Comparative Analysis of Ensemble Learning Techniques for Solar Panel Capacity Forecasting in KSA

This research investigates the efficacy of ensemble learning techniques for enhancing solar panel capacity forecasting in a specific location in Saudi Arabia, a nation rapidly advancing in renewable energy. The study, conducted under the aegis of various academic and research institutions, employs crystalline solar panels in Dharan to collect environmental data at high frequency. The dataset is rigorously evaluated using to-fold cross-validation to assess the performance of three ensemble machine learning algorithms: Bagging, Additive Regression (boosting), and Stacking. The findings demonstrate that Bagging consistently outperforms other models in predicting solar panel efficiency, as evidenced by high Coefficient of Determination (CC) values and low Root Mean Squared Error (RMSE) across key performance metrics (Umax, Pmax, and Imax). The research underscores the potential of Bagging for accurate solar power forecasting and suggests future exploration of temporal factors, weather forecasting, and hybrid energy systems to further optimize renewable energy solutions.