(22f) Optimization of A Complex System of Crude Oil Distillation Units Using NLP and Metamodels

The objective of this paper is to show the optimisation of a complex system of Crude Oil Distillation Units (CDU) belonging to a Colombian refinery of ECOPETROL S.A. This system is composed by crude oil blending scheme and all distillation units installed in the refinery. Distillation units are made up by 5 atmospheric towers, 5 atmospheric furnaces and 4 vacuum towers; crude oil blending is integrated for 14 Colombian crude oils, which were, at first, mixed in 4 mixtures (pre-mixtures) which, at second, were used for preparing the crude oil diets.

Optimisation results were optimum feedstock composition and operating conditions for the CDU system and was carried out through Non Linear Programming (NLP) Model developed by the research institute belonging to ECOPETROL S.A. NLP model optimized the system economic profit and uses Metamodels approach to represent Atmospheric Distillation Towers (ADTs) and Assays cutting with Operating Modes to model the Vacuum Distillation Towers (VDTs). This Model has been used for short, middle and long term planning.  

In NLP model, crude oil blending restrictions were mass balances, mixing rules, property boundaries like acidity, sulfur content, specific gravity. Atmospheric distillation process was modeled through Metamodels calculating volumetric flows, properties and temperatures for each ADT product and pumparound; product properties were API gravity, sulfur, neutralization number, conradson carbon, nickel, vanadium, total nitrogen. Metamodels were based on rigorous simulations of distillation units in PRO/II.

Optimization model was formulated in GAMS 23.3, was solved with CONOPT and had a user interface developed in Microsoft Office Excel 2007. NLP model found new crude oil diets and process variable points for each CDU with better economic results than those obtained with the feed compositions of normal short term planning and operation in the refinery. NLP model was able to evaluate new crude oil blending schemes such as segregation of crude oil streams in the current pre-mixtures and inclusion of new pre-mixtures for the refinery; for these new cases, the model calculated the best performance of distillation system.