Modeling Global Solar Radiation over Iran based on Meteorological Data Using ANN Technique

Among the renewable sources, Iran has a high potential of solar energy. The main step of designing new solar plants is sites selection. Monthly mean daily global solar radiation data are essential to achieve this important goal. However, these data are not available as a function of geographical and meteorological parameters. In this case, an ANN algorithm was engaged to establish a forward/reverse correspondence between the latitude, longitude, altitude, month of the year, minimum atmospheric temperature, maximum atmospheric temperature, minimum earth temperature, maximum earth temperature, relative humidity, wind speed, participation, atmospheric pressure, sunshine duration and monthly mean solar irradiation. For this purpose, the meteorological data of 31 stations of Iran along the years 1983–2005 were used as training (27 stations) and testing (4 stations) data. The Stepwise Multi Non-Linear Regression (MNLR) method was applied to determine the most suitable input variables. In order to investigate the effect of each meteorological variable, ten ANN-models were developed by using different combinations of the most suitable variables as inputs. The results showed that the ANN10 has a very good architecture for the prediction of monthly mean daily global solar radiation in Iran with an average correlation coefficient of more than 99.5% that performs a more accurate prediction than the other ANN models. It is concluded that the proposed approach can be used as an efficient tool for prediction of solar radiation in the remote and rural locations with no direct measurement equipment

کلیدواژه ها:

Solar energy potential ، Artificial neural network ، Daily global solar radiation ، Iran