Akademik Veri Yönetim Sistemi
ERZURUM 1ST INTERNATIONAL CONFERENCE ON APPLIED SCIENCES, Erzurum, Türkiye, 9 - 11 Mayıs 2025, ss.120-130, (Tam Metin Bildiri)
Solar energy forecasting is crucial for efficient energy management and grid stability. This study compares the performance of two gradient boosting algorithms, CatBoost and Gradient Boosting, in predicting solar power production using meteorological and temporal features. A solar power plant dataset with 8760 hourly observations was utilized, containing features such as wind speed, radiation, air temperature, and lagged production values. The dataset was preprocessed by scaling features and adding temporal attributes such as hour, day, and season. Both models were trained with 50 iterations and evaluated using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and R² metrics. Results show that CatBoost achieved an R² of 0.8993, RMSE of 0.0586, and MAE of 0.0235, outperforming Gradient Boosting, which yielded an R² of 0.8894, RMSE of 0.0614, and MAE of 0.0232. The study highlights the effectiveness of CatBoost in solar energy forecasting and provides insights into feature importance, with radiation showing a high correlation (0.89) with system production. Future work could explore additional features, such as cloud cover, to further enhance forecasting accuracy.