Prediction Of Spatiotemporal Distribution And Guidance Of Passenger Flow In Rail Transit Networks

Rail transit has played a key role in the comprehensive transportation system,due to its advantages in safety,speed,punctuality,large volume,and low energy consumption.At present,as an most popular travel mode for urban residents,rail transit plays an important role in meeting residents’ travel needs and alleviating traffic congestion,which can also be regarded as an effective way to build green city and smart city.With the continuous improvement of the rail transit network,the operating environment is becoming increasingly complex,and the super-large passenger flow is becoming more and more normal.Accurately obtaining the spatiotemporal distribution of rail transit passenger flow and the evolution of passenger flow in real time,guiding passengers to choose appropriate travel routes in emergencies and reducing the operating pressure of the passenger flow of the road network are effective measures to ensure the safe and efficient operation of rail transit.This paper focuses on the spatiotemporal crowd flow prediction as well as the guidance and control of passenger flow in urban rail transit network,carrying out researches on key issues such as station type identification and passenger flow evolution mechanism,prediction of the spatiotemporal distribution of passenger flow under normal operation,and evacuation guidance control of passenger flow under emergencies.The main research content of this article can be summarized into the following four aspects:1.Considering the evolution characteristics of passenger flow at stations,a method for identifying station types and passenger flow evolution mechanism based on second-order clustering is proposed.With extracting various feature values such as network structure,passenger flow time series morphological features and structural features,and passenger flow statistical features,this paper applies the principal component analysis method to reduce the dimensionality of the feature values,and proposes a station type recognition method based on two-step clustering.Besides,from the perspective of different types of stations and different rainfall intensity levels,a passenger flow impact analysis method based on oscillation coefficient is proposed.Taking Beijing rail transit as an example,the stations are divided into 6 categories,and the empirical analysis of the evolution mechanism of inbound passenger flow under typical scenarios shows that rainfall has a significant impact on rail transit travel and different types of stations are affected by rainfall as well as present different passenger flow evolution mechanisms.2.Considering the characteristics of non-linearity,non-stationarity,periodicity and randomness of passenger flow,this paper proposes a passenger flow prediction method based on Decomposition Neural Network Composition.Firstly,use the time series decomposition method to transform the inbound passenger flow into multiple sub-time series with weaker volatility and strong regularity.Secondly,The normalization method is used to preprocess the time series,which can solve the problem of low prediction accuracy of neural network due to the large gap in the mean value of the time series data.Then based on the long and short-term memory neural network model prediction and reconstruction,the final prediction result is obtained.Finally,simulations of passenger flow forecasting examples are carried out for different types of stations.The results show that the performance of the passenger flow forecasting model proposed in this paper is better than the traditional passenger flow forecasting model,and the accuracy of passenger flow forecasting is improved.3.Considering the problem of passenger transfer path identification,a method for predicting the spatiotemporal distribution of passenger flow based on mobile positioning trajectory reconstruction under normal operation is proposed.Aiming at the shortcoming that automated fare collection(AFC)data does not contain passenger transfer information,a passenger travel path identification model based on reconstruction of mobile positioning trajectory reconstruction is proposed.Besides,based on the principle of stochastic user equilibrium,considering multiple factors such as passenger retention and crowded carriages,a prediction model for the spatiotemporal distribution of passenger flow in the rail transit network is constructed.Finally,a simulation analysis of a calculation example is carried out for the rail transit network,and the results show that the indicators such as the passenger flow rate output by the model are consistent with the actual results,which verifies the validity and accuracy of the model.4.Considering the impact of passenger travel delays and transit network congestion caused by emergencies,a passenger flow evacuation guidance control method based on a two-level programming model is proposed.Considering the delay and psychological effects of emergencies on passenger travel route selection,this paper establishes a passenger travel route selection model based on cumulative prospect theory.Besides,considering the total travel impedance of passengers and network congestion,this paper designs a discrete particle swarm algorithm to optimize the guidance and control strategy of road network passenger flow evacuation.Finally,a simulation analysis of a case study is carried out for the urban rail transit network.The results show the passenger travel route model based on the cumulative prospect theory are consistent with the actual survey results,and can accurately describe the passenger’s decision-making behavior during route selection.The proposed passenger flow evacuation guidance control strategy effectively reduces the overall travel time of passengers and alleviates the congestion of the urban rail transit network.