You are on page 1of 8

Investigation of Emulsion Flow in

Steam-Assisted Gravity Drainage
C.C. Ezeuko, J. Wang, and I.D. Gates, University of Calgary
Summary
We present a numerical simulation approach that allows incorpo-
ration of emulsion modeling into steam-assisted gravity-drainage
(SAGD) simulations with commercial reservoir simulators by
means of a two-stage pseudochemical reaction. Numerical simu-
lation results show excellent agreement with experimental data
for low-pressure SAGD, accounting for approximately 24% defi-
ciency in simulated oil recovery, compared with experimental
data. Incorporating viscosity alteration, multiphase effect, and en-
thalpy of emulsification appears sufficient for effective represen-
tation of in-situ emulsion physics during SAGD in very-high-
permeability systems. We observed that multiphase effects appear
to dominate the viscosity effect of emulsion flow under SAGD
conditions of heavy-oil (bitumen) recovery. Results also show
that in-situ emulsification may play a vital role within the reser-
voir during SAGD, increasing bitumen mobility and thereby
decreasing cumulative steam/oil ratio (cSOR). Results from this
work extend understanding of SAGD by examining its perform-
ance in the presence of in-situ emulsification and associated flow
of emulsion with bitumen in porous media.
Introduction
Commercial development of heavy oil (defined as API gravity
less than 22

API) and bitumen is becoming increasingly wide-
spread but remains challenging and often energy-intensive. Heavy
oil currently contributes more than 7% of the total world daily oil
production, and this is expected to continue increasing (Dusseault
2002). SAGD is the most common commercial in-situ recovery
method in Canada for extremely viscous hydrocarbons, with vis-
cosity sometimes exceeding 10
6
cp. SAGD describes a thermal re-
covery method involving steam injection by means of a horizontal
well a few meters (typically 5 m) above another (horizontal) pro-
duction well, close to the bottom of a reservoir. For typical Atha-
basca (Canada) bitumen, the viscosity correlation developed by
Mehrotra and Svrcek (1986) predicts that a 100

