Implications of Residual Fuel Oil Phase Out
By David Ramberg, Center for Energy and Environmental Policy Research and the Joint Program on the Science
and Policy of Global Change, Massachusetts Institute of Technology, 77 Massachusetts Ave., Cambridge, MA
02139, E19-411, Phone: 1.503.442.2133/Fax: 1.617.253.9845, [email protected], and Sam Van Vactor,
Economic Insight, Inc., 1.503.222.2425, [email protected].
Abstract
By 2020, the International Convention for Prevention of Pollution from Ships (MARPOL),
adopted by the International Maritime Organization (IMO), will mandate that no marine vessel
will be allowed to consume fuel with a sulfur content greater than 0.1%. Since current bunker
fuels generally contain more than 3% sulfur, residual fuel oil may very well be phased out of the
markets. Whether (and to what extent) this occurs depends on the interaction between the
decisions of refiners, the demand for boiler fuels in general, and the supply/demand balance of
boiler fuel alternatives. This paper examines refiners’ decisionmaking processes with regard to
the destruction of residuum, which can be either further refined into light products or blended
into residual fuel oil. A simple residuum upgrading model is developed and its logic is briefly
tested with empirical data. Current trends in the usage of residual fuel oils and bunker fuels, as
well as the relative supply of heavy vs. light crude oils, are examined in the context of the refiner
model to determine the likely fate of residual fuel oils. The focus is on background and a
characterization of market forces and influences rather than a rigorous test of alternative
hypotheses.
1.
Introduction
This paper briefly examines the trends in residual fuel oil demand as well as the most
common methods of disposal for refinery residuum and characterizes a plausible scenario for the
future of residual fuel oil demand. The focus is on background and a characterization of market
forces and influences rather than a rigorous test of alternative hypotheses.
Residual fuel oil demand has been declining and is set to decline even more dramatically
in the coming decade if current trends prevail. At the same time, refiners are finding ever-heavier
crude oils on their slates, with an increasing amount of residuum content. Other than asphalt
production, the two principal methods for residuum disposal are blending into residual fuel oil
and coking. Thus, further decline in the residual fuel oil market suggests a strong incentive for
refiners to add additional coking capacity.
2. Status of the heavy fuel oil market
2.1 Types of refineries and the refining process
Petroleum refineries are complex operations. In general, they are classified as simple or
sophisticated, but within the sophisticated category there are a wide variety of options. All
simple refineries have distillation towers in which crude oil is heated until various “cuts” boil
off. When condensed, many of these cuts are similar to refined products and can be marketed
37th IAEE International Conference, New York, NY, June 15-18, 2014
1
with a minimum of further processing. Simple refineries have limited upgrading facilities and
they sometimes sell their output to other refineries for further processing. Declining petroleum
consumption since 2008 has caused many simple refineries to shut down.
Sophisticated refineries start processing like the simple ones, with a distillation tower.
However, they add additional steps and process the resulting cuts into marketable products with
a variety of upgrading facilities. The most common of these processes are thermal and catalytic
“cracking.” Complex refineries often have sulfur removal units and a variety of other facilities
required to meet strict product standards for gasoline and low-sulfur diesel. Some sophisticated
refineries go a step further and add a coking unit. Coking allows refiners to take the heaviest cut
of oil, referred to as residuum, from the distillation tower and break it into lighter products such
as diesel and gasoline by applying heat and catalysts. The process also produces coke, a solid
carbon product not unlike coal. Cracking can also upgrade some of the residuum, but at higher
marginal cost, mostly because the cracking of residuum fouls up the sensitive catalysts in the
cracking units.
Academically there have been few studies characterizing refinery economics. Part of the
reason for this is that each refinery is uniquely designed to use specific crude oils as feedstocks
and to produce a slate of products in demand in its targeted markets. That makes “generic”
characterizations of refinery economics difficult, if not impossible. Some investigators have
nonetheless produced some useful methodological contributions. Dixon produced a model-based
approach for a modern coking refinery (Dixon, 2009).1 Others have developed a model using
graph theory based on the characterization of a refinery as a complex interconnected system
(Sadhukhan, et al., 2003). 2 Some forward-looking researchers have also investigated the
possibility of upgrading refineries to both reduce emissions and eliminate the residuum entirely
(McKenna and Sheikh, 2010). Interestingly, this paper advocated the addition of a coking unit as
well as a gasifier to dispose of both residuum and petroleum coke.3 The gasified coke would be
used to provide the refinery’s electricity needs. In general, the literature on refinery economics is
rather sparse, and is largely confined to a couple of chapters in textbooks with a broader focus on
refinery processes (e.g., Gary, Handwerk, and Kaiser, 2007 and Fahim, et al., 2010).
2.2 Types of products
A modern sophisticated refinery produces several dozen products. Output varies by
season and in response to market prices. Refiners maintain sophisticated linear programming
models that compute daily profitability based on marginal processing costs, product market
prices, and crude oil costs.
Every refinery begins crude processing with an atmospheric distillation tower that heats
the crude feedstock to boiling. Gasified petroleum products rise through the tower, and different
1
Dixon is employed at a ConocoPhillips refinery.
Graph theory uses mathematical modeling to characterize relationships between elements. It is made up of hubs,
called “nodes” or “vertices”, which are connected by linkages called “edges”. Each edge contains the information
about how each node is connected to adjacent nodes. In the Sadhukhan et al. model, nodes represent refinery units,
and edges track the physical flow of product between units, as well as the energy needs and other relevant
information for each.
3
Increasing production of Canadian tar sands has produced a coke surplus in the Midwest.
