Professional Documents
Culture Documents
Fractured Reservoirs
M. Bagheri,* SPE, and A. Settari, SPE, University of Calgary
冤 冥
1
⌬v 0 0
kn
n = , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (1) ⌬␦N ⌬⌫N
冦 冧 冦 冧
a − b⌬v 1
where n is the normal stress, ⌬v is the change in fracture aperture, ⌬␦S = 0 0 ⌬⌫S . . . . . . . . . . . . . . . . . . . . . . . (11)
ks
and a and b are the constants that have physical meaning. For value ⌬␦T ⌬⌫N
of n→⬁, the closure reaches its maximum: 0
1
0
ks
a
⌬vmax = . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (2) Capital indices represent the local coordinate system of the
b
fracture, which consists of normal, strike, and dip directions (de-
Joint normal stiffness, which is the ratio of change of normal noted by N, S, and T ). This equation will be used to find the
stress to normal closure, is defined as constitutive relation of jointed rocks next.
dn For fluid-saturated fractures, ⌫ in Eq. 9 is replaced by effective
kn = . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (3) stress traction vector ⌫⬘, defined as
d⌬v
⌫⬘ = ⌫ − p. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (12)
The physical meaning of a is found from the derivative of
This form of effective stress is different than that of the matrix
Eq. 1.
shown in Eq. 8, which includes Biot’s coefficient ␣. Difference in
⭸n 1 structure of porous materials and joints is the reason for recom-
冉 冊
kn = = . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (4)
⭸⌬v b 2 mending this simple form of effective stress.
a 1 − ⌬v Huang et al. (1995) developed the constitutive relation for a dry
a rock mass represented by an equivalent continuum, using a com-
At zero normal stress, normal stiffness is called initial normal bination of both constitutive relations of matrix and fractures and
stiffness and is shown to be based on the principle of energy conservation. For saturated rock
冏
masses, this approach is valid for cases where pressure difference
⭸n 1
kni = = . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (5) in fractures and matrix (with high-grain compressibility) is small
⭸⌬v n=0 a (Bagheri 2006). The following paragraphs explain this method for
Substituting Eqs. 2 and 5 into Eq. 4 will result in saturated fractured rocks.
Change in the total strain of the continuum because of effec-
kni tive-stress changes consists of two components: one from the in-
冉 冊
kn = . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (6)
⌬v 2 tact rock and the other because of fractures:
1−
⌬vmax ⌬ij = ⌬ijI + ⌬ijJ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (13)
turn is the product of aperture and the total fracture area of the
fracture set. The exact calculation of the porosity is nontrivial.
However, regardless of fracture orientation, total fracture area Af
can be approximated by:
Lx Ly Lz
Af = , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (22)
S
where S is fracture spacing of a fracture set and Lx , Ly , and Lz are
the length, width, and height of the gridblock, respectively. The
accuracy of this approximation was tested in a separate code that
calculates the exact total area of each fracture set (Bagheri 2006).
Considering the uncertainty regarding the properties and the dis-
tribution of fractures in reservoirs, it is observed that this approxi-
mation is sufficiently accurate even for low-intensity fracture sets.
The error inherent in this approximation approaches zero when
fracture intensity increases.
The final part of initialization concerns fracture-permeability
calculations. Permeability of fractures in fractured blocks is cal-
culated on the basis of the information we have about the orien-
tation and spacing of fracture sets, and the aperture of fracture sets
from the previous step. Details of the method of calculation of
fracture permeability used in this study can be found in Gupta et al. Fig. 2—Dual-media treatment in the solution phase.
(2001). To account for deviation from the cubic law (resulting
mainly from fracture roughness and tortuosity), permeability can
be multiplied by a correction factor. porosity changes, and fracture apertures are used for the calcula-
Calculated fracture porosity and fracture permeability, which tion of fracture permeability and porosity.
now represent characterization of the geological structure of the At the end of the geomechanical iteration, matrix porosity m
fractured reservoir, are passed back to the flow model, where these is always updated using the new effective stress associated with
properties are used for flow calculations. the matrix, which is −␣pm. However, the fracture porosity f and
Dynamic Coupling. The dual-porosity iterative coupling ap- fracture and matrix permeabilities km and kf , respectively, may or
proach of this study is shown schematically in Fig 2. This figure may not be updated. Fig. 2 shows the case where the geomechani-
highlights the stages of calculation of fracture and matrix flow cal iteration involves only m.
