本文提出一種先進的模組化聚落式微電網（Modular Rundling Microgrid, MRM）架構，以及一項基於能量的韌性指標，以提升多微電網系統於主電網故障期間的可靠性與適應性。MRM 架構靈感來自德國「環形村落（Rundling）」布局，具備可擴展的環狀拓撲，並由核心微電網協調內部及跨微電網之能源流動。為了量化評估並優化微電網韌性，本研究引入創新之韌性指標，著重於關鍵負載之能源連續性。此外，為進一步增強運作韌性與穩定性，本研究開發了Kolmogorov-Arnold 雙重深度 Q 網路（Kolmogorov-Arnold Double Deep Q-Network, KADDQN）策略。該方法結合Kolmogorov-Arnold 網路（KAN）強大的非線性逼近能力與DDQN的穩定性優勢，能有效管理動態系統拓撲與高維度決策空間。最後透過OPAL-RT即時模擬器建構的硬體迴圈(Hardware In the Loop, HIL)系統，充分驗證和展現所提出方法的有效性。本整合框架為新一代韌性微電網提供一種實用且智能化之能源管理解決方案。;In this study, an advanced modular rundling microgrid (MRM) architecture and an energy-based resilience index to enhance the reliability and adaptability of multi-microgrid systems during main grid outages are proposed. Inspired by the German “rundling” village layout, the MRM features a scalable, circular topology with core microgrids coordinating internal and inter-microgrid energy flows. To quantitatively evaluate and optimize microgrid resilience, an innovative resilience index is introduced, focusing on energy continuity for critical loads. To further strengthen operational robustness and stability, a Kolmogorov-Arnold double deep Q-Network (KADDQN) strategy is developed. By combining the powerful nonlinear approximation capability of Kolmogorov-Arnold network (KAN) with the stability of double DQN, the proposed method efficiently manages dynamic system topologies and high-dimensional decision spaces. Finally, the proposed method is validated using a Hardware-in-the-Loop (HIL) system built on the OPAL-RT real-time simulator, demonstrating its effectiveness. This integrated framework provides a practical, intelligent energy management solution for next-generation resilient microgrids.
