Study And Application Of Short-term Traffic Flow Prediction And Regional Traffic Congestion Identification Method

Urban road traffic congestion has become a problem for people to travel daily.In order to effectively solve traffic congestion,facilitate people's travel,and enhance urban happiness,real-time and accurate traffic congestion prediction is particularly important.Short-term traffic flow prediction and urban road network congestion identification are the key technologies of congestion prediction,which directly affect the prediction effect.Therefore,the research on short-term traffic flow forecasting method and road network congestion identification method has certain theoretical and practical value.Due to the instability and complexity of the actual traffic flow changes in the urban road network,how to establish a fast and accurate traffic flow prediction model,how to choose the road network congestion discrimination index and establish a traffic congestion measurement model is an important content of this paper.The main research work of the thesis is as follows:1.Aiming at the nonlinear characteristics of traffic flow data and the noise caused by the interference of the acquisition environment,a wavelet-based denoising method is proposed to smooth the original traffic flow data.The idea of this method is to use the wavelet multi-resolution analysis theory,based on the Mallat algorithm to decompose the original traffic flow data,and then use the wavelet threshold denoising principle to select the threshold of each layer coefficient obtained by the decomposition,and finally use the wavelet coefficient after the noise reduction.The original data is constructed to effectively separate the data from the noise to achieve the purpose of smoothing and denoising the original data.2.The existing short-term traffic flow prediction method is studied and analyzed.Based on this,a short-term traffic flow prediction model based on BP neural network is established.The model principle is a multi-layer network based on the Widrow-Hoff learning rule for the forward transmission of the working signal to reverse the error signal.The original traffic flow data is used to train the network,the neural network algorithm is used to build the mathematical model,and the model is used for prediction.The experimental model is compared with the Kalman filter and GM(1,1)prediction model.The results show that the BP neural network prediction model is better than the Kalman filter and GM(1,1)prediction model.An effective traffic flow prediction model.3.The urban road network traffic congestion measurement index system and traffic congestion identification method are studied.Based on the experimental comparison analysis,a comprehensive measurement of urban road network traffic congestion based on "point","line" and "network" is presented.Model,which measures TCI(Traffic Congestion Index)for a single road,intersection,and area to identify road congestion conditions.In order to achieve the purpose of dynamic identification of congestion in a certain area,the method of distinguishing regional traffic congestion conditions is analyzed and discussed.4.In order to visualize the traffic congestion prediction results,the visualization method based on electronic map was studied,and the existing traffic information monitoring point data was used,and Baidu map API interface technology was used to construct the city of Xi'an Second Ring Road by using JavaScript.The road network topology map realizes the automatic display function of the area.5.Based on the research results of this paper,combined with the application requirements,a traffic congestion prediction system is designed by using MATLAB GUI.The system can predict short-term traffic flow and identify and predict regional traffic congestion.