Detecção, classificação e localização de faltas de curto-circuito em sistemas de distribuição de energia elétrica usando sistemas inteligentes
Carregando...
Data
2012-09-28
Autores
Decanini, José Guilherme Magalini Santos [UNESP]
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
Neste trabalho propõe-se uma metodologia alternativa para o diagnóstico automático de faltas de curto-circuito em sistemas de distribuição de energia elétrica. Esta ferramenta, de auxílio à tomada de decisão, acelera os procedimentos para restabelecimento das condições normais de operação propiciando maior segurança, confiabilidade e lucratividade às concessionárias. O sistema de diagnóstico foi concebido integrando modernas técnicas de processamento de sinais e sistemas inteligentes. O processo contínuo de detecção de distúrbio é realizado através de análises estatística e direta multinível dos sinais de corrente no domínio wavelet. Na etapa de classificação de faltas de curto-circuito as principais características dos sinais de corrente e tensão (assinatura) são extraídas empregando a análise multirresolução e o conceito de energia. Estes índices comportamentais correspondem aos vetores de entrada de três conjuntos independentes de redes neurais artificiais da família ART (Adaptive Resonance Theory), ARTMAP Fuzzy, cujo treinamento dar-se-á de forma supervisionada. Esta arquitetura de rede executa o treinamento com alto desempenho computacional e apresenta duas características fundamentais para aplicação em modernos sistemas de energia elétrica, estabilidade e plasticidade, permitindo a inclusão do treinamento continuado, o que representa uma importante vantagem quando comparada às recentes metodologias. Os estímulos de entrada do primeiro conjunto de redes correspondem às características extraídas dos sinais de corrente, o segundo conjunto analisa as características relativas aos sinais de tensão e o terceiro conjunto tem como vetor de entrada os índices comportamentais referentes aos sinais de corrente e tensão. A teoria da evidência de Dempster-Shafer é empregada...
The present project proposes an alternative methodology for the automatic diagnosis of short- circuit fault in distribution systems. This support tool for decision making accelerates the restoration process providing greater security, reliability and profitability to electric power companies. The diagnostic system was designed through the combined use of modern signal processing techniques and intelligent systems. The continuous procedure for disturbance detection is performed based on direct and statistical multilevel analysis of the current signals in wavelet domain. In the classification step, the main features of the current and voltage signals (signature) are extracted using multi-resolution analysis and energy concept. These behavioral indexes correspond to the input vectors of three parallel sets of ART family neural network, Fuzzy ARTMAP, which is trained in a supervised manner. This architecture performs the training with high computing performance and presents two key characteristics for application in modern electric power systems, stability and plasticity, enabling the inclusion of continuous training, which represents an important advantage when compared with recent methodologies. The first and second network set evaluates the features extracted from current and voltage signals, respectively. The third set performs the diagnosis based on both behavioral indexes. The Dempster-Shafer theory is employed to integrate the evidence (outcomes) of the network sets, providing quantitative information about the diagnosis and its reliability. Finally, the faulted section is located using, a priori, voltage and current data measured at the substation. To demonstrate the performance of this method, an application is submitted considering a practical distribution feeder from a Brazilian electric utility, composed of... (Complete abstract click electronic access below)
The present project proposes an alternative methodology for the automatic diagnosis of short- circuit fault in distribution systems. This support tool for decision making accelerates the restoration process providing greater security, reliability and profitability to electric power companies. The diagnostic system was designed through the combined use of modern signal processing techniques and intelligent systems. The continuous procedure for disturbance detection is performed based on direct and statistical multilevel analysis of the current signals in wavelet domain. In the classification step, the main features of the current and voltage signals (signature) are extracted using multi-resolution analysis and energy concept. These behavioral indexes correspond to the input vectors of three parallel sets of ART family neural network, Fuzzy ARTMAP, which is trained in a supervised manner. This architecture performs the training with high computing performance and presents two key characteristics for application in modern electric power systems, stability and plasticity, enabling the inclusion of continuous training, which represents an important advantage when compared with recent methodologies. The first and second network set evaluates the features extracted from current and voltage signals, respectively. The third set performs the diagnosis based on both behavioral indexes. The Dempster-Shafer theory is employed to integrate the evidence (outcomes) of the network sets, providing quantitative information about the diagnosis and its reliability. Finally, the faulted section is located using, a priori, voltage and current data measured at the substation. To demonstrate the performance of this method, an application is submitted considering a practical distribution feeder from a Brazilian electric utility, composed of... (Complete abstract click electronic access below)
Descrição
Palavras-chave
Sistemas de energia eletrica, Sistemas inteligente de controle, Energia elétrica - Distribuição, Eletric power systems
Como citar
DECANINI, José Guilherme Magalini Santos. Detecção, classificação e localização de faltas de curto-circuito em sistemas de distribuição de energia elétrica usando sistemas inteligentes. 2012. 124 f. Tese (doutorado) - Universidade Estadual Paulista Julio de Mesquita Filho, Faculdade de Engenharia de Ilha Solteira, 2012.