-
Keywords:review of SAGD processpitfalls in numerical and
laboratory modelseffective parametersoperational problemsfuture of
SAGD
. . . .
Journal of Petroleum Science and Engineering 68 (2009)
135150
Contents lists available at ScienceDirect
Journal of Petroleum Science and Engineering4. Mechanics of
SAGD. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . 1374.1. Steam
chamber rise . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . 1374.2.
Steam ngering theory. . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1374.3.
Co-current and counter-current displacement . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1374.4. Emulsication . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. 1384.5. Residual oil saturation in steam chamber . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . 1384.6. Heat transfer and distribution through steam chamber
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . 1384.7. Analytical models . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . 139
5. Effects of reservoir properties on SAGD performance . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . 1395.1. Porosity. . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . 1395.2. Thickness . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . 139
5.3. Gas saturation. . . . . . . . . .5.4. Permeability . . . .
. . . . . .5.5. Viscosity and API . . . . . . . .5.6. Wettability .
. . . . . . . . . .
Corresponding author.E-mail address: [email protected] (T.
Babadagli).
0920-4105/$ see front matter 2009 Elsevier B.V.
Adoi:10.1016/j.petrol.2009.06.011. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
136
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . 136
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . 136
2. Background on SAGD process . . . . . .3. Mechanism pitfalls .
. . . . . . . . . .Contents
1. Introduction . . . . . . . . . .accompanying it.Two major
techniques, namely thermal and miscible, have been considered in
HO-B development, along withseveral other auxiliary methods
(chemical, gas, electromagnetic heating, etc.) for different well
congura-tions, with steam assisted gravity drainage (SAGD) being
the most popular. Miscible techniques are nothighly recognized as a
commercial option, while thermal techniques have by far a more
stable foundation inthe industry.Despite a remarkable amount of
laboratory experiments and computational studies on thermal
techniquesfor HO-B, specically SAGD, there was no extensive and
critical literature review of the knowledge gainedover almost three
decades. We believe that this kind of review paper on the status of
the SAGD process willshed light on the critical aspects,
challenges, deciencies and limitations of the process. This will
open doorsto further development areas, and new research
topics.This paper focuses mainly on laboratory and numerical
simulation studies, not eld experiences. The attemptis to draw a
picture of the developments on the physics and technical aspects of
the process and its futureneeds. Specic attention, was given to (a)
the effect of geological environment on the physics of the
process,(b) evaluation of the laboratory scale procedure and
results, (c) problems faced in numerical modelling(capturing the
physics of the process, relative permeability curves, dynamics of
gravity controlled counter-current ow), and (d) operational and
technical challenges.
2009 Elsevier B.V. All rights reserved.Yet the in-situ recovery
of HO-B is still not a simple process and there are many technical
challengesReceived 9 July 2008Accepted 7 June 2009
a tremendous demand on tReview
SAGD laboratory experimental and numerical simulation studies: A
review of currentstatus and future issues
Al-Muatasim Al-Bahlani, Tayfun Babadagli University of Alberta,
Department of Civil and Environmental Engineering, School of Mining
and Petroleum, 3-112 Markin CNRL-NREF, Edmonton, AB, Canada T6G
2W2
a b s t r a c ta r t i c l e i n f o
Article history: With around 7 trillion-barrel reserves and
recent increases in oil demand, there is no doubt that there will
behe development of heavy oil/bitumen (HO-B) reservoirs in the
coming decades.
j ourna l homepage: www.e lsev ie r.com/ locate /pet ro l. . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . 139
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . 140
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . 141
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . 141
ll rights reserved.
-
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . 141
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . 142
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . 142
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
expansion of SAGD methods to world wide applications. To
achieve
136 A.-M. Al-Bahlani, T. Babadagli / Journal of Petroleum
Science and Engineering 68 (2009) 135150this, we believe that there
is a need to compile and analyze thepublished data about SAGD
studies at different scales and for differentpurposes. This
analysis will provide not only clarications foruncertainties faced
so far, but also an extensive summary of unclearpoints that
eventually lead to dening further research areas.
2. Background on SAGD process
SAGD is an abbreviation for steam assisted gravity drainage. It
wasrst developed by Roger Butler and his colleagues in Imperial Oil
inthe late 1970s. Its main characteristic is introducing steam
intoreservoirs and producing heated oil using two horizontal wells.
Butlerdescribed the technique as when steam is injected, a steam
saturatedzone, called a steam chamber is formed, in which the
temperature isessentially that of the injected steam. The steam ows
towards theperimeter of the steam chamber and condenses. The heat
from thesteam is transferred by thermal conduction into the
surroundingreservoir. The steam condensate and heated oil ow by
gravity to theproduction well located below. As the oil ows away
and is produced,
He then raised some critical issues to be considered in the
projectdevelopment and eld performance assessment:
1. condensate ow: with so much condensate owing, convectionwould
be expected to be the dominant heat transfer mechanism,
2. geology: geology of the formation can have a profound inuence
onsteam chamber growth (sideway in underground testing
facility(UTF) case),
3. 2D vs. 3D models: two important missing factors are ow in
thetwo horizontal wells, and the effect of wellbore, when the wells
aredrilled from surface rather than from tunnels (refereeing to UTF
Devor project),
4. geomechanical effects: the effects of SAGD on reservoir
geome-chanics is not well understood.
Singhal et al. (1998) mentioned1 that Farouq-Ali pointed
outpotential problems and limitations of SAGD such as: (1) sand
control,
1 Singhal et al. (1998) refer these comments to a talk about
SAGD given by Farouq Aliin Calgary Jun 17 1998. Unfortunately we
could not obtain any recorded material of this5.7. Heterogeneity .
. . . . . . . . . . . . . . . . . . . . . .5.8. SAGD in carbonate
reservoir . . . . . . . . . . . . . . . .5.9. SAGD geomechanics . .
. . . . . . . . . . . . . . . . . .
6. SAGD operation . . . . . . . . . . . . . . . . . . . . . . .
. .6.1. The start-up procedure . . . . . . . . . . . . . . . . . .
.
7. Steam quality . . . . . . . . . . . . . . . . . . . . . . . .
. . .8. Length, spacing and placement of horizontal wells . . . . .
. . . .9. Subcool temperature (steam trap control) . . . . . . . .
. . . . .
9.1. HP (high pressure) vs. LP (low pressure) SAGD . . . . . .
.9.2. Steam chamber monitor and volume size estimation . . . .
.
10. Numerical simulation . . . . . . . . . . . . . . . . . . . .
. . .11. Experimental pitfalls . . . . . . . . . . . . . . . . . .
. . . . .12. SAGD improvement . . . . . . . . . . . . . . . . . . .
. . . . .
12.1. Geometrical attempts . . . . . . . . . . . . . . . . . .
.12.2. Chemical attempts . . . . . . . . . . . . . . . . . . . .
.
13. Summary . . . . . . . . . . . . . . . . . . . . . . . . . .
. .Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . .
. . .References . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . .
1. Introduction
Reservoir heating is essential in heavy/ultra heavy oil and
bitumen(HO-B) recovery. Steam injection is a proven thermal
technique to beused for this purpose and it can be achieved through
continuous orcyclic (huff-and-puff) injections. Field experience
and simulationsstudies show that performing these techniques are
associated withtechnical difculties and usually low recovery
factors. The steamassisted gravity drainage (SAGD) method was
proposed by Butlermore than 30 years ago (Butler, 1994b, 1998,
2004a). Due to increasedcontact area through two horizontal wells,
the process was believed tobe successful from a technical point of
view although economicalstandpoints are still sceptical. Over a
thirty-year period, this techniquehas been tested successfully,
which led many to think of it as astandard technique in HO-B
recovery. Obviously, it has some technicaland physical restrictions
which will be discussed in this paper.
Alternatives to this technique have been proposed for
unsuitablereservoirs. Those techniques include miscible ooding
(VAPEX) ormodied versions of SAGD through different congurations of
wells orusing additives to steam.
Due to its suitability for unconsolidated reservoirs that
displayhigh vertical permeability, the SAGD technique has received
attentionin countries with huge HO-B sand reserves like Canada and
Venezuela.Although it is a highly promising technique, many
uncertainties andunanswered questions still exist and they should
be claried forthe steam chamber expands both upwards and sideways
(Butler,. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . 142
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . 142
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . 143
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . 143
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . 144
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . 144
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . 145
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . 145
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . 146
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . 146
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . 146
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . 146
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . 147
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . 148
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . 148
1994b). Two types of ow then exist during this process: One at
theceiling of the steam chamber (ceiling drainage; bitumen is
pulledaway from the front immediately after mobilization where
steam riseusually impedes liquid drainage) and the other one along
the slopes ofthe steam chamber (slope drainage; gravity holds
mobilized bitumenagainst the slope where bitumen mobility is
controlled by conductionheating from the steam zone) (Edmunds et
al., 1989; Edmunds et al.,1994; Nasr et al., 2000). Fig. 1
illustrates the SAGD mechanism.
Das (2005a) stated that most of commercial SAGD wells
areexpected to produce in the range of 6001500 m3/day of the
totaluid under normal operating conditions after the initial ramp
upperiod. The corresponding injector well should have the capacity
of4001200 m3/day cold water equivalent (CWE) of steam
injection.
3. Mechanism pitfalls
Although the SAGD process looks simple at rst sight,
severalauthors pointed out some pitfalls/concerns on the theories
of themechanism. For example, Farouq-Ali (1997) stated that:
1. the theory pertains to the ow of single uid,2. steam pressure
is constant in the steam chamber,3. only steam ows in the steam
chamber with oil saturation being
residual, and4. heat transfer ahead of the steam chamber to cold
oil is by
conduction only.talk.
