i−Y i|| ) where, γ = 2σ 2

(5.12)

This expression maps available data sets into higher dimensional plane. The training of SVM can be accomplished by Eq. 5.10 to find maximum separating distance for hyperplane. In order to ensure appropriate separation of all data, Eq. 5.10 can be expanded in the form of support vectors,  1 T w w+C ∈i 2 i=1 N

v(w, ∈) =

(5.13)

These support vectors are used for training of SVM classifier. This nonlinear support vectors can be transformed in to higher dimensional nonlinear vector using,   Yi wT (X i) + b ≥ 1− ∈i where, ∈i > 0

(5.14)

where ∈i = 1, 2, …, N are slack variables and C is called regulation coefficient. The feature vectors as shown in Figs. 5.3 and 5.4 are applied to SVM model for training during fault and power swing conditions, respectively. The decision obtained

Fig. 5.3 Feature vectors during faulty conditions for training and testing of SVM

90

5 Sequence-Space-Aided Disturbance Classifier …

Fig. 5.4 Feature vectors during power swing conditions for training and testing of SVM

from the support vector classification is mapped in the hyperplane; in this case it is also called sequence-space as shown in Fig. 5.5. The sequence-space is used to classify the disturbance occurring in the power system network into two categories: Fault and power swing. The accuracy of classification depends on the value of kernel function parameter σ and regulation coefficient C. In the proposed soft computing algorithm, the optimum values of these parameters are computed using tenfold crossvalidation technique [134]. After extensive analysis of cross-validation as outlined in the next section, it has been observed that with C = 5000 and σ = 0.025, the proposed classifier outperforms with highest classification accuracy of 99.78%. SVM-based machine learning algorithms normally use two types of technique for finding regulation parameters, i.e., fivefold or tenfold cross-validation techniques. Here, in this investigation, tenfold cross-validation technique is used as outlined in next section. Fig. 5.5 Mapping of support vectors in the sequence-space

5.4 Tenfold Cross-Validation

91

5.4 Tenfold Cross-Validation Proper understanding of regulation parameters and issues associated with training forms the basis for correct classifier settings. To find optimum value of these parameters, following steps are performed. i. Randomly divide all of the available data sets into ten equal subsets. ii. For each subset, train SVM architecture ten times; keeping one subset on hold while applying others for training of SVM by varying parameters as shown in Table 5.1. iii. The trained architecture is then tested using the held-out subset and its accuracy has been recorded. iv. Finally, model with highest accuracy is adopted. The regulation coefficient C and σ of RBF are varied over a wide range to obtain maximum accuracy. The range of C is varied from 1 to 10,000 and that of σ from 20 to 0.001. v. For optimum separation of support vectors in the sequence-space shown in Fig. 5.5, the value of C should be at optimum maxima and value of σ should be at optimum minima which give classical accuracy for the single-layer SVM techniques [134]. The identification accuracy (η) ´ is given by, η˙ =

N umber o f corr ect pr ediction during testing ∗ 100 T otal number s o f pr ediction during testing

(5.15)

Numerous simulations have been performed to determine optimum parametric settings using tenfold cross-validation. This process is called parameter scanning. The accuracy obtained during each scan, are tabulated in Table 5.1. It can be narrated that with C = 5000 and σ = 0.025, the proposed classifier outperforms with highest classification accuracy of 99.78% during cross-validation. These values of regulation parameters perform with highest classification accuracy. In next sections, the data mining for different fault and power swing scenario is discussed followed by results obtained for each case with suitable waveforms. Also, comparative analysis with many other existing algorithms is also presented in Sect. 5.8. Table 5.1 Accuracy variation during SVM parameter settings C

RBF kernel parameter σ

1

20

10

10

83.42

86.67

100

84.55

88.23

5000

84.12

90.12

10,000

87.18

91.18

1

0.01

0.025

0.001

90.11

96.45

97.61

96.28

94.91

97.02

98.27

96.43

96.75

99.14

99.78

97.01

96.56

98.83

98.69

96.62

92

5 Sequence-Space-Aided Disturbance Classifier …

5.5 Data Mining Using SVM Classifier for Power Swing Cases 5.5.1 Power Swing Due to Load Switching Table 5.2 shows validation for power swing scenario due to electrical load switching at MEB. In the proposed validation, both linear and nonlinear load variations have been considered. The training of SVM classifier is performed for total 512 data sets by developing user defined functional model in PSCAD software package. Using the developed model the load can be switched from 25 to 125% sequentially during various multiple runs. The testing of SVM classifier has been performed for 288 multiple runs for checking the effectiveness of proposed algorithm. The correct classification of –1 (Swing) has been achieved for all cases with disturbance identification accuracy of 100% indicating advantages of proper setting of regulation parameters in SVM. The proposed algorithm is also validated for wide variation in fault and system parameters along with load switching as outlined in Table 5.2. Table 5.2 Power Swing cases due to switching of electrical load connected at MEB Power system element/parameter

SVM training patterns for power swing cases

SVM testing patterns for power swing cases

Variation in parameter

Numbers of variation

Variation in parameter

Numbers of variation

Permanently ON

1

Permanently ON

1

Compensation level 0% at the sending end of (uncompensated), transmission line L1 10, 20, 40% (compensated)

4

0% (uncompensated), 10, 30, 40% (compensated)

4

Load variation

50, 75, 100, 125%

4

25%, 50%, 80%, 110%

4

Load switching

OFF to ON and Vice versa

2

OFF to ON and Vice versa

2

Bus at which load is MEB connected

1

MEB

1

Power flow angle (z) 0, 5, 10, 15

4

3, 8, 12

3

Power swing inception angle (PIA)

4

30°, 75°, 100°

3

512

Total testing patterns

288

Line L1

0°, 45°, 90°, 135°

Total training patterns for power swing due to load variation Total correct data classification (−1)

288

Identification accuracy

100%

5.5 Data Mining Using SVM Classifier for Power Swing Cases

93

Table 5.3 Power swing cases due to variation in mechanical input Power system element/parameter

