Analysis and Comparison of Tree-Based Shallow Learning Methods for Forecasting Solar Energy Production

Accurate forecasting of solar energy production is essential for maintaining energy supply-demand balance, improving planning efficiency, and ensuring grid stability in renewable energy systems. This study applies three tree-based regression algorithms, Decision Tree, Random Forest, and Gradient Boosting, to predict solar power output based on meteorological and temporal data. The dataset combines 70 days of solar power measurements from a photovoltaic system in Lahore, Pakistan, with corresponding atmospheric data obtained from NASA’s MERRA-2 reanalysis platform. To enhance prediction accuracy, the hyperparameters of each model are optimized by GridSearchCV, selecting the coefficient of determination ($\mathbf{R}^{\mathbf{2}}$) as the optimization metric. The predictive performance of the models was evaluated using four error metrics: MAE, RMSE, $\mathbf{R}^{\mathbf{2}}$, and RSE. Among the models, the Random Forest regressor achieved the highest overall accuracy, with an $\mathbf{R}^{\mathbf{2}}$ value of $\mathbf{0. 9 2 6 3}$. The results highlight the effectiveness of ensemble learning techniques, particularly Random Forest and Gradient Boosting, in short-term solar energy forecasting applications.