(755b) Simultaneous Scheduling of Crude-Oil Unloading, Transferring, and Blending Based On Continuous Time Formulation: Generic Model and Extensions
AIChE Annual Meeting
Thursday, November 7, 2013 - 3:35pm to 3:55pm
Among all the refinery operations, scheduling of crude oil activities is of utmost importance. Crude oils vary a lot from each other in terms of sulfur contents, light oil yields and properties. At the planning level, a decision maker will refer to the product demand forecast and updated market information to formulate a monthly production plan. According to this plan, operational scheduling is conducted to determined crude unloading, transfer and mixing schedules. For crude unloading, it determines the allocation of limited docking stations to crude vessels and connection between the vessel and tank; crude transfers refer to the material transfers between tanks; crude mixing prepares blends at appropriate sulfur contents and property indexes. During this process, if any vessels demurrage on the sea shore, extra costs will be incurred. And also, each set-up from a source unit to a destination unit calls for special care of configuring pipeline and maintaining pump, which is highly costly. Moreover, tanks introduce inventory costs due to holding a volume of crudes during the scheduling horizon. Generally, refineries seek to blend the premium crudes which have less undesirable components such as sulfur, aromatics and residue, with low-quality crudes, in the sense of exploiting the value of cheaper crudes.
In recent years, scheduling of crude oil operations receive a lot of attentions, which stimulates explorations on different mathematical models based on discrete and continuous time formulations. However, literature review indicates there still lacks systematic considerations of pipeline hold-ups which will introduce time delay for crude transfers between different units and thus, impact the downstream unit operations. In addition, as Achille's Heel in general pooling problem, the bilinear terms in blending constrains bring in non-convexity and thus, lead to non-optimal solutions. Correspondingly, our study utilizes the most recent relaxation techniques to construct convex envelops for bilinear terms, targeting global optimality of solutions.
The scope of this paper includes single-parcel vessels that carry different crudes, together with storage and charging tanks which are used to unload and mix the crudes, respectively, and also distillation units which receive the blended crudes and carry out further processing. The given information that serves as inputs to the scheduling problem can be summarized as such. 1) The number of vessels and their respective arrival dates, the type of crudes carried. 2) The number of tanks, and their initial crude inventories. 3) The feed specifications and crude demand required by distillation columns. 4) The crude pipeline hold-up and its initial crude volumes. 5) The flow rate limits between different units. 5) Economic data including demurrage costs, crude unloading costs, tank inventory costs and column changeover costs. It should be noticed that crude operations is a multi-objective problem inherently for which the top priority is to ensure the smooth and continuous operation of distillation columns throughout the scheduling horizon. The scheduling model in our study is based on continuous time formulation and incorporates a lot of real-life characteristics such as unloading sequence of vessels based on their relative arrival times, brine settling time, pipeline hold-up and multiple feeds to distillation columns and etc. The generated schedule determines a large body of information including crudes transfer sequences and routes, how much volume each transfer is, and what crude compositions will be in the transferring-involved tanks and streams, and what is the crudes distribution inside a pipeline, and etc. This model minimizes the usage of bilinear terms through big-M formulations while modeling the logic conditions. As the first attempt to rigorously model pipeline hold-up combined with many other real life features, our model is proven to be robust and efficient enough to help arrange the short term crude oil operations. And the solution quality is also improved by using the piecewise under- and over-estimators.