The Method Of Subspace Model Identification And Predictive Control For High-speed Train

With the increment of train speed, the dynamic interaction between the train and otherobjects such as pantograph-catenary system, wheel-rail, air, becomes more and more stronger.The qualitative changes in the dynamic environment of the train, i.e., mechanical and/orelectrical dominated actions are turned to aerodynamics domination. And higher requirementsfor effective modeling and control method of high speed train, which includes time-varying,nonlinear and uncertainty model of the process for high speed train. Deu to the problems ofeffective modeling and control method of high speed trains, the main content of the paper isas follows:⑴A predictive controller, based on multivariate state space model with date drivenmethod, is designed for high-speed train. The train multivariate dynamic system of high-speedtrain is established in view of the analysis of the dynamics system of the high-speed train. Theprediction model of high-speed, based on subspace model identification method, and isobtained directly from input/output data. Moreover, the design of the predictive controller isgiven in detail.⑵With the characteristic of the uncertainly description on the model for high-speedtrain operation process, an adaptive subspace predictive control method based on modelswitching strategy is designed for high-speed train. The subspace prediction initial model ofhigh-speed train is obtained from observation data, and the adaptive prediction model ofhigh-speed train is presented by sliding window formulations of the fast LQ decompositionfor online update, which improve the identification speed of on-line subspace method. Amodel switching strategy is presented by taking consideration of predictive error ofhigh-speed train, and the design method of the adaptive subspace predictive controller ofhigh-speed train is given in detail.⑶Due to the uncertainly characteristics of observational data sets online for high-speedtrain, the incremental subspace prediction model of high speed train is established on the statespace model, and an adaptive subspace predictive controller based on time-varying forgettingfactor is designed for high-speed train. In order to avoid the unsafe controlling factors such asno control output within missing of observations data online for high-speed train operationprocess, a incremental model based on the subspace prediction initial model of high-speedtrain. Using the time-varying forgetting factor method to reduce the old data and increase theamount of information of the new data in the train modeling, and an adaptive subspaceprediction model is proposed. Then the design approach of adaptive subspace predictive controller for high speed train is given in detail, at the same time, the control algorithm is alsopresented.⑷With the penturbation model which is constuced based on the noise characterstics ofhigh-speed train dynamic process, an adapive subspace prediction controller is designed forthe high-speed system within persistent noise penturbation. Basd on the subspace predictioninitial model, a persistent noise set is estimated in the different conditions process ofhigh-speed train. With the method of the first-order approximation, the relation between noisesequence and the perturbation of subspace prediction model is revealed by special matrices,and the perturbation model of high-speed train is obtained. Moreover, the design of thesubspace predictive controller of high-speed train is given in detail and the control algorithmis also presented.⑸The simulation experiments of state space model and predictive control method aredesigned for high-speed train control system. In order to illustrate the effectiveness of theproposed method, model validation and control simulation study of high-speed train based ona high-speed train similar to CHR3is implemented, and the results show the enhancedperformance of the proposed method.