Prediction Of Passenger Flow Distribution In Urban Rail Transit Based On Simulation Method

With the continuous expansion of the network scale and the rising passenger volume,the development of responsive operation and organization measures based on the dynamic changes of passenger flow has become the core content of the network operation of urban rail transit system.How to accurately predict the change of passenger flow distribution in the network has become an urgent problem in the study of operation organization measures developing.In this paper,the prediction of passenger flow distribution in the metro network is divided into three core sub problems: short-term od passenger flow prediction,passenger travel path matching and passenger travel train matching.(1)For the problem of short-term od passenger flow prediction,this paper takes Hangzhou Metro system as the research object,by using the aggregate mode and the disaggregate mode,passenger flow in the last hour and the same period passenger flow of last week are determined as the prediction characteristics of the prediction model.The optimal time granularity and spatial granularity selection of short-time od passenger flow prediction are determined in this paper,15 minutes is determined as the best prediction time interval of short-term od passenger flow.A multi input and multi output prediction structure for the whole metro network is designed.Three kinds of passenger flow prediction models,including multiple linear regression model,BP neural network model and Bi-LSTM network model,are used to construct the temporal subnetwork.Forecasting the passenger flow in Hangzhou metro network,the results show that the optimal prediction performance can be obtained by using Bi-LSTM network to design time sub network.(2)For the problem of passenger travel path matching,by analyzing the questionnaire data of passenger subway travel path decision-making behavior,a feasible path search algorithm between od based on electronic map API is proposed.Through the analysis of OD travel records,the normal distribution form of passenger travel time is determined,and the linear regression of travel time expected value fitting model and random forest of travel time standard deviation fitting model are constructed respectively.A passenger travel path matching method based on Bayesian algorithm is proposed.The matching analysis of passenger travel records between Jiangling road and east railway station shows that the matching result of the method is consistent with the actual passenger travel path.(3)For the problem of passenger travel train matching,the passenger flow simulation model of urban rail transit is constructed to realize the passenger train matching.By analyzing the simulation process and logic of passengers and trains,a passenger flow simulation model of urban rail transit is developed based on anylogic discrete event simulation framework to simulate the interaction between passengers and trains.(4)In order to realize the end-to-end prediction of urban rail transit passenger flow distribution status,a distributed urban rail transit passenger flow distribution status prediction system is constructed by integrating the short-term od passenger flow prediction model,passenger travel path matching algorithm and passenger flow simulation model.The system can visually display the dynamic change process of passenger flow distribution status.