Research On NO₂ And SO₂ Prediction Based On Data Frequency Decomposition Hybrid Model

For a long time,the harm of air pollution to human health has been closely watched by the public.Many scholars have studied the impact of air pollution on human health.SO₂ and NO₂ are two common and important air pollutants and are the main forming substances of acid rain.Acid rain poses a threat to buildings,crops and even human health.Therefore,the development of effective SO₂ and NO₂ prediction and early warning models is of great significance for reducing or even avoiding the harm of this pollutant.However,it is hard to obtain accurate predictions of irregular SO₂ and NO₂ data sequences.Based on the current data decomposition,whether the high-frequency sequence needs modeling analysis and whether high-frequency sequence and low-frequency sequence modeling need to introduce intelligent optimization algorithm,This study takes SO₂ and NO₂ as examples to conduct in-depth research.The Central China region has long been the most severe area of acid rain in China.Therefore,the daily average SO₂ and NO₂ data sequences of Xinyang,Jingmen,Yueyang and Changde in Central China are selected as modeling data to explore the hybrid modeling problem based on complementary set empirical mode decomposition?CEEMD?.The proposed procedure is called two-step hybrid model.The detailed process of the proposed model can be summarized into three steps:First of all,the original SO₂ and NO₂ sequences are decomposed into high frequency and low frequency sequences by CEEMD;Then,Support Vector Regression?SVR?model combined the Cuckoo Search algorithm?CS?and Grey Wolf Optimizer algorithm?GWO?are employed to model the high-frequency and low frequency sequences,respectively;Thirdly,the low frequency prediction data and the high frequency prediction data are added to obtain the final prediction results of SO₂and NO₂,respectively.In terms of model selection and evaluation process,the proposed model and other models are tested and compared by predictive error evaluation indicators such as mean absolute error?MAE?,mean absolute percentage error?MAPE?and root mean square error?RMSE?.By comparing the prediction results,compared with other prediction models,the two-step hybrid model based on data preprocessing and intelligent optimization algorithms?CS and GWO?has higher prediction accuracy.In particularly,the CEEMD-CS-GWO-SVR hybrid model is the best model for predicting SO₂ and NO₂.This paper proposes important theoretical guiding ideas for establishing urban air pollution models in Central China,and also provides theoretical support for air pollution warning system.