This study proposes an integrated methodology for the technical sizing and performance optimization of photovoltaic (PV) systems under variable climatic conditions. The approach combines machine learning (ML)-based predictive modeling with physical system design to enhance accuracy and adaptability. Three regression-based ML algorithms, which are Polynomial Regression (PR), Gradient Boosting Regression (GBR), and eXtreme Gradient Boosting (XGBoost), were implemented to estimate the recoverable solar energy (E sri ). Among them, the XGBoost model achieved the highest predictive performance, with an R² value of 97.33%, demonstrating an excellent agreement between measured and predicted values. The analysis further examined the optimization of the tilt angle (β), highlighting that higher inclination angles are optimal during winter months, whereas smaller angles yield better results in summer. The predicted E sri values exhibited strong consistency with experimental data obtained from a meteorological station, confirming the robustness of the proposed predictive framework. Moreover, the daily variation in PV energy production closely matched the recoverable solar energy, validating the analytical energy model. Finally, based on the assessment of load profiles, the optimal number of PV panels required for each month was determined, enabling efficient system sizing and reducing dependence on the electrical grid.
Said et al. (Sun,) studied this question.