Publication: Aplicaciones de Deep-Learning a la predicción de radiación solar
Loading...
Official URL
Full text at PDC
Publication Date
2019
Authors
Advisors (or tutors)
Tenllado Van der Reijden, Christian
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citations
Exportar
Abstract
El establecimiento de un sistema energético sostenible es uno de los grandes desafíos a los que debe enfrentarse la humanidad en el siglo XXI, y la energía solar presenta la principal fuente de energía renovable en la naturaleza. Este trabajo trata de predecir, a corto plazo, la radiación solar en determinadas áreas de la superficie, lo cual cobra vital importancia en la optimización de las plantas eléctricas y favorece la integración de la energía solar en los actuales sistemas de abastecimiento eléctrico.
Este estudio emplea la tecnología de redes neuronales fully-connected y Convolucionales, usando lenguaje de programación Python con librerías de Keras, Matplotlib, Numpy, Pandas y Pykriging, para la creación de dos modelos de Machine Learning que tratan de predecir la radiación solar en una ubicación concreta o varias simultáneamente en una determinada superficie terrestre donde se hayan desplegado sensores. Por otra parte se han creado modelos de Machine Learning para predecir la radiación solar, en un rango que puede llegar a varios minutos, en cualquier ubicación entre los sensores desplegados usado la técnica de kriging.
The establishment of a sustainable energy system is one of the greatest challenges that humanity must face in the 21st century, and solar energy presents the main source of renewable energy in nature. This work tries to forcast, on the short horizons, solar radiation in certain areas of the surface, which is of vital importance in the optimization of power plants and supports the integration of solar energy in power supply systems. This study uses the technology of Fully-Connected and Convolutional neural networks, employing Python programming language with libraries as Keras, Matplotlib, Numpy, Pandas and Pykriging, for the creation of two models of Machine Learning that try to predict irradiance in a specific location or several, simultaneously, in a certain area where sensors have been deployed. On the other hand, Machine Learning models have been created to forecast irradiance, in a range that can reach several minutes, at any location among the sensors deployed using the Kriging technique.
The establishment of a sustainable energy system is one of the greatest challenges that humanity must face in the 21st century, and solar energy presents the main source of renewable energy in nature. This work tries to forcast, on the short horizons, solar radiation in certain areas of the surface, which is of vital importance in the optimization of power plants and supports the integration of solar energy in power supply systems. This study uses the technology of Fully-Connected and Convolutional neural networks, employing Python programming language with libraries as Keras, Matplotlib, Numpy, Pandas and Pykriging, for the creation of two models of Machine Learning that try to predict irradiance in a specific location or several, simultaneously, in a certain area where sensors have been deployed. On the other hand, Machine Learning models have been created to forecast irradiance, in a range that can reach several minutes, at any location among the sensors deployed using the Kriging technique.
Description
Trabajo de Fin de Grado, Universidad Complutense, Facultad de Informática, Departamento de Arquitectura de Computadores y Automática, Curso 2018/2019