SVM training patterns for power swing cases

SVM testing patterns for power swing cases

Variation in parameter

Numbers of variation

Variation in parameter

Numbers of variation

Change in mechanical torque (Tm)

±30, 60, 100%

6

±10, 50%

4

Power swing inception angle (PIA)

0°, 45°, 90°, 135°, 180°

5

0°, 30°

2

Compensation level 0% at the middle of (uncompensated), transmission line L1 10, 30, 40% (compensated)

4

0% (uncompensated), 10, 30, 40% (compensated)

4

Load in the system

4

25, 50, 80, 110%

4

Bus at which load is MEB connected

50, 75, 100, 125%

1

MEB

1

Power flow angle (z) 0, 5, 10, 15, 20

5

3, 8, 12

3

Total training cases for power swing due to removal of fault on adjacent line

2400

Total testing patterns

384

Total correct data classification (−1)

384

Identification accuracy

100%

5.5.2 Power Swing Due to Mechanical Disturbances Table 5.3 shows power swing scenario due to variation in mechanical input to the turbine of the synchronous generator G1 in Fig. 5.1. Simulation is performed for 2400 data sets for training followed by validation on 384 multiple runs using SVM classifier. The correct classification of –1 has been achieved for all cases with identification accuracy of 100% indicating advantages of proper setting of regulation parameters in SVM. The mechanical input has been varied as ±30, ±60 and ±100% during training of SVM algorithm whereas change in mechanical input is varied as ±10 and ±50% during testing of SVM classifier algorithm.

5.5.3 Power Swing Due to Fault Isolation on Adjacent Line Different kinds of faults are applied on line L2 at various locations which are cleared by the distance relays at line L2. It generates the power swing scenario on the line L1. The developed scheme is trained and tested for wide range of power system disturbances. Table 5.4 shows data mining for power swing scenario due to the application of fault on Line L2 with different fault context. The SVM model has

94

5 Sequence-Space-Aided Disturbance Classifier …

Table 5.4 Power swing cases due to fault isolation on adjacent line Power system element/parameter

SVM training patterns for power swing cases

SVM testing patterns for power swing cases

Variation in parameter

Numbers of variation

Variation in parameter

Numbers of variation

Fault applied on line L2

1

L2

1

Fault type

L-G, LL, LL-G, LLL-G

10

L-G, LL, LL-G, LLL-G

10

Fault location (km)

5, 30, 70, 95

4

50

1

Compensation level 0% at the middle of (uncompensated), transmission line L1 10, 20, 30% (compensated)

4

0% (uncompensated), 10, 20, 30% (compensated)

4

Load in the system

50, 75, 100, 125%

4

25, 50, 80%

4

Power flow angle

5, 10, 15

3

8

1

Fault inception angle

0°, 45°, 90°, 135°, 180°

5

0°, 30°

2

9600

Total testing patterns

320

Total training cases for power swing due to removal of fault on adjacent line Total correct data classification (−1)

314

Identification accuracy

98.12%

been trained for 9600 multi-runs and tested for 320 runs with correct classification for 314 runs giving an identification accuracy of 98.12%.

5.5.4 Power Swing Due to Adjacent Line Switching Line switching can be done in electrical power network due to maintenance and line outage because of fault. Table 5.5 shows power swing cases due to switching of line L2 and L3 from ON to OFF and vice versa with variation in system parameters. Due to switching of adjacent lines, power swing scenario will be generated at the relay of line L1 and it can mal-operate, if it is not trained for the same. The SVM model has been trained for 1024 multi-runs and tested for 128 runs with identification accuracy of 100%.

5.6 Data Mining Using SVM Classifier for Fault Cases

95

Table 5.5 Power swing cases due to adjacent line switching Power system element/parameter

SVM training patterns for power swing cases

SVM testing patterns for power swing cases

Variation in parameter

Numbers of variation

Variation in parameter

Numbers of variation

Line L1

Permanently ON

1

Permanently ON

1

Line L2 and L3 switching

ON and OFF from OFF and ON conditions, respectively

4

ON and OFF from OFF and ON conditions, respectively

4

Compensation level 0% at the middle of (uncompensated), transmission line L1 10, 20, 30% (compensated)

4

0% (uncompensated), 10, 20, 30% (compensated)

4

Power flow angle (z) 0, 5, 10, 15

4

8

1

Power swing inception angle (PIA)

0°, 45°, 90°, 135°

4

30°, 75°

2

Load connected at MEB

50, 75, 100, 125%

4

50, 75, 100, 125%

4

1024

Total testing patterns

128

Total training cases for power swing due to line switching Total correct data classification (−1)

128

Identification accuracy

100%

5.6 Data Mining Using SVM Classifier for Fault Cases 5.6.1 Solid Faults on Transmission Line Table 5.6 shows generation of fault data by simulating various types of faults on line L1 with a wide variation in fault contexts. The SVM architecture has been trained for total 9600 data sets, and validation has been performed for 960 test cases. The proposed algorithm has been also tested for unseen fault data. The testing patterns have been varied by changing the fault location, power flow angle, and fault inception angle as shown in Table 5.6. It has been verified that correct classification has been achieved for 952 cases with disturbance identification accuracy of 99.16%.

