Daily total global solar radiation modeling from several meteorological data

Abstract

This paper investigates the modeling of the daily total global solar radiation in Adana city of Turkey using multi-linear regression (MLR), multi-nonlinear regression (MNLR) and feed-forward artificial neural network (ANN) methods. Several daily meteorological data, i.e., measured sunshine duration, air temperature and wind speed and date of the year, i.e., monthly and daily, were used as independent variables to the MLR, MNLR and ANN models. In order to determine the relationship between the total global solar radiation and other meteorological data, and also to obtain the best independent variables, the MLR and MNLR analyses were performed with the “Stepwise” method in the Statistical Packages for the Social Sciences (SPSS) program. Thus, various models consisting of the combination of the independent variables were constructed and the best input structure was investigated. The performances of all models in the training and testing data sets were compared with the measured daily global solar radiation values. The obtained results indicated that the ANN method was better than the other methods in modeling daily total global solar radiation. For the ANN model, mean absolute error (MAE), mean absolute percentage error (MAPE), correlation coefficient (R) and coefficient of determination (R 2) for the training/testing data set were found to be 0.89/1.00 MJ/m2 day, 7.88/9.23%, 0.9824/0.9751, and 0.9651/0.9508, respectively.