-
the
137A.-M. Al-Bahlani, T. Babadagli / Journal of Petroleum Science
and Engineering 68 (2009) 135150(2) hot efuent/high water-cut
production, (3) frequent changes inoperating regime (making
management of SAGD projects labourintensive), (4) deterioration of
production at late stages, and (5) highoperating costs. Butler and
Yee (2002) stated that despite itsattractiveness, SAGD tends to be
wasteful of steam because the entirepart of the reservoir that is
depleted becomes heated to steamtemperature, whereas to avoid steam
coning, only the reservoir nearthe production well has to be
heated. Heating the upper part of thereservoir to steam temperature
is undesirable because of the resultinghigh loss of heat to the
overburden as well as the high heat require-ments for the chamber.
The loss to the overburden is also increased bythe tendency of the
steam chamber to creep laterally beneath thereservoir cap (Butler,
1997).
Deng (2005) outlined the disadvantages of SAGD as: (1)
intensiveenergy input and excessive CO2 emission, and (2) costly
post-production water treatment.
After reviewing published eld data, McCormack (2001)
listedseveral problems faced in the eld applications: (1) lower
than ex-pected drainage rates from average to poor sands; (2)
difculty withinstallation of liners into the horizontal section of
the well; (3) sandproduction, wellbore scaling, and (4) uid removal
limitations.
Based on these evaluations, one can divide SAGD challenges
intomicro- and macro-scale. Some of these challenges are not
inherent toSAGD; they can also be applied to other steam injection
techniques,such as geological effects, upscaling (2D/3D models),
high SOR, andoverburden heat losses. However, such challenges have
a relativelyprofound effect on SAGD; they need more micro/lab-scale
studies tofurther understand the physics of the process and to use
them forbetter industrial applications.
4. Mechanics of SAGD
4.1. Steam chamber rise
Fig. 1. Illustration ofThe SAGD concept is based on steam
chamber development, asproduction is mainly from the
chamber/heated-oil interface. Thus, thedevelopment and analysis of
the steam chamber growth has receiveda great deal of attention by
scientists studying SAGD. Yet, it seemsthat the complete picture of
the steam chamber development processis not fully represented due
to different processes occurring at thesame time; namely,
counter-current ow, co-current ow, water im-bibition, emulsication,
steam ngering and dimensional movement(lateral vs. vertical).
In other words, the complexity is due to the fact that a lighter
uid(steam) is trying to penetrate into a heavier uid (HO-B) above
it. Itoand Ipek (2005) observed from eld data that the steam
chambergrew upwards and outwards simultaneously like the expansion
ofdough during baking. The recent understanding of the SAGD
processendorses the idea that steam chamber is not connected to the
pro-ducer; rather a pool of liquid exists above the production
well. Gateset al. (2005) identied the advantage of having such a
pool bypreventing the ow of injected steam into the production
well. Thus,it is of primary importance to clarify the relative
effects of eachparameter affecting the movement and the shape of
the steamchamber.
4.2. Steam ngering theory
From a sand pack lab experiment, Butler (1994b) observed that
therise of the steam chamber does not advance as a at front, rather
as aseries of separate and ragged ngers. He referred to the
occurrence ofthese ngers as being due to instability created by
rising lighter steambelow the heavy oil. Thus, understanding steam
nger theory iscrucial to understanding steam chamber rise
processes. Depicting arectangular boundary, Butler (1987)
hypothesized steam chamberdevelopment as follows:
Steam ows upward from the lower boundary providing heat torise
the reservoir temperature to steam temperature.
Heatedmaterial drains through the lower boundary as a number
ofstreams.
The velocity at which the residual oil leaves the system is that
ofthe steam chamber rise.
The entering steammoves at a higher velocity than the chamber
inorder to pass through the lower boundary.
At the very top of the chamber steam ngers move into
therelatively cold reservoir and heat the cold oil through
conduction.
According to Ito and Ipek (2005), many observations in the
UTFPhase A and B, Hanginstone and Surmount projects are now
clearlyunderstood through the steam ngers concept. They
hypothesizedthat the specic nature of steam ngering phenomena
during SAGDoperation may cause steam chamber deviation from usual
behaviour(stop and resume, shrinkage or even disappearance). Sasaki
et al.(2001) provided images where steam ngering can clearly be
seen on
SAGD mechanism.their 2D experimental model. They also showed an
increase in theceiling instability, hence ngering, due to
intermittent steam stimu-lation of the lower horizontal
producer.
These observations imply that it is very important to
considersteam ngering as a method of steam movement inside the
reservoir.Steam ngering occurs in a vertical manner whereas a
buoancyadvantage of steam to oilmainly drives the process. However,
chambergrowth disturbance does not only occur vertically and this
triggers aquestion: can steam ngering occur laterally? This
question may bebacked up with geomechanical investigation results
of increasingwater relative permeability ahead of the steam
chamber.
4.3. Co-current and counter-current displacement
Nasr et al. (2000) stated that the uniqueness of the SAGD
recoveryprocess lies in the salient role of moving condensing
boundaries andcounter-current ows. Counter-current owbetween heated
heavy oil
-
138 A.-M. Al-Bahlani, T. Babadagli / Journal of Petroleum
Science and Engineering 68 (2009) 135150and bitumen occurs at the
top of a rising steam chamber where steamngers rise and heated
heavy oil falls (Chung and Butler, 1987; Butler,1987, 1994a). Nasr
et al. (2000) published a paper highlighting steamoil emulsion
counter-current ow and the rate of propagation of thesteam chamber.
They used a cylindrical experimental model andadiabatic control
system, and a numerical model to simulate SAGDcounter-current ow
and to determine the sensitivities of differentparameters. They
concluded that for a givenpermeability, the counter-current steam
front propagation rate is a linear function of time. Theyobserved
that the time taken for counter-current steam front topropagate to
a specic distance is muchmore than time taken by a co-current
front, where drainage condensate was impeding the advanceof the
counter-current front. After a history matching of the steam-water
counter-current and co-current relative permeability curves,they
found that there was a signicant difference between counter-current
and co-current relative permeabilities. They argued that thismay be
the result of a coupled ow between the phases. Hence, thefollowing
question arises: can the existing formulation of
numericalsimulation capture this important character? The steam
interfaceadvancement seems to be due to a combination of both
co-current andcounter-current ow, which shows the importance of a
clear pathpresence for clean sand with no bufes, as well as how
reservoirheterogeneity would have a profound impact on such
processes.Another issue we question is whether or not
counter-current move-ment also happens within a pore. When steam
rises up inside thepore, preferentially in the middle portion, does
the oil drain downthrough the sides closer to the grain due to
wettability and connatewater issues? Or does the process take place
in a convective manner?What is the effect of injection
pressure/rate on this process?
4.4. Emulsication
Chung and Butler (1987) stated that the production of
in-situthermal recovery of heavy oils always occurs in the form of
water in oilemulsion. This is much more viscous than the oil
itself. Theyconducted a laboratory study to elucidate the
geometrical effect ofsteam injection on the water/oil emulsion of
the produced uid froma SAGD process. They performed their
experiment in two schemes.The rst scheme consisted of a steam
injector slightly above aproducer at the base of the formation. The
second scheme consisted ofa producer at the base of the formation
and a vertical circulating steaminjector perforated near the top of
the formation. They concluded thatmuch higher water/oil emulsion
content was found in the produceduid when the steam chamber was
rising in the experiment withbottom steam injection than with
injection at the top. The rate ofrecovery was higher in the
operation with top injection. This isprobably due to the fact that
an increase in water/oil emulsion ratioincreases the uid viscosity,
hence a reduction in oil production isexpected. However, they also
noticed that when the steam chamberspreads sideways, a two phase
stratied ow of steam and heatedheavy oil occurs at which steam ows
sideways to the interface, andheated heavy oil ows down, below and
along the interface,which dramatically reduces the water/oil
emulsion ratio (Chungand Butler, 1987). They later extended their
work to include otherfactors which may affect water/oil
emulsication ratio such as initialconnate water (0% and 12.5%),
steam quality, and pressure variation(153 kPa3.55 MPa) (Chung and
Butler, 1989).
For initial connate water, they noticed a higher water/oil
ratioemulsion when Swi=0% than Swi=12.5%. They commented thatthere
is less tendency for water to condense as droplets on the surfaceof
oil when enough connatewater is available. As the droplets of
watercondense on oil, they become buried because of the
spreadingcharacteristics of oil. It is worth mentioning that Sasaki
et al. (2002)observed this process in a microscopic visualization
experiment.For steam quality effect, Chung and Butler (1989)
noticed no major
difference in injecting steam wet or dry. They argued that this
isbecause the interfacial activity at the steam front and the
heatingmechanism of the bitumen are the same for both cases. They
observedno major difference as a result of pressure variation. This
also appliesto the effect of particle size in the porous matrix
where no signicantvariationwas found. Sasaki et al. (2001)
visualizedwater/oil emulsionat the boundary of the steam chamber in
a 2D laboratory scaledmodel. They noticed a 25% uctuation in the
ratio after steambreakthrough and chamber rise. Understanding the
water/oil emul-sion is crucial not only from a reservoir
engineering point of view, butalso from a production technology
perspective as well.
4.5. Residual oil saturation in steam chamber
Butler (1994a) observed that major oil ow happened on thechamber
sideways rather than through it. He explained this observa-tion by
two hypotheses; (1) residual oil saturation is too low insidethe
steam chamber to allow for any oil movement, and (2) due towater
condensate between steam and oil, water imbibition andinterfacial
tension support the oil to drain laterally.
