Selecting Time Granularity Of Short-term Passenger Flow Forecast In Urban Rail Transit

In recent years,urban rail transit has been developing rapidly,and the scale of network operation is becoming larger and larger.The problem of passenger flow management is becoming more and more serious.High precise short-term passenger flow prediction is very important for scientific and rational organization of urban rail transit network operation.According to differences of previous passenger flow information in previous studies,short-term passenger flow for ecasting methods are divided into the Same-Period Method and Ring-Ratio Method,and the predicted objects studied in this paper are determined as passenger flow entering the station and the OD passenger flow.The choice of time granularity is a basic problem for short-term passenger flow forecasting,which directly affects the accuracy of the forecasting results.In response to this basic problem,this paper uses Monday,Wednesday,and Friday as the weekday feature days,and Sunday as the weekend feature day.The following research was conducted using AFC data of five consecutive weeks:Firstly,time series model of passenger flow was constructed.The selected 15 time granularity divides the 1140 min(5:00-24:00)operating time throughout a day into 15 intervals.The passenger flow time series extraction algorithm is used to extract the passenger flow entering the station,the OD passenger flow,and the station's spatial passenger flow at each time granularity,which provide the basic data for similarity measurement and stability test.Secondly,the similarity measurement and stability test method are constructed.The similarity measure of the Same-Period Method is to calculate the Pearson coefficient between the time series of passenger flow in the same period of history,and then obtain the similarity evaluation index R after standardization.The Ring-Ratio Method stability test uses the ADF unit root test to evaluate the volatility of the passenger flow time series,and the corresponding estimated value P of the test result is an index of stability.A priori passenger flow information with higher similarity and less volatility will make the short-term passenger flow prediction results more reliable.Thirdly,a case study was conducted.Taking the 2016 Beijing subway as an example,the similarity measurement and the stability test of the passenger flow entering the station,the station's spatial passenger flow,and the OD flow of 5,000 OD pairs at the 278 stations of the entire network were performed respectively.The results show that the Same-Period Similarity of the passenger flow entering the station and the Ring-Ratio Stability of the weekday increase with the increase of time granularity,and the Ring-Ratio stability on weekends decreases with the increase of time granularity.The Same-Period Similarity of OD passenger flow increases with the increase of time granularity,and the Ring-Ratio Stability decreases with the increase of time granularity.The similarity of the spatial passenger flow at the same time increases with the increase of the time granularity,and the similarity in the peak period is much larger than the off peak period and the entire weekend.Finally,the short-term passenger flow forecasting level is determined.According to the results of the case study,the threshold for the Same-Period Similarity and the Ring-Ratio Stability of the passenger flow entering the station were set to be 0.9 and 0.05 respectively,and for the OD passenger were 0.75 and 0.05 respectively.The threshold for the similarity of station spatial passenger flow at the granularity of 10min,20min,30min and 60min is 0.75.The result of division shows that with a reasonable selection of forecasting methods type,90%of the station's short-term passenger flow entering the station and 70%of the short-term OD passenger flow at a 10-minute granularity on the working day can be accurately predicted.Under the two-day weekend,90%of the station's short-term passenger flow entering the station in 60-minute granularity,and 90%OD passenger flow can be accurately predicted in 10-minute granularity,.By reasonable selection of time granularity,90%of the station spatial passenger flow can be accurately predicted in the peak period of working day,but the spatial passenger flow in the off peak period and the two-day weekend is more difficult to predict.