-
Rodrigo P. Ramos
Colegiado de Engenharia Elétrica - CENEL, Universidade Federal do Vale do São Francisco - Univasf, PE
-
Ricardo M. Prates
Colegiado de Engenharia Elétrica - CENEL, Universidade Federal do Vale do São Francisco - Univasf, PE
-
Jadsonlee da S. Sá
Colegiado de Engenharia da Computação, Universidade Federal do Vale do São Francisco - Univasf, PE
-
Adeon C. Pinto
Colegiado de Engenharia Elétrica - CENEL, Universidade Federal do Vale do São Francisco - Univasf, PE
-
Eubis P. Machado
Colegiado de Engenharia Elétrica - CENEL, Universidade Federal do Vale do São Francisco - Univasf, PE
-
Wêdson P. da Silva
Colegiado de Engenharia Elétrica - CENEL, Universidade Federal do Vale do São Francisco - Univasf, PE
-
José Bione M. Filho
Companhia Hidro Elétrica do São Francisco - CHESF, PE
-
Alcides Codeceira Neto
Companhia Hidro Elétrica do São Francisco - CHESF, PE
-
Alex C. Pereira
Companhia Hidro Elétrica do São Francisco - CHESF, PE
-
Eduardo B. Jatobá
Companhia Hidro Elétrica do São Francisco - CHESF, PE
Keywords:
Photovoltaic generation, Soiling, Artificial intelligence, Deep learning, Image processing
Abstract
Inspection in photovoltaic generation systems is of utmost importance for maintenance and corresponding generation efficiency. The combined use of unmanned aerial vehicles and deep learning computational algorithms can be considered a potentially strong technique for the automatic inspection of non-conformities present in photovoltaic panels. In this work, a methodology was developed for inspecting a ground-based solar power plant by capturing video collected by drones and analyzing and classifying the resulting images using convolutional neural network models. Subimages cropped from the extracted images were processed and classified into one of four classes: intact panel and with mild, medium and severe non-conformities. Accuracies of up to 96.42% in the classification of non-conformities were obtained.
