Adaptive Clustering and Optimization With Hardware-in-the-Loop Validation for Enhanced Overcurrent Relay Protection in Power Distribution Networks

This study presents an adaptive clustering and optimization approach to enhance overcurrent relay (OCR) protection in power distribution networks. Network topologies were initially clustered into four groups based on total operating times using both K-means and Particle Swarm Optimization (PSO) algorithms. The optimal number of clusters was determined using validation techniques, including the Elbow method, Silhouette Coefficient, and confusion matrix analysis, ensuring strong cluster cohesion and separation under non-standard operating conditions. The clustering process revealed four distinct clusters. Subsequently, the Cuckoo Search (CS) optimization technique was applied to fine-tune relay settings within representative network topologies from each cluster, optimizing the Time Multiplier Settings (TMS) to minimize relay tripping times while maintaining system reliability. Comparative analyses demonstrated that non-standard relay characteristics significantly improved fault detection and protection coordination compared to standard ones. To validate the proposed methodologies, both ETAP simulations and Hardware-in-the-Loop (HIL) testing were conducted, achieving effective Coordination Time Intervals (CTI) in real-time scenarios. The detailed setup, procedures, and validation results of the HIL experiments demonstrate the practical applicability of the proposed scheme under realistic operating conditions. The integration of advanced clustering, metaheuristic optimization, and HIL validation offers a comprehensive and reliable framework for adaptive OCR protection in modern power systems.