Development of PSO-based SVM model for Fault Detection in Power Distribution Systems

Abstract

In this paper, a new mutant particle swarm optimization (mPSO) algorithm for optimizing support vector machine (SVM) parameters is propose to detect short circuit faults in power distribution systems. Further, time domain reflectometry (TDR) with pseudo-random binary sequence (PRBS) excitation has been considered to generate fault simulation datasets. The proposed technique has been tested on a typical two-lateral radial distribution network to identify ten different types of short circuit faults. To demonstrate superiority of the proposed mPSO, comparative studies of fault diagnosis have been performed using SVM whose parameters are selected using cross-validation and classical PSO. The obtained high classification accuracy and the comparative results demonstrate the superiority of the proposed mPSO in classifying short circuit faults.