Experimental Evaluation of Microgrid Energy Management Using Surrogate-Assisted Optimization on PHIL and Smart Grid Systems

In this study, we present an integrated surrogate-assisted multi-objective optimization for an energy management system, ensuring physical feasibility with real system constraints. The hybrid framework incorporates knee-guided selective physical replay and a stochastic survival strategy to maintain both convergence and diversity of the search process. The method is used to evaluate grid-forming and grid-following modes using the OPAL-RT and Lucas-Nülle platforms in three different stages to address the technical and economic performance, and the reliability of the system. The proposed method reduces 116 generated surrogate candidates to 7 physically feasible non-dominated solutions based on physical replay. In the direct evaluation stage without replay, the system achieves high renewable utilization (PV ≈ 97%), reliable load coverage (>99%), and minimal supply–demand mismatch (≈1 W), supported by controlled battery usage. In the extended EMS evaluation, the proposed method reduces the number of true evaluations from approximately 54,600 to 16,895 (≈69% reduction) while maintaining stable performance. Despite the reduction in the number of evaluations, the method preserves stable convergence behavior and a consistent Pareto spread (≈0.0124). Statistical tests, such as Wilcoxon (p≈1.9×10−6) and Friedman (p≈2.9×10−7), show a significant difference and consistent performance across runs. This demonstrates the framework’s ability to provide a compact, decision-relevant set of feasible operating solutions under real system constraints and its practical applicability to real-world EMS decision making.