Akademik Veri Yönetim Sistemi
Engineering Applications of Artificial Intelligence, cilt.145, 2025 (SCI-Expanded, Scopus)
The aim of this research is to improve electricity generation efficiency and contribute to reducing energy costs by positioning solar panels at the optimal tilt angle in Turkey, a country with high solar energy potential. To achieve this goal, more advanced methods such as machine learning techniques are preferred instead of traditional methods calculated by mathematical formulae. In a typical Particle Swarm Optimisation (PSO) application, mathematical equations are used as the fitness function, but such an approach can be time, labour, and performance consuming. In this study, a new method called ExtraPSO (Extra Tree with Particle Swarm Optimisation) is developed by using K-Nearest Neighbors, AdaBoost (Adaptive Boosting), Gradient Boosting, Random Forest, ExtraTrees algorithms instead of a mathematical model for the fitness function. As a result of the testing of ExtraPSO, it was observed that a powerful predictive hybrid model was obtained in addition to saving labour and time. The prediction success of the developed ExtraPSO model is determined as 99.0549% with the Accuracy metric. In addition, the optimum tilt angle and maximum exergy amount were determined for Turkey according to the seasons. Considering all seasons and annual average for Turkey, the maximum exergy amount that can be obtained for a solar panel placed at the optimum tilt angle is determined as 3.871 W/m2, 5.713 W/m2, 7.048 W/m2, 5.272 W/m2 and 5.285 W/m2 for winter, spring, summer, fall and annual average, respectively.