AI-DRIVEN FAULT DIAGNOSIS IN MODERN ELECTRICAL POWER DISTRIBUTION NETWORKS

Abstract

The increasing penetration of distributed energy resources, smart meters, power electronic devices, and automated control systems has significantly changed the operation of modern electrical distribution networks. Although these developments improve flexibility and efficiency, they also make conventional fault-diagnosis techniques less effective under changing network conditions. This study proposes an artificial intelligence-driven framework for rapid detection, classification, and localization of faults in modern power distribution networks. Electrical measurements, including three-phase voltage and current signals, sequence components, signal energy, and transient characteristics, are processed to extract meaningful fault-related features. A hybrid convolutional neural network–long short-term memory (CNN–LSTM) model is employed to identify both spatial and temporal patterns associated with different fault conditions. The proposed approach is evaluated through simulated normal and faulty operating conditions involving single-line-to-ground, line-toline, double-line-to-ground, three-phase, and high-impedance faults. Variations in loading conditions, distributed generation penetration, fault resistance, fault location, and measurement noise are also considered.