Documentos de Académico
Documentos de Profesional
Documentos de Cultura
Summary outward, losing its latent heat when it comes into contact with the
SAGD is one successful thermal recovery technique applied in cold bitumen at the edge of a depletion chamber. As a conse-
the Athabasca and Peace River reservoirs in central and northern quence, the viscosity of bitumen falls several orders of magnitude,
Alberta, Canada. In SAGD, steam is injected into a horizontal and the bitumen flows under gravity toward a horizontal produc-
injection well and is forced outward, losing its latent heat when it tion well located several meters below and parallel to the injection
comes into contact with the cold bitumen at the edge of a deple- well (i.e., 5 m, but drilling tolerances often leave variations
tion chamber. As a consequence, the viscosity of the bitumen falls between 3 and 7 m). As the oil flows away and is produced, the
several orders of magnitude, its mobility rises several orders of steam chamber expands both upward and sideways (see Sections
magnitude, and then it flows under gravity toward a horizontal B and C in Fig. 1a). A cross-section of the SAGD process is dis-
production well located several meters below and parallel to the played in Fig. 1a. Section A shows the circulation stage, Section
injection well. Heat-transfer mechanisms are pivotal to the SAGD B presents the early phase in which the chamber is not well devel-
process. Though heat energy is transferred from steam to reservoir oped, and Section C presents the mature steam chamber in the
by conduction and convection, heat transfer by convection is not injection phase.
considered in the classic SAGD mathematical models such as The term steam-assisted gravity drainage was first devel-
Butlers. Researchers such as Butler and Stephens (1981), Reis oped by Roger Butler and his colleagues at Canadas Imperial Oil
(1992), Akin (2005), Liang (2005), Nukhaev et al. (2006), and in the late 1970s (Al-Bahlani and Babadagli 2009). Butler and
Azad and Chalaturnyk (2010) considered conduction from steam Stephens (1981) proposed the first closed-form solution for the
to cold reservoir to be the only heat-transfer component. How- prediction of oil production rate in the SAGD process. In his
ever, because the heat capacity of water is typically two to five model, known as the Butler theory, Butler described the SAGD
times that of bitumen, convection caused by the mobile conden- process as when steam is injected, a steam-saturated zone, called
sate flow in the reservoir may contradict these studies. Farouq-Ali a steam-depletion chamber or simply a steam chamber is
(1997) was the first to criticize the assumption that there is only a formed, in which the temperature is that of the injected steam
thermal conduction mechanism in the SAGD process. He pointed (Tchamber Tsteam). The steam flows toward the interface of the
out that with so much condensate flowing, convection would be steam chamber, where it condenses and loses its latent heat by
expected to be the dominant heat-transfer mechanism, which can flashing to bitumen. The latent heat from steam is transferred by
be plausible at high temperatures. In response, Edmunds (1999a) thermal conduction into the surrounding reservoir and mobilizes
stated that on the basis of the associated change in enthalpy, the the bitumen. The steam condensate and mobile bitumen flow by
heat transfer into a cold reservoir because of convection is prob- gravity to the production well located below the injector from
ably less than 5% of that because of conduction. Ito (1999) chal- side-drained paths (see Section F in Fig. 1b).
lenged Edmunds (1999a) statement, on the basis of Ito and Suzuki The physics differs at the top and sides of the steam chamber.
(1996, 1999) and Ito et al. (1998), pointing out that this number, The top of the chamber rises because of steam fingering (ceiling
5%; i.e., ratio between convection to conduction presented by drainage) (Gotawala and Gates 2008), and the mobilized bitumen
Edmunds (1999a) is unrealistically low, (and) it should be in the falls from the sides of the steam finger, where steam rise usually
range of 50%. This study examined the relative roles of convec- impedes liquid drainage. Along the sides of the chamber, bitumen
tive and conductive heat transfer at the edge of SAGD steam mobility is controlled by both conduction and convection heat
chambers. In summary, the mathematical model developed in this transfer on the interface, and thus, condensate and mobilized bitu-
study considered both conduction and convection, and the result- men are drained by gravity drive (slope drainage) (Edmunds et al.
