The selection of a process configuration for a refining and petrochemical production facility is
a task that requires both the skill of an engineer and the
business acumen of a finance practitioner. The selection
process usually requires facility yields and feedstock requirements, relevant
prices, unit capacity required, utilities consumption, capital
costs for the configuration (adjusted for location, timing and
location tax policies), and a financial model. The financial
metrics for each of the candidate configurations are calculated
and used to rank the different opportunities.
This article explores the use of linear program (LP)
modeling techniques to select process technology configurations based on
economic drivers (margin and capital), with the inclusion of a
simplified financial model built into the LP. The financial
model is used to estimate the key metrics and give the project developer an initial view of
the viability of the configuration.
Configuration selection problem.
The demand for additional product in local and selected
export markets are the usual drivers for new process capacity
located in an area that offers economic benefits. The
advantages of a potential facility location can include
feedstock availability, cost, location, byproduct markets, etc.
The project developer’s problem is how to
procure—with minimum capital investment—the
lowestcost feedstocks to produce the needed
finished products, and then generate sufficient economic
returns to justify the project.
Traditional problem solution.
The process configuration that will produce the desired
finished products has many variations. The goal is to
select—on an unbiased basis—the configuration that
will meet the economic returns needed at minimum cost. The
traditional methodology to perform this analysis is:
1. Construct an LP or other model of
the facility that incorporates all of the potential process
technologies envisioned for the project. The model should
include the following:
• Product specifications for transportation fuels, as
well as forecast changes to these specs as a result of environmental requirements
• Prices for crudes, other feedstocks and product prices in
constant dollars on a consistent basis
• Outside utility purchase availabilities and
costs
• Known or expected market limitations
• Process technology capacity limitations
• The latest crude assays for potential crudes
processed, with sufficient detail to support product
specifications of transportation fuels and process yield
drivers.
2. Establish a set of case studies to
evaluate each of the process technology configurations selected
by the design team, including:
• Configurations based on previous configuration
studies
• Input from subject matter experts (SMEs)
• Developer company’s requests.
3. Run the model for each of the cases
to determine the yields, gross margin and individual unit
capacity needed.
• Model process yields to determine sufficient
information to create an income statement for each case.
• Process unit capacities are based on the economic
drivers and technology licensor limitations provided for the
study.
4. Use highlevel capital cost
estimation procedures—i.e., cost
curves.
• Estimate the inside battery limits (ISBL) of each
process technology and other ancillary units.
• Estimate the ISBL cost at the selected site(s)
using a location factor.
• Apply a factor to account for the offsites,
owners’ costs and contingencies.
5. Use a financial model to determine
the metrics for each case.
• Use the appropriate tax, depreciation method, debt
interest rates, percent equity and accepted metric calculation
methods for each case.
• Metrics should include internal rate of return
(IRR), net present value (NPV), and aftertax payback (ATP)
period.
6. Rank various configurations by
capital expenditure (CAPEX) with the financial metrics to
select the final configuration.
This methodology has been used in industry and has provided
project developers with sufficient
data to make an informed process configuration selection.
However, this approach does have potential problems and
requires many model runs to fully explore the configuration
possibilities.
The usual start to a configuration study is an examination
of what the company did in its last analysis for similar
products. This can maximize the use of previous work and
potentially lower the project engineering costs. An
oftenoverlooked, but important fact is that every
configuration study is based on a unique set of feedstocks, market conditions and
economic drivers that can lead to different solutions.
A better approach to configuration selection.
The use of modern LP modeling systems with the inclusion of
capital costs can provide a more balanced and less biased view
of the process technology configuration selection.
The following sections present a method of adding the impact of
capital on the configuration selection problem, which is
implemented in a process industry modeling system (PIMS).
This system permits the user to model key points like ISBL
as a function of capacity, economyofscale exponents,
different stream factors for each process unit, different
location factors, different owner’s costs and contingency
factors. A calculation of IRR, NPV and ATP is also included,
using consistent pricing and capital costs (same dollar basis)
to give the developer an initial view of the project cost and economic
viability.
Methodology. LP solution techniques include
the concept of “recursion,” which is really a version
of the successive substitution method of solution to ensure
that the model is optimized and converged. The concept is that
the model starts with an estimated crude and feedstock composition and uses it to
determine the yields and properties of all model internal
streams. The LP solver then determines the optimal solution
based on this data. The solution can have different
crude/feedstock compositions and rates. The model internal
stream yields and properties are then recalculated by the PIMS
and compared to the previous values. The model is converged if
all of the internal stream property differences and stream
allocation dispositions are within the userspecified
tolerances. This technique, depicted in Fig. 1, is used by all
commercial modeling systems.

Fig. 1. Recursion
in a process industry modeling system
(PIMS).