2
37th IAEE International Conference, New York, NY, June 15-18, 2014
2
components of the crude oil condense in stacked trays positioned within the tower at decreasing
temperature levels. The heavy components with higher boiling points condense in the lower trays
of the tower. The components with the lowest boiling points rise to the top as gases. The top of
the distillation tower produces propane, isobutene and other gases. The next set of cuts includes
naphtha, which is the primary building block for gasoline. Following these light products are the
middle distillates – mainly kerosene and diesel. As the name implies, they are the middle part of
the barrel and in many respects are the most flexible of all the cuts. That is because middle
distillates can be refined into a variety of products – gasoline, motor diesel, jet fuel, and heating
oil. The flexibility of middle distillates makes them the most valuable part of a crude oil barrel.
Once the useful cuts have boiled off, the refiner is left with the residuum, a tar-like
sludge. Moreover, impurities, like sulfur, tend to collect in this last cut. In the past residuum has
been a waste product. Refiners burned it or paid someone to cart it away. Except in a few
jurisdictions, environmental regulations prevent such activity today. Residuum is not directly
marketable. There are three main options for its disposal: 1) further processing into asphalt, 2)
blending with middle distillates to market as FO-380 (the fuel is also known as residual, bunker
or heavy fuel oil), and 3) upgrading residuum into coke and light petroleum products – mainly
middle distillates and gasoline.
This paper takes the asphalt market as more or less given and focuses on the tradeoff
between blending residuum for the FO-380 market or, alternatively, upgrading residuum into
light petroleum products. We do not attempt to model the operations of an entire refinery, which
is much more complex than the distillation tower and coker we have described. In fact, after
atmospheric distillation, the residuum is passed through a vacuum distillation tower where more
cuts can be extracted. The residuum of the vacuum distillation tower (eventually) becomes the
feedstock for the coker. For the purposes of analysis here, it is sufficient to consider only that the
refiner will always have some amount of residuum it needs to destroy, and that aside from
asphalt production, the only options are coking to gases, middle distillates, and petroleum coke
or blending into residual fuel oil. Before proceeding with that analysis, however, it is worth
describing the FO-380 product and its associated costs of processing.
Most, if not all, developed economies prohibit burning high-sulfur fuel oil, with one
major exception – bunker fuel for ships operating on the high seas. A small amount of lowsulfur (less than 1%) FO-380 is burned for power generation, but even this is being phased out.4
The reason is simple economics. Diesel or light fuel oils are more efficient for use as motor oils
or power generation. If refiners have to process heavy fuel to remove sulfur it is not that much
more expensive to upgrade the material through coking to a middle distillate, where sulfur
removal is part of the built-in process.
High sulfur FO-380 is the cheapest fuel oil a refiner can make. Generally, it is about
two-thirds residuum and one-third diesel. Product specifications are met by blending the two
fuels. Blending is often cheaper than upgrading, but it is not costless. The blender must
maintain adequate storage, including heated tanks. Moreover the market is small and sporadic,
with high transaction costs.
4
A good example is Hawaii’s power generation company, Hawaii Electric Co. It has announced its intention
to stop using residual fuel oil in the near future and replace it with diesel (HECO, 2013).
37th IAEE International Conference, New York, NY, June 15-18, 2014
3
2.3 Who uses heavy fuel oil
Thousand!barrels!per!day!(MB/D)!
Figure 1 illustrates the trend in heavy fuel oil consumption as collected by the Energy
Information Administration (EIA, 2014). Overall consumption declined 31% from 1986 through
2010. The vast share of the
Figure!1:!!Residual!Oil!ConsumpIon!
drop,
however,
was
!14,000!!
experienced
by
North
America,
Europe
and
!12,000!!
Japan: a 54% decline. In
the remainder of the world
!10,000!!
the decline was only
11.4%; consumption rose
!8,000!!
and then fell in China
!6,000!!
during the period, muting
the decrease.
!4,000!!
Thousand!barrels!per!day!(MB/D)!
Residual fuel oil use
!2,000!!
has declined even as world
petroleum consumption has
!"!!!!
increased. In 1986 heavy
1986!
1989!
1992!
1995!
1998!
2001!
2004!
2007!
2010!
fuel oil accounted for
North!America!
Europe!
Japan!
China!
Other!
Source:(EIA(
nearly 21% of total
petroleum consumption. By 2010 the share had fallen to around 10% (EIA, 2014). The IEA
calculates similar figures – 20% in 1986 and 9.5% in 2010 (IEA, 2013).5 There have been two
principal causes of the decline: first, cheaper fuels - coal and natural gas - easily substitute for
heavy fuel oil in the boiler
Figure!2:!!Bunkering!and!Non"Bunkering!Residual!Oil!
fuel market. Second are the
ConsumpKon!
!14,000!!
environmental constraints
placed on the use of the fuel
!12,000!!
in
most
developed
countries.
This
constraint
!10,000!!
on use is due to sulfur
!8,000!!
emissions arising from
burning the fuel. Heavy
!6,000!!
fuel oil has a slightly higher
carbon emission compared
!4,000!!
to other oils and its footprint
is significantly higher than
!2,000!!
natural gas, but the impact
of CO2 emission regulations
!"!!!!
1986!
1989!
1992!
1995!
1998!
2001!
2004!
2007!
2010!
is not yet an important
North!America!
Europe!
Japan!
China!
Other!
World!Bunker!
consideration.
Source:(EIA(
5
The IEA and the EIA use different methodologies to calculate residual fuel oil and bunker fuel usage.
Rather than arbitrarily decide which data set is more accurate, we have chosen to present the data of each
organization to provide the reader with a range of reasonable estimates.
