1 CHAPTER 1 INTRODUCTION 1.1 BACKGROUND The increased use of non-linear loads and the occurrence of fault on the power system have resulted in deterioration in the quality of power supplied to the customers. The commonly occurring power quality disturbances include voltage sag, voltage swell, harmonics, transients and flicker. In order to improve the quality of electric power supplied, it is essential to detect and identify the power quality problem. This has resulted in significant advances in power quality monitoring equipment. Power quality disturbance waveform recognition is often troublesome, as it involves a broad range of disturbance categories or classes. This necessitates the development of automatic recognition system to classify the disturbance waveform. This thesis explores the application of signal processing techniques and Artificial Neural Networks for power quality disturbance recognition. In this chapter, different types of power quality problems and their causes are discussed. The problem statement and objective of the research work are also presented. Further, the research methodology adopted in this work is briefly stated.
2 1.2. POWER QUALITY PROBLEM Any deviation in the perfect sinusoidal waveform of voltage or current that can result in failure or mis-operation of customer equipment is called as power quality problem. The most common power quality disturbances that occur in electrical power distribution system include the following (Math H. J. Bollen 2001): Voltage Sag Voltage swell Transients Harmonics and Flicker The characteristics and causes of each of the power quality disturbances are presented in the following subsections: Important standards dealing with power quality issues have been developed by the International Electro technical Commission (IEC) and Institute for Electrical and Electronics Engineers (IEEE). These standards define the acceptable limits for the power quality disturbances (Roger C. Dugan et al 2004). 1.2.1 Voltage sag Voltage sag is the brief decrease in r.m.s voltage from 10% to 90% of the rated system voltage for the duration of 0.5 cycle to 1 minute (IEEE 1159). Figure 1.1 shows a waveform with sag.
3 1 Sag 0.8 0.6 0.4 Amplitude 0.2 0-0.2-0.4-0.6-0.8-1 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 Time (s) Figure 1.1 Voltage sag The common causes of voltage sag are short circuits (faults) in the electric power system, motor starting and large load addition in the utility area. Sag causes computers and other sensitive equipment to malfunction. 1.2.2 Voltage swell Voltage swell is the brief increase in r.m.s voltage from 110% to 180% of the rated system voltage for the duration of 0.5 cycle to 1 minute. A waveform with swell is shown in figure 1.2. 1.5 Swell 1 0.5 Amplitude 0-0.5-1 -1.5 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 Time (s) Figure 1.2 Voltage swell They are caused due to the abrupt reduction in load on a circuit, with a poor or damaged voltage regulator and due to damaged or
4 loose neutral connection. The occurrence of voltage swell may damage the electrical equipment. 1.2.3 Harmonics Harmonics are sinusoidal voltage or current having frequencies that are whole multiples of the frequency at which the supply system is designed to operate. IEEE standard recommends a limit of 3 % harmonic distortion for an individual frequency component and 5 % for Total Harmonic Distortion (THD). Figure 1.3 shows a waveform with harmonics. 1 Harmonic 0.8 0.6 0.4 Amplitude 0.2 0-0.2-0.4-0.6-0.8-1 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 Time (s) Figure 1.3 Harmonics They are caused due to the usage of equipment with non-linear characteristics. These harmonic distortions cause the computers to malfunction and cause motors, transformers and wires to heat up excessively. 1.2.4 Transients Transients are sudden but significant deviation from normal voltage or current levels. It can be a unidirectional impulse of either polarity or a damped oscillatory wave with the first peak occurring in
5 either polarity. Transients typically last from 200 millionth of a second to half a second. Figure 1.4 shows a signal with transient. 4.5 Transient signal 4 3.5 3 Amplitude 2.5 2 1.5 1 0.5 0-0.5 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 Time (s) Figure 1.4 Transient signal They are caused due to lightning, switching on and off equipment, starting of heavy loads and due to capacitor switching. The transients can erase or alter computer data resulting in difficulty to detect computational errors. Transients can destroy electronic circuitry and damage electrical equipment. 1.2.5 Flicker Voltage flickers are systematic variations of the voltage envelope or a series of random voltage changes (IEC 161-08-05). The presence of flicker is shown in figure 1.5. 1.5 Voltage flicker 1 0.5 Amplitude 0-0.5-1 -1.5 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 Time(sec) Figure 1.5 Voltage flicker
6 Voltage flickers are caused by arc furnaces, welding machines, alternate and reciprocating loads and wind generators. These voltage disturbances will cause various problems in power systems like electrical equipment damage, malfunction of computers and other sensitive equipment. 1.3 ADDRESSING THE POWER QUALITY PROBLEM To address the power quality problem, the following tasks need to be carried out. Monitoring the electric power systems Capturing the disturbance waveforms Classifying the captured waveforms into known categories and Proposing remedial action based on the type of disturbance. 1.4 PROBLEM STATEMENT The increased requirements on supervision and control in modern power systems make power quality monitoring a common practice for utilities. In order to improve the quality of power, electric utilities continuously monitor power delivered at customer sites. Power quality monitoring is necessary to categorize and analyze the power quality disturbances.
