Identificação de funcionamento atípico de painel fotovoltaico

• Ano de defesa: 2020
• Autor(a) principal: Silva, André
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal do Espírito Santo; BR; Mestrado em Energia; Centro Universitário Norte do Espírito Santo; UFES; Programa de Pós-Graduação em Energia
• Palavras-chave em Português: Monitoramento não intrusivo; detecção de falhas; localização de falhas; energia solar fotovoltaica; redes neurais convolucionais; subject.br-rjbn; Engenharia/Tecnologia/Gestão
• Link de acesso: http://repositorio.ufes.br/handle/10/15014

Solar energy is a cheap and clean source, being an excellent option to replace fossil fuels, reducing the generation of greenhouse gases. To achieve this goal, the cost of energy must be competitive with other sources. The solar modules have a long service life, but maintenance actions are necessary to reach this time. Due to exposure to adverse environmental conditions, equipment is subject to failures, which are usually difficult to detect and locate. This dissertation aims to identify atypical functioning in a photovoltaic string using only the voltage and current at its terminals, without using environmental data of solar irradiation and temperature. The methodology used in the collection of voltage and current data from the string in normal operating conditions and four types of failure conditions: shading of an entire panel, shading of a panel sector, and short-circuit of a whole module, and electric arc. The panel shading, sector shading, and short-circuit failures have subdivided into 6, one for each panel generating 20 different operating conditions, to identify the abnormal module. For each of the 20 states, data have collected under diverse climatic conditions. The collected data has used to train, validate, and test a convolutional neural network, which can learn the characteristics that allow the identification of each of the 20 operating conditions with an accuracy of 94,03%.