Research On Spatiotemporal Characteristics And Prediction Of Urban Small Motor Vehicle Traffic Emissions Based On GPS Trajectory Data

With the rapid development of China’s economy and accelerated urbanization,the prosperity of the transportation industry has brought great convenience to people’s travel and goods transportation,but also caused a significant negative impact on the environment.Motor vehicle emissions have become one of the main sources of urban air pollution.Therefore,how to accurately predict the temporal and spatial evolution of urban traffic emissions is of great significance for urban planning and policy making.The existing traffic emission estimates tend to be based on historical and current traffic emissions estimates,and are mostly limited to the overall fuel-based estimates,ignoring the uneven spatial and temporal distribution of traffic emissions,which cannot meet the needs of future urban management and planning.With the development of information technology,theoretical methods for using traffic big data to solve urban traffic problems are becoming increasingly mature.In this paper,we analyze the GPS data of cabs,focus on the main line of " basic data processing and spatiotemporal matching-traffic emissions estimation using modified MOVES models-spatiotemporal distribution characteristics of traffic emissions-traffic emissions prediction",and carries out the research on the traffic emission prediction.The main research content and findings of this paper include the following four aspects:Firstly,this paper cleans GPS trajectory data and converts coordinates.based on which the spatiotemporal step of the grid is determined with the number of GPS trajectory points in the invalid grid.It spatiotemporally matches the grid and trajectory data,and extracts traffic conditions such as vehicle flow,average speed,and average acceleration from the grid.Secondly,based on the extracted operational attributes of small motor vehicles,this paper calculates the corresponding specific power,and combines the geographical information,vehicle information,fuel information,and deterioration index of Xi’an City to locally modify the MOVES model,providing a calculation method for the traffic emission factors and traffic emissions of small motor vehicles in Xi’an City.Thirdly,based on the traffic status extracted from GPS trajectories,this paper uses the modified MOVES model to calculate the traffic emission factors and traffic emissions of small motor vehicles in Xi’an.Through the GA-FCM clustering algorithm and spatial autocorrelation analysis,the spatiotemporal characteristics of the traffic emission factors and traffic emissions are analyzed.The results show that the temporal and spatial distribution of traffic emission factors and traffic emissions has obvious regularity,and there are significant differences in different regions and times.Finally,this paper based on the spatial and temporal distribution characteristics of traffic emissions,selects 14 variables,such as time,space,weather,and history,to construct a multifactor variable input matrix,and performs smoothing and standardization processing.The random forest method was used to filter out eight more important variables.Through genetic algorithms,the optimal LSTM layers,the number of Dense layers,the number of Dense layer neurons,and the number of hidden layer neurons are found.An LSTM model is constructed to predict the traffic emissions in the study area.Comparison with models such as GRU reveals that the GA-LSTM model has the best prediction results with an accuracy of 85% on.