7 Power quality monitoring is the process of gathering, analyzing and interpreting raw measurement data into useful information. The process of gathering data is usually carried out by continuous measurement of voltage and current using power monitoring recorders over an extended period. Disturbance waveforms are recorded continuously using power monitoring instruments, producing yearly data in giga byte range. The large amount of data cause several practical problems in the storage and communication of a data from local monitors to the central processing computers. The monitoring instruments allow a transient event to be detected by fixing a threshold value. The disadvantage of these kind of instruments is that their operation strongly depends on the calibration of thresholds and also require large amount of memory to store samples of disturbances occurred. A more suitable solution to reduce the data size would be to extract automatically the distinctive features of the disturbance waveforms. This thesis explores the suitability of various signal processing techniques to extract relevant features from the signals. Moreover, the power monitoring instruments lack the ability to distinguish between events. The user has to categorize the collected events for further analysis manually. There is usually a high volume of data to be processed and classified. This makes it very tedious and time consuming to interpret the data. It is highly desirable, if the analysis is automated. This thesis proposes Artificial Neural Network (ANN) based models for automatic classification of power quality disturbances.
8 1.5 AIM AND OBJECTIVE The aim of the research work is to develop an automatic disturbance recognition system based on signal processing techniques and Artificial Neural Networks. To achieve this aim, the following objectives have been identified. Development of Transformation-based and Amplitude-frequency estimation based signal processing tools for the extraction of the distinctive features of the power system disturbances. Development of suitable feature selection technique for dimensionality reduction. Analysis and localization of disturbances in the noisy environment. Development of automatic disturbance recognition system using Artificial Neural Networks. Performance evaluation of the developed recognition systems. 1.6 RESEARCH METHODOLOGY Research methodology is the way of conducting the research work to achieve the desired objectives. There are four distinct phases of study in this research work: Phase 1 involves literature review; Phase 2 deals with the different methods in disturbance data collection.
9 Phase 3 focuses on signal processing based feature extraction methods and phase 4 comprises the development of ANN based models for automatic disturbance recognition and their performance evaluation. An outline of the research methodology is presented below: 1.6.1 Phase 1: Literature review Phase 1 is aimed at reviewing the present status of art on power quality disturbance detection, analysis and recognition and identifying appropriate techniques for developing automatic disturbance recognition system. 1.6.2 Phase 2: Data collection In this phase, the data required to develop the automatic disturbance recognition system are collected/generated. 1.6.3 Phase 3: Feature extraction In this phase, signal processing techniques are applied for extracting the relevant features of the disturbance waveforms. Further, feature section techniques are applied for dimensionality reduction. 1.6.4 Phase 4: Development of ANN model This phase focuses on the development of Artificial Neural Network based models for the automatic classification of disturbances. The performance measures are used for evaluating the developed ANN model.
10 1.7 SCOPE OF THE RESEARCH WORK The present work focuses on the development of automatic power quality disturbance recognition system based on signal processing techniques and Artificial Neural Networks. The research work will cover the application of signal processing techniques for disturbance detection, feature extraction and analysis of disturbances under normal and noisy conditions. Development of neural network models for categorizing the disturbance waveforms will also be carried out. 1.8 ORGANIZATION OF THE THESIS The organization of the thesis is outlined below: Chapter 2 presents the literature review on the application of signal processing techniques and Artificial Neural Networks for the detection, analysis and classification of power quality disturbances. Chapter 3 describes the research methodology adopted in the study. In this chapter, four different phases followed in this research work have been discussed. The methodologies adopted in each stages of the study are also described. Chapter 4 presents the details of transformation based signal processing techniques namely, Wavelet Transform and S-Transform and their application to power quality disturbance detection and identification. The detection of disturbances in the noisy environment is discussed. An improved wavelet based denoising technique for the detection of disturbance with low SNR is proposed.
11 Chapter 5 explains the details of Wavelet Transform based feature extraction technique and its application to the analysis and detection of power quality disturbances. The automatic disturbance recognition using Artificial Neural Network (ANN) trained by back propagation algorithm is presented. A novel feature selection technique based on wavelet entropy for classifying the disturbance waveforms is also explained in this chapter. Finally the results of simulation and conclusion of this investigation based on performance measures are given. In chapter 6, feature extraction using S-Transform is explained. The detection of disturbances based on statistical curves is discussed. Modified S-Transform for enhancing the energy concentration of the disturbances is proposed. This chapter also discusses the development of Radial Basis Function Neural Network (RBFNN) for classifying the disturbance waveform and the performance of the network under noisy environment. Chapter 7 explains the details of Hilbert Transform, one of the amplitude-frequency estimation based signal processing technique. The trajectory of the analytic signal based on Hilbert Transform is used for the detection of the disturbances. Feature extraction based on the envelope of the signal and automatic recognition using RBF neural network are presented. In chapter 8, the application of Hilbert-Huang Transform and Hilbert spectrum for the analysis of power quality disturbances is discussed. The development of RBF network using Fuzzy-C-Means
12 (FCM) clustering algorithm is also discussed. The simulation results are presented to demonstrate the validity of the proposed method. Chapter 9 presents the conclusion of the research work and suggestions for future research.