AI-BASED FAULT DETECTION AND PREDICTIVE MAINTENANCE IN SMART POWER DISTRIBUTION SYSTEMS

Abstract

AI-based fault detection and predictive maintenance have emerged as transformative technologies in modern smart power distribution systems by enabling real-time monitoring, intelligent fault classification, and proactive asset health assessment across distributed electrical infrastructures. As distribution networks become increasingly digitalized through IoT sensors, phasor measurement units, smart meters, and advanced automation, traditional rule-based and periodic maintenance approaches fail to address the rising complexity, dynamic load behavior, and nonlinear fault characteristics of modern grids. Artificial intelligence, deep learning, and edge-cloud analytics offer a scalable and data-driven solution for detecting incipient faults, analyzing transient disturbances, and predicting equipment degradation before critical failures occur. The increasing availability of high-frequency electrical signals, asset condition parameters, and historical maintenance logs has accelerated the demand for intelligent systems capable of integrating heterogeneous grid data for early warning and reliability optimization. This paper presents an AI-driven predictive maintenance framework that leverages hybrid deep learning architectures, multi-sensor grid data, and edge-enhanced anomaly detection to support real-time fault localization, equipment health forecasting, and operational decision-making. The study evaluates system performance, detection accuracy, computational efficiency, and predictive reliability across diverse distribution scenarios, emphasizing the role of AI as a foundational technology for next-generation smart power distribution systems.