Study On Passenger Flow Forecast And Operation Scheduling Method Of Urban Rail Transit

With its advantages on quantity of transportation, speed, safety, cleanness and highly comprehensive benefits, urban rail transit is gradually becoming the key point of urban public transport facilities construction and management. Along with the rapid progress of the transport network, the ever-growing interaction of passenger flow highlighted its real-time character. Due to these reasons, it is very important to analyze historical passenger flow data and forecast future passenger flow data for operation department to make a better scheduling and improve the quality of service.Based on the data of real passenger flow from the Automatic Fair Collection for urban rail transit, related theories of road and public traffic systems are considered by combining characteristics of rail transit. Therefore, Beijing urban rail transit system is chosen to be the research object and related theoretical problems are researched and applied in practice. Main contents are summarized as follows:(1) The nonequilibrium time and space characteristics of passenger flow are analyzed by combining characteristics of passenger flow. The mathematical formulas of time, section, direction, station nonequilibrium factor are used to describe the nonequilibrium characteristic.(2) The features of several long-term and short-term passenger flow forecast models are analyzed. The advantages of BP neural network to solve nonlinear and uncertain problems are indicated. So it is feasible to use BP neural network to study on the problem of short-term passenger flow forecast of urban rail transit.(3) Based on the theory of BP neural network to forecast, the features of common BP algorithm and faster training algorithms are analyzed. The characteristics of transfer functions and the methods to improving generalization are discussed. The design method of BP neural network model is determined. Cluster analysis is a useful method to ensure the consistency of the training samples.(4) Take the section passenger flow of line 2 and the station passenger flow in and out of XiZhiMeng station of Beijing urban rail transit system as example, BP neural network models of time series forecast and regression analysis forecast are built up. Models can be used to forecast time and date passenger flow. The correlation coefficients of input data and output data are analyzed. Through the test, the architecture of every BP neural network forecast model is determined. Through the example of regression analysis forecast, it validates that this method opens out the relation of the short-term future passenger flow and its influence factors in some error bound.(5) The method which is used to determine the operating strategy and scheduling of the train is discussed. Take the line 1 and line 2 in Beijing as example, this article makes a daily transportation organization scheme, and the dispatching plan which is used to solve the unbalanced problem of passenger flow is also discussed.(6) The Matlab graphical user interface is used to design the system. Using the system, kinds of temporal and spatial distribution curves of passenger flow and scheduling of the train can be displayed, passenger flow can be forecasted and the result of forecast can be analyzed.