«Zahir Javed Paracha MEngstud., MS(TQM), B.Sc., G.Cert (Commercialisation) for the degree of DOCTOR OF PHILOSOPHY School of Engineering and Science ...»
Design and Development of Intelligent
Computational Techniques for
Power Quality Data Monitoring and
A thesis submitted
Zahir Javed Paracha
MEngstud., MS(TQM), B.Sc., G.Cert (Commercialisation)
for the degree of
DOCTOR OF PHILOSOPHY
School of Engineering and Science
Faculty of Health, Engineering and Science
To my (late) father
The most important requirement of power system operations is sustained availability and quality supply of electric power. In Electrical Power Distribution System (EPDS), non-linear loads are the main cause of power quality (PQ) degradation. The PQ problems generated by these non-linear loads are complex and diversified in nature. The power system which is not capable to handle nonlinear loads faces the problem of voltage unbalance, sag, swell, momentary or temporary interruption and ultimately complete outage of EPDS.
The PQ problems have motivated power system engineers to design and develop new methodologies and techniques to enhance EPDS performance. To do so, they are required to analyse the PQ data of the system under consideration. Since, the density of the monitoring nodes in EPDS is quite high, the aggregate analysis is computationally involved. In addition, the cost involved with the PQ shortcomings is significantly high (for domestic consumers and rises exponentially for industrial consumers), hence it also becomes mandatory to project /predict the undesired PQ disturbance in EPDS. This will provides power system engineers to formulate intelligent strategy for efficient power system operations.
This objective of the research is to identify and exploit the hidden correlation in PQ data with minimal computational cost and further use this knowledge to classify any PQ disturbance that may occur. The technique and the methodology developed in this research employ the actual PQ data of United Energy Distribution (UED) system in Victoria which is owned by Jemena Ltd. Australia. This power distribution network consists of 27 zone substations and is responsible for
I delivering electricity to over 600,000 customers in Melbourne Australia (United Energy Limited, Last updated 2002).
The techniques applied in this work analyse the PQ parameters for 66/22kV zone substations. The PQ data of the UED system is carefully pre-processed to highlight the principal determinants of undesired PQ disturbances with the help of principal component analysis (PCAT) technique model. The processed data is used for classification of major PQ disturbances such as power factor, sag, swell and harmonics in EPDS. The tool used for classification of these disturbances is Artificial Neural Network (ANN).
Further this research also investigates the power distribution system behaviour considering the relationship of main PQ disturbance harmonics in conjunction with the other major PQ parameters i.e. voltage unbalance, sag, swell and frequency. The work is aimed at applying fuzzy clustering techniques to marginalise out the undesired harmonics from the PQ data having multiple PQ attributes of the UED system.
The results reveal that the nuisance of dimensionality in PQ data can be evaded with the help of PCAT model. It also efficiently classifies the main PQ disturbances with appreciable accuracy of 93%-95%. The results establish the fact that in a resource constraint environment harmonics in EPDS can be used as single basis for PQ analysis.
This thesis aims to provide a framework for PQ data analysis. The methodology adopted here is not only helpful for united energy network but also applies to the EPDS across the board.
“I, Zahir Javed Paracha, declare that the PhD thesis entitled “Design and Development of Intelligent Techniques for Power Quality Data Monitoring and Management” is no more than 100,000 words in length including quotes and exclusive of tables, figures, appendices, bibliography, references and footnotes. This thesis contains no material that has been submitted previously, in whole or in part, for the award of any other academic degree or diploma. Except where otherwise indicated, this thesis is my own work”.
Zahir J. Paracha Dated the 29th day of April, 2011
First and foremost I am grateful to God Almighty Allah for giving me the strength and courage to complete this project. It is a pleasure to thank all those who have provided assistance and support during the period of this research.
It is my great honour to thank my supervisor Prof. Dr. Akhtar Kalam for his all-time support for me and my family. He has been instrumental in helping me to stay focused and determined towards my research and was at my rescue for any problem which I faced during my stay in Victoria University. This thesis would never have been possible without his continuous guidance, technical mentoring, inspiration and valuable suggestions. I have learnt many things from Prof. Kalam and greatly admire his dedication and hard work. He has become an invaluable mentor.
