Research On The Detection Method Of Abnormal Congestion Based On Traffic Flow Forecasting

With the rapid development of urban transport and road network is becoming more complex,traffic congestion is also an increasing number, so that the entire road network to run slowly, and reduced operating efficiency, which caused many problems such as socio-economic losses,casualties and environmental pollution have become increasingly serious.The harm to the road traffic which caused by abnormal congestion in traffic congestion is extremely huge because of the their own uncertain and unpredictable characteristics. In order to minimize the loss of abnormal congestion, the traffic control department and traffic researchers are looking for accurate and efficient methods to detect abnormal congestion,which can timely detect abnormal congestion events in the road network,and adopt traffic management measures.By studying the existing detection algorithm, this paper proposed a abnormal congestion detection method based on traffic flow forecasting. First of all,select the ARIMA model and BP neural network model which is optimize by genetic algorithm as a single prediction model,Use the measured data collected by a traffic detector of a section in Nanjing to model the single models,then verify the validity of the model by the measured data.Then test the forecasting effect of combination forecasting model combined by the two single models,and compare with the two single models.Next,the prediction error method is introduced, a Integrated detection model of abnormal congestion based on travel flow integrated forecasting. Finally,simulate traffic conditions with abnormal congestion events by the VISSIM traffic simulation software,and varify the effectiveness of the integrated detection model by the data obtained in the simulation. The results show that the detection model can accurately detect abnormal congestion event in simulation. Through the analysis of the simulation, Improve the overall performance of the detection algorithm by making a corresponding improvement on the lack of detection.