Study On Mechanism And Prediction Of Site Passenger Flow Uncertainty For Urban Rail Transit

Accurate and reliable short-term urban rail transit (URT) passenger flow forecasting is an important indicator to evaluate URT service level and system’s operating status. Howerer, the majority of current researches mainly focus on forecasting the mean of passenger flow, as existing researches on the reliability of passenger flow forecasting methods are quite few, a forecasting method of site passenger flow uncertainty is mainly studied, based on the research of passenger flow uncertainty mechanism. In addition, the forecasting performance of this method is evaluated.First, the uncertainty mechanism of passenger flow is analyzed, taking operation period, site location, type of passenger flow, ticket price, weather, and other related factors into consideration. And the fluction of site passenger flow which is in different characteristics of day and in different station location is qualitatively studied. From the view of the composition of passenger flow, Weekend and Holiday have a greater influence on the passenger flow uncertainty than Workingday, the station which is next to the train station or commercial centre has a greater influence on the passenger flow uncertainty than the station which is next to working area or residential district. Then, on the basis of the background knowledge of the GARCH model and the SV model, the parameters of the ARIMA-GARCH model and the ARIMA-SV model are calibrated based on data stationary test, identification of ARMA model and heteroscedasticity test using six different types of passenger flow collected by Automatic fare collection (AFC) system, then the ARIMA-GARCH model and the ARIMA-SV model are constructed. Finally, the confidence interval (CI), the kickoff numbers (KN) and the kickoff percentage (KP) are used to evaluate and compare the uncertainty of the ARIMA model, the ARIMA-GARCH model and the ARIMA-SV model.According to the evaluation results of urban rail transit passenger flow uncertainty forecasting based on actual data, for the passenger flow that has different types of fluctuation, the uncertainty forecasting performances of both the ARIMA-GARCH model and the ARIMA-SV model are significantly better than the ARIMA model, and the ARIMA-SV model is better than the ARIMA-GARCH model. Compared to the ARIMA model, both the GARCH model and the SV model proposed in this thesis can fit the fluctuation of passenger flow better, and provide a more reliable theoretical basis for the operation and management of URT.