properties, and the way in which they are coupled to the geome- New fracture porosities are based on the current gridblock bulk
chanics module. For example, in the initialization stage, matrix and volume. Bulk volume is a dynamic parameter, which changes be-
fracture porosity and pressure are transferred to the geomechanics cause of applied stress by the amount of volumetric strain. Math-
module in each geomechanical iteration loop to build a new con- ematically, this is shown as:
tinuum pore pressure using Eq. 16. This pressure is treated as an
external force in the equations. BV = 共1 + v兲 ⭈ BVi , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (23)
New fracture-normal stiffness and flow properties are calcu-
lated in each geomechanical iteration loop using the effective where BV, BVi, and v are the current and initial bulk volume and
stress on the fracture plane, which is a function of the new fracture the total volumetric strain (compressive strains are negative), re-
pressure transferred from the flow model and total stresses. It spectively. The same adjustment of bulk volume must be per-
should be noted that the fracture flow properties are calculated formed in the flow model as well.
before the new total stress field is calculated. As a result, the total Coupling of fracture-flow properties can be treated in two
stresses used in these calculations are lagging one geomechanical ways. Like matrix porosity, fracture-flow properties can be up-
iteration behind the pore pressures. dated in every geomechanical iteration and passed back to the flow
After the calculation of continuum pore pressure, an equivalent model. The highly nonlinear nature of these properties makes the
constitutive matrix is built using the last normal stiffness of frac- convergence process slow. To alleviate convergence difficulties,
tures. This matrix is composed of static Young’s modulus and an alternative approach for coupling fracture-flow properties was
Poisson’s ratios of the rock, static fracture spacing, and dynamic found to be effective. In this method (shown in Fig. 2), fracture
normal stiffness of fractures. Using this matrix in stress/strain porosity and permeability are updated only at the end of each
calculations provides the total stress and strain of the pseudocon- timestep and used in the next timestep. In other words, these prop-
tinuum. The overall deformation is then decoupled as described erties do not partake in coupling iterations; they use the converged
below. Matrix strains are the basis for the calculation of matrix- coupling variables (such as pressure and stress).
冤 冥
n2x n2y n2z 2nxny 2nynz 2nznx kPa at the depth of 7262.5 m.
Different elastic moduli and Poisson’s ratios were assigned to
nxsx nysy nzsz nysx + nxsy nzsy + nysz nzsx + nxsz different parts of the model. The model was divided into seven
nxtx nyty nztz nytx + nxty nzty + nytz nztx + nxtz regions with different rock properties, as shown in Table 1. All of
the regions have the same grain modulus of 2.07×109 kPa. Initial
The entries of the transformation matrix Tm are the components horizontal stresses are 109 137 kPa, and initial vertical stress is
of normal, strike, and dip vectors of the joint set in the global 110 240 kPa at a depth of 4876 m. The gradients of horizontal and
coordinate system. The normal deformation of fractures is then vertical stresses are 22.378 kPa/m and 22.604 kPa/m, respectively.
obtained using Eq. 9 where ⌬⌫J is substituted for the fracture Three identical fracture sets, parallel to the coordinate axes are
normal effective stress, ⌫⬘N⳱nn−pf (the shear terms are ignored). assumed in all reservoir gridblocks. They have constant spacing of
This effective stress is also used to update the normal stiffness of 2 m, initial stiffness of 39.62 GPa/m, joint aperture at zero effec-
fractures (for the next geomechanical iteration) in every gridblock. tive stress of 0.44 mm, and minimum aperture of 0.022 mm (maxi-
This approach satisfies Eq. 13 and does not require explicit cal- mum of 95% closure). These parameters are used in calculation of
culation of fracture strains. equivalent moduli of fractured rock according to Eq. 21 and to
Matrix effective stress is obtained by the substitution of pm and calculate shape factor in the transfer function in the flow model
the total stress in Eq. 8. In the next step, matrix strains are obtained (for example, see Kazemi et al. 1976).
