High-Speed Train Delay Prediction Method And System Based On Extreme Learning Machine

China has gradually formed a high-speed railway network with "eight vertical and eight horizontal" as the short-term planning.At the same time,high-speed railway is favored by people on account of its fast operation speed,high punctuality rate and strong transportation capacity.However,due to various unavoidable factors,train operation often deviates from its planning timetable.If the train operation plan is not adjusted timely and effectively,it may endanger the safety of train operation.In order to provide more accurate train delay information for dispatchers and effectively alleviate the problems of railway transportation efficiency and safety reduction caused by disturbances and disruptions,this paper realizes the accurate prediction of train arrival and departure delay.The specific research contents are as follows:(1)The basic theories of train delay were analyzed and summarized.Starting from the causes and types of train delay,this paper then introduced the mechanism of propagation and absorption of train delay,which provided theoretical support for the research of this paper.(2)Statistical analysis of train arrival and departure delays.Based on the actual operation data of trains on Wuhan Guangzhou railway(W-G HSR)line,the statistics of the deviation of train arrival and departure time in each station were completed from the perspective of station,and the order of train delay rate at station and section were completed from the perspective of station and interval,so as to study the spatial characteristics and laws of train arrival and departure delay.(3)Train delay prediction method based on particle swarm optimization(PSO)algorithm and extreme learning machine(ELM)algorithm(PSO-ELM)were proposed.Firstly,after defining the prediction scene of train delay in this paper,the spatial and temporal characteristics and infrastructure characteristics which may affect train delay time were extracted,including the station delay rate obtained in Chapter 3,train departure and arrival delays,and so on.Then,the importance of these features was sorted by decision tree(DT)algorithm,and the feature data sets which affect train arrival and departure delay were determined.Finally,the case study was carried out by using the actual train operation data on the Wuhan Guangzhou high-speed railway(W-G HSR)line,which proved that the prediction model proposed in this paper was better than the prediction model based on k-nearest neighbors(KNN)algorithm and the gradient boosting decision tree(GBDT)algorithm.The efficiency and accuracy of particle swarm optimization(PSO),bayesian optimization(BO)and genetic algorithm(GA)were compared.The results showed that PSO possessed the highest efficiency and accuracy.(4)On the basis of the research in Chapter 3 and Chapter 4,combined with the train operation data on Wuhan-Guangzhou high-speed railway(W-G HSR)line and the user needs of high-speed railway field investigation,the high-speed railway trains delay analysis and prediction system was developed,which realized the visualization of the accurate prediction and statistical analysis of train arrival and departure delay,supported the high-speed railway managers to grasp the delay situation of trains,and assist to make relevant rescheduling plan.There are 39 figures,26 tables,and 54 references in this paper.