Data-Driven Intermittent Earth Fault Detection in Compensated and Isolated MV Networks

Abstract

Finnish Distribution System Operators (DSOs) have extensive experience in operating compensated or isolated Medium Voltage (MV) networks. Intermittent earth faults, which can potentially lead to permanent ones, are a common phe-nomenon in MV underground cables, which can be attributed to a variety of factors including the natural aging process of the equip-ment, electrical overstress, mechanical deficiencies, unfavorable environmental conditions, chemical pollution, moisture ingress, poor insulation, and loose connections. Condition monitoring and early detection of such faults are crucial, especially with the increasing use of underground cabling to enhance the security of electricity supply. These measures can enable DSOs to carry out preventative maintenance, which in turn can reduce system interruptions and improve the delivery of MV electricity. This research aims to explore the effectiveness ML-based techniques and supervised learning namely multilayer perceptron (MLP), support vector machines (SVM), Long short term memory (LSTM) and Decision Tree algorithms in classification and detection of intermittent earth fault in an MV 20kV distribution system.