Research On Real-time Prediction Method Of Urban Rail Transit Network Passenger OD

Under the condition of urban rail transit network operation,the operator needs to timely and accurately predict the passenger travel demand and change when making dynamic transportation organization decisions.This paper aims at the prediction demand of real-time passenger flow OD under the normal operation scenario,and analyzes the passenger flow change rule by using the automatic fare collection(AFC)data.Combined with the historical passenger flow rule and the passenger flow situation in real-time operation,the passenger flow OD is predicted in real time.The main contents of this paper are as follows:(1)Analysis of passenger flow characteristics and proposal of research scheme.Based on AFC data,this paper analyzes the passenger flow change rule from the two dimensions of station inbound passenger flow and passenger flow destination structure,summarizes the characteristics of common passenger flow OD prediction methods,and forms the basic idea and research framework of real-time passenger flow OD prediction of road network,including the analysis and recognition of passenger flow change patterns,real-time passenger flow OD prediction and adaptive passenger flow OD correction.(2)Analysis and recognition of passenger flow change patterns.An abnormal passenger flow detection algorithm based on local outlier factor(LOF)algorithm is proposed to detect abnormal passenger flow under abnormal operation conditions.This paper proposes to use the complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN)algorithm to analyze and recognize the change patterns of the time series components of the inbound passenger flow at different time scales,and use the K-means++ algorithm to analyze and recognize the change patterns of the category at different time dimensions of passenger flow destination structure.The change patterns of passenger flow are analyzed and recognized from the perspective of inbound passenger flow and passenger flow destination structure.(3)Real-time passenger flow OD prediction.On the basis of passenger flow change patterns,based on the long short-term memory neural network(LSTM),the prediction models of each time series component are constructed respectively to realize the real-time prediction of station inbound passenger flow.Based on the K-nearest neighbor(KNN)algorithm,the date category prediction model of passenger flow destination structure is constructed,and the prediction model of passenger flow destination structure under date category is proposed to realize the real-time prediction of passenger flow destination structure.Combining the prediction results of inbound passenger flow and passenger flow destination structure,the real-time prediction of passenger flow OD is realized.(4)Adaptive passenger flow OD correction.Combined with the AFC card swiping information that can be collected in quasi real time in actual operation,an adaptive correction model of passenger flow destination structure prediction result based on state space model is constructed,and the particle filter algorithm is used to solve the model,which realizes the adaptive correction of passenger flow structure prediction result and improves the accuracy of passenger flow OD prediction.(5)Taking Beijing urban rail transit system as an example,the prediction method proposed in this paper is verified,and the results show that the prediction method proposed in this paper has good prediction performance.There are 45 figures,16 tables and 60 references.