SPE 113351

Calculation of a Critical Steam Injection Rate for Thermally-Assisted Gas-

Oil Gravity Drainage
Abdul S. Al-Rabaani, Martin J. Blunt, and Ann H. Muggeridge, Imperial College London
Copyright 2008, Society of Petroleum Engineers

This paper was prepared for presentation at the 2008 SPE/DOE Improved Oil Recovery Symposium held in Tulsa, Oklahoma, U.S.A., 1923April2008.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents of the paper have not been
reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect any position of the Society of Petroleum Engineers, its
officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Society of Petroleum Engineers is prohibited. Permission to
reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.


Abstract
Heavy oil contained in naturally fractured reservoirs is becoming an important resource as conventional oil reserves are
depleted. However, maximizing recovery from such reservoirs is problematic due to the low flow rate of oil and the poor
understanding of recovery mechanisms. One EOR method that is of particular interest is Thermally Assisted Gas-Oil Gravity
Drainage (TA-GOGD). In this process, steam is injected into the reservoir. This heats the rock matrix blocks through the
higher permeability fracture network and improves oil recovery principally by reducing the oil viscosity and thus increasing
the rate of gravity drainage through the matrix.
This paper shows that there is a critical steam injection rate for TA-GOGD in fractured reservoirs. If the injection rate is
below this critical rate, the time to heat the matrix will increase and oil recovery increases with the square of steam velocity
in the fracture. If the steam injection rate is greater than the critical rate then there is no significant increase in oil recovery
with rate.
A simple formula for calculating the critical steam rate in the fractures is derived analytically. We show that it correctly
predicts the critical injection rate for detailed simulations of steam injection into an element of a fractured reservoir
containing live oil. The predictions are shown to be correct for all fluid and rock properties investigated

Introduction
Gravity drainage is one of the most important mechanisms by which oil is recovered from fractured reservoirs (Saidi, 1998).
Gas is injected into the fracture system. The gas then enters the matrix once its pressure exceeds the entry pressure for the
matrix. As a result, oil drains out of the bottom of the matrix block (or stack of matrix blocks if there is capillary continuity
between the blocks) driven by the density contrast between the gas and the oil. The time taken for the oil to drain from the
matrix can be estimated from the expression:

t
GD
=

o
l
z
gK
v
(1)
where is the oil viscosity (N s m
-2
), k is the rock permeability (m
2
), is the density difference between the oil and
gas (kg m
-3
), g is the acceleration due to gravity (m s
-2
),

