You are on page 1of 14

SPE

Society of Petroleum Engineers

SPE 18752

Reservoir Simulation of Cyclic Steam Stimulation in the


Cold Lake Oil Sands
by C. I. Beattie,* Esso Resources Canada Ltd.; T.C. Boberg,* Exxon Production Research Co.;
and G.S. McNab, Esso Resources Canada Ltd.
* SPE Members

Copyright 1989, Society of Petroleum Engineers, Inc.

This paper was prepared for presentation at the SPE California Regional Meeting held in Bakersfield, California, April 5-7, 1989.

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,
as presented, have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material, as presented, does not necessarily reflect
any position of the Society of Petroleum Engineers, its officers, or members. Papers presented at SPE meetings are subject to publication review by Editorial Committees of the Society
of Petroleum Engineers. Permission to copy is restricted to an abstract of not more than 300 words. Illustrations may not be copied. The abstract should contain conspicuous acknowledgment
of where and by whom the paper is presented. Write Publications Manager, SPE, P.O. Box 833836, Richardson, TX 75083·3836. Telex, 730989 SPEDAL.

ABSTRACT This geomechanical representation allows


the simulator to match field observations
The Cold Lake reservoir is an that are otherwise difficult to reproduce,
unconsolidated sand containing extremely including injection pressures, flowback
viscous bitumen. Steam injectivity during times, and production pressures. Also,
cyclic steam stimulation can be achieved the model appropriately handles the
only by injecting at pressures high enough recompaction process which provides drive
to mechanically fail the formation. energy in the Cold Lake reservoir.
Simulation of the complex fracturing and
reservoir deformation behavior that The water-oil relative permeability
results is very challenging. In addition, hysteresis model is based upon laboratory
the reservoir exhibits water-oil relative measurements. Bounding imbibition and
permeability hysteresis, which must also drainage curves are input; gridblock
be properly modeled. This paper describes relative permeabilities, which depend upon
enhancements made to a thermal reservoir both saturation and saturation history,
simulator to incorporate these Cold Lake are determined such that calculated values
physics. always lie on or between the bounding
curves. The hysteresis model makes it
Rigorous geomechanical modeling is not possible to use laboratory-derived
economical, so an empirical approach has relative permeabilities when simulating
been developed that is consistent with the cyclic steam stimulation and still match
behavior of unconsolidated sands. field water-oil ratios.
Fracturing is modeled by allowing the
permeability in a plane of gridblocks to INTRODUCTION
increase rapidly when the pressure exceeds
a specified fracture pressure. Reservoir The Cold Lake oil sand deposit in northern
deformation in all blocks is modeled by Alberta contains over 40 billion cubic
first allowing dilation, during which meters (250 billion barrels) of
porosity increases when the pressure bitumen-in-place. The extremely high oil
exceeds a specified failure pressure. viscosity and low native water mobility in
Subsequent pressure decline causes the this reservoir result in negligible
reservoir to recompact, and porosity initial injectivity or productivity.
decreases. However, recompaction is not Steam injection at commercial rates
the reverse of dilation, and a fraction of requires injection pressures high enough
the total dilation is permanent. While to cause both localized fracturing and
all gridblocks have similar deformation widespread pore volume increases in the
properties, the history of each individual formation. The resulting complex
block plays a role in determining its geomechanical behavior determines the
exact behavior. initial steam flow paths. In addition,
recompaction of failed regions of the
reservoir during the production period
References & illustrations at end of paper provides significant drive energy during

69
2 RESERVOIR SIMULATION OF CYCLIC STEAM STIMULATION IN THE COLD LAKE OIL SANDS SPE 18752
cyclic ~team stimulation (CSS) formation. The phenomena of both dilation
operations. The reservoir also exhibits and the subsequent recompaction during
relative permeability hysteresis which production occur when wells are steam
affects production behavior. These stimulated at Cold Lake. Details of the
phenomena must be adequately represented modeling of these phenomena are discussed
in the simulator to properly model Cold in turn below.
Lake reservoir performance during css.
This paper describes the deformation, Dilation
fracturing, and relative permeability
hysteresis behavior in the Cold Lake Two observations lead to the conclusion
·clearwater reservoir, and the methods used that significant dilation of the reservoir
by Esso Resources Canada Limited (ERCL) to occurs during steam injection at Cold
include these phenomena in our simulators. Lake. The first is surface heave. Using
The physics and modeling aspects of each surface benchmark arrays at Cold Lake,
of these are described in turn, and ERCL has measured surface uplifts as great
methods of determining values for the as 4 5 em ( 17 in) during injection, far
additional input required by the simulator larger than can be attributed to thermal
are discussed. Examples illustrate how expansion or tensile fracturing of the
the model enhancements improve the formation. The second observation is that
simulator matches of field observations. steam injectivity at Cold Lake is greater
than might be expected based on the native
The simulator enhancements described here reservoir properties. A common problem
have been incorporated into both 2 when simulating steam injection into tar
commercially-available thermal simulator sands is that of calculating injectivity
and a new thermal version of Exxon as high as is observed in the field while
Production 3 Research Company's MARS using reasonable £racture lengths and rock
simulator. Although different
formulation, both simulators are fully
in compressibilities. The compress!£ility
Cold Lake sands is about 1.0 GPa Cax10
2s
implicit models, and give virtually psi- 1 ) at reservoir conditions. 7 ' In
identical results when used to simulate order to match observed injectivities,
the same problem. most published simulations of Cold Lake
CSS have used a rock compressibility one
or. t"9o 1 6'r.Pfrs of magni \'2de large;r t~an
RESERVOIR AND OPERATIONS DESCRIPTION th1s. ' ' Coats et al used th1s h1gh
compressibility method for a California
ERCL's Cold Lake operations target the heavy oil reservoir, and referred to it as
Clearwater formation, a near-shore deltaic the "spongy rock" approach. The only
sand of Cretaceous age at a depth of about published Cold Lake simulation that did
450 meters (1500 ft). The sands are not use an enhanced compressibility
thick, often in excess of 40 meters (130 approach required unreasonably long - 240
ft), with a high net-to-gross thickness meter (790 ft) - fr~<jtures to match the
ratio. Porosity ranges from 30 to 35 observed injectivity.
percent, with oil saturations that average
70% of pore volume. At the initial A major problem with the spongy rock
reservoir temperature of 13°C (55°F), the approach is that it results in an
oil viscosity is about 100 Pa-s ( 100, ooo ~njection pressure which steadily
cp). Oil viscosity decreases to 0.002 1ncreases with time, even in first cycle,
Pa-s (2 cp) at 250°C (480°F). whereas the field pressures increase
rapidly when injection is started, and
ERCL uses cyclic steam stimulation at Cold then level off for most of the cycle.
Lake. Steam is injected at 225 m3/day This is illustrated in . Figure 1, which
(1400 bbl/day) for 30 to 60 days, followed shows typical field and simulation results
by a production period of 120 to 400 days for first cycle injection. The spongy
depending on cycle number. Average rock simulation useg 1 a ro2f
prod~cing day oil rates decline from over com121essibility of 35 Gpa (2.4 x 10
20m /day 3 (125 bbl/day) in first cycle to psi ).
about 7 m /day (45 bbljday) in the eighth
cycle; water-oil ratios increase from 1-2 surface heave and the high observed
to over 4 during the same period. Total injectivity can be explained by two
daily production from almost 1800 pilot mechanisms that increase porosity. First,
a~d commercial wells averaged over 14,000 oil sands demonstrate non-linear
m /day (90, 000 bbljday) in 1988. More compressibility behavior. Figure 2 shows
detailed ~s~riptions of both pilot 4 and how the compressibility increases
commercial ' operations are available. dramar£cally as the effective stress nears
zero. In addition, shear failure can be
DEFORMATION MODELING induced in the formation at sufficiently
low effective stresses in the presence of
High-pressure injection into a tar sand anisotropic stresses. Shear failure
reservoir causes both widespread pore caused by increasing pore pressure results
volume increases (dilation) and fracturing in dilation of the pore system. Recent
(which is discussed later) in the

