AI-Based Fault Detection and Isolation for Reliability in Modern Power Systems

Abstract

Fault detection and isolation (FDI) play a critical role in ensuring the reliability and stability of modern power systems. With increasing complexity, integration of renewable energy sources, and growing demand, traditional methods for fault management face challenges in terms of accuracy, speed, and adaptability. Artificial intelligence (AI)-based techniques offer promising solutions to these issues by enabling automated, fast, and accurate fault diagnosis through advanced data analysis and pattern recognition. This paper presents a detailed investigation into AI-driven approaches for fault detection and isolation in power systems. Techniques such as artificial neural networks (ANN), support vector machines (SVM), fuzzy logic, and hybrid models have been explored for their effectiveness in identifying faults including short circuits, line-to-ground faults, and equipment malfunctions. The study evaluates the performance of these AI algorithms using historical and simulated power system data, emphasizing robustness in noisy and dynamic environments. The proposed AI-based framework significantly improves detection speed and accuracy compared to conventional rulebased and threshold methods. It adapts to varying operating conditions and learns from evolving system behaviors, enhancing fault isolation precision and reducing downtime. Moreover, AI integration facilitates predictive maintenance and real-time monitoring, thereby improving system resilience. The paper outlines the research methodology encompassing data acquisition, feature extraction, model training, and validation. Key findings demonstrate that hybrid AI models combining fuzzy logic and neural networks outperform single-method models in fault classification accuracy. The workflow integrates sensor data preprocessing, AI inference engines, and fault localization modules. The study also discusses advantages such as adaptability, scalability, and automation, alongside challenges including data quality dependence and computational requirements. Finally, future directions include incorporating deep learning architectures, edge computing, and cybersecurity considerations to further enhance fault management systems. This work contributes to advancing reliable and intelligent fault detection in modern power systems critical for sustainable energy infrastructure.