A daily PV Plant Power Forecasting Using eXtreme Gradient Boosting Algorithm

Abstract

Accurate forecasting of photovoltaic (PV) plant power generation is essential for optimizing energy management and ensuring grid stability. This study proposes a novel approach to daily PV power forecasting using the eXtreme Gradient Boosting (XGBoost) algorithm, known for its robustness and high predictive accuracy. The dataset consists of hourly PV production and meteorological parameters such as solar irradiation, temperature, humidity, and wind speed. The XGBoost model is trained and validated using real-world data, demonstrating its ability to capture nonlinear relationships and temporal dependencies in the data. The results indicate that our model achieves a significant improvement, with a Mean Absolute Percentage Error (MAPE) reduction of 15% compared to ARIMA and 10% compared to LSTM. Additionally, the Root Mean Square Error (RMSE) is reduced by 12% compared to conventional methods. These improvements demonstrate the effectiveness of XGBoost in enhancing forecasting accuracy. Furthermore, our approach enhances computational efficiency, making it a viable option for large-scale deployment.