C increase in
temperature induces approximately four orders of magnitude
decrease in viscosity. Experimental studies have proved important
in the advancement of SAGD modeling and commercial field
applications. Clearly, the effective impact of phenomena or proc-
esses observed in laboratory experiments is likely to be different
at field scale. Some important processes that have been observed
during SAGD experiments by independent researchers (Chung
and Butler 1987; Nasr et al. 2000; Sasaki et al. 2001a, b; Moham-
madzadeh and Chatzis 2009; Mohammadzadeh et al. 2010) are
summarized in Fig. 1. The processes presented in this figure are
still areas of ongoing research, with a view of developing a
consistent physical understanding. However, incorporating these
processes in numerical modeling, based on current understanding,
can provide important insight for interpretation of SAGD experi-
mental and field data. Our work focuses on the emulsification pro-
cess observed during SAGD experiments, which is also reported
in produced fluids at field scale.
Emulsion is a colloidal system of immiscible fluids, with one
fluid as the dispersed phase (usually micrometer-sized drops) and
the other as the continuous (nondispersed) phase. In petroleum
systems, emulsion is composed primarily of oil and water as the
interacting fluids. Emulsions are stabilized by surface-active
agents that adsorb at the oil/water interface. Natural emulsifiers,
such as asphalthenes, are common in heavy oils and bitumen;
asphalthene levels as high as 18 wt% have been reported in Atha-
basca bitumen (Zhao et al. 2009). Although other surface-active
components (e.g., resins, fatty acids) can be found in heavy oils,
asphalthenes are considered to have the strongest potential to pro-
duce stable emulsions, forming “skins” at the oil/water interface.
A combination of factors, such as the presence of asphalthenes
and sand and the low density difference between oil and water,
promote in-situ emulsification during production of heavy-oil sys-
tems. Emulsion stability can also be decreased by asphaltene solu-
bilization by resins. According to Bancroft’s rule (Bancroft 1913),
water-in-oil (WO) emulsion is most likely to form when oil and
water mix in the presence of asphalthenes (as the primary active
agent), because asphalthenes are significantly more soluble in oil
than in water. This rule appears to hold strongly for cold-produced
heavy oil. Although emulsion is consistently reported in SAGD
field operations, the exact details of produced emulsion from
heavy-oil fields undergoing SAGD [i.e. WO, oil-in-water (OW),
complex mixture] is seldom reported. Field reports mostly focus
on the impact of emulsion on separation and transportation of pro-
duced liquid. It has been reported in some instances that the pro-
duction of oil-in-water (reverse) emulsion is common in SAGD
(Bosch et al. 2004). It is still unclear whether the OW emulsion
originated from within the reservoir or was a result of inversion
within the flow systems. Experimental investigations, however,
suggest that liquid produced during SAGD is predominantly WO
emulsion (Chung and Butler 1987; Sasaki et al. 2001a, b; Noik
et al. 2005; Mohammadzadeh and Chatzis 2009). In this respect
and similar to other processes identified in Fig. 1, laboratory evi-
dence appears definitive that WO emulsion is predominant in situ.
It has also been observed that produced WO emulsion during
SAGD is a microemulsion type with average droplet sizes of the
dispersed (water) phase ranging from 1.1 to 5 lm (Mohammadza-
deh and Chatzis 2009; Noik et al. 2005), which implies that in-
situ-generated emulsion during SAGD is reasonably stable. In
addition, the exceptionally large (darcy scale) permeabilities of
Canadian bitumen reservoirs suggest that “straining capture” of
dispersed-phase droplets and consequent pore-blocking effect
should be negligible because of a high pore-size/droplet-size
ratio.
Emulsification is a complex process evident in contradictory
experimental observations. Its prominence in SAGD experiments
(both laboratory and pore scale) suggests that it should be consid-
ered during development of analytical and numerical simulations
of SAGD. Why does it matter whether reservoir fluids are in
native or emulsion state during SAGD? Emulsion properties (e.g.,
viscosity, heat capacity) can differ significantly from those of con-
stituent components. In addition, emulsion formation can appreci-
ably change the multiphase status in porous media. Emulsions
have been rigorously explored and exploited in pipeline transport
of highly viscous oils (Yaghi and Al-Bemani 2002), but sparse in-
formation is available regarding in-situ-emulsion studies during
thermally enhanced heavy-oil and bitumen recovery. Numerical
modeling of low-pressure SAGD experimental data also presented
in Sasaki et al. (2001b) showed that the traditional SAGD-model-
ing approach (excluding emulsion effect) was insufficient at
describing the SAGD process when significant emulsification
Copyright VC 2013 Society of Petroleum Engineers
This paper (SPE 157830) was accepted for presentation at the SPE Heavy Oil Conference
Canada, Calgary, 12–14 June 2012, and revised for publication. Original manuscript
received for review 17 April 2012. Revised manuscript received for review 12 September
2012. Paper peer approved 31 October 2012.
440 June 2013 SPE Journal
occurs. Azom and Srinivasan (2009) coupled a mechanistic emul-
sion model with numerical SAGD simulation and obtained an
improved match of the final cumulative produced oil for the
experiment of Sasaki et al. (2001b). However, their numerically
predicted cumulative oil performed poorly in capturing the trend
of experimental cumulative oil. Here, we document a new mecha-
nistic emulsion model, requiring mostly parameters that can easily
be obtained from field or experimental data.
In this work, we present a mechanistic and practical approach
to coupling emulsion modeling with numerical SAGD simulation
by use of a commercial reservoir simulator (STARS 2011). Reac-
tion kinetics is inferred from history match of experimental oil
production data. WO emulsion alone is considered in the pre-
sented simulations. In-situ formation of emulsion is modeled as a
two-stage process describing transfer of steam condensate bubbles
across bitumen interface and encapsulation of condensate bubbles
in bitumen. Previous authors have shown that the formation of
WO emulsion is an exothermic process (Moulik et al. 1992; Batov
and Kartsev 2006), which has been adopted in our model.
Numerical Simulations
In this work, emulsification influences SAGD performance by
three mechanisms: alteration of bitumen viscosity, enthalpy of
emulsification and alteration of effective thermal properties of the
oleic phase, and modification of relative flow potential of bitumen
at any water saturation.
In-Situ Emulsification. We approximate in-situ emulsification
with a two-step pseudochemical reaction process assuming suffi-
cient and instantaneous hydrodynamic instability at bitumen/con-
densate-bubble interfaces. Generated emulsion is also assumed to
be stabilized by natural emulsifiers in the continuous phase. These
assumptions are supported by experimental evidence that indi-
cates significant instability at steam-chamber edges (Sasaki et al.
2001a, b) and high asphalthene content of heavy oils and bitumen
(Zhao et al. 2009). Eq. 1 describes transfer of condensate micro-
bubbles across bitumen interface. Heat release into bitumen, asso-
ciated with emulsion generation, is described as enthalpy of
emulsification and is corroborated by experimental studies (Mou-
lik et al. 1992; Batov and Kartsev 2006). In Eq. 1, the stoichio-
metric constants X and Y are such that mass is conserved:
X ðH
2
OÞ ! YðEmulsiÞ ½DH
emulsification
Š; ð1Þ
where DH
emulsification
is the enthalpy of emulsification that is in-
dicative of an exothermic process.
The second step models the engulfing or encapsulation of con-
densate bubbles by bitumen (associated with a complex mixing
dependent on hydrodynamic, thermal, and chemical variables) to
form an emulsion with a known volume fraction of the dispersed
phase, as follows:
AðEmulsiÞ þ BðBitumenÞ ! DðWO
Emuls
Þ ð2Þ
The stoichiometric constants A and B represent the number of
moles of each component required to produce an emulsion of an a
priori known volume fraction. Experimental analysis should pro-
vide reliable values of average volume fraction of dispersed phase
representative of emulsions for case-specific SAGD operation. D
is a reaction-product coefficient that ensures mass conservation
and depends on the molecular mass of produced emulsion. Ther-
mophysical properties of emulsion are strongly dependent on the
volume fraction of dispersed phase in the continuous phase. Eq. 2
must satisfy conditions of mass conservation.
Transfer of steam-condensate microbubbles into bitumen is
assumed to occur instantaneously—initiation does not require
critical activation energy. The reactions in Eqs. 1 and 2 are mod-
eled as first-order reactions dependent only on emulsion concen-
tration; reaction order may be case dependent. In Eq. 2, because
the molecular mass of bitumen is significantly larger than that of
steam-condensate bubbles with which it forms an emulsion, more
condensate microbubbles are required to generate emulsion for
each mole of bitumen. Therefore, Eq. 1 (i.e. transfer of condensate
microbubbles across bitumen interface) is the rate-limiting pro-
cess. Emulsion is assumed to be very stable, such that coalescence
can be ignored. The abundance of natural emulsifiers in heavy oil
and bitumen makes this assumption practicable.
Viscosity of Emulsions. The viscosity behavior of WO emul-
sions is complex and is influenced by several factors, including
crude components, temperature, viscosities of oil and water, sur-
factant concentration, fines, and shear rate. Viscosity of WO
emulsion is typically higher than that of oil or water. A WO emul-
sion becomes increasingly unstable at high water volume fraction
and can “invert” to OW emulsion at very high water fractions of
approximately 80% and greater (Kokal 2005). In this work, we
model viscosity of emulsion on the basis of Taylor’s (1932) em-
pirical relationship (Taylor 1932):
l
Ã
l
¼ 1 þ 2:5
/ þ 0:4
/ þ 1
_ _ _ _
v; ð3Þ
which estimates emulsion viscosity (l*) as a function of dis-
persed-phase viscosity, continuous-phase viscosity (l), ratio of
dispersed- to continuous-phase viscosity (/), and dispersed-phase
volume fraction (). We incorporate emulsion viscosity in our nu-
merical simulation assuming a priori knowledge of representative
volume fractions of the composite phases in the emulsion. Emul-
sion viscosity predicted by Taylor’s equation for a specified dis-
persed-phase volume fraction is calculated at each temperature
step for an appropriate temperature range based on initial reser-
voir temperature and steam-injection temperature. Fig. 2 shows
the magnitude of deviation of emulsion viscosity calculated by
Taylor’s model for heavy oil of Sasaki et al. (2001b), at 100