l
z
is the height of the matrix block (or stack of blocks) in m and
is the porosity. This equation shows that the time taken for oil to drain from a matrix block increases in proportion to the
oils viscosity, thus a very viscous, 200cP oil will take 200 times longer to drain from a given matrix block than a light 1cP
oil.
TA-GOGD (Thermally Assisted Gas-Oil Gravity Drainage) aims to improve the rate of oil recovery from fractured
reservoirs containing heavy (more viscous) oil (e.g Sahuquet and Ferrier, 1982; Dreher et al, 1986; van Wunnik and Wit,
1992; Shahin Jr et al., 2006;Bahonar et al, 2007; Das, 2007). Steam is injected into the reservoir and flows through the
fracture network heating the rock matrix and the oil contained within that matrix. The heated oil evolves gas and also
becomes less viscous. The presence of the gas in the matrix initiates gravity drainage of the oil and the reduction in oil
viscosity due to the heating means that the oil drains more quickly than would have been the case if only gas at reservoir
temperature had been injected. Experimental evidence (Sumnu et al, 1996) suggests that the steam does not enter the rock
matrix, partly because it is believed to condense in the fracture and partly because the rock in the matrix is typically oil wet.
Recovery may be further enhanced due to other effects resulting from the heating of the matrix including expansion of the
2 SPE 113351
rock, expansion of the oil and wettability alteration (Ayatollahi et al., 2005; Al-Hadhrami and Blunt, 2001; Rao, 1999;
Babadagli, 1996; Briggs et al, 1992; Briggs et al., 1988; Reis, 1990; Willman et al., 1961).
The effectiveness of TA-GOGD is thus principally a function of the rate at which the rock matrix is heated by the steam
and the rate at which oil drains from the matrix. These in turn will be determined by the thermal conductivity of the rock and
fluids, the change in oil viscosity with temperature, the permeability of the rock, the density difference between the oil and
the gas evolved from the oil during heating, the rock porosity and of course the size of the matrix block. However all these
data and thus the rate of heat transfer from injected steam into the matrix are uncertain (Penney et al, 2005).
Field-scale prediction of TA-GOGD performance, like any flow simulation in fractured reservoirs, is difficult because it
is impossible to model on the reservoir scale the exact fluid flow through the network of fractures and matrix blocks, let alone
the details of the heat flow. Instead the average behaviour of the system is modelled using a dual porosity or dual
permeability approach (Warren and Root, 1963, Gilman and Kazemi, 1983; Oballa et al., 1993; Winterfeld, 1996; Cicek,
2005) ). It is possible to model the details of matrix-fracture fluid and heat flow for small numbers of matrix blocks (Dreher
et al, 1986; Briggs et al, 1992; Sumnu et al, 1996) but even these simulations can be challenging due to the very small
fracture aperture (<1mm) compared to the typical size of a matrix block (>1m). Either prohibitively large numbers of grid
blocks are needed or there is a large contrast in grid block pore volumes which can lead to convergence problems.
A number of authors (Lesser et al, 1966; Satter, 1967; Satman, 1988; van Wunnik and Wit, 1992; Pooladi-Darvish et al,
1994) have used simple analytic solutions to investigate the details of matrix heating and subsequent oil recovery in order to
validate dual porosity and dual permeability simulations as well as to determine the parameters controlling the effectiveness
of TA-GOGD. Early workers (Lesser et al, 1966; Satter, 1967; Satman, 1988) simply looked at the rate of reservoir heating
by conduction and neglected the impact of heating on fluid flow, as their objective was simply to assess the feasibility of
using steam injection to heat the sub-surface. Van Wunnik and Wit (1992) and Pooladi-Darvish et al (1994) also investigated
the impact of heating on oil recovery although both made simplifying assumptions when modelling the transfer of heat from
fracture to matrix. Van Wunnik and Wit assumed that the time taken to heat each individual matrix block was small
compared with the gravity drainage time for the oil in order to more easily investigate field wide heat transport and oil
recovery. Pooladi-Darvish et al (1994) investigated the transient heating of a single matrix block and the subsequent oil
drainage from that block to determine whether the main mechanism for matrix heating was via conduction or convection.
Their analytical solution, based on a constant temperature in the fracture, showed that convective heat transfer was negligible
compared with conduction and that only the early stages of gravity drainage are affected by the heating thus justifying the
assumption of van Wunnik and Wit. In contrast the Satman (1988) and Lesser et al (1966) solutions for heating by
conduction showed that the temperature distribution in fractures changes with time and location.
In this paper we investigate the rate of heating within a matrix block and its impact on gravity drainage using a
combination of analytic solutions for heat conduction into a matrix and detailed numerical simulation of heat transport and
gravity drainage. We show that, for the fluid system considered, the time taken to heat the matrix block may be similar or
greater than the time taken for the heated oil to drain from that matrix block. In these circumstances oil recovery is controlled
by the rate at which the matrix block is heated. We use one of the analytic solutions to show that there is a critical steam
injection rate. If the injection rate is below this critical rate, the time to heat the matrix will increase and oil recovery
increases with the square of steam velocity in the fracture. If the steam injection rate is greater than the critical rate then there
is no significant increase in oil recovery with rate. A simple formula for calculating the critical steam velocity in the
fractures is derived and we show that it correctly predicts the critical injection rate for detailed simulations of steam injection
into an element of a fractured reservoir containing live oil.

Example Reservoir Properties
The rock and fluid data used in the analytical and the simulation models are summarized in table 1. The temperature
dependence of viscosity (shown in Figure 1) and most of the other rock parameters were taken from the paper published by
Penney et al (2005). The oil viscosity at the reservoir temperature of 50
o
C is 240 cP, as the temperature is increased the oil
viscosity reduces to ~ 2 cP at 250
o
C. This relationship between viscosity and temperature can be described by equation
= 44795T
3.101
(2)
where is the viscosity (N.s/m
2
) and T is temperature (
o
C).

Table 1a: Matrix and fracture properties used in the analytic and simulation studies

Matrix Fracture
Porosity (fraction) 0.32 0.999
Permeability (mD) 10 10,000
Initial water saturation 0.1 0
Initial oil saturation 0.9 1





SPE 113351 3
Table 1b: Fluid and rock properties used in both analytic and simulation studies
Oil density (kg m
-3
) 965
Gas density (kg m
-3
) 35
Volumetric capacity of water (J m
-3
C
-1
) 410
6

Volumetric heat capacity of steam (J m
-3
C
-1
) 55.710
3

Steam latent heat of condensation (J kg
-3
) 1.7810
6

Steam density (kg m10
6
) 16
Thermal conductivity of rock (J m
-1
s
-1
C
-1
) 1.74
Volumetric heat capacity of rock (J m
-3
C
-1
) 2.3510
6

Rock compressibility (kPa
-1
) 4.310
-7


Table 1c: Reservoir properties
Initial reservoir pressure (kPa) 3200
Initial reservoir temperature (

C) 50
Steam quality (%) 100



Figure 1: The reduction of oil viscosity with temperature relation (after Penney et al, 2005)

Analytical Studies
Figure 2 shows a conceptual model of a fractured reservoir undergoing steam injection. Steam enters the reservoir through
the injection well and flows through the fracture system to the production well, losing heat to the cooler matrix blocks as it
flows. As the steam cools it loses
s
C J m
-3
K
-1
to the matrix where C
s
is the volumetric heat capacity of steam until it
condenses. At this point it loses
s
L J m
-3
, where L is the latent heat of condensation of the steam and
s
is the steam
density, and becomes hot water flowing through the fracture. The hot water continues to flow and loses
w
C J m
-3
K
-1
where
C
w
is the volumetric heat capacity of water. If we neglect heat losses to the overburden then the heat lost from the steam is
transferred to the matrix and the fluids within it. Ignoring the gas evolved from the oil during heating the heat gained by the
matrix is given by [ ] { } [ ] ( )
o
C
wc
S
w
C
wc
S
m
C + + 1 1 J m
-3
K
-1
, where is the porosity, S
wc
is the connate water
saturation in the matrix, C
m
and C
o
are the volumetric heat capacities of the rock and oil respectively. We have assumed that
the gas evolved from the oil can be neglected due to the low gas volumetric heat capacity.

