A nonlinear two-parameter model for the spatial analysis of solar irradiation

Nowadays, energy estimation in various application areas is a major research topic. Additionally, various machine learning techniques, especially regression methods and artificial neural networks, have been developed in recent decades to improve the accuracy of such estimates. This article presents a nonlinear compact regression model for estimating the yearly solar irradiation in Africa and Europe by considering only the latitude and mean temperature of the locations as input parameters. The definition of the values of the coefficients is based on the least-square method constrained by the maximum absolute error. The results of 16 conventional regression models, using the same number of predictors, were compared with the result of the model proposed. Our model minimizes the root mean square error by at least 15%.