基于时空神经网络的汽车乘员舱温度预测  

Prediction of the Temperature in Vehicle Cabin Based on Spatio-temporal Neural Network

作　　者：陈伦江 苏楚奇[1,2,3] 刘珣 CHEN Lun-jiang;SUChu-qi;LIU Xun(School of Automotive Engineering,Wuhan University of Technology,Wuhan 430070,China;Hubei Key Laboratory of Advanced Technology for Automotive Components,Wuhan University of Technology,Wuhan 430o70,China;Hubei Collaborative Innovation Center for Automotive Components Technology,Wuhan University of Technology,Wuhan 430070,China)

机构地区：[1]武汉理工大学汽车工程学院,武汉430070 [2]武汉理工大学现代汽车零部件技术湖北省重点实验室,武汉430070 [3]武汉理工大学汽车零部件技术湖北省协同创新中心,武汉430070

出　　处：《武汉理工大学学报》2024年第3期134-141,共8页Journal of Wuhan University of Technology

基　　金：湖北省科技重大专项(2021AAA006)。

摘　　要：驾乘人员附近的局域温度预测是优化汽车空调系统控制并实现乘员舱区域化热管理的有效措施。针对乘员舱局域温度的动态预测问题,综合考虑了乘员舱温度场的空间特征与时间特征,并融合汽车空调的风速特征,建立了属性增强的时空图卷积网络模型(AST-GCN)。通过在实车上收集的多组数据集进行模型训练和验证试验表明,在预测较长时间范围的温度变化时,AST-GCN模型的预测精度相比时间图卷积网络模型(T-GCN)和门控循环单元网络模型(GRU)更高。此外,扰动分析试验表明,AST-GCN模型具有较好的鲁棒性。To optimize the control of automobile air conditioning system and realize the regional thermal management of vehicle cabin,the prediction of local temperature near the driver and passenger is an effective measure.Aiming at the problem of dynamic prediction of cabin local temperature,the spatial and temporal characteristics of cabin temperature field and the air speed characteristics of air conditioning system were comprehensively considered to establish an attributeaugmented spatio-temporal graph convolution network(AST-GCN).The model training and validation tests on multiple sets of data that collected on real vehicle show that the prediction accuracy of AST-GCN is higher than that of the temporal graph convolutional network(T-GCN)and gated recurrent unit network(GRU)when predicting temperature change over a long time range.In addition,the perturbation analysis shows that the AST-GCN has good robustness.

关 键 词：汽车空调系统 区域化热管理 温度预测 时空图卷积网络 

分 类 号：U469.72[机械工程—车辆工程]