Sistema de clasificación de fallas en tiempo real de paneles fotovoltaicos mediante uso de imágenes RGB y aprendizaje profundo

This study presents a proposal for a classification system for faults that occur in photovoltaic modules from the use of RGB images and a deep learning approach implemented in a portable embedded device (Raspberry Pi). Convolutional Neural Networks (CNN) are used for photovoltaic module detection us...

Full description

Autores:
Flórez Sierra, Andrés Felipe
Tipo de recurso:
Trabajo de grado de pregrado
Fecha de publicación:
2020
Institución:
Universidad de los Andes
Repositorio:
Séneca: repositorio Uniandes
Idioma:
spa
OAI Identifier:
oai:repositorio.uniandes.edu.co:1992/51548
Acceso en línea:
http://hdl.handle.net/1992/51548
Palabra clave:
Paneles solares fotovoltáicos
Redes neuronales convolucionales
Fallas de sistemas (Ingeniería)
Energía solar
Ingeniería
Rights
openAccess
License
https://repositorio.uniandes.edu.co/static/pdf/aceptacion_uso_es.pdf
Description
Summary:This study presents a proposal for a classification system for faults that occur in photovoltaic modules from the use of RGB images and a deep learning approach implemented in a portable embedded device (Raspberry Pi). Convolutional Neural Networks (CNN) are used for photovoltaic module detection using semantic segmentation and additionally for fault classification. This study considers 2 specific cases of classification; Binary (Fault and not Fault) and quaternary (Cracks, Shadows, Dust and Not Fault). The models presented show an average accuracy greater than 90% for the binary classification and 74% for the quaternary classification. The implementation of these models in low cost embedded devices corresponds to a viable alternative for the continuous monitoring of photovoltaic modules.