Developing new solar radiation estimation models with machine learning techniques and testing their efficiency in various places

Abstract

Renewable energy sources have become a popular topic all over the world in terms of cost, efficiency, and environmental pollution. Solar energy is one of the most important renewable energy sources. As the use of solar energy sources increases, modeling the solar radiation (SR) intensity, which is costly and difficult to measure, has become an essential issue. The main aim of this study is to determine the Angstrom coefficients for SR estimation using machine learning (ML) techniques. The linear regression and support vector machine (SVM) regression approaches were used in the MATLAB program to determine the Angstrom coefficients. To examine how the performance of this new approach varies by region, four different regions were identified. To evaluate the performance of the developed models, four different statistical tests were applied. According to these test results, it has been concluded that the developed models with the ML approach in SR estimation are highly successful in general. The SVM and linear regression methods used in the ML algorithm have estimation results very close to the real values in all selected regions. When the results were evaluated in general, it was seen that method SVM regression performed better than method linear regression in all selected regions.