Incoming Passenger Flow Prediction Of Urban Rail Transit Considering Stable Passenger Flow

Today,the first choice of urban residents for short-distance travel is urban rail transit due to its fast speed,efficient passenger flow,and punctual driving.The passenger flow prediction of urban rail transit has long been a hot topic in urban rail transit industry,which has attracted the attention and research of many scholars at home and abroad.Reliable,accurate and realtime passenger flow prediction is an indispensable part of smart city,which is of great significance to passenger travel planning and vehicle selection.It is also a key factor in the process of subway operation management and scheduling.The current passenger flow prediction methods are mainly based on the massive AFC historical passenger flow data,while the existing passenger flow prediction methods mainly forecast weekday passenger flow and weekend passenger flow separately.The reason is that the morning and evening commuting peak of weekday passenger flow makes the weekday passenger flow and commuter passenger flow have a great difference.However,none of these forecasting methods takes into account the relationship between weekday steady passenger arrivals excluding peak commuter arrivals and weekend passenger arrivals.Therefore,this paper adopts the compressed sensing theory to construct the stable passenger flow on weekdays,and then establishes the prediction model according to the stable passenger flow and the historical passenger flow on weekends.At the same time,considering the causality of the morning and evening commuting peak,the commuting peak prediction model is proposed.The main work done is as follows:Firstly,for the inbound passenger flow of "single peak" and "double peak" stations,the lost passenger flow during commuting peak hours is simulated.Then,according to the characteristics of stable passenger flow throughout the day,the reconstruction algorithm based on compressed sensing theory is adopted to recover the stable passenger flow during commuting peak hours and obtain the stable passenger flow throughout the day.Then we use the clustering method to analyze the stable inbound passenger flow from Monday to Friday and weekend passenger flow to find their similarity and provide input data for the subsequent prediction model.Secondly,the weekday steady passenger flow and weekend passenger flow are taken as the historical data,and the inbound passenger flow sequence is analyzed by using wavelet multi-resolution analysis and single branch reconstruction.Then,appropriate models are used to forecast according to the characteristics of passenger flow components at different scales,including quadratic exponential smoothing method,LSTM model and CNN-LSTM model.Finally,each layer component is fused to get the passenger flow prediction results under the original scale.Through several groups of experiments,the combined model based on wavelet transform has higher prediction accuracy than the single model.At the same time,it is also verified that weekday steady passenger flow and weekend passenger flow are both historical data,which can improve the prediction accuracy of weekday steady passenger flow and weekend passenger flow.Finally,aiming at the absence of wavelet transform prediction model in commuting peak hours,a combined commuting peak prediction model based on ARIMA model and EncoderDecoder model is proposed,and the predicted commuting peak passenger flow is utilized to improve the working day passenger flow.The experimental comparison shows that the prediction method proposed in this paper can improve the prediction accuracy of weekday inbound passenger flow.The accuracy and reliability of the short-time passenger flow prediction model of urban rail transit based on compressed sensing and wavelet analysis proposed in this paper are further verified.