ant output from the model is reasonably consistent with published 1989, 1994; Nasr et al. 2000).
field data. This study supports the idea that although convection Further research includes that of Reis (1992), who modified
can dominate near the chamber edge in high-water-saturation res- Butlers model by revising the steam chamber configuration from
ervoirs, in bitumen-rich reservoirs, its contribution to heat transfer an S-curve shape suggested by Butler to a triangular shape (linear
is less than 1% and can be neglected. geometry). Akin (2005) modified Butlers model by including the
effects of steam distillation (hence, asphaltene deposition) and
Introduction considering viscosity reduction and production rate increase as
results of deasphalting. Liang (2005) presented an analytical model
SAGD is one successful thermal recovery technique applied in for cyclic steaming. Nukhaev et al. (2006) modified Butlers model
the Athabasca and Peace River reservoirs in Alberta, Canada. In to consider real-time production rate variation. Azad and Chalatur-
SAGD, steam injected into a horizontal injection well is forced nyk (2010) modified the work of Reis (1992) to suggest a semiana-
lytical model for considering geomechanical effects with a limit
*
Now with RPS Energy equilibrium model. They named this the model of slices.
Copyright V
C 2013 Society of Petroleum Engineers Though the SAGD process looks simple, there are some theo-
Original SPE manuscript received for review 23 January 2012. Revised manuscript received
retical pitfalls. All of the analytical models pertain to the flow of a
for review 13 August 2012. Paper (SPE 163079) peer approved 21 August 2012. single fluid (Farouq-Ali 1997). They assume constant steam
A B C
a. Cross section of SAGD process; Section A presents circulation phase, Section B presents early phase, and Section C presents
steam injection phase.
Caprock
Drained Flow
b. Cross section of s-shaped SAGD process in Butler model and this study; Section D presents circulation phase, Section E
presents early phase, and Section F presents steam injection phase.
Fig. 1Cross-section of SAGD process; Section A represents a real behavior, and Section B represents a behavior modeled in the
Butler model and this study. Cross section of SAGD process; Section A presents circulation phase, Section B presents early
phase, and Section C presents steam injection phase. Cross section of S-shaped SAGD process in Butler model and this study;
Section D presents circulation phase, Section E presents early phase, and Section F presents steam injection phase.
pressure inside the chamber (Farouq-Ali 1997), with no steam flow that can be done to raise it. In addition, surface area is not control-
outside of it (Farouq-Ali 1997); constant oil saturation equal to the lable in the SAGD process. Because of these limitations in con-
residual inside the steam chamber (Farouq-Ali 1997); and heat duction heat transfer, convective heat transfer and controlling
transfer ahead of the steam chamber by conduction only (Farouq- convection can enhance the thermal flux rate at the edge of steam
Ali 1997). None of these models include geomechanics in closed- chambers. Despite this, it remains unclear how large convection is
form solution. [It must be noted that the work of Azad and Chala- relative to conduction. Farouq-Ali (1997) criticized the idea that
turnyk (2010) must be programmed, and cannot be considered a only thermal conduction exists in SAGD, pointing out that with
closed-form mathematical solution.] Most analytical models deal so much condensate flowing, convection would be expected to be
with slope drainage, and little (aside from Das and Butler 1996) has the dominant heat-transfer mechanism; this is plausible at high
been published dealing with the mechanisms of drainage at the temperatures. Farouq-Ali (1997) also supported his idea by refer-
ceiling of the steam chamber (Al-Bahlani and Babadagli 2009). ring to Ito and Suzukis (1996) numerical simulations, which
A key control on production rate in the SAGD process is the show that convection is more important than conduction. This
heating of oil sands at the edge of the steam chamber (Sharma convection is associated with condensate carrying heat into the
and Gates 2011a); the higher the flux rate, the hotter and mobile oil zone. In response to Farouq-Alis (1997) critique,
more mobile the bitumen. In analytical models presented by most Edmunds (1999a) stated that on the basis of the associated change
researchers (e.g., Butler and Stephens 1981; Butler 1985, 1994b; in enthalpy, 22% of total heat transfer is carried by convection
Ferguson and Butler 1988; Reis 1992, 1993; Akin 2005; Liang (i.e., 18% by liquid water and 4% by mobile bitumen). Edmunds
2005; Nukhaev et al. 2006; Gotawala and Gates 2008; Azad and concluded that because of the water streamline being nearly per-
Chalaturnyk 2010), heat conduction is the major mode of heat pendicular to the temperature gradient (and almost parallel to iso-
transfer between the edge of steam chambers and the outer cold therms), the convective heat flux perpendicular to the steam
bitumen. On the basis of Sharma and Gates (2011a), conduction chamber edge must be reduced by the sine of the angle between
heat transfer at the edge of steam chamber can be enhanced by the streamlines and the isotherms. That is, if the condensate flows
raising steam chamber temperature, increasing the thermal con- exactly along the isotherms, there is zero net convection. He
ductivity of oil sands, and increasing heat-transfer surface area. finally concluded that heat transfer into cold reservoirs caused by
Because steam chamber temperature is a function of steam pres- convection is probably less than 5% of that caused by conduction.
