Real-Time Fault Classification in Electrical Networks Using Explainable Artificial Intelligence

Abstract

The increasing complexity and interconnected nature of modern electrical power systems have intensified the need for accurate and rapid fault detection mechanisms to ensure system reliability, operational stability, and uninterrupted power delivery. This study presents a real-time fault classification framework for electrical networks using Explainable Artificial Intelligence (XAI) techniques to enhance fault diagnosis accuracy, transparency, and decision-making efficiency in smart power systems. The proposed framework integrates real-time sensor data acquisition, signal processing methods, machine learning algorithms, and explainable AI models to identify and classify various electrical faults, including line-to-ground, line-to-line, and three-phase faults under dynamic operating conditions. A data-driven methodology was employed in which voltage, current, and frequency parameters were continuously monitored and analyzed using intelligent classification models. Explainability techniques such as feature importance analysis and interpretable decision models were incorporated to improve the transparency and reliability of fault prediction outcomes. Experimental and simulation results demonstrated significant improvements in fault classification accuracy, response time, and detection reliability compared with conventional fault analysis approaches. The proposed XAI-based framework also reduced false classification rates and enhanced operator understanding of critical fault conditions through interpretable analytical insights. Furthermore, the intelligent monitoring system exhibited strong adaptability and scalability for integration into smart grid and industrial power distribution infrastructures. The study concludes that explainable artificial intelligence provides an effective and reliable solution for real-time fault classification in electrical networks by enabling accurate predictive analysis, transparent decision-making, and enhanced operational security. The findings contribute to the advancement of intelligent power system protection technologies for resilient and sustainable electrical network management.