Research On Urban Traffic Pollution Intelligent Prediction Algorithm Based On Deep Residual Networks

Urban traffic pollution is an important source of pollution affecting urban air quality,directly threatening human health and the global climate.With the rapid increase in the number of motor vehicles in the city,traffic pollution has gradually become the main influencing factor of urban air pollution.However,due to the complexity and variability of urban traffic,urban traffic emissions are difficult to effectively control and predict,thus affecting the overall city air quality.At present,the research on traffic emission forecasting mainly focuses on the prediction of small-area areas or the forecast by the annual unit,and can not predict the spatial distribution of traffic emissions in the whole city.This thesis investigates and analyzes the research results and existing problems of predecessors in urban traffic emissions forecasting.Aiming at the problem of the impact of traffic pollution on urban air pollution and the problem of urban traffic emission prediction,this thesis carried out data analysis and model construction.And this thesis has carried out experimental verification,which laid the foundation for the air pollution prevention and control system of smart cities.Aiming at the problem of the impact of traffic pollution on urban air pollution,this thesis collected data on air pollution source emissions and air pollutant concentration in Beijing.Correlation analysis was carried out on the degree of influence by using the grey correlation analysis method,so that the correlation between the emission of each pollution source and the concentration of each pollutant was obtained.And through the ranking of relevance,it is determined that traffic pollution has become the most important factor affecting air pollution in Beijing.Aiming at the problem of urban traffic emission prediction,this thesis constructs a traffic emission data conversion model based on the mechanism analysis and formula derivation of traffic emissions.The model realizes the conversion of vehicle GPS trajectory data to traffic flow data,and then to the conversion of traffic emission data.This provides historical data for urban traffic emissions projections.Then,an urban traffic emission prediction model based on deep residual network method and time series prediction knowledge is proposed,which is referred to as deep residual network prediction model.After pre-processing the historical data of a large number of traffic emissions,we conducted model training.By comparing the prediction effects of different residual unit number models,we determined the optimal depth residual network prediction model,which makes the urban traffic emission prediction value closer to the true value.Finally,it is verified by experiments that the deep residual network prediction model can predict the traffic emissions of all regions of the city at the same time,and complete the spatial distribution prediction of traffic emissions.And through the analysis of experimental results,we prove the feasibility of the urban traffic emission intelligent prediction algorithm.