37th IAEE International Conference, New York, NY, June 15-18, 2014
4
Overall, the decline in residual fuel oil usage is masked by an increase in its use as a fuel
for maritime bunkering. The decline in residual fuel oil usage excluding its use as a bunker fuel
is over 50% globally. Among OECD countries, the decline is nearly 70% from 1986 through
2010; among the non-OECD nations, the decline is over 37% (EIA, 2014). Only the use of
residual fuel oil as a bunker fuel has increased over this period. Bunkering increased from just
9.3% of global residual fuel oil consumption in 1986 to over 35% in 2010 (EIA, 2014). Data
from the International Energy Agency show the bunkering percentage of total residual fuel oil
consumption increasing from about 12.4% in 1971 to nearly 39% by 2011 (IEA, 2013). Figure 2
uses the EIA data to depict the decline in residual fuel oil usage for non-bunkering purposes, and
to superimpose the rising global bunker fuel consumption trend over the 1986-2010 period. This
suggests that non-bunker fuel usage of residual fuel oils is declining even in developing nations.
According to the IEA, usage of heavy fuel oil used for shipping – bunker fuel – increased
from about 1.6 million barrels per day (mmb/d) in 1986 to over 3.2 mmb/d by 2010 (IEA, 2013).
The EIA calculates a similar increase, from nearly 1.2 mmb/d to over 3.1 mmb/d in the same
period (EIA, 2014). Generally, the breakdown of heavy fuel oil consumption in 2010 is as
follows:
Table&1:&Breakdown&of&Residual&Fuel&Oil&Usage&in&2010,&million&b/d
Use
EIA
IEA
EIA&%&Total IEA&%&Total
Bunkers
3.1
3.2
35.1%
38.8%
OECD
1.7
1.6
19.1%
19.6%
Non8OECD
4.1
3.5
45.7%
41.6%
Total
8.9
8.3
100.0%
100.0%
Sources:?IEA?World,Energy,Statistics,and,Balances,?EIA,International,Energy,Statistics
2.4 Heavy fuel oil pricing
Since heavy fuel
oil is an inferior petroleum
product (containing many
impurities and higher
handling
costs),
its
expected price ought to be
below crude oil prices.
That is normally the case,
but there has not been a
consistent
relationship
between the two price
series. At times FO-380
prices come close to parity
with crude oil. At other
times they are deeply
discounted.
Figure$3:$Comparison$of$Crude$Oil$and$FO-380$Prices$
$140.00$
$120.00$
$100.00$
$80.00$
$60.00$
$40.00$
$20.00$
$0.00$
Jan-1987$
Jan-1991$
Jan-1995$
Heavy$Fuel$Oil$
Jan-1999$
Jan-2003$
Brent$Crude$Oil$
Jan-2007$
Price$Difference$
Jan-2011$
Figure 3 illustrates
Source: EIA 37th IAEE International Conference, New York, NY, June 15-18, 2014
5
the relationship between average wholesale high-sulfur fuel oil prices in the U.S. as compared to
the European price for Brent North Sea crude oil in US dollars per barrel (EIA, 2014b, c). The
chart illustrates that heavy fuel oil prices were deeply discounted in two periods: during first
Gulf War and in the 2007-2008 oil price runup. In both cases there was a rapid increase in
petroleum product and crude oil prices, with heavy fuel oil prices lagging.
3. Simplified model of refinery upgrading
A simplified model can be used to explain how the dynamics of petroleum product and
crude oil prices work in the refinery system. The price differential between light and heavy
petroleum products generally correlates with differentials between light and heavy crude oils.
That is, the wider the differential between heavy and light refinery feedstock, the greater the
incentive to upgrade.
There are many different technological options for upgrading heavier oil into lighter
products and petroleum product demand growth is not a prerequisite for additional upgrading
capacity. The need actually depends on the relative mix of petroleum products in demand.
There are a number of ways to model the economics of a refinery that processes feedstocks from
a specific slate of crude oils and produces a specific slate of products based on the process units
that it contains.
Sadukhan, Zhang and Zhu presented a model of a complex refinery based on graph
theory that does not require optimization for identification of the lowest-cost method of
producing the highest-value slate of products (Sadukhan, et al., 2003). Their model is adaptable
to a variety of refinery types, but its complexity rivals that of the optimization-based linear
programs employed by the refiners themselves. Dixon produced a much more tractable model for
refinery production of anode-grade petroleum coke using a simplified coking refinery model
(Dixon, 2009). Dixon’s coking refinery contains an atmospheric crude distillation unit, a vacuum
distillation unit, a delayed coker, a gas oil hydrotreater, a fluidized catalytic cracker, a distillate
hydrotreater, a reformer/isomerizer, and an alkylation unit. Its purpose was to model the
changing product slate when shifting toward a heavier, more sulfur-laden crude slate. Both
papers relied on economic and operating data for refinery process units included in alternative
editions of Petroleum Refining Technology and Economics (Gary, Handwerk, and Kaiser, 2007).
For the model presented in this paper, we have a much narrower focus: given that refiners
will always have some amount of residuum produced by their refining activities, what are the
incentives to upgrade to higher-valued products? The model we use to explore this question is
much simpler than those described above. The decision on whether to coke residuum into highervalued products is based on the difference in value between the feedstock and the slate of higher
valued products that coking produces. The wider the difference, the greater the value to the
refiner of coking. Heavy vs. light crude oil price differences can serve as a proxy where product
prices are not available.
Figure 4 illustrates the classical Marshallian relationship between feedstocks and
products, but with a key modification. Ordinarily the vertical axis measures price. In Figure 4,
however, the vertical axis measures the price or value difference between heavy and light crude
37th IAEE International Conference, New York, NY, June 15-18, 2014
6
oil or between heavy and light petroleum products. The horizontal axis measures the quantity of
residual oil left over from the simple distillation process that is upgraded into lighter, highervalued products.
