Power System Fault Detection Using Explainable Artificial Intelligence Techniques

Abstract

The growing complexity of modern electrical power networks has increased the need for dependable and interpretable fault detection mechanisms. Conventional protection approaches, while effective in structured grids, often struggle to adapt to systems influenced by distributed generation and renewable energy sources. This study presents an interpretable machine learning–based framework designed to improve fault identification while maintaining decision transparency. Time-domain electrical parameters, including voltage and current measurements, are analyzed under diverse operating conditions to distinguish between normal and faulty states. An explainability layer is incorporated to clarify the reasoning behind model predictions, enabling engineers to better understand system behavior during disturbances. Simulation results indicate that the developed approach achieves high detection performance while offering meaningful insights into feature contributions. The framework therefore supports informed decision-making and demonstrates strong potential for application in intelligent power system protection environments.