Hybrid machine learning-based framework for predicting the output of daily solar irradiation

Providing guidance for solar energy conversion systems, selecting appropriate places, and even future investment policies all rely heavily on accurate predictions of daily sun irradiation (DSI) for the regions. The purpose of this paper is to conclude the DSI values of Turkish provinces with notably diverse patterns of exposure to the sun. The study recommended using a machine learning (ML) technique for DSI prediction; this technique was the hybrid Adaboost optimized decision tree (HAB-DT) algorithm. The algorithm is trained using the provinces' minimum and maximum daily ambient temperatures, cloud coverage, daily extraterrestrial solar irradiation (SI), day length, and SI. The records go back two years and come from the Turkish State Meteorological Service (Jan-1-2018 to Dec-31-2019). Throughout every scenario tested, HAB-DT demonstrated the best performance in terms of R2, RMSE, and MAPE. In conclusion, this research reports that the suggested and existing algorithms examined in this study can be employed in the prediction of DSI data with high accuracy; however, the HAB-DT method is the best proper algorithm among all algorithms. Following comes k-NN, then ANN, then DL, and finally SVM.