70
SPE 18752 C.I. BEATTIE, T.C. BOBERG & G.S. McNAB 3

publications describe thi 9}1ear dilation approach, we have observed porosity


mechanism in more detail. 4 ' increasing to a maximum of 110-120% of the
original value, which compares f<r~orably
As steam is injected into a relatively to a published prediction of 123%.
incompressible reservoir, pore pressure
increases and the effective stress A similar treatment of porosity increases
decreases. At pore pressures with pressure during injection was
corresponding to low effective stresses, proposed by Ito 16 as part of his "sand
the formation compressibility increases by deformation concept". However, Ito
about two orders of magnitude, and shear limited the maximum porosity to 104% of
failure may also occur. Both of these the original value, and did not assume the
phenomena result in dilation with rapid existence of a fracture in his simulation.
increases in porosity and permeability. In our approach, both fracturing and
This greatly increases the steam deformation are assumed to occur, as this
injectivity, and injection pressures best fits availab!~ field observations and
increase slowly thereafter. The dilation published theory. Ito also assumed that
is reflected by uplift of the ground dilation was reversible. As shown in the
surface above the injection location. next section, we have found that this is
not a good assumption.
We model this phenomenon by specifying a
dilation pressure (P ) below which
behavior is elastic ana a low original Recompaction
compressibility value is used. Above the
dilation pressure, a higher dilation Little has been published on the
compressibility is used enabling the recompaction behavior of an oil sand which
porosity to increase rapidly with has undergone dilation. However, the
pressure, as shown by the line labelled general behavior can be inferred. The
"dilation" in Figure 3. A maximum Cold Lake area was covered during the last
porosity is also specified; once this ice age by an ice sheet over 1800 meters
value is reached, further dilation is not (5900 ft) thick. The sands are therefore
permitted, and the compressibility reverts more compacted than might be expected for
to a low value. The improved match of an unconsolidated sand at a depth of 450
injection pressure which results from use meters ( 1500 ft) . Because of this
of this dilation model is shown by the depositional history, recompaction to the
line labelled "dilation" in Figure 1. original . porosity following dilation
cannot occur at Cold Lake. This is borne
The functional form used for all of the out by field observations. Surface heave
porosity-pressure relationships in our is cyclical, increasing during injection
deformation model is: and decreasing during production.
However, some of the surface uplift at
~=¢REF exp{ C(P-PREF)} .•..... (1) ERCL's operation is permanent. Hence,
while partial reversal of the dilation
where porosity .¢REF occurs at reference occurs, recompaction to the original
pressure PRE , ¢ TS a porosity at some porosity is not occurring. The oldest
pressure P, Knd C is compressibility. A area of the Leming pilot, for example, has
linearized approximation to Equation 1 is a permanent residual uplift of 15 em ( 6
often used in reservoir simulators, but in) which has remained unchanged for
for the large compressibilities typical of several years. While injection and
dilation, this linearization is not valid production have continued during that
and the exponential form must be used. time, injection has been at pressures
below the failure pressure.
Use of the dilation model requires
estimates of the pressure at which Field observations also suggest that
dilation begins (P ), the dilation recompaction does not begin immediately
compressibility, and tRe maximum allowable when the pressure begins to decline
porosity. The dilation pressure is close following dilation. When a Cold Lake css
to or slfQ~tS-Y lower than the fracture well is initially placed on production
pressure. ' The dilation following steaming, the bottomhole
c9mpr~ssibility can be estimated f~~m pressure is sufficiently high that the
F~gure 2 to be 100 - 1000 GPa . well flows without pumping. After
Experience shows that results are flowback ends at a bottomhole pressure of
relatively insensitive to variations about 4 MPa (580 psi), the pump is seated
within this range. However, an to continue production. In a first cycle
excessively high value can cause numerical well, flowback normally lasts for less
instability in the simulator. We than two weeks. During this time the
generally use a large value for the pressure declines from over 10 MPa ( 1450
maximum allowable porosity, and find that psi) to 4 MPa (580 psi). A thermal
computed porosities never reach this simulator cannot compute a pressure change
value. Other regions of the reservoir of this magnitude with a relatively small
become the predominant failure zone before fluid withdrawal unless an initial period
the most highly pressured gridblocks reach of elastic behavior with little
their maximum porosity. Using this recompaction to support the pressure is
71
4 RESERVOIR SIMULATION OF CYCLIC STEAM STIMULATION IN THE COLD LAKE OIL SANDS SPE 18752

