Otimização do balanço termelétrico de uma refinaria de petróleo: MILP x MINLP

• Ano de defesa: 2017
• Autor(a) principal: Alves, Rafael Victor Ferreira
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal do Rio de Janeiro; Brasil; Instituto Alberto Luiz Coimbra de Pós-Graduação e Pesquisa de Engenharia; Programa de Pós-Graduação em Engenharia Química; UFRJ
• Palavras-chave em Português: Engenharia química; Programação matemática; Balanço termelétrico de refinaria; CNPQ::CIENCIAS EXATAS E DA TERRA::QUIMICA
• Link de acesso: http://hdl.handle.net/11422/9760

This work presents two algebraic models to represent the thermoelectric system of an oil refinery, which are optimized by mathematical programming techniques. The first model is mixed integer linear programming (MILP), while the second is mixed integer nonlinear programming (MINLP), in a way that the main nonlinearities are the variations in boiler efficiency with regrets to load and steam headers temperatures. The nonlinear models were optimized using different solvers (DICOPT, BONMIN, COUENNE and SCIP) and two initial points. The results of the optimizations of the models in different operating and cost scenarios allowed to verify that the MINLP model, optimized by the DICOPT and using the MILP model results as the starting point, was a solution with better results for optimization of the thermoelectric balance, reaching the lowest operating cost.