期刊名称：International Journal of Energy and Environment
印刷版ISSN：2076-2895
电子版ISSN：2076-2909
出版年度：2012
卷号：3
期号：5
页码：775-786
出版社：International Energy and Environment Foundation (IEEF)
摘要：Estimation of global solar radiation (GSR) is important in most solar energy applications, particularly in design methods, in system characterization and in decision making for energy management. In this paper, a new methodology based on artificial neural networks (ANN) for generating daily GSR data is presented. By modeling GSR in regions where historical records are available, solar potential map for other sites that GSR have not been recorded was generated. In order to examine the ANN models, meteorological data throughout the year 2008 belonging to Karaj city in Alborz province of Iran were used to develop GSR predictors. Input parameters were maximum temperature, relative sunshine duration and extraterrestrial solar radiation while the output parameter was the solar radiation. Various networks were designed and tested and the most accurate model was selected. The best network was found as one hidden layer network with 3-4-1 topology, i.e., a network having four neurons in its hidden layer. To estimate the differences between the measured and predicted values, root mean square error (RMSE), mean absolute error (MAE), mean absolute percentage error (MAPE) and coefficient of determination (R2) were computed as0.66, 0.52, 4.46% and 0.978, respectively. The optimum ANN model was then used to predict GSR in other cities in the province. Data from three stations located in Hashtgerd, Taleghan and Chitgar cities were used as production set. The GSR values for production sites of Hashtgerd, Taleghan and Chigrar were calculated as 4.93, 4.35 and 5.08 kWh m-2 day-1, respectively. Finally, the predicted solar potential values in all stations were integrated and represented in the form of a map. While results are site-specific, the methodology introduced here is general and provides an inexpensive means for GSR prediction based on readily available data.
