Enhancing Photovoltaic Energy Forecasting with Machine Learning: A Comparison Study of XGBoost, LightGBM and LSTM Models

This paper investigates the integration of machine learning (ML) models—XGBoost, LightGBM, and LSTM—in forecasting photovoltaic (PV) energy generation. The study highlights the advancements in photovoltaic technology and the role of automation in enhancing system efficiency and reliability. By leveraging historical data on photovoltaic energy generation and relevant meteorological conditions, the research evaluates the performance of the three ML models using Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and R-squared (R²) metrics. The results show that XGBoost provides the most accurate forecasts, followed by LightGBM, while the LSTM model underperforms in comparison. These findings underscore the potential of ML techniques in improving energy forecasting accuracy, supporting better grid integration and energy management. The study also suggests that future work should explore hybrid methods to further optimize model parameters and enhance predictive performance. Historical and projected values are presented, along with detailed analysis, to offer a comprehensive understanding of the models’ capabilities. The implementation code used in this study is referenced for reproducibility and further research.