Figure'4:''Refinery'Upgrading'Economics'
All&
Upgrading&&
Capacity&
Coker&
Capacity&
Price&Difference&
Light&Products&
&&Residual&Oil&&
Or&Heavy&&&
Light&Crude&&
Oils&
Other&Upgrading&
Processes&
SRMC&
ΔP2&
LongIRun&&
Coker&&
Cost&
ΔP1&
D2&
Variable&
Coker&&
Cost&
D1&
Quan<ty&of&Light&Products&Refined&from&Residuum&
As a general rule,
refiners have rising short-run
marginal
cost
curves
associated with upgrading
using thermal or catalytic
cracking. Coking, on the
other hand, is far more
capital intensive but has a
relatively flat marginal cost
curve until the coker is at full
capacity. As a general rule,
the variable cost of coking,
and therefore its marginal
cost, is proportionally lower
than that of the other
processes until fully used.
Under this logic, as long as a
coking refiner has access to
residuum feedstock, cokers
will be utilized to full
capacity before turning to the
other options.
The model can be expressed mathematically as follows:
The supply curve is segmented between coking units and other upgrading processes:
𝑆𝑅𝑀𝐶 = 𝑀𝐶! when 𝑄 ∈ (0, 𝑄! ],
𝑆𝑅𝑀𝐶 = 𝑓 ∆𝑃, 𝑄 when 𝑄 ∈ 𝑄!! , 𝑄! , and
𝑆𝑅𝑀𝐶 = ∞ when 𝑄 > 𝑄! .
𝑆𝑅𝑀𝐶 denotes short-run marginal cost, 𝑀𝐶! is the marginal (variable) cost of operating a coking
unit, 𝑄 represents the quantity of light products produced specifically from residuum upgrading,
𝑄!! is the coking capacity in the market, 𝑄! is the total residuum upgrading capacity from all
processes, and ∆𝑃 is the price difference between the slate of lighter products produced and
residual fuel oil, for which the difference between heavy and light crude oils can serve as a rough
proxy. 𝑓 ∆𝑃, 𝑄 is a steeply increasing function.
One might ask why the price difference is between light products and residual fuel oil,
rather than residuum. The reason is because if the price difference between the light products and
37th IAEE International Conference, New York, NY, June 15-18, 2014
7
residual fuel oil is less than the short-run marginal cost of upgrading residuum to lighter
products, then the least-cost option would be to blend residuum into residual fuel oil rather than
upgrade it to lighter products through coking. (Note, however, if a refiner chooses not to run its
coker it must also account for the opportunity costs associated with possible underutilization of
other refinery processes. Thus, the calculation of short run marginal cost of upgrading is highly
complex. In some cases it may actually be negative.)
The demand curve is a decreasing function of the price differential between light
products and residual fuel oil:
𝑄∗ = 𝑓(∆𝑃)
If ∆𝑃 is very large, then consumers will not demand light products beyond what is produced
from less intensive refinery operations. The logic is that if residual fuel oils are heavily
discounted, they will be preferred to the lighter products even though they are dirtier, more toxic
and cause more damage to equipment on combustion. As the price differential narrows, the
difference in fuel quality becomes more important, and consumers will increasingly prefer the
light product to the heavy one.
In this model, the equilibrium quantity is fully determined by the demand curve’s
intersection with ∆𝑃 (𝑄∗ = 𝑓(∆𝑃)). By definition, the differential between heavy and light
products determines the equilibrium quantity at which consumers are satisfied with the relative
production of various types of petroleum products.
A number of elements of the equation are difficult to estimate from outside of the
industry due to a lack of data. For example, the SRMC curve for non-coking processes is not
known because the markets rarely provide the incentive for refiners to re-tool their cracking and
other upgrading units to process residuum into higher-valued products. Certainly, the refiners do
not announce which process units are upgrading residuum during the course of their operations.
But we serendipitously do have an insight into the variable and long-run costs of coker
operations thanks to open testimony on the setting of the Trans Alaska Pipeline System (TAPS)
quality bank tariff.
On the TAPS system, drillers feed crude oil of varying qualities to the pipeline at the
North Slope, where they are commingled so that a crude of average quality is received at the
southern terminus where it is loaded onto tankers. In order to compensate the producers who
supplied higher-than-average quality crude oil into the pipeline, the producers who supplied
lower-than-average quality crude pay into the Quality Bank (QB). The high-quality producers
receive payments from the QB to reflect the actual value of the crude they supplied.
Since 1995, the Federal Energy Regulatory Commission (FERC) has mandated that the
crude streams be valued using a distillation method. The distillation method values crude streams
by using assay data to construct what each crude stream would produce in petroleum products
after simple atmospheric distillation – a process unit present in all refineries. The yield-weighted
value of these distillation “cuts” comprises the crude stream value. In 2005, the FERC decided
that the residuum cut, lacking any market price by which to assess its value, should be priced at
37th IAEE International Conference, New York, NY, June 15-18, 2014
8
the yield-weighted values of the cuts after processing residuum in a coker, less the coker’s costs.
A return on capital was included, as was an 87% utilization rate and a 20-year lifespan (FERC,
2005). During public hearings, an estimated cost breakdown for coking residuum was revealed in
publicly-available testimony.
It emerged that as of the year 2000, for a West Coast refiner processing Alaska North
Slope (ANS) crude oil, the per-barrel cost of residuum processing through a coking unit was
$8.24. $5.54 was the per-barrel capital cost for a West Coast-based coking unit including a 4drum coker and associated equipment as well as two sulfur plants (to remove 100% of the sulfur
from the stream) and a downstream hydrotreater. The fixed cost per barrel was $1.83, and the
variable operating cost per barrel was $0.87 (FERC, 2006, pp. 23-24).6 The total cost has been
periodically updated on the order of one or two times per year, but the cost breakdown has not
been reported since 2006. Proportionally, based on the 2006 filing, we can use the following
breakdowns to provide proxy values for variable, fixed and capital costs: in 2000, variable costs
were approximately 11% of total costs, fixed costs were approximately 22% of the total costs,
and capital costs were approximately 67% of total costs.
