On Prediction Of PM2.5 Concentration In Wuhan Based On Improved Grey Wolf Algorithm And SVR

In recent years,with the rapid development of various fields in the country,the living standards of all aspects of China have been greatly improved,and people's material needs have been rising steadily which resulting in frequent new peaks in industrial production.Therefore,under the highly frequent activities of human beings,the proportion of environmental problems caused by human factors is increasing rapidly,making air pollution increasingly serious and smog appearing frequently,which has become a concern of the current society.Atmospheric particulate matter PM2.5 is only a component of the Earth's atmospheric composition,but it has an important impact on air quality and visibility.Moreover,the particle has a long residence time in the atmosphere,is not easy to diffuse,has strong viscosity,and is easy to be attached with toxic and harmful substances,and thus is highly susceptible to damage to human health.According to research,fine particles(such as PM2.5)can directly affect the ventilation function of the lungs after entering the human alveoli,making the body easily in anoxic state.Therefore,it is of great practical significance to make scientific and effective predictions of PM2.5 concentration,so that people can take timely protective measures to reduce or even avoid harm to the human body.This paper takes Wuhan as an example to select the daily historical data of PM2.5 concentration and its influencing factors from January 1,2014 to October 31,2018.Moreover,the data is preprocessed with missing value interpolation and abnormality detection.Then,from the three perspectives of seasonal replacement,air quality pollutants and meteorological conditions,we can find the variation of PM2.5 concentration and the correlation between them,so as to find the significant factors affecting PM2.5 concentration for next accurate prediction.In order to construct a prediction model with higher accuracy and confidence,this paper uses the original gray wolf algorithm(GWO),improved gray wolf algorithm based on nonlinear convergence factor and dynamic weight to optimize support vector regression(SVR),and finally form four GWO-SVR hybrid prediction models to predict PM2.5 concentrations.Finally,in order to provide a reference for the selection of the best predictive model in practical applications and obtain higher precision PM2.5 concentration,we will evaluate,contrast,and analyze the performance of all aspects of the hybrid forecasting model from the qualitative and quantitative aspects.The results show that the improved GWO-SVR hybrid model has the characteristics of high precision,fast convergence and good universality in predicting PM2.5 concentration.So it can be seen that the improvement of the two algorithms for grey wolf optimization is feasible and meaningful.This paper mainly uses the grey wolf optimization algorithm and support vector regression to construct a hybrid prediction model of PM2.5 concentration,and compares the mixed models to select the optimal hybrid prediction model.The characteristics of this paper are mainly reflected in the combination of the improved gray wolf optimization algorithm and SVR to form a hybrid prediction model,thus the uncertainty of the use of a single SVR algorithm to make predictions only by consulting the literature or previous experience when selecting parameters is overcoming.Compared with the original GWO-SVR hybrid prediction model,the improved hybrid prediction model improves the prediction accuracy,confidence,and the convergence speed is fast and the generalization ability is strong.