Study On The Evolution Of Passenger Flow Distribution In Urban Rail Transit Network

In recent years,with the expanding of urban rail transit network in China,the passenger flow demands of metro networks also increase dramatically.Many cities in China have begun to build intelligent network command platform for metro management,which is an important component of smart city.However,the accurate and real-time passenger flow distribution state of metro network is not easily available,which plays an important role in monitoring and early warning of network operation status,collaborative operation optimization,joint emergency response and coordination.In view of this,this paper attempts to explore an accurate and realtime estimation method of urban rail transit network passenger flow distribution,by systematically analyzing the temporal and spatial distribution of urban rail transit passenger flow and travel behavior characteristics.This paper focuses on three key scientific issues,including inferring historical passenger travel route,dynamic OD passenger flow estimation and route choice behavior.It is of theoretical and practical significance to improve the intelligent level of urban rail transit operation and management in China.First of all,based on the smart card data of AFC system in working days,the characteristics of urban rail transit passenger flow,namely inbound/outbound volume and OD matrix,are extracted.Then,the distribution characteristics of passenger flow are analyzed in detail from the time and space dimensions.The space–time variation and the characteristics of passenger travel can lay the foundation of data analysis for the subsequent model construction,parameter calibration,verification and application research.Secondly,an accessible route searching algorithm between OD pairs based on limited search depth was proposed.The OD pairs with only single accessible route were taken as the analysis object,and four different distribution functions were used to fit the travel time distribution of single route.The travel time distribution of single route presents the features of right skew and fat tail,and obeys the log-normal distribution.Based on the characteristics of travel time distribution for single route,a parameter estimation method of mixture distribution was constructed for the travel time of multi-route OD pairs.Besides,an adaptive mechanism was introduced to realize the dynamic updating of distributed parameters,and the Bayesian classifier was used to reconstruct the travel route of historical passengers.According to a case analysis of travel time for a single route,it was found that the dynamic updated distribution parameters based on adaptive mechanism can better describe the trend change and uncertainty level of route travel time.The results of consistency test shown that the travel route inferring model of historical passenger proposed in this paper is able to provide highly reliable route estimation.Then,by combing the complex relationship between observable and to be estimated od in large-scale transportation network,the dynamic flow conservation relationship was established in the form of matrix.In this paper,we broke the Gaussian assumption of system noise in traditional dynamic OD estimation models,and constructed a dynamic OD passenger flow estimation model based on particle filter algorithm.Besides,the periodic impact intensity of OD passenger flow is dynamically captured by particle updating and pre-sampling.The results shown that compared with the traditional historical value estimation method and Kalman filter method,the dynamic OD passenger flow estimation method based on particle filter algorithm is more accurate in both the overall index and the time interval index,and has very good operation efficiency.The average estimation time for a single time period is only 1.37 seconds,which fully meets the dynamic requirements of urban rail transit enterprises.Finally,the paper analyzed the characteristics of passenger route choice behavior of urban rail transit and the differences between urban rail transit and road traffic system.12 potential influencing factors were selected from two aspects of inherent route attributes(transfer times,route travel time)and crowd-related attributes,which corresponding quantitative indexes were also established.Based on the smart card data and detailed information obtained from the travel route inferring model,a passenger route choice model based on conditional multiple Logit model was constructed,and the model parameters were calibrated.The results shown that,compared with the route choice model considering only simple and static factors,the real-time passenger route choice model constructed in this paper not only has better fitting ability,but also reveals the decision-making mechanism of passenger route choice behavior more deeply and comprehensively.It can provide fine and valuable reference for the formulation of transportation organization plan.