Research Of Short-term Traffic Flow Prediction Based On Spatio-Temporal Characteristics

With the acceleration of urbanization,the problem of urban road traffic congestion has become increasingly serious.It is suggested that accurate and real-time short-term traffic flow information are basic support for excellent performance of transportation system.Therefore,it is of great significance to establish an efficient short-term traffic flow prediction model.However,subsequent researches are mostly illustrated on single-section historical data,ignoring the impact of upstream and downstream traffic flow.Based on the analysis of temporal and spatial characteristics of traffic flow,this paper discretizes the spatial-temporal equations of traffic flow,and establishes a short-term traffic flow prediction model.The main research content is as follows:First,the background and significance of short-term traffic flow forecasting are elaborated,At the same time,the existing forecasting methods are classified in detail.Besides,a literature review is made on the current research status of each type of method.The technical route,chapter arrangement and research content of this paper are also introduced.Secondly,the basic parameters definition,the relationship between time occupancy rate and density and the influencing factors of urban traffic flow are elaborad.in this article.Based on the basic map and three-phase traffic flow theory research,the traffic flow is divided into three states: free flow,congested flow,and blocked flow.Traffic flow data processing technology uses threshold method to identify abnormal data,and uses adjacent data average method to repair abnormal data.Following that,the temporal characteristics of the traffic flow are analyzed in the free-flow state and the C-C algorithm is used to solve the phase space reconstruction parameters for unsteady time series and the phase reconstruction.Moreover,the spatial characteristics are analyzed by calculating the Pearson correlation coefficient.The results show that the section traffic is predicted.The flow is affected by the upstream traffic flow,and a segmented road resistance function is established based on the classical traffic flow parameter model.Combined with the traffic wave theory,the spatial-temporal discretization of conservation equations of traffic flow was learned by using the method of solving partial differential equations,and a short-term traffic flow prediction model was established.A short-term traffic flow prediction state space model is established by using the traffic-occupancy function relationship and Kalman filtering is used to solve it.Using the traffic flow data of the expressway in Beijing to conduct empirical research,traffic flow predictions were conducted for data intervals of 2 minutes,4 minutes,and 6 minutes,respectively.Finally,2 minutes is determined as the best prediction time interval,which was proved that the model after phase space reconstruction had more Good predictive performance.Finally,in the blocked flow state,the time series of the traffic flow is analyzed as a stationary sequence;the Pearson correlation coefficient calculation results show that the predicted cross-section traffic flow is affected by both the upstream and downstream trafficflow.Based on the distribution of spatio-temporal nodes in traffic flow prediction for analog congestion state,the space-time equations of traffic flow conservation are discretized according to the second-order precision difference scheme of partial differential equations,As a result,a traffic flow prediction model is established.Combined with the flow-occupancy function relationship,a state-space model is established and Kalman filtering is used to solve it.An empirical study was conducted using Beijing Expressway traffic flow data.By comparing the results of multiple time interval data predictions,it was determined that 2minutes was the best prediction time interval.Meantime,it was proved that the model was of excellent effect in traffic flow prediction.