Prediction Of High-speed Rail Braking Process Speed Based On Gaussian Process Regression

High-speed railway is one of the most important infrastructures in China and plays an important role in the national economy,train operation control system is the key to ensure the safety and fast operation of trains,as the safety core of the whole train control system,Automatic Train Protection(ATP)system realizes the interval control and speed control of the train through the overspeed protection curve,so as to ensure the safety of the train operation.During the braking of the train,the existing ATP overspeed protection curve algorithm is to build the model based on the train's own parameters.However,in the actual braking process,due to the influence of objective conditions such as different line conditions,complex weather conditions and the change of its own parameters,the operation control based on the parameter model is often different from the actual situation,affecting the driving safety.Therefore,the non-parametric model can be used to calculate the train braking process speed with higher accuracy,which is of certain reference significance for the research on the algorithm of train overspeed protection curve.In this paper,gaussian process regression is applied to the algorithm of ATP overspeed protection curve,and the calculation and prediction of train braking process speed in the speed monitoring curve are studied.The main work of this thesis is as follows:(1)The control process and speed monitoring curve of train operation are analyzed.According to the dynamic model of train braking,the braking curve of train entering station is simulated by means of iterative forward pushing and iterative backward pushing.(2)The relevant theories of gaussian process regression are derived,and the structure and applicable data types of different single kernel functions are analyzed,the influence of different kernel functions and their hyperparameter on the prediction model is studied,the rules of choosing kernel function in the face of different types of data are summarized.Gaussian process regression model based on kernel function is applied to the prediction of train arrival curve,the results show that the gaussian process regression model with the combined kernel function is more accurate in predicting the train braking curve.By comparing the calculation results of gaussian process regression model and traditional parameter model,it is shown that the gaussian process regression model is more consistent with the actual braking of the train,while the result of the parameter model is a certain limit on the running speed of the train.(3)In view of the continuous generation of speed-distance data during the current braking of the train and the failure of the standard gaussian process regression method to include new samples into the model in time,compared with the standard gaussian process regression method,the iterative gaussian process regression method is used to predict the braking curve of the train entering the station.Compared with the standard gaussian process regression method,this method not only ensures the timeliness of the algorithm,but also improves the progress of the prediction.New samples can be added to the training model in time to improve the reliability of the model.