Comparative Analysis of Machine Learning Techniques for Detection and Classification of Stator Winding Fault

This paper presents a novel approach for condition monitoring and fault detection in synchronous generators using co-simulation and machine learning. A 2-kVA synchronous generator was modeled and simulated in Ansys Maxwell and Simplorer under healthy and turn-to-turn fault conditions. The simulation data were validated with laboratory experiments. Using a sliding window technique, various machine learning algorithms were applied to identify and classify fault conditions, with k-nearest neighbors achieving the highest accuracy of 98.6%, highlighting its effectiveness in fault diagnosis for synchronous generators. The findings of this research emphasize the potential of combining co-simulation tools with machine learning for real-time and robust fault detection in synchronous generators.