An Overview of Extreme Learning Machine-Based Intelligent Fault Localization Approaches

The reliable operation of modern power systems depends heavily on the rapid detection and accurate localization of faults occurring in transmission and distribution networks. With the increasing integration of renewable energy resources, smart grid technologies, distributed generation, and advanced communication infrastructures, conventional fault localization methods face significant challenges related to network complexity, dynamic operating conditions, and measurement uncertainties. In recent years, Artificial Intelligence (AI)-based techniques have emerged as promising alternatives for enhancing the accuracy and efficiency of fault localization processes. Among these techniques, the Extreme Learning Machine (ELM) has gained considerable attention due to its fast learning capability, low computational complexity, excellent generalization performance, and suitability for real-time applications. This paper presents a comprehensive overview of ELM-based intelligent fault localization approaches developed for power system protection and monitoring. The review discusses the fundamental principles of fault localization and the theoretical foundations of ELM, including its architecture, learning mechanism, and major variants such as Online Sequential ELM, Kernel ELM, Weighted ELM, and Deep ELM. Furthermore, existing research contributions employing ELM for transmission lines, distribution networks, microgrids, renewable energy-integrated systems, and smart grid environments are systematically analyzed and compared. The paper also examines various signal processing and feature extraction techniques used in conjunction with ELM, including wavelet transforms, empirical mode decomposition, and phasor measurement unit-based approaches. Performance metrics, implementation challenges, and comparative advantages of ELM over traditional machine learning methods are critically evaluated. Finally, current research gaps and future directions are identified, highlighting opportunities in explainable artificial intelligence, federated learning, digital twins, edge computing, and hybrid intelligent fault localization frameworks. The findings indicate that ELM-based approaches offer a promising and computationally efficient solution for next-generation intelligent fault localization systems, contributing significantly to the development of reliable, adaptive, and resilient smart power grids.