Optimisation: one word, so many interpretations. Dynamic or steady state? Non-linear or linear? Constrained or unconstrained? Open or closed loop? Since our focus is on refinery operations, the most relevant and practical assumptions seem to be closed loop, constrained, dynamic and linear.
For more than three decades, advanced process control (APC) and multivariable control (MVC) have been applied within refineries and petrochemical plants. The relatively narrow scope has limited the gains, and a great deal of money is being left on the table.
Recent advances in processing power now enable planners to run 20 to 30 scenarios almost simultaneously to aid in their refinery optimisation. As a result, refineries have the ability to pivot to an optimal product mix on a daily basis.
Accenture estimates refineries could widen margin/barrel of crude by roughly a dollar, representing gains of 8 to 12 %, by implementing broader optimisation. In the case of a 300,000 barrels/day refinery, making an extra 50 cents to a dollar/barrel more would mean US$ 150,000 to $300,000 more per day.1
Focus on 4 critical areas
A number of constraints and variables, described in four areas below, makes optimisation a complex yet worthwhile endeavour.
Molecules are transformed and can therefore change the product stream in which they end. Optimal planning is hard to achieve because of the many sensitivities (or degrees of freedom). In addition, operational issues (e.g., fouling on heat exchangers, ambient air-temperature changes, equipment maintenance) also affect refinery capability and typically are not included in planning. Planning has to include predefined commitments and be built around the crude assay. Achieving the feasible plan is possible but invariably, within hours, the published plan is dated.
Operators typically are held accountable for the performance of their units. With the application of APC, they have become more aligned to a pilot on a long-haul plane, with the autopilot (in this case APC) performing minute-by-minute changes. The overall direction and local profitability of the unit, however, remains within their ownership. As the refinery plan is created in the office, this typically cascades its way to the control room. In the event that the optimal operation is determined by a computer and then directly communicated to the APC on the process unit, there is frequently pushback from the operator due to lack of involvement or understanding of the targets being set.
As the plan is created, it is based on both hydrocarbon lift commitments and costs/prices (estimated or known) of raw materials/products. Shadow prices—the effective value or cost of intermediate streams—are also developed. Ultimately, the aim is to deliver maximum value to the organisation. There comes a point, however, where the same value can be made through different scenarios, some of which have hidden costs. For example, increasing the severity on the fluidised catalytic cracker (FCC) typically involves a higher catalyst circulation rate that causes additional catalyst attrition and subsequent loss as fine particles, which is expensive for FCC zeolite-type catalysts. Inclusion of this “local” knowledge is something that would increase the robustness of any plan but has been very difficult to achieve. The more complex the plan and more degrees of freedom it has possessed, the longer it would take to solve. Historically, it has taken days to reach a solution.
The creation of an optimal plan requires inclusion of operational boundaries and limits, coupled with feasible dynamics. There are times when the plan developed, in theory, is optimal but achievement is virtually impossible. This is where digital meets physical, and the inclusion of reaction kinetics/process dynamics in the plan is not something that typically happens. The plan is a steady-state solution whilst the process is dynamic, and the trajectory of how to get to the optimum is not defined.
Considering all of these variables, is it possible to achieve refinery wide optimisation? The answer is yes, but with a caveat: It will not be perfect but it will be close enough. Given the magnitude of potential gains, leading refiners of the future are likely to be those rigorously pursuing wide-scope optimisation.
1 Based on Accenture experience and benefits documented from higher-level optimisation projects in limited areas of operations.