(373at) A Cluster-Based Strategy to Integrate the Operational Management of Crude Oil Supply
Ideally, each crude oil arriving at the terminal would have a dedicated storage tank to be unloaded. If this is the case, the crude oils would only be mixed at the charging or feeding tanks in order to produce blends (of crude oils) feasible at the specifications required by the processing in the crude oil distillation units (CDUs). Nevertheless, the number of crude oils available in the market are much higher than the number of storage tanks in a terminal. Also, during the unload of crude oil into the storage tanks, there are no specific operational restrictions to rule out where a certain crude oil should be stored. This means that vessels may unload in all tanks and consequently all types of crude oils can be mixed in the storage tanks. In this case, there are two main consequences: (1) as shown in the work of  this leads to large non-convex MINLP problems since unload activities are not constrained; and (2) crude oils with different properties (intended to be segregated) can be mixed (i.e., crude oils with low and high content of sulfur) .
The scheduling problem in this work is formulated as a non-convex MINLP. To tackle this problem, we propose an MILP clustering formulation to be solved before the main non-convex MINLP scheduling problem. The clustering problem considers the properties of crude oils, storage capacity of resources, flow rate between resources, and has the main objective to minimize the standard deviations from the mean properties of a set of crude oils assigned to a cluster of storage tanks.
The result of solving the optimization problem has two consequences for the main non-convex MINLP. The first one is to restrict the trips and unload possibilities from vessels to storage tanks (i.e., restrict binary variables,). Next, by knowing the set of crude oil raw materials assigned to each cluster of storage tanks, we can derive lower and upper bounds on the properties of the mixed crude oils in each storage tank, making it possible to linearize the non-convex bilinear terms associated to blending . This linearization is used in an MILP-NLP decomposition strategy to tackle the original non-convex MINLP formulation.
 de Assis, L.S. , Camponogara, E. , 2016. A MILP model for planning the trips of dynamic positioned tankers with variable travel time. Transp. Res. Part E 93, 372â388 .
 Zimberg, B., Camponogara, E., Ferreira, E., 2015. Reception, mixture, and transfer in a crude oil terminal. Comput. Chem. Eng. 82, 293â302.
 de Assis, L.S. , Camponogara, E. , Menezes, B.C. , Grossmann, I.E. , 2019. An MINLP formulation for integrating the operational management of crude oil supply. Comput. Chem. Eng. 123, 110â125.
 Kelly, J.D., Menezes, B.C., Grossmann, I.E., Engineer, F., 2017. Feedstock storage assignment in process industry quality problems. Foundations of Computer Aided Process Operations: Tucson, United States.
 MÃ©ndez, C.A., Grossmann, I.E., Harjunkoski, I., KaborÃ©, P., 2006. A simultaneous optimization approach for off-line blending and scheduling of oil-refinery operations. Comput. Chem. Eng. 30, 614 â 634.