高速列车无线网络自适应滑模容错控制研究  

Research on Adaptive Sliding-Mode Fault-Tolerant Control for Wireless Network in High-Speed Trains

作　　者：刘洋[1] 李帅 李常贤[1] LIU Yang;LI Shuai;LI Changxian(School of Electrical Engineering,Dalian Jiaotong University,Dalian Liaoning 116028,China)

机构地区：[1]大连交通大学电气工程学院,辽宁大连116028

出　　处：《中国铁道科学》2024年第6期202-211,共10页China Railway Science

基　　金：辽宁省教育厅基本科研项目(JYTMS20230038)。

摘　　要：为消除高速列车无线网络控制过程中网络时延和执行器故障对控制性能的叠加影响,对高速列车无线网络进行容错控制研究。首先,搭建列车无线网络控制试验台采集时延数据,采用卷积神经网络(CNN)提取时延数据的空间特征,并利用改进粒子群算法(IPSO)优化门控循环单元(GRU)以提高预测精度;其次,通过反向传播神经网络(BPNN)学习故障状态下的列车参数,对列车牵引/制动执行器进行健康诊断;最后,设计自适应滑模容错控制器对时延和执行器故障进行补偿。结果表明:与PSO-LSTM预测模型相比,IPSOCNN-GRU模型具有更高的预测精度,其最大、最小和平均预测相对误差分别降低94.15%,17.24%和74.39%;在网络时延和执行器故障条件下,所提模型相较于RBF神经网络和反演控制,其速度跟踪平均绝对误差、均方误差和标准差的值均降低近95%。该模型能够精确地预测网络时延,可确保在各种操作条件下列车的平稳运行。To eliminate the superimposed effects of network delays and actuator failures on control performance in the context of high-speed train wireless network control,a fault-tolerant control study is conducted.First,a train wireless network control testbed is built to collect delay data,a Convolutional Neural Network(CNN)is used to extract the spatial features from the delay data,and an Improved Particle Swarm Optimization(IPSO)is employed to optimize the Gated Recurrent Unit(GRU)for enhanced prediction accuracy.Second,the train traction/brake actuator health diagnosis is performed by learning the train parameters under fault conditions through a Back Propagation Neural Network(BPNN).Finally,an adaptive sliding mode fault-tolerant controller is designed to compensate for time delays and actuator faults.The results show that compared to the PSO-LSTM prediction model,the proposed model has higher prediction accuracy,with reductions of 94.15%,17.24%,and 74.39%in maximum,minimum,and average prediction relative errors,respectively.Under conditions of network delay and actuator failure,the proposed model reduces the relative errors of mean absolute error,mean square error,and standard deviation of speed tracking by nearly 95%when comparing with the RBF neural network and inverse control.The model accurately predicts network latency,thereby ensuring smooth train operation under various operating conditions.

关 键 词：列车运行控制 高速列车 自适应滑模容错控制 健康诊断 改进的粒子群算法 

分 类 号：U285.4[交通运输工程—交通信息工程及控制] TP273[交通运输工程—道路与铁道工程]