sure (because steam chamber is at saturation condition), it is lim- Edmunds (2000) also emphasized that heat transfer in the mobile
ited by the range of practical steam pressure. The thermal zone is dominated by conduction, except very near the liner or
conductivity of cold oil sands is nearly constant, and there is little anywhere that live steam penetrates. Ito (1999) challenged the
e
tur
oil saturation (ROS); on the basis of the Butler model, the oil satura-
era
T tion inside the chamber at its residual]; and lint
st Te w is viscosity of water
mp
Di mpe at the steam interface [at steam temperature (on the basis of the But-
Te
str
ibu ratu ler model, the chamber including the interface temperature is equal
tio re to the steam temperature). Sharma and Gates (2011a) explained that
n
T their correlation is indirectly a function for the pressure difference
r
Steam Temperature = Tst Di required for convective flow, because given an original reservoir
sta
nc pressure, the higher the steam pressure is, the greater the corre-
e
Ux sponding steam temperature is, the lower the bitumen viscosity is,
and, consequently, the higher the oil rate and Ux are.
On the basis of Coreys relative permeability correlation, the
relative permeability of the water inside the reservoir can be cal-
Drained Flow
culated using the following equation:
r n
res ro Sw Swc
Initial Reservoir Temperature = Tr krw krw . . . . . . . . . . . . . . . . . . . 12
1 Swc Sor
Condensate Velocity Calculation The problem with Model 1 relates to condensate velocity eval-
uation, which is the main controller of convective heat transfer.
Because water mobility is substantially higher than that of bitumen
The Model 1 presented by Sharma and Gates (2011a) deals with
[the mobility ratio between water and bitumen is greater than 1,000,
the moving velocity (Ux) of a steam chamber interface as a moving
which means water mobility is three orders of magnitude greater
boundary problem, such as water and gas sweeping problems. The
than bitumen mobility (this will be discussed in detail in the begin-
steam chamber interface moving velocity (Ux) is mainly controlled
ning of the Model 3 subsection] and water condensate also flows at
by the rate of exposing a new oil sands element to steam hot inter-
higher rates than mobile oil on the surface edge [steam/oil ratio
face (ablation), which is, in turn, controlled by thickness and flow
(SOR) in SAGD projects are mostly above 3], condensate steam is
rate of the bitumen/condensate layer. The other debate about this
the major component of convection and mobile bitumen is less im-
model regards linear distribution of oil saturation beyond the
portant. It must be noted that the part of condensate velocity that is
steam chamber edge, which is not supported by any analytical
parallel to the temperature gradient (perpendicular to isotherms) is
work. Inconsistency is the main disadvantage of this model:
that responsible for convective heat transfer. Condensate velocity
Although Eq. 10 must be solved to evaluate temperature variation,
normal to the steam chamber edge (Vc) is the major controller and
in oil saturation calculations, the Butler model temperature varia-
only unknown in the convective heat-transfer evaluation. Conden-
tions were used, though these are calculated on the basis of con-
sate density and condensate heat capacity are known and nearly
ductive heat transfer only.
constant for the duration of SAGD processes. In this section, three
models are presented for condensate velocity evaluation.
