Summary

Coupled geomechanical and reservoir modeling is becoming feasi-

ble on a full-field scale. This paper describes the advances of the
coupled model described previously,
1
its extensions for modeling
compaction, and the application of the model in full-field studies.
The advances of the theory and numerical implementation since
the original work
1
make it practical to perform full-field coupled
studies with complex, realistic descriptions of the geomechanical
behavior of the reservoir and shales, which allows prediction of
stress changes, reservoir compaction, and surface subsidence.
These capabilities are demonstrated in field examples that analyze
and predict classical pressure-induced compaction in a gas field
and thermally induced compaction in a heavy-oil field. In both
cases, the methodology of interpreting the geomechanical labora-
tory and field data and their integration in the coupled modeling
process was the key to obtaining a realistic predictive tool. The
examples demonstrate that the technology is maturing to the point
that conventional studies can be converted to coupled modeling on
a fairly routine basis.
Introduction
The geomechanical behavior of porous media has become increas-
ingly important to hydrocarbon operations. Numerical modeling of
such processes is complex and has been carried out historically in
three separate areas: geomechanical modeling (with the primary
goal of computing stress/strain behavior), reservoir simulation
(essentially modeling multiphase flow and heat transfer in porous
media), and fracture mechanics (dealing in detail with crack prop-
agation and geometry). Amodular system has been developed cou-
pling these three modeling components in such a manner that the
already highly developed modeling techniques for each component
can be used fully.
1
This model has been applied to several geome-
chanical/reservoir problems assisting in reservoir development.
The paper will first discuss the theory of different degrees of
coupling and its consequences for the formulation of the constitu-
tive models and running efficiency of the software. Next, the
modeling of compaction by rigorous means (plasticity) and its
simplifications, which lead to a considerable increase of computa-
tional efficiency, will be presented. In addition to classical
pressure-depletion-induced compaction, the paper will describe the
theoretical and modeling aspects of thermal compaction phenomena,
which have been observed in some applications.
Methods of Coupling
The key idea in the modular coupled system is the reformulation of
the stress-flow coupling so that the conventional stress-analysis
code can be used in conjuction with a standard reservoir simulator.
This is termed a partially coupled approach because the stress and
flow equations are solved separately for each time increment.
However, the method solves the problem as rigorously as a fully
coupled (simultaneous) solution if iterated to full convergence.
The coupling takes place through the use of interface code
developed to allow communication between simulators. In a
geomechanics/reservoir problem, for instance, the pressure and
temperature changes occurring in the reservoir simulator are
passed to the geomechanical simulator. The updated strains and
stresses are passed back to the reservoir simulator and used to com-
pute coupled parameters in the reservoir formulation (i.e., porosity
and permeability). An iterative method then must be used to obtain
convergence. The interface is flexible enough to allow the user to
choose several degrees of coupling. The degree of coupling may
affect the accuracy of the solution as well as the computational
efficiency; therefore, tradeoffs may be made to optimize run times.
To see the different degrees of coupling, consider first the
general formulation of the coupled problem in a finite-element set-
ting. After discretization in space and time, such a system can be
written in matrix form as
2,3
where [K]=the stiffness matrix,

=the vector of displacements,

[L]=the coupling matrix to flow unknowns, [E]=the flow matrix,
and

P=the vector of reservoir unknowns (i.e., pressures, satura-

tions, and temperatures). On the right side,

F=the vector of force

boundary conditions, and

R=the right side of the flow equations.
The symbol At denotes the change over timestep; i.e.,
Note that in the conventional reservoir simulation notation,4
[E]=[T]-[D], where [T]=the symmetric transmissibility matrix,
[D]=the accumulation (block diagonal) matrix, and

R=

Q-[T]

P
n
,
where

Q=the vector of boundary conditions (well terms).

Decoupled. Consider now the flow part of the coupled system
only, by assuming that At

=0. This is the assumption made in

reservoir simulation (i.e., stresses do not change), which gives the
familiar matrix equation
Conversely, if we assume that At

P=0, we obtain the classical elas-

ticity equations. In many stress analysis packages, pressure and/or
temperature can be imposed as external loads, which corresponds
to assuming that At

P is known. Then the top half of Eq. 1 can be

decoupled and written as
In practice, decoupled simulations can be carried out in several ways.
With a reservoir model only. To account for at least zero-
dimensional effects of stress changes, the compressibility terms in
the [D] matrix must be modified to account for the expected type
of containment in terms of deformations and the fact that the reser-
voir simulator uses a nondeforming grid (i.e., constant bulk
volumes of gridblocks). As a result, if the bulk compressibility cb is
. . . . . . . . . . . . . . . . . . . . . . . . . . . (4)
t t
F P =

K L
H H H
.
. . . . . . . . . . . . . . . . . . . . . . (3)
( )
n
t
P Q P =

T D T
KH KH KH
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . (2b)
1 n n
t
P P P
+
=
KH KH KH
. and
, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (2a)
1 n n
t
+
=
KH KH KH

. . . . . . . . . . . . . . . . . . . . . . . . (1)
t
t
F
R P

=


K L
L E
H H
H H
T
,
Advances in Coupled Geomechanical
and Reservoir Modeling
With Applications to Reservoir Compaction
A. Settari,* SPE and Dale A. Walters,* SPE, Duke Engineering and Services Inc.
* Now with Taurus Reservoir Services Ltd.
Copyright 2001 Society of Petroleum Engineers
This paper (SPE 74142) was revised for publication from paper SPE 51927, first presented
at the 1999 SPE Reservoir Simulation Symposium, Houston, 1417 April. Original manu-
script received for review 16 December 1999. Revised manuscript received 14 March 2001.
Manuscript peer approved 30 March 2001.
334 September 2001 SPE Journal
known, one can calculate compressibility c
R
I
(corresponding to free
deformation) or c
R
II
(corresponding to laterally constrained defor-
mation) to use in the reservoir simulator.
1
While this implicitly
assumes stress changes in each block, the stresses are not calculated.
Using a reservoir solution (Eq. 3), one can compute the entire
time history of

