基于马尔可夫决策过程的电动汽车充电行为分析

针对电动汽车充电行为不确定性问题，建立了基于出行链理论的电动汽车出行及电池电量变化模型，提出了引入马尔可夫决策过程（Markov decision processes, MDP）的电动汽车用户充电行为分析方法。该方法将用户充电行为作为马尔可夫决策集，根据车辆在各区域间的转移概率构造状态转移矩阵，设置用户满意度指标作为决策过程报酬函数，通过求解有限阶段总报酬准则得到电动汽车用户在每个决策点处的最优充电决策。算例部分根据抽取电动汽车特征量数据进行马尔可夫决策过程仿真，得出充电负荷的时间与空间分布情况，通过与传统蒙特卡洛方法进行对比表明该文所提方法可以较好地模拟用户整个出行过程中的充电行为，反映充电需求的时空分布特点；同时，分析了不同区域、不同停车时长情况下电动汽车的不同充电行为，能够为电动汽车充电桩的规划建设提供参考。

Analysis on Charging Behaviors of Electric Vehicles Based on Markov Decision Processes

In view of the uncertainty of electric vehicle charging behavior, models of electric vehicle travel and battery electricity change based on trip chain theory are established, and an analysis method for charging behavior of electric vehicles by introducing the Markov decision processes（MDP） is proposed. The method takes the user's charging behavior as a Markov decision set, constructs a state transfer matrix according to the transfer probability between various regions. The user satisfaction index is set up as the reward function of the decision process. The optimal charging decision of the electric vehicle users at every decision point is obtained by solving the finite stage total reward criterion. The example is simulated by extracting characteristic data of electric vehicles, results show the time and space distribution of electric vehicle charging load. Compared with the traditional Monte Carlo method, the proposed MDP method can simulate user charging behavior more accurately and reflect the temporal and spatial distribution characteristics of charging demand. At the same time, different charging behavior of electric vehicles in different areas and different parking hours is analyzed, which can provide support for the planning and construction of electric vehicle charging piles.