assumed. recompaction compressibility in all


gridblocks.
Recompaction, therefore, has two phases:
an initial elast.ic period when no recovery The quadrilateral shape of the deformation
of dilation occurs and porosity is model illustrated in Figure 3 is
changing only due to the original consistent w~ experimental observations.
compressibility, followed by a Roscoe et al describe experimental and
recompaction period with an enhanced theoretical results on the behavior of
compressibility that allows recovery of soils. Figure 4 is schematically similar
some of the dilation which occurred during to the behavior their model would describe
injection. We model this behavior as for a constant total stress condition in
shown schematically in Figure 3. When the Cold Lake Clearwater Formation. The
pressure in a dilated gridblock begins to sands had a very high porosity when
decline, the recompaction model is used. initially deposited. With further
Note that the dilation model is not deposition and burial, the sands became
reversible. After the pressure begins to more compacted, and the porosity moved
decline but before it reaches the down the consolidation curve. Maximum
recompaction pressure (PR) , the slope of compaction was achieved during the last
the porosity-pressure function is ice age, as described earlier. After
determined by the low original melting of the ice sheet reduced the
compressibility as illustrated by the effective stress acting on the sands, the
upper line labeled "elastic" in Figure 3. current virgin reservoir conditions were
The simulator tracks the maximum pressure reached. The path that is then followed
and porosity reached by each gridblock during injection and subsequent production
during dilation, and these values are used is shown in the Figure. The similarity to
as the reference values in Equation 1 our model (Figure 3) is apparent.
during the initial elastic period. During Rigorous modeling of the behavior
this period, effective stress is described by Roscoe et al would require
increasing, and this continues until the solving for the three-dimensional in situ
stress is high enough (pressure low stresses as additional simulator
enough) to allow recompaction to begin. variables. However, the current level of
understanding of oil sand behavior does
Below P , the reservoir begins to not warrant this additional complexity.
recompac~ A single compressibility Our model therefore assumes that the total
cannot be used to describe recompaction stress is constant. This limits the
because this could cause the porosity to model's capability in cases where total
fall below the original reservoir porosity stress changes are important. However,
in gridblocks that had only slightly success in applying this model to
dilated. Our model assumes that blocks simulation of Cold Lake css suggests that
which undergo large dilations will the stress changes are usually of
recompact more than blocks that dilate secondary importance in controlling the
only slightly. This is modeled by magnitude of dilation and recompaction.
defining the "residual dilation fraction"
(F ) to be the fraction of the total At present, the pressure at which
di~ation in a gridblock which is permanent recompaction begins (PR) and the residual
and unrecoverable, as shown in Figure 3. dilation fraction ( ~R! must be obtained
When pressure declines to PR, the minimum from history matching, as no published
allowable porosity for each block is estimates of their values exist.
calculated from its historical maximum Fortunately, each parameter has a unique
dilation porosity and F . The minimum effect on the simulated results, and it is
porosity applies at a plessure value of therefore relatively straightforward
zero. With the coordinates of both ends determine their values.
of the recompaction function specified,
Equation 1 can be used to calculate the It has been estimated using this model
recompaction compressibility for each that recompaction provides about 60% of
block. The method thus allows more the drive fnergy in early cycles of Cold
recompaction in blocks that dilated more Lake css. The value used for the
and guarantees that the porosity will recompaction pressure PR determines the
never fall below the original value. pressure range over whi~h this drive is
obtained, and hence controls the simulated
The dashed 1 ines on Figure 3 show model production pressures. Figure 5 shows the
calculated recompaction behavior for two simulated bottomhole pressure during the
gridblocks where less dilation occurred first production cycle for different
than in the case represented by the solid values of the recompaction pressure. Also
lines. The resultant reduced recompaction shown are representative field data. A
compressibilities are readily apparent. relatively low recompaction pressure must
(All recompaction cases shown in Figure 3 be used to reproduce the observed pressure
used an F value of 0. so.) This method decline, which supports the need to model
has the adaed advantage of requiring only an elastic period before recompaction
a single F fraction to define begins following dilation. We have also
history-dependa~t behavior of the observed that failure to match the

72
SPE 18752 C.I. BEATTIE, T.C. BOBERG & G.S. McNAB 5

production pressure decline will result in k = k 0 exp { ~L <%-¢0 )/(1-¢0 )} •... (2)
the simulator incorrectly predicting
interwell communication events during where k and Jt are the permeability and
multiwell simulations, since the pressure porosity values, k and ¢ are the
difference between injecting and producing original permeabilit~ and po~osity, and
wells significantly affects this Kmr is a user-defined multiplier.
communication. Interwell communication Eqtiktion 2 is applied separately in the x,
performance ' 8
can have a signifit~nr impact on y, and z directions, and different values
production and must of ~T may be specified in each gridblock
therefore be accurately modeled. for g~~ direction.
The residual dilation fraction FR largely The relationship of permeability r~
controls the amount of fluid wnich the porosity is not well defined, but Kozeny
reservoir can produce. Figure 6 shows how predicts that the permeability of an
total fluid production varies with F unconsolidated sand is proportional to the
during a first cycle simulation. Althoug§ cube of the porosity. This yields a
other parameters (such as relative permeability increase of 70% for a
permeabilities) affect the total fluid porosity ~rease of 2 0% • Dusseau! t and
production, the residual dilation fraction Rothenburg predict a permeability
has the largest impact without increase of 50% for a porosity increase of
substantially changing the produced 25% during shear failure. We have found
water-oil ratio. The value used for FR is that the simulation result is not
therefore chosen to provide a match of the particularly sensitive to changes on the
observed total fluid volume. We have order of the difference between these two
found that an F value in the range of predictions. We use a ~TL, value th~t
o. 4-0.5 usually \rovides the best match results in about a 70~u 1ncrease 1n
for Cold Lake. permeability at the highest porosi~y .t~at
our simulations reach. Permeab1l1t1es
Behavior in Subsequent Cycles decrease with porosity during
recompaction, but a residual permeability
When the pressure begins to increase increase exists as a result of the
following recompaction (during the next residual porosity increase described
injection cycle, for example),, the earlier.
reservoir initially behaves elast1cally
and the original compressibility is used.
This continues until the dilation function FRACTURE MODELING
is intersected, at which time dilation
again occurs. The subsequent dilation and It has long been known that steam
recompaction behavior is similar to that injection into the Cold Lake reservoir
described previously. If the pressure initially ga~~es tensi~e fracturing ~f t~e
begins to once again decline before the formation. ' Inclus1on of fractur1ng 1n
dilation line is intersected, then the reservoir simulators is required to obtain
elastic line is reversible, and the a reasonable representation of reservoir
gridblock will move back along it until behavior. This is particularly important
the recompaction line is intersected and to match interwell communici~i~g
recompaction continues. Figure 7 shows observations from the field. '
the behavior of a gridblock over three Published approaches to fracture modeling
cycles. In cycle two the block dilates have ranged from simple manual methods to
more than it did in the previous cycle. attempts at fully rigorous
In cycle three, the block· does not reach representations. The most ambitious
the dilation envelope, and returns along approaches involve direct computation of
the elastic line until it intersects the fracture dimensions in addition to
recompaction line and recompaction multiphase fl_ltf.~ 1 flow and heat transfer
resumes. calculations. ' A less advanced
approach is to incorporate fra2~~2t
In summary, the deformation model consists behavior using source/sink methods.
of non-reversible dilation and Simpler still are methods which assume an
recompaction lines which are joined by enhanced pressure-dependent fracture
reversible lines describing elastic permeability in a ~~ied column or
behavior. All of these paths are layer of gridblocks. ' The simplest
history-dependant and therefore behavior fracture modeling approach is to impose an
is unique for each gridblock. increased transmissibility during
injection on a predetermined group of
blocks in the fracture plane, returning
dur1ng product1on. ' 1 ' 1
Permeability Changes the.transmissi~ili~y Eo ~ts original value