The latest TAPS Quality Bank tariff filing pegs total West Coast coker costs at $13.46
per barrel (ConocoPhillips, 2014, pp. 28). Under the ratios calculated above, the cost breakdown
is $1.48 per barrel for variable costs, $2.96 per barrel for fixed costs, and $9.02 per barrel for
capital costs in 2014.
Market dynamics are evident from Figure 4. Assume there is a demand shift away from
heavy oils to light oils in the product market: the schedule shifts from D1 to D2. Average
petroleum product prices might not change, but light product prices should rise relative to heavy
products. The price difference, ΔP1, rises to ΔP2. This incentive causes refiners to recycle heavy
oils through cracking equipment, etc. at increasing cost. If the differential goes above long-run
coker costs, refiners will have an incentive to increase coker capacity, i.e., to invest in new
facilities. They will do so in order to displace less efficient upgrading equipment. Such
investments allow refiners to reduce costs and increase profits, but they will not necessarily add
new overall refining capacity.
3.1 Empirical Evidence
If the coker model holds in practice, we can expect to see two phenomena: we should see
coker utilization increase whenever the price differential between heavy fuel oil and light crudes
is greater than the variable cost of coking; second, we could expect to see coking capacity
additions whenever the price differential is greater than the long-run coker cost for an extended
period of time. However, we should consider that the FERC coker costs are broadly generic and
will not necessarily reflect coker costs across the entire globe or across the entire industry.
Keeping this shortcoming in mind, we use the following data:
6
The definition of “fixed” costs, used by the FERC for regulatory purposes includes some costs that could be
avoided and, thus, ought to be categorized as marginal costs. For this analysis, however, we utilized the FERC
categories.
37th IAEE International Conference, New York, NY, June 15-18, 2014
9
•
•
•
•
A “heavy-light spread” price series, for which we can use either the Brentresidual fuel oil price spread data reported in Figure 3 (provided by the EIA), or
we can compare Los Angeles Ultra Low-Sulfur Diesel prices with Los Angeles
380CST High-Sulfur Fuel Oil prices. Both of the latter are provided by
Bloomberg.7 We can also use another light-heavy spread proxy: the difference
between the Brent crude oil price and the Mexican Maya crude oil price (the latter
is a benchmark heavy crude oil (Dixon, 2010, pp. 944)).
A figure for coker utilization. The EIA provides data on both coker capacity and
processing of fresh feed input for the US and PADD 5 (the West Coast) (EIA,
2014d and 2014e). Feed Input divided by capacity is capacity utilization. We
should see this figure increase whenever the heavy-light price differential is
greater than the variable cost of coking. We also use the raw coker feed input data
in thousand barrels per day.
A figure for coker capacity. In addition to the U.S. data provided by the EIA
(EIA, 2014d), the Oil & Gas Journal has compiled global coker capacity by year
since 1997 (OGJ, 1996-2013).
Figures for the variable and total cost of coking as calculated by the FERC for the
TAPS quality bank. These are provided on the website of ConocoPhillips
Transportation Alaska, Inc. (CPTAI, 2005-2014) and in the FERC filing of 2006
which retroactively calculated costs back to 2000 (FERC, 2006, pp. 26). The
monthly data extend from February 2000 through April of 2014.
There are a number of potentially confounding factors that will make it difficult to
decisively determine the impact of an increasing light-heavy price spread on coker utilization.
First, each refinery and each coking unit are unique; the FERC’s calculation was for a “typical”
coking unit at a West Coast refinery that processes ANS crude oil. That fact alone will make the
FERC’s coker costs unlikely to be representative of the global market. Second, refiners may
make their processing decisions well in advance of the actual month of production. The refinery
system is so complex that making ad-hoc changes in one process might affect others adversely,
and so negatively impact profitability over the entire refinery despite an improvement in the
economics of the coking unit. One might expect there to be some kind of lag in the price signal,
or a period of sustained price spreads, before the refinery responds. Another issue is that a
refinery may choose to coke its residuum into middle distillates regardless of short-term price
signals because of the need to maintain a minimum level of crude throughput in the refinery to
meet marketing obligations, and/or to keep other processing components of the refinery fully
utilized. A related issue is that, to ensure access to crude oils, an independent refinery may have
contracted a set quantity of crude oil for delivery in advance through forward contracts. If the
negative margin from coking excess residuum is less than the cost to breach the crude delivery
contract, the refinery will continue to operate its coker regardless of the price signal.
Additionally, a refiner may take a coker down for needed maintenance independently of price
signals. This would show up as a decrease in coker utilization even though the price signal had
nothing to do with the curtailment. Furthermore, capacity utilization may appear to decrease in
7
Bloomberg prices retrieved from Bloomberg Terminal on March 23, 2014. Price series are monthly. LA ULSD
Ticker code is DIEILAAM Index, Ticker code for LA 380 CST Fuel Oil is N6LA380 Index. All prices converted to
dollars per barrel using 42 gallons per barrel for diesel prices and 6.66 barrels per metric ton for FO 380CST
(conversion factor URL: http://www.eia.gov/cfapps/ipdbproject/docs/unitswithpetro.cfm, accessed April 4, 2014).
37th IAEE International Conference, New York, NY, June 15-18, 2014
10
the months immediately after a capacity expansion as the refiner gets the coking unit running up
to speed. These fluctuations likewise have nothing to do with responding to price signals.
Furthermore, the coking capacity data are annual as of January 1st each year, regardless of when
in the year the capacity was actually added; this makes identifying when an increase in coker
feed inputs is due to an increase in overall capacity vs. when it is due to price signals impossible.