5.6.2 Faults During Power Swing Distance relays must be blocked during power swing to ensure reliability of the power transfer. On the other hand, it must generate the trip signal when the fault is

96

5 Sequence-Space-Aided Disturbance Classifier …

Table 5.6 Fault cases for fault on transmission line L1 Power system parameter

SVM training patterns

SVM testing patterns

Variation in parameter

Numbers of variation

Variation in parameter

Numbers of variation

Fault type

L-G, LL, LL-G, LLL-G

10

L-G, LL, LL-G, LLL-G

10

Fault location F L (Km)

5, 30, 50, 70, 90

5

15, 60

2

Compensation level 0% (uncompensated), 10, 20, 40% (Compensated)

4

0% (uncompensated), 10, 20, 40% (compensated)

4

Fault resistance RF ()

0, 5, 10, 20

4

0, 8

2

Power flow angle

5, 10, 15

3

8, 12

2

Fault inception angle FIA (z)

0°, 45°, 90°, 135°

4

0°, 30°, 80°

3

9600

Total testing patterns for fault

960

Total training patterns for fault Total correct data classification (+1)

952

Identification accuracy

99.16%

detected. Due to the symmetric nature of signals during the power swing, symmetrical faults are very much difficult to be detected. To test the proposed algorithm, several asymmetrical and symmetrical faults are simulated on line L1 during diverse swing conditions as shown in Table 5.7. In all simulated cases, initially power swing is generated by switching linear and nonlinear loads at MEB as shown in Table 5.7. Due to switching of load, the power swing scenario is generated at the relay on line L1. When the power swing condition is prevailing in the system, different kinds of symmetrical and asymmetrical faults are applied on line L1 in order to validate the proposed algorithm. The developed algorithm has been analyzed by applying all ten types of faults at various locations on line L1 ranging from close-in faults to far end faults. The proposed algorithm is also verified for the high resistance transient and permanent faults with wide variation in power system context. The proposed scheme is also validated for CT saturation during the fault. The compensation level has been also varied from 0% (uncompensated) to 40% which incorporates wide range of nonlinearity in the system under consideration. The SVM training is performed for 9600 multi-runs followed by testing on 320 test cases. The correct classification (+1) has been achieved for 318 test cases reflecting an identification accuracy of 99.37%. The disturbance identification accuracy is found improved as compared to existing algorithms for fault detection during power swing as outlined in Sect. 5.8.

5.7 Result Discussions

97

Table 5.7 Evaluation of fault during power swing on line L1 Case

Power system SVM training patterns for element/parameter fault during power swing Variation in parameter

SVM testing patterns for fault during power swing

Numbers Variation in of parameter variation

Numbers of variation

Compensation level at the sending 0% 4 end of transmission line L1 (uncompensated), 10, 20, 30% (compensated)

0% 4 (uncompensated), 10, 20, 30% (Compensated)

Fault cases (with and without CT saturation)

Power swing cases

Fault location (km)

5, 30, 50, 70, 90

5

50, 80

2

Fault type

L-G, LL, LL-G, LLL-G

10

L-G, LL, LL-G, LLL-G

10

Fault resi. RF ()

0, 5, 10, 20

4

8

1

Power flow angle

5, 10, 15

3

10

1

Load variation

50, 75, 100, 125%

4

25, 50, 80, 110%

4

Load switching

OFF to ON

1

OFF to ON

1

bus at which load is connected

MEB

1

MEB

1

9600

Total testing cases

320

Total training cases for faults inception during power swing Total correct data classification (+1)

318

Identification accuracy

99.37%

5.7 Result Discussions 5.7.1 Power Swing Detection Figure 5.6 shows the sample test case of power swing due to load switching. The power swing can be observed from Fig. 5.6a after application of disturbance at 0.5 s. As seen in Fig. 5.6b, as the disturbance is symmetrical, the positive-sequence component of the fault current rises above I T 1 which triggers the sampling process as shown in Fig. 5.6c. SVM outputs –1 recognizing disturbance as power swing in the system as shown in Fig. 5.6d and blocks the trip signal generation. Figure 5.7 shows the validation of proposed algorithm for increase in mechanical input by 50% to the generator with 10% compensation at 0.5 s. The waveforms of sequence components are shown in Fig. 5.7b. The model signal M(k) goes high each time when Qj (k) crosses its respective preset threshold I Tj . The proposed algorithm responds correctly during each scan of disturbance as shown in Fig. 5.7d. Figure 5.8 shows the validation of proposed algorithm during L-G fault applied on adjacent line L2 at 0.5 s and removed at 0.51 s by the relay of the same line. The

98

5 Sequence-Space-Aided Disturbance Classifier …

Fig. 5.6 Validation for 50% load switching at MEB with 8% compensation and PFA = 12° at 0.5 s

absolute values Qj (k) rises above the preset threshold as shown in Fig. 5.8b. When the fault is cleared, the zero-sequence components of line currents become zero and variation in positive-sequence component of line current has symptoms of power swing, and hence, the proposed SVM-based algorithm classifies the disturbance correctly. The model signal M(k) (Fig. 5.8c) and response of SVM are shown in Fig. 5.8d which is used to block the trip signal. Figure 5.9 shows the validation of proposed algorithm during switching of parallel line L2 from ON to OFF at 0.5 s. The waveform of line current can be observed from Fig. 5.9a and that of absolute values from Fig. 5.9b. The model signal M(k) is generated as shown in Fig. 5.9c whenever the positive-sequence component crosses its respective threshold limit to start the sampling process for one complete cycle.

5.7 Result Discussions

99

Fig. 5.7 Validation for 50% change in mechanical input to generator at 0.5 s

The sampled data set is applied to SVM-based classifier. The response of SVM-based disturbance classifier recognizes the applied disturbance as swing in the system and outputs −1 as shown in Fig. 5.9d indicating the reliability of protective actions.

5.7.2 Validation for Fault Cases Figure 5.10 shows the validation of proposed algorithm for the LL-G fault applied at 0.5 s. The response of proposed algorithm is shown below.

100

5 Sequence-Space-Aided Disturbance Classifier …

Fig. 5.8 Validation for fault isolation of adjacent line

When the fault is applied, fault current rises abruptly and phase of currents change due to current inversion as the fault location is after the series compensation (Fig. 5.10a). The absolute value of sequence components Qj (k) computed using Eq. 5.6 rises above the preset threshold value as shown in Fig. 5.10b. When Qj (k) crosses the threshold value I Tj , model signal M(k) goes high as shown in Fig. 5.10c. The rising edge in M(k) triggers the sampling process for one complete cycle (80 samples), and then it is applied to trained SVM architecture for testing purpose. As shown in Fig. 5.10d, SVM outputs +1 indicating correct classification of disturbance as a fault in the system and hence relay generates the trip signal.