An increase in the absolute permeability


of the reservoir will accompany the Some multiwell simulations we have run
porosity increase that results from have required as many as 4500 gridblocks.
dilation. We model the permeability It is therefore not practical to slow the
changes using Equation 2: performance of our simulator by use of the

73
6 RESERVOIR SIMULATION OF CYCLIC STEAM STIMULATION IN THE COLD LAKE OIL SANDS SPE 18752

more complex fracture modeling approaches. for the two methods, our approach has two
We simulate fracturing by use of special advantages. First, the fracture
porosity-dependent transmissibility gridblocks undergo deformation phenomena
multipliers in a specified plane of similar to all other blocks. All dilated
gridblocks. This approach allows the blocks will have the permanent residual
fracture length to vary from cycle to permeability increase that occurs at Cold
cycle depending upon leak-off conditions. Lake where not all dilation is
The method also allows the fracture to recoverable. If the fracture permeability
stop growing if breakthrough to a were instead modeled as a direct function
higher-mobility region near a neighboring of pressure, the fracture permeability
well occurs. could revert to the original reservoir
value at pressures below the fracture
The fracture plane permeability is pressure (unless special simulator logic
increased when the pressure exceeds the were used to prevent this). The fracture
fracture pressure by use of Equations 1 plane would then be less permeable than
and 2. The dilation pressure P is set adjacent dilated regions of the reservoir
equal to the desired fracture pre~sure for a condition which seems unreasonable. '
gridblocks in the specified fracture
plane. When the block pressure exceeds Most approaches to pressu25-ggpendent
this value, dilation occurs. However, transmissibility multipliers ' use a
much larger than normal values of ~T are permeability-pressure relationship which
specified for fracture plane blo~~~ so has a large discontinuity in slope at the
that the permeability calculated by fracture pressure. This causes numerical
Equation 2 reaches values much greater instability. Another advantage of our
than in non-fracture plane blocks when the method is that Equation 2 results in a
pressure is above fracture pressure. more smoothly increasing
permeability-pressure relation which
Use of this fracture model requires the significantly improves simulator
user to specify the fracture orientation, stability. In one single-well simulation
the fracture pressure, and the fracture of five css cycles, we obtained a
~nT value. Fractures can be either reduction of 4 0% in CPU time by using
vg~~ical or horizontal depending upon the Equation 2 compared to that observed when
in situ stress state. The orientati~~ is using a direct permeability-pressure
usually determined by stress testing, or functionality to model the fracture plane
from temperature observations made 2g permeability.
producing or observation wells.
Fracture pressure can also be obtained by RELATIVE PERMEABILITY HYSTERESIS
in situ stress tests. Fracture pressure
for horizontal fractures can be determined Many simulation studies of css in heavy
by integration of the density log to oil reservoirs result in the observation
determine overburden stress. I<vrr values that the use of experimentally-derived
are obtained by history matchil\~.L A low water-oil relative permeability data leads
value of ~L. will keep injected fluid to production cycle water-oil ratios that
close to ~he well, resulting in are much g!ia~gr 3 0than those observed in
unreasonably high injection pressures and the field. ' ' Matching the field
high produced water/oil ratios. Larger observed WOR requires reduction of the
~T values will cause steam to move water relative permeability to well below
fg~ber away from the well, lowering both the measured values, especially at low
the injection pressures and the computed water saturations. However, when this is
WORs during production. For single well done, it is difficult to obtain simulated
models, we have found that simulation injectiy~t~gs that are as high as observed
results are relatively insensitive to Kmr values. ' It has been concluded from
values as long as sufficiently la~g~ these observations that heavy oil
values are used. Better estimates of ~nT reservoirs must exhibit !i12;i~ 0
values are obtained from multiwell modef~~ permeability hysteresis. ' '
in which proper ~k values are critical Hysteresis has been experimentally
to matching vt~erved interwell observed (Reference 3 o provides a review
communication events. In order to match of experimental results). -Hysteresis in
these events we have found it necessary to both oil and water phases has been
use ~nrL values that result in fracture reported for a heavy oil reservoir. 31 we
trans~~ssibilities (at the injection have observed oil and water relative
pressure) that are 100-500 times those of permeability hysteresis experimentally for
the reservoir transmissibility. Cold Lake cores, and therefore include it
in our simulations.
In other reported applications of
pressure-dependent transmissib~si~~ Several methods of modeling relative
permeabil!~y 29
'3 '
multipliers for fracture modeling, ' sy~~eresis have been
the permeability is a direct function of reported. ' Some workers have
pressure. In our approach, permeability switched manually from the imbibition to
is a function of porosity, which is in ~h7 ~rai12~gEJo curve at the end of
turn a function of pressure. While the 1nJect1on. ' However, this approach
ultimate effect on permeability is similar produces a discontinuity in the relative