4 SPE 113351
Matrix
block
Fracture
No flow
boundaries (by
symmetry)
Steam

Figure 2: Schematic of matrix fracture system with steam flowing in fractures. The solid line shows the line of symmetry in the
fracture and the dashed line shows a line of symmetry in the matrix. Provided steam only flows from left to right there is no heat or
fluid flow across these lines.

In practice the contribution of the steam cooling to the matrix heating is small compared with the contribution from the
condensation of the steam and the cooling of the condensed water (see table 1b showing the values for steam and water
properties). Comparing the contributions from steam condensation and the cooling of resulting hot water it would appear that
the major contribution to the matrix heating comes from the steam condensation as this term is an order of magnitude greater
than that due to the cooling of the water. In this case we could model the heating of the matrix block as though it were
immersed in a bath of hot water at the condensation temperature of the steam i.e. constant temperature boundary conditions.
However the volume of the steam in the fracture is typically so much less than the volume of the matrix block that we
cannot neglect heating from the condensed water in the fracture. If the matrix is a cube of side l and the fracture aperture is
h, then the ratio of the fracture volume to the matrix volume is given by:
3
10 3
3
3
2
2
6
~

=
l
l
l
h
h
m
V
f
V
(3)
for typical matrix blocks, m 1 l and a typical fracture aperture, h<1mm. In contrast the volume of steam need to heat
the matrix block by 1 K is given by
08 . 0 ~
L
s
m
C
m
V
s
V

= where V
m
is the volume of the matrix block (using the rock and steam properties given in table 1).
The above calculation shows that many fracture volumes of steam need to be injected to heat the matrix block and the oil
contained within it, thus we need to consider the heat transport within the fracture system as well as the matrix system.
For simplicity we shall analyze the heat transfer from steam in the fracture to oil in the matrix for a small symmetry
element of the schematic matrix-fracture system, shown in figure 3. The results will be compared with the matrix heating
time obtained by assuming that the matrix block is immersed in steam at a constant temperature, as would be the case if
heating were mainly due to condensation for the steam. For the symmetry element of the matrix-fracture system shown in
figure 3 the conservation equation for heat flow is
0
2
2
2
=

t
m
T
z
m
T
h
z
z
m
T
m
t
f
T
w
C
x
f
T
w
UC

(4)
Where U is the steam velocity in the fracture, T
f
is the temperature of the steam in the fracture, T
m
is the temperature of
the matrix,
m
is the thermal conductivity of the matrix (in practice an effective value for matrix rock and the fluids contained
within it) and is the thermal diffusivity of the matrix, given by
SPE 113351 5
[ ] { } [ ]
o
C
wc
S
w
C
wc
S
m
C
m
+ +
=
1 1

(5)
In most cases we can approximate this as
m
C
m

This assumes that
1. Heat transfer within the matrix is dominated by conduction (as found by Pooladi-Darvish et al, 1994) in the
direction perpendicular to the fracture. Conduction in the matrix is negligible parallel to the fracture.
2. The steam velocity U

is constant in time.
3. The fracture aperture is sufficiently small that temperature changes across the fracture aperture can be neglected
4. All thermal properties are independent of temperature, and in particular the steam does not undergo any phase
changes as it cools.
5. Changes in steam volume with temperature are negligible.
6. There is no conduction of heat within the steam in the direction of flow.
7. The matrix has zero conductivity in the direction parallel to the flow in the fracture.
The assumption that steam does not undergo any phase change so we can neglect the heat from condensation considerably
simplifies the analysis but may affect the validity of the resulting solution. This will be discussed in more detail in later
sections of this paper.
If we impose the following boundary and initial conditions
( )

= =
= > =
= =
t
h
z x
m
T t x
f
T
t
h
z x
mi
T
m
T
t , x
fi
T
f
T
,
2
, ,
0 ,
2
,
0 at
(6)
then the solution for temperature changes in the fracture and the matrix can be found on page 396 of Carslaw and Jaeger
(1959).
( ) { }
( ) { }

+
=

=
2
1
2
2
1
2
U
x
t U
zU kx
erfc
mi
T
fi
T
mi
T
m
T
Dm
T
U
x
t U
kx
erfc
mi
T
fi
T
mi
T
f
T
Df
T

(7)
where
w
hC
m
k
2
= . This solution has been presented previously for radial flow around a well by Satman (1988) and in the
context of geothermal engineering by Bodvarsson and Tsang (1982). It should also be possible to use this equation to
describe mass transfer between matrix and fracture during miscible gas injection.

