Akademik Veri Yönetim Sistemi
APPLIED THERMAL ENGINEERING, cilt.298, ss.1-5, 2026 (SCI-Expanded, Scopus)
This study investigates the performance of a small-scale solar chimney power plant (SCPP) experimentally and through machine learning modelling under real Saharan conditions in Ouargla, Algeria. The experimental setup consists of a 4 × 4 m solar collector and a 6 m-high chimney instrumented with calibrated thermocouples, anemometers, and solar sensors. Measurements performed under solar irradiance levels between 700 and 1080 W/m2 showed that the internal updraft velocity ranged from 0.9 to 3.4 m/s, with significant influence from ambient wind. A dataset comprising 452 samples was used to train several machine learning algorithms, including 20 input features representing thermal, geometric, and meteorological variables and one target variable. Among them, the XGBoost model achieved the best performance (R2 = 0.753, MSE = 0.046, MAE = 0.167 m/s), outperforming all other algorithms tested. Feature-importance analysis identified solar irradiance and ambient wind velocity as the dominant predictors. The inclusion of crosswind directionality introduced a novel aerodynamic perspective, demonstrating that machine learning can determine optimal collector inlet orientations to enhance airflow stability. The proposed hybrid experimental-AI framework offers a reliable and adaptable approach for optimising solar chimney systems in arid and semi-arid regions.