Walls et al. (2003) studied residual oil saturation in
steamchambers using a numerical model. Their work consisted of
twomain parts: (1) sensitivity tests done on the shapes and the end
pointsof the two phase-relative permeability curves, and (2) krog
relativepermeability curve adjustment to match theoretically
determinedresidual oil saturation. They concluded that water
relative perme-ability and oil relative permeability in the gasoil
system are the mainfactors that determine the magnitude and the
shape of the oil satura-tion curve as a function of time. They also
concluded that residual oilsaturation increases at lower SAGD
operating pressures. Many of thenumerical simulationmodels reported
failed to show their applicationof changes in relative permeability
curves due to temperature change.This will be discussed further in
the Simulation section.
Pooladi-Darvish andMattar (2002) stated that some of the
reasonsfor larger residual oil saturation are that steam at higher
pressures hasless latent heat, more heat will leave the reservoir
through theproduced uids at higher temperatures, and more heat will
be left inthe steam chamberwhere oil is no longer present.What is
critical hereis the amount of oil left inside the steam chamber. As
the steam/oilinterface passes through the reservoir, production
occurs mainly byco/counter-current. As the interface progresses,
the reduction ofresidual oil saturation is mainly due to the
steamoil gravity dif-ference, which is a very slow process.
Eventually, residual oil satura-tionwill become nil, but the
question is, how long will it take to reachthat point?
4.6. Heat transfer and distribution through steam chamber
Understanding the heat transfer through the steam chamber is
alsocritical. As mentioned earlier, Farouq-Ali (1997) criticized
theassumption that only thermal conduction exists in SAGD. In
responseto that critique, Edmunds (1999) stated that based on the
associatedchange in enthalpy, the liquid water could carry and
deposit 18% of theheat of condensation of the same water.
Convection due to oil isaround 1/5 of this and conduction to carry
the remaining 78%. He thenevaluated the convection role due to
water streamline being almostparallel to isotherms of less than 5%.
This was also emphasized byEdmunds (1999) who stated that except
for the very near vicinity ofthe liner or anywhere live steam
penetrates, heat transfer in themobile zone is dominated by
conduction, not convection.
Gates et al. (2005) provided images of steam quality
andtemperature of the steam chamber from a simulation study.
Compar-ing these pictures, one can clearly see that temperature is
almostconstant while steam quality varies signicantly. This
supportsthe claims of varying steam pressure throughout the steam
chamber,i.e., that steam chamber pressure is not constant. In their
work, they
provided a novel method for visualizing heat transfer within
the
-
boundaries of the steam chamber. They stated that the usefulness
ofthis method is that steam quality proles provide the means
toexamine convective heat transfer in the reservoir.
Using a hypothetical example of hotwell analysis, Butler
(1987)provides a heat distribution table for a typical Athabasca
SAGD project.We reproduced it into a pie chart as shown in Fig. 2.
Butler commentson the outcome by stating that in general the heat
remaining withinthe steam chamber, per unit production of oil, will
be lower if thesteam temperature is lower (i.e., the chamber
pressure is lower) or ifthe oil saturation is higher (i.e., there
is less reservoir to be heated per
139A.-M. Al-Bahlani, T. Babadagli / Journal of Petroleum Science
and Engineering 68 (2009) 135150m3 of oil). The latter is very
important in determining the performanceof a reservoir; high
(initial) oil saturation is always desirable.
Yee and Stroich (2004) showed that, after 5 years of Dover
projectphase B, the amount of injected heat in the chamber was
32.2%, andoutside of the chamber, 34.7%. The rest (33.1%) was
reproduced. It canbe seen that almost one third of heat injected is
reproduced. Webelieve that this may have a benecial effect as heat
may prevent waxdeposition inside the tubing andmaymaintain a lower
oil viscosity foruplifting if no emulsion is created.
4.7. Analytical models
Butler (1987) developed an approximate expression to predict
thesteam chamber rising rate and the dimensions of the steam
ngers.He concluded that the rise rates are proportional to
reservoir per-meability and are strong functions of steam
temperature and oilviscosity. Later work was done by Edmunds et al.
(1989). They cri-ticized the dependency of Butler's (1987) approach
on certaingeometric simplications which restrict the generalization
of themodel. They continued their numerical analysis based on the
UTFPhase A test. They presented a good analysis of 1-D ceiling
drainageand its relation to SAGD cases.
Butler (1994a) described the result of their work as a
prediction ofthe drainage rate using Darcy's equationwith
counter-current ow andthe incorporation of relative permeability
effects. However, he criticizedEdmunds et al.'s (1989) work in
terms of geometrical and theoreticalaspects. Fromone of
Butler'sworks (1997)we can identify the evolutionof Butler's theory
in three stages: (1) the original model, (2) theTanDrain &
LinDrainmodel, and (3) the steam risingmodel. Despite theextensive
theoretical input to evolve these equations, all
modicationspresentedare concernedwith the shape, height and
growthof the steamchamber. To bemore specic, the original Butler
theory concentrated onobtaining a relationship between the drainage
rate and the drainageheight independent of interface shape or its
horizontal extension. Later,the dependency on the shape of the
interface and boundaries weretaken into account. Butler then
provided a guideline on how toanalytically calculate oil production
through the following set ofequation which we summarized on the ow
chart given in Fig. 3.
As seen in Fig. 3, the equations tend to have quite a number
ofsimplications, some of which mentioned by Farouq-Ali
(1997).However, a very important feature can be drawn from Butler's
theoryevolution: the huge dependency of analytical models to steam
cham-ber growth and shape. These factors as we have seen earlier
areheavily dependent on reservoir characteristics, especially in
hetero-geneous reservoirs. Butler's formulation also includes the
effect of
Fig. 2. Reproduction of Butler's (2001) heat distribution for a
typical Athabasca oil SAGD
project.reservoir properties such as porosity, thickness and
initial oilsaturation.
Chen et al. (2007) also commented on Butler's theory where
theyaddressed that the theory is based on simplifying assumptions,
suchas that the steam chamber pressure remains at the original
reservoirpressure and the chamber must remain connected to the
producer.
Birrell (2001), on the other hand, stated that Butler's
equationwasshown to accurately predict the performance of SAGD in
the eld.
Reis (1992, 1993) stated that the limitation to the Butler's
model isits complexity; it requires an iterative solution to a set
of equations tocalculate the production rate. He provided linear
and geometricalmodels for oil production where he introduced a
dimensionless tem-perature coefcient to the denominator. He showed
that Butler'smodel overpredicts oil production compared to his
model. He alsoprovided energy balance and steam oil ratio
equations. However, hismodel does not predict production during the
rise of the steamchamber.
5. Effects of reservoir properties on SAGD performance
5.1. Porosity
Few studies were presented to show the effect of porosity on
SAGDperformance. By reviewing the analytical models provided,
however,one can observe that they all have cumulative production
and dailyproduction proportional to porosity whichmeans that higher
porositywould analytically promote SAGD performance. This was
observedin the analytical study by Llaguno et al. (2002) where they
reportedthat accumulation properties (thickness, porosity and oil
saturation)have a greater effect on SAGD performance than ow
properties(permeability, viscosity, API, and reservoir pressure).
The micro-scalepore structure could be considered as a critical
issue if the assumptionof counter-current displacement occurs
within in a single pore,(i.e., steam rises through the center of
the pore while heated oil drainsthrough its edge closer to the
grain), is valid. In this case, one has tounderstand what impact
the pore characterstics (shape, pore andthroat size) have on the
counter-current gravity drainage process.
5.2. Thickness
Several studies report that an increase in oil production
wasnoticed with an increase in oil pay thickness (Sasaki et al.,
2001; Chanet al., 1997; Shin and Polikar, 2007; Singhal et al.,
1998; Edmunds andChhina, 2001; McCormack, 2001). Edmunds and Chhina
(2001) statedthat zones less than 15 m thick are unlikely to be
economic. Most ofthe work done to draw this conclusion is based on
the fact that thinreservoirs increase thermal losses resulting in
higher SOR. However,this conclusion is subject to a variable
understanding of what is thickand what is thin. Also, the steam
chamber growth behaviour dueto other geological parameters may have
an effect on such con-clusions. For example, cupcake like steam
chamber growth (laterallyand sideways) would not see much effect of
reservoir thickness, whilehand fan like steam chamber growth
(laterally then sideways) in athick reservoir might take much more
time for the steam chamberto grow. Other complicated process such
as steam ngering, emul-sication, and prevailing counter-current ow
may result in theuctuation/decrease of oil production. Thus, there
may be an ultimatethickness for each reservoir which may dictate
where the injector isplaced from the top of the reservoir. This is
governed by thesteamchamber growth behaviour which is in turn
governed by otherreservoir charecterstics such as kv/kh.
5.3. Gas saturation
Nasr et al. (2000) studied the effect of initial methane
saturation
on the advancement of a steam front in an experimental sand
packed
-
140 A.-M. Al-Bahlani, T. Babadagli / Journal of Petroleum
Science and Engineering 68 (2009) 135150model. With the presence of
initial methane saturation, they noticedthe faster movement of
steam front at present temperature valueson a given time. However,
as the steam front entered the region ofmethane saturation, the
propagation rate declined as the methanemole fraction increased in
the gas phase. Canbolat et al. (2002)observed that the initial
presence of n-butane had a positive effect onthe process. They
explained this by the reduction of oil viscosity due togas
presence. Bharatha et al. (2005) conducted a study on dissolvedgas
in SAGD by means of theory and simulation. They stated in
theirconclusion that the effect of dissolved gas on SAGD is to
reduce thebitumen production rate. They also showed that operating
pressureplays a greater role in reducing the effect of dissolved
gas saturationpresence.
