Study On Short-term Traffic Flow Forecasting Method Of Urban Expressway Based On Spatial-temporal Correlation

With the rapid development of economy and the acceleration of urbanization process,traffic volume increase dramatically,resulting in traffic congestion often occurs in many major cities of the road network,and even the city expressway also occurs crowded phenomenon,seriously affect the function of expressway system.Intelligent transportation system as an effective way to relieve the pressure on urban traffic,its core is the traffic control and traffic guidance system,accurate and real-time traffic flow forecast as the basis of traffic control and traffic guidance system is the key to the implementation of the intelligent transportation system.By summarizing the research status of short-term traffic flow forecasting at home and abroad,it is concluded that the existing forecasting methods mostly take the traffic flow time series on a single section as the research object,difficult to adapt to the time-varying and nonlinear problems of complex traffic flow,prediction accuracy needs to be improved.On this basis,this paper put forward a prediction method considering the spatial and temporal correlation characteristics of traffic flow.Firstly,analyze the characteristics of urban expressway traffic flow,the temporal and spatial correlation characteristics of traffic flow data are analyzed emphatically,which laid a theoretical foundation for short-term traffic flow forecasting model,and a set of preprocessing methods for microwave data is proposed for the original traffic flow data collected by the microwave detector on the expressway.,which solves the difficulty of the subsequent processing due to the lack of data.Secondly,based on the spatial and temporal characteristics of traffic flow of urban expressway,two short-term traffic flow forecasting methods are proposed,which are based on the time and space characteristics of traffic flow.One is to construct four kinds of time-space dimension state vectors considering spatial and temporal factors of traffic flow,establish a traffic flow forecasting model based on GA-Elman neural network;The other is to establish multidimensional time series model based on the traffic data of multiple cross sections with strong spatial correlation and transform it into the state space model,With the unscented kalman filter algorithm to solve the state space model and achieve short term traffic flow forecasting for multiple sections.Finally,combined with the measured data of Beijing second Ring Expressway to verify the two forecasting models established in this paper,simulation results show that the neural network model and the state space model based on spatial and temporal correlation can obtain satisfactory prediction results with accuracy of more than 90%,the prediction result based on spatial and temporal correlation of traffic flow is better than that of the single section based on time series.