基于随机森林算法的电动汽车充放电容量预测
Charging and Discharging Capacity Forecasting of Electric Vehicles Based on Random Forest Algorithm
-
摘要: 文中提出一种电动汽车充放电容量的组合预测方法。首先,基于电动汽车历史充电数据和用户参与电动汽车与电网互动(V2G)意愿的调查数据,分析车辆荷电状态(SOC)特性、出行时间特性以及用户对价格的敏感度,建立随机森林分类模型,判断车辆是否参与V2G调度,并对影响用户决策的特征因素进行重要性评估。其次,采用蒙特卡洛方法模拟电动汽车出行和充放电情况,并分别预测充放电容量。最后,以办公区为例进行仿真,对比分析多种充放电模式下的电动汽车充放电行为与负荷分布。所构建的随机森林分类模型的准确率为0.917,能够有效区分V2G计划时段内电动汽车的充放电行为,仿真结果验证了所提预测框架的有效性。Abstract: A combined forecasting method for the charging and discharging capacity of electric vehicles（EVs） is proposed. Firstly,based on the historical charging data of EVs and the survey data of users’ willingness to participate in vehicle-to-grid（V2 G）, the characteristics of vehicle state of charge（SOC）, travel time and users’ sensitivity to price are analyzed. Then, a random forest based classification model is established to determine whether the EV participates in V2 G scheduling, and the importance of characteristic factors that affect users’ decision-making is evaluated. Secondly, the Monte Carlo simulation method is used to simulate the situations of EV travelling and charging/discharging, and the charging and discharging capacities are predicted respectively. Finally, a simulation is carried out by taking an office area as an example to compare and analyze the EV charging and discharging behaviors and load distribution with multiple charging and discharging modes. The constructed random forest classification model has an accuracy of 0.917, which can effectively classify the charging and discharging behavior of EVs during the V2 G planning period. Simulation results also verify the effectiveness of the proposed forecasting framework.
下载:
