Documente Academic
Documente Profesional
Documente Cultură
REFINING
L
inear programming (LP), a petrochemical feedstock preparation
technology first applied during “As barriers that once units. These refineries have developed in
the Second World War to help stages over a long period of time, with
solve troop-supply problems, continues separated individual the last modernisation completed in
to be the premier tool for determining refineries continue to fall in 1988.
the optimal distribution of limited Finished products from the refineries
resources. Nowhere is this more evident an attempt to improve are transferred to tankers berthed at the
than in the petroleum refining industry. overall industry health, North Pier, New South Oil Pier, SHU Oil
As barriers that once separated Pier, and Sea Island. MAA products are
individual refineries continue to fall in linear programming is used exported to North Pier and South Pier,
an attempt to improve overall industry to identify the synergies and MAB products to Sea Island and SHU
health, LP is used to identify the products to SHU Oil Pier. All of the
synergies and operational improve- operational improvements refineries have been integrated for better
ments that result. that result” feedstock management and product
sharing through six inter-refinery
Inventory disruptions transfer (IRT) lines. These include two
Historically, the way to alleviate product detailed tankage and hydraulic study 24in lines for black oil products, two
or feedstock supply problems in for the Kuwait National Petroleum 20/24in lines for white oil products, one
petroleum refining was to “build Company (KNPC). This assessed the 14in line for motor gasoline compo-
another tank”. These tanks provide adequacy of existing tank farms at all nents and one 20in line for naphtha/
extra storage capacity, which effectively three KNPC refineries for operations kerosene. The refineries continuously
reduces the time element in operations through 2010, and developed an exchange products for use in their
planning, lessening the impact of optimised solution to the refineries’ various process activities or blending
disruptions, both planned and tank farm needs. In addition to dealing operations in order to create products
unplanned. Tanks allow stocks to be with routine operations such as planned that are ready to export. KNPC defined
more readily available before they are turnarounds, the study had to assure the objectives of the study as follows:
needed, or held when their transfer is that the tank farm capacity would be — Outline an interim solution for
delayed. They give refiners more time to sufficient to handle unplanned events tankage deficiencies in KNPC refineries
bring their contents to specification such as ship delays, power failures, before anticipated upgrading and
quality or to prepare for expected weather problems and unit outages, as expansion projects are completed in
outages. Tanks also provide for easier well as an expected increase in refined 2008/2010
stock segregations when enough exist to product diversification. This would — Carry out a simulation study to assess
designate each in limited services. allow for optimum utilisation of the the future tankage requirements for
However, new tanks are expensive to available storage capacity, while also KNPC
build ($26.50–40.00 per barrel installed) minimising capital cost for the — Check for each refinery, and for
and maintain, and difficult to justify upgrading facilities. KNPC in general, the adequacy of
both economically and environ- existing intermediate and finished
mentally. In addition, most refiners are Inter-refinery supply study product storage capacity, number of
faced with limited or no available real KNPC operates three refineries in tanks, IRT systems, unit charge systems,
estate. Therefore, even though Kuwait with a total crude throughput of blending and dispatch facilities, and
throughputs and product diversification approximately 900 000bpd. The refiner- identify any modifications required
are on the rise, refiners are challenged to ies are Mina Al Ahmadi (MAA), Mina — Check the adequacy of the
operate within the same or less available Abdulla (MAB) and Shuaiba (SHU). hydraulics and flow metering systems to
inventory capacities. While SHU is KNPC’s oldest refinery, ensure efficient operation of the
So how much tankage capacity is MAB is KNPC’s most modern. MAA, refineries and export operations
enough? How much is excessive? Foster however, is the largest and most — Identify the additional storage
Wheeler (FW) recently completed a complex, with FCC, hydrocracker and capacity, tankage requirements and
1
P T Q Q2 2005
w w w. e p t q . c o m
08 haverly 13-05.qxd 5/13/05 1:09 PM Page 2
REFINING
2
P T Q Q2 2005
08 haverly 13-05.qxd 5/13/05 1:27 PM Page 3
REFINING
and viscosity for fuel oil). emergency process unit outages and
shipping delays, a more dynamic aspect
Model execution “By defining the known was needed. The randomness provided
A simplified depiction of the overall linkages within a system, through the Monte Carlo simulation
model structure is shown in Figure 2. created this dynamic representation of
The multi-period/multi-refinery LP
the simulation model is real-world events.
