(300a) Refinery-Wide Scheduling for Optimization of Multiple Unit-Operations in the Supply, Production, and Demand Chains in Fuels, Lubes, Asphalts and Petrochemicals Industries

We
present a refinery-wide scheduling optimization problem found in every process
industry with mixing, transforming and separation where the transforming and
separation activities in the crude-oil refinery industry are referred to as reacting/converting
processes and distilling/fractionating processes, respectively. For this study
we highlight only the quantity and logic aspects or phenomena of the problem
(what we term logistics) using
mixed-integer linear programming (MILP). The quantity and quality details of
the problem (what we term quality)
using nonlinear programming (NLP) are not described. The NPL attributes such as
densities, components, properties and conditions require special processing
subtypes such as blenders, splitters, separators, reactors, fractionators and
black-boxes [1].

In
the proposed refinery-wide scheduling approach, the goal is to investigate the limits
or capabilities of an enterprise-wide optimization (EWO) by solving simultaneously
the supply, production, and demand chains using MILP. This finds time-varying setups
of unit-operations and their connections that construct the flowsheet operations
to further NLPs by fixing the MILP results. Such EWO setups maybe used in detailed
calculations within other edges such as depooling of
aggregated tanks by considering their actual topology and operations [2] or decision-automation for production of lubes and
asphalts using sequence-dependent switchovers [3].

Figure
1 shows the oil-refinery flowsheet using the unit-operation-port-state
superstructure (UOPSS) [4], [5] in a discrete-time modeling. The rectangular
shapes (⊠) with the cross-hairs are continuous-process types of unit-operations
with modes. The triangular shapes are pool unit-operations and the diamond
shapes are called perimeters and represent the points where resources or
materials enter and leave the flowsheet i.e., sources and sinks. The circles
without cross-hairs are inlet ports and with cross-hairs are outlet ports where
ports are the flow interfaces to the unit-operations. The lines with
arrow-heads (→) connecting outlet-ports to
inlet-ports are called external streams or in graph theory they are called arcs
or directed-edges. The lines without arrow-heads are internal streams and refer
to inlet ports connected to unit-operations and unit-operations connected to
outlet ports. The rectangular shapes without cross-hairs are pipeline
unit-operations and the upside-down triangular shape is a parcel unit-operation
which is modeled as a batch-process and represents the flow of material from
one place to another and may represent road, rail and marine modes of
transportation [6]. The dotted line boxes surrounding several of the shapes
implies that the physical unit has more than one procedural operation
associated with it (has multiple modes of operation) and together forms what we
call a projectional unit-operation i.e., a physical
unit times a procedural operation equals a projectional
unit-operation.

Figure 1. Refinery Process Scheduling Optimization UOPSS Flowsheet.

Each
unit-operation and external stream have both a quantity and a logic variable
assigned or available and represent either a flow or holdup if quantity and either
a setup or startup if logic. Continuous-processes
have flows and setups and pools have holdups and setups where perimeters only
have a logic setup variable. The internal streams have neither explicit nor
independent flow and setup variables given that their flows are uniquely
determined by the aggregation of the appropriate external streams and their
setups are taken from the setup variables on the unit-operation they are
attached to. Although Figure 1 shows an almost complete and integrated oil-refinery
excepting the crude-oil and product blending and storing areas, we highlight
only the Crude-oil Distillation Unit (CDU), Vacuum Distillation Unit (VDU) and
Delayed-Coking Unit (DCU). The other included major process units known as the
Catalytic Reforming Unit (CRU), Fluidized Catalytic Cracking Unit (FCCU), Hydrocraking Unit (HCU), Gas Processing Units (GPU1 and
GPU2), Alkylation Unit (AKU), Hydrogen Producing Unit (HPU) and Steam Producing
Unit (SPU) are not discussed further.

The
two-mode CDU with operations of maximum gasoline (MAXGSL) and maximum diesel
(MAXDSL) is of interest given that we have included a pipeline with significant
holdup where two segregated crude-oil mixes labeled MAXGSL and MAXDSL are
transported from the crude-oil blending and storing area located several
kilometers away from the oil-refinery typically at an off-site
terminal. Following the pipeline modeling found in Zyngier and Kelly [6]
and given 30 one-day discrete time-periods, a pipeline with two modes (MAXGSL
and MAXDSL) can be arbitrarily shutdown at any time (i.e., has zero or no flow)
in any time-period and when flowing has a constant nominal flowrate. As can be
seen in Figure 2 for the Gantt chart resource row "PIPELINE,MAXGSL,o," in time-period 3 or day 3
there is a flow out of the pipeline of the MAXGSL crude-oil mix even though the
flow into the pipeline in time-period 3 is MAXDSL i.e., see Gantt chart row
"PIPELINE,MAXDSL,i,".  In the next
time-period (time-period 4) there is a flow of MAXGSL into the pipeline but
instead MAXDSL flows out of the pipeline. This is directly related to the fact
that a pipeline is modeled as a first-in-first-out (FIFO) queue or renewable
resource.

