Akademik Veri Yönetim Sistemi
ERZURUM 1ST INTERNATIONAL CONFERENCE ON APPLIED SCIENCES, Erzurum, Türkiye, 9 - 11 Mayıs 2025, ss.110-120, (Tam Metin Bildiri)
Accurate prediction of AC power production in solar plants is crucial for efficient energy management and grid stability. This study compares the performance of two advanced gradient boosting models, CatBoost and LightGBM, in forecasting AC power production using meteorological and temporal features. A publicly available solar plant dataset, comprising 8760 hourly observations, is utilized, including features such as daily yield, total yield, ambient temperature, module temperature, irradiation, and DC power. The dataset is preprocessed by extracting temporal attributes (e.g., hour, day, month) and scaling numerical features. Both models are trained on 80% of the data and evaluated on the remaining 20% using Mean Squared Error (MSE), Root Mean Square Error (RMSE), and R² metrics. Results show that LightGBM outperforms CatBoost, achieving an R² of 0.9984, MSE of 2.41, and RMSE of 1.55, compared to CatBoost’s R² of 0.9923, MSE of 4.09, and RMSE of 2.02. Feature importance analysis reveals that irradiation is a key predictor of AC power production. This study provides practical insights for improving solar energy forecasting and optimizing renewable energy systems.