Abstract:

Artificial intelligence is revolutionizing power system operations through its faster, more precise, more intelligent monitoring and control strategies.  New generation power systems can become more stable and dependable with the use of AI in failure detection and diagnosis (FDD).  Classical FDD methods employ rule-based or model-based techniques, which can be challenged in handling real-time flexibility, nonlinearities, and massive data processing. These limitations cause delay in fault detection, mislabeling issues, and degrade diagnostic performance under complex operating conditions.  This paper presents an Machine Learning-Based Fault Detection and Diagnostic Modeling (ML-FD-DM) paradigm to address these challenges.  The framework employs robust machine learning techniques such as SVM, RF, and ANN to detect and classify fault types from past and real-time data. The model identifies small failure patterns through feature extraction, data preprocessing, and supervised learning.  The new ML-FD-DM framework is evaluated on different power system scenarios with simulated and real-world data.  It adapts to grid conditions and enhances diagnostic efficiency and accuracy.  Findings indicate that the ML-FD-DM framework performs better compared to other existing methods in terms of detection accuracy, response time, and reliability of fault classification. This indicates that fault management powered by AI can revolutionize power system fault management and facilitate the development of more durable and intelligent energy infrastructures.