Restrições de estabilidade no controlador preditivo robusto baseado em modelo

• Ano de defesa: 2015
• Autor(a) principal: Amanda Goncalves Saraiva Ottoni
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; UFMG
• Palavras-chave em Português: Modelo linear; Incerteza de modelo; Controle preditivo; Função de transferência; Branch and Bound; Controle robusto; Incerteza politópica; Estabilidade; Engenharia elétrica; Incerteza; Transferência de funções
• Link de acesso: http://hdl.handle.net/1843/RAOA-BC9GUF

Model Predictive Control (MPC) is a modern control technique that has several variations and many applications in industry. A new robust MPC strategy - RMPC-SC (Robust MPC with Stability Constraints) - is proposed in this work. A set of constraints to the increment of the control variables, for which guarantees the stability of a monovariable discrete time linear dynamical system, described by a known transfer function, but with polytopic uncertainty in the parameters, is explicitly calculated. The condition developed is enough to stabilize the system at every instance of the transfer function within the admitted uncertainty range and can be used in conjunction with any robust MPC technique that admits inclusion of restrictions on the control variable. A new optimization method Branch Bound is proposed in order to improve the optimization procedure inherent to the RMPC algorithm, developed in this work. This method can be used in the global optimization of functions that can be described as difference between increasing functions, constrained to a box.