Distributionally robust energy management for multi-microgrids with grid-interactive EVs considering the multi-period coupling effect of user behaviors

The increasing penetration of renewable energy sources (RESs) in multi-microgrids (MMGs) poses significant challenges to stable operation of the systems, and exploring grid-interactive functionalities of electric vehicles (EVs) is receiving increasing attention. However, current distributionally robust energy management models suffer from convergence inefficiencies when exposed to large amounts of historical data, and typically neglect the multi-period coupling effect of EV user behaviors, which hinder the effective utilization of the highly-potential EV resources. In this paper, a novel distributionally robust energy management model for MMGs is proposed to accommodate the uncertainties of RESs and loads, with the grid-interactive EVs operating in an efficient vehicle-to-grid (V2G) mode. Firstly, a multi-period dynamic EV-connection matrix is formulated to determine the connection and dwell times for EVs interacting with the power systems, which enables the cross-cycle continuity of SOCs. Further, the multi-period coupling uncertainties of accidental EVs disconnections are taken into account. Secondly, the Kohonen neural network-based ambiguity set is constructed without including the entire historical scenarios, where the ambiguous distribution is characterized by the representative scenarios with weights. On this basis, a two-stage distributionally robust optimization model is finally developed, which can be solved iteratively by the extended column-and-constraint generation method until the worst-case cost expectation is obtained. The proposed model was evaluated through simulations on a system comprising four interconnected microgrids from the Hainan provincial power grid. The results demonstrate that the proposed model achieves superior cost efficiency, convergence performance and robustness compared to alternative approaches.