Análise de estabilidade transitória de sistemas elétricos por redes neurais ARTMAP nebulosas modulares
Carregando...
Data
2003-10-10
Autores
Silveira, Maria do Carmo Gomes da [UNESP]
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
Esta pesquisa apresenta uma metodologia para a análise de estabilidade transitória (de primeira oscilação) de sistema de energia elétrica usando uma rede neural baseada na arquitetura ART (Adaptive Resonance Theory), designada rede neural nebulosa ARTMAP modular para aplicações em tempo real. A margem de segurança é empregada como critério da análise de estabilidade transitória, considerando-se faltas tipo curto-circuito trifásico com saída de linha de operação. O funcionamento das redes neurais é constituído por duas fases fundamentais: treinamento e análise. A fase de treinamento requer uma grande quantidade de processamento para a sua realização, enquanto que a fase de análise é efetivada, praticamente, sem esforço computacional. Esta é, portanto, a principal justificativa para o uso de redes neurais para a resolução de problemas complexos que exigem soluções rápidas, como é caso de aplicações em tempo real. As redes neurais ART, possuem como características primordiais, a plasticidade e a estabilidade, as quais são qualidades essenciais para a execução do treinamento e para a análise de modo eficiente. A rede neural ARTMAP nebulosa modular está sendo proposta visando proporcionar um desempenho superior, em termos de precisão e rapidez, se comparada à formulação ARTMAP convencional, muito maior ainda quando comparada ao emprego de redes neurais com treinamento realizado via técnica retropropagação (que é um benchmark em termos de precisão no contexto de redes neurais). O treinamento da rede neural será realizado usando como “professor” um método de energia para o cálculo das margens de segurança. A metodologia proposta nesta pesquisa é um procedimento que incorpora inovações em relação à literatura existente, em destaque: (1) desenvolvimento de um novo modelo apropriado para a resolução... .
This work presents a methodology for transient stability analysis (first swing) of electrical energy systems, using a neural network based on ART (adaptive resonance theory), called modular fuzzy ARTMAP neural network for applications in real time. The security margin is used as a criterion for transient stability analysis, considering three-phase short circuit with outage of a transmission line faults. The neural network functioning is composed of two fundamental phases: training and analysis. The training phase needs a great quantity of processing, while the test phase is realized almost without computational effort. Therefore, this is the principal reason for using neural networks to solve complex problems that demand fast solutions, as the real time applications. The ART neural networks have as prime characteristics the plasticity and the stability, which are essential qualities for the training execution and for analysis in an efficient way. The modular fuzzy ARTMAP neural network is proposed to provide a superior performance, in precision and velocity, when compared to the conventional ARTMAP formulation, and much faster when compared to the neural networks using backpropagation training (benchmark in precision in a neural network context). The neural network training is realized using as a “teacher” an energy method to calculate the security margin. The proposed methodology is a proceeding that incorporates innovations in relation to the existent literature emphasizing: (1) development of a new appropriated model for solving the transient stability analysis by neural networks; (2) theoretical-practical development of the neural network output variable, represented in this work by the IMS (security margin interval) variable in a binary way. It is a convenient representation for the transient stability analysis approach; (3) proposition of a... (Complete abstract click electronic address below).
This work presents a methodology for transient stability analysis (first swing) of electrical energy systems, using a neural network based on ART (adaptive resonance theory), called modular fuzzy ARTMAP neural network for applications in real time. The security margin is used as a criterion for transient stability analysis, considering three-phase short circuit with outage of a transmission line faults. The neural network functioning is composed of two fundamental phases: training and analysis. The training phase needs a great quantity of processing, while the test phase is realized almost without computational effort. Therefore, this is the principal reason for using neural networks to solve complex problems that demand fast solutions, as the real time applications. The ART neural networks have as prime characteristics the plasticity and the stability, which are essential qualities for the training execution and for analysis in an efficient way. The modular fuzzy ARTMAP neural network is proposed to provide a superior performance, in precision and velocity, when compared to the conventional ARTMAP formulation, and much faster when compared to the neural networks using backpropagation training (benchmark in precision in a neural network context). The neural network training is realized using as a “teacher” an energy method to calculate the security margin. The proposed methodology is a proceeding that incorporates innovations in relation to the existent literature emphasizing: (1) development of a new appropriated model for solving the transient stability analysis by neural networks; (2) theoretical-practical development of the neural network output variable, represented in this work by the IMS (security margin interval) variable in a binary way. It is a convenient representation for the transient stability analysis approach; (3) proposition of a... (Complete abstract click electronic address below).
Descrição
Palavras-chave
Redes neurais (Computação), Lógica difusa, Sistemas de energia eletrica - Estabilidade, Teoria da ressonância adaptiva, Redes neurais ARTMAP nebulosas
Como citar
SILVEIRA, Maria do Carmo Gomes da. Análise de estabilidade transitória de sistemas elétricos por redes neurais ARTMAP nebulosas modulares. 2003. v, 92 p. Tese (doutorado) - Universidade Estadual Paulista, Faculdade de Engenharia de Ilha Solteira, 2003.