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

作　　者：邓艺璇 黄玉萍 黄周春 DENG Yixuan;HUANG Yuping;HUANG Zhouchun(Guangzhou Institute of Energy Conversion,Chinese Academy of Sciences,Guangzhou 510640,China;Key Laboratory of Renewable Energy,Chinese Academy of Sciences,Guangzhou 510640,China;University of Chinese Academy of Sciences,Beijing 100049,China;College of Economics and Management,Nanjing University of Aeronautics and Astronautics,Nanjing 210016,China)

机构地区：[1]中国科学院广州能源研究所,广东省广州市510640 [2]中国科学院可再生能源重点实验室,广东省广州市510640 [3]中国科学院大学,北京市100049 [4]南京航空航天大学经济与管理学院,江苏省南京市210016

出　　处：《电力系统自动化》2021年第21期181-188,共8页Automation of Electric Power Systems

基　　金：国家自然科学基金青年科学基金资助项目(71801114);能源基金会(美国)资助项目(G-2011-32631)。

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。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.

关 键 词：V2G 电动汽车 充放电分类 负荷预测 随机森林 

分 类 号：U491.8[交通运输工程—交通运输规划与管理] TM910.6[交通运输工程—道路与铁道工程]