C.
The effective viscosity of emulsion mixture with the oleic phase
(l
mix
) is then calculated by using a linear mixing rule, dependent
on the concentration of emulsion component in the oil phase:
lnl
mix
¼ xlnðl
Ã
Þ þ ð1 À xÞlnðlÞ ð4Þ
where x is the mole fraction of emulsion in the oleic phase and l
is oil viscosity.
. . . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
SAGD
performance
Geomechanics
Emulsification
Steam
fingering
Mass transfer
(molecular and
dispersion)
Heat transfer
(convection
and
conduction)
Multiphase
flow
(Kr and IFT)
Co-current and
counter-
current flow
Fig. 1—A summary of complex processes observed during
SAGD experiments that are key to accurate numerical model-
ing. These processes affect SAGD performance to varying
degrees, depending on factors such as operational conditions,
and are interrelated in several cases.
June 2013 SPE Journal 441
Multiphase-Flow Behaviors in the Presence of Emulsion. In
addition to viscosity influence, formation of emulsion alters pore-
level flow such that the influence of steam condensate on oil flow
path is reduced. At the pore level, emulsification leads to a reduc-
tion of the thickness of condensate flow path. A fraction of water
volume that would otherwise be flowing as a separate phase is
now encapsulated within the oleic phase. As a result, multiphase
effect is minimized or eliminated, and effective oil relative flow
at any water saturation within the emulsion layer should be
higher. A pore-level visualization of the concept described here is
presented in Fig. 3. In this work, emulsion flow is calculated with
bitumen relative permeability. In this regard, the WO emulsion is
assumed to maintain a Darcy flow, which is strictly for Newtonian
fluids. This is a reasonable assumption because experimental evi-
dence shows that 50% or more volume fraction of dispersed phase
is required for non-Newtonian flow of emulsion to occur (Soo and
Radke 1984). In our numerical model, emulsion is produced at the
production well as a component of the oleic phase and volume
fractions of the reactants (which constitute the emulsion) defined
in Eq. 2 are used to decouple produced emulsion into bitumen and
condensate volumes. For cumulative produced emulsion (Q
emul-
sion
), constituent bitumen volume (Q
b
) is calculated by
Q
b
¼ Q
emulsion
n
b
M
b
q
b

n
i¼1
n
i
M
i
q
i
_
¸
¸
¸
_
_
¸
¸
¸
_
ð5Þ
where M is molecular mass, q is density, n is number of moles
taken from the stoichiometry in Eq. 2, and subscripts b and i rep-
resent bitumen and reactants counters of emulsion component,
respectively.
Simulation-Model Design. We used the experimental data in
Sasaki et al. (2001b) as an analog for our numerical studies. The
experiment was a 2D laboratory-scale SAGD with saturated steam
at 105

