Research On Air Pollutant Concentration Prediction Based On SSA And PSO Hybrid Optimization Algorithm

Air is essential for the survival and development of all life on Earth,affecting people’s physical health,life,and economic development.Nowadays,due to the continuous development of industrialization and the increase in the number of private cars,air quality is continuously declining,and many air pollutants such as PM2.5,PM10,and SO2 are dispersed in the atmosphere.The concentration of PM2.5 in these air pollutants has attracted much attention.However,the change of PM2.5 concentration is related to multiple influencing factors,which also leads to significant challenges in the prediction of PM2.5 concentration.In this paper,a gated cyclic unit neural network(GRU)prediction model based on a hybrid optimization algorithm of particle swarm optimization(PSO)and sparrow search algorithm(SSA)is proposed to solve the problems of insufficient prediction accuracy and insufficient convergence speed for current PM2.5 concentration prediction models.The main work done in this article is summarized as follows:Firstly,in order to prevent input redundancy in neural networks,the original data in this paper is analyzed and processed.In this article,a series of processing operations were performed on the original data,including removing abnormal values from the data,adding missing values from the data,and standardizing the data.Subsequently,the correlation between the data,namely,the relevant influencing factors of PM2.5concentration,including the correlation and periodicity of PM2.5 concentration itself,the impact of other air pollutant factors and meteorological factors,and the spatial correlation of PM2.5 concentration,was analyzed.Secondly,by analyzing the advantages of PSO algorithm and SSA algorithm in previous experiments,we introduced SSA algorithm into the execution process of PSO algorithm,which not only optimizes the position and vector of particles in PSO algorithm,but also preserves the diversity of population.Finally,we constructed a SSA-PSO hybrid optimization algorithm.And then,Testing the optimization algorithm’s performance with ten test functions,the results demonstrate its good optimization effect,rapid optimization speed,and ability to converge to the global optimal solution at a quicker rate.The hyperparameters in neural networks greatly affect the prediction performance of the model.Therefore,in this paper,we use the SSA-PSO optimization algorithm to optimize the hyperparameters of GRU.Finally,an experiment is conducted using the SSA-PSO-GRU prediction model constructed in this article.The prediction results of the model were compared with the results of seven models,namely,support vector regression model,radial basis function model,K-nearest neighbor algorithm,BP neural network model,GRU neural network model,PSO-GRU neural network model,and SSA-GRU neural network model,as well as the comparative analysis of the three evaluation indicators.The results show that the prediction performance of the model proposed in this paper is better than that of the comparison model,with root mean square error(RMSE)and mean absolute error(MAE)decreasing to varying degrees,and goodness of fit(R~2)improving to varying degrees.