6 SPE 113351

No flow (ie heat and advection) boundary in fracture
No flow (ie heat and advection) boundary in matrix
Steam
h/2
l
x
/2
Matrix
(Temperature T
mi
at time t=0
Fracture
x
z

Figure 3: Element of matrix-fracture system used for investigating matrix heating by steam injection

Alternatively if the steam temperature in the fracture does not change with time or location (as assumed by Pooladi-
Darvish et al (1994)) then the time taken for the matrix block (dimensions
x
l by
y
l by
z
l ) to heat up to mean temperature
T is given by (derived from p184 Carslaw and Jaeger, see Appendix A)

+ +

=
mi
T
fi
T
T
fi
T
z y x
t
512
6
ln
2
1
2
1
2
1
2
1

l l l
(8)
We can now substitute equation (2) that relates oil viscosity to temperature into equations (7) and (8) to obtain
expressions for time taken to heat the oil to a given viscosity and compare these with the time taken for oil of that viscosity to
drain from the matrix block (Equation (1)). The results are shown in Figure 4 for a 1m matrix block (Figure 4a) and a 10m
matrix block (Figure 4b).
SPE 113351 7
10
100
1,000
10,000
100,000
1,000,000
0.000 0.010 0.020 0.030 0.040 0.050
Viscosity (N.s/m
2
)
T
i
m
e

(
h
r
),Constant fracture temperature
)Equation (8
Gravity Drainage
Heating due to steam flowing in
))fracture (Equation (7

Figure 4a: Comparison of the time for oil of different viscosities to drain from a 1m grid block as described by Equation (1) with the
time predicted for heating the same matrix block using Equation (7) (steam flowing in fracture) and Equation (8) which assumes a
constant temperature in the fracture. Oil drains very slowly from the matrix block unless its viscosity is less than 5cp (0.005 N s/m
2
).
The time taken for the matrix to heat is negligible in comparison if equation (8) is used although the time taken for heating is greater
if equation (7) is used when oil viscosities are lower than 0.005 Pa s

10
100
1,000
10,000
100,000
1,000,000
0.000 0.010 0.020 0.030 0.040 0.050
Viscosity (N.s/m
2
)
T
i
m
e

(
h
r
),Constant fracture temperature
)Equation (8
Gravity Drainage
Heating due to steam flowing in
))fracture (Equation (7

Figure 4b: As Figure 4a but for a 10 m matrix block. Heating time as calculated by equation (7) becomes more significant compared
with gravity drainage time for higher oil viscosities than for the 1m matrix block case.

The first observation from figure 4 is that oil recovery by gravity drainage is very slow until the oil viscosity is reduced to
below 0.005 Pa s. For higher viscosities the time for gravity drainage is so slow that it is unlikely that any oil would be
recovered over the typical lifetime of a field (~200,000 hours), particularly for larger matrix blocks. Comparison of the two
8 SPE 113351
expressions for heating the matrix block show that if the steam temperature in the fracture is constant then, as previously
calculated by Poovadi-Darvish et al (1994), the time taken to heat the matrix is generally negligible compared with the time
taken for the oil to drain out of the matrix. In contrast equation (7), which allows the steam temperature to change with time,
suggests that the time taken to heat the matrix block may be equal to or greater than the time taken for oil to drain out of the
matrix for oil viscosities less than 0.005 Pa s. This is the viscosity below which oil recovery is likely to occur at economic
rates. As might be expected equation (7) tends to equation (8) when the matrix is only heated by a small amount.
These analytic investigations provide conflicting results. Equation (7), with the more realistic boundary conditions in the
fracture suggests that oil rate during TA-GOGD may depend directly on the rate of heating of the matrix. However this
equation neglects the contribution of latent heat of condensation to the matrix heating. Equation (8) suggests that the time
taken to heat the matrix block is negligible compared with the time taken for oil to drain from the matrix block for all oil
viscosities. However the boundary conditions of constant matrix temperature for this expression do not seem plausible given
the small pore volume of a fracture compared with the matrix. We shall evaluate which of these solutions is more applicable
to TA-GOGD by performing detailed simulations of this process in a small number of matrix blocks surrounded by fractures.

Simulation study
The process of matrix heating was first investigated for a single matrix block using the grid illustrated in Figure 5. There is a
single horizontal fracture above a matrix block into which steam is injected in the top left hand corner. A production well is
completed in the top right hard corner. A second production well was completed in a fracture underneath the matrix block.
The model was intended to represent the system shown in figure 3 and was designed to determine whether the simulated heat
transfer from the steam into the fracture matched that predicted by Equation (7) as well as to evaluate whether oil rate was
indeed directly related to heating time as suggest by figure 4.
The grid used in this study was formed of 5112 cells in the X, Y and Z directions respectively. The grid blocks
representing the matrix were 10 m long in all directions. The fracture aperture was 0.01m. Other rock and fluid properties
were taken from Table 1. Steam was injected into the top fracture with different steam injection rates and a temperature of
250
o
C.

Fracture
Fracture
Matrix

Figure 5: Grid geometry used to investigate the heating of a single matrix block by steam injection in the top fracture. This
geometry is designed to replicate the system described by equation (7) and shown in Figure 3.

