Based On VMD Decomposition And Support Vector Regression PM2.5 Concentration Prediction

Fine particulate matter(PM2.5)mainly exists in the air and will enter the lungs of human body through breathing,affecting the respiratory function,and even affecting the nervous system,respiratory system,cardiovascular system,etc.Thus,for PM2.5,It is necessary to control the daily concentration.Once the potentially harmful concentration has been predicted,the relevant departments can remind the public ahead of time and take action to minimize the loss.The goal of this paper was to establish a PM2.5 prediction model based on variational mode decomposition(VMD)and support vector regression(SVR),and the forecast method of decomposition and integrationm was used to forecast PM2.5 concentration,and realized the short term forecast of PM2.5 concentration in Wuhan.To improve the accuracy of the PM2.5mass concentration prediction,the model first uses the VMD to perform a decomposition of the original PM2.5 mass concentration time series,and obtains multiple relatively stable components with different time-scales;then uses the SVR algorithm to predict each of the components separately;The final step is to calculate the sum of the predicted values for each component as the predicted result of the original PM2.5 mass concentration.The data used is taken from the daily mean PM2.5 mass concentration from 14 May2016 to 24 September 2017 published by Wuhan Meteorological Bureau as a set of experimental samples.We adopt five forms of prediction,namely autoregressive integrated moving average(ARMA),BP neural network,random forest(RF),support vector regression(SVR),and support vector regression based on variational mode decomposition(VMD-SVR)proposed in.PM2.5 concentration is predicted in Wuhan,and the accuracy of the prediction is compared and analysed.Experimental results show that the method of this paper can indeed improve the mass concentration prediction accuracy of PM2.5.