HIGH IMPEDANCE FAULT IDENTIFICATION AND DETECTION USING DISCRETE WAVELET TRANSFORM AND ARTIFICIAL NEURAL NETWORKS

Abstract

This report presents the application of neural network technique as pattern recognition to high impedance faults (HiZs) classification and location identification scheme for radial power distribution feeders. The proposed scheme is capable to obtain precise fault location estimations for both linear and non linear high impedance faults. The scheme is based on ANNs whose local measurement voltages and currents are obtained from the substation during normal and abnormal feeder operation. The fault detection and location can be estimated by a set of characteristics extracted from the voltage and current signals measured at the substation. This characteristic set is classified by an artificial neural network based scheme whose output determines fault detection at a location and its classification is also done. Discrete wavelet transformations (DWT) and artificial neural networks (ANN) have been widely used in power system research. The method proposed in this thesis uses digital signal processing, where the application of Discrete Wavelet Transform (DWT) is used for extracting the features of the distorted waveforms caused by HiZ's, and the neural network back propagation algorithm is used for fault identification. After capturing the voltage and current waveforms from MATLAB power system simulation, they are analyzed by DWT. The DWT output coefficients are converted into RMS values in various frequency ranges, then These characteristic set is fed to an ANN pattern classifier to distinguish between fault cases and nofault cases, and also to classify linear or nonlinear fault if the fault exists and to find the location of the fault in the line.