C (100% quality) injected through a horizontal well 10 cm
above a production well. No preheating was performed. The total
model dimensions were 38 Â38 Â4.45 cm (width Âheight Â
thickness, respectively). Effective reservoir dimensions were
30 Â30 Â0.45 cm, with the rest made of acrylic resin plates. Av-
erage porosity and permeability were 38% and 142 darcies,
respectively. Sustained production of emulsion (WO) was
reported after steam breakthrough into the producer well. Cumu-
lative produced oil at the end of the 550-min experiment was 64
cm
3
[i.e., 41.6% of oil originally in place (OOIP)]. In their subse-
quent numerical model of the experiment (Sasaki et al. 2001b),
the history-matched numerical model predicted a final cumulative
oil production of 49 cm
3
(31.8% of OOIP). In their history match,
the cumulative injected steam was constrained to the laboratory
values. They also used linear relative permeability curves. Never-
theless, their numerically predicted final cumulative oil recovery
from the history match was approximately 24% less than that of
the experimental data. Table 1 shows a summary of the experi-
mental rock/fluid data used for the construction of the Sasaki
et al. (2001b) numerical model. For the temperature dependence
of heavy-oil viscosity, see Sasaki et al. (2001b).
We constructed a numerical representation of the Sasaki et al.
(2001b) experimental model in a commercial thermal reservoir
simulator (STARS 2011). The discretized model was 341 Â
52 Â341 cm (i-, j-, and k-directions, respectively). Similar to the
experiment, the effective reservoir block was 30 Â0.45 Â30 cm,
and both wells were placed centrally in the model, the production
well being approximately 10 cm below the injection well. The
same fluid, rock, endpoints, and initial properties used in Sasaki
et al. (2001b), presented in Table 1, were adopted in our simula-
tions. The steam-injection rate into the top well was constrained
to experimental injection data, whereas the production well was
constrained by a minimum bottomhole pressure equal to 101 kPa.
The fluid model, summarized in Table 2, consisted of three
phases (gaseous, oleic, and aqueous) with four components. Com-
ponent “Emulsi” represents steam condensate microbubbles that
have been transferred from aqueous to oleic phase because of
hydrodynamic instability and can therefore be reasonably ap-
proximated with steam properties. Emulsion formed by high-
energy mixing of water and oil is represented by the “WO
Emuls

component in oleic phase, and Eq. 3 describes its viscosity.
The gas/liquid equilibrium correlation and constants used for
WO emulsion partitioning into oil are presented in Table 3. Stoi-
chiometric constants of Eqs. 1 and 2 used for history matching
of the experimental data in Sasaki et al. (2001b) are presented in
Table 4.
Results and Discussions
History Match. Traditional SAGD. The first part of our simula-
tions explored the possibility of improving numerical calculated
oil production by history matching of experimental data in Sasaki
et al. (2001b) using the traditional SAGD-modeling approach
(i.e., excluding the emulsion model). Previous authors (Azom and
Srinivasan 2009) investigated the sensitivity of several rock and
. . . . . . . . . . . . . . . . . . . . .
0 0.1 0.2
Dispersed Water Volume Fraction
V
i
s
c
o
s
i
t
y

a
t

1
0
0
º
C
,

(
c
P
)
Oil Viscosity
Emulsion Viscosity
0.3 0.4 0.5
200
220
240
260
280
300
320
340
360
Fig. 2—Viscosity of single oil phase and emulsion phase [based
on Taylor (1932)] at different dispersed-phase volume fractions
and 100
º
C. Considering that several SAGD experiments
reported WO emulsion, we restrict water fraction evaluated in
our simulation to a value less than the expected fraction for
inversion (typically greater than 50%water fraction).
Mobile Oil Mobile Oil
Steam
(a) (b)
Steam
Steam Condensate
Steam Condensate
Rock Grain
Rock Grain
WO-Emulsion
Fig. 3—Schematic of pore-scale fluid occupancies: (a) emulsion SAGD, (b) traditional concept of SAGD.
442 June 2013 SPE Journal
fluid parameters, including absolute permeability, porosity, and
heat-transfer properties on cumulative-oil-production data of
Sasaki et al. (2001b). They observed that no reasonable match
could be obtained even with a 200% change of these parameters
from the measured values. Similar sensitivities are therefore not
presented in this paper. However, considering the reasonable
uncertainty associated with relative permeability measurements,
sensitivity of cumulative produced oil to relative permeability is
presented. A comparison of numerical simulation results for dif-
ferent relative permeability curves is shown in Fig. 4. As reported
in Sasaki et al. (2001b), there is a significant sensitivity of cumu-
lative oil production to the shape of relative permeability curves.
However, nonlinear relative permeability curves produce the clos-
est match to the final value of experimental data. It is therefore
evident that traditional SAGD modeling is deficient at capturing
important physics of the experiment and therefore produces a
poor prediction of experimental data. In addition to incorporating
the WO-emulsion model, subsequent sections used the nonlinear
relative permeability curves with nonzero endpoint for numerical
simulations.
Emulsion SAGD. History matching of the Sasaki et al.
(2001b) experiment with the WO-emulsion SAGD model also
used experimental injection data to constrain the injection well.
Production well had a minimum bottomhole pressure of 101 kPa.
Fig. 5 shows spatial distributions of gas saturation, oil saturation,
and experimental steam-chamber images at different times. Simu-
lated gas- and oil-saturation distributions excellently predict key
attributes of the experimental steam chamber: lateral and vertical
growth rate. In Fig. 6, the spatial locations of numerical emulsions
and temperature profiles are presented at different times. Experi-
mental temperature image at 240 minutes is also shown in Fig. 6.
This figure shows that emulsion is the dominant component at
simulated steam-chamber edges. This is also evident in experi-
mental images of Fig. 5. As described in Fig. 3, multiphase effects
caused by water partitioning into bitumen to form WO emulsion
enhance the effective flow of bitumen. In addition, emulsification
enhances thermal effect on perimeter bitumen by enthalpy of
emulsification. Fig. 6 also shows that the rate of growth of numer-
ical temperature profile in the vertical and lateral directions com-
pares strongly with experimental observation.
In Fig. 7, experimental cumulative oil production is compared
with that from different numerical SAGD models. Clearly, the
WO-emulsion SAGD model gives an excellent match of both the
change in slopes and final cumulative oil recovery. As reported in
Sasaki et al. (2001b), the SAGD model without emulsion under-
predicts experimental cumulative oil production by approximately
TABLE 1—SUMMARY OF EXPERIMENTAL PARAMETERS
(SASAKI ET AL. 2001B) USED FOR BASE-SAGD NUMERICAL
SIMULATION
Parameters
Heavy Oil
Molecular weight 490 g/gmol
Density 0.998 g/cm
3
Compressibility 7 Â10
À7
kPa
À1
Thermal-expansion coefficient 6Â10
À4
C
À1
Heat capacity 411.7 J/gmolÁ

