Prediction and analysis of PM2.5 in Fuling District of Chongqing by artificial neural network

Abstract

The meteorological data, measurements of aerosol optical depth (AOD) and PM2.5 concentration from 2016 to 2017 in Fuling District of Chongqing were selected to study their correlation. The back propagation (BP) artificial neural network (ANN) was used to build a PM2.5 prediction model with the meteorological factors as input, and the predicted PM2.5 values were compared with the measured ones. The results show that PM2.5 concentration has a piecewise linear relationship with temperature attributed to diffusion rate and premise conversion rate, a positive correlation with relative humidity, and a significant inverse correlation with wind speed, but no apparent linear relationship with rainfall, although rainfall has a significant purification effect on PM2.5. The similarity in the influence mechanism of AOD and PM2.5 concentration leads to a certain positive correlation between them. The predicted PM2.5 by the BP ANN model shows a similar trend with the measured one, but has some significant differences in numerical values. Therefore, it is feasible to establish BP artificial neural network to predict PM2.5 by using meteorological data.