I am greatly indebted to Mr. Peter Wong, Manager Electricity Asset Management, Jemena Ltd. Australia for providing the necessary support and data for the experimental work at 66/22kV zone substations of United Energy network throughout the project. Many thanks to Mr. Raman Luthra (Protection Engineer, Jemena) for the field visits, sharing information and hands on training with the united energy distribution network (UED) power quality monitoring system.
Special thanks to my colleagues Mr. Ahmed M. Mehdi and Mr. Waqas Ahmed for their useful technical input and innovative ideas.
I greatly acknowledge the initiative and drive of our new Head of School Associate Professor Iwona Miliszewska in fostering a culture of quality research in the School of Engineering and Science.
Tony Davis and Mr. Rahamathulla Mohammad from Victoria University for their free and informal discussion and support in completion of this work. I wish to thank my previous and present colleagues: Mr. Hassan Al-Khalidi, and Dr. M.T.O.
Amanullah, Mr. Faizan Dastgeer, Mr. Nur Ashik Hidayatullah, Mr. Mohammedreza Pourakbar and Mr. Rizwan Ahmad for their support and friendship. I owe special thanks to Ms. Harpreet Kaur Bal and Mr. Hadeed A. Sher for technical discussions and assistance in formatting and editing. I would also like to acknowledge the support of staff at office of the Postgraduate Research and the faculty office. I am thankful to the Department of Innovation Industry Science and Research (DIIR) and Victoria University for the APA scholarship.
Finally I owe a lot to my parents, wife, children, father-in-law, brothers and sisters for their love, prayers, encouragement and support. I started this project after getting motivation from my late father Prof. S. M. Javed Paracha. He greatly valued education and research and wanted me to explore the field of engineering. While he only lived for 3 months after the start of this research project my mother Zeenat Begum came to my rescue with the message to fulfil the dream of my father. She has stood by me each day and it is because of her great support and prayers that I have been able to achieve this milestone. I am greatly thankful to my wife Saeeda for being tolerant and looking after our children throughout this challenging period of research project. Big thanks to my father in law Mr. Abdul Hameed Butt for keeping my morale high throughout this work.
TABLE OF CONTENTS
LIST OF FIGURES
LIST OF TABLES
LIST OF ABBREVIATIONS AND SYMBOLS
1.3 SUMMARY OF MAIN CONTRIBUTIONS AND PUBLICATIONS
JOURNAL PUBLICATIONS AND BOOK CHAPTERS
INTERNATIONAL CONFERENCE PUBLICATIONS
NATIONAL CONFERENCE PUBLICATIONS
EARLIER PUBLICATIONS THAT FACILITATED THIS RESEARCH
1.4 OUTLINE OF THE THESIS
2.2 ELECTRICAL POWER DISTRIBUTION SYSTEM (EPDS)
2.3 PQ ISSUES IN EPDS
2.4 IMPORTANCE OF PQ
2.4.1 Utility Perspective
2.4.2 Consumer’s Perspective
2.4.3 Equipment Manufacturers’ Perspective
2.5. PQ DISTURBANCES
2.6 PQ MONITORING
2.6.1 Conventional Methods of PQ Monitoring
2.6.2 PQ Monitoring In Present Power Distribution Networks
2.6.3 PQ Monitoring in Future Power Distribution Networks
2.7 PQ STANDARDS
2.7.1 IEEE 1195 Standards
2.7.2 IEC 61000 Series of Standards
2.7.3 New Zealand Standards
2.7.4 Commonly used PQ Standards in Saudi Arabia
3.2 PQ MONITORING SYSTEM
3.3 EXPERIMENTAL SETUP FOR MEASUREMENT OF POWER QUALITY DISTURBANCES............. 37
3.4 POWER QUALITY DATA
3.5 PRE-PROCESSING OF PQ DATA
3.5.1 Principal Component Analysis Technique (PCAT)
3.5.2 Steps for Implementation of Principal Component Analysis
(a) Plot of PQ data and Calculation of data Mean
(b)Shifting of PQ Data to Mean
(c)Establishment of New Data Axis
(d)Calculation of Data Covariance
(e)Calculation of Eigen Values and Eigen Vectors from Covariance Matrix45
