COMPARISON OF MEASURED GLOBAL SOLAR RADIATION FOR SIVAS PROVINCE WITH COMMON RADIATION MODELS BY ESTIMATING WITH ARTIFICIAL NEURAL NETWORKS (ANN)

Solar Radiation (SR) values measured for Sivas between June 2018 - May 2019 were modeled using an ANN. Performances of common models in literature to predict the daily total global solar irradiation values were examined. Some meteorological data such as daily average air temperature were obtained from General Directorate of State Meteorology Affairs (GDSMA). SR and geographical variables were measured with a pyranometer. For the statistical analysis of the calculation methods used to prediction the monthly average total SR and duration, the measured data were used as reference. Root of mean squares error (RMSE) was compared using the mean bias error value (MBE) and t-statistics (t-st) methods for performance evaluation of the equations. The most suitable equality was found by comparing the measured and calculated values. As a result, Model 1 (Angstrom-Prescott-Page) with a determination coefficient (R2) of 0.979 showed the best prediction performance among the models examined. This model compared to other models; MBE was found to be 0.380 and RMSE was found to be 0.936. if α = 0.01 statistical significance level is below the acceptable t-critical = 3.106 value (0.164-1.748), it shows that this model will be a suitable model for Sivas province.