5.4. Permeability
McLennan et al. (2006) stated that the predicted ow
performanceof SAGD well pairs is sensitive to the spatial
distribution of per-
Fig. 3. Flow chart of Butler's analytical model for
calmeability. After experimental (sand packed core) and
numericalmodel investigations, Nasr et al. (2000) noted that the
effect of liquidconvection ahead of the steam front can provide
better heating for the10 Darcy permeability case than for the 5
Darcy case. They also ob-served that there was evidence that steam
temperature inside 5 Darcysand was lowered by about 3 C than that
for the 10 Darcy sand for agiven steam injection temperature. They
argued that this might be aresult of higher capillary pressure for
the 5 Darcy case. They alsoreported that the propagation rate of
the steam front is not a linearfunction of permeability.
In a 2D simulation model investigating SAGD in a
carbonatereservoir, Das (2007) reported no signicant change in
productiondue to matrix permeability at earlier stages and faster
decline for lowpermeability at later stages. He referred this to
the possibility ofmatrix production which occurs primarily by
imbibition and thermalexpansion. However, by looking at the
examination range (1050 mD)it can be seen that the range is too
small to study the effect ofpermeability. Kisman and Yeung (1995),
on the other hand, found
culating oil production from a SAGD operation.
-
141A.-M. Al-Bahlani, T. Babadagli / Journal of Petroleum Science
and Engineering 68 (2009) 135150from a simulation model that
decreasing the vertical permeabilityresulted in a signicant
decrease in CDOR (calendar day oil rate) andOSR initially. But an
increase in both CDOR and OSR was noticed atlater stages. It was
also shown by Shanqiang and Baker (2006) in a 3Dsimulation model
that decreasing permeability reduced the initial oilproduction but
later increased it dramatically.
McLennan et al. (2006) presented a permeability modeling
pro-cedure. Their methodology consists of two major steps: (1)
debiasingand re-scaling the by-facies core horizontal permeability,
kH vs.porosity relationships using mini-models (seven working
phases),and (2) assigning permeability to the geological grid. They
outlinedtwo key features of their methodology as (1) the
integration ofmissing lower porosities or increased shaliness into
measurementsfrom dilated and preferentially sampled corewhich is
also dilated, and(2) the translation of porosity-permeability
relationships at the corescale to the SAGD ow simulation scale
(McLennan et al., 2006).
Nasr et al. (1996) showed a decrease in OSR due to a decrease
inpermeability through their numerical modeling study. Collins et
al.(2002) stated that laboratory tests on specimens of undisturbed
oilsands have conclusively proven that absolute permeability
increasesdramatically with dilation. They also showed that shear
dilation of oilsands enhances permeability in the SAGD process.
Shin and Polikar(2007) observed that higher permeability resulted
in a higherultimate recovery as well as lower CSOR. They also
noticed that ningupward sequence showed better SAGD performance due
to lateralsteam propagation (cupcake growth). Nasr et al. (1997)
reported froma 2D sand packed model that for low permeability
reservoirs, thesteam zone was localized around the injection well.
The low per-meability reduced the drainage of oil and growth of the
gravity cell.Mukherjee et al. (1994) observed that the presence of
a low per-meability zone between the injector and producer may
cause waterhold up between the wells where water is not well
drained.
Butler (2004b) studied the effects of reservoir layering. He
statedthat in layered reservoirs with permeability ratios less than
about two,the height average permeability should be used in the
Lindrainequation. He then suggested that in the situation described
above,steam should be injected in the more permeable area. He also
statedthat if the more permeable layer is at the bottom, a steam
swept zonewill tend to undermine the upper layer. If the more
permeable layer isat the top and the permeability ratio is greater
than two, thepenetration of the steam into the lower layer will be
delayed and oilwill move through the lower region driven by the
imposed pressuregradient. Effects on oil rate are not very severe
at least until the upperlayer is exhausted.
5.5. Viscosity and API
Das (2007) studied the effect of oil viscosity in a 2D
model,investigating SAGD in a carbonate reservoir. He found that
recoveryrate and injectivity improved with lower viscous oil.
Shanqiang andBaker (2006) studied the effect of oAPI on SAGD
performance andobserved that increasing oAPI reduced oil
production. Singhal et al.(1998) outlined the effects of viscosity
on geometrical and opera-tional parameters in a screening study.
They advised that fromviscosities less than 35,000 mPa s and
thickness more than 15 m,using vertical steam injectors staggered
around horizontal producerswas a feasible recovery strategy. Also,
the relaxation of subcool con-strain under certain circumstances
may be feasible. For viscositiesabove 65000 mPa s, the use of
horizontal injectors and subcoolconstrainwas determined to be
critical. Larter et al. (2008) studied theimpact of variation in
heavy oil heterogeneity in reservoirs through 3Dsimulation. They
concluded that the impact of dramatic oil viscosityvariations in a
heavy oil reservoir on reservoir productivity dependson the
recovery method. They showed that in terms of productivityand
compositional varations of the oil phase, the impact is large
on
SAGD and CSS when initial viscosity gradient with depth is taken
intoaccount. They observed a reduction of 30% for the top oil end
memberand 75% for the bottom oil end member.
5.6. Wettability
Few studies were conducted to study this crucial
reservoirproperty. Das (2007) used a 2D numerical simulator and
observedthat lower oil recovery is obtained with
oil-wet-carbonate-reservoirsand with no capillary pressure.
However, the role of wettabilityalteration from water-wet to
oil-wet was demonstrated to have apositive impact in thermal
recovery around the production wellboreregion (Isaacs et al., 2001;
Yuan et al., 2002). In their patent documentIsaacs et al. (2001)
demonstrated that oil-wet sand in the near regionof the production
well (by treatment with wettability alterationchemicals), when
coupled with SAGD, causes an increase in recoverycompared to
classical SAGD. Following up on that patent, Yuan et al.(2002)
studied the potential impacts of altering wettability near
aproductionwell on SAGD using a eld scale numerical model to
clarifythe possible key parameters. They concluded that (1) the
bigger theregion around the production well being oil-wet, the
better the oilproductionwas, at least in early stages of the steam
chamber, (2)morethan near well effect was observed from alternating
wettability in alocal zone near the production well, (3) SOR was
lowered due toconstant bottom hole pressure, and (4) it might be
benecial not tokeep the oil-wet zone at its wettability status for
the entire operationperiod to reduce the cost of
wettability-changing agent. However, theynoticed water accumulation
between the water-wet and oil-wet zone.This water blockage
phenomenon was caused by creation of oil-wetzone. This diagnose is
very relevant since water ow through the oil-wet region will be
impeded due to the absence of phase lubricantwhich may also be a
factor inuencing SOR. This water blockagewould probably have a
negative impact on steam chamber growth andmaintenancewhichmay be
another reasonwhy oil-wet regionmay bea temporary solution.
These observations lead us to raise a few ags on the role of
ES-SAGD in oil-wet reservoirs and how solvent addition with
hightemperature effect would cause wettability alteration, and
henceaffect gravity drainage performance. These observations and
thoughtsshould prompt further effort to understand the effect of
wettabilityand/or wettability alteration on SAGD and potential
performanceoptimization.
5.7. Heterogeneity
Yang and Butler (1992) studied the effect of reservoir
hetero-geneities on heavy oil recovery by SAGD. Their approach was
to use atwo dimensional sand packed model. They limited their study
to twoeld conditions: (1) a reservoir with thin shale layers, and
(2) areservoir containing horizontal layers of different
permeabilities. Forthe two layer reservoir they studied two cases:
(1) a high/low per-meability reservoir, and (2) a low/high
permeability reservoir. Theynoticed that the high/low permeability
was acting like a whole highpermeability reservoir. In the low/high
permeability case, theynoticed an undermining of steam in the lower
(high permeability)layer. This effect decreased with time. They
then compared the cumu-lative oil production from the previous
setup to all low permeabilitysetups and noticed little difference.
They noted two effects cancellingeach other, namely (1) undermining
steam enhanced gravity drainageof bitumen above the inter-layer
surface, (2) a higher water/oil emul-sion was expected above the
undermining steam causing an increasein the viscosity of the
produced uid. For reservoirs containing ahorizontal barrier, they
conducted two cases with: (1) top steaminjection, and (2) bottom
steam injection. They also studied thereservoir dipping effect for
the low/high permeability setup. Theycontrolled the dipping by
tilting the model upward and downward
by 5 . A reservoir dipping upward gave a higher production than
a
-
142 A.-M. Al-Bahlani, T. Babadagli / Journal of Petroleum
Science and Engineering 68 (2009) 135150reservoir dipping downward.
The reason for this, they commented, isthat the production rates
are mainly controlled by the total drainageheight and placing the
production well at the lowest location of a dipreservoir obtained
the maximum height. They then studied the effectof barrier length
for each case (short horizontal barrier, and longhorizontal
barrier). With a top steam injection, the presence of a
shorthorizontal barrier had no effect on the general performance.
They thenconducted several experiments on long barriers by changing
thelocation of the injection and production well relative to the
barrier.They concluded that a long horizontal barrier decreases the
produc-tion rate but not as much as expected in some congurations.
