插电式混合动力汽车车速预测能量管理策略  

Energy management strategy for speed prediction of plug in hybrid electric vehicles

作　　者：赵爽 罗勇[1,2] 孙强 ZHAO Shuang;LUO Yong;SUN Qiang(Key Laboratory of Advanced Manufacturing Technology for Automobile Parts,Chongqing University of Technology,Chongqing 400054,China;Ningbo Shenglong(Group)Co.,Ltd.,Technical Center,Ningbo 315199,China)

机构地区：[1]重庆理工大学汽车零部件先进制造技术教育部重点实验室,重庆400054 [2]宁波圣龙(集团)有限公司技术中心,浙江宁波315199

出　　处：《重庆理工大学学报（自然科学）》2023年第9期79-87,共9页Journal of Chongqing University of Technology：Natural Science

基　　金：国家自然科学基金项目(51305475);重庆市教委科学技术研究项目(KJQN201801143);重庆市自然科学基金面上项目(cstc2019jcyj-msxmX0308);重庆理工大学重大科研项目同步开展基础及应用基础研究项目(2022TBZ003)。

摘　　要：针对混合动力汽车预测型能量管理策略中车速预测精度不足,使得燃油经济性降低问题,提出一种基于小波分解(wavelet decomposition,WD)和双通道卷积神经网络(convolutional neural network,CNN)的车速预测方法,以提高车速预测精度。采用小波分解将原始车速序列分解为多个分量,以降低原始车速序列的非平稳性;将各分量送入2个并行的卷积神经网络进行特征提取,经特征融合后送入长短期神经网络(long short-term memory neural network,LSTM)进行预测;将各分量预测结果叠加,得到最终的车速预测结果。最后,基于车速预测结果,以燃油经济性最优为目标,建立了基于模型预测控制的能量管理策略,对预测时域内动力源输出进行滚动优化。仿真结果表明:在CLTCP工况下,所提出的车速预测方法预测精度比CNN-LSTM神经网络模型高58.96%;比动态规划策略的油耗增加13.3%,比基于规则的策略油耗降低了18.98%,从而验证了该车速预测方法和预测能量管理策略的有效性。In view of the lack of speed prediction accuracy in the predictive energy management strategy of hybrid electric vehicles,the fuel economy is reduced,a speed prediction method based on wavelet decomposition(WD)and dual channel convolutional neural network(CNN)is proposed to improve the speed prediction accuracy.Firstly,wavelet decomposition is used to decompose the original vehicle speed sequence into multiple components to reduce the nonstationarity of the original vehicle speed sequence;Secondly,each component is sent to two parallel convolutional neural networks for feature extraction,and then sent to long short-term memory neural network(LSTM)for prediction after feature fusion;Then,the final speed prediction result is obtained by superimposing the prediction results of each component.Finally,based on the results of vehicle speed prediction,the energy management strategy based on model predictive control is established to optimize the power source output in the prediction time domain.The simulation results show that the prediction accuracy of the speed prediction method proposed in this paper is 58.96%higher than that of the CNN-LSTM network model under CLTCP conditions.The fuel consumption of the predictive control strategy proposed in this paper is 13.3%higher than that of the dynamic programming strategy,but the fuel consumption is 18.98%lower than that of the rule-based strategy,which verifies the effectiveness of the speed prediction method and the predictive energy management strategy.

关 键 词：能量管理 车速预测 小波分解 卷积神经网络 长短期神经网络 

分 类 号：U469.72[机械工程—车辆工程]