Research On Subway Passenger Flow Forecast And Warning Based On Information Fusion

With the gradual growth of passenger flow scale,efficient organization and safe scheduling of passenger flow become increasingly difficult.Therefore,the research on realtime detection and accurate prediction of subway passenger flow has become more and more urgent.However,the traditional single-passenger flow detection methods have problems such as poor accuracy and uncertainty.At the same time,if only relying on a single source of data,once the device not working,it will cause abnormal passenger flow data records,thereby greatly affect passenger flow analysis,forecasting and early warning.In addition,most of the current researches on short-term passenger flow prediction of rail transit only consider the historical data of stations,and lack of comprehensive consideration of external factors such as climate and holidays,resulting in low accuracy of passenger flow prediction.Based on this,the paper takes the short-term passenger flow prediction of rail transit as the research object,and takes the passenger flow detection method,multi-source detection passenger flow information fusion and short-term passenger flow prediction and early warning as the research content,aiming to provide decision support for subway operators to carry out passenger flow warning and traffic scheduling.This paper is devoted to the research on rail transit passenger flow data processing and short-term passenger flow prediction and early warning.Firstly,collect and analyze the passenger flow data.Consider the research object and data acquisition cost,AFC gate technology and surveillance video technology are selected to detect the inflow of the station,and the error of inflow detection is analyzed.The analysis results show that both AFC gate detector and surveillance video detector have certain errors in the collected inflow due to their own physical characteristics and external factors.Next,the data fusion of subway inflow is carried out.In order to solve the problem of passenger flow detection accuracy existing in a single detection source,a fusion algorithm of multi-source inflow based on federal Kalman filter is designed in this paper to implement the fusion of inflow detected by AFC gate and surveillance video,more accurate estimates of inflow are obtained.By comparing the errors of inflow before and after fusion,it can be concluded that the accuracy of inflow after fusion is significantly higher than that of a single detector before fusion.Then,the transfer flow of the station is analyzed.This paper focuses on the modeling principle and algorithm process of fare clearing model,and calculates the transfer flow of transfer stations based on this model.The passenger volume of the station is calculated by the fused inflow and transfer flow so as to provide relatively exact data support for subsequent passenger flow prediction and early warning.Finally,a short-term passenger flow prediction model RC-Attention GRU is designed based on deep learning technology.In order to improve the accuracy of passenger flow prediction,before designing the passenger flow prediction model,the relevant characteristics of the station passenger flow,such as periodicity,neighbor,as well as the external factors that affect the passenger flow,such as climate,holidays and so on,are analyzed.Then,a passenger flow prediction model RC-Attention GRU is designed based on deep learning technology and analysis results of passenger flow characteristics.The Xipu station of Chengdu metro is taken as an example to verify the model.The experimental results show that the prediction performance of this model is better than other baseline models.Finally,based on the passenger flow prediction results and the set station passenger flow threshold,the passenger flow warning level is designed,and the prediction and warning example is given.