Theyalso observed that heated bitumen above the barrier may not
beproduced even though it is hot because of the steam pressure
holdingup the oil at the bottlenecks to the ow (Yang and Butler,
1992). Thisconrms that SAGD is heavily dependent on a good
communicatingreservoir.
Chen et al. (2007) conducted a numerical simulation study
onstochastic of shale distribution Near Well Region (NWR) and
AboveWell region (AWR). They stated that drainage and ow of hot
uidwithin the NWR has a short characteristic length and is found
tobe very sensitive to the presence of shale that impairs vertical
per-meability. The AWR affects the (vertical and horizontal)
expansion ofthe steam chamber. It is of characteristic ow length on
the order ofhalf of the formation height. SAGD performance is
affected adverselyonly when the AWR contains long continuous shale
or a high fractionof shale. They also studied the potential
improvement of SAGDperformance by hydraulic fracturing by
identifying three cases:horizontal fracture, vertical fracture
parallel to, and vertical fractureperpendicular to the well. In
some cases, they observed an improve-ment in the oil steam ratio by
a factor of twowhen a vertical hydraulicfracture was introduced.
They also concluded that vertical hydraulicfractures are predicted
to enhance SAGD performance more drama-tically in comparison to
horizontal hydraulic fracture. Finally, theystated that a vertical
hydraulic fracture along the well direction issuperior to one
perpendicular to the well direction.
Zhang et al. (2005) showed 4D seismic amplitude and
crosswellseismic images of steam chamber growth at the Christina
Lake SAGDproject which identied the effects of reservoir
heterogeneity.
The SAGD performance in the presence of water leg was studied
byDoan et al. (1999). They concluded that the presence of water
sandshinders oil recovery. Birrell (2001) advised stepping away
fromsimulation models to achieve an understanding of steam
chamberdevelopment in a heterogeneous reservoir and applying the
actualresults from pilot data if available.
Yang and Butler (1992) showed that long reservoir barriers such
asshales can cause differences in the advancement velocity of
theinterface above and below the barrier. This difference is
reduced bythe drainage of heated bitumen through conduction above
the barrier.
5.8. SAGD in carbonate reservoir
Unlike with clastic reservoirs, very few attempts were made
toexplore the applicability of SAGD in carbonate reservoirs. Yet,
eventhose existing attempts are extremely simplied to the extent
thatjumping to commercial conclusions would be a fallacy. No
studyon any laboratory experiments investigating the physics of
theSAGD process in carbonate reservoir (tight matrix with fractures
orextremely heterogeneous structure with signicant
permeabilitychange) has been identied in the literature. All
presented attemptsare of a numerical simulation nature. However,
there is no doubt thatthese attempts open a wide window for further
investigation.
Das (2007) conducted a 2D simulation model investigating
CSS,conventional SAGD and Staggered SAGD in carbonate reservoirs.
Hismodel had an extreme heavy fractured reservoir with fracture
spacingof 0.54 m. One of his interesting observations was that more
steam
went into the system with wider fracture spacing. He attributed
thatto a higher fracture tomatrix steam invasion. Beside this,
largermatrixis present with wider spacing which implies the need
for more energyto heat up the matrix, hence more steam would be
injected. He thenreported an average oil rate of 400 bbl/d and 34%
recovery in 8 years,which is very optimistic. He also reported an
increase in SOR withhigher fracture spacing.
It is widely known that production from fractured
carbonatereservoirs is mainly due to matrix-fracture drainage. The
productionmechanism in fractured carbonates would be different from
the con-ventional SAGD process in loose sands. The matrix-fracture
inter-action could be enhanced by two horizontal wells, the upper
oneinjecting steam and the lower one collecting the oil, if a good
verticalcommunication exists.
5.9. SAGD geomechanics
Ito and Suzuki (1996) observed a large amount of oil
drainsthrough the steam chamber when geomechanical changes occur
inthe reservoir. Hence, they agged the role of geomechanical change
offormation during SAGD as very important. Chalaturnyk and Li
(2004)hypothesized that, in a SAGD process the combination of
porepressure and temperature effects (resulting from steam
injection)creates a complex set of interactions between
geomechanics and uidow. In their work they studied, using a coupled
reservoir simulation,major geomechanical/reservoir factors which
include: (1) initial in-situ effective stress state, (2) initial
pore pressure, (3) steam injectionpressure and temperature, and (4)
process geometry variables such aswell spacing and wellpair
spacing. They stated that it was difcult tobe conclusive about
specic geomechanical process relative to themultiphase
characteristics of SAGD fromwork at that stage. However,they
provided some observations including enhancement of
absolutepermeability occurrence in the zones of shear failure. Ito
(2007),referring to Chalaturnyk and Li's observations on the
geomechanicaleffects, mentioned that: (1) steam chambers stop
rising or shrinkingwhen injection pressure is reduced, and (2)
steam chambers resumerising when pressure is increased. Ito
emphasizes that it is critical tostudy the geomechanical properties
of oil sands to understand theSAGD process.
Collins et al. (2002) modied a geomechanical/reservoir
simula-tion to incorporate the absolute permeability increase
resulting fromthe progressive shear dilation of oil sands. Li and
Chalaturnyk (2006)emphasized the shearing process inducing
improvement to absolutepermeability. This causes an improvement of
effective permeability towater and thereby, the water relative
permeability increases due tothe isotropic unloading and shearing
process (Li and Chalaturnyk,2006). The movement of uid ahead of the
steam chamber was alsoreported by Birrell (2001). Although he did
not identify the type ofuid, such geomechanical observations (=
water relative perme-ability increase) suggest that this uid
movement is of hot water.Singhal et al. (1998) stated that the
application of the sanddeformation concept (effect of SAGD
geomechanics) to the UTFprojects helped explain the shape and
location of the steam chamber,and the strong oil rate performance
at the central well AP2, which wasmainly due to ceiling drainage of
oil through the steam chamber,rather than gravity drainage along
its sides.
6. SAGD operation
6.1. The start-up procedure
Vincent et al. (2004) dened start-up as the period of
timebetween the introduction of steam into both the injection
andproduction well and when the well pair is converted to
SAGDoperation. Proper initialization procedures are required to
bringthe entire length of a well pair into active drainage (Nasr et
al.,
2000). It is also known that a proper start-up procedure
(especially
-
143A.-M. Al-Bahlani, T. Babadagli / Journal of Petroleum Science
and Engineering 68 (2009) 135150circulation) will heat up the
wellbore ensuring a better steam qualityat the sand face. The space
between the injection and productionwell is heated via conduction.
This is achieved by circulating steamin the tubing and out of the
annulus (Nasr et al., 2000). Anotherinteresting feature of wellbore
steam circulation introduced by Grillset al. (2002) was the
recovery of drilling mud lost to the formation.Sasaki et al. (2001)
reported that increasing vertical well spacingbetween horizontal
wells made the lead time for the gravitydrainage to initiate oil
production longer. This was also observedby Hamm and Ong
(1995).
Doan et al. (1999) stated that depending on the reservoir, a
blowdown phase may be necessary. During the blow down period,
thereservoir is depressurized so that the subsequent injection of
steamensures higher latent heat. The SAGD process follows the blow
downperiod where both injection and production wells are operated
atconstant pressure. In general, the initialization phase is slow
and oilrates during this phase are low (Nasr et al., 2000). Thismay
be becauseof oil being drained through oil expansion only. Chen et
al. (2007)showed that start-up time is sensitive to shale presence
near the wellregion.
Vincent et al. (2004) conducted a coupled wellbore
thermalreservoir simulation study to explore the communication
initia-tion for the MacKay River SAGD project. They investigated
differentvariables for operating strategy development including:
steamcirculation rate and pressure, the magnitude and timing of
pressuredifferential implementation between the injector and
producer, andoptimum timing for SAGD conversion. Maxwell et al.
(2007) used acombination of microseismic and surface deformation
monitoringwith an array of tiltmeters to monitor the warm-up phase
of a SAGDwell pair. For the case studied, they reported the
possibility offracture network creation which is then lled with
steam at laterstages. Another way to conduct start-up was done in
the Cold Lakeproject where wells were pressurized with a solvent
followed by hotwater to push the solvent into formation. Then,
normal steamcirculation was conducted in both wells (Donnelly,
1997). Throughtheir numerical model study, Shin and Polikar (2007)
found thatthe start-up period increased with decreasing
permeability andincreasing well spacing.
By installing a ber-optic distributed temperature system
(DTS),Karwchuk et al. (2006) noticed a signicant thermal gradient
existsacross the producing well's diameter. From a thermal model
(Joshi's)they showed that temperature and magnitude of the ow
fromupstream have an impact on the wellbore recorded
temperaturesdownstream, and that this cooling has an impact on the
sub-coolmeasurements calculated. Thus, the sub-cool temperature may
notreect the true rock temperature surrounding the producing
welldownstream of an inow, and thus the inow performance at
thatpoint. They also noticed that ow is greater initially at the
toe of thewell; however, ow improved as the well was further
heated.
Nasr et al. (1991, 1998) commented that the steam
circulationphase delays oil production. They then proposed two
novel methodsto accelerate this phase. The rst is by linking
transversely the pairedhorizontal wells with vertical channels to
improve liquid drainage,and the second is the addition of naphtha
as a steam additive toaccelerate oil drainage. They reported an
increase in oil productionand a lower SOR. Showing their experience
gained from the Celticeld, Saltuklaroglu et al. (2000) reported
expected problems with thesteam circulation method for
communication achievement. Theimportant one to mention is that
production through annulus wouldresult in high pressure (up to 4600
kPa) above the fracture pressure,which would result in excessive
heating of the intermediate casingand cement. They thus decided to
go for cyclic start-up process.