5.7 Result Discussions

101

Fig. 5.9 Waveform during switching of line L2

5.7.3 Validation for Fault During Power Swing In order to realize the fault dynamics during power swing, a mechanical disturbance has been applied at 0.5 s (at 2000 sample) which generates the power swing as shown in Fig. 5.11. The waveform of line currents is shown in Fig. 5.11a. The absolute value of sequence components Qj (k) (Fig. 5.11b) increases above the threshold value. Hence, model signal M(k) goes high (Fig. 5.11c), this initiates the sampling process. The sampled data in form of feature vector is applied to SVM architecture. It recognizes the disturbance as power swing and responds as –1 (Fig. 5.11d). Afterward, at 1.2 s (4800 sample number) a solid LLL-G fault is applied at 50 km in the presence of power swing, on the line L1 because of which line current and

102

5 Sequence-Space-Aided Disturbance Classifier …

Fig. 5.10 Validation for LL-G fault applied at 60 km with RF = 8, FIA = 0°, PFA = 12° with 30% compensation level applied at 0.5 s

absolute value of sequence component go high. The model signal M(k) goes high again when Qj (k) crosses its respective preset threshold I Tj and initiates sampling. The sampled signals are applied to trained SVM model continuously. The proposed algorithm recognizes the disturbance as fault and responds +1 as shown in Fig. 5.11d. Once fault is detected in the system, the output of SVM is locked everlastingly.

5.8 Comparative Analysis The identification accuracy (η) ´ is summarized in Table 5.8 as shown below for all cases of fault as well as power swing. The developed scheme gives identification accuracy of 99.53 and 99.46% in case of fault and power swing, respectively. The

5.8 Comparative Analysis

103

Fig. 5.11 Validation for fault during power swing scenario

overall efficiency is found to be 99.50% which shows effectiveness of proposed classifier scheme in practical scenario. Table 5.9 shows comparative analysis of proposed algorithm with the existing schemes applied for the discrimination between power swing and faults. It can be narrated that overall efficiency of proposed algorithm is found superior as compared to the existing algorithms. Methods [95, 98, 100, 104] are investigated for uncompensated transmission line. Also, the accuracies obtained from these methods are less as compared to proposed algorithm. In [97], the variation in compensation level is limited to 0–15% and it is also not validated for high resistance faults. Traveling wave-based schemes as outlined in [99, 103] show better performance as compared to proposed SVM-based classifier but it is found that scheme outlined in [99] is validated for 20% fix compensation and the one outlined in [103] is validated for only two levels 30 and 70% fix compensation. The proposed algorithm outlined in this

104

5 Sequence-Space-Aided Disturbance Classifier …

Table 5.8 Overall accuracy for fault and swing conditions Condition

Fault

Power swing

Total cases

Cases

Number of training cases

Number of test cases

SVM classifier response

Overall identification accuracy (η) ´ %

(+1)

(−1)

Fault on line L1

9600

960

956

4

99.58

Fault during power swing

9600

320

318

2

99.37

Change in mechanical input

2400

384

0

384

100

Load switching

512

288

0

288

100

Line switching

1024

128

0

128

100

Fault isolation on adjacent line

9600

320

6

314

98.12

32,736

2400

2388 (correct classification)

99.53

99.46

99.50

research has been validated for wide range of power system contexts with varying degree of compensation ensuring classifier accuracy of 99.50% along with diversified pattern of feature vectors applied to SVM classifier scheme. From the outlined comparative analysis, advantages of developed scheme can be summarized as follows: • Remarkable improvement in the accuracy of disturbance classification as compared to existing algorithm because of tenfold cross-validation technique used in SVM. • Fast and reliable discrimination of disturbance during transient stage. • The developed scheme has been validated for compensation level between 0 and 40%; thus, it can be applied to both compensated and uncompensated transmission line. • It perfectly detects high resistance fault, fault with CT saturation, and symmetrical fault during power swing indicating sensitivity, stability, and reliability of the protective system under consideration.

5.9 Effective Outcome This research presents a novel sequence-space-aided SVM classifier for discrimination between fault and power swing for protection of series compensated transmission

Three-phase line currents

1

0.5

Process signal

Sampling frequency (kHz)

Required post-fault data (cycle)

0.5

4

Three-phase line currents

DFT- and ANN-based

NR

Low

Not affected

Not investigated

97.5%

Response time (cycle)

Calculation burden

High resistance fault (RF )

Degree of compensation

Identification accuracy

99.11%

0–15% (TCSC-based)

Not investigated

Low

NR

Performance parameters (NR indicates not reported)

DFT- and SVM-based

[97]

Comparative methods

[95]

Methodology

Parameter

99%

Not investigated

Not affected

Low

1

0.5

1

Three-phase line currents

Phase-space-based

[98]

100%

20% (Fix)

Not Investigated

High

1–2

1

10

Three-phase voltages and currents

Traveling wave and MMG-based

[99]

Table 5.9 Performance comparison of proposed algorithm with existing schemes

95%

Not investigated

Not affected

High

0.5

0.5

1

Three-phase voltages and its harmonics

Kalman filter and SVM-ANN-based

[100]

NR

30 and 70% (TCSC-based)

Not affected

High

NR

1

1

Three-phase voltages and currents

DFT and Universal pilot relaying

[103]

NR

Not Investigated

Not affected

High

2

1

1

1

4

Sequence components of line currents

MFCDFT and SVM-based

Proposed method

99.50%

0–40% (in step of 10%)

Not affected

Low

1

Three-phase voltages and currents

Hilbert transform and RFFD-based

[104]

5.9 Effective Outcome 105

106

5 Sequence-Space-Aided Disturbance Classifier …

line. The proposed algorithm utilizes current signals for computation of absolute values of sequence components. The 400 kV power system network is developed in PSCAD/EMTDC, whereas algorithm is designed and validated using MATLAB software. The phasor estimation of input signals has been performed using MFCDFTbased technique to derive sequence-space of modeled network. The derived feature vectors are applied for training of SVM architecture. Extensive simulations have been performed on total 32,736 multiple run cases. The regulation parameters (C, σ ) of SVM have been extracted using tenfold cross-validation technique, and validation has been done for 2400 test cases with wide range of system dynamics. Moreover, the proposed scheme is also validated for symmetrical and asymmetrical fault in the presence of power swing including CT saturation during the fault. The proposed technique is proved to be superlative with identification accuracy of 99.50% showing remarkable improvement in the performance of protection scheme as applied to series compensated transmission line.

