Research on Prediction and Source Analysis of Ozone Pollution

For researchers and practitioners in the environmental field, accurate ozone pollution prediction and source analysis play an important role in smart air monitoring systems. However, there are some challenges in predicting and analyzing the source of ozone pollution in the air. Ozone data are spatio-temporal data, so when analyzing spatio-temporal data, the spatio-temporal correlation of the data should be considered. In this paper, we use temporal graph convolutional network (TGCN) model to solve the problem of ozone prediction. The TGCN model mainly includes two parts. The first part is the graph convolution model (GCN). The main feature of this part is to use graph convolution to process the site structure of the site distribution to extract the spatial correlation between sites; The second part is the gated recurrent unit structure (GRU) network model, which is mainly used to effectively capture the time dimension correlation of ozone data; Combining aforementioned two models can effectively consider the temporal and spatial correlation of ozone data and obtain more accurate prediction results. Meanwhile, this paper proposes to use the Apriori algorithm in the field of statistics to solve the source analysis task. This algorithm takes advantage of the matching correlation between the pollution of the national station and the surrounding micro station area to obtain the correlation between sites, so as to infer the source of pollution after the pollution occur.In this paper, we use TGCN model to learn the temporal and spatial dependence of ozone data in the ozone pollution prediction process, and analyzes the ozone source based on the Apriori algorithm. Experiments show that the prediction accuracy and source analysis results have achieved better results.