It is noticed that the essence of the start-up procedure is
tocreate a gravity drainage seed which will grow into a chamber.
Thus,conducting successful start-up is essential for a successful
SAGD
application.7. Steam quality
Gates and Chakrabarty (2005) stated that the quality of
theinjected steam should be as high as possible at the sandface
becauseany condensate in the injected uids falls under gravity from
theinjector towards the producer and does not deliver a
signicantamount of heat to the oil sand. Gates et al. (2005)
provided an imageshowing variation in steam quality throughout the
steam chamber.This may be a good indication of the temperature
prole inside thesteam chamber. In these images, steam quality drops
as steam movestowards the edge of the chamber which supports claims
that pressureinside the steam chamber is not constant. In terms of
the steamquality effect on emulsication, Chung and Butler (1989)
reportedfrom a 2D experimental model no signicant difference on
emulsi-cation with wet or dry steam. They attributed that to the
interfacialactivities and the heating mechanism being the same at
the steamfront. The comparison was done on a low pressure
injection, and theydid not report any high pressure.
8. Length, spacing and placement of horizontal wells
Wellbore ow restriction at eld conditions was studied by Ongand
Butler (1990). They found that the effect of well length on
thegravity head was not as signicant as the effect of well size.
They alsoreported that the steam chamber slope may be caused by a
wellborepressure drop. Nasr et al. (2000) stated that pressure drop
along thehorizontal wellbore causes a slope in the steam chamber
along thewell. Singhal et al. (1998) suggested that, as the well
length in SAGDoperations vary between 5001000 m and because steam
chambers inmany situations are unlikely to spread more than 50 m
lateral to thewells on either side, exploitation by 500 m long well
pairs placed100 m apart may be considered.
Sasaki et al. (2001) observed from a laboratory 2D scaled
reservoirvisualization model that setting larger vertical spacing
betweeninjection and production horizontal wells resulted in
quicker genera-tion of the steam chamber and increased oil
production. However, italso led to longer breakthrough time. They
concluded from this thatthe interwell spacing (L) can be used as a
governing factor to evaluateproduction rate and lead time in the
initial stage of the SAGD process.Canbolat et al. (2002) observed
through a series of 2D experimentsthat a larger recovery efciency
was achieved for smaller injector-to-producer well separations.
This was also observed by Chan et al (1997)from a numerical
simulation model. However, in their case the reser-voir was thin
(20 m) and the injector was placed 3 meters below thetop of the
reservoir, thus such results are expected. They also reportedthat
even when oil pay is increased, the injector is
preferentiallyplaced above the midpoint of the oil pay section.
They concluded thatinjector offset may capture an incremental 1015%
recovery.
Terez (2002) studied the effect of well placement in a 40 ft
thickmodel. He conducted 26.4 ft and 36.6 ft interwell spacing runs
andresults did not show a noticeable difference. However, Terez's
workwas not conducted on a classical SAGD basis, rather it was a
generalstudy of horizontal wells in thermal application for
displacement andgravity drainage where SAGD comes as a tertiary
recovery process.Shin and Polikar (2007) found from a 2D simulation
model thatby increasing the spacing between the injector and the
producer,CSOR decreased due to enhanced thermal efciency. By
reducing thespacing, the bitumen recovery reached its highest point
and thendecreased. They commented that I/P spacing does not affect
theultimate recovery. However, we think that such factors are
subjectedto time constraints and unfortunately the authors did not
providegraphs to make such a comparison.
Das (2005a) stated that over 80% of steam is injected at the
heel ofthe well and the remaining steam is injected at the toe,
while uid isproduced either from the heel or the toe or from both.
This can explain
the tilted steam chambers and raises questions about the
uniformity
-
gave a good comparison between the effects of steam pressure
onsteam temperature and hence, its effect on oil production and
SOR.
144 A.-M. Al-Bahlani, T. Babadagli / Journal of Petroleum
Science and Engineering 68 (2009) 135150of steam chamber growth
along the horizontal well. This is backed upwith another graph
presented by Das (2005a) which shows that thepressure difference
between the injector/producer heels is higherthan between the toes.
He comments that this situation may impose agreat potential for
steam breakthrough around the heel area. His studyprovided good
insight into the crucial role of wellbore design toachieve a
successful SAGD. An interesting feature shown in this studywas that
45% of injected heat was produced back to the surface in
aconcentric wellbore design during startup. He commented
thatwithout the temperature data inside the wellbore, it is difcult
todecide whether the vapour quality at the surface is due to
counter-current heat transfer or due to the excess steam in the
horizontalsection.
Yet there is no optimum well spacing proposed. This is due to
thefact that thickness, viscosity, permeability and heterogeneity
might begoverning factors in choosing an optimum well spacing.
However, itseems that common practice is somewhere between 515 m
apart.
9. Subcool temperature (steam trap control)
Doan et al. (1999) stated that steam trap control is used as
anoperational control to reduce or prevent steam withdrawal from
thesteam zone in the reservoir. Das (2005b) identied three
mainadvantages of steam breakthrough prevention to the SAGD
process:(1) energy conservation and SOR reduction, (2) reduction of
highvapour ow which negatively affects the lifting capacity of the
welland surface facilities, and (3) reduction of sands and ne
movementthrough the liner which may cause erosion. He then added
that due tothe uneven nature of the well trajectory in the eld, it
is very difcultto identify, rest alone, and control steam
breakthrough.
Ito and Suzuki (1996) reported the optimum temperature for
asteam trap control to be between 30 to 40oC. They referred to
uiddrainage ahead of the steam chamber, which is 30 to 40 C lower
thansteam saturation temperature. Das (2005b) noticed a positive
effect ofsub cool temperature after exceeding 20 C. Chen et al.
(2007) showedthat when coupling hydraulic fracturingwith steam trap
control of theproducer well, injectivity is dramatically improved
and an effective oilproduction rate in the reservoir with poor
vertical communication isachieved. Edmunds (2000) investigated this
feature on 2D and 3Dnumerical models. He found that in a specic
case, a steam trap of2030 C was optimum. However, he reached a very
importantconclusion, stating that the utility of the (mixed) BHT
(Bottom HoleTemperature) as an operating control parameter is
doubtful. Thisconclusion was drawn from eld observations where an
increasingproduction rate caused the BHT to drop. This observation
as hestates contradicts with a widely known assumption that
the(mixed) producing temperature is always a monotonic
(increasing)function of the production rate. He then advised that
operators arebetter off to be cautious when setting production
rates with availablehandling capacity of the plant. Another 2D and
3D dynamic modelswere conducted by Ivory et al. (2008) to compare
ES-SAGD and SAGDoperating at low pressure. They found that an
introduction of 10 Csubcool temperature minimised oil production
and SOR. When nosubcool temperaturewas introduced into the system,
an increase in oilproductionwas observedwith a greater SOR. They
also showed that anincrease in BHP created gas saturation around
the production well,which implies the need for a greater subcool
temperature setpoint.From screening of Tangleags type projects,
Singhal et al. (1998)advised that steam trap constrain on
production could be ignoredunder certain circumstances, especially
during early periods of steaminjection, to achieve an optimal
performance.
9.1. HP (high pressure) vs. LP (low pressure) SAGD
One of the controversial issues in SAGD operations is whether
to
operate at high or low pressures. Pooladi-Darvish and Mattar
(2002)They stated that a higher steam pressure leads to higher
steam tem-perature and lower oil viscosities. This in turn, leads
to a higher oilow rate. On the other hand, higher steam pressure
leads to lowerthermal efciency and higher steam oil ratio (SOR).
Some of thereasons for larger SOR are that steam at higher pressure
has less latentheat, more heat will leave the reservoir through the
produced uids athigher temperature, and more heat will be left in
the steam chamberwhere oil is no longer present. They also added
that a higher pressurewould allow natural lift of the produced
uids. Edmunds and Chhina(2001) analytically showed the relationship
between LP-SAGDand low CSOR. They illustrated a six fold increase
in oil rate be-tween atmospheric pressure and 10 MPa. They also
conducted aseries of simulation and economic analyses and concluded
in favour ofLP-SAGD to HP-SAGD for the following reasons: (1) SAGD
economics(mainly due to gas price) is sensitive to CSOR and LP-SAGD
improvesCSOR, and (2) ESPs capability improves with LP-SAGD. Ito
and Ipek(2005) observed that high pressure operation is important
foractivating steam ngers.
Das (2005b) conducted a simulation study where he examined
theeffects of lower operating pressure. He identied two
advantagesof LP-SAGD over HP-SAGD due to lower operating
temperature:(1) reduced silica content in produced uids and (2)
lower H2Sproduction. He reached the following conclusions: (1) low
pressureoperations appear to be energy efcient, and (2) low
pressure opera-tion is more amenable to the application of articial
lift. He alsomentioned that reservoir characteristics may lead to
uid losseswhich may dictate the operating pressures. These
observations by Dasare in agreement with Edmunds and Chhina's
(2001) conclusions.
Butler and Yee (2002) reported for Imperial Oil SAGD pilot HWP1
agradual decrease in CSOR with time, which is an indicator of
aneconomic SAGD project. Initial operation pressure was of the
sameorder of reservoir pressure (5 MPa). Two years later, chamber
pressurewas kept at 12 MPa using periodic steaming. They expected
that this,along with low operating pressure, made sufcient gas
available toproduce improved steam economy through the SAGP2
effect, eventhough gas was not added to the steam.
