Research And Application On Key Technologies Of Metro Passenger Flow Monitoring System

With the rapid development of urban rail transit,the passenger flow forecasting of metro stations is beneficial to the operating department to observe the real-time trend of passenger flow,adjust the train scheduling strategy,and improve the service level.In view of the current passenger flow monitoring system,there is a problem that the short-term inbound and outbound passenger flow forecasting and the passenger flow forecasting of the hub station are not accurate enough,the universality is poor,and the stability is low.Based on the basic structure of the original passenger flow monitoring system,a passenger flow monitoring system based on AFC data is designed to improve the predictive performance of the system.The passenger flow monitoring system studied in this thesis is mainly composed of two models: a short-term inbound and outbound passenger flow prediction model based on SVM and a passenger flow distribution prediction model based on a hybrid approach at the hub station.The main research contents are as follows:(1)The spatio-temporal characteristics of short-term passenger flow and the research status of passenger flow forecasting are analyzed.Combined with the characteristics of passenger flow and the demand of passenger flow monitoring system,the overall framework and software part design of the system are given.The ideas and workflows of the key technologies of the system are introduced.Finally,passenger flow monitoring system is implemented.(2)The short-term inbound and outbound passenger flow prediction technology is studied.According to different passenger flow waveform characteristics of different stations,the aggregated aggregation algorithm is used to analyze the weekly passenger flow of the station,and the weekly passenger flow is classified.The correlation time series of each type of passenger flow is analyzed,and the sequence with strong correlation is selected to perform SVM regression prediction.The dual-population adaptive chaotic firefly algorithm is proposed to optimize the SVM model parameters.The algorithm uses the dual population mechanism to improve the population diversity and random distribution ability.The chaotic attraction degree is introduced to improve the global search ability of the algorithm,and bad initial value is avoided.Besides,adaptive search step size of the algorithm is good to the convergence speed and accuracy of the algorithm.The experimental results show that the short-term passenger flow prediction model can effectively predict the passenger flow at different sites,and the performance of the improved algorithm is obviously improved,which satisfies thehigh-precision and universal performance requirements of the passenger flow monitoring system for passenger flow prediction.(3)The technique of predicting the distribution of passenger flow at the hub station by the maximum entropy model is studied.Aiming at the unpredictable problem in the traditional passenger flow distribution prediction model when there is no complete OD matrix,a hybrid prediction method based on improved maximum entropy model and gravity model is designed.The improved maximum entropy model introduces the transfer distance and transfer disgust duration to solve the problem,and improves the prediction ability when there is no complete OD matrix.The entropy method is used to calibrate the line impedance and improves the prediction stability of the gravity model.Finally,combining the advantages of the two models,a hybrid forecasting method is formed to effectively predict the passenger flow distribution at the hub station.The experimental results show that the hybrid passenger flow forecasting model of the hub station can predict transfer passenger flow with or without complete OD matrix,which satisfies the high-precision and stable performance requirements of the passenger flow monitoring system for passenger flow prediction.