ENHANCING FAULT DETECTION AND LOCALIZATION IN TRANSMISSION LINES USING ARTIFICIAL NEURAL NETWORKS AND LEARNING ALGORITHMS: A COMPREHENSIVE ANALYSIS

Abstract

Accurately detecting and pinpointing faults in power cable circuits presents a substantial challenge in ensuring the reliability and efficiency of electrical systems. The growing expansion of underground cable networks, along with the increasing complexity of modern power grids, has heightened the challenge of accurately locating faults. This challenge is further compounded by the diverse range of fault types encountered, including short circuits, insulation breakdowns, and intermittent faults, which add complexity to the localization process. This research explores the effectiveness of artificial neural networks (ANNs) in fault detection, classification, and localization. A total of 16 examples were studied by examining different power system topologies, using multiple backpropagations learning algorithms (e.g., LM (Levenberg-Marquardt) and SCG (Scaled Conjugate Gradient), and utilizing data from both endpoints of a transmission line. The findings show that using the LM learning algorithm and combining input from both terminals improves fault detection and localization in the majority of instances. Although the SCG algorithm exhibits faster convergence in fault location, both algorithms yield comparable results in fault identification and classification only. Integrating data from the remote end as additional inputs to ANNs enhances the fault localization process by offering a more comprehensive dataset for analysis.