Research On Passenger Flow Forecasting And Scheduling In Urban Rail Transit

The rapid development of urban rail transit technology provides a reasonable travel plan for the increasing urban population.However,due to the imbalance between passenger transport demand and transport capacity,it often causes unpredictable traffic congestion on urban rail transit networks.Because of the dynamic evolution of traffic,various urgent and influential passenger safety problems,it has brought great challenges to security operations and great trouble to residents.At present,due to lack of accurate grasp and analysis of passenger flow distribution under congested conditions,lack of accurate knowledge and assessment of passenger flow evolution,and lack of effective grooming and control of passenger traffic congestion,the urban rail transit passenger flow congestion is mainly based on judgement.The existing passenger traffic forecasts are mostly concentrated on road traffic.The researches on urban rail transit passenger flow forecast are still not enough,and the proposed models do not perform well.These issues have been studied in this thesis:1.Aiming at the lack of accurate grasp of the passenger flow distribution under congested conditions,in this thesis,we propose a combination of hierarchical clustering algorithm and LSTM(Long Short-Term Memory)networks for urban rail transit traffic prediction model.We study the characteristics of Shanghai rail transit and passenger flow data,and analyze the time and space characteristics of the passenger flow.Then we use the hierarchical clustering method to cluster the week features,effectively classify the sample types,and further improve the prediction accuracy.In this thesis,we combine the results of hierarchical clustering and the advantages of LSTM networks to predict the passenger flow of urban rail transit.At last,the model proposed is compared to the existing models.2.Aiming at the lack of effective grooming and control of traffic congestion,an improved genetic algorithm is proposed to optimize the scheduling of urban rail transit trains.In this thesis,we consider the travel time of passengers on the subway and the cost of subway operation,and optimize the train schedule.Since travel time and operating costs often conflict with each other,the number of train may decrease as the headway is maximized within a reasonable range.In this case,passenger travel time will increase and operating costs may decrease.On the other hand,when the headway is minimized,the two targets may change in opposite directions.Therefore,it is proposed to maintain the balance between passenger travel time and operating costs in the model.In order to find a near-optimal solution,a genetic algorithm is designed to find a suitable train scheduling plan.