C
Critical pressure 1115 kPa
Critical temperature 494

C
Capillary pressure 0.0 kPa
Phase equilibrium constant 0.0
Properties of Physical Model
Porosity 0.38
Average permeability 142Â10
5
md
Thermal conductivity, k 0.70 J/cmÁminÁ

C
Volumetric heat capacity 1.99 J/cm
3
Á

C
Resin thermal conductivity 0.13 J/cmÁminÁ

C
Resin volumetric heat capacity 1.67 J/cm
3
Á

C
Overall convective-heat-transfer
coefficient (resin to
surrounding air)
0.041 J/cm
2
ÁminÁ

C
Initial Conditions/Saturation Endpoints
Temperature 20.0

C
Pressure 101.3 kPa
S
o
1.0
S
w
0.0
S
g
0.0
Reference pressure 101.3 kPa
Reference temperature 20.0

C
S
or
0.05
S
wc
0.10
S
gc
0.05
TABLE 2—FLUID MODEL DESIGN FOR NUMERICAL SAGD
SIMULATIONS AND GAS/LIQUID PARAMETER
FOR WO EMULSION
Components Aqueous Oleic Gaseous
H
2
O 3 3
Bitumen 3
Emulsi 3 3
WO
Emuls
3 3
TABLE 3—GAS/LIQUID EQUILIBRIUMCONSTANT
FOR WO
Emuls
Units Correlation (Reid et al. 1977) Values
T in

C; p in kPa K ¼
k
v1
p
e
k
v4
TÀk
v5 k
v1
1 Â10
7
k
v4
À4300
k
v5
À206
TABLE 4—REACTIONS
Equation
Stoichiometric
Constants Values
Rate
Constant
(minutes
À1
)
DH
emulsification
(kJ/mol)
1 X; Y 1.0; 1.0 0.011 33.5
2 A 82.50 110 0.0
2 B 1.0
2 D 83.50
0 100
0
0
0.2
0.2
0.4
0.4
0.6
0.6
0.8
0.8
1
1
Saturation (Water, Liquid)
0
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0.2
K
row
K
rog
K
rw
K
rg
K
rg
;K
row
K
rg
;K
row
K
rw
;K
rog
K
rw
;K
rog
0.4 0.6 0.8 1
Saturation (Water, Liquid)
0 0.2 0.4 0.6 0.8 1
Saturation (Water, Liquid)
R
e
la
t
iv
e

P
e
r
m
e
a
b
ilit
y
,

K
rR
e
la
t
iv
e

P
e
r
m
e
a
b
ilit
y
,

K
r
0
0.2
0.4
0.6
0.8
1
R
e
la
t
iv
e

P
e
r
m
e
a
b
ilit
y
,

K
r
0
20
40
80
60
300 200
Time (min)
C
u
m
u
l
a
t
i
v
e

O
i
l

(
B
I
T
U
M
E
N
)

