基于驾驶行为的纯电动汽车剩余续驶里程预测  

Prediction of Remaining Mileage of Electric Vehicle Based on Driving Behavior

作　　者：李骏 李继秋 孙亚诚 单丰武 曾建邦 LI Jun;LI Jiqiu;SUN Yacheng;SHAN Fengwu;ZENG Jianbang(School of Electromechanical and Vehicle Engineering,East China Jiaotong University,Nanchang 330013;School of Automobile Studies,Tongji University,Shanghai 201804;New Energy Vehicle Corporation,Jiangxi Jiangling Motors Group,Nanchang 330013)

机构地区：[1]华东交通大学机电与车辆工程学院,南昌330013 [2]同济大学汽车学院,上海201804 [3]江西江铃集团新能源汽车有限公司,南昌330013

出　　处：《机械工程学报》2025年第4期262-272,共11页Journal of Mechanical Engineering

基　　金：国家自然科学基金(51206171);载运工具与装备教育部重点实验室基金(KLCE2021-08);江西省自然科学基金(20192BAB206033);江西省创新创业大学生训练计划省级重点(202110404029)资助项目。

摘　　要：针对目前纯电动汽车在实际行驶过程中仪表显示剩余续驶里程预测不准这一难题,提出一种基于驾驶行为的纯电动汽车剩余续驶里程预测方法。首先,对车辆行驶数据进行片段切分并进行数据维度扩充,利用最大信息系数法提取与纯电动汽车平均百公里能耗相关的驾驶行为特征参量;然后,针对行驶片段数据探讨特征参量选取方式对K-Means聚类结果的影响,并将驾驶行为分类标签、电池荷电状态(State of charge,SOC)、行驶工况、温度等参数作为预测输入,使用BP和长短时间记忆(Longshort-termmemory,LSTM)网络模型分别对纯电动汽车行驶过程中剩余续驶里程进行预测,发现相比未考虑驾驶行为,考虑驾驶行为时模型预测结果与真实值之间的误差更小,且相比BP神经网络模型,LSTM网络模型预测结果与真实值之间的误差更小;最后,结合纯电动汽车实际行驶过程中的行驶数据对预测结果进行验证,发现原车剩余续驶里程预测准确度确定系数由0.8539提高至0.9822。所取得研究成果对提高纯电动汽车实际行驶过程中的剩余续驶里程预测准确度,减轻驾驶员里程焦虑具有重要的意义。Aiming at the problem of inaccurate prediction of remaining driving range of electric vehicle in actual driving process,a prediction method of remaining driving range of electric vehicle based on driving behavior is proposed.Firstly,segment the vehicle driving data and expand the data dimensions,and extract the driving behavior characteristic parameters related to the average energy consumption of 100 km of electric vehicles using the maximum information coefficient method;Then,based on the driving segment data,the influence of the selection method of characteristic parameters on the K-Means clustering results is discussed.With the driving behavior classification label,SOC,driving conditions,temperature and other parameters as the prediction input,BP and LSTM network models are used to predict the remaining driving range of the electric vehicle during driving.It is found that the error between the model prediction results and the real value is smaller when driving behavior is considered than when driving behavior is not considered,Compared with the BP neural network model,the error between the predicted results of LSTM network model and the true value is smaller;Finally,the prediction results are verified with the actual driving data of the electric vehicle,and it is found that the determination coefficient of the prediction accuracy of the remaining driving range of the original vehicle is increased from 0.8539 to 0.9822.The research results obtained in this study are of great significance to improve the prediction accuracy of the remaining driving range in the actual driving process of electric vehicles and reduce the driver's mileage anxiety.

关 键 词：纯电动汽车 剩余续驶里程预测 驾驶行为 最大信息系数法 K-MEANS聚类算法 

分 类 号：U463[机械工程—车辆工程]