<![CDATA[Accurate prediction of solar irradiance plays a critical role in the planning and operation of renewable energy systems, particularly for photovoltaic integration and energy management. This study investigates the use of a deep learning approach based solely on Convolutional Neural Networks (CNN) to forecast short-term solar irradiance values. The model is trained using normalized multivariate time series data, which include several meteorological parameters as input features. The CNN architecture is designed to extract temporal patterns from the input sequences and predict radiation intensity at the next time step. Experimental results show that the proposed model achieves strong predictive performance, with a Mean Squared Error (MSE) of 0.0006, Root Mean Squared Error (RMSE) of 0.0242, Mean Absolute Error (MAE) of 0.0184, and a coefficient of determination (R²) of 0.9607. These findings demonstrate that CNN, despite its simplicity, is capable of effectively learning complex temporal relationships in solar irradiance data. Furthermore, the loss curves for both training and validation sets indicate stable convergence without signs of overfitting. The results suggest that CNN-based forecasting models can offer a lightweight and accurate solution for real-time solar prediction applications, especially when computational resources are limited.]]>
Copyrights © 2025
