Research On Air Pollutant Concentration Prediction With Three-population Coevolutionary Algorithm

Air pollution will seriously affect human life and health.How to control air pollution and ensure people’s daily life and economic activities has been one of the essential issues that the government and environmental protection departments have paid attention to in recent years.Accurate and efficient prediction of pollutant concentration is the premise and foundation for effective air pollution control.However,in air quality prediction,many factors affect the pollutant concentration of the target site.Valid factors can improve the operation efficiency and prediction performance of the model.This paper proposes a prediction model of air pollutants based on a triplepopulation coevolutionary algorithm.The model takes the extreme learning machine as the basic prediction model.It uses the cooperative coevolutionary method to establish three subpopulations to realize the selection of important climate factors,the selection of monitoring sites,and the parameter optimization of the extreme learning machine.The coevolution of three populations completes the prediction of air pollutant concentration.A limited number of feature subsets are used to train the model to improve the prediction accuracy and efficiency.This paper takes the actual air quality data set as an example and uses the proposed model and comparison algorithm to predict the pollutant concentration at 24 hours,48 hours,and 72 hours.Experimental results show that feature selection can effectively improve the prediction effect.In addition,the proposed model can use smaller feature subsets to obtain higher prediction accuracy than other compared algorithms,which verifies that the model can generate better prediction performance in the prediction of air pollutant concentration.