P and use it to compute a transient stress solution
with Eq. 4. This approach is being employed frequently because
conventional models can be used;
5,6
it essentially amounts to com-
puting two independent time histories with no coupling (at best, the
porosity or permeability relationships can be updated in specific
time intervals manually
7
).
A significant improvement to the reservoir model only
approach can be obtained by employing the zero-dimensional
stress solution assuming zero lateral strain. In an idealized case, the
change of average horizontal stress Aohavg can be computed from
change of average reservoir pressure Apavg by
Then, the coupling term can be formally included in Eq. 3.
In practice, because the reservoir model is usually finite-difference
and the stress model is finite-element, the coupling through the
[L]
T
matrix is reflected in porosity and permeability dependence on
stress. One would calculate the local effective stresses from block
pressure and average stresses and define porosity and permeability
as a function of the effective stress state. This approach accounts
for overall depletion effects and has been used quite early
8
to
model stress-dependent permeability.
Explicitly Coupled. Explicit coupling can be achieved by lagging
the coupling terms one timestep behind. Starting with the reservoir
solution and known change of stress over the previous timestep,
At

n
, we first solve
Then, using the flow solution, AtP
n+1
, the stress solution is
advanced by
Again, the implementation does not involve the use of the [L]
matrix because the discretizations are usually different. Instead,
the porosity in Eq. 7 is expressed as a function of both pressure
and mean stress. The discretization of the accumulation terms in
Eq. 7 can then be done in a mass conservative fashion. The stress-
dependent terms in the [T] matrix can be treated explicitly in
terms of stress. A variation of this approach, where only the
coupling through the [T] matrix is considered (and the volume
coupling ignored), has been used for modeling cases where the
dominant feature is the stress-dependent enhancement of perme-
ability (i.e., waterflooding
9
).
The explicit coupling is a special case of the iteratively coupled
system,
1
in which only one iteration per timestep is performed (see
the next section).
Iteratively Coupled. The iterative method consists of a repeated
solution of the flow and stress equations during the timestep
according to
Obviously, when the iteration (Eq. 9) converges,

P
(v)
=

P
n+1
, and

(v)
=

n+1
the solution is identical to the fully coupled system, Eq.
1, provided that any iterative processes in either formulation have
been converged. Again, including the coupling term in Eq. 9a is
equivalent to expanding the porosity in the reservoir model as a
function of p, T, and mean stress, or I1, as detailed in Ref. 1. Thus,
the porosity in the reservoir model is determined directly by the
volumetric strain computed from the stress model, rather than by
some compressibility relation (this point has often been misunder-
stood). The coupling through flow properties (i.e., effect of stress
on matrix [T]) can be explicit ([T]=[T]
n
) or implicit ([T]=[T]
(r)
).
Convergence of the iteration on the volume coupling has been
established.
10
The iterative method as implemented in this work is
the most flexible. It includes the explicit coupling (one
iteration/timestep) and can be simplified further by specifying the
porosity vs. effective stress relation directly in the reservoir model.
This latter approach often increases computational efficiency with-
out sacrificing accuracy, as discussed in the next section.
Fully Coupled. The fully coupled approach has the advantages of
internal consistency, as the full system (Eq. 1) can be solved simul-
taneously with the same discretization (usually finite-element).
There are only a few models that currently treat multiphase
flow.
2,11,12
Large development efforts will be needed to bring their
flow-model capabilities on par with existing commercial (finite-
difference) simulators.
Modeling of Compaction
Modeling reservoir deformation is of considerable importance in
soft and/or thick reservoirs, where the results of compaction may
provide an important production mechanism, cause well failures,
and/or cause ground subsidence or heave with environmental con-
sequences. Field development of large compacting fields such as
Groningen, Wilmington, the Bolivar coast of Venezuela, or Ekofisk
led to the development of techniques for estimating compaction,
starting with the work of Geertsma
1315
and followed by a number
of modified reservoir models.
1619
The common feature of such
reservoir-compaction models was that the compaction is treated as
a 1D problem (uniaxial strain) by assuming that (a) only vertical
deformations take place and (b) each vertical column of blocks
deforms independently. Consequently, the porosity changes were
calculated by modifying the conventional compressibility cR based
on the results of uniaxial strain laboratory experiments, and the
stress problem was not solved. The relation between reservoir
compaction and surface subsidence was typically obtained by an
independent solution of a stress problem using the computed com-
paction as a boundary condition.
In the context of the thermal coupled model presented here,
both compaction and subsidence are obtained naturally as part of
the solution. Typically, the stress part of the model would be
extended upward to the surface as well as laterally so that any arching
effects of the over- and sideburden would be captured. Moreover,
the laboratory-compaction data are used to calibrate the
stress/strain (constitutive) model of the skeleton rather than to
define cR. Therefore, the coupled model can represent the material
behavior under general triaxial loading paths and can include
effects of shear and temperature. Finally, the model provides the
stresses and displacements necessary to analyze casing failure,
a frequent problem in compacting reservoirs.
20
In particular, the
casing-failure study in thermal operations
21
was performed with
the coupled system described here.
The constitutive model is the key element of the compaction
model [analogous to the pressure/volume/temperature (PVT)
model in compositional simulation, for example]. Elastic and
plastic deformations may occur in the reservoir, but we typically
associate compaction with the plastic deformations. Therefore, its
modeling requires an elastoplastic constitutive model. However,
useful approximations can also be obtained using a nonlinear
elastic model with hysteresis. These two approaches will be
described and compared next.
Elastoplastic Constitutive Model. A variety of plasticity models
may be used when modeling compaction. For the example consid-
ered here, a generalization of the Drucker-Prager model, including
. . . . . . . . . . . . (10)
( ) ( ) ( ) ( )
,
n n
t t
P P P

