Akademik Veri Yönetim Sistemi
Thermal Science and Engineering Progress, cilt.66, 2025 (SCI-Expanded, Scopus)
Accurate prediction of photovoltaic (PV) performance under varying thermal and environmental conditions is crucial for reliable energy assessment and investment planning. Commercial software often fails to fully capture the influence of cooling strategies on electrical output. This study examines three PV configurations—uncooled, aluminum fin-cooled (PV_AF), and heat pipe-cooled (PV_HP) using experimental measurements, validated numerical simulations, and machine learning (ML) models including KNN Regressor, AdaBoost, Random Forest, Gradient Boosting, and CatBoost. The tested PV/T collector has a rated maximum power of 50 W (tolerance ± 5 W), open-circuit voltage of 24.62 V, short-circuit current of 2.57 A, maximum power voltage of 20.84 V, and maximum power current of 2.46 A. Experiments were conducted in Gaziantep, Türkiye (37°02′16.6″N, 37°18′51.8″E) during January, April, July, and October 2024 to capture diverse outdoor conditions, including solar radiation, ambient temperature, and wind speed. Results demonstrate significant performance improvements with cooling, particularly under hot conditions: PV_AF and PV_HP increased efficiency by 4.98 % and 5.83 % in January, 5.08 % and 7.50 % in April, 6.36 % and 9.28 % in July, and 4.81 % and 7.68 % in October, respectively. Among ML algorithms, CatBoost achieved the best predictive accuracy (R2 > 0.999). The findings highlight the synergistic benefits of thermal regulation and data-driven modeling in enhancing PV performance.