Umudike Journal of Engineering and Technology

The integration of Distributed Generation (DG) in distribution networks introduces bidirectional power flows and fluctuating fault currents, severely compromising the selectivity of traditional overcurrent relays. To address this problem, an adaptive Multi-Agent System (MAS) integrated with Explainable Artificial Intelligence (XAI) using Shapley Additive Explanations (SHAP) is proposed. The proposed scheme dynamically recalculates Time Multiplier Settings (TMS) and Tap Settings (TS) using decentralized, local decision-making system. This eliminates the communication delay typical of centralized SCADA architectures while providing transparent, human-readable justifications for relay coordination. The proposed model was validated on the IEEE 69-bus test system using ETAP and MATLAB. Simulation results across 16 topological scenarios demonstrate that, whereas conventional fixed-setting approaches require a compromised TMS of 0.22 (delaying operation during low-fault events), the MAS autonomously adjusts the TMS between 0.15 and 0.22. This dynamic adaptation successfully clamps the fault-clearing time to a target of 0.20 seconds, significantly minimizing electrical stress during islanded or weak-grid operations. Ultimately, the proposed hybrid MAS-XAI framework outperforms conventional adaptive schemes in operational speed, coordination accuracy, and systemic reliability.