Statistical Analysis And Forecasting Of Suspended Particulate Matters,PM10and PM2.5

With the rapid urbanization and industrialization, fossil fuel consumption in China is gradually increasing, resulting in the worsening of air quality. Recently, pollutant hazes occur more frequently and widely around the country, which makes a negative influence on people’s normal life and activities. It has become a social concern and needs to be researched and solved urgently. Suspended particulate matters, PM10and PM2.5, is the main reasons of the haze, resulting in harmfulness to human beings. Thus, it is necessary to detect and record the concentrations of suspended particulates, and based on the recorded data we need to forecast its future trends and variations scientifically and effectively.Firstly, from the prospect of meteorological factors and air pollutants, this paper explores the key factors influencing the concentration of PM10and PM2.5. Secondly, the Granger causality test tool is proposed to examine the dynamic relationship between particulate matters and meteorological factors together with air pollutants. Thirdly, in order to obtain better prediction of PM10and PM2.5, this paper proposes ARIMA model and artificial intelligence algorithm as well as Neural Networks based hybrid models to conduct PM10and PM2.5forecasting. Moreover, considering the fact that most traditional predictions are generally deterministic, that is, a certain predictive value. It is not reasonable to be regardless of the possible variations of the forecasts, which may convey detailed information. Therefore, this paper initially proposes ARIMA and CS-based BPNN model to construct interval forecast models, aimed at supplying reasonable fluctuation ranges of the PM10and PM2.5concentrations.