Figure 6 compares the time taken to heat the point x=50 m, y=10 m and z= 100 m to a temperature of T
Dm
=0.75 as a
function of steam velocity predicted by the simulation and Equation (7). It can be seen that heating is an order of magnitude
quicker in the simulation than predicted by the analytic solution. This is due to the additional heating in the simulation from
steam condensation in the fracture that is not included in the analytic solution. However the shape of the curve is the same
SPE 113351 9
for the simulation and the analytic solution showing that matrix block heating is a function of steam velocity in the fracture.
Thus equation (8) which assumes a constant steam temperature in the fracture is not valid.
Figure 7 shows the time taken to recover 50% of the oil in the matrix as a function of steam velocity in the fracture. The
shape of this curve is remarkably similar to that shown in Figure 6. For steam velocities greater than 0.001 m/s the time
taken to recover 50% OIP is constant, but for lower steam velocities the time taken to recover the oil increases. Thus oil
recovery appears to be directly related to the time taken to heat the matrix which in turn is a function of the steam velocity in
the fracture.
1.00E+00
1.00E+01
1.00E+02
1.00E+03
1.00E+04
1.00E+05
1.00E+06
1.00E+07
1.00E+08
0.000001 0.00001 0.0001 0.001 0.01 0.1 1 10 100
Steam velocity (m/s)
T
i
m
e

t
o

h
e
a
t

0
.
7
5
T
D
(
d
a
y
s
)
Simulation Results Equation 7

Figure 6: The time to heat the matrix block to 75% the difference between the initial temperature of the injected steam and the initial
matrix block temperature as a function of steam velocity in the fracture. The simulation predicts that heating is an order of
magnitude faster than the analytic solution as the analytic solution omits heating due to steam condensation. However both curves
show the same behaviour in that heating rate decreases with increasing steam velocity up to a steam velocity of 10
-5
m/s after which
the heating rate is constant.
1
10
100
1000
10000
100000
1000000
0.00001 0.0001 0.001 0.01 0.1 1 10 100
Steam Velocity (m/s)
T
i
m
e

t
o

r
e
c
o
v
e
r

5
0
%

o
f

O
I
P

Figure 7: The time to recover 50% of the oil in place as a function of steam velocities

The simulation investigation was extended to the multi-block case shown in Figure 8. This multi-block model is intended
to represent a section of a fractured reservoir midway between injection and production wells. As in the single block model
steam was injected in the top left corner of the top fracture. Two production wells were used, one in the top right and one in
the bottom left. The top production well was used to ensure a uniform flow of steam along the top of the model, replicating
the steam zone seen in field scale simulations. The bottom production well was intended to recover oil draining from the
10 SPE 113351
matrix blocks. This simulation used the same data as in the single block investigation. A 3D grid formed of 62126 cells in
the X, Y and Z direction respectively was used with matrix size dimensions of 124850 m. The model was formed of 4
matrix blocks of dimensions 124850 m separated by vertical fractures. The fracture aperture was 0.01m.

Fracture
Matrix Matrix Matrix Matrix
Steam injection
Oil production
Steam production

Figure 8: 2D schematic diagram of the matrix fracture geometry and well locations used in the multi-block simulation of TA-GOGD.
There was also a vertical fracture along front and back faces of the model. The fracture aperture was0.01m in this simulation and
each matrix block was 12m long, 12m wide and 48m deep.

The time taken to heat all the matrix blocks to an average 0.75T
D
as a function of steam velocity is shown in Figure 9
whilst Figure 10 shows the time taken to recover 50% of the oil in place, again as a function of steam velocity. Crucially the
shape of the curves in both Figure 9 and Figure 10 is very similar to that seen in figure 6 with the time taken to heat the
matrix block and recover the oil decreasing as steam velocity increases up to a critical steam velocity. For steam velocities
greater than this velocity the time taken to recover 50% of the oil does not decrease.
1.00E+00
1.00E+01
1.00E+02
1.00E+03
1.00E+04
1.00E+05
1.00E+06
1.00E+07
0.000001 0.00001 0.0001 0.001 0.01 0.1 1 10
Steam velocity (m/s)
T
i
m
e

t
o

h
e
a
t

0
.
7
5
T
D
(
d
a
y
s
)
Simulation Results Equation 7

Figure 9: The time to heat the multiple matrix block model to 75% the difference between the initial temperature of the injected
steam and the initial matrix block temperature as a function of steam velocity in the fracture. The simulation predicts that heating is
an order of magnitude faster than the analytic solution as the analytic solution omits heating due to steam condensation. However
both curves show the same behaviour in that heating rate decreases with increasing steam velocity up to a steam velocity of 10
-4

m/s after which the heating rate is constant.

SPE 113351 11

1
10
100
1000
10000
100000
0.00001 0.0001 0.001 0.01 0.1 1 10
Steam Velocity (m/s)
T
i
m
e

t
o

r
e
c
o
v
e
r

5
0
%

o
f

O
I
P

(
d
a
y
s
)

Figure 10: The time to recover 50% of oil in place as a function of steam velocities

These detailed simulations show that oil recovery from TA-GOGD appears to depend directly in the time taken to heat the
matrix blocks and that this in turn is a function of the steam velocity in the fracture. Equation (7) predicts the shape of the
time to heat the matrix block as a function of steam velocity in the fracture although it over predicts the exact time by at least
an order of magnitude as it neglects heating from steam condensation. In particular it appears to correctly predict the critical
velocity for the steam in the fracture, above which no further decrease in heating time and time to recover the oil is seen.