= =
H H H H H H
. where
. . . . . . . . . . . . . . . . (9b)
( 1) ( 1)
, 1,
t t
F P

+ +
= =

K L
H H H
and
, . . . . . . . . . . (9a) ( )
( 1) ( )
T
n
t t
P Q P
+
=

T D T L
H H H H

. . . . . . . . . . . . . . . . . . . . . . . (8)
1 1 n n
t t
F P
+ +
=

K L
H H H
.
. . . . . . . . . . . . . . (7)
( )
1
T
n n n
t t
P Q P
+
=

T D T L
H H H H
. . . . . . . . . . . . . (6)
( )
avg
T
n
t t
P Q P =

T D T L
H H H H
.
. . . . . . . . . . . . . . . . . . . . . . . . . (5)
avg avg
1 2
1
h
p

.
September 2001 SPE Journal 335
a compressive cap, will be used. The failure criterion is formulated
in terms of invariants of the effective stress tensor, with compres-
sive stresses defined as negative values. The following standard
stress invariants are used.
The Drucker-Prager yield surface is shown in Fig. 1. The yield
surface has two sections. One is the standard Drucker-Prager fail-
ure surface, denoted here as the cone. On the cone, the expression
for the yield surface is
The expression for the cap portion of the yield surface is
The equation F=0 defines the yield surface. The cap and cone
portions of the surface are constrained to meet at their common
point of tangency. If the friction angle is constant, the value of a
and the location of the transition point can be computed from the
following expressions.
The Drucker-Prager model is a so-called two-invariant model
because the yield function depends only on two of the stress invari-
ants. Hardening of the cap is an important part of the model
because this is the region of compaction. It is controlled in the
model by a user-defined relationship between the volumetric plastic
strain and the mean effective stress. The change of the porosity
during plastic hardening then corresponds to the observed com-
paction compressibility.
Nonlinear Elastic Constitutive Model. The nonlinear elastic
model used for the analyses of this paper is a modified hyperbolic
model.
22
The model varies the Youngs modulus E and bulk mod-
ulus B as a function of the mean stress as follows.
where Ei=the initial Youngs modulus, Bm=the tangential bulk
modulus, or=the reference stress (may be either the minimum
effective stress or the mean effective stress), Ke, Kb and ne, nb=the
constant and exponent parameters for describing the magnitude
and shape of the Youngs and bulk moduli, and pa=atmospheric
pressure. The tangential Youngs modulus Et is then computed with
the following formulas.
where od=the deviatoric stress (o1-o3), Rf =the failure ratio rep-
resenting the maximum deviator stress (calculated with a Mohr-
Coulomb failure criterion) to the ultimate deviator stress predicted
from the hyperbolic fit, and e1 and e2=the exponent parameters
23
that govern the behavior of the tangential Youngs modulus as the
deviator stress increases. The classical hyperbolic model
22
is
obtained by setting e1=2 and e2=1. Thus, for the classical model,
as the deviator stress increases, a reduction or softening of the
Youngs modulus occurs according to the value of Rf (normally
between 0.5 and 0.9). The friction angle d, used to calculate the
maximum deviator stress according to a Mohr-Coulomb criterion,
was allowed to decrease with increasing stress level as follows.
where d1=the value of the friction angle at a confining stress of pa
(1 atm), Ad=the reduction in friction angle for a 10-fold increase in
confining stress (1 log cycle), and o3=the minimum effective stress.
Comparison. The two constitutive models described previously
are both capable of predicting nonlinear stress/strain behavior. The
elastoplastic formulation is a much more rigorous approach in
dealing with the post-failure material behavior. The nonlinear
elastic formulation was historically developed for prefailure
behavior. It is a good representation for the stress-strain response
for many soils and soft rocks under standard triaxial loading at con-
stant confining stress up to a shear-induced failure.
22
Once shear
failure occurs (i.e., the failure criterion is reached), the hyperbolic
model is unable to implement post-failure phenomenon (i.e., strain
hardening or softening); rather, the stress path is restricted to the
elastic stress space.
This limitation in post-shear-failure behavior is not an issue in
the majority of reservoirs experiencing compaction effects because
the mechanism is a cap failure. Although the deviatoric stress may
increase during pressure depletion (considering uniaxial strain con-
ditions), the shear stress developed is likely still below the shear
failure criterion. Therefore, the hardening behavior associated with
compaction may be captured by the hyperbolic formulation by
using the mean effective stress as the reference stress. Thus, as the
mean effective stress increases owing to pressure depletion, the
moduli values will increase (Eqs. 18 and 19). Also, as the pressure
decreases, the deviatoric stress will increase (assuming uniaxial
strain conditions) according to
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . (23)
(1 )
(1 2 )
d
p

, . . . . . . . . . . . . . . . . . . . . . . . . . . . . (22)
3
1
log
a
p

=
, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (21)
( )
( )
max
ult
d
f
d
R

=
, . . . . . . . . . . . . . . . . . . . . . . . (20)
( )
1
2
max
1
e
e
d
t i f
d
E E R

1

= 1
, ;
1

]
)
, . . . . . . . . . . . . . . . . . . . . . . . . . . . . (19)
b
n
r
m b a
a
B K p
p

=


and
, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (18)
e
n
r
i e a
a
E K p
p

