Short-term Traffic Flow Forecast Of Freeway Combining Toll Station And Detector Data

The mastery of changes in traffic flow is a prerequisite for the effective management of freeways and the timely implementation of control measures.Real-time and accurate prediction of freeway traffic can provide not only information support for the related departments to implement road management,but also provide real-time path guidance for the travelers.Most existing researches are based on single vehicle detector data,and the prediction accuracy is generally insufficient since the models only consider certain period statistic of one constant section.Therefore,how to make reasonable use of various types of existing test data and establish an effective predictive model to improve the quality of flow prediction is of great significance to the improvement of the operating efficiency of freeways.A short-time flow prediction model of freeway section integrated the spatio-temporal correlation analysis and the support vector machine regression(SVR)is developed by analyzing the data of freeway toll station and vehicle inspection device,investigating the characteristics of freeway traffic,and using the ramp-flow estimation algorithm that based on the field collected flow data,which improves the traditional SVR model and enhances the prediction accuracy of highway section flow.The main contents are as follows:(1)Based on the ramp flow estimation of the toll station data,the key problems of the ramp flow estimation of the toll station,and the distribution of the ramp flow in the toll station and the downlink direction were analyzed initially.After that,according to the distribution characteristics of the ramp flow in the upward and downward direction,a ramp flow estimation model based on the support vector machine regression was established.Finally,this model was validated by comparing with the actual traffic data of Chongqing Caojie toll station.(2)The short-term traffic volume prediction of freeway based on spatial-temporal correlation and support vector regression.Firstly,the deficiencies of traditional SVR prediction model were analyzed,and the idea of spatio-temporal correlation analysis was applied.The flow prediction function based on spatio-temporal correlation degree was developed by studying the selection of the adjacent upstream sections,the influence of the ramp flow on the section flow and the measurement of the time and space correlation degree of the adjacent multiple sections.Finally,the short-term traffic prediction model of freeway section was established by combining the prediction results based on the spatio-temporal correlation degree and SVR using the entropy method.(3)Model validation and results analysis.Based on the data of the toll station of Chongqing Yuwu Freeway and the vehicle inspection device,the section of the salt well microwave inspection device was validated with traditional SVR model by prediction the flow of one workday and one holiday.The application results showed that the short-term traffic prediction model of freeway section,which combined the space-time correlation and support vector regression have better prediction ability in both workday and holiday.In addition,the prediction quality of short time traffic flow is improved.