Artificial Neural Network for High-Impedance-Fault Detection in DC Microgrids

Abstract

In this paper, we present a novel method for detect-ing high-impedance faults (HIFs) in DC microgrids. HIFs are more difficult to detect than other types of faults because their voltage and current values are not significantly different from those under normal operating conditions. We propose a recurrent neural network (RNN)-based method that can detect events from the temporal behaviour of a current signal, including HIFs and load changes. The method proves to be accurate, distinguishing between HIFs and other waveforms with a high score above 95% on accuracy and Fl-score metrics.