Application Of Elman Neural Network On Estimating Daily Irradiation Exposure Of Global Radiation

With the energy crisis approach,humans began to face an energy resource depletion and environmental pollution problems.Countries all over the world try to develop their own new energy technologies,and try to hold the opportunities in the new energy technology innovation.Solar energy is clean and safe energy,and it is major energy sources in the future.Using of solar energy requires a reasonable and effective assessment of solar energy resources.At present,the number of Meteorological Station in most areas of China is relatively small,and the distribution is uneven,which can not satisfy the application of scientific research and engineering project.Therefore,it is especially important to establish a daily total radiation exposure estimation model to predict the daily total radiation exposure.This paper intends to study the daily total radiation exposure prediction in theory and method.In this paper,we discussed the daily total radiation exposure estimation in theory and method,and the daily radiation exposure of Shandong Province is estimated by neural network model.Firstly,the principal component analysis method was used to analyze the data from 2000 to 2003 in Jinan,Fushan and Juxian of Shandong Province,to establish a weight calculation mode.By comparing the weight of each factor,we select five main factors affecting the daily total radiation exposure,namely daily average temperature,sunshine hours,daily average pressure,daily average vapor pressure,and daily relative humidity.Secondly,using the above five factors as the input of the Elman neural network model.Considering the important influence of urban air pollution on solar radiation,add aerosol optical depth as the input of the model.Using the data of 1461 groups from 2000 to 2003 in Jinan,Fushan and Juxian,the Elman neural network model is established to estimate the daily total solar radiation exposure of three meteorological stations in Jinan,Fushan and Juxian of Shandong Province.The results show that the average percentage errors are 21.3%,12.3%,19.7%.The root mean square error is2.20 MJ·m^-2,1.70 MJ·m^-2,1.99 MJ·m^-2.Comparing the prediction results with the prediction results of the GRNN,the Elman neural network model's prediction results are better than GRNN.The average percentage error decreased by 5%-18%,and the root mean square error decreased by 0.506 MJ·m^-2 on average at the three Meteorological Stations.In addition,we obtain an improved Elman neural network model,by adding feedback from the output layer to the hidden layer.The improved Elman neural network model is established to estimate the daily total radiation exposure.The results show:The average percentage error and the root mean square error of the improved Elman neural network model is lower 6%²2%and 0.89 MJ·m^-2 than that of Elman neural network model,respectively.Compared with Elman neural network model,.the goodness of fit the improved Elman neural network model increases.Consequently,the improved Elman neural network model is more reasonable for estimating daily total radiation exposure.Furthermore,it is also helpful for the effective utilization of solar energy resources.Finally,the accuracy of the model prediction results is studied for the training times and training scale of the Elman neural network.The results show:Elman neural network has a stable range of values for the number of training times,when estimating the daily total radiation exposure.This demonstrates the stability and superiority of the Elman neural network in the process of estimating the daily total radiation exposure.With the increase of the training scale of Elman neural network,the average percentage error and root mean square error of the total solar radiation exposure are decreasing.