A Joint Model-Data-driven Economic Dispatch Method for Multi-Microgrid Distribution Systems

To address the dispatch challenges and node voltage violations caused by renewable energy output fluctuations and electric vehicle load variations in multi-microgrid distribution systems， a joint model–data-driven economic dispatch and voltage control method is proposed. First， the multi-microgrid distribution system is transformed as a Stackelberg game model， in which dynamic electricity pricing is designed to improve power interactions and benefit coordination among multiple agents. An embedded active price-updating loop is introduced to enhance the nonlinear adaptability of the day-ahead dispatch model while improving computational efficiency. Subsequently， deep reinforcement learning （RL） is employed to track day-ahead dispatch commands and implement real-time decentralized voltage control. This enables the distribution system to maintain autonomous reactive power regulation capability in the presence of unmodeled power disturbances while ensuring multi-agent economic dispatch. Simulation results demonstrate that the proposed method mitigates local power imbalance and dispatch burdens caused by excessive economically driven operation of energy storage systems under peak–valley price incentives， improves power interactions among different entities in the distribution system， and achieves active voltage control under source–load uncertainty conditions.