Artificial Intelligence for Fault Detection and Diagnosis in Power Systems: Methods, Evaluation Challenges, and Deployment Considerations

This review examines artificial intelligence methods for fault detection and diagnosis in electrical power systems. It discusses conventional machine learning, deep learning, signal-processing and hybrid approaches, evaluation metrics, robustness, interpretability, cybersecurity, and deployment challenges. The manuscript highlights the importance of transparent datasets, realistic validation, and safe integration with conventional protection systems. This work is a literature-based review and does not report new experimental results.