Data-Driven Fault Detection in Power Systems: Clustering Analysis for Lightning and Switching Overvoltage

Lightning strikes and sudden overvoltage during switching operations can disrupt power networks. This work introduces a machine learning-based strategy that uses non-supervised learning and clustering algorithms to find and classify temporary faults. Data from a Simulink model of a physical system starts the operating sequence. Data synchronization and feature extraction are carefully done to prepare for clustering analysis. Transient fault occurrences are clustered using K-means and hierarchical clustering. The model is implemented in MATLAB. The actual results show that the suggested method finds and categorizes transient faults. The silhouette score shows that the model can distinguish lightning and switching problems. This work proposes an algorithmic strategy that uses data analysis to improve system robustness and proactive maintenance. This work uses machine learning and clustering analysis to improve electrical power systems’ dependability and resilience to lightning and switching overvoltage.