Research On Air Quality Index Prediction Based On EMD And LSTM

With the continuous improvement of living standards,people pay more and more attention to their own health problems,and the frequent occurrence of respiratory diseases leads to people’s increasing concern about atmospheric environmental pollution.Due to the nonlinear and complex characteristics of air quality data,the inherent relationship of the original data can not be fully learned in the training process of the deep learning model,which makes the prediction of urban air quality very difficult.In response to the above problems,this thesis proposes an air quality index prediction model based on deep learning.The complex air quality index sequence is decomposed into multiple high-frequency components and low-frequency components by using empirical mode decomposition(EMD),and then the different components are predicted by using long and short term memory neural network(LSTM),Finally,the final prediction result is obtained by summing the predicted values of all components.Finally,aiming at the instability problem of a single model,a combined prediction model based on multiple neural networks is designed.By comparing the prediction results of multiple models with the real results,different weights are assigned to the model,and the final prediction results are obtained after the weighted sum of the prediction values of each prediction model.The main work of this thesis is as follows:(1)Data preprocessing.Because air quality monitoring stations are often disturbed by uncertain factors such as environment and human factors,there will be missing values or abnormal values.This thesis fills in the missing values by means of mean interpolation,and uses Pearson correlation coefficient to analyze the main factors affecting air quality index after filling the complete data,and determine the final input variables into the model.(2)Based on the construction of LSTM neural network model.Aiming at the non-stationary and nonlinear characteristics of air quality data,an air quality prediction method is proposed.This method uses the components processed by EMD as the input of neural network,and uses LSTM as the prediction model to fully learn the useful information in the original data.For the problem that the parameters of LSTM model are usually set manually,the improved particle swarm optimization(IPSO)algorithm is used to optimize the parameters,find the most suitable parameters for the model,and verify the effectiveness of the model using real data sets from multiple cities.(3)The construction of combined forecasting model based on multiple neural networks.In order to solve the problem that the predicted value of a single model is unstable at a certain time,a combined prediction method for air quality using four prediction models EMD-LSTM,EMD-GRU,EMD-BILSTM and EMD-BIGRU was proposed.The results show that this method breaks the limitation of single prediction model and improves the prediction accuracy.