Theoretical Research For Short-Term Traffic Flow Prediction On Urban Expressway Based On Periodic Component Extraction

The rapid increasing of trip volumes and vehicle ownership brings a series of traffic problems.The urban expressway plays an important role in satisfying long-distance trips of motor vehicle and improving the operational efficiency of the city.Reliability analysis and prediction of short-term traffic flow can induce the travelers to choose the proper route to improve the traffic efficiency and safety and reduce traffic load environment,to maximize the use of traffic resources.The existing research of traffic flow prediction is as a whole,without separating the periodic and nonperiodic parts,which is difficult to meet the forecasting precision requirement well.This paper introduces the temporal characteristics of traffic flow and studies the application of the short-term forecasting models based on periodic components extraction,combining the emerging long-term memory neural network model and the classic models for comparison of urban expressway major lanes and the ramp.The main contents of this thesis are as follows:(1)Combining with the traffic flow theory,three traffic flow parameters were selected to analyze traffic flow characteristics and data preprocessing methods,including the macroscopic traffic flow parameters?the relationship between parameters,the method of identification repairment and stabilization for fault data,which can be used to provide reliable data for short term traffic flow prediction.Then the Pearson coefficient on traffic flow data was calculated for verifying the cyclical feature.(2)Considering the the cyclical feature in the traffic flow,using the Fourier series theory,a short-term traffic flow forecasting model of urban expressway was established based on periodic component extraction.The residuals of the periodic components were separated and analyzed by AutoRegressive Integrated Moving Average,Support Vector Machine and Long Short-Term Memory Neural Network models.Then the residuals output results and the periodic components were combined into the predicted values.The new prediction model may theoretically make the best of the periodic influence on improving prediction accuracy.(3)Three improved models were proposed to identificate the importance of the cyclical feature.Four major lanes and the ramp between You'anmen&Baizhifang bridge were chosen as research objects.Short-term traffic flow prediction was realized by ARIMA,SVM and LSTM models based on periodicity component extraction,including single-step prediction and multi-step prediction.The empirical research findings of this paper were analysis in the end,which can improve and perfect the fundamenta research of traffic flow guidance system.At the same time,the prediction result can provide the theoretical rationale for the establishment of urban expressway entrance control strategy.