Abstract: The driving range of electric vehicles is lower than traditional fuel vehicles at the same level price. The driving range problem seriously limits the promotion of electric vehicles in China. Therefore, predicting the energy consumption of electric vehicles can estimate the driving range accurately, reduce drivers' mileage anxiety and enhance consumers' purchase intention. Based on the real vehicle data collected by the electric vehicle monitoring platform, this paper proposes an energy consumption prediction method of multiple linear regression and radial basis function neural network. Firstly, the energy flow direction of electric vehicle is systematically analyzed, which mainly includes three parts: driving resistance energy consumption, braking recovered energy and accessory energy consumption. Based on the analysis of energy flow direction and considering the influence of temperature on the energy consumption, the data collected by the real vehicle are used to identify the parameters of the regression model, and the multiple linear regression model for energy consumption prediction is established. Compared with the actual value,the nonlinear error is obtained. Finally, the RBF neural network is used to fit the nonlinear residual of the regression model. The results shows that the model can improve the prediction accuracy of single model, and it is of practical significance in estimating the driving range of electric vehicles.

Key words: Electric vehicle; Energy consumption prediction; Linear regression; RBF neural network

摘要： 纯电动汽车的续驶里程低于同价位传统燃油车，续航问题严重限制了纯电动汽车在中 国的推广，因此精确预测纯电动汽车能耗，可以准确估计续驶里程，降低司机里程焦虑，提 升消费者购买意愿。文章基于纯电动汽车监控平台的实车采集数据，提出了一种多元线性回 归和径向基函数神经网络组合能耗预测方法。首先系统分析纯电动汽车能量流向，其主要包 括行驶阻力能耗、制动回收的能量和附件能耗三部分。在能量流向分析基础上，考虑温度对 整车能耗的影响，利用实车采集数据，辨识回归模型参数，建立能耗预测多元线性回归模型， 对比实际值得到非线性误差，最后利用 RBF 神经网络拟合回归模型的非线性残差。对比结果 表明，该模型能够改善单一模型预测精度不足的问题，对估计纯电动汽车续驶里程具有工程 应用意义。

关键词: 纯电动汽车；能耗预测；线性回归；神经网络