model was run over a total of five years unconstrained by complexity Also, the nature of LPs themselves is
(five years corresponding to the such that they do their best to anticipate
maintenance planning cycle for the
and quite accurate” and prepare for upcoming circum-
refining system). To achieve sufficient stances. This feature had to be
resolution for proper analysis, this time overridden to deal with unplanned
frame was broken down: first, into five automatically run using the GRTMPS events. By applying the unplanned
annual models of 52 time periods each; Evaluation Wizard. element of the study to the one-week,
and second, into 260 one-week single An interesting side note concerns the single-period models only, any
periods. One week was chosen as the speed at which the 52-week models ran. preparation behaviour was eliminated,
minimum resolution time required for Initial solution attempts required in restoration time minimised and tankage
evaluating inventory utilisation within excess of 40 minutes per run, which was forced to deal with any upsets caused by
the tank farms of the refineries and still unacceptable given the number of runs the disturbances. Final inventory
providing reasonable run-times for the needed to establish a good baseline. positions from each one-week, single-
model. However, a change in the LP solution period model were then passed onto the
The 52-week models were run with all technique, from Simplex to the Barrier, next week, so imbalances were worked
planned outages and conditions to or interior point, technique, brought off over several weeks when necessary.
establish a baseline for each year, with the run-times down to just under four In addition, inventory capacity was
closing inventory from one year passed minutes. Although the Barrier modelled so that some specific limits
on as opening inventory in the technique has been shown to perform could be violated at a cost, but total
following year. Then, the 260 one-week better than Simplex on very large, sparse inventory capacities were firm.
models were run numerous times to LP models, an improvement of this size Modelling these specific limits as “soft”
record the effects of unexpected outages was quite unexpected. allowed the model to indicate where
and shipping. A special utility was A multi-period/multi-refinery LP and when it could be advantageous to
written to automatically apply the model alone would have been sufficient consider changing the service of tanks.
Monte Carlo simulation results to the to perform the study if plant operations In the course of execution, the model
one-week models and define the cases. were predictable and steady. However, produced a tremendous amount of data
The cases were then stacked and due to the nature of unplanned or that had to be compiled and analysed.
3
P T Q Q2 2005
08 haverly 13-05.qxd 5/13/05 1:10 PM Page 4
REFINING
Figure 3 Typical model output for a tank group (current tank farm operations)
With Haverly’s assistance, FW modified presentation. for each case scenario to assess facilities
the report-generation feature of utilisation. The scenarios that were
GRTMPS to capture the relevant Case summaries examined include the following:
information related to tank farm FW used a stepwise approach to analyse — Base case Current operations: the
inventory management and download the tank farm operations, with a stated current refineries’ tankage, blending
it into Microsoft Access. The raw data goal to keep recommended capital and shipping configurations were
was then exported into Microsoft Excel investments to a minimum. A series of examined, given planned 2010
for numerical analysis and graphical simulation model runs was performed operations. This model represents the
Figure 4 Typical model output for a tank group (after implementing all proposed recommendations)
4
P T Q Q2 2005
08 haverly 13-05.qxd 5/13/05 1:32 PM Page 5
REFINING
current tankage condition at all three line. also allows for future expansion and
refineries and incorporates all current If the upper band line exceeds 100% upgrading of the refineries without being
interconnections between the refineries, capacity utilisation, this would indicate constrained by tank farm operations.
based on data collected during FW’s site that available tankage capacity is not
visit and output from KNPC’s own LP sufficient and hence a potential Conclusion
model. The model is initially populated restriction to tank farm operations. The primary conclusion of the KNPC
with current tankage, planned unit Conversely, a band line that falls below study was the determination that no
outages, planned tank maintenance and 100% capacity utilisation indicates that additional tankage would be required
pump limits there is sufficient tankage available to for them to operate comfortably into
— Case 1 Re-allocation of existing deal with normal tank farm operations. the next decade. Additionally, sufficient
tankage: based on constraints identified The following results were captured cost savings have been identified that
in the base case runs, existing tankage is from the tank farm simulation model, can pay back the cost of system
reallocated to allow reasonable levels of measured and trends recorded: improvements within two years.
