Método para detecção de falta de arco elétrico serie no lado CC de sistemas fotovoltaicos

• Ano de defesa: 2024
• Autor(a) principal: Pozzobon, Eugênio Piveta
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Santa Maria; Brasil; Engenharia Elétrica; UFSM; Programa de Pós-Graduação em Engenharia Elétrica; Centro de Tecnologia
• Palavras-chave em Português: Arco elétrico série; Sistemas fotovoltaicos; Transformadas wavelet; Máquina de vetores de suporte; F28379D; Series arc fault; Photovoltaic systems; Wavelet transforms; Support vector machine; CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA
• Link de acesso: http://repositorio.ufsm.br/handle/1/31999

With the growth of photovoltaic (PV) installations, there is an associated increase in the risk of fire due to faults on the DC side. In this context, this work proposes and implements an algorithm based on Wavelet Packet Decomposition (WPD) and Support Vector Machine (SVM) to detect series arc fault on the DC side, sampling data from the alternating current component of a string. Consequently, a fault database was established for the SVM training using supervised learning. This database included data obtained through laboratory tests standardized by IEC 63027 and UL 1699B, as well as tests conducted in PV plants. From this database, WPD was utilized to compute features of signals both with and without faults. Subsequently, the SVM algorithm was trained to classify and identify series arc faults. The algorithm was then implemented in an embedded system using the Texas Instruments F28379D microcontroller. The results demonstrate the very good performance of the proposed algorithm, indicating that the implementation is feasible on low-cost microcontrollers.