Detection Method for High‐Impedance‐to‐Ground Faults in Low‐Resistance Grounding Systems Based on GAF ‐ CNN

Abstract

In urban medium‐voltage distribution networks, the proportion of cable laying continues to rise. In view of this, the low‐resistance grounding system has become a widely adopted technical solution in urban distribution networks due to its unique advantages of quickly cutting off faults and reducing overvoltage levels. To solve the problem of detecting high‐impedance‐to‐ground faults (HIGF), this work suggests a way to detect HIGF using Gramian Angular Fields‐ Convolutional Neural Network (GAF‐CNN). Firstly, a neutral line zero‐sequence current threshold is set to detect possible HIGF. ZSC amplitude over the threshold are turned into a one‐dimensional time series. Then, Gramian Angular Fields (GAF) are used to turn these series into two‐dimensional pictures that are labeled correctly. After that, the two‐dimensional pictures that have been labeled are fed into a Convolutional Neural Network (GAF‐CNN) so that it can learn how to tell the difference between HIGF and system asymmetry events. Finally, a simulation model built in MATLAB/Simulink is used to create a dataset and make sure the suggested method works. The results show that the method successfully detects HIGF, with a fault detection capability exceeding 3000 Ω, which solves the problem of detecting HIGF. © 2026 Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.