Prediction Of PM2.5 Concentration Based On A New Hybrid Model Of CEEMD

With the development of urban industrialization and technological advance-ment,the problem of environmental pollution is getting seriously,causing worldwide attention.Especially in recent decades,China’s economy has developed rapidly and the consumption of various energy sources has become larger and larger.Therefore,we have paid a heavy price for environmental degradation.The smoggy weather has occurred in various parts of China in recent years is due to the excessive combustion of some energy fuels,and the amount of various pollutants discharged into the air is seriously caused by the extent that the atmospheric environment can carry.It has had a serious impact on various aspects of personal health and the country’s economic development.At present,PM2.5 is one of the main air pollutants.If we can make accurate predictions about PM2.5 concentration,we can make correspond-ing effective protective measures,effective prevention and control are conducive to human production and life.This paper is based on Complementary Ensemble Empirical Mode Decompo-sition(CEEMD),the Moth-flame Optimization Algorithm(MFO),Support Vector Regression(SVR),Gray Relational Analysis(GRA)and Back Propagation Neural Network(BPNN)and other algorithms.These algorithms are used to predict the PM2.5 concentration in Guiyang,Lijiang,Guangzhou and other cities with differ-ent cultural environments and geographic locations in 2017.An intelligent hybrid model CEEMD-MFO-SVR-GRA-BPNN is proposed.First,the original PM2.5 con-centration data is decomposed into CEEMD and decomposed into three intrinsic mode functions and one residual term.Secondly,the MFO optimization algorith-m is used to optimize the parameters(c,g)in the SVR.Subsequently,GRA is used to screen for atmospheric factors affecting the concentration of PM2.5.Fi-nally,the BPNN model is built for the residual.The proposed model is compared with six models such as CEEMD-MFO-SVR,CEEMD-WOA-SVR,CEEMD-PSO-SVR,EEMD-MFO-SVR,EMD-MFO-SVR and MFO-SVR.The results show that the proposed model is superior to the comparison model in accuracy and general-ization ability,so the proposed hybrid model CEEMD-MFO-SVR-GRA-BPNN can accurately predict the PM2.5 concentration.