A Hybird Optimization Models For PM_2.5 Concentration Forecasting

With the rapid development of economy,the process of urbanization has been accelerated,in the meantime environmental problems have been paid more and more attention.PM_2.5 concentration is an important indicator of air quality,it is the general name of the solid or liquid droplets in the air with a diameter less than or equal to 2.5um.High concentration of PM_2.5 in the atmosphere indicates serious environmental pollution.PM_2.5 will bring great harm to people's health.Therefore,it is of great significance to predict the concentration of PM_2.5 reasonably and effectively.By studying the PM_2.5 concentration of four sites in Hefei,a hybrid prediction model is proposed to improve the prediction accuracy and stability of the model.By using EEMD,WD and SSA to remove the high frequency noise of the original data in data preprocessing,and the use of fuzzy entropy measure greatly saves computing resources.Several commonly used artificial neural networks and support vector machine have been used in the forecasting process.The neural network ELMAN,which shows excellent performance at all stations,is selected as the final prediction method through the experiment.In order to avoid the traditional optimization algorithm which may easily fall into the local optimal,a hybrid optimization algorithm GASPO has been used.The optimization algorithm can optimize the weights of the input and hidden layers and the threshold between the hidden layer and the output layer of the neural network.By comparing the prediction results of three hybrid optimal prediction models FEEMD-ELMAN-GA?FEEMD-ELMAN-PSO and FEEMD-ELMAN-GAPSO at 4 sites show that the prediction accuracy is better than the single prediction model by using the data preprocessing method and the hybrid optimization algorithm.The model proposed in t paper has achieved good result in the prediction of PM_2.5 concentration.