A PHYSICS-INFORMED AI-ASSISTED FRAMEWORK FOR FAULT DETECTION AND LOCALIZATION IN CONTAMINATED XLPE UNDERGROUND CABLES

Abstract

Underground power cables are widely used in modern power transmission and distribution systems due to their improved reliability, safety, and reduced environmental impact. However, insulation degradation caused by contaminants in cross-linked polyethylene (XLPE) cables remains one of the major challenges affecting the reliable operation of underground cable networks. Al-though conventional fault detection techniques have been widely employed; they often face limitations in accurately identifying and localizing faults under varying contamination conditions. This paper presents a physics-informed artificial intelligence framework for fault detection and localization in contaminated XLPE under-ground cables. Initially, a comprehensive review of conventional diagnostic methods and recent artificial intelligence techniques is carried out to identify existing research trends, challenges, and knowledge gaps. A finite element model of a high-voltage XLPE cable is then developed using COMSOL Multiphysics to investigate the influence of different contaminant materials and their locations on the electrical behaviour of the insulation system under pulse voltage excitation. Variations in electric potential, electric field in-tensity, and polarization are analysed to identify physics-based features that can support future AI-driven fault diagnosis and localization. The study demonstrates that contaminant characteristics significantly influence the electrical field distribution within the insulation, providing valuable information for intelligent condition monitoring. The proposed framework offers a systematic approach for integrating physics-based simulation with artificial intelligence and serves as a foundation for the future development of reliable cable diagnostic and predictive maintenance systems.