Research On Urban Rail Transit Passenger Flow Assignment Model Based On Smart Card Data And Its Application

Urban rail transit is an important way to relieve the pressure on traffic in the process of urbanization.In China,major cities are actively promoting the construction of urban rail transit.With the opening of new lines,many cities have formed the pattern of network operation,in which many problems arise.The distribution of passenger flow is the basis for solving these problems.Traditional passenger flow assignment methods,mainly based on equilibrium theory and Logit model,still have problems in practical applications due to various reasons.With the development of information technology,mass operation data collected and stored in rail transit information system provides a new approach to solve the problem of passenger flow assignment.This paper mines passengers' smart card data recorded in the AFC system of urban rail transit and finds out the pattern of passenger flow to achieve the purpose of passenger flow assignment.The first part of this paper is the literature review on passenger flow assignment,followed by a brief description of the composition of urban rail transit.To store the rail transit network,an adjacency matrix whose weights are from the timetable is used.Next,this paper introduces the frequently-used path search algorithms and hierarchical structure of an AFC system,then calculates passengers' travel time according to smart card data.Based on the assumption that the travel time of a single path obeys the normal distribution,a mixture model consisting of multiple normal distributions is constructed to describe the distribution of travel time on the effective paths between the origin and the destination,namely the OD.According to the characteristic of the model,an iterative non-gradient optimization method—EM algorithm—is used to estimate parameters.This paper also conducts an empirical analysis of the model based on data from Nanjing Metro.The normal distribution test of data in multiple OD pairs and multiple time buckets shows that the hypothesis of the model is valid.Then this paper gives a method to determine the number of normal distribution in mixture model and the solution that EM algorithm can not guarantee the global optimal solution.To test the correctness and effectiveness of the mixture model,data of each OD pair is divided into two parts.The first one is used to estimate parameters in the model and the other is used to test whether it obeys the probability density function obtained from the first part data.Results show that the mixture model in this paper is correct and effective.Finally,two practical cases—calculating the number of passengers for transferring in transfer station and ticket clearing—are used to illustrate the application of the model.