Chapter 6

Auto-Reclosing Scheme with Adaptive Dead Time Control Based on Synchro-Check Principle

6.1 Introduction Series compensation plays important role in smart power grid to improve power transfer capacity and voltage profiles. Majority of faults occurring in the grid are transient in nature. However, discrimination between transient fault and permanent fault are contemporary problems in the field of distance protection of transmission line. Auto-reclosure is one of the foremost solutions for the same. This research presents development of auto-reclosure scheme using modified full-cycle discrete Fourier transform (MFCDFT) with adaptive dead time control. The fault detection logic is based on monitoring impedance trajectory in the R-X diagram of distance relay. The discrimination between transient and permanent fault is done using differential voltage across the contacts of circuit breaker. Reclosing instance is identified by comparing absolute value of positive-sequence voltage, phase difference, and frequency difference with respective preset threshold using synchronism check relay. The proposed scheme has been validated by simulating power system considering wide variation in fault location, fault clearance angle, fault resistance, load flow, and compensation level in PSCAD/EMTDC. The algorithm for auto-reclosing scheme is validated in MATLAB software. The real-time validation is also performed using digital signal controller on laboratory prototype. The outcome of simulation and emulation indicates substantial reduction in dead time indicating the effectiveness of the developed scheme. The critical requirements for digital protection of series compensated transmission line (SCTL) are accurate and fast fault identification and clearance in order to maintain power system stability. Majority of the faults occurring in the transmission line are transient in nature and cleared by successful re-energization of the line using auto-reclosure technology. This would be appreciated if the fault is of a transient nature or it is an arcing fault. However, if it is a persisting fault, excessive breaking duty is undesirably applied to the interrupter unit. Recently, numerous auto-reclosing techniques have been developed by technocrats for proper discrimination between © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 U. Patel et al., Futuristic Trends in Numerical Relaying for Transmission Line Protections, Energy Systems in Electrical Engineering, https://doi.org/10.1007/978-981-15-8465-7_6

107

108

6 Auto-Reclosing Scheme with Adaptive …

permanent and transient fault as outlined in auto-reclosure technology review in Sect. 1.7. In this research, as outlined in the previous chapters, phasor estimation of input analog signals has been performed by MFCDFT [24] which yields fast and accurate estimation of input signal required for high-speed auto-reclosing. A new adaptive multi-phase auto-reclosure based on synchro-check measurement is introduced in this work. Different aspects that influence the reclosure performance are simulated in PSCAD environment and validated in MATLAB. PSCAD is more purposely designed for simulating transient phenomenon as in this case for multi-machine auto-reclosing scheme. Furthermore, it is commercial software designed for power system simulation contains a built-in library for capturing different types of fault dynamics [122]. These things are more tedious to do in MATLAB. However, in order to implement any complex control algorithms, it can better be done in MATLAB [121]. For real-time implementation, programming in PSCAD using FORTRAN language is more complex as compared to m-file coding in MATLAB. In the proposed work, different types of voltage and current waveforms reflecting fault dynamics are captured from the PSCAD with a sampling rate of 4 kHz (80 samples/cycle) for a system frequency of 50 Hz. The captured waveforms in digital form are then migrated in MATLAB software for phasor estimation using MFCDFT algorithm. Further, this phasor estimated values are applied to auto-reclosing algorithm for adaptively taking relaying decision. Hence, the use of PSCAD for capturing fault dynamics followed by scripting of m-file in MATLAB for implementation of auto-reclosing algorithm is ideal for real-time validation. The test results including real-time tests are investigated and a suitable adjustable dead time for the proposed auto-reclosure is introduced in the case of transient faults. The proposed algorithm is able to ascertain reclosing time from knowledge of the post-fault voltage phasor at the measurement point after clearance of transient fault. Numerical relay should be capable of processing digital and analog input signals in order to improve response of protective systems [120]. The implementation of the proposed algorithm is performed in the numerical processor written in embedded ‘C’ using ATmega 328P AVR microcontroller provided by ATMEL. It contains digital signal controller (DSC) core in addition to on-chip peripherals like 8—channel 10 bit ADC, analog comparator, two-cycle multiplier, and digital communication ports on a standalone integrated package which offers throughput of 20 MIPS (million instructions per second) at 20 MHz for real-time performance [135]. The response time of the emulated reclosing system is found around one cycle time during transient faults after clearance of fault. On the other hand, it blocks reclosing command during permanent fault avoiding potential problems of system insecurity and instability. Section 6.2 outlines system modeling of one of the Indian power system networks. Section 6.3 reveals the proposed methodology of auto-reclosing scheme followed by hardware implementation in Sect. 6.4. The result analysis of the proposed algorithm during transient and permanent fault conditions has been illustrated in Sect. 6.5. The real-time validation of developed scheme is presented in Sect. 6.6 followed by research outcome in Sect. 6.7.

