基于LSTM的电动汽车剩余续驶里程预测  

Prediction of Remaining Driving Range of Electric Vehicles Based on LSTM

作　　者：王焕焕 赵慧勇 Wang Huanhuan;Zhao Huiyong(School of Automotive Engineering,Hubei University of Automotive Technology,Shiyan 442002,China)

机构地区：[1]湖北汽车工业学院汽车工程学院,湖北十堰442002

出　　处：《湖北汽车工业学院学报》2025年第1期28-32,39,共6页Journal of Hubei University Of Automotive Technology

基　　金：汽车零部件技术湖北省协同创新项目(2015XTZX0403);湖北汽车工业学院博士科研启动基金(BK201410)。

摘　　要：基于LSTM神经网络和实车运行数据,开展了电动汽车剩余续驶里程预测的研究。采用滑动平均和赋值填充等方法进行原始数据处理,通过皮尔逊相关性分析选出总电压、当前剩余电量及其与放电结束电量的差值3个特征参数,增加功率、功率变化率、车速、加速度4个特征参数,分别输入到模型进行剩余续驶里程预测和对比分析,并在行驶片段上进行验证。结果表明,7个特征参数整体误差较小,效果较好。Based on the LSTM neural network and real vehicle operation data,the research was carried out on the prediction of the remaining driving range of electric vehicles.The raw data were processed by moving average and assignment filling.Through Pearson correlation analysis,three characteristic parameters of total voltage,current remaining power,and its difference from the end-of-discharge capacity were selected.Additionally,four characteristic parameters of power,power change rate,vehicle speed,and acceleration were added.These parameters were respectively put into the model for predicting the remaining driving range and comparative analysis,with validation performed on driving segments.The results show that the seven characteristic parameters exhibit an overall small error and relatively good performance.

关 键 词：LSTM神经网络 电动汽车 剩余续驶里程 

分 类 号：U461[机械工程—车辆工程] TM911[交通运输工程—载运工具运用工程]