Research On The Delay Recovery Mechanism Under The Wuhan-Guangzhou High-speed Train Operation Adjustment Strategy

High security,high speed,and high density of competitive advantages make high-speed railways more and more popular with passengers,but at the same time,passengers are increasingly concerned about the operational reliability of high-speed railways.Providing high quality and reliable services under the premise of ensuring safety is the primary task and important content of the development of the railway transportation industry.However,the train will encounter a lot of random interference during operation,which will cause the train to be delayed.The delayed recovery ability of the train after the disturbance directly affects the service quality of the high-speed railway.Efficiently recovering high-speed trains from being late is the top priority of dispatchers’ daily work.Selecting and implementing a reasonable dispatching strategy is an important manifestation of dispatching command level.Therefore,based on the actual train operation results,the train operation rules are extracted and the train delayed recovery model is established.The influence characteristics and action mechanism of the delayed recovery of high-speed trains under the operation adjustment strategy are investigated mathematically and physically.The effect of the operation adjustment strategy can be estimated according to the relevant parameters,so as to assist the decision-making of scheduling decisions and make a more effective operation adjustment plan.Based on the occurrence process of train delayed recovery,the environment of high-speed train operation process is treated as a "grey box",the train status is analyzed according to the actual data,and the operation law is explored by mining data.Firstly,the mathematical statistics method is used to analyze the distribution law of delay duration,the distribution law of planned buffertime and the utilization law of buffertime that affect the recovery of train delay.Secondly,the distribution law of delayed recovery under different operation adjustment strategies is explored to explore the relationship between the recovery of delay and the distribution of delay duration and buffertime.Finally,different operation adjustment strategies are established by using machine learning method in order to describe and explain the mechanism of delayed recovery under different adjustment strategies.First,the non-parametric estimation method is used to determine the probability density distribution model of the delay duration.On this basis,the parameter estimation method is used to fit the temporal and spatial distribution law of the delay.It is found that the probability density distribution of the delay under different stations,different sections,different holidays and different periods is subject to the two parameter negative index model.Then,the spatialtemporal distribution of planned buffertime and the utilization of buffertime in stations and sections are statistically analyzed.It is found that no matter the planned value or the utilization value of buffertime,the planned value and the utilization value of stations are larger than those in sections.Moreover,the planned value and the utilization value of buffertime are concentrated in the period of 07:00-21:00 in stations and sections.At the same time,the planned value and the utilization value of buffertime in Changsha south station are larger than other stations,the planned value and utilization value of the buffertime between the west of Zhuzhou and the south of Changsha are larger than that of other sections.Then,the spatial-temporal modeling of the delayed recovery time under different operation adjustment strategies was found.Whether at the station or the section,regardless of holidays or periods,the delayed recovery follows a two-parameter negative exponential distribution model.Exploring the correlation between delayed recovery and delay duration and planned buffertime,K-Means algorithm was used to cluster the delayed recovery of stations and sections under different adjustment strategies,and the delayed recovery between different categories was significantly different after classification.The classification results are good.Finally,based on the gradient boosting regression tree algorithm,a delayed recovery regression prediction model under different scheduling strategies is established.Firstly sort out the six variables related to delayed recovery,then use stepwise regression to select multiple variable features,select the combination of variable features that have an impact on delayed recovery,and use the gradient boosted regression number(GBRT)algorithm to establish the delayed recovery.Regression prediction model,and choose random forest model for comparison during prediction.The results show that the accuracy of GBRT regression prediction model is higher than that of random forest prediction model.