Research On PM2.5 Concentration Prediction Model Based On EMD-SVR

In recent years,China’s economic development has greatly improved,but at the same time,it has also caused serious pollution,among which PM2.5 pollution is particularly serious.It is important to control the PM2.5 pollution that how to effectively analyze the degree of influence of these factors on PM2.5 concentration,and then achieve the prediction of PM2.5 concentration.This paper takes PM2.5 data of Shijiazhuang City as the research object,through the obtained data,fully considers the impact of pollutant factors and meteorological factors on PM2.5 concentration changes,and establishes a gray correlation analysis model to extract the factors that affect the PM2.5 concentration,and proposes a hybrid prediction model to predict PM2.5 concentration.First,this paper takes statistical analysis methods to analyse the changes of pollutant concentration and meteorological factors in different months,in order to explore the difference between atmospheric pollutant concentration and meteorological factors during heating and non-heating periods.At the same time,the scatterplot is used to analyze the general trend of the influence of atmospheric pollutant concentration and meteorological factors on PM2.5 concentration changes.It lays a theoretical foundation for feature extraction and prediction of influencing factors of PM2.5 concentration during heating and non-heating periods.Secondly,a grey correlation analysis(GRA)model is established to extract the main factors under taking into account the impact of pollutant factors and meteorological factors on PM2.5 concentration changes during the heating and non-heating periods.Firstly,the influence degree value of each influence factor to the change of PM2.5 concentration is calculated by using this model;secondly,the influence degree value is sorted;finally,the main influence factors are analyzed and extracted based on sorted influence degree.Third,based on the extracted main influencing factor characteristics,a mixed EMD-SVR model is constructed to predict PM2.5 concentration in heating and non-heating periods.First,for the nonlinear problem of PM2.5 concentration data,the EMD algorithm is used to decompose it into relatively stable component sequences;secondly,the SVR algorithm is used to predict the component sequences;finally,the component sequence prediction values are added to obtain PM2.5 Predicted concentration.Aiming at the problem that the EMD-SVR model takes a long time to execute,an improved EMD-CFSVR model is proposed on the basis of this model.The Concurrent.Futures parallel framework is used to reduce the running time through parallel execution.Finally,the model proposed in this paper is experimentally verified and analyzed from multiple angles.The model proposed in this paper verifies that it is more practical for predicting PM2.5 concentration.