6.2 System Modelling

109

6.2 System Modeling The system modeling of an Indian power network with two generators G1 and G2 connected by 220 kV, 120 km long transmission line has been shown in Fig. 6.1. The power line is realized using the Bergeron line models in PSCAD [122]. Realization of generators, transformer, and transmission line is done according to specifications outlined in Appendix in Sect. 6.8. The system under study implements mainly user-defined components coded in Fortran along with conventional data flow models available in PSCAD. Ratings of instrument transformers (ITs) have been decided based on power flow in the power system network. The current and voltage signals stepped down by the ITs are fed to relay through MFCDFT block. Potential transformer (PT) of bus and capacitance voltage transformer (CVT) of line provide voltage signals to ensure the system healthy condition through synchro-check relay during auto-reclosing operation. The proposed technique has been validated for numerous temporary and solid ground faults by scripting the algorithm outlined in the next section using MATLAB. In order to validate the proposed technique, extensive simulations have been performed by varying fault context like L-G, LL, LL-G, LLL-G (10 types) at varying locations ranging from 0.1 to 110 km to cover the effect of close-in fault and far end zone fault. Moreover, faults are simulated with a different fault clearance angle (0°, 30°, 45°, 90°, 135°, 180°), power flow angle (5°, 10°, 15°), and fault resistance (0.01–300 ). The load in power system is varied by adjusting the variable load angle between the two generators. The voltage and current signals are taken for numerous simulation cases using multi-run facility in PSCAD/EMTDC. The proposed algorithm is successfully validated in MATLAB for discrimination between transient and permanent faults as outlined in the next section. In order to implement the proposed auto-reclosing scheme in real power system, the real-time implementation of the proposed algorithm has been performed using AVR ATmega 328P controller on laboratory prototype. The real-time validation has been performed for all 10 types of fault with the variation in fault location and fault resistance at different fault inception angles. During real-time validation, the generator is synchronized with infinite busbar through a transmission line. The generator and line parameters are also shown in Appendix in Sect. 6.8. The detained discussion on the proposed methodology is presented in the next section. 13.8/220 kV

∆Y 13.8 kV

TL (120km)

B1 CT

G1

PT

B2

G2 ∆Y

R CVT

Fig. 6.1 System modeling in PSCAD

Fault

110

6 Auto-Reclosing Scheme with Adaptive …

6.3 Proposed Fault Detection and Auto-Reclosing Technique Fault detection and auto-reclosing algorithm is divided into two sections: (i) Fault detection using MFCDFT and (ii) proposed auto-reclosing algorithm. Figure 6.2 represents logical flow of the proposed algorithm. For phasor estimation, current and voltage samples measured by ITs are acquired from PSCAD for different simulation cases. For the detection of fault, sliding window of 83 samples has been selected for estimating the phasor values at any instance using MFCDFT [137]. It is further used to estimate RMS value and phase angle, respectively, for fundamental complex phasor after removal of all the harmonics and decaying DC component as outlined in Sect. 3.2.2. The derived phasors are applied to adaptive numerical relaying scheme to calculate fault resistance at the time of

Fig. 6.2 Proposed technique for auto-reclosing

6.3 Proposed Fault Detection and Auto Reclosing Technique

111

fault using slope tracking method as presented in Sect. 3.4. The fault resistance is subtracted from total impedance which ultimately yields in actual line impedance at the time of fault. In order to facilitate relaying decisions, the estimated line impedance is compared with quadrilateral relay setting and if it falls within the characteristics, then the proposed numerical relaying technique issues the trip signal. Subsequently, auto-reclosing function is to be initiated during the transient fault and it is to be blocked during permanent fault. After detection of fault, trip signal is issued by relay operates breaker B1 (Fig. 6.1). When the contacts of CB are isolated, breaker auxiliary switches are opened which are further used for auto-reclosure operation. The main relay (R) is an auto-reset relay which resets in few milliseconds (less than 40 ms). AND logic initiates ‘dead time’ of auto-reclosure (79) by checking relay reset and CB contact isolation conditions. In conventional auto-reclosing scheme, dead time is kept fixed in the order of 5–30 cycles (0.1–0.6 s) depending on system voltage [108]. Normally, it is provided by a separate timer which provides some additional time delay after extinction of secondary arc which ultimately results in system instability. The conventional auto-reclosure scheme applies the reclosing signal after the completion of fixed dead time interval. In the proposed work, the main emphasis is given to the reclosing operation with adaptive dead time control. In order to make the dead time adaptive, initially, samples of voltage signals of line CVT and bus PT are acquired by data acquisition system as shown in Fig. 6.2 to derive system healthy condition using synchro-check relay. It takes the samples from CVT and PT continuously as shown in Fig. 6.1 to calculate phasors of differential voltages (phase wise) VX Y Z (k) as given by Eq. 6.1. When circuit breaker trips, the differential voltage increases from zero to its nominal value as shown in Fig. 6.3a. VX Y Z (k)∠θD = VB (k)∠θ B − VL (k)∠θL

(6.1)

where VX Y Z (k)∠θD = voltage across the contacts of circuit breaker (of each phase, i.e.,VX (k), VY (k) and VZ (k)), VB (k)∠θB = bus voltage from PT, VL (k)∠θL = line voltage from CVT. The phasor estimation of phase difference θD at any kth instance is represented by A(k). The proposed algorithm uses positive-sequence component of these voltages obtained using the equation, ⎤ ⎤ ⎤⎡ ⎡ V1(k) V x(k) 1 a a2 1 ⎣ V2(k) ⎦ = ⎣ 1 a 2 a ⎦⎣ V y(k) ⎦ 3 V z(k) V0(k) 1 1 1 ⎡

(6.2)

where, a = −0.5 + 0.866i. Figure 6.3a shows the waveform of voltage across the contacts of circuit breaker for L-G fault applied at 0.2 s (800 sample number) which is cleared at 0.3 s (1200 sample number). The absolute value of the positive-sequence component, Q(k), can be obtained using,

112

6 Auto-Reclosing Scheme with Adaptive …

Fig. 6.3 Waveforms during transient L-G fault a differential voltage across contacts of circuit breaker V x (k), b phasor estimation of Q(k) and A(k), c frequency estimation F(k), and d Reclosing signal R(k)

 1  Q(k) =  N

k  n=k−N +1

   V1(n)  

(6.3)

where k = in-progress sample number of Q(k). Figure 6.3b shows the variation in Q(k) and A(k) which indicates gradual increment from zero to nominal value after tripping of circuit breaker. In the proposed

6.3 Proposed Fault Detection and Auto Reclosing Technique

113

algorithm, the fundamental frequency is derived using MFCDFT from the original complex phasor to calculate frequency difference between bus and line voltages. In the proposed reclosing scheme, the system healthy condition is derived when following three conditions are satisfied as specified in [137]. (i)

