Computational Model for Discriminating Defective Photovoltaic Cells in Electroluminescent Images.

EasyChair Preprint 11289

Abstract

The generation of photovoltaic energy has been growing worldwide due to being clean and cost-effective. The most vulnerable component of photovoltaic generation systems is the photovoltaic cell (PV), which operates exposed to adverse environmental conditions such as wind, rain, salinity, and dust. Therefore, there is a demand for technologies that contribute to improving the efficiency and reliability of photovoltaic generation systems through the automation of inspection processes and detection of defects in PV cells. Thus, we propose a model to discriminate defective monocrystalline silicon PV cells in electroluminescent images using texture attributes and a custom convolutional neural network. We evaluated the performance of this model through cross-validation with images from a widely used database, obtaining competitive results compared to those found in the literature.

Keyphrases: Convolutional Neural Network, Local Binary Pattern, Photovoltaic cell, fault detection

@booklet{EasyChair:11289,
  author    = {Alan M. da Rocha and Marcelo M. S. de Souza and Francilândio L. Serafim and Carlos A. R. Fernandes and Ícaro B. Viana},
  title     = {Computational Model for Discriminating Defective Photovoltaic Cells in Electroluminescent Images.},
  howpublished = {EasyChair Preprint 11289},
  year      = {EasyChair, 2023}}