Automatic Rescheduling Of High-speed Train Operation With Temporary Speed Restriction Under Strong Winds

With the advantages of safety,convenience,comfort and punctuality,the high-speed railway has become a crucial part of promoting the development of national economy and strong transportation network.The increasing complexity of high-speed railway network and the continuous growth of passenger volume have put forward higher requirements on operation service quality and traffic command capabilities.During the daily operation of high-speed railway,trains run safely and punctually according to a stage plan.When natural disasters,equipment failures and other emergencies occur,if the train stage plan deviates,it requires to be adjusted through manual experience.As a consequence,dispatchers will be in larger work intensity,and emergency response efficiency also needs to be improved.Nowadays,with the trend of intelligent rail transit around the world,automatic rescheduling of high-speed train operation is also developing towards automation and autonomy.It is vital of significance in theory and practice to study the automatic rescheduling of high-speed train operation with temporary speed restriction under strong winds,which can reduce the work intensity of dispatchers and improve the efficiency of railway operation.The automatic rescheduling of high-speed train operation based on reinforcement learning will be studied in this paper for the following typical scenarios including train arrival and departure delay,temporary speed restriction for a single section and multiple sections under the influence of strong winds.The main contents are as follows:Firstly,a train departure sequence rescheduling approach addressed by Monte Carlo tree search is proposed in train arrival and departure delay scenario under strong winds.Considering train operation constraints in stations and block sections,an automatic rescheduling model of train operation is utilized to mine Spatio-temporal constraint characteristics in the train diagram so as to establish a reinforcement learning environment.In this way,the designed Markov decision process involves state space,action set,state transition probability and reward function.Moreover,it is the proposed approach named Monte Carlo tree search that presents an optimal train departure sequence,which can effectively reduce delays in train arrival and departure delay scenario.Secondly,an automatic rescheduling approach of adjusting train departure sequences and operation time is presented with temporary speed restriction in a single section under strong winds.Based on redundant time characteristics of the train diagram,the automatic rescheduling model of train operation is improved to optimize buffer time in a station and recovery time in a block section.Aiming at improving the train actual arrival and departure time in the reinforcement learning environment,the calculated method of train operation time is constructed in stations and block sections.Then the method of detecting and resolving conflicts is proposed in the high-speed train diagram.Given by Monte Carlo tree search,the automatic rescheduling strategy adjusts train departure sequences and operation time in order to reduce train delay and suppress delay propagation effectively.Finally,an automatic train rescheduling strategy based on delay prediction is further investigated with temporary speed restriction in multiple sections under strong winds.According to maximum traction and maximum braking characteristics,train running situations including speed and time is predicted in different temporary speed restriction sections.What's more,inner relations between recovery time in a block section,train delays and running conditions are revealed.Considering the train delay prediction method based on maximum traction,maximum braking and delay propagation rule under the running situations,a train operation rescheduling strategy is automatically selected between Monte Carlo tree search and first-in-first-out for different speed levels and sections lengths of temporary speed restrictions.The paper has 38 figures,31 tables and 110 references.