S
C

(
c
m
3
)
400 500 600
Fig. 4—Cumulative oil production from traditional SAGD for dif-
ferent relative permeability curves.
June 2013 SPE Journal 443
24%. This strongly suggests that the numerical emulsion model-
ing procedure presented in this work provides a good representa-
tion of the physics of emulsification and flow associated with the
experimental SAGD. Recall that in the preceding sections, we
described three key factors of interest: in-situ emulsification, vis-
cosity modification, and multiphase flow. Emulsification leads to
an increase in the effective viscosity of bitumen. However, at the
steam-injection temperature for this simulation (and in the major-
ity of SAGD operations), oil viscosity is sufficiently low, as
shown in Fig. 2. Consequently, the influence of emulsion viscosity
on SAGD performance is dominated by the multiphase-flow
effect that now allows bitumen (in emulsion form) to flow at high
relative permeability in the presence of larger water saturation. In
summary, this result appears somewhat counterintuitive until one
considers the multiphase effect (comparable with foamy oil, in
which large saturation of disconnected gas has little influence on
oil relative permeability). Moreover, numerical SAGD cumulative
oil production is strongly sensitive to relative permeability satura-
tion endpoints. As a result, good care must be exercised during
experimental measurement of relative permeability saturation
endpoint values. Therefore, the uniqueness of SAGD history
match is improved by avoiding a significant altering of endpoint
values. Relative permeability saturation endpoint values presented
in Table 1 were therefore honored throughout our numerical his-
tory matching. Fig. 8 compares experimental oil-production rate
with that from different numerical models. Again, the WO-emul-
sion model shows an excellent agreement with experimental data.
Experimental cumulative water production is compared with dif-
ferent SAGD models in Fig. 9 and shows a satisfactory match
with numerical simulations. It can be observed that the WO-emul-
sion model also outperformed traditional SAGD in terms of
cSOR.
We investigated the effect of dispersed- (water-) phase volume
fraction of WO emulsion on numerical SAGD with emulsion
flow. To clarify, stoichiometric coefficients depend on volume
fraction of water in the emulsion considered, particularly because
the molecular weight of emulsion is expected to differ from that
of the individual components. In our simulation, molecular weight
of emulsion is calculated from the mole fraction of the reactants,
and resultant molar weight of emulsion approaches that of water
as the water volume fraction in emulsion increases. Results of cu-
mulative oil production are shown in Fig. 10. For the range of
water volume fraction presented, final cumulative oil production
decreased with an increase in water volume fraction. An increase
Time
(min)
Gas Saturation Oil Saturation
Experimental
Steam Chamber
170
0.00
0.20
0.40
0.60
0.80
1.00
0.00
0.20
0.40
0.60
0.80
1.00
240
0.00
0.20
0.40
0.60
0.80
1.00
0.00
0.20
0.40
0.60
0.80
1.00
360
0.00
0.20
0.40
0.60
0.80
1.00
0.00
0.20
0.40
0.60
0.80
1.00
480
0.00
0.20
0.40
0.60
0.80
1.00
0.00
0.20
0.40
0.60
0.80
1.00
Fig. 5—Spatiotemporal images of simulated gas saturation, oil saturation, experimental steam chamber. Numerical simulation
based on history-match SAGD experiment coupled with an emulsion model. (Experimental images courtesy of Sasaki et al., VC
2001
Society of Petroleum Engineers.)
444 June 2013 SPE Journal
Time
(min)
Global Mole Fraction
(WOEmuls)
Temperature (C)
Temperature (C)
Experimental
170
0.00
0.20
0.40
0.60
0.80
1.00
20
37
54
70
87
104
240
0.00
0.20
0.40
0.60
0.80
1.00
20
37
54
70
87
104
360
0.00
0.20
0.40
0.60
0.80
1.00
20
37
54
70
87
104
480
0.00
0.20
0.40
0.60
0.80
1.00
20
37
54
70
87
104
240 min
Fig. 6—Spatiotemporal images of simulated global mole fraction (WO
Emuls
), temperature, and experimental temperature. Numerical
simulation based on history-match SAGD experiment coupled with an emulsion model. (Experimental images courtesy of Sasaki
et al., VC
2001 Society of Petroleum Engineers.)
Time (min)
C
u
m
u
l
a
t
i
v
e

O
i
l

S
C

(
c
m
3
)
0 100 200 300 400 500 600
0
20
40
60
80
Azom and Srinivasan
Emulsion-SAGD
SAGD
Experiment (Sasaki et al.)
Fig. 7—Comparison of different numerical models’ calculated
cumulative oil production with experimental data of Sasaki
et al. (2001b).
Time (min)
0
0
0.2
1.2
1.4
1
0.4
0.6
0.8
100 200 300 400 500 600
O
i
l