All of these factors together make it highly unlikely that such a simple model will be able to
account for a very high proportion of the changes in coker throughput as a function of the lightheavy product (or crude) price spread.
3.1.1. The monthly model on coker variable costs
Though the above caveats should dampen expectations, we ran a number of formulations
of the model using monthly data on the light-heavy crude (or product) price spread, the FERC’s
proxy for coker variable costs, and the change in coker throughput. We modeled from zero to
three-month lags on the following formulation as a “trigger”: (∆𝑃 − 𝑀𝐶! ). The idea was that
trigger values above zero should produce positive changes in coker throughput, and values below
zero should produce negative changes. Using the trigger in levels did not result in very good fits
– perhaps because the trigger value was nearly always positive given the small variable cost for
cokers. Using the change in the trigger value ∆(∆𝑃 − 𝑀𝐶! ) was more promising, although
almost none of the monthly formulations were good fits. The highest R2 figure was less than
0.05, meaning that the price spread alone accounted for less than 5% of the total variation in
coker inputs. Our best result, which was statistically significant at the 2% level on 131 monthly
observations, was the 3-month lag of the ∆(∆𝑃 − 𝑀𝐶! ) trigger on the monthly change in coker
throughputs in PADD 5 (the US West Coast). The key model outputs are as follows:
Table 2: 3-Mo. Lagged Change in ΔP less Coker Variable Costs on Coker Throughputs
Intercept
ΔP-MCc, t-3
R Square
Adjusted R Square
# observations
Coefficients Standard Error
-1.189
2.533
1.749
0.73
LCL
-7.155
0.03
UCL
4.778
3.468
t Stat p-level H0 (2%) rejected?
-0.469
0.64 No
2.397
0.018 Yes
0.043
0.035
131
In short, a $1 increase in the “trigger” (the spread between Brent crude oil and the EIA’s
generic residual fuel oil prices, less the FERC’s variable cost proxy value) will increase the coker
throughput in PADD 5 by 1.7 thousand barrels per stream day 3 months later, on average. The
95% confidence interval of this effect is from 30 barrels per stream day to nearly 3.5 thousand
barrels per stream day. The coefficient has a p-value of 0.018. Most importantly, we see that the
results are directionally consistent with the theory. That we only found significant results in
PADD 5 makes sense, since we were using PADD 5 coker costs as a proxy for all coker variable
costs. Costs outside of PADD 5 are not likely to reflect the FERC proxy values.
3.1.2. The annual models on coker costs
For annual data, we were left with only 12-15 observations if we included the coker cost
proxies, which begin in 2000. Even without them, we would have a maximum of 24-27
observations (depending on the price series used to construct the light-heavy spread). In these
cases, we were never left with enough observations to reject the null hypothesis that the
37th IAEE International Conference, New York, NY, June 15-18, 2014
11
outcomes recorded could have come from random processes. While we did find a relatively
strong fit in the case of the Brent-Maya spread on the change in worldwide coker capacity (an R2
of 0.2), the results were not statistically significant due to a lack of observations. More data,
encompassing a wider range of years, would be required to test this model rigorously using
annual data.
3.2 Model conclusions
Although the results drawn from the model testing are tenuous, the data are at least
weakly directionally consistent with our logic. The logic in the model supports two principal
conclusions: 1) Typically refiners will always use their cokers, if available. If product demand
shifts towards heavy fuel oil, refiners will back down other less efficient processes, but still use
the coker. This is because processing residuum through a cracking unit, while technically
possible, will destroy the catalysts much more quickly. The result is a greater need for
maintenance and an increase in the time a cracker needs to be taken offline; 2) The demand for
coker investment depends on the balance of light and heavy crude oil production as well as
regulatory constraints on the use of heavy fuel oil.
4. Impact of new environmental regulations
High sulfur heavy fuel oil has been effectively phased out in developed countries. As
noted earlier, in 2010 developing countries consumed about 3.5 mmb/d and the bunker market
consumed 3.2 mmb/d of residual fuel oils (IEA, 2013). The bunker market is poised for big
changes by 2020 in a series of steps.
The International Maritime Organization (IMO) amended the International Convention
for the Prevention of Pollution from Ships (MARPOL) in 2010. The MARPOL agreement
designated specific portions of U.S., Canadian and French waters as Emission Control Areas
(ECAs) (EPA, 2010, pp. 1). The new regions join areas of the Baltic Sea and North Sea already
included in similar ECAs. Current regulations allow oil containing 1% sulfur or less to be
burned in ECAs. However, requirements get much stricter in 2015. According to the IEA:
“More importantly, in 2015 ECA‐wide sulphur limits are due to be further reduced to 0.1%. This
new, tighter ECA sulphur limit is expected to result in a sharp upturn in OECD marine gasoil
demand, at the expense of marine fuel oil.” (IEA, 2014, pp. 13).
Although this diminishes demand for high sulfur heavy fuel oil along European and
North American coasts it does not impact the bulk of the market. The primary vessels that use
high sulfur oil are large ships typically involved in international trade. High sulfur FO-380 of
less than 3.5% sulfur is sold to these ships in U.S. and European ports. Although it cannot be
burned in the ECAs, once outside an ECA ships are free to burn it. However, by 2020 the
MARPOL agreement will reduce sulfur content in all global bunker fuels to 0.5% (EPA 2010,
pp. 4). If these regulations come about, the demand for high sulfur fuel oil for bunkering will
disappear in favor of diesel grades, and total residual fuel oil demand will drop by nearly half if
current trends for non-bunkering uses persist.
