Abstract. Atmospheric particulate matter is one of the most harmful pollutants in the atmospheric environment, and has an important impact on climate change, reduced visibility, environmental hazards (such as acid rain, smoke) and health hazards. PM2.5 is the main factor causing haze weather, reducing visibility and affecting traffic safety. PM2.5 enters the alveoli through the respiratory tract and endangers human health. The basic characteristics of PM2.5 are small size, light weight, long residence time in the atmosphere, and can be transported to a long distance by the atmospheric circulation, causing a wide range of air pollution.In 2018, the proportion of days with excellent and good air quality in Guilin City in January and February was 71.2%, ranking 10th among the 14 cities in the Guangxi Autonomous Region. As a famous tourist city, its air quality should be paid more attention. The main air pollutant in Guilin City is PM2.5, which refers to fine particles with a diameter of 2.5 μm or less in the air, it has great harm to human health and reduces the visibility of the atmosphere. Air quality data, meteorological data, and GPS tropospheric delay data (ZTD: zenith tropospheric delay) were collected to calculate the correlation between PM2.5 concentration and influencing factors by the gray correlation model, and the relationship between the main influencing factors and the variation characteristics of PM2.5 concentration was analyzed. The result shows that PM2.5 has strong correlation with other air pollutants SO2, NO2, wind direction, relative humidity, ZTD, temperature and wind speed; moderately correlated with rainfall and pressure. The degree of gray correlation between the impact factor and PM2.5 varies with the seasons.Based on this research and analysis, air quality pollutants and meteorological factors are used as input factors and the concentration of PM2.5 values as the output factor of the neural network model. The model is trained by training sample data to establish a neural network model to predict the concentration value of PM2.5 in Guilin. The accuracy of the model is verified by the accuracy index. The summer neural network model has the best precision. The MAE is 4.51 μg/m3, the RMSE is 5.73 μg/m3, and the MRE is 15.1%. The correlation coefficient between the predicted value and the measured value reaches 0.908. It shows that the neural network model based on meteorological factors and ZTD has a good predictive effect and has certain guiding significance for the prevention and control of air pollution in Guilin.

Abstract. There exists the shortage of low accuracy when using BP neural network model to predict PM2.5 concentration in air. An improved particle swarm optimization (IPSO) algorithm combined with BP neural network was proposed. Using the advantages of improved PSO algorithm global optimization ability, the weight and threshold of BP neural network are optimized, pollutant data and meteorological data are used as input data, PM2.5 concentration is used as output data, and IPSO-BP model is established for simulation prediction. Comparing and analyzing the IPSO-BP model, PSO-BP model and BP model, the results show that the MAE and RMSE of the IPSO-BP model are 6.94 and 8.47, respectively, and the R2 is 0.77. The accuracy test indicators are better than the PSO-BP model, and the BP model, PM2.5 concentration has the highest prediction accuracy, validating the validity of the model's prediction of PM2.5 concentration.