电动汽车的能耗预测对于车辆路径规划与充电行为至关重要.提出一种考虑充电行为的多模型融合能耗预测方法,首先构建基于实车稀疏数据与有限参数的能耗计算模型,在此基础上构建充电行为模型,分析并提取能耗强相关的充电行为特征,最后基于长短期记忆循环神经网络(Long short-term memory neural network,LSTM)搭建能耗预测模型.使用实车数据对所提方法进行验证,结果表明,该方法可以精准预测相同车型不同起始电池荷电状态(State of charge,SOC)、不同温度、不同时间段下的汽车能耗,均方根误差(Root mean square error,RMSE)为1.27,与现有方法相比,RMSE至少降低4.5%.
Energy Consumption Prediction Method for Electric Vehicles by Integrating Charging Behavior with Data-driven Method
Accurately predicting the energy consumption of electric vehicles is essential for efficient vehicle path planning and charging.A multi-model fusion method for energy consumption prediction that takes into account charging behavior is proposed.Firstly,based on the limited parameters and sparse real vehicle data,the energy consumption calculation model is constructed.Then,the charging behavior model is created to analyze and extract features closely related to energy consumption.Finally,long short-term memory neural network(LSTM)is used to construct the energy consumption prediction model.The method is validated with real vehicle data.Results indicate that the proposed method accurately predicts the energy consumption for the given car model with differing starting battery states of charge(SOC),temperatures,and periods.The root mean square error(RMSE)recorded is 1.27,which shows a reduction of no less than 4.5%compared to the existing methods.
Energy consumption predictionelectric vehiclescharging behaviorlong short-term memory neural network
NETL
NSTL
万方数据
维普
