The Application Of Long Short-Term Memory In Air Quality Prediction

With the improvement of social and economic level,people are more and more concerned about air quality and the current changes of air quality.Therefore,how to accurately and scientifically predict air quality has important strategic and practical significance for environmental protection departments and the public.However,most of the existing air quality prediction studies do not consider the significant impact of emergency policy factors on air quality.Based on this,this paper forecasts the air quality index(AQI)of Hangzhou by adding the improved EMD-GA-LSTM prediction model of empirical mode decomposition(EMD),and analyzes the impact of relevant emergency policies on the change of air quality.The research framework of this paper can be divided into four stages:the first stage is a general analysis of the change of air quality in Hangzhou.The environmental control policies of the 2016’s G20 summit and the influence of COVID19’s isolation policy on air quality in 2020 are emphatically analyzed by the phased and backward comparison method.In the second stage,the meteorological data and air quality data of Hangzhou are preprocessed.After the first stage of policy impact analysis,add policy characteristics;At the same time,considering the influence of seasonal factors,seasonal characteristics are added.In the third stage,the EMD-GALSTM prediction model is constructed based on the above data.In this model,the original AQI sequence is decomposed into finite eigenmode function(IMF)components and one residual component by EMD.Secondly,the LSTM prediction model is built for each component sequence,and the genetic algorithm is used to find the optimal number of layers and neurons in the hidden layer and full connection layer of LSTM model.Then,the cumulative value of the prediction results of each component series is taken as the prediction value of the original AQI series.Finally,the EMD-GA-LSTM prediction model is compared with GA-LSTM model,single LSTM model and RNN model.In the fourth stage,the EMD-GA-LSTM model with and without policy features is used to predict the air quality index(AQI)of Hangzhou from January 10,2022 to March 10,2022 respectively,and the prediction results are compared with the real value of Hangzhou AQI under the "welcoming the Asian Games" policy.The results show that compared with GA-LSTM model,single LSTM model and RNN model,EMD-GA-LSTM prediction model proposed in this paper has the lowest Mae and RMSE and the highest R2 when predicting AQI,which proves that this model improves the performance of the original model;Using the EMD-GA-LSTM model with and without emergency policy characteristics to predict the AQI during the Asian Games,the results show that the prediction model with emergency policy characteristics has a better prediction effect on the AQI during the Asian Games,which proves that the emergency policy affects the air quality to a great extent.Therefore,when formulating the air quality control plan,in addition to the long-term environmental control policy,we can also consider adding phased emergency policies to speed up the improvement of air quality.