utilisation within tank groups — Process unit capacity utilisation As demonstrated by this project,
— Case 2 In-line finished product — Tankage utilisation as a percentage of stochastic studies such as the one
blending: automated, in-line blending working capacity presented in this article can now be
facilities allow finished products to be — Frequency that tank filling is effectively and confidently conducted
blended as required, rather than the constrained by maximum working using today’s LP tools. Coupling
current practice of building a full batch capacity statistical modelling methods with LP is
of inventory prior to shipping release, — Product properties and transfer flow an approach promising great returns,
thus reducing hold time in tankage rates within the refineries, via inter- not only in inventory studies such as
— Case 3 Inter-refinery transfers of refinery transfers and to final product this but also in studies involving price
finished products: by permitting shipping facilities and quality sensitivities, feedstock
transfers of finished products between — Ship movements as a result of two- selections, market place dynamics, and
refineries (ie, tank-to-tank transfers), port loadings investments of capital.
distribution system flexibility is — Product movement operating costs
improved — Final product blend values
— Case 4 Elimination of dual-port — Improvements in all of these as a
product loading: this is similar to Case result of defined system modifications GRTMPS is a trademarked technology of
Haverly Systems. Microsoft Access and
3, except finished products from one and expansions. Microsoft Excel are trademarked products
refinery can be shipped directly to the For each case, FW identified, of Microsoft Corporation.
loading port of another refinery without quantified and compared the benefits to
going through finished product tankage the base case. Estimated capital costs for References
first recommended system improvements as 1 Al-Otaibi G A and Stewart M D, Simulation
— Case 5 Addition of new tankage: if well as expected revenue increase and/or model determines optimal tank farm
design, Oil & Gas Journal, Vol 102.7, 50–55,
reasonable system utilisation cannot be cost savings were used to determine a 16 February 2004.
achieved by the changes in facilities, project payout period.
practices and procedures previously Weekly operating costs, as
identified, additional refinery tankage determined by the tank farm simulation
will be considered. model runs, were aggregated into Michael D Stewart is chief process
annual average costs over a five-year engineer for Foster Wheeler NA Corp,
Model results period. Annual credits for each case were Houston, Texas, USA, and manager for
The base case model output shown in the difference in the average annual planning activities and studies associated
Figure 3 is for a typical tank group operating costs vs the base case. FW with refinery and petrochemical process
before any system configuration estimated capital costs for each case plants. He has over 27 years of
changes have been made. Figure 4 from in-house data and previous international experience. He holds a BS
shows the improvement after experience with similar projects and degree in chemical engineering from West
implementing all recommendations. equipment. Virginia University, an MBA from Marshall
Individual lines represent specific Implementation of the recommended University and has completed graduate
Monte Carlo simulation runs. The bands system improvements resulted in a studies in economics at the University of
indicate the statistical average and project payout period slightly in excess Oklahoma. He is a registered professional
standard deviation for all runs. of two years. The total overall project engineer in West Virginia and Texas, and a
Methods other than visual cost was reduced due to the early member of the American Institute of
observation of the resulting charts were decision to minimise installation of any Chemical Engineers.
used to analyse tankage utilisation. For new tankage. Email: Michael_stewart@fwhou.fwc.com
analysis of tank group capacity The results of the aforementioned L Dean Trierwiler is the business and
utilisation, the mean tank group analysis were presented to KNPC as well technical manager at Haverly Systems Inc,
~
utilisation (u) and the standard as the economic benefits for upgrading Houston, Texas, USA. He joined Haverly in
deviation (δ) of the data set are the tank farms at the three refineries. 1990 and has since worked in the
calculated and displayed on the charts. Consequently, KNPC is now management, support and application of
Assuming normal distribution of implementing the upgrading project Haverly’s planning, scheduling and crude
probability, lines are then drawn on the based on the system improvements as assay software tools. Trierwiler holds a BS
~ ~
chart to represent u and u+2δ, which defined in Cases 1 through 4, confident degree in engineering from Washington
indicates that 95.45% of tank group that the optimal capital investment State University.
utilisation fits below the upper band strategy has been chosen. This project Email: dean@haverly.com
5
P T Q Q2 2005