The Prediction Model Of PM2.5 Concentration Based On Attention Mechanism

In recent years,many factors have brought unprecedented pressure to the atmospheric environment,such as the rapid development of China's social economy and industrialization,the continuous expansion of traffic scale,and the unreasonable energy structure,and the air quality has been highly concerned by environmental managers and citizens.Therefore,how to accurately predict the environmental pollution weather before it comes,and take corresponding measures to reduce the harm of the bad environment is an urgent research problem.People gradually realize that,in the case of more and more serious pollution,it is of great significance to study the prediction of air quality.Carrying out air quality prediction research can not only have a better understanding of the changing trend of air pollution,but also understand the situation of air quality timely,accurately and comprehensively.In addition,the results of effective and accurate air quality prediction can provide important reference for urban environmental pollution control,urban construction and public health.Because the atmospheric environment system is complex and variable,a large number of monitoring data have been accumulated in the past decades,but the traditional prediction model is difficult to capture effective information in a large number of historical monitoring data,resulting in the prediction results are not good.Therefore,it is a reliable tool to establish an effective air pollution prediction model in the air quality prediction research to avoid negative health effects and to formulate effective prevention policies.In recent years,deep learning method has been widely used in all kinds of time series prediction problems,in which attention mechanism shows the powerful ability of processing time series.In this paper,a time series prediction model is proposed by combining the attention mechanism with the long and short time memory network,which is applied to the field of air quality prediction.By virtue of its powerful nonlinear processing capacity and noise tolerance capacity,the efficient prediction of air quality is realized.Based on the long short term memory network and attention mechanism,this paper establishes an Attention-LSTM model for predicting the concentration of PM2.5.Combined with the hourly air quality data of Beijing in 2017-2019 and the meteorological elements data of the same period,two experiments with different input dimensions are designed to predict PM2.5 concentration,using root mean square error(RMSE)and mean absolute error(MAE)Pearson correlation coefficient(P)is used to evaluate the performance of the model.The results show that attention mechanism has obvious advantages in predicting air quality.On the other hand,based on the data set introduced in this paper,through correlation analysis to explore the important factors affecting air quality,combined with the accurate prediction results,to provide more timely and accurate air quality information for the relevant departments to strengthen air pollution prevention and control and prevent serious pollution events.