Separação cega de fontes aplicada ao compartilhamento de responsabilidades sobre as distorções harmônicas

• Ano de defesa: 2022
• Autor(a) principal: Pereira, Fabiana Alves
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Uberlândia; Brasil; Programa de Pós-graduação em Engenharia Elétrica
• Palavras-chave em Português: análise de componentes Independentes; contribuições harmônicas; método não-invasivo; qualidade de energia; separação cega de fontes; Blind source separation; independent component analysis; harmonic contributions; non-invasive methods; power quality; CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA; Engenharia elétrica; Sistemas de energia elétrica; Energia elétrica - Transmissão - Corrente contínua; Abastecimento de energia
• Link de acesso: https://repositorio.ufu.br/handle/123456789/36709; http://doi.org/10.14393/ufu.te.2022.5047

One of the greatest challenges in power quality has been, and still is, the establishment of a reliable non-invasive method for qualifying and quantifying the harmonic contributions of an electric utility and consumer at the Point of Common Coupling (PCC). This difficulty arises from the fact that the harmonic distortions measured at the PCC are unidentifiable mixtures from both parties. In this way, it becomes evident the correlation between sharing harmonic responsibility and the classical Blind Source Separation problem (BSS), which seeks to separate the signals that composed an observed mixture, without, however, having previous information about the original ones and the way they were mixed. In this context, this work aims to develop and evaluate a non-invasive methodology, based on BSS methods, capable to effectively promote the sharing of harmonic responsibilities between electric utility and consumer at the PCC. More specifically, due to the nature of electrical loads, the methodology presented here will be based on BSS methods based on Independent Component Analysis (ICA) and will be evaluated both computationally and experimentally for different operating conditions and then, the results will be compared with other methods proposed in the literature.