R
a
t
e

S
C

(
c
m
3
/
m
i
n
)
Azom and Srinivasan
Emulsion-SAGD
SAGD
Experiment (Sasaki et al.)
Fig. 8—Comparison of different numerical models’ calculated
oil-production rate with experimental data of Sasaki et al.
(2001b).
June 2013 SPE Journal 445
in water volume fraction influences simulations in two main
ways: It increases emulsion viscosity and effective bitumen vis-
cosity and increases the volume of water produced with emulsion,
thereby reducing the oil content of produced emulsion. However,
because oil viscosity at steam-injection temperature was suffi-
ciently low (see Fig. 2), a 40% increase in water volume fraction
did not result in a 40% decrease in cumulative oil production.
Conclusions
We presented an approach to incorporate WO-emulsion physics
within SAGD numerical simulations. Key phenomena accounted
for in the emulsion model are viscosity effect, energy release dur-
ing emulsification, and multiphase effect. Numerical simulation
results produced by use of our model showed excellent agreement
with SAGD experimental data. Specific conclusions from this
work include
• The physics of WO emulsion during SAGD can be satisfacto-
rily integrated within commercial reservoir simulation models.
• Accounting for viscosity alteration, multiphase effect, and en-
thalpy of emulsification is a promising mechanistic approach to
sufficiently describe emulsion physics during SAGD, particu-
larly in very-high-permeability reservoirs. In this respect, large
permeability values (on the order of 10 darcies) similar to those
of Athabasca bitumen reservoirs allow dispersed-phase emul-
sion bubbles to freely flow through pore throats in the porous
media.
• Although emulsification poses an important separation problem
at oil-production plants, it appears to play a vital role within the
reservoir during SAGD—increasing bitumen mobility and
thereby decreasing cSOR.
• We therefore recommend that, subject to a rigorous effort to
accurately measure endpoint relative permeability and satura-
tion values, the degree of emulsification expected in specific
SAGD operations should be evaluated and incorporated in sim-
ulations against modifications of relative permeability and end-
point values.
Nomenclature
A ¼ stoichiometric constant
B ¼ stoichiometric constant
M ¼ molecular mass
n ¼ number of moles taken from the stoichiometry
in Eq. 2
Q
b
¼ constituent bitumen volume
Q
emulsion
¼ cumulative produced emulsion
S
g
¼ gas saturation, fraction
S
gc
¼ critical gas saturation, fraction
S
o
¼ oil saturation, fraction
S
or
¼ residual oil saturation, fraction
S
w
¼ water saturation, fraction
S
wc
¼ connate water saturation, fraction
x ¼ mole fraction of emulsion in the oleic phase
X ¼ stoichiometric constant
Y ¼ stoichiometric constant
DH
emulsification
¼ enthalpy of emulsification
l ¼ continuous-phase viscosity
l* ¼ emulsion viscosity
l
mix
¼ effective viscosity of emulsion mixture with the
oleic phase
¼ dispersed-phase volume fraction
q ¼ density
/ ¼ ratio of dispersed- to continuous-phase viscosity
Subscripts
b ¼ bitumen counter of emulsion component
i ¼ reactants counter of emulsion component
Acknowledgments
The authors acknowledge Computer Modeling Group for use of
its thermal reservoir simulator and the Natural Sciences and Engi-
neering Research Council of Canada for funding support.
References
Azom, P. and Srinivasan, S. 2009. Mechanistic Modeling of Emulsion
Formation and Heat Transfer During the Steam-Assisted Gravity
Drainage (SAGD) Process. Paper presented at the SPE Annual Techni-
cal Conference and Exhibition, New Orleans, Louisiana, 4–7 October.
http://dx.doi.org/10.2118/124930-MS.
Bancroft, W.D. 1913. The Theory of Emulsification V. J Phys Chem 17
(6): 501–519. http://dx.doi.org/10.1021/j150141a002.
Batov, D.V. and Kartsev, V.N. 2006. Formation Enthalpies of Water-o-xy-
lene-triton X-100 Microemulsion at 298.15 K. Colloid J. 68 (5):
648–650. http://dx.doi.org/10.1134/S1061933X06050188.
Bosch, R., Axcell, E., Little, V., et al. 2004. A Novel Approach for
Resolving Reverse Emulsions in SAGD Production Systems. Can.
J. Chem. Eng. 82 (4): 836–839. http://dx.doi.org/10.1002/cjce.
5450820424.
Chung, K.H., and Butler, R.M. 1987. Geometrical Effect of Steam Injec-
tion on the Formation of Emulsions in the Steam-assisted Gravity
Drainage Process. Paper SPE 873822 presented at SPE Annual Techni-
cal Meeting, Calgary, Alberta, Canada, 7–10 June. http://dx.doi.org/
10.2118/87-38-22.
Time (min)
0
0
100
100
200
200
300
300
400
400
500
500
600
600
C
u
m
u
l
a
t
i
v
e

W
a
t
e
r

S
C

(
c
m
3
)
Emulsion-SAGD
SAGD
Experiment (Sasaki et al.)
Fig. 9—Comparison of different numerical models’ calculated
cumulative-water-production data with experimental data of
Sasaki et al. (2001b).
Time (min)
0
0
10
20
30
40
50
60
70
100
40%_Water
20%_Water
30%_Water
10%_Water
200 300 400 500 600
C
u
m
u
l
a
t
i
v
e