Calculation of critical rate
We now examine equation (7) more closely to derive an expression for the critical steam velocity in the fractures. For small
x, ( ) x x erfc 1 . Therefore for late times when the matrix temperature approaches the fracture temperature i.e. 1
Dm
T ,
equation (7) can be written as
( )
( )
2
2
1 4
2
U
Dm
T
Uz kx
U
x
t

+
+ = (9)
Following the derivation given in Appendix B we find a critical steam velocity given by
w
C
z
h
m x
cr
U
l
l 2
(10)
Thus the optimum steam velocity to maximize oil recovery rate is dependent only on the matrix block size, the matrix
thermal conductivity and the water volumetric heat capacity.

Discussion
The detailed simulation results presented here indicate that gravity drainage occurs at roughly the same rate as the matrix is
heated and as a result the rate of oil recovery increases with steam flow rate up to a critical value. Beyond this flow rate the
rate of recovery remains constant and the time to heat the matrix can be estimated by equation (8), which assumes constant
temperature in the fracture. Close examination of the results of previous studies also shows that oil rate increases with steam
velocity (Dreher et al , 1986; Cicek et al, 2005; Penney et al, 2005) and thus is dependent on the rate of heating of the matrix.
Moreover the results of Penney et al (2005) indicate that there is an optimum steam injection rate. They noted that recovery
in their dual permeability simulations increased with steam flow rate up to 18,000 tonnes/day but increased little for higher
steam flow rates. Thus the results of van Wunnik and Wit (1992), evaluating reservoir performance on the basis that matrix
heating is instantaneous compared with oil drainage rate should be treated with caution. Similarly the analytic solution of
Pooladi-Darvish et al (1994), which is based on an assumption of constant temperature in the fracture, may not be reliable in
all cases.
Equation (10) provides a means of estimating the optimum steam injection rate to maximize oil recovery rate, however
this injection rate may be difficult to achieve operationally as it assumes that steam is recycled through the reservoir. It also
neglects heat losses to the formation so in practice a higher steam injection rate may be required if maximum oil recovery rate
12 SPE 113351
is desired. Lower steam injection rates will result in fewer pore volumes of steam being injected for the recovery of the same
pore volumes of oil recovered albeit over a longer period of time.

Conclusions
This paper has investigated the impact of the rate of matrix block heating on oil recovery rate by gravity drainage during
Thermally Assisted Gas Oil Gravity Drainage (TA-GOGD). Both analytic solutions and detailed numerical simulation have
shown that, for the rock and fluid properties considered here, oil recovery rate is directly dependent on the rate of heating of
the matrix block. In particular oil recovery rate increases with increasing steam injection rate up to a critical steam injection
rate. For higher steam rates oil recovery rate remains constant. An analytic expression has been derived for estimating this
optimum or critical steam injection rate that depends only on the matrix block size, water volumetric heat capacity and matrix
block thermal conductivity.

Acknowledgment
The authors would like to thank the Computer Modelling Group (CMG) for providing the simulator STARS and for their
continuous technical support. The authors would like also to thank Petroleum Development Oman (PDO) for funding
Abduls PhD.

Nomenclature
T= Temperature
= Thermal diffusivity
K=permeability
=Average density
c= Specific heat
h= Fracture aperture

l=Matrix block length
t=Time
U= Velocity
= Porosity
=Viscosity
L= Latent heat of condensation
C= Volumetric heat capacity
S= Saturation
= Thermal conductivity
V= Volume
Subscripts
m= matrix
s= steam
r= reservoir
i= initial
cr= critical
f= fracture
o= oil
w=water
c=connate
v=vertical
D=Dimensionless
Unit conversion
1mD=1x10
-15
m
2

