Model Applications to Refinery Production Planning

Refiners have typically solved this problem by using linear programming (LP) methods, which have been used extensively in refineries since 1950. Gary et al. [7] state that a site-wide model of the refinery is therefore usually required to in order to properly determine refinery economics . 

Linear programming involves the maximization of a linear objective function of many variables subject to linear constraints on each variable [57]. In the context of a refinery, the objective function can refer to the overall profit generated from processing a particular set of crudes. The variables that affect this objective function are typically the amounts of different crudes purchased. The goal is to determine an optimal set of crudes that maximize the profit margin of the refinery. This scenario is an example of crude oil evaluation. Refiners typically use LP methods 

We consider the above equation to represent the base yield of the unit. In Aspen PIMS and other similar LP software, the base yield is called the base vector. 

We typically encode the base vector in a form shown in Table 4.16. The negative signs arise from moving all the terms from the right-hand side of the equation to the left-hand side. 

This base vector is sufficient to model a FCC unit that processes a single type of feed at fixed operating conditions. However, most FCC units do not operate this fashion. They accept multiple feed with varying composition and may operate at different conditions. To account for variations in feed composition, the concept of the DELTA vector is useful. Every attribute (specific gravity concarbon, sulfur content, etc.) of the feed that can affect the yield of the unit has its own DELTA vector. The DELTA vector can be thought of a slope that modifies the base yield of each product. If we consider the specific gravity (SPG) of the feed as an attribute that can change the product yields, we can now rewrite the yield equation as  

---

Model Applications to Refinery Production Planning 307 Table 5.21 Reformer yields at various N+2A and C5+ reformate RON from rigorous model. 

In this section, we present two examples with different scenarios. The first example illustrates the performance of the model on a single site total refinery planning problem where we compare the results of the model to an industrial scale study from Favennec et al. (2001). This example serves to validate our model and to make any necessary adjustments. The second example extends the scale of the model application to cover three complex refineries in which we demonstrate the different aspects of the model. The refineries considered are of large industrial-scale refineries and actually mimic a general set-up of many areas around the world. The decisions in this example include the selection of crude blend combination, design of process integration network between the three refineries, and decisions on production units expansion options and operating levels. 

This chapter differentiates itself from the reported studies in the literature through the following contributions (1) detailed kinetic model that accounts for coke generation and catalyst deactivation (2) complete implementation of a recontactor and primary product fractionation (3) feed lumping from limited feed information (4) detailed procedure for kinetic model calibration (5) industrially relevant case studies that highlight the effects of changes in key process variables and (6) application of the model to refinery-wide production planning. 

---