This research presents a Long Short-Term Memory (LSTM) Recurrent Neural Network (RNN) model designed to predict photovoltaic generation in San Antonio, Pichincha Province, Ecuador. With the growing importance of renewable energy, accurately predicting solar radiation is essential for optimizing photovoltaic systems. This study focuses on using deep learning techniques to improve prediction accuracy. Traditional forecasting models struggle to handle high climatic variability, leading to inconsistent energy forecasts. The challenge is to enhance prediction reliability for better energy management. The main objective is to develop an LSTM-based model to predict solar radiation with high accuracy. The novelty of this research lies in applying deep learning techniques to Ecuador’s specific climatic conditions, improving energy planning and grid stability. Historical solar radiation data from 2017 to 2023 were used, with 80% for training and 20% for validation. The model achieved an accuracy of 97.71% and a mean squared error (MSE) of 0.04771. Predictions for the first three months of 2024 indicate radiation values exceeding 400 W/m2 with a determination coefficient (R2) of 0.7194, demonstrating the model’s reliability for photovoltaic energy management.