Research On Urban Rail Passenger Flow Component Division Model And Its Application

With the rapid development of the national economy and the accelerating urbanization process,daily travel activities between cities are becoming more frequent.Residents' demand for timeliness and speed of travel has gradually increased.The urban rail transit system is rapidly developed with its fast,punctual and comfortable features,and plays a significant role in the urban public transport system.With the increase of passenger traffic in the subway network,the composition of passenger,flow has become complex and diverse,and the flow of passengers has changed.The change law of rail transit passenger flow based on the overall passenger flow analysis is not enough to capture the influence mechanism of different events on various passenger groups.Based on the overall passenger flow forecast,the travel rules of different passenger groups are not well captured.Therefore,it is necessary to classify the passenger component within complex urban rail transit,and study the travel characteristics and laws of various passenger components to refine the control of urban rail transit network dynamics.In this paper,the component division principle is proposed to classify the passenger flow with different travel time characteristics,and the structural component division model is used to determine the division method.Based on the theory of component division,the different travel rules of various types of passengers and the correlation degree with the influence factors of the exterior are studied.Finally,a short-term passenger flow forecasting model is constructed based on the component dividing model to achieve accurate prediction of time-sharing passenger traffic in future dates.Firstly,based on the research on the time-varying law of the chronological passenger flow data,this paper determines the different passenger travel time distribution groups,and constructs the passenger component division model based on the distribution characteristics of each group.According to the distribution of travel time,the whole-day passenger flow is divided into five components:morning peak,all-day peak,noon peak,evening peak and night peak.The genetic algorithm is designed to solve the model and determine the parameters of optimal component partition model.The classification result of the model has high precision,and the distribution of the whole-day passenger flow is well fitted,the fitting error is about 11%.Secondly,in order to deeply explore the actual meaning of each component,the time-varying laws of each component and the correlation between the components are analyzed according to the results of component division.Based on the continuous card number identification of AFC data,the relationship between different component traffic and passenger card frequency is studied,and the destination selection preferences of different types of passengers in each week are analyzed according to the site frequent degree.Further,the degree of influence of changes in the external environment on the changes of the passenger flow of each component is analyzed,so as to reveal the passenger flow characteristics of each component in depth.Finally,based on the analysis of the time-varying characteristics of each component,the component partitioning model constructed in this paper is combined with the Radial Basis Function(RBF)Neural Network to construct a short-term passenger flow prediction model.The RBF neural network is constructed for each component to capture the variation law of the passenger flow,and the detailed prediction of the composite passenger flow is realized.The predicted passenger flow of each component is reduced to the predicted value of the time-sharing passenger flow according to the historical passenger flow distribution law,and the predicted values of the time-sharing passenger flow of each component are superimposed and combined to obtain the predicted daily-time passenger flow forecast value.In the case study,the model has a good prediction of the variation law of the whole-day time-sharing passenger flow.The relative error of the 10min whole-day passenger flow forecast is 12.7%,and the absolute error is only 27 people.