If the voltage at both sides of CB are equal or if the differential phasor Q(k) is less than threshold voltage V T which is set to 10% of rated phase voltage (12.7 kV). (ii) The phase difference between bus side and line side voltage, A(k) should be within threshold limits of ±30°(PT1,2 ). (iii) The frequency difference between bus side and line side voltage, F(k) should be within frequency threshold limit of F T = 2 Hz. The voltage magnitude and angle difference check characteristic for high-speed auto-reclosing supervision are shown in Fig. 6.4. It can be observed that voltage magnitude should be less than 0.52 pu and the phase difference should be within ±30° for conducting auto-reclosing action. In the proposed algorithm, these limits are set to 0.1 pu and ±30° for avoiding erroneous judgment. The frequency difference F(k) is also compared with frequency threshold (F T ) as shown in Fig. 6.3c. When all the three conditions (magnitude, phase difference, and frequency) are within their respective thresholds, then it indicates system is absolutely healthy. This indicates that the fault is transient in nature and no longer persisting. Absolute value Q(k) is compared with preset threshold V T as shown in Fig. 6.3b to derive model signal for magnitude, i.e.,

Fig. 6.4 Voltage magnitude and angle difference check characteristic for high-speed auto-reclosing supervision

114

6 Auto-Reclosing Scheme with Adaptive …

 Mm(k) =

1 0

Q(k) ≥ VT Q(k) < VT

(6.4)

Similarly, the phasor estimation of angle A(k) and F(k) and also compared with their respective threshold limits to derive the model signal for phase and frequency, respectively, using the equations, 

1 |A(k)| > PT 0 |A(k)| ≤ PT  1 F(k) ≥ FT M f (k) = 0 F(k) < FT

M p(k) =

(6.5)

(6.6)

The adaptive reclosing signal R(k) can be derived as shown in Fig. 6.3d at the negative going of all the above model signals derived from Eqs. 6.4, 6.5 and 6.6 using negative edge triggered S-R flip-flop with S = 1 and R = 0 when all parameters are within acceptable standards. The waveforms of the line side CVT voltage for the same L-G fault are shown in Fig. 6.5. It can be observed that transient fault is applied at point P and in response to this relay issues trip signal at instance Q to isolate the system. Hence, the primary arc establishes as contacts of the circuit breaker are separated. The primary arc diminishes in few cycles and their remains secondary arc which exhibits re-striking characteristics. The duration of primary arc and secondary arc depends on energy available in the power system network, system voltage, and frequency. The secondary arc extinguishes completely within few cycles and there remains a very small negligible system frequency voltage component on the line, which is due to electrostatic

Fig. 6.5 Waveforms of line side CVT voltage during L-G transient fault applied at 0.2 s

6.3 Proposed Fault Detection and Auto Reclosing Technique

115

coupling between faulted phase and the two healthy phases. The applied transient fault is cleared at instance R and hence voltage rises to recovery voltage and hence magnitude, phase, and frequency difference starts decreasing which are used to issue reclosing signal. It is obvious that during the dead time, capacitive and inductive coupling from the other two phases induces a voltage into the open phase conductor which feeds secondary arc. The magnitude of recovery voltage due to capacitive coupling of a line is given by [3], Vr =

C1 − C0 Vm √ (2C1 + C0 ) 3

(6.7)

where C1 Line positive-sequence capacitance. C0 Line zero-sequence capacitance. Vm Maximum system phase-to-phase voltage. The recovery voltage induced due to capacitive coupling produces secondary arc current. Its effect is negligible in the uncompensated transmission line [138]. The effect of the mutual coupling will be profound when the line is equipped with compensation. In order to reduce secondary arc current, shunt reactors are used to damp the effect of mutual coupling in the compensated transmission line. Before proceeding to the next section, the concepts of dead time, permanent fault time, and reclaim time should be cleared. Dead Time: Dead time is defined as the interval between two events; one is opening of breaker and the other is reclosing of breaker to resume transmission of electric power [108]. It depends on system voltage and circuit configuration. As outlined in Sect. 6.2, in order to avoid erroneous judgment, in conventional auto-reclosing scheme, dead time is kept fixed in the order of 5–30 cycles (0.1– 0.6 s) for a system frequency of 50 Hz [108]. Normally, it is provided by a separate timer which provides some additional time delay even after extinction of secondary arc which ultimately results in system instability. Whereas, in the proposed scheme, no need to fix minimal dead time in order to make it adaptive. In the proposed scheme, system healthy condition is monitored by synchro-check relay as explained in Fig. 6.2 (algorithm). Hence, in the proposed scheme, the required dead time is estimated adaptively to maintain high stability of protected system. Permanent Fault Timer: As given in many literatures, any fault with duration more than 0.1 s can be considered as permanent fault. In the proposed scheme for safety viewpoint, permanent fault timer is set to 1 s so that it can be identified properly. Hence, the fault with duration of less than 1 s is considered as transient fault. If the fault persists for longer duration than stipulated time (Q(k) > V T ), thus permanent