3.6 EXPERIMENTAL RESULTS AND DISCUSSION
COMPUTATIONAL ANALYSIS OF PQ DATA USING NEURAL NETWORKS........... 51
4.2 NEURAL NETWORK METHODOLOGY
4.2.1 Feed Forward Back Propagation (FFBP)
4.3 PQ DISTURBANCES IN EPDS
4.3.1. Power Factor
4.3.2. Sag and Swell
4.4 IMPLEMENTATION OF NEURAL NETWORKS ON PQ DATA
4.4.1 Use of PCAT MODEL for data refining
4.4.2 Estimation of Power Factor
4.4.3 Estimation of Sag and Swell
4.4.5 Estimation of Harmonics
CLUSTERING OF UNDESIRED PQ DATA USING FUZZY ALGORITHM
5.2 PQ MEASUREMENT AND FEATURE SELECTION
5.4 MATHEMATICAL ALGORITHM
5.4.1 Fuzzy C- Mean Clustering
5.4.2 GK based Clustering
5.5 EXPERIMENTAL RESULTS
CONCLUSION AND FUTURE WORK
6.2 FUTURE WORK
Figure 2-1 Sag, Swell and Normal waveform (Paracha & Kalam, 2009)
Figure 2-2 Normal and harmonic waveforms (Paracha & Kalam, 2009)
Figure 2-3 Waveforms of fundamental, 3rd and 5th Harmonic (Paracha et al., 2009a)
Figure 2-4 Normal interruption and surge waveforms (Paracha & Kalam, 2009).. 18
Figure 2-5: GE Curve for Voltage Flicker (IEEE Standards Number 141-1993, 1994)
Figure 3-1 PQ Monitoring Set-up (Jemena Electricity 2008)
Figure 3-2 PQ centralised recording system (Jemena Electricity 2008)
Figure 3-3 United Energy Distribution PQ Monitoring (Jemena Electricity 2008).. 36 Figure 3-4 Experimental set up of 66/22kV zone substation in Melbourne Victoria
Figure 3-5 Quality disturbances at 66/22kV Glen Waverley zone substation (Information Technology Industry Council, 2008)
Figure 3-6 Plot of PQ data in two dimensions
Figure 3-7 Plot of mean of PQ data
Figure 3-8 New axes showing the maximum variation of PQ data
Figure 3-9 Final dimension of PQ data based on Eigen vectors
Figure 3-10 Block diagram of PCAT model for processed PQ data
Figure 3-11 Plot of Eigen Values Vs Principal Component
Figure 3-13 Plot of PQ data in 2 dimensions (Paracha et al., 2009c)
Figure 3-14 Loss of PQ data (Paracha et al., 2009c)
Figure 4-1 A simple architecture of FFBP-NN
Figure 4-2 The architecture of Feed Forward Back Propagation NN
Figure 4-3 A simple architecture of recurrent layer neural network
Figure 4-4 The architecture of recurrent neural network (RNN)
Figure 4-5 Architecture of 2 Layer Feed Forward back Propagation Neural Network
Figure 4-6 The training error curve for estimation of power factor using FFBPNN (Paracha et al., 2009c)
Figure 4-7 The training, testing and validation error curves for swell using FFBPNN (Paracha et al., 2009d)
Figure 4-8 The training, testing and validation error curves for sag using FFBP-NN (Paracha et al., 2009d)
Figure 4-9 The training, testing and validation error curves for sag and swellusing RNN (Paracha et al., 2009d)
Figure 4-10 The training, cross validation and testing error curves for harmonic currents in Phase A (Paracha et al., 2009b)
Figure 4-11 The training, cross validation and testing error curves for harmonic currents in Phase B (Paracha et al., 2009b)
Figure 4-12 The training, cross validation and testing error curves for harmonic currents in Phase C (Paracha et al., 2009b)
Figure 5-2 Total harmonic distortion and corresponding voltage sag and swell..... 77 Figure 5-3 Total harmonic distortion and corresponding values of frequency........ 78 Figure 5-4 Intelligent PQ monitoring strategy (Paracha & Kalam, 2010)
Figure 5-5 FCM clustering for harmonics and voltage unbalance
Figure 5-6 GK based extended fuzzy clustering for harmonics and voltage unbalance
Figure 5-7 FCM clustering for harmonics and sag/swell
Figure 5-8 GK based extended FCM for harmonics and sag/swell
Figure 5-9 FCM clustering for harmonics and frequency
Figure 5-10 GK based extended FCM for harmonics and frequency
Table 2-1 PQ disturbances, their typical duration and voltage magnitude in per unit for electrical power system as defined in IEEE-1159-95 (Faisal, 2007)