Figure 2. Gantt Chart for CDU with 1-day and 30-day Past and Future
Horizons.

There
are two other necessary logic features of the CDU's feed operation and they are
what we call multi-product and standing-gauge tanks. The multi-product logic is
implemented for TANKCDU2 which can store MAXGSL and MAXDSL but not at the same
time or in the same time-period. This means that when the optimizer decides to
switch the tank's material service from one to the other, the holdup or heel in
the tank must be at a certain amount specified by the user and in this case, it
is set to zero. The standing-gauge logic, also known as a dead-tank, ensures
that if there is flow into the tank then there can be no flow out of the tank
and vice-versa. This is configured for the three crude-oil tanks and is
apparent from the Gantt chart where the grey trend lines below the black horizontal
bars indicates the flow in (port-state "i,")
and flow out (port-state "o,"). The standing-gauge can be
generalized to restricting that if there is flow in then the flow out be at
least some number of time-periods after and this can be used to model mixing-
or certification-delays for example.

Figure
3 highlights the operation of the VDU which receives atmospheric residue or
reduced crude-oil (ATR) from both the upstream CDU and from an off-site
oil-refinery imported by rail. In this situation, the VDU feed ATR is
transported by what are known as unit-trains which will usually have around 100
tanker rail-cars per train. There is a round-trip or travel-time in this
case of 4-days where on the first day the ATR is loaded, on the second day it
is hauled from the source oil-refinery or terminal to the destination
oil-refinery, on the third day it is unloaded and on the fourth day it travels
back to the source and this is called industrial-shipping as opposed to liner
or tramp shipping. On the Gantt chart row labeled "TANKVDU,i," we see the extra
amount of ATR being received into TANKVDU from the parcel unit-operation
representing the unit-train deliveries of ATR.

Figure 3. Gantt Chart for VDU with 1-day and 30-day Past and Future
Horizons.

Figure
4 displays the Gantt chart for the DCU operation which converts the vacuum
residue labeled as VR into pyrolysis naphtha (PN) and distillate (PD) as well
as light gases such as methane, ethane (C1C2), propane, propylene, butanes and butylenes (C3C4) including a solid coke material (COKE).
The interesting aspect of the DCU is the production of coke which is modeled as
a semi-batch arrangement using pool unit-operations. The two physical coke
drums COKEDRUM1 and COKEDRUM2 have two operations or modes of FILL and DRAW
where the FILL operation can only charge material and the DRAW mode can only
discharge material. This means that if COKEDRUM1 is filling then it cannot
be drawing which implies that COKEDRUM2 must be drawing. Conversely, if
COKEDRUM1 is drawing then it cannot be filling and therefore COKEDRUM2 must be
filling. We can observe this behaviour or functionality on the Gantt chart
where we see that when COKEDRUM1 is in the FILL mode COKEDRUM2 is in the DRAW
operation and vice-versa.

Figure 4. Gantt Chart for DCU with 1-day and 30-day Past and Future
Horizons.

[1] Kelly,
J. D.; Zyngier, D. “Unit-operation nonlinear modeling for planning and
scheduling applications. Optimization and Engineering”, v. 18, p. 133–154,
2017.

[2]
Menezes, B.C., Kelly, J.D., Grossmann, I.E., “Logistics optimization for dispositions and depooling of distillates in oil-refineries: closing the
production scheduling and distribution gap”, ESCAPE 2017, June,
2018.

[3]
Menezes, B.C., Kelly, J.D., Grossmann, I.E., “Decision Automation for Lubes
and Asphalt Production Scheduling using MILP with Sequence-Dependent
Switchovers”, AIChE Spring Meeting,
Orlando, April, 2018.

[4]
Kelly, J.D., "The unit-operation-stock superstructure (UOSS) and the
quantity-logic-quality paradigm (QLQP) for production scheduling in the process
industries", In: MISTA 2005 Conference Proceedings, 327, (2005).

[5]
Zyngier, D., Kelly, J.D., "UOPSS: a new paradigm for modeling production
planning and scheduling systems", ESCAPE 22, June,
2012.

[6]
Zyngier, D., Kelly, J.D., "Multi-product inventory logistics modeling in
the process industries", In: W. Chaovalitwonse,
K.C. Furman and P.M. Pardalos, Eds., Optimization and
Logistics Challenges in the Enterprise", Springer, 61-95, 2009.