Research And Application Of Hybrid Model For Short-term Prediction Of Urban PM_2.5

With the continuous enhancement of China’s comprehensive national strength and the steady improvement of residents’living standards,people are more and more yearning for a green and healthy lifestyle,and pay more attention to the daily air quality.Fine particulate matter(PM_2.5)is one of the representative components of air pollutants.It is of great practical significance to carry out targeted treatment by studying the periodic change of its concentration and to issue environmental prediction to urban residents.However,the traditional physical simulation prediction method needs to consider many natural factors to be measured,while the classical linear statistical method can not deal with the highly nonlinear and non-stationary characteristics of PM_2.5 series data,and may not perform well in predicting the future concentration change and increase or decrease trend.In order to build a reasonable and scientific early warning system,this paper first uses the modified variational mode decomposition（CVMD）algorithm based on the correlation entropy criterion to smooth the sequence,and then uses the support vector regression（SVR）model optimized by the state transition simulated annealing algorithm（STASA）to make nonlinear fitting prediction.Based on the idea of"Decomposition-Hybrid",a new model called"CVMD-STASA-SVR"is established Urban PM_2.5 short-term forecast hybrid model.The model does not need to measure external factors,only needs to consider historical PM_2.5 data,and has high prediction accuracy,low error and strong robustness.The main work of this paper is as follows:（1）This paper introduces the concept of correntropy criterion into the basic variational mode decomposition,and constructs a novel algorithm called CVMD for decomposing time series signals.The correntropy criterion can reasonably determine the number of decomposed intrinsic mode functions,ensure the accuracy of signal decomposition,and avoid the occurrence of redundancy and excessive decomposition.（2）SVR model is used for prediction,and STASA algorithm is used to optimize the parameters selection of SVR.The performance of STASA on complex test functions shows that it has the ability to solve complex modeling parameter optimization problems.Therefore,CVMD is used to decompose the historical concentration series,and STASA-SVR is used to predict the PM_2.5 value in the future.After combining the two,a new hybrid model is proposed for the short-term prediction of urban PM_2.5.（3）The historical data of two regions in Beijing are selected to verify the effectiveness of the model.In the process of modeling,the multi-input prediction strategy is considered,and five indexes are selected to build a reasonable and scientific prediction evaluation system.Three level simulation experiments are carried out with 10 comparison models.The results show that the hybrid model has the best prediction performance,and can be used as an effective prediction model.（4）After verifying the reliability of the model,the multi-step ahead prediction mechanism is added to further improve the model,and the actual prediction is carried out at 9 PM_2.5monitoring points in Taiyuan city.In the future time prediction experiment,the results show that the prediction error increases with the increase of the number of steps,but compared with the existing prediction methods,the hybrid model has excellent performance and strong robustness,and can obtain satisfactory prediction results under different conditions.Again,the model can be used as a reliable solution for short-term prediction of urban PM_2.5.