Detection and classification of faults in renewable energy penetrated stand alone microgrids using SVM and DWT techniques

Abstract

This paper deals with the problems of fast and accurate detection of any type of faults as well as their bus location those may occur in a renewable energy sources (RESs) penetrated standalone microgrid (MG) system. The developed technique relies only on the current signal measured from every generator bus of standalone MG. The Discrete Wavelet transform (DWT) based decomposition upto 9th level of the current signal is performed to detect the fault. The detailed and approximate coefficients from the DWT are utilized to develop the kurtosis patterns. These kurtosis patterns are analyzed to detect the presence of faults in the MG. The types of faults are classified using Support Vector Machine (SVM) along with principal component analysis (PCA) based dimensionality reduction technique. The proposed approach remains independent of fault bus location and partial fault impedance as well. The fault detection scheme is made intelligent enough so that it can distinguish the variation in renewable generations and the faulty condition. The PSCAD software is used to simulate various types of fault in different buses of MG. The DWT decomposition and PCA enabled SVM of current data are utilized for detection and classification of faults as illustrated in the result section. The evidence from the obtained accuracy in detection of fault as well as its type shows the efficacy of the proposed method. Besides, the time of execution in detecting and classifying the faults shows the fastness of the proposed algorithm.