Research On Forecasting Model Of Vehicle Emission Factor And System Implementation

Since the 21 st century,the number of motor vehicles in China has increased year by year,which has brought about a large amount of toxic gases emitted by motor vehicles,which not only harms the health of travellers but also is the leading factor in environmental problems,such as smog.In order to regulate traffic flow and vehicle driving conditions,relevant personnel need to use the vehicle emission model to specify a scientific and effective road network-based control strategy.Among them,the correct understanding of the characteristics of vehicle emission factors is the premise of building a vehicle emission model.However,at present,China’s existing vehicle emission models are used for localization in the study of motor vehicle emission factors.The process is complicated and difficult to operate.Therefore,it is of great significance to study a vehicle emission model suitable for China’s traffic conditions.Based on the existing experimental environment,this thesis constructs a predictive model of vehicle emissions data collected by the project through machine learning algorithms and deep neural networks,and compares it with the MOVES(Motor Vehicle Emission Simulator)model developed by the US Environmental Protection Agency.The results show that the vehicle emission model based on deep neural network is superior to the localized MOVES model in accuracy and easier to operate.This thesis mainly does the following three tasks:(1)In order to solve the problem of vehicle emission factor prediction,this thesis firstly uses the Lasso(Least absolute shrinkage and selection operator)regression algorithm and Bayesian regression algorithm in machine learning algorithm to construct the prediction model.The results verify that the machine learning algorithm has certain feasibility in dealing with this problem.Moreover,the results of the comparative experiment show that the Bayesian regression prediction effect is slightly better than the localized MOVES model.(2)This thesis firstly proposes the use of deep neural networks to construct predictive models,and compares them with the predictive models constructed by machine learning algorithms and the localized MOVES models.The results show that the prediction models constructed using deep neural networks are more accurate.Through this experiment,it can be proved that the deep neural network has certain feasibility in dealing with the emission factor of motor vehicles,which not onlyprovides new methods and ideas for the relevant researchers in dealing with the emission factors of motor vehicles,but also for future Research lays the foundation.(3)This thesis designs and implements a vehicle emission factor prediction platform based on B/S architecture.It encapsulates the model trained by deep neural network into an algorithm service,provides API interface externally,and adopts the micro-service design concept.The service is deployed separately,and a highly available and easy to maintain platform is built.The practicality of the vehicle emission model proposed in this thesis has been greatly improved,making this research more practical.