Model 2. The second model is implemented in this study on the
basis of heat energy conservation. Heat conservation at the steam
Model 1. The first model is implemented by Sharma and Gates chamber interface is formulated as
(2011a). They suggest the following equation for evaluation of
condensate velocity normal to the steam chamber edge (Vc): Conductive Heat Flux Convective Heat Flux
kres =lres Total Heat Transferred Into Cold Bitumen 16
Vc Ux rw w
int =lint
. . . . . . . . . . . . . . . . . . . . . . . . . . 11
krw w
where
res
where krw is relative permeability of water (condensate) in the res-
ervoir, which is at connate water saturation; lres @T @T
w is viscosity of water Conductive Heat Flux K K . . . . . . . . . 17
int
(condensate) in the reservoir; krw is relative permeability of water at @x @n
Model 3. If water saturation is greater than irreducible water sat- where n is distance normal to the advancing front of the steam
uration, water mobility in Athabasca and Peace River bitumen chamber at any time. The condensate velocity normal to the steam
reservoirs is many times greater than that of bitumen because chamber interface is formulated as
Evaluated Temperature, C
ur
250
P
s
es
st
Pr
Pr
Di essu 200
str
ibu re
tio 150 krw = 0.05
n krw = 0.02
P
r 100
Di krw = 0.01
Steam Pressure = Pst sta
nc 50
e Butlers
Co Model
0
nv 0 2.5 5 7.5 10 12.5 15 17.5 20
ec
tio Distance Normal to Interface, m
Vc nD (B)
ire
Temperature Difference, C
cti 1E+2
on
1E+1
1
Initial Reservoir Pressure = Pr 1E1
1E2
1E3
Fig. 5Illustration of pressure distribution vs. advance of
krw = 0.001
krw= 0.0001
krw = 0.01
krw = 0.1
steam front. 1E4
1E5
1E6
kkrw @P 0 2.5 5 7.5 10 12.5 15 17.5 20
Vc . . . . . . . . . . . . . . . . . . . . . . . . . . 29 Distance Normal to Interface, m
lw @n
By differentiating Eq. 28, the condensate velocity normal Fig. 6Comparison of temperature distribution estimated on
to the steam chamber interface (Eq. 29) can be presented as the basis of Model 3 for different relative permeabilities. Section
follows: B shows temperature difference error caused by neglecting the
convective term. See Table 1 for fluid and reservoir properties.
kkrw @P /l c
Vc Ux / c Pst Pr exp w Ux n
lw @n kkrw the temperature profile with convection included) vs. Eq. 21 (i.e.,
Ux / c P Pr 30 the temperature profile that is based on the Butler model excluding
convection) is illustrated in Section A of Fig. 6; here, the differ-
For the steam chamber interface we can set n equal to zero and ence between Eq. 33 and Eq. 21 is shown for the possible range of
formulate the condensate velocity normal to the steam chamber reservoir properties. Section B of Fig. 6 shows the error in temper-
interface at the edge: ature profile including the condensate convection (i.e., using Eq.
33) and the Butler model (i.e., using Eq. 21) for several orders of
Vc Ux / c Pst Pr . . . . . . . . . . . . . . . . . . . . . . 31
magnitude of the relative permeability (i.e., from 0.0001 to 0.1).
For solving the heat transfer, the Vc is substituted by its value As shown in Section B of Fig. 6, the temperature difference,
from Eq. 30: including convection, is only noticeable for relative permeabilities
2 larger than 0.01. Accordingly, for the range of reservoir properties
@ T /lw c (i.e., relative permeabilities from 0.0001 to 0.01), the difference
K U x / c P st P r exp U x n qc cpc between Eqs. 21 (Butler model) and 33 is negligible.
@n2 kkrw
@T
Ux qr cpr 0 32 Conductive and Convective Heat-Transfer
@n
Calculation for Different Presented Models
On the basis of Appendix A, the solution for Eq. 32 is sug- The main focus of this study is to compare conductive and con-
gested in Eq. A-19, as follows: vective heat transfer. For this purpose the conductive and convec-
tive heat-transfer equations will be presented for different models.
