High Impedance Fault Detection in Distribution Systems: An Approach Based on Fourier Transform and Artificial Neural Networks

Abstract

Several challenges for generation, transmission, and distribution of electricity arise with the expansion of electrical Distribution Systems (DSs). Continuity and the ability to serve end consumers represent a significant challenge for companies responsible for supplying electricity. Therefore, the study of the improvement of effective fault identification techniques, more specifically of High Impedance Faults (HIFs), has grown substantially. HIFs are not identified by conventional protection as this type of fault has currents with values close to those in the steady-state condition in DSs. Furthermore, due to the electric arc, the HIFs offer danger to living beings and network devices. To the date, there is no fully effective protection to detect it. Therefore, this paper aims to propose a methodology capable of identifying HIFs and classifying power quality events. To this end, the proposed technique uses low-order harmonics extracted by Fourier Transform (FT) from the current signals registered at the substation of a DS as input data for a multilayer Artificial Neural Network (ANN). The tests consisted of simulating HIFs along with several system buses and evaluating false positives (non-HIF events) by simulating frequent events in DSs. The results proved the proposed technique is promising, with a high detection rate even with the addition of noise to the signals.