Research On Design And Performance Monitoring Of High-speed Train ATO Speed Controller Based On Data Driven

As an important train equipment,the ATO system can replace manual driving,reduce the error rate of manual operation and improve train operation efficiency;At the initial stage of operation of ATO,its speed tracking performance is good.With the increase of train operating mileage,affected by the controller itself and the controlled system,the train speed tracking performance will decrease,which will affect the efficiency of train operation.How to ensure that ATO system has good control performance and can continue to be a research hotspot.This paper conducts research from the aspects of train modeling,control algorithm update and maintainability of ATO.The specific research content is as follows:Firstly,from the perspective of train modeling in the study of traditional ATO,aiming at the problems of dynamic model mismatch and poor online identification real-time performance of traditional predictive controllers when high-speed trains are running in complex environments,a train predictive controller based on the RBF-ARX model is proposed.The controller analyzes a large amount of historical data generated during train operation,inputs the data into the offline identification parameters of the RBF-ARX model and uses the SNPOM algorithm to optimize the parameters to obtain the predictive model of the train,and then construct the predictive controller to perform online train target curve tracking simulation.Secondly,from the perspective of the control algorithm update of the traditional ATO research,when the high-speed train is running in a complex environment,the traditional PID controller is affected by unmodeled dynamics and unknown external interference,which leads to large train speed tracking errors.A high-speed train speed control algorithm based on active disturbance rejection control is proposed.The state space equation of the train is established based on the single-mass model,and the unknown part of the train equation is used as an extended state to design a second-order active disturbance rejection controller,which is compared with the nonlinear PID and PD control in terms of tracking error and parking accuracy.Thirdly,from the perspective of the maintainability of ATO,the performance monitoring of the predictive controller designed in above is firstly studied,and the data-driven method is used to design the control performance benchmark based on cloud similarity from a large amount of input and output data generated at the train operation site.Use support vector machine to classify and simulate four factors that cause controller performance degradation.At the same time,in order to solve the problem of the performance degradation of the subspace prediction controller when the high-speed train is operating in a complex and changeable environment,a train prediction controller performance monitoring algorithm based on the subspace linear quadratic Gaussian(LQG)benchmark is proposed.