Research On Parameter Identification And State Estimation Methods Of High-speed Train Based On Iterative Kalman Filtering

The demand for high-speed,high-density,automation,and intelligence of railways is constantly increasing.How to ensure that the trains can maintain the advantages of safety,stability,energy saving and environmental protection in the complex and changing environment has become an urgent problem to be solved in the field of rail transportation.Therefore,this thesis focuses on train data-driven modeling,online real-time identification of unknown time-varying system parameters based on small sample data,and accurate estimation of train speed state.Aiming at the modeling problem of high-speed trains,firstly,the dynamic model of high-speed trains is established,and the time-varying and uncertainties of trains in complex operating environments are analyzed.Secondly,considering the influence of unmodeled dynamics on the train state,based on the historical data of traction/braking force and speed of the train,a data-driven model of the high-speed train is established,and the nonlinear system data with uncertainty is completed through dynamic linearization technology.The transformation of the model to a time-varying autoregressive model with exogenous input and then to a system state space model.Aiming at the identification problem of unknown time-varying parameters in the train system,considering that the high-speed train system has the characteristics of repeated operation in a limited time interval,the idea of iterative learning is introduced,and the adaptive iterative learning is combined with the Kalman filter algorithm.Considering the correlation in the iterative direction and the dynamic characteristics in the time direction at the same time,an identifier suitable for identifying the unknown timevarying parameters of a dynamic system operating in a repetitive environment is designed,and the parameter estimation update rate is derived,which can be performed online and in real time.Optimal identification of fast time-varying parameters in an iterative domain.Then,based on the Lyapunov function,the stability and convergence of the parameter identifier are analyzed.Aiming at the problem of high-precision estimation of train speed state,considering the limited observation of local sensor nodes in the train sensor network and the communication noise between sensor nodes,the network topology information is fully excavated and utilized,and the information weight of sensor nodes is dynamically adjusted.Design distributed direct and indirect measurement models.Then,combined with the Kalman consistency filtering algorithm,a new state estimator is designed,and the update rate of the speed state estimation is derived to realize the optimal consistent convergence estimation of the train speed and filter out the influence of measurement and communication noise.Finally,the stability analysis based on Lyapunov function is given.This thesis studies the data-driven high-speed train nonlinear system parameter identification and state estimation integration technology,avoids the problems of high modeling cost and poor identification accuracy of complex nonlinear systems,and realizes high-speed train system synchronization parameter identification and state estimation.Finally,the proposed algorithm for parameter identification and speed estimation is simulated and verified based on the CRH3 C high-speed train,which proves the superiority and effectiveness.