116

6 Auto-Reclosing Scheme with Adaptive …

fault timer (1 s) times out and blocks the reclosing command. However, one can set their own limit for permanent fault timer depending on the system to be protected. Reclaim Timer: Reclaim time is the time interval between generation of reclosing signal by the reclosure and instance of resetting of auto-reclosing (79) scheme for monitoring existence of new fault in the power system. During reclaim time, reclosing signal will be continuously applied to the control circuit and after completion of the same, reclosing signal will be removed because of resetting of auto-reclosing system. It can be set between 0.5 and 100 s depending on operating voltage and network configuration [136]. In the proposed algorithm, the reclaim timer is set to 15 s for safety view point and to quickly relieve the burden from the auto-reclosure thus avoiding continuously duty on the same. When auto-reclosure generates the reclosing signal, then ‘reclaim timer (T R )’ is initiated. When the reclaim timer times out, the auto-reclosure scheme (79) resets after clearance of transient fault. If any fault occurs after the timeout of reclaim timer, then it is considered as new fault. After reclosing of CB, if fault reappears before completion of reclaim time, then relay senses the fault and opens the CB. However, because of opening of CB before completion of reclaim time, auto-reclosure enters into the lockout state. Hence, CB remains open everlastingly until manual closing. The reclosing signal is applied to the phase(s) in which voltage across the contacts of the circuit breaker is more than zero. Hence, the proposed scheme is applied to single-phase auto-reclosing scheme for L-G fault and three-phase auto-reclosing during multi-phase faults. In all existing EHV substations, bus PT is present for voltage measurement and protection purpose (over/under voltage). Moreover, CVT is also available in all the EHV transmission line to measure line side voltage for line synchronizing with bus and power line carrier communication. Hence, the proposed algorithm based on voltage difference across the contacts of CB can be implemented in the present structure of real power system. The hardware and software realization of developed scheme have been performed by generating temporary and solid ground faults with wide variation in fault and system conditions as outlined in the next section.

6.4 Hardware Implementation Dedicated digital signal controller (DSC), AVR ATmega 328P as a computational tool is employed in the present work for realization of protection scheme. To process computationally intensive algorithms; apart from the basic features like timer, serial communication, and interrupts, numerical processor should have higher clock rate, on-chip analog peripherals with higher sampling frequency and large amount of memory. ATmega 328P is equipped with large memory capacity of 2 K words of onchip SARAM, 32 K words on-chip flash memory, and 64 K words off-chip SARAM memory that is sufficient to store large program. The high-performance, 10-bit, 8

6.4 Hardware Implementation HOST COMPUTER

ANALOG INPUTS

SCC

ADC BUFFER 8 CH

117 USB CONNECTOR

DIGITAL SIGNAL CONTROLLER

OUTPUT DRIVER CIRCUIT

AUXILIARY TRIPPING & RECLOSING CIRCUIT

Fig. 6.6 Block diagram for emulation of the proposed algorithm

channels ADC has a minimum conversion time of 500 ns. The auto-sequencing capability of the ADC allows maximum of 8 conversions in a single conversion session without any CPU overhead [135]. All these specifications are appropriate to implement auto-reclosure relaying algorithm. Block diagram of hardware implementation of the proposed scheme in the form of laboratory prototype is shown in Fig. 6.6. The hardware setup consists of a signal conditioning circuit (SCC), AVR microcontroller, Intel Core i3, 3-GHz host PC with Windows 8 and relay drivers. The SCC is used to scale down the input current and voltage signals before supplying to the ADC which is available within the controller itself. This is due to the fact that the analog input supplied to the ADC must be between 0 and 3.3 V to limit noise immunity. The outputs of SCC are sampled at a rate of 4 kHz and applied to the processor for relaying decision. It generates trip signal at the time of permanent fault and issues reclose signal during transient fault. The power circuit diagram of hardware implementation of the proposed scheme in laboratory is illustrated in Fig. 6.7. For execution of algorithm, code written in C using embedded coder tool box available in MATLAB is loaded in the memory of processor to represent the auto-reclosure. The communication between PC and DSC is done by programmable Universal Serial Bus (USB) which is used to monitor the real-time measurements in PC. The written code is complied, linked, and downloaded using USB port to the program memory of the processor. Input signals are buffered in the processor before computing RMS values derived from MFCDFT algorithm to take protective decision about fault detection and auto-reclosing as discussed in the previous section. Test results of software simulation and hardware validation have been presented in the next section. In order to realize the power system network, a three-phase generator is connected with three-phase load through transmission line network. The transmission line is realized in laboratory prototype using rheostats and inductors. The system and line parameters are given in Appendix at the end of this chapter. Hardware validation has been carried out on uncompensated lines as series compensation is not possible to realize in laboratory prototype. In order to take the protective actions, the voltage and current signals are captured by instrument transformers which are applied to SCC to scale the signals at appropriate levels. The scaled analog signals are supplied to the ADC of the controllers. The proposed auto-reclosing algorithm is executed in the processor for taking protective actions as outlined in Sect. 6.2. The trip signals and reclosure signals are applied to the contactor (circuit breaker) through relay drivers and a well-developed control circuit. The

118

6 Auto-Reclosing Scheme with Adaptive …

Current Sensor probes

DSO Output

Contactor / CB

Host PC Voltage Signal Current Signal Relays with Drivers

Controller

Fault Signal

Fig. 6.7 Hardware setup of the proposed scheme in laboratory environment

voltage signals, current signals, and fault signal are captured using a high resolution digital storage oscilloscope (DSO). The hardware implementation of the proposed auto-reclosing scheme requires control circuit as shown in Fig. 6.8. The positions of all contacts in the proposed control circuit are shown with the normal de-energized condition of the line. Initially, the CB (B1 of Fig. 6.1) is energized (closed) manually by pressing the spring loaded push button (PB1) with the condition that the spring is fully charged (SW1 previously closed) and one of the CB auxiliary switches (CB1) is closed under breaker open condition. It is to be noted that the closing spring is charged electrically immediately following closing operation by 220 V AC motor (Fig. 6.8). The motor is automatically operated to keep the closing spring in a charged state. The CB is de-energized (opened) manually by using stop push button (PB2). The opening spring is charged by acquiring the energy released from closing spring during the closing operation of CB. Hence, an open-close-open sequence remains stored in CB operating mechanism. As shown in Fig. 6.8, one CB auxiliary switch (CB2) connected in series with tripping coil (TC) remains closed under the closed condition of CB. Whenever fault occurs, the fault impedance fall in the operating characteristics of distance relay R and hence it operates to closes its two contacts R1 and R2 simultaneously. With the closing of contact R1, tripping coil of CB is energized and thus opens the line breaker. Closing of R2 energizes the auxiliary relay (33). One of the contacts of auxiliary relay (33–1) provides hold on path to itself (33) as the R2 opens due to

6.4 Hardware Implementation

119

Fig. 6.8 Proposed control circuit for auto-reclosing scheme

automatic resetting (