Sasaki et al. (2001) studied the effect of steam injection
pressure ina 2D laboratory scaled visualization model. They stated
that highersteam-injection pressure leads to a shorter breakthrough
time andhigher expansion rate of the steam chamber as the higher
pressuredrop between the injection and production wells (p) results
in alarger driving force for oil mobilization. It is worth
mentioning thatthey studied the effect of injection pressure by
studying (p) sincethey set production pressure to a constant.
Wiltse (2005) introduced a hydraulic gas pump (HGP) as
articiallift system solution for LP-SAGD. After eld testing, he
reported thatHGP had outperformed the reciprocating rod pump system
installedearlier in that well. Li et al. (2006) conducted a coupled
simulationbetween EXOTHERM and FLAC to examine favourability
between HPand LP-SAGD. They concluded that high pressure SAGD
induced higherporosity and permeability with higher oil production.
In order toobserve the effects of geomechanics variation and hence
permeabilityenhancement due to shear dilation, steam chambers
should be opera-ted at or near the minimum total stress, which
implies HP-SAGD(Collins et al., 2002; Chalaturnyk and Li 2004; Li
and Chalaturnyk2006; Li et al., 2006).
Another study showing favourability of HP-SAGD was conductedby
Robinson et al. (2005), who also reported an increase in oil
pro-duction. Kisman and Yeung (1995) also showed that operating at
lowpressure decreased oil production and improved OSR in a
simulationmodel for the Burnt Lake oil sand conditions. Shin and
Polikar (2005)
2 SAGP (steam and gas push) an addition of non-condensable gas
to steam in order
to minimize heat loss from steam chamber.
-
145A.-M. Al-Bahlani, T. Babadagli / Journal of Petroleum Science
and Engineering 68 (2009) 135150observed through a 2D simulation
model study that high pressureinjection gave a better CSOR and
CDOR. However, the LP-SAGD gavethe highest NPV. Gates et al. (2005)
stated that high injection pressureimplies a relatively high
saturation temperature that leads tofavourable bitumen viscosities.
This explains the favourability of HP-SAGD in models which do not
take into consideration geomechanicaleffects. Bharatha et al.
(2005) reported from a simulation andtheoretical study that HP-SAGD
reduces the effect of dissolved gassaturation in a SAGD operation.
Card et al. (2006) suggested changingthe operating pressure in
twomanners: (1) operating at high pressureuntil steam chamber
contacts the overburden, and (2) then operate atlow pressure to
minimize heat losses. Collins (2007) stated that amajor benet of
HP-SAGD is that the produced uids ow to thesurface under reservoir
pressure, as long as the pressure differentialbetween the steam
chamber and thewellhead exceeds the hydrostatichead of the
production uids. He also discussed the failure of LP-SAGDin the
Peace River Shell project where SAGD injectionwas at 2700 kPaand
SOR ranged from 510. The performance improved afterconversion to
CSS with an injection pressure of 11,000 kPa. Heidentied the effect
of well depth as a factor in the higher pressurerequirement for
deeper projects. He then provided a comparison ofLP and HP-SAGD
factors such as lift, heat exchange, water treat-ment, viscosity,
wells, heat losses, geomechanical enhancement, andresidual oil.
Thimm (2005) stated that for the most part, the scaling
aspectsdue to produced water in SAGD tend to favour LP-SAGD. In
unusualcases, where naturally occurring radioactive materials are
involvedwith sulphate scales, or where signicant amounts of
phosphate arepresent, a higher pressure might be favourable.
9.2. Steam chamber monitor and volume size estimation
In a SAGD process, pressure and temperature monitoring indicate
areference of heat transfer process occurring in the reservoir and
ofsteam displacement along the completion to the reservoir
(Herrera,2001). The inection method (using thermocouples) is
considered tobe the classical method in determining the top of the
SAGD steamchamber (Birrell and Putnam, 2000). Birrell and Putnam
identied thedrawbacks of this method, where in the eld,
thermocouple spacingcan limit the ability to make accurate steam
rise rate determinations.Also, drops in steam chamber pressure (and
in turn, temperature)may result in a situation where temperature in
the bitumen saturatedsands above the steam chamber is hotter than
in the steam chambergiving the false impression that the steam
chamber has risen. Thus,they applied a graphical method utilizing
Inverse Conjugate ErrorFunction (ICEF) incorporated with natural
log plots to interpretthermocouple data which allowed for the
determination of steamchamber position to the centimetre scale.
They corrected for transienttemperature effect resulting from steam
chamber pressure andtemperature variation by simplifying the
operating temperature to anite number of values with step change
(Birrell and Putnam, 2000).
Zhang et al. (2005) used 4D seismic and crosswell seismic
imagingto monitor and understand steam chamber growth. The
imagesobtained showed that less than 100% well length was swept by
thesteam chamber and a non uniform steam chamber growth
occurred.Shamila et al. (2005) conducted a study aimed towards
investigatingthe applicability and subsequent accuracy of the
pseudo-steady statemethod in estimating the swept volume/size (uid
injectionchamber) under steam injection conditions, with the
application ofhorizontal wells using a commercial 3D thermal
reservoir simulator.Herrera (2001) suggested the use of
microseismic measurements forsteam chamber monitoring by combining
it with 3D visualizationtechnology (Herrera, 2001). An example of a
well monitored SAGDproject is Phase B at UTF, where eleven
temperature observationwellsare available with twenty thermocouples
in each well spaced
throughout the McMurry succession (Birrell, 2001).10. Numerical
simulation
Edmunds (2000) conducted a 2- and 3D numerical simulationstudy
on steam trap control. He reported that 2D simulations werefound to
be unrealistically optimistic for general SAGD problems.He also
agged a very important fact that well pairs in SAGD are
nothomogeneous as assumed, for several reasons including
geology,spacing, and skin factor.
Another comment was made by Collins et al. (2002)
aboutconventional reservoir simulations; they do not account for
geome-chanical enhancement explicitly, but implicitly include the
effects byusing permeabilities from highly disturbed core. Li et
al. (2006), and Liand Chalaturnyk (2003) worked on a coupled
simulation ofEXOTHERM and FLAC and showed a higher oil production
than inuncoupled simulation. They inferred that this difference
took intoaccount the enhancement on both porosity and permeability
incoupled simulation.
An important feature related to the variation in relative
perme-ability curves due to temperature was not incorporated in
most of thenumerical simulations presented in the literature.
In their steam chamber volume estimation by well test
analysis,Shamila et al. (2005) observed that increasing the grid
density of thesimulation model greatly increases the precision and
accuracy ofswept volume estimation using the pseudo-steady state
method. Thisshows the sensitivity of grid density on the accuracy
of the results,which is not incorporated in some numerical
simulations reported. Infact, some simulation studies used huge
coarse grids for eld simu-lation. Yet having a ne grid for eld
scalemodels will increase the runtime and depending upon CPU
capabilitymay not converge. Thus,dynamic gridding may be a good
solution for such cases. Christensenet al. (2004) showed that
dynamic gridding reduced the CPU timethree folds while keeping good
accuracy in the simulation results.However, they also commented on
a gure where they observed anger shape above the steam chamber with
dynamic gridding whichwas not shown with ne gridding. They
explained this as being aninaccuracy probably introduced by the
simple upscaling technique.However, comparing that shape to Sasaki
et al.'s (2001) 2D visualiza-tion observations, we wonder if that
could actually be a steam ngerrepresented by dynamic gridding,
which was not introduced in negridding.
Another remark we can make is that many simulation models,which
tried to draw conclusions for SAGD performance in specicreservoirs,
use a simple Cartesian model which does not depict actualreservoir
heterogeneity. Thus, one has to ensure to use
representativereservoir models in SAGD performance analyses using
numericalsimulators, as the reservoir heterogeneity may have
considerableeffects on the accuracy of the process. A good
demonstration of suchan approach is shown by Robinson et al.
(2005). McLennan andDeutsch (2005) stated that ow modeling is a
transfer functionconverting the geological uncertainty to
production uncertainty. Thus,the importance of having a good
reservoir static model emergeswhere they further state that the
main objective of using geostatisticsto characterize a potential
SAGD reservoir is to quantify the uncer-tainty in production
performance (Oil Production Rate and SOR) dueto geological
uncertainty. They conducted a study where they imple-mented nine
different static ranking measures for a potential SAGDreservoir.
They concluded that static measures of local connectivitywere the
most effective; they referred that to their correlation of
OilProduction Rate and SOR being the highest. They stated in their
con-clusion that dynamic ranking measures suffer from
simplifyingassumptions that mask geological heterogeneity where
static rankingmeasures explicitly account for geological
heterogeneity modeled bygeostatistics.
Nestor et al. (2001) referred the complexity of optimizing
theSAGD process to the time consuming and limited number of
objective
function (performance measure) evaluations, a potentially
high
-
146 A.-M. Al-Bahlani, T. Babadagli / Journal of Petroleum
Science and Engineering 68 (2009) 135150number of parameters, and a
non-linear solution space. They alsonoted that the performance
measures (such as net present value,cumulative oil production, and
cumulative steam injection), geome-trical parameters (e.g. I/P
spacing, well length), and operational para-meters (e.g.
subcooling, steam injected enthalpy etc.) requireexpensive
numerical simulations beside run time/number consump-tion. This
diagnosis seems to be accurate since Gates et al. (2005)reported
executing over 100 runs to achieve a SAGD optimization.Nestor et
al. (2001) thus proposed a solution for optimisation calledNEGO
(neural network based efcient global optimization) whichwould
optimize geometrical and operational parameters in a SAGDprocess.