74
SPE 18752 C.I. BEATTIE, T.C. BOBERG & G.S. McNAB 7

permeability of a gridblock, which causes have significantly different oil or water


numerical instability. The approach is permeabilities as a result of different
physically unrealistic as well. Others saturation histories.
have made relative permeability a function
of pressure, assuming that imbibition The 'improved simulator match of produced
occurs only during injecr~on, and drainage water-oil ratios made possible by using
only during production. However, this relative p~~W:tJ1 hysteresis is well
assumption is not always valid. water documented. ' ' ' The use of
saturations can decrease locally during hysteresis also affects the calculated
injection when an oil bank develops and is fluid distributions and failed zone
moved away from the well by injected geometry. Figure 9 shows the effect of
steam. Similarly, water saturation relative permeability hysteresis on
increases can occur during production calculated profiles of water saturation
after the oil bank is produced. In and porosity normal to a vertical fracture
addition, this method can fail when following first cycle steam injection.
injection pressures are low, and may give The fluid front and dilated zone (as
misleading results for simulations of reflected by increased porosity) move
conversion from css to low-pressure farther into the reservoir when hysteresis
steamfloods. is included. This results in greater
injectivity and less extensive fractures
Since imbibition and drainage correspond during steam injection than when relative
to increasing and decreasing water permeability hysteresis is not used.
saturation respectively (for a water-wet Because the advance of the fluid and
sand), relative permeability hysteresis porosity fronts determines the timing and
should be modeled as a function of magnitude of interwell communication
saturation. our approach is to define events, inclusion of relative permeability
bounding imbibition and drainage curves hysteresis is particularly important in
with common endpoints, as shown in Figure multiwell simulations.
8. When a saturation reversal occurs, the
simulator calculates a scanning curve that The imbibition and drainage bounding
defines how the relative permeability will curves and the scanning curve exponents
move towards the other bounding curve. (see Appendix) must be defined to use this
Care must be taken to calculate the relative permeability hysteresis model.
scanning curves using a method which The imbibition curve is best obtained
guarantees that calculated relative experimentally, by injecting water into a
permeabilities always remain on or between core initially containing oil and
the bounding curves. While some of the irreducible water. (This is an imbibition
earlier methods did not always ensure that process in a water-wet system.) The
this would happen, the method presented experimental curve is then adjusted if
here has been devised to accomplish this. necessary to honor any available data for
The Appendix provides mathematical details the effective permeability to water at
of our approach. We use hysteresis only original reservoir conditions. This can
for the water-oil two-phase relative be derived from analysis of observation
permeabilities. Three-phase well presE1we responses to nearby steam
permeabilities are calculated using the injection. The bounding drainage curve
gas-oil two-phase curves, the hysteretic is obtained experimentally by injecting
water-g~l curves, and Stone's first oil into a core initially containing water
model. and residual oil. However,
history-matching of the water-oil ratios
Figure 8 shows the water relative observed in the field will normally
permeability paths followed by a gridblock require adjustment of the experimental
over several cycles in one of our css drainage bounding curve.
simulations. The gridblock water
saturation initially increases from 0. 25 The scanning curve exponents determine how
to 0.49. During this time the water rapidly the relative permeability
relative permeability moves along the approaches the bounding curve after a flow
imbibition bounding curve. When the reversal occurs. Imbibition and drainage
saturation reaches a value of 0.49, a scanning curve exponents must be provided
saturation reversal occurs and the for both water and oil. These values are
relative permeability moves along a extremely difficult to determine
scanning curve towards the drainage experimentally, and are normally used as
bounding curve. When the saturation falls history-match parameters in the
to 0. 34, a second reversal occurs, water simulation. We have found it best to use
saturation begins to increase, and a new scanning curve exponents that move the
scanning curve is followed as the relative permeabilities rapidly toward the
imbibition curve is approached. Further drainage bounding curve, but more slowly
scanning curves are followed after toward the imbibition bounding curve. The
saturation reversals occur in later relative permeabilities actually employed
cycles. This example demonstrates the therefore tend to be closer to the
history-dependent nature of the relative drainage boundi.ng curve than the
permeabilities, and illustrates how imbibition bounding curve throughout much
gridblocks with the same saturation can of the simulation.
75
8 RESERVOIR SIMULATION OF CYCLIC STEAK STIMULATION IN THE COLD LAKE OIL SANDS SPE 18752

CURRENT RESEARCH CONCLUSIONS

The simulator incorporating the 1. A reservoir deformation model has


fracturing, deformation, and relative been developed which appropriately
permeability hysteresis models presented represents the main features of oil
herein appropriately represents these key sand d~lation and recompaction
aspects of reservoir behavior. An occurring during cyclic steam
illustration of the excellent match of stimulation at Cold Lake. The model
well performance that can be achieved in a allows the simulator to match
single well simulation using these models injection and production pressures
is given in Reference 1. These models are and flowback times that are otherwise
nevertheless largely empirical. The difficult to reproduce.
greatest deterrent to rigorous modeling is
the current lack of complete understanding 2. The relative permeability hysteresis
of these phenomena. Even if the detailed model presented herein provides a
physics were better understood, we would simple and effective means of
still be faced with the inability of modeling the history-dependent
current computers to handle the rigorously hysteretic behavior exhibited by the
defined problem for an adequate number of Cold Lake Clearwater reservoir.
gridblocks. Esso Resources Canada Limited
is currently proceeding with research 3. The reservoir simulator is an
programs in both the laboratory and the important tool in developing an
field to measure deformation behavior with understanding of the complex physical
the aim of better defining reservoir phenomena occurring during cyclic
deformation during css. Efforts are also steam stimulation in tar sand
continuing to improve experimental formations.
measurement of relative permeability
hysteresis behavior, and in particular to NOMENCLATURE
better define movement between the
bounding curves. A constant in Equation A4