=


. . . . . . . . . . . . . . . . . . . . . . . (17)
2
1transition 2 2
1
p
p
X a K R
I
R

+
=
+
. and
. . . . . (16)
( ) ( )
2
2 2 2 2 2 2
1
p p p p
a K R R X R X K R = + + + + .
. . . . . . . . . . . . . . . . . . . . . . (15)
( )
2
2
1 2
cap 2
1
I X a R J
F
a
+
=
. . . . . . . . . . . . . . . . . . . . . . . . . . . . (14b)
( )
6 cos
3 3 sin
c
K

=

and
, . . . . . . . . . . . . . . . . . . . . . . . . . . (14a)
( )
2sin
3 3 sin
p

=

with
, . . . . . . . . . . . . . . . . . . . . . . . . . . (13)
cone 2 1 p
F J I K =
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (12b)
3
det( )
ij
J s = and
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (12a)
1
2 2 ij ij
J s s = ,
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (11c)
ij ij ij
s = ,
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (11b)
1
3
I
= ,
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (11a)
1 ii
I = ,
Fig. 1The Drucker-Prager yield surface.

J
2
X= -I1
a
b
R=
a
b
I
1transition
cone
cap
hardening
336 September 2001 SPE Journal
The negative sign indicates that the deviatoric stress will increase
with a decrease in pressure (negative Ap) and will soften the mod-
ulus according to Eq. 20. Therefore, the two mechanisms are in
competition. The hyperbolic parameters may be adjusted so that
one or the other dominates, as the lab data dictate.
An example lab data set has been used to illustrate that both con-
stitutive models may be adjusted to obtain a good representation of
the material behavior. Fig. 2 compares the raw data and best fits of
both models. The comparison shows that both constitutive models
may be used to match common compaction data. Tables 1 and 2
contain constitutive model parameters used for the history match.
The hyperbolic model cannot match the elastic and the plastic
part of the typical compaction path simultaneously, as shown in
Fig. 3. Although in some reservoirs both elastic and plastic defor-
mations occur (e.g., Fig. 4 of Ref. 7), sometimes the reservoir
stress is already close to the cap [i.e., the original elastic zone up to
the preconsolidation stress, opre, is very small (see Fig. 15 of Ref.
24)]. In this case, the hyperbolic model can be matched to the plas-
tic path, and the initial elastic loading can be ignored. The model
implemented also includes hysteresis of mechanical properties,
whereby a completely different set of parameters will be used (in
an incremental fashion) when the effective stress starts to decrease.
The hysteresis is used to model the elastic behavior along the
rebound curves, as shown in Fig. 3.
The hysteresis also offers a means to treat the general case in
which the initial part of the loading is elastic. If the preconsolidation
stress, opre, at which the cap is reached, is greater than the initial stress
state of the reservoir, it is sufficient to set the maximum historical
stress to opre and then initialize the model on the hysteresis curve.
The main advantage of using the nonlinear elastic model vs. the
elastoplastic model is computing efficiency. For example, a simple
seven-spot element-of-symmetry model with a 61016 grid was
run through a 10-year depletion scenario with both constitutive
models described earlier. Both models were run with the explicit
timestep coupling. In addition, the elastic model was run iterative-
ly coupled (converged with a pressure tolerance of 0.001). The
comparison of the surface displacement uz at the surface and run
times is shown in Table 3.
The elastoplastic model run times were approximately twice as
long as those of the nonlinear elastic model, and the fully coupled
model was also 3.5 times slower, while all the results were within
5%. Of course, run times will be problem-dependent, but for many
compaction problems, plastic failure begins at the onset of pressure
depletion. Thus, all elements are in plastic failure throughout
depletion, resulting in significant computing costs if the elastoplas-
tic model is used. There are some instances in which the nonlinear
elastic model may not be applicable (i.e., more complicated stress
paths that may occur near the wellbore), but for the majority of the
reservoir, a nonlinear elastic model can be used. Thus, significant
gains in computational efficiency can be obtained with little loss in
the accuracy of the solution.
Thermal Compaction
The constitutive models presented so far have neglected thermal
effects on the stress/strain behavior. In certain reservoir materials,
this may not be valid. Coussy
25
has presented a brief description of
incorporating thermal effects in the elasticity domain for a general
TABLE 1ELASTOPLASTIC PARAMETERS FOR
HI STORY MATCH
Youngs modulus 1.510
5
psi
Poissons ratio 0.3
Grain bulk modulus 310
9
psi
Cohesion 14.5 psi
Friction angle 30
X, initial intersection of cap with I1 axis 45.0 psi
User-defined hardening function
(
vp vs. mean)
vp
mean
(psi)
0 1.510
1
1.8010
2
2.610
2
5.3010
2
9.010
2
8.5010
2
1.910
3
1.2310
1
3.310
3

Fig. 2Compaction curve-example data.
27
28
29
30
31
32
33
34
35
36
37
0 500 1,000 1,500 2,000 2,500 3,000
mean
' , psia
P
o
r
o
s
i
t
y
,

%
Raw data
Nonlinear elastic model
Elasto plastic model

Fig. 3Compaction curve for nonlinear elastic model.

