A Study On Traffic Congestion Detection And Forecasting For Dangerous Goods Transportation

Road transportation is the main means of dangerous goods transportation. The dangerous goods with great variety and large number have brought tremendous hidden danger to the transportation. Therefore, how to grasp the status of traffic timely and accurately is very important to the route selection of dangerous goods transportation, the reduction of the occurrence of traffic accident and the decision making of emergency rescue after traffic accident. So the traffic congestion detecting and forecasting is an important part of ensuring the safety of dangerous goods transportation. In order to improve the accuracy of the detecting and forecasting, the thesis proposed a set of models of traffic congestion detection and forecasting based on multi-source data and verified the scientificity, rationality and effectiveness of the models by a series of case studies. Finally, the thesis put forward some methods and suggestions on the practical application of the results on the dangerous goods transportation.At first, the thesis analyzed the characteristics of the transportation of dangerous goods, the dangerous goods transportation accidents occurred in recent years and the reasons for the accidents. Besides, the thesis emphasized the importance and requirements of the traffic congestion detecting and forecasting for the route selection of dangerous goods transportation and the decision making of emergency rescue after traffic accident.Then, the thesis analyzed the classification and characteristics of traffic data. Combined with the reliability of data source, the thesis chose induction coil and video detector which could provide site traffic data as the research object. And then, the thesis identified and eliminated the abnormal data with threshold method and the traffic flow mechanism method, which are the two commonly used traffic data processing methods. After that, the thesis fixed the missing data with adjacent data average method.Furthermore, the thesis designed the traffic congestion index for every single detector based on the characteristics of traffic flow, vehicle speed and time occupancy ratio. The traffic congestion detecting model based on the above two kinds of detectors established according to the reliability and the availability of traffic data of every single detector. Then the thesis verified the validity of the model by an instance. Finally, the thesis established the traffic forecasting model based on multi-source data fusion. The model combined the BP neural network algorithm with Kalman filtering algorithm and grey forecasting method. The model used the data preprocessed by Kalman filtering algorithm and grey forecasting method as the original data of BP neural network forecasting model. Also, the thesis verified the validity and reliability of the two kinds of combined forecasting model by an instance. In the end, the thesis substituted the combined forecasting results into the traffic congestion detecting model based on multi-data fusion, used for forecasting the status of the traffic congestion in the next time period, and then put forward some methods and suggestions on the practical application of the results on the dangerous goods transportation.