Modeling And Simulation Of Dynamic Passenger Flow Distribution In Urban Rail Transit Network Under Real-time AFC Data Access

With the rapid development of urban rail transit and the continuously increasing passenger flow,the operation of urban rail transit is becoming increasingly complex.How to timely and accurately grasp the current system operation state of network to provide a quantitative basis for scientifical and rational network operation decision-making strategy becomes a problem that needs to be solved.This paper focuses on the modeling method of dynamic distribution of passenger flow in urban rail transit network under quasi-real-time data access conditions and constructs a corresponding simulation system.The main research content includes:(1)Based on the analysis of passenger flow characteristics and its influencing factors in urban rail transit network and comparison of the main methods for deducing the dynamic distribution of passenger flow in urban rail transit network,the basic ideas for the dynamic network passenger flow distribution based on online simulation are formed.The two key issues that need to be solved are determined:the dynamic estimation of passenger OD flow and the simulation of dynamic network passenger flow distribution.(2)Aiming at the problem of dynamic passenger OD flow estimation,on the basis of analyzing the connotation of quasi-real-time AFC data,the characteristics of OD dynamic estimation of passenger flow and the applicability of commonly used state estimation methods,a dynamic passenger OD flow estimation method which combined machine learning and recursive Bayesian was proposed.The LSTM-based passenger OD flow prediction model,LSTM model embedded passenger OD flow recursive Bayesian estimation model were constructed.The high-order extension of particle filter algorithm was studied faced with third-Markov process for passenger OD flow state transition process.The implementation algorithm was given and verified by practical case.(3)Aiming at the simulation problem of network dynamic passenger flow distribution and based on the analysis of simulation requirements,system core simulation elements and their dynamic role relationships,a simulation model based on"transport flow" status network,passenger travel process and train operation process was constructed.A system state advancement method was proposed based on state network traversal.In order to give full play to the fast computing capability of multi-core computers,the key technologies of parallel simulation in terms of simulation task decomposition and clock coordination were studied to support rapid simulation of network dynamic passenger flow distribution.(4)A simulation system of dynamic network prassenger flow distribution is designed and developed using the actual Metro as an application case,and the simulation data verification and case analysis are also completed.