M e a n E f f e c t i v e S t r e s s ,

i

p r e
Rebound
Plastic compaction


TABLE 3COMPARISON OF DIFFERENT
SOLUTION METHODS
Model uz (ft) Run time (minutes)
Elastic, expl. coupling 3.64 6.75
Elastoplastic, expl. coupling 3.48 14.59
Elastic, fully coupled 3.45 24.37
TABLE 2HYPERBOLIC PARAMETERS
FOR HISTORY MATCH
Hyperbolic parameter Primary Curve Hysteresis Curve
Ke 485.0 700.0
ne 0.3 0.6
Kb 404.2 583.3
nb 0.3 0.6
Rf 0.85 0.85
Cohesion (psi) 14.5 14.5
1 (degrees) 40 40
(degrees) 12 12
pa (psi) 14.7 14.7

September 2001 SPE Journal 337

thermoporoelastoplastic formulation. In short, the equation gov-
erning the free energy of a system was modified to include thermal
effects in the hardening parameter governing the evolution of the
elastic domain. Therefore, temperature changes are incorporated
into a general loading function, which may cause thermal hardening
and plastic strain (depending on the direction of the temperature
change). This rigorous formulation was simplified and applied to
the nonlinear elastic model described earlier. The modifications to
the nonlinear elastic formulation are described next.
Modified Formulation of Thermoporoelasticity. The general
equation of thermoporoelasticity (compression is negative) may be
formulated as
where oij and :ij=total stress and strain components, p and T=pore
pressure and temperature; other symbols are defined in
the Nomenclature.
Eq. 24 was modified to include terms associated with thermal
compaction and hardening. The following equation was used to
represent the strain for an increment in temperature.
Here, Ko and no describe the shape of the thermal compaction
curve (as a function of temperature), and Tref=a reference temper-
ature (usually ambient or initial reservoir temperature). The vari-
able ocomp represents essentially the same material behavior as the
coefficient of linear thermal expansion, although ocomp may have a
different sign and is a nonlinear function of temperature. The ther-
moporoelastic formulation (Eq. 24) is modified so that oL is
replaced with the sum of the thermal expansion and compaction
effects of the material.
It is expected that any thermal compaction occurring (a nonrecov-
erable plastic compression of the matrix) would induce hardening
of the Youngs and bulk moduli of the material. Thus, the nonlin-
ear elastic constitutive model discussed previously was modified
as follows.
Amodification of the moduli will occur only when the reservoir is
heated or cooled from the initial reference temperature. The expo-
nents me and mb describe the shape of the modulus multiplier.
Hysteretic behavior may be used for these modified variables and
is based on the direction of temperature change as opposed to pres-
sure or effective stress.
Calibration With Lab Data. A hypothetical example of test data
will be used to demonstrate how to calibrate the model. The test
(similar to actual tests that led to the development of the method)
consists of a series of temperature loads applied to a sample under
isotropic confining stress conditions. The volumetric strain is
measured throughout the test. To calibrate the hardening parame-
ters, mechanical loading stages should follow each temperature
load. Fig. 4 presents a comparison of the hypothetical raw data and
a best fit using the nonlinear elastic model. The thermal com-
paction parameters required to history match the lab data are listed
in Table 4. The values for the thermal hardening parameters were
not calibrated using the data presented in Fig. 4 because alternating
mechanical and thermal loading is needed. All other parameters
required for the nonlinear elastic model are contained in Table 3.
The loading data shown in Fig. 4 are contrary to what would be
expected based only on the thermal expansion of the solid (which
would follow the unloading path of Fig. 4). A number of mecha-
nisms can be responsible for the compaction, including weakening
of bonds, grain rearrangement, and shrinkage of some mineral com-
ponents. The plastic hardening is evident from the curvature on Fig.
4. Finally, it should be noted that the amount of thermal compaction
will be dependent on the loading path and type of material, and the
case shown here should not be regarded as typical.
Field Examples
Thermally Sensitive Heavy-Oil Field. The following example
illustrates the thermal effects caused by steamflooding a heavy-oil
reservoir. The field consists of a flat, five-layer reservoir system
with 32 production wells interspersed with 17 water and steam
injectors. Table 5 lists the general reservoir properties used in the
example. The production wells were perforated in Layers 2
through 4, while the injectors were only perforated in Layer 4.
Lab Data Match. The lab data used for the example are pre-
sented in Figs. 3 and 4. As discussed, the nonlinear elastic model
provides a good match of the data and was used for this example.
The material shows sensitivity to both pressure decreases and
temperature increases, indicating that compaction may be an issue
depending on the magnitude of pressure depletion and heating of
the reservoir.
Full-Field Compaction. The full-field simulation consisted of
a 10-year production/injection period. All wells in the full-field
model were set to constant fluid-rate constraints for the full simu-
lation period. The production wells were set to produce at 450
BOPD with a minimum bottomhole pressure (BHP) of 15 psia. The
water injectors were set to 650 BWPD (at Tinj=Tres), and the steam
injectors were set to 1,000 BWPD (0.8 steam quality at
Tinj=550F).
The results of the simulation are illustrated in Figs. 5 through 7.
It is apparent that the thermal component of the compaction is
dominant, as the total compaction pattern follows the temperature
pattern. This is, of course, a function of the material behavior as
expressed by Fig. 4. For comparison, Fig. 8 shows the case in
which the thermal compaction component was removed and the
normal thermal expansion coefficient was used. This case shows
. . . . . . . . . . . . . . . . . . . . . . . . (28)
ref
b b
n m
m
m b a
a
T
B K p
p T

=


. . . . . . . . . . . . . . . . . . . . . . . . . (27)
ref
e e
n m
m
i e a
a
T
E K p
p T

=


. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (26)
tot comp L
= +
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (25)
comp
ref
n
T
K
T

=


, . . . . . . . (24)
1
2
1 2
ij ij v L ij ij
G T p



1 + | |
= +
1 ]

\
]
TABLE 4THERMAL PARAMETERS FOR HISTORY MATCH
Parameter Primary Curve Hysteresis Curve
L (1/ F) 5.610
6
5.610
6
K (1/ F) 8.3610
5
0.0
n 0.33 0.0
me 1.85 1.5
mb 1.85 1.5

Fig. 4Thermal compaction curve-example data.

