Advanced Clustering and Optimisation Techniques for Adaptive Protection in AC Microgrids With Hardware‐in‐the‐Loop Testing

ABSTRACT This study introduces an advanced adaptive protection approach for AC power systems, designed to address key limitations in current clustering‐based relay coordination methods. Using K‐means, hierarchical, and spectral clustering techniques, the proposed scheme classifies grid operating conditions more effectively, ensuring that relay settings adapt to different network scenarios. A key feature of this study is a practical method for identifying the optimal number of clusters and selecting the fine‐tuning relay settings using the Genetic and Tug of War Optimisation algorithms, which improves the speed of fault detection and isolation over twenty different network topologies. The proposed scheme supports standard and non‐standard Overcurrent Relay (OCR) characteristics, in which the IEC inverse‐time curve is expanded beyond its conventional pickup‐current limit to capture higher fault‐current levels in renewable‐integrated networks, ensuring faster and more selective relay operation. In Clustering 1 (Topology 1), the total tripping time was reduced to 23.15 s, while Clustering 2 (Topology 16), involving more relays, recorded 29.77 s. The total tripping time for Clustering 3 (Topology 2) was 29.34 s. Hardware‐in‐the‐Loop (HIL) testing verified the real‐time performance of the proposed scheme, showing high performance with simulation results and less than 2% deviation in relay tripping times. These outcomes demonstrate the scheme's ability to deliver reliable and responsive protection across various grid environments.