Modelo para la predicción de la radiación solar a partir de redes neuronales artificiales en la ciudad de Nauta
The present study addresses the problem of solar radiation prediction in the city of Nauta, considering its importance for applications in renewable energy and climate management. The main objective was to develop a predictive model based on LSTM artificial neural networks and the RandomForestRegres...
| Authors: | , |
|---|---|
| Format: | master thesis |
| Publication Date: | 2025 |
| Country: | Perú |
| Institution: | Universidad Nacional De La Amazonía Peruana |
| Repository: | UNAPIquitos-Institucional |
| Language: | Spanish |
| OAI Identifier: | oai:repositorio.unapiquitos.edu.pe:20.500.12737/11508 |
| Online Access: | https://hdl.handle.net/20.500.12737/11508 |
| Access Level: | Open access |
| Keyword: | X https://purl.org/pe-repo/ocde/ford#2.02.04 |
| Summary: | The present study addresses the problem of solar radiation prediction in the city of Nauta, considering its importance for applications in renewable energy and climate management. The main objective was to develop a predictive model based on LSTM artificial neural networks and the RandomForestRegressor machine learning algorithm, seeking to improve accuracy by reducing the root mean square error (RMSE) and increasing the correlation coefficient (R). The methodology included the collection of historical meteorological data from SENAMHI, which were preprocessed, normalized and segmented into training, validation and test sets. The models were configured and trained using machine learning techniques and recurrent neural networks. The results showed that the LSTM model obtained an RMSE of 25.94 and an R of 0.764 during training, while the RandomForestRegressor achieved an RMSE of 27.25 and an R of 0.740. However, when using untrained data, the RandomForestRegressor showed a higher generalization capacity with an RMSE of 27.38 compared to 30.04 for the LSTM. It is concluded that both models are effective for predicting solar radiation, highlighting the robustness and adaptability of the RandomForestRegressor and the ability of the LSTM to capture complex temporal patterns. It is recommended to expand the database and adjust hyperparameters to improve the performance of the model in different geographical environments. |
|---|