31
32
33
34
35
36
0 100 200 300 400 500 600
Temperature,
o
F
P
o
r
o
s
i
t
y
,

%
Raw data
Nonlinear elastic model
338 September 2001 SPE Journal
general compaction resulting from pressure reduction and local
thermally induced expansion around steam injectors. In relative
terms scaled to the maximum compaction in the injector locations
of Fig. 7, the case of Fig. 7 shows general compaction of 0.08 to
0.1 units, while the case of Fig. 8 shows general compaction of
0.04 to 0.06 and heave of 0.06 to 0.09 around the injectors.
Gas Field. The model described here has been used in develop-
ment planning of a large offshore gas-condensate field. The model
has been used directly to forecast both reservoir compaction and
seafloor subsidence. The work to date has shown the importance of
obtaining correct laboratory data, as well as detailed simulation of
the reservoir surroundings.
While the complexity of the model does not allow detailed
description here, selected results are shown to illustrate these points.
Lab Data Match. Owing to considerable reservoir heterogeneity,
lab samples were obtained and tested for a wide range of porosi-
ties. The results were grouped for modeling into three material
types based on permeability ranges. An example of the stress/strain
and volumetric-strain match is shown in Fig. 9.
Example of Predictions. The reservoir model consists of a
521212 block grid with heterogeneous properties and consid-
erable structure. This model represents one of the fault blocks in
the field. Initially, a stress model was built with the same
521212 areal grid, assuming free deformations at the top of the
reservoir. This is the base-case scenario.
Amore rigorous model was created by extending the finite ele-
ment (FEM) grid for the stress solution above and below the reser-
voir grid, modeling compaction transferred to the seafloor as well
as any rebound below the reservoir. The resulting grid consisted of
521218 elements. This model (both the reservoir and FEM
grids) is shown in Fig. 10. Finally, a model was extended to the
flanks of the reservoir (parallel to the faulting), resulting in a
601218 grid.
The fault block is produced by six vertical wells completed over
six layers so that the reservoir is depleted from the initial pressure of
3,400 psi to approximately 1,800 psi after 20 years of production.
The time history of predicted compaction from these three
cases [at the areal location (27,3) for the sum of all reservoir lay-
ers] is shown in Fig. 11, normalized to the base case. It can be seen
that the compaction is not sensitive to the inclusion of the over-
burden in this case. This is because of the soft properties of the
overburden; in cases in which the reservoir is surrounded by hard-
er rock, significant arching may occur. To demonstrate this, anoth-
er case was run with stiffer overburden, which is also shown in Fig.
11 and results in decreased compaction.
The 20-year simulation took 176 minutes for the base case, 183
minutes for the overburden case, and 255 minutes for the overbur-
den+flanks case, on a 450 MHz Pentium with 500 MB of random
access memory (RAM). This compares with 13 minutes for an
uncoupled run with the reservoir model only. Although the coupled
modeling requires an order of magnitude more time, the run times
Fig. 5Pressure distribution after 10 years.
INJ1
INJ10
INJ11
INJ12
INJ13
INJ14 INJ15
INJ16
INJ17
INJ2
INJ3
INJ4
INJ5
INJ6 INJ7
INJ8
INJ9
PROD1
PROD10
PROD11
PROD12
PROD13
PROD15
PROD16
PROD17
PROD18
PROD19
PROD2
PROD20
PROD21
PROD22
PROD23
PROD24
PROD25
PROD26
PROD27
PROD28
PROD29
PROD3
PROD30
PROD31
PROD32
PROD4
PROD5
PROD6
PROD7
PROD8
PROD9
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
0 2,000 4,000 6,000
0
2
,
0
0
0

4
,
0
0
0

6
,
0
0
0

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
200
250
300
350
400
450
500
550
600
650
700
750
Layer 4, time = 3,650 days
TABLE 5RESERVOIR PROPERTIESEXAMPLE 1
Initial average reservoir pressure 559.5 psi
Depth to reservoir top 1300 ft
Reservoir temperature 123F
Horizontal stress gradient 0.665 psia/ft
Vertical stress gradient 0.95 psia/ft
Grid size: 32325
Layer Kh (md) Kv (md) z (ft) h (psia) z (psia)
1. Shale .001 .001 0.15 25 334.8 708.9
2. Reservoir 1,300 130 0.28 37.5 341.6 724.6
3. Reservoir 800 80 0.34 12.5 347.2 737.3
4. Reservoir 1,200 120 0.25 18.75 351.6 746.2
5. Shale .001 .001 0.15 31.25 357.2 758.9

September 2001 SPE Journal 339
(under 5 hours) are manageable and show that full-field coupled
simulations are quite feasible.
Conclusions
1. A coupled thermal reservoir/geomechanical model with modu-
lar coupling and elastic or elastoplastic rock behavior was
developed and applied to several field studies. The system is
now sufficiently advanced so that geomechanical modeling can
be used for reservoir planning in the same manner as conven-
tional reservoir simulators.
2. The coupling method developed offers different degrees of cou-
pling from independent flow and stress solutions to a fully cou-
pled system, with corresponding differences in run times. The
timestep explicit coupling provides the best compromise between
computing efficiency and accuracy for compaction problems.
3. Correct representation of the stress/strain (constitutive) model
of the skeleton is the key for realistic predictions. A nonlinear
elastic model with hysteresis can be used to approximate an
elastoplastic model for modeling compaction at a fraction of its
computer requirements.
Fig. 6Temperature distribution after 10 years.
INJ1
INJ10
INJ11
INJ12
INJ13
INJ14 INJ15
INJ16
INJ17
INJ2
INJ3
INJ4
INJ5
INJ6 INJ7
INJ8
INJ9
PROD1
PROD10
PROD11
PROD12
PROD13
PROD15
PROD16
PROD17
PROD18
PROD19
PROD2
PROD20
PROD21
PROD22
PROD23
PROD24
PROD25
PROD26
PROD27
PROD28
PROD29
PROD3
PROD30
PROD31
PROD32
PROD4
PROD5
PROD6
PROD7
PROD8
PROD9
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
0 2,000 4,000 6,000
0
2
,
0
0
0

