Algoritmo diferencial evolutivo aplicado à identificação de sistemas dinâmicos não-lineares

• Ano de defesa: 2013
• Autor(a) principal: Morais, Josué Silva de
• 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: Engenharia elétrica; Sistemas não lineares; Identificação de sistemas; Métodos de simulação; Diferencial evolutivo; NARMAX expoente reais; Differential evolution; Identification systems; NARMAX real exponent; CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA
• Link de acesso: https://repositorio.ufu.br/handle/123456789/18418

The main objective of this work is to contribute in the structure’s selection stage in system identification process using artificial intelligence techniques, especially the Differential Evolutionary (DE) algorithm. To reach this goal a new NARMAX representation is proposed where exponents with real values are used. The evaluation of the proposed method was performed through an application developed specifically for this work with a QT package using C + + and parallel programming techniques. Different kinds of artificial systems were tested to evaluate the proposed methodology. Finally the methodology is applied to a dc-dc converter Boost -Buck and a biochemical polymer reactor and the results are presented.