Real-time Forecasting Of Channel Section Passenger Flow In Urban Rail Transit Station Based On Spatial-temporal Coupling Characteristics

With the rapid development of urban rail traffic in our country,many big city has reached the level of network operation.With the continuous expansion of passenger flow scale,there is not only a higher demands to the overall network capacity of urban rail traffic,but also a huge challenge to station which undertake the function of passenger distribution.In the station,the channel is closed and narrow,which is the prone area of the congestion in the station.In order to improve the scientific and intelligent level of the station passenger flow management,and provide the important data support for the station operators,it is necessary to predict the short-term passenger flow in real time and accurately.The real time and accuracy of the passenger flow forecast is of great significance for the operators to predict the state of the passenger flow in advance,to take control measures in time and to ensure the safety of the operation.This paper focuses on the research of real time prediction method of city rail transit station passenger,a single section prediction model based on time characteristics and a multi section prediction model based on spatial feature are established.The research contents and main achievements of this paper include the following four parts:(1)This paper analyzes the advantages and disadvantages of the existing passenger flow acquisition technology,select the automatic acquisition technology to collect the required data.By using the technology of data fault identification and repair,the original data is pre processed to get a complete and reliable input data.The spatial and temporal characteristics of the passenger flow data in the station are analyzed,and the characteristics of the fluctuation,periodicity and unbalance of the passenger flow are defined.This paper puts forward the concept of "pool area" and "channelization area",and analyzes the characteristics of passenger flow,solved a major difficulty of the station passenger flow collection.(2)According to the strong fluctuation of passenger flow data in the station,the wavelet packet is used to separate the data trend and the random quantity.Based on the comparison of various prediction methods,SVM and RBF are chosen as the basic prediction model.And then use the improved " variable weight combination method based on fixed length step algorithm and time weighting " to combine two basic model,which is the base to setting up the SCW model.(3)Taking into account the temporal and spatial characteristics of the passenger flow,the spatial factor is introduced into the passenger flow forecasting model.Based on the analysis of the characteristics of the passengers in the station,this paper presents a method for calculating the spatial correlation of passenger flow with fine delay parameters,and gives the quantitative description of the spatial interaction of passenger flow.The influence of the flow line,the degree of congestion and data aggregation on the spatial correlation is studied.The multi section passenger flow data is added into the input space to construct the RTS model.(4)Based on the analysis of passenger flow data in Jianguomen station of Beijing urban railway,SVM,RBF,wavelet+RNF,wavelet+SVM,SCW and RTS model are respectively used to forecast the passenger flow in the station,and the results show that this method can improve the accuracy and stability of passenger flow forecast to a certain extent.Finally,the application of the predicted data is analyzed,and the practical significance of the real time passenger flow forecast is proved.