4
,
0
0
0

6
,
0
0
0

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
140
160
180
200
220
240
260
280
300
320
340
360
380
400
420
440
460
480
500
Layer 4, time = 3,650 days
Fig. 7Compaction distribution-case with thermal compaction
effects.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
0 2,000 4,000 6,000
0
2
,
0
0
0

4
,
0
0
0

6
,
0
0
0

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
Layer 1, time = 3,650 days
Fig. 8Compaction distribution-case without thermal com-
paction effects.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
0 2,000 4,000 6,000
0
2
,
0
0
0

4
,
0
0
0

6
,
0
0
0

1
2
5
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
Layer 1, time = 3,650 days
340 September 2001 SPE Journal
4. Thermally induced compaction has been formulated and imple-
mented in the model. This effect is potentially important in
thermal projects.
5. Proper modeling of reservoir surroundings in terms of stress
can be important for the accuracy of compaction and/or subsi-
dence predictions.
Nomenclature
a = coefficient in the cap yield surface
B = bulk modulus
Bm = tangential bulk modulus
cb = bulk compressibility
cR = conventional rock compressibility
c
R
I
= equivalent cR in unconstrained system
c
R
II
= equivalent cR in laterally constrained system
[D] = flow accumulation matrix
[E] = flow equations matrix
e1, e2 = exponents in the generalized Youngs modulus
equation (hyperbolic model)
E = Youngs modulus
Ei = Initial Youngs modulus (Eq. 27)
Et = tangential Youngs modulus
Fcap = yield surface function
Fcone = yield surface function

F = vector of force/displacement boundary conditions

G = shear modulus
I1 = 1st stress invariant
J1,J2, = 2nd and 3rd stress invariants
[K] = stiffness matrix
Kb = constant for bulk modulus (hyperbolic model)
Ke = constant for Youngs modulus (hyperbolic model)
[L] = coupling matrix
me = thermal hardening exponent for modulus
mb = thermal hardening exponent for bulk modulus
ne,nb = exponents for Ei and Bm (hyperbolic model)
p = pore-fluid pressure
pa = atmospheric pressure (scaling constant)

P = vector of flow unknowns

Q = vector of source terms (wells)

[R] = residual of the flow equations

R = right side of the flow equations (Eq. 1)

Rf = failure ratio (hyperbolic model)
sij = total deviatoric stress components, sij=oij-ij o
[T] = transmissibility matrix
T = temperature
Tini = initial reservoir temperature
Tref = reference temperature
ux,y,z = displacements
X = intersect of the cap surface with I1 axis
oL = coefficient of linear thermal expansion
o = Biot constant, o=1-cs/cb
ocomp = thermal compaction coefficient
op = plastic parameter, Eq. 14
At = change over time interval (timestep)
Aoh avg = change of average horizontal stress
Apavg = change of average reservoir pressure

= vector of displacement unknowns

ij = Kronecker delta
:ij = strain components
:v = volumetric strain
:vp = volumetric plastic strain
r = Poissons ratio
d = friction angle for shear failure
od = deviatoric stress,=o1-o3
Fig. 9Example of stress/strain lab data match.
0
1,000
2,000
3,000
4,000
5,000
6,000
7,000
8,000
9,000
10,000
0 0.01 0.02 0.03 0.04 0.05
Uniaxial strain (=volumetric strain)
D
e
v
i
a
t
o
r

