Prediction Of Surface Net Solar Radiation In Aksu Area Of Xinjiang Based On EEMD-BP Combination Model

The transformation of energy and material on the earth system can not be separated from the drive of solar energy.With solar energy,plants can carry out photosynthesis to produce oxygen to maintain the survival of animals.Photosynthesis is only one of many cycles on the earth,so solar energy plays an indispensable role in the energy and material transfer of the whole earth system.With the more and more serious pollution of the earth and the large consumption of surface energy,the pursuit of sustainable development has become an urgent demand of all countries.In view of how to use the clean energy on the earth more effectively,many countries have introduced a lot of policies to reduce the loss of energy and promote sustainable development.However,there are few solar radiation observation and prediction stations in the world,and they are distributed in different countries.It is not enough to make better use of the solar radiation resources only by using the data of these stations.The lack of data limits the research in this field.As Aksu Area in northwest Xinjiang is located in the high-energy solar radiation area with good data quality,this paper takes this area as the experimental area,establishes the prediction model to verify the prediction accuracy of different models,so as to improve the effective utilization rate of surface net solar radiation energy.In order to improve the prediction accuracy of the net solar radiation in Aksu area,this paper uses the method of combining the EEMD and BP neural network algorithm based on the monthly average data of the surface net solar radiation from January 1961 to December 2010provided by ecwmf,A new EEMD BP neural network combined prediction model is proposed to predict the historical data of net radiation of the surface sun.Firstly,EEMD algorithm is used to extract signals from the original data and layer the data,so as to obtain the eigenmodule function(IMF)and the residual components of different frequencies;The data after decomposition is reconstructed,and the data is divided into three components:high,medium and low frequency.Then,the data of the four coordinate points,i.The model has been applied and verified in the test of the monthly average value of surface net solar radiation in Aksu area,Xinjiang.In the case study,the accuracy and applicability of EEMD BP neural network to predict net solar radiation on the surface are analyzed by comparing with the prediction results of single BP neural network and EMD BP neural network,taking mean percentage error(MPE),mean deviation error(MBE),root mean square error(RMSE)and correlation coefficient as evaluation indexes.Through the above research results show that,compared with the single BP neural network prediction model(R~2=0.7185)and EMD-BP neural network combination prediction model(R~2=0.7357),the prediction value of EEMD-BP neural network combination prediction model is closer to the actual data,has better correlation number(R~2=0.9124),and the error analysis index of prediction result is smaller.The prediction accuracy of the combined prediction model is improved a lot.The first mock exam is applied to the prediction of surface solar radiation in Akesu area of Xinjiang province.The advantages of the two algorithms are overcome,and the defects of single model prediction are overcome.The combined forecasting model of EEMD-BP neural network achieves the desired results.It shows that this model has higher accuracy in the prediction of surface net solar radiation.This combined prediction method can not only be applied to the prediction of surface net solar radiation data,but also provide a new reference method for the prediction of other meteorological data.