The development of the reservoir B constant in Equation A4


deformation and relative permeability
hysteresis models for simulation of Cold C compressibility (MPa- 1 {psi- 1 })
Lake css illustrates an important utility
of reservoir simulation. We originally FR =
residual dilation fraction
conceptualized the deformation behavior by
speculating on how the simulator would k = absolute permeability (m 2 {md})
have to be modified to improve on the
pressure matches provided by the spongy krw = relative permeability to water
rock approach. This was done largely
without the benefit of published K = ratio for scanning curve calculation
information on oil sand geomechanical defined by Equation A3
behavior. Much of our early geomechanics
research was aimed at investigating the RMDL = permeability multiplier (constant
validity of the initial conceptual model in Equation 2)
being used in the simulations. The bulk
of the simulation-based theorv has n = scanning curve exponent
subsequently been verified.I 4 , 15 ,~ 1
Similarly, our experiments on relative p pressure (MPa {psi})
permeability hysteresis were initiated as
a. resul ~ o~~~ <j<?f.:S\rmation ~f published pressure at which dilation begins
s~mulat~ons ' ' ' suggest~ng that the (MPa {psi})
hysteresis phenomenon in heavy oil
reservoirs is required to explain the pressure at which recompaction begins
observed high injectivity and low produced (MPa {psi})
water-oil ratio~! our experiments and
those of others verified the existence SW = water saturation (fraction)
of the hysteresis phenomenon for heavy oil
sands. This illustrates how reservoir ~ = porosity (fraction)
simulation can be used to improve the
understanding of reservoir physics even Subscripts
when the representation of these physics
in the simulator is initially overly i = imbibition
simplified. The inability of the
simulator to match field results, or the irr = value at irreducible water
changes required to achieve a match, can saturation
provide significant insight into the
actual physics occurring. This is a d = drainage
useful vehicle for identification of the
need for new research projects to better MAX = maximum value
define reservoir physics.

76
SPE 1.8752 C.I. BEATTIE, T.C. BOBERG & G.S. McNAB 9

0 value at original reservoir conditions 6. de Souza, J.F.C "Commercial Success


for Esso at Cold Lake", present"ed at
p value when saturation reversal occurs the AOSTRA Seminar on Advances in
Petroleum Recovery and Upgrading
ro = value at residual oil saturation Technology, Calgary, Alberta, June
12-13, 1986.
REF = reference value
7. Scott, J.D. and Kosar, K.M.
"Geotechnical Properties of Athabasca
Superscripts Oil Sands", paper presented at the
Western Research Institute u.s.
* = normalized value D.O.E. Tar Sand Symposium, Vail,
Colorado, June 26-29, 1984.
ACKNOWLEDGEMENTS 8. Agar, J.G. "Behavior of Oil Sands at
Elevated Temperatures", Ph.D. thesis,
Many people provided valuable comments and University of Alberta, Edmonton,
suggestions during the course of this Alberta (1984).
work. In particular, the contributions of
P.R. Kry, T.W. Miller, M.Y. Kwan, S.V. 9. Dietrich, J. K. "Cyclic Steaming of
Bharatha, R.C.K. Wong, E.S. Denbina and Tar Sands Through Hydraulically
M.B. Rotter are appreciated. The Induced Fractures", SPE Res. Eng. ,
permission of Esso Resources Canada May, 1986, 217-229.
Limited to publish this paper is
gratefully acknowledged. 10. Duerkson, J.H., Cruickshank, G.W.,
and Wasserman, M.L. "Performance and
Simulation of a Cold Lake Tar Sand
REFERENCES Steam Injection Pilot", JPT, October,
1984, 1781-1790.
1. Denbina, E.S., Boberg, T.C. and
Rotter, M.B. "Evaluation of Key 11. Settari, A. and Raisbeck, J.M.
Reservoir Drive Mechanisms in the "Analysis and Numerical Modeling of
Early Cycles of Steam Stimulation at Hydraulic Fracturing During Cyclic
Cold Lake", paper SPE 16737 presented Steam stimulation in Oil Sands" ,JPT,
at the 62nd Annual Technical November, 1981, 2201-2212.
Conference of the Society of
Petroleum Engineers, Dallas, Texas, 12. Coats, K.H., Ramesh, A.B., and
September 27-30, 1987. Winestock, A.G. "Numerical Modeling
of Thermal Reservoir Behavior", The
2. Coats, K.H. "A Highly Implicit Oil Sands of canada-Venezuela. 1977,
Steamflood Model", SPEJ, October, CIM Special Volume 17, 399-410,
1978, 369-383. Montreal (1977).
3. Kendall, R.P., Morrell, G.o., 13. Pethrick, W.D., Sennhauser, E.S., and
Peaceman, D.W., Silliman, W.J., and Harding, T.G. "Numerical Modelling
Watts, J.W. "Development of a Optimization of Cyclic Steam
Multiple Application Reservoir stimulation in Cold Lake Oil Sands",
Simulator for use on a Vector paper 86-37-21 presented at the 37th
Computer", paper SPE 11483 presented Annual Technical Meeting of the
at the Middle East Oil Technical Petroleum Society of CIM, Calgary,
Conference of the Society of Alberta, June 8-11, 1986.
Petroleum Engineers, Manama, Bahrain,
March 14-17, 1983. 14. Settari, A., Kry, P.R., and Yee,
c.-T. "Coupling of Fluid Flow and
4. Gallant, R.J., and Dawson, A. G. Soil Behavior to Model Injection into
"Evolution of Technology for Unconsolidated Oil Sands", paper no.
Commercial Bitumen Recovery at Cold 88-39-72 presented at the 39th Annual
Lake", paper no. 227 presented at the Technical Meeting of the Petroleum
4th UNITAR/UNDP International Society of CIM, Calgary, Alberta,
Conference on Heavy crude and Tar June 12-16, 1988.
Sands, Edmonton, Alberta, August
7-12, 1988. 15. Dusseault, M.B. and Rothenburg, L.
"Shear Dilatancy and Permeability
s. Mainland, G.G., and Lo, H.Y. Enhancement in Oil Sands", paper no.
"Technological Basis for Commercial 32 presented at the 4th UNITAR/UNDP
In Situ Recovery of Cold Lake International Conference on Heavy
Bitumen", paper RTD3(1) presented at Crude and Tar Sands, Edmonton,
the 11th World Petroleum Congress, Alberta, August 7-12, 1988.
London, England, August 28
September 2, 1983.