a
n
d

M
e
a
n

E
f
f
e
c
t
i
v
e

S
t
r
e
s
s
,

p
s
i
aObserved
Calculated
Deviator
Mean
Fig. 10Reservoir and FEM grid for the gas-field-compaction example.
September 2001 SPE Journal 341
oij = total stress components
opre = preconsolidation stress
or = reference stress
Superscripts
T = transpose (matrix)
Subscripts
max = maximum deviator stress
ult = ultimate deviator stress
Acknowledgments
The authors want to acknowledge a number of people who con-
tributed, directly or indirectly, to this work, particularly Wash
Wawrzynek of FAC, Alda Behie, Peter Puchyr, and Vladimir
Zhitomirsky. The plasticity model development was also support-
ed by Imperial Oil Research Ltd.
References
1. Settari, A. and Mourits, F.M.: A Coupled Reservoir and
Geomechanical Simulation System, SPEJ (September 1998) 219.
2. Lewis, R.W. and Schrefler, B.A.: The Finite Element Method in the
Deformation and Consolidation of Porous Media, John Wiley & Sons,
New York City (1987) 1344.
3. Vaziri, H.H.: Nonlinear Temperature and Consolidation Analysis of
Gassy Soils, PhD dissertation, U. of British Columbia, Vancouver,
British Columbia, Canada (1986).
4. Aziz, K. and Settari, A.: Petroleum Reservoir Simulation, Applied
Science Publishers, New York City (1979) 1476.
5. Chin, L.Y. and Boade, R.R.: Full-Field, 3-D Finite Element
Subsidence Model for Ekofisk, 1990 North Sea Chalk Symposium,
Copenhagen, Denmark, 1112 June.
6. Fredrich, J.T. et al.: Three-Dimensional Geomechanical Simulation of
Reservoir Compaction and Implications for Well Failures in the
Belridge Diatomite, paper SPE 36698 presented at the 1996 SPE
Annual Technical Conference and Exhibition, Denver, Colorado,
69 October.
7. Sulak, R.M., Thomas, L.K., and Boade, R.R.: 3D Reservoir
Simulation of Ekofisk Compaction Drive, JPT (October 1991) 1272;
Trans., AIME, 291.
8. Settari, A. and Price, H.S.: Simulation of Hydraulic Fracturing in
Low-Permeability Reservoirs, SPEJ (April 1984) 141.
9. Koutsabeloulis, N.C., Heffer, K.J., and Wong, S.: Numerical geome-
chanics in reservoir engineering, Computational Methods and
Advances in Geomechanics, Siriwardane and Zeman (eds.), Balkema,
Rotterdam, The Netherlands (1994) 2097.
10. Settari, A. and Mourits, F.M.: Coupling of geomechanics and reservoir
simulation models, Computational Methods and Advances in
Geomechanics, Siriwardane and Zeman (eds.), Balkema, Rotterdam,
The Netherlands (1994) 2151.
11. Gutierrez, M. and Lewis, R.W.: The Role of Geomechanics in
Reservoir Simulation, paper SPE 47392 presented at the 1998
SPE/ISRM Eurock, Trondheim, Norway, 810 July.
12. Koutsabeloulis, N.C. and Hope, S.A.: Coupled Stress/Fluid/Thermal
Multi-Phase Reservoir Simulation Studies Incorporating Rock
Mechanics, paper SPE 47393 presented at the 1998 SPE/ISRM
Eurock, Trondheim, Norway, 810 July.
13. Geertsma, J.: Problems of rock mechanics in petroleum production
engineering, Proc., First Congress of the Intl. Soc. of Rock
Mechanics, Lisbon, Portugal (1966) 1, 585594.
14. Geertsma, J.: A basic theory of subsidence due to reservoir com-
paction: the homogeneous case, Verhandelingen Kon. Ned. Geol.
Mijnbouwk. Gen. (1973) 28, 43.
15. Geertsma, J. and van Opstal, G.: Anumerical technique for predicting sub-
sidence above compacting reservoirs, based on the nucleus of strain con-
cept, Verhandelingen Kon. Ned. Geol. Mijnbouwk. Gen. (1973) 28, 63.
16. Finol, A. and Farouq Ali, S.M.: Numerical Simulation of Oil
Production With Simultaneous Ground Subsidence, SPEJ (October
1975) 411; Trans., AIME, 259.
17. Merle, H.A. et al.: The Bachaquero StudyAComposite Analysis of
the Behavior of a Compaction Drive/Solution Gas Drive Reservoir,
JPT (September 1976) 1107; Trans., AIME, 261.
18. Rattia, A.J. and Farouq Ali, S.M.: Effect of Formation Compaction on
Steam Injection Response, paper SPE 10323 presented at the 1981
SPE Annual Technical Conference and Exhibition, San Antonio, Texas,
57 October.
19. Chase, C.A. Jr. and Dietrich, J.K.: Compaction Within the South
Belridge Diatomite, SPERE (November 1989) 422.
20. Bruno, M.S.: Subsidence-InducedWell Failure,SPEDE(June1992)148.
21. Smith, R.J.: Geomechanical Effects of Cyclic Steam Stimulation on
Casing Integrity, MS thesis, U. of Calgary, Calgary (1997).
22. Duncan, J.M. and Chang, C.Y.: Nonlinear Analysis of Stress and
Strains in Soils, J. Soil Mechanics and Foundation Division, ASCE
(1970) 96 (SM5) 1629.
23. Settari, A. et al.: Geotechnical Aspects of Recovery Processes in Oil
Sands, Cdn. Geotech. J. (1993) 30, 22.
24. Colazas, X.C.: Subsidence, compaction of sediments and effects of
water injection, Wilmington and Long Beach offshore fields, MS the-
sis, U. of Southern California, Los Angeles (1971).
25. Coussy, O.: Mechanics of Porous Continua, John Wiley & Sons,
Chichester, U.K. (1995).
SI Metric Conversion Factors
bbl 1.589 873 E - 01 = m3
ft 3.048* E - 01 = m
F (F-32)/1.8 = C
psi 6.894 757 E + 00 = kPa
*Conversion factor is exact.
A. (Tony) Settari is president of Taurus Reservoir Solutions Ltd., a
petroleum and geomechanics engineering firm based in
Calgary, and holds the Petroleum Engineering chair at the U.
of Calgary. e-mail: Asettari@TaurusRS.com. He is a leading
expert in the area of reservoir engineering and computer sim-
ulation of petroleum reservoirs, and in the analysis of fracturing
and geomechanical processes in reservoirs. He has been
involved in a wide range of simulation applications, including
naturally fractured reservoirs, enhanced recovery projects,
hydraulic fracturing and acidizing, in-situ thermal processes in
oil sands, perforation mechanics, and geomechanics. Settari
holds a BS degree from the Technical U. of Brno,
Czechoslovakia and a PhD degree in mechanical engineer-
ing from the U. of Calgary. Dale Walters is a senior engineer
with Taurus Reservoir Solutions Ltd. in Calgary. e-mail:
Dwalters@TaurusRS.com He has been involved in a wide range
of simulation applications, including reservoir compaction,
hydraulic fracturing, thermal processes in oil sands, and
enhanced recovery projects. Walters holds BS and MS degrees
from the U. of Calgary, both in civil engineering.
SPEJ
Fig. 11Total reservoir compaction predicted with different
stress-boundary conditions.
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000
Time, days
N
o
r
m
a
l
i
z
e
d

R
e
s
e
r
v
o
i
r

C
o
m
p
a
c
t
i
o
n
Base case
Overburden
Over + sideburden
Stiffer overburden
342 September 2001 SPE Journal