The other large sector of use is in developing countries. As their economies improve
many are likely to adopt environmental regulations similar to those in the OECD. This is more
37th IAEE International Conference, New York, NY, June 15-18, 2014
12
likely in countries that have access to shale gas that is potentially cheaper and much cleaner. As a
boiler fuel, natural gas is preferable to residual fuel oils not only because of its favorable
emissions profile, but also because burning natural gas causes less damage to equipment than
burning residual fuel oil. However, it is possible that a glut of residual fuel oil will decrease the
prices sufficiently that low-income nations will increase their use of it as a boiler fuel.
5. Balance of heavy and light crude oil
There has been a general expectation that global crude oil production will shift towards a
greater proportion of heavy relative to light crude oils. The expectation is based on simple
economics: light crude oils are somewhat cheaper to produce and transport and, of course, they
require less sophisticated refining. This is reflected in higher prices for high-gravity low-sulfur
(light sweet) crude oils. Thus, when given a choice, producers will develop light before heavy
crude oil fields. The deposits of light crude oils should therefore be depleted more quickly than
the deposits of heavy crudes. Moreover, there are known abundant deposits of heavy oils –
notably Canadian tar sands and Venezuela’s Orinoco belt.
This expectation has been somewhat offset recently with the development of light shale
oil in the U.S.’s Bakken and Eagle Ford deposits. While shale oil is viable in the U.S. it remains
unclear as to whether the technology can be exported to other countries. Even if it is possible to
develop oil shale elsewhere it is likely to be in existing oil provinces, dominated by OPEC’s
members. They will certainly choose to develop lower cost conventional oil first. In any case,
the expectation of increased light crude oil in the U.S. is more than offset by existing plans for
Canadian oil sands and other heavy oils. Thus, it is unlikely that average refinery feedstocks will
become lighter, either in North America or the rest of the world.
6. Need for Refinery Investment
According to the Oil and Gas Journal’s annual refining survey, global coking capacity
increased by 1.0 mmb/d from 1997 through 2010 and now totals 4.7 mmb/d (OGJ, 1996-2013).
EIA and IEA data indicate that heavy fuel oil consumption declined by 2.2-2.4 mmb/d over the
same period (EIA, 2014 and IEA, 2013). There were, of course, other investments made in
cracking capacity that allow further bottoms upgrading. Nonetheless, these data underscore the
likely impact of further reductions in high-sulfur heavy fuel oil. If MARPOL regulations are
implemented as planned in 2020, approximately 1.5 mmb/d of coking capacity ought to be added
to accommodate the surplus of residuum.
Adding a coker to an existing refinery typically costs around $1 billion, for a capacity of
40 thousand b/d. This suggests a required investment of around $35 billion between now and
2020, although some of the capacity will likely be built in tandem with new refineries.
7. Conclusion
It is highly likely that the trend depicted in Figure 1 will continue with the resulting
impact on refiner configuration being an increase in coking capacity. The trends in various key
factors – residual fuel oil demand, environmental protection, and the “heaviness” of the crude
slate – all suggest that the most effective and least expensive way to dispose of residuum in the
37th IAEE International Conference, New York, NY, June 15-18, 2014
13
future will be through a substantial increase in global coking capacity. According to the logic of
the residuum upgrading model presented in section 3, the lowest-cost method of disposal of
residuum is likely to be the coker in the absence of a market for bunker fuel blending and
assuming current trends in asphalt demand.
However, it is still possible that developing countries that have not yet adopted strict
policies of environmental protection could increase their usage of residual fuel oils once the
bunker fuels market dries up. If refiners do not prepare for the 2020 phase-out of high-sulfur
residual fuel oils in advance, there could be a glut of the fuel, which would provoke a price
collapse. At low enough prices (and a high enough price differential), consumers may opt to use
residual fuel oil instead of cleaner alternatives such as diesels or natural gas. The environmental
benefits expected with the enforcement of the MARPOL 2020 limits may not materialize unless
similar constraints on land-based use of residual fuel oils are enacted worldwide.
8. References
ConocoPhillips Transportation Alaska, Inc. (CPTAI). (2005-2014). Pipeline Tariffs, TAPS.
URL: https://www.conocophillipsalaska.com/tariffs/conocophillips.asp, accessed April 3, 2014.
ConocoPhillips Transportation Alaska, Inc. (ConocoPhillips). (2014). Oil Pipeline Filing FERC
No. 21.6.0 (cancels FERC No. 21.5.0). January 27, 2014. URL:
https://www.conocophillipsalaska.com/tariffs/conocophillips/2014/Current%20CPTAI%20Quali
tyBank%20Tariff%20Filing%20at%20FERC%20%201-27-14.pdf, accessed April 3, 2014.
Dixon, Todd W. (2009). US Refining Economics – A Model Based Approach. Light Metals
2009. The Minerals, Metals, & Materials Society, 2009, pp. 941-944.
Fahim, Mohamed A., T.A. Alsahhaf, and A. Elkilani (2010). Fundamentals of Petroleum
Refining. First Edition. Elsevier, 2010.
Gary, James H., G.E. Handwerk, and M.J. Kaiser (2007). Petroleum Refining: Technology and
Economics, Fifth Edition. Taylor and Francis Group. Boca Raton, FL, USA.
Hawaiian Electric Companies, June 28, 2013, (HECO, 2013). 2013 Integrated Resource
Planning Report. Appendix I, Hawaii Electric Companies Fuels Master Plan, to 2013. Retrieved
from Hawaii Electric Co. website December 2013:
http://www.heco.com/vcmcontent/IntegratedResource/IRP/PDF/IRP-2013-Report-Filed.pdf.
Oil & Gas Journal (OGJ) (1996-2013). OGJ Survey Downloads. Worldwide Refineries Capacities as of January 1, Coking. Pennwell. URL: http://www.ogj.com/ogj-surveydownloads.html, accessed 4/3/14.
