基于数据融合的混合动力汽车速度轮廓预测  

Hybrid Electric Vehicle Speed Profile Prediction Based on Data Fusion

作　　者：高凯 罗攀[1] 谢进 胡林 陈彬[1] 杜荣华 GAO Kai;LUO Pan;XIE Jin;HU Lin;CHEN Bin;DU Ronghua(College of Automotive and Mechanical Engineering,Changsha University of Science&Technology,Changsha 410114;Hunan Key Laboratory of Smart Roadwasy and Cooperative Vehicle-Infrastructure Systems,Changsha University of Science&Technology,Changsha 410114)

机构地区：[1]长沙理工大学汽车与机械工程学院,长沙410114 [2]长沙理工大学智能道路与车路协同湖南省重点实验室,长沙410114

出　　处：《机械工程学报》2024年第6期342-353,共12页Journal of Mechanical Engineering

基　　金：国家杰出青年科学基金(52325211);中瑞国际合作(52211530054);湖南省教育厅科学研究重点(21A0193);长沙理工大学智能道路与车路协同湖南省重点实验室开放基金(kfj190701);湖南省研究生科研创新(QL20230207)资助项目。

摘　　要：能量管理策略(Energy management strategy,EMS)可优化不同场景下混合动力汽车(Hybrid electric vehicles,HEV)的能量消耗,降低排放,提高燃油经济性。优化的关键之一是汽车速度轮廓的预测:预测未来区间的速度轮廓,进而计算需求功率,优化发动机和电动机的功率分配。因此提出一种融合驾驶行为和激光雷达数据的车辆速度轮廓智能预测方法,可应用于满足驾驶意愿的装备激光雷达的混合动力车辆能量优化。首先,构建一个基于门控循环单元(Gated recurrent unit,GRU)网络的驾驶意图识别模型,以从车辆状态中识别驾驶员的驾驶意图,实时考虑驾驶员的驾驶需求。其次,在智能车辆和传统车辆混行的交通场景中,依赖于车辆通信的速度轮廓预测方法可能不可用,研究的交通流速度由车辆上配备的激光雷达估计。算法采用联合概率数据关联跟踪器和交互多模型方法,实时得到前方车辆相对本车的速度,无需道路、工况等先验知识。最后,融合驾驶意图和前方车辆相对本车的速度,预测未来1 s内驾驶员期望的速度曲线。试验结果表明,GRU模型训练的准确率在85%~95%,识别的准确率可达到88%,可有效的识别驾驶员的驾驶意图。提出的方法具有良好的测速精度。预测的速度和实际速度的差距在较小范围内,可用于EMS中,进而为动力电池管理系统(Battery management strategy,BMS)合理控制动力电池的能量输出提供依据。Energy Management Strategy(EMS)can optimize energy consumption,reduce emissions and improve fuel economy of hybrid electric vehicles(HEV)in different scenarios.One of the keys of optimization is the prediction of vehicle speed profile:predict the speed profile of the future interval,and then calculate the power demand,and optimize the power distribution of the engine and motor.Therefore,an intelligent prediction method of vehicle velocity profile based on driving behavior and lidar data is proposed,which could be applied to energy optimization of lidar equipped hybrid electric vehicle to satisfy driving intention.Firstly,a driving intention recognition model based on gated recurrent unit(GRU)network is constructed to identify driving intentions from vehicle states,and driver's driving needs are considered in real time.Secondly,in the traffic scene where intelligent vehicles and traditional vehicles are mixed,the speed profile prediction method relying on vehicle communication may not be available.In this study,the traffic flow speed is estimated by lidar equipped on vehicles.The algorithm used the Joint Probability Data Association tracker and Interactive Multipul Model method to get the speed of the vehicle in front relative to the vehicle in real time,without prior knowledge of road and condition.Finally,the driver's expected speed curve in the next second is predicted by integrating the driving intention and the relative speed of the vehicle ahead.Experimental results show that the training accuracy of GRU model is between 85%-95%,and the recognition accuracy can reach 88%,which can effectively identify the driver's driving intention.The proposed method has good velocity measurement accuracy.Prediction speed and actual speed of the gap in a smaller range,can be used for EMS,then for power battery management system(BMS)provides the basis for reasonable control of power battery energy output.

关 键 词：能量管理 速度轮廓预测 数据融合 驾驶意图 激光雷达 

分 类 号：U461[机械工程—车辆工程]