Ux
n
X1
expj n ja
gn Model 1. Because Sharma and Gates (2011a) do not present the
T Tr n 0
n! Ux =ja n
T ( ) solution for temperature variation, Butlers model is used for tem-
Tst Tr ja X
1
1 perature variation in the conductive heat-transfer calculation:
gn
Ux n 1
n! Ux =ja n @T Ux
Conductive Heat Flux K K Tst Tr
Ux @n a
X
1 ja n
P Ux Ux
gn exp n K T Tr 36
n 0
n! Ux =ja n a a
( )
ja X1
1 33 The convective heat transfer can be calculated using the conden-
gn
Ux n 1 n! Ux =ja n sate velocity normal to the steam chamber edge (Vc) from Eq. 15:
krw = 0.0001
1.85
krw = 0.001
krw = 0.01
krw = 0.1
krw = 0.5
260 1.80
Pressure
1.70
240 1.65
1.60
230
1.55
220
1.50
Reservoir Pressure
210 0
0 2.5 5 7.5 10 12.5 15 17.5 20
Distance Normal to Interface, m
120
Condensate Velocity Normal to Interface
104
krw = 0.0001
krw = 0.001
krw = 0.01
krw = 0.1
100 krw = 0.5
80
60
5 105
40
20
105
0 0
0 2.5 5 7.5 10 12.5 15 17.5 20
Distance Normal to Interface, m
Fig. 7Comparison of pressure distribution calculated by theory for Model 3 (Eq. 28) for different relative permeabilities in Section
A; and comparison of condensate velocity normal to interface distribution calculated by theory for Model 3 (Eq. 30) for different
relative permeabilities in Section B. See Table 1 for fluid and reservoir properties.
enhanced convective flow near the interface. This may be why range between 0.001 and 0.01. The latter value is within the
Birrells (2001) results for high temperatures (i.e., near the steam convection-active zone near the chamber interface. The tempera-
chamber interface) are identical to heat flux curves associated ture enhancement for water relative permeabilities ranging
with high water saturations. between 0.001 and 0.01 is plausibly displayed only in log scale (see
Convective flux in Model 1 is dominated by the relationship Section B in Fig. 6), and the error ranges between 0.2 and 0.002%.
between water dynamic-viscosity variation and temperature. These The results from Fig. 6 show that although convection can
are compared in Section B in Fig. 8 for two different correlations enhance the thermal flux rate at the edge of the steam chamber, it
(i.e., Reid et al. 1977; Butler 1991). Multiple correlations can be happens over very short distances. For practical Athabasca reser-
used, because water dynamic-viscosity variation is not very sensitive. voir parameters, its effect is lower than 1%; this supports the
The heat flux caused by convection, although at times large at work of Edmunds (1999a, b). However, it does not contradict Ito
the interface, happens over very short distances. Fig. 6 compares and Suzuki (1996, 1999) and Ito et al. (1998), which pointed out
temperature profiles beyond the edge of the steam chamber. In the that convective heating mainly occurs above 140 C but is not the
conduction-only case, these are calculated using Butlers model, dominant heat-transfer mechanism. Convective heat transfer
whereas the case that couples conduction and convection is calcu- becomes minor below 100 C.
lated on the basis of Model 3 (Eq. 33). The results reveal that in Overall, Edmunds (1999a, b) argument on convective heat
water relative permeabilities ranging between 0.001 and 0.0001, transfer is valid, and the assumptions behind Butlers theory can
the temperature difference between Model 3 and Butlers model be ignored, causing errors less than 1% (i.e., that heat is trans-
ranges between 0.004 and 0.30 C. This difference increases ferred from the steam chamber to the cold oil zone by thermal
rapidly [i.e., 31 C (36%) for krw of 0.05, 17 C (19%) for krw of conduction alone, and that no steam condensate flows ahead of
0.02, and 9 C (9%) for krw of 0.01 (see Section A in Fig. 6)]. the steam chamber).
Although there are significant differences for higher water relative It must be noted that the theory of Model 3 assumes a stable
permeabilities, this happens only over small distances. The water steam/bitumen front. Steam fingering or instability can accelerate
relative permeabilities at ROS are approximately 0.02, reaching convection by inducing steam fingers, which penetrate the cold bi-
0.05 in extreme cases, but under reservoir conditions the water tumen at the edge of the chamber and increase channeling flow. A
relative permeabilities beyond the edge of the steam chamber steam-finger length is one of main unknowns in this phenomenon.
