Identification and Detection of Electrical Fault Location in Underground Cables Using Machine Learning Algorithms

Abstract

Underground power cables are key components of advanced power systems because of their safety, reliability, and beautiful advantages. However, the identification and localization of electrical faults in these cables remain challenging because of their difficulty in nature and the complication of fault behaviour. This paper introduces a machine learning–based approach using the random forest classifier and random forest regressor algorithms for efficient fault detection and location estimation in underground power cables. Voltage and Current data are collected from both sending and receiving ends under different fault conditions, such as conductor-to-ground (CG), conductor-to-conductor (CC), double conductor-to-ground (DCG), and three-conductor to ground (3CG) faults. The random forest classifier is developed to identify fault types based on extracted electrical features, while the random forest regressor predicts the exact fault distance using ohm’s law–based calculation. The model is developed and validated using MATLAB Simulink and hardware setup. Tests confirm that this system performs with greater accuracy and stability than previous methods, quick and validated fault detection in underground power cables.