Controle neuro-fuzzy para eficiência energética de sistemas de abastecimento de água com demanda variável

• Ano de defesa: 2020
• Autor(a) principal: Moreira, Hugo Augusto Marinho
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal da Paraíba; Brasil; Engenharia Mecânica; Programa de Pós-Graduação em Engenharia Mecânica; UFPB
• Palavras-chave em Português: Eficiência energética; Identificação de sistemas; Automação; Inteligência artificial; Sistemas hidráulicos; Energy efficiency; Systems identification; Automation; Artificial intelligence; Hydraulic system; CNPQ::ENGENHARIAS::ENGENHARIA MECANICA
• Link de acesso: https://repositorio.ufpb.br/jspui/handle/123456789/18509

This work aims to apply the Neuro-Fuzzy technique for pressure control of a water pumping system with variable demand. The control acts on the variation speed of the motor pump set (CMB) and its main objective is to raise the energy efficiency of the system. The controller were tested in an experimental system of water supply installed at the Energy and Hydraulic Efficiency Laboratory in Sanitation at the Federal University of Paraíba (LENHS / UFPB). The controller is analyzed regarding performance (permanent regime, transient and disturbances) and economy of energy. Therefore, the hourly water demand curve was implemented and indicators hydroenergetics (CMB yield and specific energy consumption coefficient) are used to measure energy efficiency gains. To validate your performance, the designed controller is compared to two other controllers, one being Neural (artificial neural networks) and a Fuzzy. As a way of analyzing and interpreting the operating conditions, non-linear multivariable computational modeling is also performed via artificial intelligence techniques. The developed models contemplate how output variables the yield of the CMB, the pressure and the flow of the system. The results show a significant energy gain from the pumping system, indicated by hydroenergetic indicators. The pressures were kept close to the set-points, with low rise time (ts <7 seconds) and without overshoot; in addition to the excellent stability (error <8%) in severe situations of demand variation. It is concluded that the Neuro-Fuzzy technique guaranteed superior results to the other controllers analyzed.