Research On Ozone Prediction And Traceability Based On Multi-Source Monitoring Data

Ozone（O₃）has gradually become one of the main pollutants of air pollution in China in recent years.It is of great significance to predict the future concentration of ozone pollution and explore its source after the occurrence of ozone pollution.Air quality monitoring sites for contaminants concentration ozone pollution study provides data support,but based on multi-source monitoring data of ozone concentration prediction and the challenge of traceability study has the following three aspects:Firstly,the ozone has space-time correlation between monitoring sites,and the correlation between affected by meteorological factors such as dynamic wind direction and wind speed.Secondly,there is a complex relationship between ozone pollution and ozone precursor air pollutants.Thirdly,it is difficult to accurately obtain pollutant emission inventory in urban areas.In view of the above challenges,this paper conducts research on ozone pollution concentration prediction and traceability analysis based on multi-source monitoring data from the perspectives of deep learning and swarm intelligence optimization algorithm.The main work content of this paper is described as follows:（1）Aiming at the problems of dynamic temporal correlation of ozone at monitoring stations and complex influence relationship between ozone pollution and its precursor air pollutants,a graph network model based on dynamic temporal attention was proposed to predict the ozone concentration at each monitoring station in the future.Model module through the use of space and time attention to the temporal and spatial correlation between dynamic capture monitoring sites,and the wind direction and wind speed and other meteorological characteristics as external input single after a feed-forward neural network to extract the feature,then with monitoring data fusion after their input sequence prediction model,the dynamic integration of the role of the meteorological factors.For the complex nonlinear relationship between ozone and its precursors,a time-gated convolution module is proposed to extract the time correlation between data and learn the complex nonlinear relationship between ozone and its precursors to improve the prediction accuracy.The experimental results show that the proposed model has greatly improved the prediction accuracy compared with other baseline models.（2）Aiming at the problem that it is difficult to accurately obtain the pollutant emission inventory in urban areas in the study of ozone pollution traceability,an improved firefly algorithm is proposed to trace the source of ozone pollution in urban areas at small scale,and find the main contribution regions of ozone pollution.Firstly,a Gaussian diffusion model was established with the air quality monitoring station as the center,and the dynamic spatial correlation matrix trained by the prediction model was integrated to calculate the ozone concentration diffused to the surrounding firefly locations.Secondly,the transfer entropy of ozone concentration at the monitoring site and firefly location was calculated to quantify the causality of ozone pollution at the two locations.Finally,the calculated transfer entropy was used to improve the firefly location transfer formula.Experimental results show that the algorithm can achieve accurate traceability results by using easily accessible monitoring station data without accurate pollutant source inventory,which provides a new idea for urban pollution traceability research.