They stated that the solution methodology includes theconstruction
of a fast surrogate of an objective function whoseevaluation
involves the execution of a time consuming mathematicalmodel (i.e.
reservoir numerical simulator) based on neural networks,DACE
modeling (design analysis of computer experiments), andadaptive
sampling.
Tan et al. (2002) conducted a simulation model to investigate
theimportance of using a discretized wellbore for SAGD simulation.
Theyconcluded that a discretized wellbore model is necessary to
predicttemperature and saturation proles for startup and production
ofSAGD well pairs. The use of discretized wellbore is becoming
acommon practice. This feature enables steam circulation during
thestartup period.
11. Experimental pitfalls
All experimental works to our knowledge were performed
onsandpack (natural sand or glass beads) models with no
exception.Therefore, the evaluation of the experimental efforts
will be only onthis type of model.
Ong and Butler (1990) developed a scaling criterion for
horizontalwells. They showed that the scaling factor for the radius
of thehorizontal well is proportional to the one-fourth power of
the productof the ratios of their respective heights of formation
and uid viscosityvalues. This geometrical scaling resulted in an
impractical (large)laboratory size well. To determine a smaller
size well, they thenstudied the wellbore ow resistance which
resulted in a more prac-tical wellbore size (Ong and Butler
1990).
Sasaki et al. (2001) found that for a 2D model, heat loss owing
tocondensate production had little effect on oil drainage process
nearthe steam-chamber interface. They explained this as due to
con-densate ow down the side walls in the central region of the
steamchamber. They concluded that scaled 2D models are possible
toanalyze steam chamber behaviour in SAGD process
investigation,except for the amount of single-phase water
condensate dissipated atheat loss. Chung and Butler (1987) had the
same approach with theintroduction of the dimensionless number. A
scaling method of Pujoland Boberg used by Nasr et al. (1996) for
their 2D model where theeld and model must be geometrically similar
(i.e., width-to-lengthratio and height-to-length ratio must be the
same). For otherparameters containing uid and rock properties,
terms related tothe transport of heat and mass, and initial and
boundary conditionsmust be equal in the eld and lab model. Unlike
Pujol and Boberg'sscaling approach, the permeability used in the
laboratory experimentwas different from that in the eld. Nasr et
al. (1996) commented onits impact as presenting inadequate scaling
of capillary pressure. Usinga 2D sand packed model, Nasr et al.
(1997) reported that transitionfrom initialization (injection in
both wells) to developed SAGD (upperwell injection, lower well
production) resulted in a temporary coolingof wells and drop in
production. This phenomenon is observed moreduring the laboratory
experiments with sand packed models than inwhat actually happens in
the eld, since poor model insulation cancause such pitfalls.
Although, they reported based on previousobservations that heat
loss was negligible, when they injected a
small amount of steam in the producer, better results were
observed.Thus, we recommend a mean of heat maintenance in the
productionwell during the classical SAGD laboratory experiment. We
add thatsand pack models would usually represent, to some extent,
sand-stones, which are cohesive by capillary forces or interlocked
sandgrains. This may favour sand beads movement with pressure
variationshowing false increasing permeability, hence higher
production, andlower SOR.
12. SAGD improvement
From what we have seen above, it is obvious that a
consistentsteam chamber growth is indispensable for a successful
SAGDoperation. Thus, different attempts were made to accelerate
andimprove the efciency of this process. From the above
observations,we can classify such attempts under two main
categories, namely(1) chemical, and (2) geometrical. The chemical
approach aimsdirectly for improving heat efciency and reducing the
oil waterinterfacial tension to achieve higher production. The
geometricalapproach attempts to alternate pressure differential
points related towell placement in order to achieve accelerated
chamber growth.
12.1. Geometrical attempts
1. Cross-SAGD or X-SAGD: The main feature of this conguration is
tocreate a mesh of injection and production wells. The
operationtechnique is then to alter the injection and production
pointsaccording to strategic timing to minimize steam short
circuitingand hence improve steam trap control and production. The
crossingpoints between wells are either plugged initially or at a
later stagein the project life. According to Stalder (2007) there
are twopenaltieswith X-SAGD: (1) Only the points where the wells
crossare effective in establishing the initial steam chamber rather
than atthe entire length of the wells, (2) the plugging operation
requiresan additional cost and poses a serious practical challenge
tooperations. One can add to those points that X-SAGD would
requirea high initial CAPEXwhere conducting a pilot would be
difcult ifnot impossible with fewer wells. This increases the
initial risk.Stalder (2007) conducted a comparison study between
X-SAGDand SAGD in a numerical simulation model representing
expansive,contiguous, and homogeneous bitumen reservoirs. His
resultsindicated that XSAGD has the advantage of accelerating
recovery,achieving higher thermal efciency by reducing CSOR,
andfavouring low pressure to high pressure SAGD.
2. Fast-SAGD (F-SAGD): This technique employs an
additionalhorizontalwell aimed to accelerate and improve the
steamchambergrowth rate. The horizontal well is placed alongside
the well pairwhich operates by CSS. Shin and Polikar (2006a,b,
2007) conducteda 2D simulation model and concluded that the F-SAGD
had a lowercumulative steam-oil ratio due to thermal efciency and
highercalendar daily oil recovery (CDOR) that was as much as 34%
ofclassical SAGD. They showed gures where higher oil recovery
wasobtained from the F-SAGD compared to classical SAGD but they
didnotmention if thiswas from the SAGDproductionwell or
combinedwith the offset well. It is obvious that two wells would
producemore oil than one. Thewaywe look at the F-SAGD is as an
attempt tocreate a pressure sink in the lower part of the
reservoirwhere steamwill compromise between its tendency to rise
and the physical factof uid movement from high to low pressure
points.
12.2. Chemical attempts
1. Expanding solvent SAGD (ES-SAGD): This novel approach
wasdeveloped by Nasr et al (2003). Its main concept is the
co-injection ofhydrocarbon additive with steam at low
concentrations. Anotherapproach was a hybrid injection of steam and
solvent. Solvent would
condense with steam around the steam chamber interface causing
oil
-
147A.-M. Al-Bahlani, T. Babadagli / Journal of Petroleum Science
and Engineering 68 (2009) 135150dilution and viscosity reduction. A
reduction was reported in thesteam oil ratio by up to 50% and
solvent recovery of 9599% in a 2Dexperiment. 2D experiments showed
an improved oil recovery,enhanced non-condensable gas production,
lower residual oil satura-tion, and faster lateral advancement of
heated zones. It was alsoreported that adding non-condensable gases
to live oil did notimprove the process because of initial methane
presence (Nasr et al.,2001; Nasr et al., 2003; Nasr and Ayodele,
2008; http://www.petro-canada.ca/en/about/636.aspx last visited Jan
12., 2008). From theimages provided by Nasr and Ayodele (2008), one
may notice that ES-SAGD temperature was uniformly centred in the
middle of the modelcompared to the classical SAGD. This adds
another point to the processwhere solvent may operate as insulator
reducing heat losses andhence reducing the amount of gas needed.
This solvent effect mayhave a greater role than viscosity reduction
since at a highertemperature, further viscosity reduction due to
solvent addition mayhave only a small effect. This observation was
also reported by Deng(2005). He pointed out that the viscosity
reduction was mainly fromsteam. He also noticed that a higher
addition of propane impedes heattransfer between the steam and the
oil zone. Deng (2005) used a 2Dmodel to simulate a steam/propane
hybrid process. He observed thatpropane's role was to maintain the
reservoir pressure, which raisessome questions about how solvent
addition would affect reservoirgeomechanics. Images provided from
Deng's (2005) simulationsshowed that the addition of propane
converted the steam growthshape from a hand fan shape to a cupcake,
where better lateralmovement was noticed compared to classical
SAGD.
Gates (2007) conducted a study to determine a suitable
injectionstrategy for higher ultimate recovery by visualizing the
process in aphase diagram. He concluded that the presence of
solvent in ES-SAGDyields a lower operating temperature due to
partial pressure effects.He also showed a solvent recovery of
around 80%. Ivory et al. (2008)conducted 2D and 3D simulation runs
where they examined the effectof diffusion and dispersion on the
ES-SAGD process. They observedthat the diffusion coefcient
increased with increasing temperature.They referred this to the
decrease in the oil viscosity, which isinversely proportional to
the diffusion coefcient. Ayodele et al.(2008) conducted laboratory
experiments investigating effect of lowpressure ES-SAGD. They
compared the single component (propane)case with multi-component
systems. They concluded that multi-component solvent yields better
eld application for low pressureES-SAGD with lower energy
consumption.
This process (ES-SAGD) was tested by Suncor Energy in Burnt
Lakeand Firebag. Petro-Canada is also planning to pilot solvent
SAGD inMacKay River.
2. Non-condensable gas (NCG) or SAGP: As it was pointed
outearlier, Butler (1997) and Butler and Yee (2002) mentioned
thepotential problems associated with wasted heat in a mature
SAGDproject. Adding non-condensable gas to steam affects both the
steamchamber growth rate and shape (Butler and Yee, 2002; Canbolat
et al.,2002). For the process of SAGP, Butler (1997) stated that
the injectionconditions are such that a very high concentration of
non-condensiblegas accumulates in the steam chamber, particularly
near the top. In theprocess proposed, the concentration of
non-condensible gas (typicallymethane) at the top of the steam
chamber is intentionally maintainedat a level over 90 mol% and the
dewpoint of this gas is much lowerthan the saturation temperature
of steam at reservoir pressure. Thesehigh gas concentrations are
maintained by the addition of natural gasto the injection steam.
The gas addition mu