Organization for Economic Cooperation and Development, International Energy Agency, IEA
(2013), "Oil demand by product for non-OECD countries", IEA World Energy Statistics and
Balances (database). doi: 10.1787/data-00511-en, URL:
http://stats.oecd.org/BrandedView.aspx?oecd_bv_id=enestats-data-en&doi=data-00511-en#,
accessed April 2, 2014.
37th IAEE International Conference, New York, NY, June 15-18, 2014
14
Organization for Economic Cooperation and Development, International Energy Agency (IEA)
(2014). Oil Market Report. February 13, 2014. URL:
http://omrpublic.iea.org/omrarchive/13feb2014fullpub.pdf, accessed March 18, 2014.
Sadhukhan, J., N. Zhang, and X.X. Zhu (2003). Value Analysis of Complex Systems and
Industrial Application to Refineries. Industrial and Engineering Chemistry Research, 2003,
42:5165-5181.
U.S. Department of Energy, Energy Information Administation (EIA). (2014). International
Energy Statistics, (Thousand Barrels Per Day), URL for Residual Fuel Oil Consumption:
http://www.eia.gov/cfapps/ipdbproject/iedindex3.cfm?tid=5&pid=66&aid=2&cid=regions&syid
=1980&eyid=2012&unit=TBPD, accessed 3/28/14.
URL for Residual Fuel Oil Consumption for Bunkering:
http://www.eia.gov/cfapps/ipdbproject/iedindex3.cfm?tid=5&pid=66&aid=13&cid=regions&syi
d=1984&eyid=2010&unit=TBPD, accessed 3/28/14.
URL for Total Petroleum Consumption:
http://www.eia.gov/cfapps/ipdbproject/iedindex3.cfm?tid=5&pid=5&aid=2&cid=ww,r1,r3,CG6,
CG5,CH,JA,&syid=1980&eyid=2012&unit=TBPD, accessed 3/31/14.
U.S. Department of Energy, Energy Information Administation (EIA). (2014b). Spot Prices for
Crude Oil and Petroleum Products, Europe Brent Spot Price FOB (Dollars Per Barrel), file
pet_pri_spt_s1_m.xls, URL: http://www.eia.gov/dnav/pet/pet_pri_spt_s1_m.htm, accessed
4/1/14.
U.S. Department of Energy, Energy Information Administation (EIA). (2014c). U.S. Residual
Fuel Oil Prices by Sales Type, U.S. Residual Fuel Oil Sulfur Greater Than 1% Wholesale/Resale
Price by All Sellers (Dollars Per Gallon), file pet_pri_resid_dcu_nus_m.xls, URL:
http://www.eia.gov/dnav/pet/pet_pri_resid_dcu_nus_m.htm, accessed 4/1/14.
U.S. Department of Energy, Energy Information Administation (EIA). (2014d). U.S. Number
and Capacity of Petroleum Refineries, U.S. Refinery Thermal Cracking, Coking Downstream
Charge Capacity as of January 1 (Barrels per Stream Day), file pet_pnp_cap1_dcu_nus_a.xls,
URL: http://www.eia.gov/dnav/pet/pet_pnp_cap1_dcu_nus_a.htm, accessed 4/4/14, and PAD
District 5 Number and Capacity of Petroleum Refineries, West Coast (PADD 5) Refinery
Thermal Cracking, Coking Downstream Charge Capacity as of January 1 (Barrels per Stream
Day), file pet_pnp_cap1_dcu_r50_a.xls, URL:
http://www.eia.gov/dnav/pet/pet_pnp_cap1_dcu_r50_a.htm, accessed 4/4/14.
U.S. Department of Energy, Energy Information Administation (EIA). (2014e). U.S.
Downstream Processing of Fresh Feed Input, U.S. Downstream Processing of Fresh Feed Input
by Delayed and Fluid Coking Units (Thousand Barrels per Day), file
pet_pnp_dwns_dc_nus_mbblpd_m.xls, URL:
http://www.eia.gov/dnav/pet/pet_pnp_dwns_dc_nus_mbblpd_m.htm, accessed 4/4/14, and
PADD 5 Downstream Processing of Fresh Feed Input, West Coast (PADD 5) Downstream
Processing of Fresh Feed Input by Delayed and Fluid Coking Units (Thousand Barrels per Day),
37th IAEE International Conference, New York, NY, June 15-18, 2014
15
file pet_pnp_dwns_dc_r50_mbblpd_m.xls, URL:
http://www.eia.gov/dnav/pet/pet_pnp_dwns_dc_r50_mbblpd_m.htm, accessed 4/4/14.
U.S. Environmental Protection Agency (EPA). (2010). Designation of Emission Control Area to
Reduce Emissions from Ships in the U.S. Caribbean. March, 2010. URL:
http://www.epa.gov/otaq/regs/nonroad/marine/ci/420f10015.pdf from page
http://www.epa.gov/otaq/oceanvessels.htm, accessed April 8, 2014.
U.S. Federal Energy Regulatory Commission (FERC). 2005. “Commission affirms initial
decision on TAPS quality bank, but limits scope of retroactive refunds for resid”. News Release,
October 20, 2005. Docket Number OR89-2, et al. URL: https://www.ferc.gov/media/newsreleases/2005/2005-4/10-20-05-G-6.asp, accessed April 3, 2014.
U.S. Federal Energy Regulatory Commission (FERC). 2006. BP Pipelines (Alaska), Inc. et al
submits FERC Oil Tariff 33 et al under IS06-466. Docket No. IS06-466. Accession No.
200600710-0020, filed July 3, 2006. URL:
http://elibrary.ferc.gov/idmws/file_list.asp?document_id=4420402, accessed April 1, 2014.
37th IAEE International Conference, New York, NY, June 15-18, 2014
16