O
i
l

S
C

(
c
m
3
)
Fig. 10—Emulsion SAGD numerical simulation, with cumula-
tive-oil-production plot for different in-situ volume fractions of
water (dispersed phase) in WO emulsion.
446 June 2013 SPE Journal
Dusseault, M.B. 2002. Cold Heavy Oil Production with Sand in the Cana-
dian Heavy Oil Industry. Edmonton, Alberta, Canada: Alberta Ministry
of Energy. http://www.energy.alberta.ca/OilSands/pdfs/RPT_Chops_
chptr2.pdf.
Kokal, S. 2005. Crude-Oil Emulsions: A State-of-the-Art Review. SPE
Prod & Fac 20 (1): 5–13. http://dx.doi.org/10.2118/77497-PA.
Mehrotra, A.K. and Svrcek, W.Y. 1986. Viscosity of Compressed Atha-
basca Bitumen. Can. J. Chem. Eng. 64 (5): 844–847. http://dx.doi.org/
10.1002/cjce.5450640520.
Mohammadzadeh, O. and Chatzis, I. 2009. Pore-level Investigation of
Heavy Oil Recovery Using Steam-assisted Gravity Drainage (SAGD).
Paper SPE 13403 presented at the International Petroleum Technology
Conference, Doha, Qatar, 7–9 December. http://dx.doi.org/10.2523/
13403-MS.
Mohammadzadeh, O., Rezaei, N., and Chatzis, I. 2010. Pore-level Investi-
gation of Heavy Oil and Bitumen Recovery Using Solvent-aided
Steam-assisted Gravity Drainage (SA-SAGD) Process. Energ. Fuel 24
(12): 6327–6345. http://dx.doi.org/10.1021/ef100621s.
Moulik, S.P., Das, M.L., and Bhattacharya, P.K. 1992. Thermodynamics
of Microemulsion Formation. 1. Enthalpy of Solution of Water in Bi-
nary (Triton X 100 þButanol) and Ternary (Heptane þTriton X
100 þButanol) Mixtures and Heat Capacity of Resulting Systems.
Langmuir 8 (9): 2135–2139. http://dx.doi.org/10.1021/la00045a012.
Nasr, T.N., Law, D.H., Golbeck, H., et al. 2000. Counter-current Aspect
of the SAGD Process. J. Cdn. Pet. Tech. 39 (1): 1–7. http://dx.doi.org/
10.2118/00-01-03.
Noik, C., Dalmazzone, C., Goulay, C., et al. 2005. Characterization and
Emulsion Behaviour of Athabasca Extra Heavy Oil Produced by
SAGD. Paper SPE 97748 presented at the SPE/PS-CIM/CHOA Inter-
national Thermal Operations and Heavy Oil Symposium, Calgary,
Alberta, Canada, 1–3 November. http://dx.doi.org/10.2118/97748-MS.
Reid, R.C., Prausnitz, J.M., and Sherwood, T.K. 1977. The Properties of
Gases and Liquids, third edition. New York: McGraw-Hill.
Sasaki, K., Akibayashi, S., Yazawa, N., Doan, Q.T., and Farouq, A.
2001a. Experimental Modeling of the SAGD Process – Enhancing
SAGD Performance with Periodic Stimulation of the Horizontal Pro-
ducer. SPE J 6 (1): 89–97. http://dx.doi.org/10.2118/69742-PA.
Sasaki, K., Akibayashi, S., Yazawa, N., et al. 2001b. Numerical and Ex-
perimental Modeling of the Steam-assisted Gravity Drainage (SAGD)
Process. J. Cdn. Pet. Tech. 40 (1): 45–50. http://dx.doi.org/10.2118/
01-01-04.
Soo, H. and Radke, C. 1984. The Flow Mechanism of Dilute, Stable Emul-
sions in Porous Media. Ind. Eng. Chem. Fundam. 23 (3): 342–347.
http://dx.doi.org/10.1021/i100015a014.
STARS Users Manual, Version 2010.12. 2011. Calgary, Alberta, Canada:
Computer Modeling Group (CMG) Ltd.
Taylor, G.I. 1932. The Viscosity of a Fluid Containing Small Droplets of
Another Fluid. Proc. R. Soc. Lond. A 138 (834): 41–48. http://
dx.doi.org/10.1098/rspa.1932.0169.
Yaghi, B. and Al-Bemani, A. 2002. Heavy Crude Oil Viscosity Reduction
for Pipeline Transportation. Energ. Source 24 (2): 93–102. http://
dx.doi.org/10.1080/00908310252774417.
Zhao, B., Becerra, M., and Shaw, J.M. 2009. On Asphaltene and Resin
Association in Athabasca Bitumen and Maya Crude Oil. Energ. Fuel
23 (9): 4431–4437. http://dx.doi.org/10.1021/ef900309j.
Cosmas C. Ezeuko is a post-doctoral fellow at the Department
of Chemical and Petroleum Engineering, University of Calgary.
His current work exploits a blend of industry experience and
research expertise and focuses on investigation of processes
affecting SAGD and expanding solvent-SAGD performances,
modelingof biofilmformation damage andbiocide treatments,
andcharacterization and reservoir simulation of thermal EOR re-
covery from complex heavy-oil carbonates. Ezeuko holds a
BEng from the University of Benin and MSc and PhD degrees in
petroleumengineering fromHeriot-Watt University. He is a mem-
ber of SPE andaregisteredprofessional engineer inAlberta.
Jingyi (Jacky) Wang is a research engineer in the Department
of Chemical and Petroleum Engineering at the University of
Calgary, specializing in unconventional resource recovery pro-
cess design and numerical simulation. He has more than 10
years of combined chemical-engineering and reservoir-engi-
neering experience working with industry and research, with
expertise in the areas of thermal reservoir simulation, cold pro-
duction, vapor extraction, CO
2
sequestration, hydrate recov-
ery, reactive numerical modeling, enhanced oil recovery, and
production optimization. Wang holds a BSc degree in chemi-
cal engineering from East China University of Science and
Technology and an MEng degree in reservoir engineering from
the University of Calgary. He is a member of the Association of
Professional Engineers and Geoscientists of Alberta and of SPE.
Ian D. Gates is an associate professor in the Department of
Chemical and Petroleum Engineering at the University of Cal-
gary. He spent 7 years in the industry before joining the Univer-
sity. His primary research interests are in thermal and thermal/
solvent methods and optimization of these technologies,
application of smart-well technologies for adaptive produc-
tion of heavy-oil fields, support of vector machine learning for
reservoir characterization, in-situ gasification (in-reservoir reac-
tion engineering), biofilm development in porous media, and
coal gasification. Gates holds a BSc degree from the University
of Calgary, an MASc degree from the University of British Co-
lumbia, and a PhD degree from the University of Minnesota, all
in chemical engineering. He is a registered professional engi-
neer in Alberta.
June 2013 SPE Journal 447