A New High Impedance Fault Detection Technique Using Deep Learning Neural Network

Abstract

Because of random, nonlinear, and low-magnitude fault currents, conventional overcurrent relay dependent protection scheme has limits to detecting high impedance faults in distribution systems. This leaves the traditional relays inadequate for high impedance fault detection. This chapter presents the application of deep learning neural network technique to detect and classify high impedance faults (HIFs). Relays are based on a novel low-frequency diagnostic vector, third and fifth harmonic function. The currents and voltages signals are fed to the deep learning neural network for fault discrimination, with a deep neural network, which is trained to satisfy the image of the third and fifth harmonic of the voltage and current, differential operator, initial condition, and boundary conditions. The aim of this chapter is to design a robust deep learning neural network-based smart relay, which can predict and classify the high impedance low-current defects in radial electrical delivery systems. To determine the reliability and sensitivity of the proposed procedure, varieties of faults and system conditions were simulated. The achieved results show the validity of the developed approach.