Interval Prediction Of Air Quality Index Based On Variational Mode Decomposition And Harmony Search Algorithm

Air Quality Index is an index that quantitatively describes the status of air quality.The establishment of an Air Quality Index prediction model can support relevant authorities to release preventive measures of air pollutants and carry out duties in public health protection.Therefore,it is also helpful to avoid the health threat caused by excessive concentration of air pollutants.In this paper,an interval prediction model of support vector regression（SVR）based on scale factor rule,variational mode decomposition（VMD）,harmony search（HS）algorithm is proposed.The main steps of constructing the model are as follows:（1）VMD technology is introduced,by selecting different values of K to decompose the series and analyzing the characteristics of the decomposed mode components,the optimal number of mode decomposition and the final mode components are obtained;（2）The mode components are grouped by sample entropy,and the random components are denoised based on singular spectrum analysis;（3）The interval prediction of Air Quality Index is carried out.Firstly,scale factor rule is used to construct each component interval.Secondly,taking the comprehensive evaluation index I_PICI as the objective function,the scale factors of each component and the parameters of different basic models（support vector regression and recurrent neural network）are optimized synchronously through different optimization algorithms（harmony search algorithm and genetic algorithm）.Finally,the predicted values of each component are added to obtain the final prediction interval.At the same time,compared with the results of other interval prediction models（probability statistics method based on a priori hypothesis and interval prediction method based on quantile regression）,it can be obtained that the model based on scale factor rule is the best in the optimized space and interval prediction evaluation.Among them,the SVR model based on VMD and HS algorithm is the best.The innovation of this paper is to propose a new interval prediction model.The model uses the comprehensive evaluation index I_PICI constructed in this paper as the objective function to optimize the scale factors and model parameters synchronously,which avoids the limitations of artificially setting the proportion coefficient and model parameters.The air quality indexes of three provincial capitals in the Yellow River Basin are predicted.Finally,the results of different evaluation indexes confirm the effectiveness and superiority of the model.