Intelligent rescheduling optimization method of high-speed railway based on deep reinforcement learning DDDQN
In the daily operation of the high-speed railway system,trains are often disturbed by various emergencies leading to delays,which seriously affects passengers'travel experience.In order to work out train rescheduling scheme in a short time and reduce train delay time as much as possible,a train rescheduling optimization method DDDQN combining deep reinforcement learning and an programming model was proposed.First,the track was divided into multiple sections connected.An integer programming model was constructed to describe the train operation process to minimize the total delay time of all trains based on the job-shop scheduling problem.Then,each train was regarded as an agent,and the state,action and reward functions of multiple agents were defined according to the actual operation requirements.Two deep neural networks were constructed to approximate the functions.Finally,combined with the above integer programming model,the training method of DDDQN was designed.In this model,the feasible solution to the problem was explored by the agent in the simulation environment,and the parameters of the neural network were updated by the"mutual feed"mechanism between the two neural networks.On this basis,the optimal solution to the problem can be obtained in a short time by solving the integer programming model.The actual track data and operation data of the Beijing-Zhangjiakou high-speed railway were used for simulation experiments,and the total train delay time and solution time obtained by three different solution methods under 10 different emergency scenarios were compared,which verified that the proposed DDDQN model could obtain the optimal solution of the problem in a short time.DDDQN can reduce train delay time by up to 30.43%and solution time by up to 68.33%.DDDQN provides an intelligent method and reference for improving the emergency handling ability and transportation organization efficiency of high-speed railway systems under emergencies.
国家科技期刊平台
NSTL
万方数据