2
Conductive Heat Flux, W/m
Steam Temperature
Steam Temperature
40 40
krw = 0.05
krw = 0.01
krw = 0.005
krw = 0.001
krw = 0.0001
20 20
0 0
0 25 50 75 100 125 150 175 200 225 0 25 50 75 100 125 150 175 200 225
Temperature, C Temperature, C
50 100 150 200 250 300 350 400 50 100 150 200 250 300 350 400
Temperature, F
(B) (E) Temperature, F
40
60
50
30
UTF Field Data Sw , %
60-100 C
80 140 80 Model 1
40
85
65
50
30
20
2
2
Sw , %
Convective Heat Flux, W/m
150-275 C
Convection
krw = 0.0001
var. w (Butler, 1991)
120
krw = 0.005
krw = 0.001
60 60
krw = 0.05
krw = 0.01
Model 1 Increasing initial reservoir
water saturation
Steam Temperature
Steam Temperature
40 100 40
const. w
20 80 20
0 60 0
0 25 50 75 100 125 150 175 200 225 0 25 50 75 100 125 150 175 200 225
Temperature, C Temperature, C
50 100 150 200 250 300 350 400 50 100 150 200 250 300 350 400
Temperature, F Temperature, F
(C) (F)
100 100
85
60 65
50 krw = 0.01
80 var. w (Butler, 1991) 80 50
40 krw = 0.005
2
30 krw = 0.001
2
Total Heat Flux, W/m
60 60 30 20 krw = 0.0001
const. w
Total
Sw , % Sw , %
60-100 C 150-275 C
Steam Temperature
Steam Temperature
40 40
0 0
0 25 50 75 100 125 150 175 200 225 0 25 50 75 100 125 150 175 200 225
Temperature, C Temperature, C
50 100 150 200 250 300 350 400 50 100 150 200 250 300 350 400
Temperature, F Temperature, F
Fig. 8Comparison of conductive and convective heat-flux components estimated from field data obtained from vertical observa-
tion wells in Dover Phase B SAGD pilot vs. predictions calculated by Models 1 and 3 at different water relative permeabilities. See
Table 1 for fluid and reservoir properties.
It is suggested to be on the order of several meters by Sharma and of the depletion steam chamber and can be used with small errors
Gates (2011b), on the order of meters by Butler (1987, 1994a) and in final results.
Ito and Ipek (2005), and on the order of millimeters to tens of
centimeters by Gotawala and Gates (2008). There is a no doubt
on the positive effect of steam-fingering, but there are many Nomenclature
unknowns associated with fingering, such as the steam finger c compressibility of the oil sand, 1/Pa
lengths and the rate of mixing and, in turn, its effect on convective cpc condensate heat capacity, J/kg C
flux. The steam fingering and its enhancing effect on convective cpr reservoir heat capacity, J/kg C
flux should be studied in detail. k absolute permeability of the reservoir, m2
krw relative permeability of water, no unit
int
krw relative permeability of water at the interface, no unit
Conclusions res
krw relative permeability of water in the reservoir, no unit
This study examines the relative roles of convective and conduc- K reservoir thermal conductivity, W/m C
tive heat transfer at the edge of SAGD steam chambers. The results n the Corey coefficient, which sets the curvature of the rel-
demonstrate that convection can transfer a relatively large amount ative permeability curves, no unit
of heat at the edges of a steam chamber. However, temperature P fluid pressure, Pa
enhancement as a result of convection is minor, and is less than Pr reservoir pressure, Pa
1% under common Athabasca reservoir conditions. Therefore, this Pst steam pressure, Pa
study supports both Edmunds (1999a, b) argument and Butlers Sio initial oil saturation in the reservoir, no unit
conductive heat-transfer assumption, which assumes that conduc- Sor residual oil saturation (ROS) in the reservoir, no unit
tion is the only source of heat transfer to the cold formation ahead Swc connate-water saturation in the reservoir, no unit