| Atmospheric haze is mainly caused by tiny particles, such as PM2.5. PM2.5 particles have a complex cause of formation, and can lead to serious consequences. In order to forecast fine particulate concentration and strengthen the control of pollution, it is very important to develop the relevant researches and promote pollution control. However, it is hard to realize accurate forecast by making full use of historical data by traditional ways. In this paper, Back Propagation Neural Network was applied to forecasting the concentration of PM2.5. And PM2.5 pollution forecast model based on the Back Propagation Neural Network in Wuhan city was built up by software MATLAB. A simulation model was able to realize with the aid of the language MATLAB.Firstly, Time series analysis model is used to forecast the concentration of PM2.5. Subsequently, the model of principal component analysis and stepwise regression were constructed after drawing the scatter-plot graphs of six indices in API and analyzing their correlation. Then we obtain the relative optimal model. And we can predict the concentration of PM2.5 of the next month.The models established by traditional methods are always linear.Thus, prediction accuracy is hard to guarantee. Back propagation neural network can offer ideal mathematical models, which shows its ability of processing non-linear problem.Last, Back propagation neural network was taken into consideration in forecasting the concentration of PM2.5. In all 365 test samples, predictive value completely conformed to the actual value was 240, 65.8% of total days; 337 days had the difference not more than one grade,which 92.3% of total days. In a word, the predictive results were basically consistent with the actual situation. This paper proved by positive practice that it was feasible to apply Back Propagation Neural Network in air pollution forecast in Wuhan city. It also provide a new way of thought and methods for air pollution forecast scientifically and effectively. |
PDF Full Text Request