Short-term crude mix and operating conditions optimization for oil refinery operations

Most of the time crude oil refineries process variety of feedstocks during a course of a week. Different crude oils are either available at the site or are scheduled to be delivered within a couple of days. Depending on the refinery configuration and the capacities of the process units, decisions how to process these feedstocks include determination of the crude mix to process each day, operating conditions, and blend recipes. Current industrial practice is to use multi-mode linear process unit models or to optimize operating conditions via real-time process unit optimization. This work optimizes the entire refinery in order to utilize in the best possible manner capabilities of the process units and maximize the profit from the feedstocks at hand. A global commercial solver and an implementation of the normalized multi-parametric discretization algorithm have been used to solve this mixed-integer nonlinear programming model.