New Empirical Models for Estimation of Diffuse Radiation using Satellite Data

Diffuse radiation is used in photovoltaic systems and other energy applications. Since global radiation is measured by local meteorological stations, it is possible to reach this radiation data. However, diffuse radiation is not usually measured, so it is not possible to obtain regular data on diffuse radiation. For this reason, efforts are underway to develop various empirical models to estimate diffuse radiation. This work aims to develop new empirical models to estimate the diffuse radiation values for Konya, Turkiye. The empirical models are used to determine the relationship between the diffuse fraction and the clearness index. Data from NASA-surface meteorology and solar energy and the measured global solar is used. The three most suitable developed models are selected, and it is suggested to estimate the diffuse radiation. The developed models consist of 2nd, 3rd, and 4th-order polynomial regression models. The proposed models are tested to evaluate their performances by using eight statistical methods. These are Mean Bias Error (MBE), Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Sum Squared Relative Error (SSRE), Relative Standard Error (RSE), Coefficient of determination (R2), and Correlation Coefficient (r). For the suggested models, the statistical parameter R2 values are calculated as 0.999705413, 1, and 1, and the RSE values are determined as 0.0084537, 0.000122, and 6.05E-06. The proposed models can contribute to the researchers working on the applications of photovoltaic systems. The approaches could be used to estimate diffuse radiation elsewhere under similar climatic conditions.