Abstract:
"The integration of renewable energy, particularly solar power, into the global power grid is critical to addressing environmental challenges. Accurate forecasting of solar power output from photovoltaic systems is essential for effective energy management and stable grid integration. However, the variability of solar generation due to meteorological factors presents significant challenges in achieving reliable predictions.
This research focuses on improving solar power forecasting using a transformer-based encoder model. Leveraging advanced attention mechanisms, the model effectively captures long-term and short-term dependencies in solar power and weather data. The methodology includes extensive data preprocessing, feature engineering, and iterative model refinement to enhance accuracy and efficiency. By employing deep learning techniques, the transformer-based model provides a robust solution to the limitations of traditional forecasting methods.
The model's performance is validated on a test dataset, achieving a Mean Absolute Error (MAE) of 0.7018, Mean Squared Error (MSE) of 0.6918, and Root Mean Squared Error (RMSE) of 0.8317. These results highlight the model's superior ability to forecast solar power output, offering a significant improvement over conventional methods. This study sets a foundation for future advancements in solar power forecasting, supporting the broader adoption of renewable energy and enhancing grid reliability."
