Predicción de nivel de energía solar usando modelos de Inteligencia Artificial.

Abstract

Abstract: This research addresses the prediction of photovoltaic solar energy generation in the city of Loja, evaluating the performance of Artificial Intelligence models under local climatic variability. By analyzing meteorological and production data at five-minute and one-hour resolutions, we compared Machine Learning algorithms (Random Forest, XGBoost), Deep Learning models (LSTM, GRU), and statistical models (SARIMA). The results show that decision-tree based models provide the highest accuracy, with Random Forest standing out for short-term forecasting and XGBoost for long-term forecasting, outperforming neural networks due to the characteristics of the dataset. The main contribution lies in validating an efficient predictive methodology for equatorial regions and in designing network architecture based on Edge Computing for operational deployment, thereby facilitating the integration of renewables and energy planning.