Otimização da programação de petróleo em uma refinaria com regras específicas

• Autor(a) principal: Jahn, Bárbara
• Data de Publicação: 2015
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional da UFU
• Texto Completo: https://repositorio.ufu.br/handle/123456789/22710; http://dx.doi.org/10.14393/ufu.di.2018.1195
• topic: Programação da produção; Programação do abastecimento de petróleo; Programação misto-inteiro linear; Batch scheduling; Crude oil scheduling; Mixed-integer linear programming; Engenharia química; Petróleo - Refinaria - Controle de produção; Planejamento da produção; CNPQ::ENGENHARIAS::ENGENHARIA QUIMICA

Process-oriented oil and chemical companies, like most industries, are becoming increasingly dependent on optimization tools. The main objective of this work is to investigate the viability of using optimization techniques applied to crude supply scheduling models in a real oil refining plant with specific operating rules of the brazilian industry. The system representative operating rule is that up to two charging tanks can feed one crude distillation unit simultaneously. One of the most promising continuous-time formulations in the literature, the multi-operation sequencing (MOS) formulation from Mouret et al. (2011), was selected to model the oil refining plant production system, a multi-stage system consisting in up to five crude parcels, nine charging tanks, three crude distillation units (CDU) and thirty six types of crude oil. The optimization problem objective function was to maximize the distilled crude mixtures gross profit margin. A mixed-integer non-linear programming (MINLP) problem was then obtained, which was solved with MILP-NLP decomposition strategy by algorithms implemented in comercial solvers. Two models were developed with the MOS formulation: one model with non-overlapping unloading operations and the other model with overlapping unloading operations. Both models were implemented in the General Algebraic Modelling System (GAMS) and applied to six different scenarios in the oil refining system. The scenarios time horizon varied from 6 to 10 days. The models represented the oil refining system and its operating rules successfully, as well as were able to find feasible solutions to the crude oil scheduling problem (COSP) in feasible computational times. The model with non-overlapping unloading operations was able to solve the optimized scenarios in up to 28.32 minutes, and the model with overlapping unloading operations in up to 13.94 minutes.