Research On Short-term Passenger Flow OD Estimation Of Urban Rail Transit Based On IC Data

The degree of networking of urban rail transit system is getting higher,which has dynamic and information requirements for the operation management of the rail transit system.Because the current information collection system is difficult to meet the dynamic management requirements,the estimation of the distribution status of passenger flow has become a realistic problem.In this paper,the short-term passenger OD flow matrix estimation is studied which is the basic input data of the dynamic management system.Firstly,the particularity of passenger OD flow estimation are analyzed based on the characteristics of urban rail transit.Then,this article sorts out and summarizes the existing research ideas and methods.Based on this,this article has formed a research route that making full use of historical data and fusing real-time data to develop estimation model.Secondly,this paper analyzes the variation and influencing factors of passenger OD flow.The main contents include:(1)Using clustering technology and data visualization to analyze the distribution of passenger.(2)A method for measuring the similarity,stationarity and complexity of short-term passenger OD flow is proposed.The predictability of OD is evaluated by similarity and complexity.It provides theoretical guidance for the choice of estimation time granularity.Thirdly,the OD estimation model of urban rail transit short-term passenger flow based on K-nearest neighbor algorithm is developed.And based on the least squares model,an estimated result correction model combining historical data and real-time data is developed.The actual case is used to introduce the parameter setting and implementation process of the prediction model in detail.Finally,the final experimental results show that the average error of all prediction sites on working days is 2.16 person-times/30 min,and non-working day is 2.02 persontimes/50 min.It proves that the prediction model proposed has good adaptability to various types of station.The prediction method is effective and the prediction results have high reliability.