77
10 RESERVOIR SIMULATION OF CYCLIC STEAM STIMULATION IN THE COLD LAim OIL SANDS SPE 18752

16. Ito, Y. "The Introduction of the Services Spring Colloquium for the
Microchanneling Phenomenon to Cyclic Geosciences, Seattle, Washington,
Steam Stimulation and Its Application 1983.
to the Numerical Simulator (Sand
Deformation Concept)", SPEJ, August, 27. Gronseth, J.M., and Kry, P.R.
1984, 417-430. "Instantaneous Shut-In Pressure and
Its Relationship to the Minimum In
17. Roscoe, K.H., Schofield, A.N. and Situ Stress", Hydraulic Fracturing
Wroth, C.P. "On Yielding of Soils", Stress Measurements, M.D. Zoback and
Geotechnique, vol 8(1958), 22-53. B.C. Hainson (ed.), National Academy
Press, Washington, D.C. (1983).
18. Vittoratos, E.S., Scott, G.R., and
Beattie, C.I. "Cold Lake Steam 28. Vittoratos, E.S. "Interpretation of
Stimulation: A Multi-Well Process", Temperature Profiles From the
paper SPE 17422 presented at the Steam-Stimulated Cold Lake
Society of Petroleum Engineers Reservoir", paper SPE 15050 presented
California Regional Meeting, Long at the 56th California Regional
Beach, Calif., March 23-25, 1988. Meeting of the Society of Petroleum
Engineers, Oakland, California, April
19. Scheidigger, A.E. "The Physics of 2-4, 1986.
Flow Through Porous Media", 3rd ed.,
University of Toronto Press, Toronto, 29. Dietrich, J.K. "Relative Permeability
Ontario (1974). During Cyclic Steam Stimulation of
Heavy Oil Reservoirs", JPT, October,
20. Farouq Ali, S.M., and Blunschi, J. 1981.
"Cyclic Steam Stimulation with
Formation Parting", paper 83-34-45 30. Bang, H.W. "Simulation Study Shows
presented at the 34th Annual Hysteresis Effect on Oil Recovery
Technical Meeting of the Petroleum During a Cyclic Steam Process", Oil
Society of CIM, Banff, Alberta, May and Gas J., February 27, 1984, 83-86.
10-13, 1983.
31. Bennion, D.W., Moore, R.G. and
21. Settari, A. and Cleary, M.P. Thomas, F.B. "Effect of Relative
"Three-Dimensional Simulation of Permeability on the Numerical
Hydraulic Fracturing", JPT, July, Simulation of the steam Stimulation
1984, 1177-1190. Process", paper no. 83-34-46
presented at the 34th Annual
22. Nghiem. L.X. "Modeling Technical Meeting of the Petroleum
Infinite-Conductivity Vertical Society of CIM, Banff, Alberta, May
Fractures Using Source and Sink 10-13, 1983.
Terms", SPEJ, August, 1983, 633-644.
32. Killough, J .E. "Reservoir Simulation
23. Geshelin, B.M., Grabowski, J.W., and with History-Dependent Saturation
Pease, E. C. "Numerical Study of Functions", SPEJ, February, 1976,
Transport of Injected and Reservoir 37-48.
Water in Fractured Reservoirs During
Steam Stimulation", paper SPE 10322 33. Stone, H.L. "Probability Model for
presented at the 56th Annual Fall Estimating Three-Phase Relative
Technical Conference of the Society Permeability", JPT, February, 1970,
of Petroleum Engineers, San Antonio, 214-218.
Texas, October 5-7, 1981.
APPENDIX
24. Lin, C.Y. "A New Approach for Derivation of the Relative
Simulation of Cyclic Steam Permeability Hysteresis Scanning Curve
Stimulation Above Fracture Pressure Equations
by Modifying a Thermal Simulator",
paper SPE 18077 presented at the 63rd There are four cases, water or oil
Annual Fall Technical Conference of relative permeabilities as water saturation
the Society of Petroleum Engineers, increases or decreases. The derivation
Houston, Texas, October 2-5, 1988. given here is for the water relative
permeability with increasing water
25. Soni, Y., and Harmon, R.A. saturation in a water-wet system. The
"Simulation of the Saner Ranch other three cases are similar. All
Fracture Assisted Steamflood Pilot", equations can be used for either an oil-wet
JCPT, vol. 25, no. 1, (Jan-Feb 1986), or a water-wet system.
57-70.
Figure 10 shows a pair of normalized
26. Coats, K.H. "Thermal Simulation of water relative permeability bounding
Tar Recovery in Hydraulically curves, one for increasing water saturation
Fractured Formations", paper (imbibition) and one for decreasing water
presented at the Boeing Computer saturation (drainage) . The point p refers

78
r~ ""';.:.. -~ .;-: i~·-
, ·~- ~ ~- ... _....-z:-._

SPE 18752 C. I. BEA'rriE, T.C. BOBERG & G.S. McNAB 11

to conditions in a block when a saturation the derivative of K with respect to Sw* be


reversal occurs. The normalized water
relative permeabilities and water
saturations in Figure 10 are defined as:
finite in the range 1 ~ s: ~ (S:)P. At

= 1, this derivative is infinite when n <


(Al) 1. Therefore, n ~ 1.

So, by combining Equations A3 and A7

(A2) and defining KP with Equation A3, the


( Sw) ro - ( Sw) irr * becomes:
expression for krw
The subscripts irr and ro refer to conditions
(~;wi)p-
at irreducible water and residual oil.
[ (krwi)p -
(k;w)p] [ 1 - s: ]"
There is an important restriction in (k;wd)p 1 (S:)p
this derivation; the bounding curves for
both oil and water depend only on water * (k;wi - k;wd)
saturation. where n -~ 1
(AS)
As the water saturation increases, the
block's water relative permeability will Since the other derivations are
attempt to rise to the imbibition bounding similar, only the final equations are
curve (Figure 10) . Define a ratio, K, that presented.
compares the distance between the block's
permeability and this bounding curve to the Normalized Water Relative Permeability.
total distance between the bounding curves, Normalized Water Saturation Decreasing
all at the same water saturation:

K
k;wi - k;w k;w = k;wd + [ (k;w)p - (k;wd)p] [~] n
(A3) (k;wi)p - (k;wd)p (S:)p
k;wi - k;wd
*(k;wi - k;wd)
where the subscripts i and d refer to the where n ~ 1
imbibition and drainage bounding curves, •••• • · • (A9)
respectively.
Normalized Oil Relative Permeability.
*

r
Now, krw can be calculated from Normalized Water Saturation Increasing
Equation A3 if the bounding curves and K
[ (k;0 )p - p]
are known. Assume a form for K (over 1 ~
k;o k;oi +
(k;od)p
(k:oi)
(kroi) p
[1 1- s; (s:)p
*(k;od - k;oi)
K = A + B (1 - s:) n (A4) where n ~ 1
(AlO)
where A, B, and n are constants.
where the normalized oil relative
There are three boundary conditions. permeability is:
First:
(All)
K = 0 at s: = 1 (AS) (kro) irr

which forces the scanning curve to approach Normalized Oil Relative Permeability.
the imbibition bounding curve. Thus, A = 0 Normalized Water Saturation Decreasing
and n > 0.

The second boundary condition forces (k;od) p - (k; 0 ) p] [


[ ( k;od) p - ( k;oi) p
s: ] n

the scanning curve to pass through the ( s:) p


initial point, p
*(k;od - k;oJ
K (A6) where n ~ 1
(Al2)

therefore: K (A7)

The third boundary condition requires that

79
1.4

1.2

......... -
12 -------------------------------------------·

11 ,,
,, . .•
,,,,''

•••• •• ···~.
0
a..
::::!1:
1.0

'-"'
,....... (/)
0.8
0 (/)
~ 10 w
............ I 0:::
I--
w I (/) 0.6
a::: :
:::::> I w
VJ 9 ~ SPONGY ROCK >
VJ
...... :: DILATION i=
(.) 0.4
a::: : w
D.. ; • FIELD DATA LL..
LL..
w
0.2
8 i
7~--------~----------~----------,---------~ 0.0
0 10 20 30 40 0~ 1 10 100 1000
INJECTION TIME (DAYS) COMPRESSIBILITY (GPa -l)
CD
C>
FIGURE 1: Simulator match of first cycle injection pressure FIGURE 2: Oil sand compressibility u

--------:INiriAl:l
¢MAXI FR = 8/A
/-------- I
c,I>M:r CONSOLIDATION /
,/
/ SAND
DEPOSITION
;
ELASTIC CURVE ~/,"

~,o~

1
/

!::::: ""
>-
1-- ~t:y..(j. (/)
/ END OF
0

~
cO~ a::: INJECTION

il
0..
~~
' )________ _
----------------
A
0
a.. END OF
PRODUCT!~~

-------------------- ¢o ~~=====~~-------------------~ ~
-- _j
-
8 - --- POINT OF VIRGIN

~J
MAXIMUM RESERVOIR
INITIAL ELASTIC COMPACTION CONDITIONS
¢o RESERVOIR
CD
CONDITIONS - - - - - - ____________ _ j
I I ~
P. P.0 INCREASING PORE PRESSURE --..
Po \.ft
R PRESSURE ..___ INCREASING EFFECTIVE STRESS
N
FIGURE 3: Reservoir deformation model FIGURE 4: Representation of Cold Lake sand behavior based on Ref. 17
12

10
PR
PR
PR
= 3.0 MPa
= 4.1
=
MPa
6.5 MPa
,..,
~

~
75001

........... 7000
w
--, ~.~.:.··"', ~

.t:':
!-,,,
.·;:~1·.,

~::.:,··

'08
a.. '\, PR =9.0 MPa
~
::::J
I
~:.~)~

' ·,.,
_J ~:.;~
::::!:
...._ • FIELD DATA 0
> \\'~;
w
0::::
::::J
6 -, 0
w
6500
(/)
(/)
''·,., (.)
::::J
w 4 0

----::-_':.~~:::-:::::::.:::::~:::~::,..,.
0:::: 0
a.. g: 6000
2

5500~------~--------~------~------~------~
0
0 30 60 90 120 150 0.0 0.2 0.4 0.6 0.8 1.0
PRODUCTION TIME (DAYS) RESIDUAL DILATION FRACTION, FR

FIGURE 5: Simulator match of first cycle production pressure FIGURE 6: Simulator predictions of first cycle produced fluid volumes

0.12~-----------------------,

END OF
CYCLE 2
,------------------------,
END OF \ ,
~

,r~---~----,.________
I
~
I 0.10 IMBIBITION /"/
BOUNDING CURVE / ,'
>- CYCLE 3 ~/ I ,
,'
I
I
/ I 0.08
!:::: \ .,,/' ,
,'
I
I
U')
I
0 3:: I
0::: I
0 ~0:. 0.06 I
a.. I
I

/ I

..
I

~
END OF 0.04 I
I

I
CYCLE 1 CYCLE 1
CYCLE 2 , / DRAINAGE
INITIAL RESERVOIR CYCLE 3 0.02 START ., ...... "'/ BOUNDING CURVE
CONDITIONS ~ CD
PRESSURE 0.00
0.1 0.2
-
0.3
---
0.4 0.5 0.6 0.7 0.8 0.9 ""N \JI

WATER SATURATION

FIGURE 7: Gridblock porosity history over three cycles FIGURE 8: Grfdblock water relative permeability history over three cycles
,""":;_''\
:··.~ ~ ...
~~··-

0.7
..,,;:
~- ~·:.
·~~>
0.6
:::!:
3: 0.5 0::
1.&.1
(/)
0..
0.4 WITH HYSTERESIS _J
----
0.3 --- 1.&.1
0::
0::
1.&.1
I-
. )p------"1.
(KRWI
0.45 <(
>- 3:
!:::: 0.42 0
(/) 1.&.1

~ 0.39
0 WITH HYSTERESIS
N
::J ------------------1
.
<(
------- --
0.. 0.36
-----
:::!:
0:: . )
co
N 0.33
0
z r-------· (KRWD P
0.30 0
0 5 10 15 20 25 30 0 (S~ )P
DISTANCE FROM FRACTURE (M) NORMALIZED WATER SATURATION

FIGURE 9: Water saturation and porosity normal to a vertical fracture FIGURE 1 0: Normalized Water Relative Permeability Curves

.•
01

"...,
\J'I

You might also like