1cP=0.001 N.s/m
2

References
Al-Hadhrami, H. S., and Blunt, M. J. "Thermally Induced Wettability Alteration to Improve Oil Recovery in Fractured Reservoirs", SPE
Reservoir Evaluation & Engineering, 179-186, June, 2001
Ayatollahi, S. , Boukadi, F. , Wadhahi, M. , Maamari, R. , and Bemani, A. , "Investigation of Thermal Gas Oil Gravity Drainage (GOGD)
in Fractured Carbonate Reservoirs ", paper SPE 93585, Proceeding of the SPE Middle East Oil and Gas Show and Conference,
Kingdom of Bahrain, Mar 12 - 15, 2005
Babadagli, T. "Heavy-Oil Recovery From Matrix During Thermal Applications in Naturally Fractured Reservoirs", In Situ, v. 20, 3, p. 221,
1996
Bahoner, M., Ataei, A., Masoudi, R. and Mousavi Mirkalaei, S.M., "Evaluation of Steam Injection in a Fractured Heavy-Oil Carbonate
Reservoir in Iran", SPE 105299, proceeding of the 15
th
SPE Middle East Oil and Gas Show, Bahrain, 11-14 March 2007.
Bodvarsson, G.S. and Tsang, C.F., "Injection and Thermal Breakthrough in Fractured Geothermal Reservoirs", J. Geophys. Res. v. 87(82),
1031-48, 1982
SPE 113351 13
Briggs, P.J., Baron, P.R., Fulleylove, R.J., and Wright, M.S. "Development of Heavy-Oil Reservoirs", Journal of Petroleum Technology,
206-214, February, 1988
Briggs, P.J., Beck, D.L., Black, C.J.J. and Bissell, R., "Heavy Oil from Fractured carbonate Reservoirs", SPE Res. Eng. pp173-179, May
1992.
Carslaw, H. S., and Jaeger, J. C., "Conduction of Heat in Solids", Oxford University Press. Second Edition 1959.
Cicek, O., "Modeling and Designing Improved Oil Recovery by Steam Injection in Naturally Fractured Reservoirs", SPE 93574,
proceeding of the SPE Western Regional Meeting held in Irvine, CA, USA, 30 March - 1 April 2005.
Das, S., "Application of Thermal Recovery Processes in Heavy Oil Carbonate Reservoirs", SPE 105392, proceeding of the 15
th
SPE Middle
East Oil and Gas Show, Bahrain, 11-14 March 2007
Dreher, K.D., Kenyon, D.E. and Iwere, F.O.,, "Heat Flow During Steam Injection into a Fractured Carbonate Reservoir", paper SPE14902,
proceeding of the SPE/DOE Fifth Symposium on Enhanced Oil Recovery, Tulsa, OK, April 20-23, 1986.
Firoozabadi, A., and Markeset, T., "Laboratory Study of Reinfiltration for Gas-Liquid System in Fractured Porous Media", paper SPE
26129, 1992
Gilman, James R., and Kazemi, Hossein. "Improvements in Simulation of Naturally Fractured Reservoirs ", SPE 10511, SPE Journal, 695-
707, August, 1983
Lesser, H.A., Bruce, G.H. and Stone, H.L., "Conduction Heating of Formations With Limited Permeability by Condensing Gases", SPEJ,
December 1966.
Oballa, V., Coombe, D.A. , and Buchanan, W, L. "Factors Affecting the Thermal Response of Naturally Fractured Reservoirs", JCPT, 32-
42, October, 1993
Penney, R., Moosa, R., Shahin, G., Hadhrami, F., Kok, A., Engen, G. and van Ravesteijn, O., "Steam Injection in Fractured Carbonate
Reservoirs: Starting a New trend in EOR", IPTC 10727, proceeding of the International Technology Conference held in Doha, Qatar,
21-23 November 2005.
Pooladi-Darvish, M., Tortike, W.S. and Farouq-Ali, S.M., "Steam Heating of Fractured Formations Containing Heavy Oil: Basic Premises
and a Single-Bloock Analytical Model", SPE28642, proceeding of the SPE 69
th
Annual Technical Conference and Exhibition held in
New orleans, LA, USA, 25-28 September 1994.
Pratap, Mahendra, Kleppe, Jon, and Uleberg, Knut, "Vertical Capillary Continuity Between the Matrix Blocks in a Fractured Reservoir
Dramatically Improves the Oil Recovery by Water Displacement", paper SPE 37725, proceeding of the Middle East Oil Show and
Conference, Bahrain, 15-18 March, 1997
Rao, D.N., "Wettability Effects in Thermal Recovery Operations", SPE Reservoir Eval. & Eng. v 2(5), October 1999
Reis, J.C., "Oil Recovery Mechanisms in Fractured Reservoirs During Steam Injection ", paper SPE 20204, proceeding of the SPE/DOE
Enhanced Oil Recovery Symposium, Tulsa, Oklahoma, 22-25 April, 1990
Sahuquet, B.C. and Ferrier, J.J., "Steam Drive Pilot in a Fractured Carbonate Reservoir: Lacq Superieur Field", JPT, 873-880, April 1982
Saidi, A. M, "Reservoir Engineering of Fractured Reservoirs: Fundamental and Practical Aspects": Paris, Total Edition Press. 1988
Satman, A., "Solutions of Heat- and Fluid-Flow in Naturally Fractured Reservoirs: Part 1 - Heat Flow Problems", SPE Production
Engineering, November 1988.
Satter, A., "A Prediction Method for Conduction Heating of Reservoirs by Steam Injection", SPE 1950 proceeding of the SPE 42
nd
Annual
Fall Meeting held in Houston, TX, Octrober 1-4 1967.
Shahin Jr, G.T. , Moosa, R., Kharus, B., and Chilek, G., "The Physics of Steam Injection in Fractured Carbonate Reservoirs: Engineering
Development Options That Minimize Risk", paper SPE 102186, proceeding of the SPE Annual Technical Conference and Exhibition,
San Antonio, Texas, USA, 24-27 September, 2006
Sumnu, M.D., Brigham, W.E., Aziz, K. and Castanier, L.M., "An Experimental and Numerical Study on Steam Injection in Fractured
Systems", SPE 35459, proceeding of the Tenth SPE Symposium on Improved Oil recovery held in Tulsa, OK, 21-24 April 1996
van Wunnik, John N.M., and Wit, Krijn. "Improvement of Gravity Drainage by Steam Injection Into a Fractured Reservoir: An Analytical
Evaluation ", SPE 20251, SPE Reservoir Engineering, v. 7(1), p. 59-66, 1992
Warren, J.E. and Root, P.J. "The Behaviour of Naturally Fractured Reservoirs", SPEJ, v. 3, 3, p. 245-255, September, 1963
Winterfeld, P.H., "Development of Dual-Permeability Thermal Simulator and Validation for Yates Field", SPE 35420, proceeding of the
SPE/DOE Tenth Symposium on Improved Oil Recovery held in Tulsa, OK, 21-24 April 1996.
Willman, B.T. , Valleroy, V.V. , Runberg, G.W. , Cornelius, A.J. , and Powers, L.W. . "Laboratory Studies of Oil Recovery by Steam
Injection ", SPE1537, Journal of Petroleum Technology, v. 13, Number 7, p. 681-690, July, 1961