The capital costs are modified during the recursion cycle via
the PIMS simulator interface, and are structured to impact the
economic solution (objective function) to account for the
differences in investment costs based on the process unit
capacities determined in the last solution cycle.
There is a major data difference in the LP that must be
addressed: the capacity per calendar day in the LP vs. the
capacity per stream day used in the capital cost estimation for
the ISBL. The technique converts the capacities from barrels
per calendar day (bpcd) to barrels per stream day (bpsd) before
estimating capital cost, and then back to bpcd before insertion
into the LP model. The total investment cost (TIC) for a given
unit is then updated based on the new capacity calculated by
the model. This is shown in algebraic form:
Unit TIC = [new capacity
(bpsd)/
base capacity (bpsd)]^{x} ×
basecapacity ISBL × factor
Where x is the economyofscale exponent, ISBL is the
insidebatterylimit cost of the process, and factor adjusts
the costs for location, offsites, owner’s costs and
contingencies. The sum of these costs for all process units in
the configuration gives an estimate of the facility TIC.
The TIC recovery for each unit is then the TIC calculated as
above, divided by all of the following: the capacity in bpsd,
multiplied by the stream factor, multiplied by 365 calendar
days, multiplied by five (years). These values are captured
using a PIMS utility row, which debits the economics (objective
function) of the model.
The methodology presented in this article uses a utility to
capture the amount of cash needed on a daily basis to pay for
the total installed process unit cost over a fiveyear period.
This technique also uses PIMS utilities to report the IRR, NPV
and ATP for the adjusted capital costs and process yields
during the recursion step of the solution. This is the simplest
method of transmitting solution information to the existing
PIMS reporting structure.
This technique also permits economics to drive the selection
of the process technologies used in the configuration. The
inclusion of the capital costs inside the LP greatly reduces
the number of LP runs and provides a reasonable estimate of the
IRR and other financial metrics for the selected
configuration.
PIMS model with cost estimates.
The general configuration is designed to produce
transportation fuels, monomers, aromatics and polymers, and has
many processing paths open; few unit capacities are limited or
at a minimum. The assumptions presented in Table 1 were used in
the creation of this model. The units have to “pay”
for the capital cost and operating costs when the capacity
limitations are “free.” The units are also limited by
the maximum product demand and licensor constraints.
Other financial model assumptions used in this example
include the following:
• Max. investment: $13 billion (B)
• Min. IRR: 13%
• Discount factor for NPV: 10%
• Depreciation: 14 years with 10% salvage value
• Construction period: four years
(yearly spending pattern of total installed cost: 24.4%, 43.2%,
25.2%, 7.2%)
• Tax rate: 25%
• Offsites: 60% of ISBL
• Owner’s cost: 10% of ISBL
• Contingency factor: 20% of ISBL
Selected process configuration.
The results of the optimal process technology arrangement determined by
the PIMS model using the economic drivers specified, the
financial assumptions noted, and the technologies available
will be presented in three sections:
• Finished product yields, feedstock consumption and
gross margin
• Process technologies selected and their
capacities
• Financial metrics from the LP and from a detailed
standalone financial model.
The facility production of salable products and the feedstocks required to produce them
are shown in Table 2.
No jet fuel or petroleum coke was produced. The
pricing supplied made it more profitable to produce only ULSD
while still meeting the flash point requirement. The only
aromatics produced were paraxylene and benzene. All of the
intermediate aromatics were converted or blended into gasoline.
The selected configuration for the refinery and petrochemical facilities is shown in Fig. 2.
Process Technologies.
The capacitylimiting process unit in the facility is the
RDS, as it limits the crude rate and sets the RFCC capacity.
The RFCC gasoline treater is not needed because the 77ppm
gasoline it produces can be blended with the other gasoline
components to meet the sulfur specification of 45 ppm max. The
RDS is more valuable in this case than the delayed coker
because the RDS configuration produces more net liquid that can
be upgraded to salable product, and the availability of
relatively inexpensive natural gas and coal mitigate the value
of petroleum coke as fuel. The process unit capacities required
to produce the finished petrochemicals and transportation
fuels are shown in Table 3.
Financial Metrics.
The financial metrics for the LP model are compared below to
those calculated by a standalone financial model with the same
capital, yields, capital spending pattern, depreciation and
other financial assumptions:
1. PIMS approximate method
• IRR: 13.3%
• NPV at 10%: $6.5 B
• Aftertax payback: 5.1 years
2. Detailed financial model
• IRR: 13.1%
• NPV at 10%: $6.3 B
• Aftertax payback: 5.2 years
3. Accuracy delta
• IRR: 1.5%
• NPV at 10%: 3.2%
• Aftertax payback: 1.9%
The financial metrics are not the same because of the
following factors:
• PIMS results contain nonsold items (nitrogen,
spent lime, combustion gases) to materialbalance the
model
• The detailed financial model excludes the nonsold
items and only uses what is sold
• These extra tons sold at a low price still impact
the PIMS view of the configuration profitability and make the
PIMS results higher for the IRR and NPV, but lower for the
payback time.
Takeaway.
The inclusion of capital costs and financial metrics in an
LP model of a potential configuration provides the user with an
unbiased view of the facility, based on economics. The
identified configuration is a viable starting point for the
selection analysis and should be part of every feasibility
study where the LP is a key analysis tool.
This technique does not eliminate the need for sound
engineering judgment nor the role of the SMEs. These are needed
when the “optimal configuration” is converted to a
practical solution with all of the elements not considered in
the LP model. These elements include sour water stripping;
practical unit capacities; multiple pressure steam levels;
ancillary utilities; flares; and health, safety and environment concerns.
HP
The author 


Timothy E.
Swaty has more than 37 years of professional
experience in most aspects of petroleum refining, and broad exposure
to the petrochemical industry. He is currently a
principal technical specialist with Fluor. His work at
Fluor includes the application of management science
techniques to refining and petrochemical problems,
process optimization and financial analysis. Mr. Swaty
has a BSChe degree from the University of Houston and an
MBA from Texas A&M University (Corpus Christi). He is
a registered professional engineer in Texas, Kansas and
California. 