基于分布式近似动态规划的电动汽车实时充电调度策略

To address the challenges about high computational burden, poor privacy and reliability in centralized scheduling of large-scale electric vehicles (EVs), this paper proposes a real-time EV charging scheduling strategy based on the distributed approximate dynamic programming (ADP). First, a polytope-based EV charging model is constructed, and a corresponding EV aggregation and disaggregation method based on polytope approximation is proposed, which can reduce computational complexity while ensuring sufficient optimization accuracy. Furthermore, a centralized scheduling model for aggregated EVs is established, where the model is reformulated under the Markov decision process so as to align with the practical real-time scheduling process. Next, this study proposes a distributed ADP algorithm to solve the scheduling model under stochastic environments. The value function is approximated using piecewise linear functions, and a consensus theory incorporating power limitation factors is applied for distributed optimization. Moreover, by integrating the ADP algorithm with consensus theory, a distributed piecewise linear function slope update method is proposed, which can embed empirical information about the system's stochastic factors into the value function, assisting decision-making by only utilizing local information and communication between neighboring elements. As a result, the approximate global optimal real-time EV scheduling strategy is obtained in a distributed manner. Finally, case studies on the modified IEEE 6-bus system and IEEE 118-bus system validate the effectiveness of the proposed polytope-based aggregation method, consensus-based distributed optimization method, and distributed ADP stochastic solving method. © 2026 Chin.Soc.for Elec.Eng.