Appendix A
Heating up simple block matrix derivation
In this appendix, a full derivation for the analytical solution for heating up a simple matrix block is given.
( )
( )( )( )
( ) ( ) ( )
r
T
s
T
r
T T
Dm
T
n m l
t
e
c
z n
b
y m
a
x l
l m n
n m l
n m l
Dm
T

=
+ + +
+ +

=
where
, ,
2
1 2
2
1 2
cos
2
1 2
cos
0 0 0
1 2 1 2 1 2
1
3
64
1

(A1)
As we are interesting for long timescales when the higher order terms are negligible. Therefore, l=m=n=0. Thus we will
have:
14 SPE 113351
t
lmn
e
c
z
b
y
a
x
Dm
T

=
2
cos
2
cos
2
cos
3
64
1 (A2)
By integrating the equation for

dxdydz
a
a
b
b
c
c
Dm
T
z y x
Dm
T

=
l l l
1
(A3)
We get the mean normalised temperature in the block
t
lmn
e
c b a
z y x
Dm
T

4 4 4
3
64 1
1
__
l l l
(A4)

t
lmn
e
abc
z y x
Dm
T

3
64
3
64 1
1
__
l l l
(A5)
We have 2a=
x
l , 2b=
y
l and 2c=
z
l


Then,
t
lmn
e T
Dm

= .
6
512
1 (A6)

And we have

+ + =
2
1
2
1
2
1
4
2
, ,
c b a
n m l

+ + =
2
4
2
4
2
4
4
2
, ,
z y x
n m l
l l l

(A7)
To get the time that will be needed to get to the average temperature we take the log. of both sides
( )
Dm
T
t
e =

1
512
6

(A8)
( )

=
512
6
1
ln
1

Dm
T
t (A9)
SPE 113351 15
( )
( )
( )

+ +

+ +

=
m
T
s
T
T
s
T
z y x
t
Dm
T
z y x
t
512
6
ln
2
1
2
1
2
1
2
1
512
6
1
ln
2
4
2
4
2
4
2
4

l l l
l l l
(A10)
Appendix B: Derivation for the critical velocity

At late times when 1

=
mi
T
fi
T
mi
T
m
T
Dm
T then
( ) { }
2
1
2
1
U
x
t U
zU kx
Dm
T

(B1)
Where

=
m
C
m

and

=
w
hC
m
k
2

Rearranging this we obtain
( )
( )
2
2
1 4
2
U
Dm
T
Uz kx
U
x
t

+
+ = (B2)
Taking the natural logarithm of both sides we get
( ) ( )
( )
( ) ( ) )
2
2
1 4 ln(
2
) (
2
1 4 ln ln
2
2
1 4
2
2
1 4
ln ln
U
Dm
T Uz kx U
Dm
T
U
Dm
T
Uz kx U
Dm
T
t

+ + =

+ +
=

(B3)
When U is small then
( ) ( )
( ) ( ) ( ) U
Dm
T kx t
U
Dm
T kx t
ln 2
2
1 4 ln ln 2 ln
2
2
1 4 ln ln 2 ln

(B4)
So the logarithm of heating time depends linearly on the log of U and the gradient of this straight line is negative
If U is large then
16 SPE 113351
( ) ( )
( )
( )
( )
large is when
2
1 4
2
ln ln
2
1 4
2
2
1 4
ln
2
2
1 4 ln
2
2
1 4 ln ln
U
Dm
T
z
t
U
Dm
T
Uz
Dm
T
U
Dm
T Uz
Dm
T U t

+
=
+

(B5)

We can use this expression to find the critical velocity by substituting for
( )

2
1 4
2
ln ln
Dm
T
z
t (B6)

In the expression derived for small U so
( )
( ) ( )

2
2
1 4 ln ln 2
2
1 4
2
ln U
Dm
T kx
Dm
T
z

(B7)
Giving,
z
kx
cr
U = (B8)
Substituting for

=
w
hC
m
k
2

we get
w
C
z
h
m x
cr
U
l
l 2
= (B9)