by substituting the static-rock-constitutive-matrix moduli and the Differences in effective stresses normal to these fractures in
calculated effective stress in Hooke’s law. It was shown by ex- different locations result in a distribution of fracture porosity and
amples for simple cases—the sum of the two deformations is equal permeability with depth, decreasing from the top to the bottom of
to the total deformation of the pseudocontinuum (Bagheri 2006). the reservoir. For instance, fracture permeability covers a range
from 88.17 md (kz) to 16.77 md (kx, ky) at the first layer. The
Results and Discussion minimum permeability in the entire reservoir is the horizontal
To demonstrate the importance of modeling fracture permeability permeability of 15.43 md at the bottom of the deepest layer, where
as a dynamic parameter, two independent dual-porosity tests are effective stress is the highest. The distribution is shown as a func-
explained here. In both examples, there are three sets of fractures, tion of reservoir block number from left to right in Fig. 4.
each parallel to one coordinate axis. In the first example, these Depletion Scenario. For the depletion scenario, a vertical well
fracture sets have the same deformation characteristics with the is located in the cell (18, 3, 2–5) at the top of reservoir, producing
same fracture spacing. In the second example, fractures deform 8000 m3/d of oil at surface conditions until the reservoir pressure
differently. is depleted to 20 000 kPa. The well is completed through the
bottom four layers of the reservoir. The model is run for 20 years.
Example 1. This example is loosely modeled after the Shearwater Three gridblocks—(1, 3, 1) at the bottom, (9, 3, 1) at the middle,
field (Kenter et al. 1998). The data, presented at the SPE Forum on and (18, 3, 1) at the top of the reservoir—are chosen to monitor the
Reservior Geomechanics (2000), are a good example of an over- behavior of the reservoir.
pressured, high-pressure/high-temperature field with large deple- To study only the effect of fracture closure on permeability,
tion. The actual field is not fractured, but the data were modified fracture porosity was decoupled from the geomechanical code and
to represent a hypothetical dual-porosity dead-oil reservoir. The its variation was calculated by use of the ordinary compressibility
structure of the reservoir and the stress model are shown in Fig. 3. method. Then the results of this model with dynamic permeability
were compared with the results of the same model using a static
fracture permeability of the same initial distribution. These cases
are referred to as “dynamic” and “static” throughout this paper.
冉 冊
WI = , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (25)
Rb demonstrate higher influence of dynamic permeability modeling in
ln +s reservoir performance.
Rw
As initial oil in place is the same in both cases, ultimate re-
where h is the thickness perpendicular to the flow direction, kx and covery would also be the same if enough time is devoted to pro-
ky are the permeabilities, Rw is the well radius, s is skin factor, and duction. However, the production rate is lower in the dynamic
Rb is the effective well radius and is given case. If production is continued, the rate of the pressure decline
will decrease more in the dynamic case, as a result of lower rate of
production. Eventually, the plots of production rate for the static
关共ky Ⲑ kx兲1 Ⲑ 2⌬x2 + 共kx Ⲑ ky兲1 Ⲑ 2⌬y2兴1 Ⲑ 2
Rb = 0.28 × . . . . . . . . . . . . (26)
关共ky Ⲑ kx兲1 Ⲑ 4 + 共kx Ⲑ ky兲1 Ⲑ 4兴
Fig. 10—Effect of fracture opening on injection. Fig. 11—Effect of fracture opening on the injector’s pressure.
response for the top and the middle of the reservoir because of the reservoir is horizontal and single layer). The only parameters that
time required for these blocks to see the pressure front. Changes in have been changed in the flow model are the permeability and
fracture permeability are larger than in the depletion case, for the porosity of fractures. Fracture permeability is reduced from its
same amount of effective-stress changes, as seen by comparing original value of 10,000 md to 161 md, and fracture porosity is
Figs. 9 and 14. Again, all gridblocks form a complete fracture- reduced from 1 to 0.14%. Also, shape factor was calculated on the
deformation path, showing the fundamental dependency on effec- basis of the fracture spacings shown in Table 2, and matrix per-
tive stress. meability was multiplied by 0.1. We have extended the test to a
At high injection rates, it is possible to generate negative (ten- coupled model using the stress model shown in Fig. 16. The stress
sile) effective stresses. Although it is clear from Fig.14 that hori- model has seven layers of gridblocks, which are under compres-
zontal effective stresses in our example remain compressive, this sive stresses. Initial homogeneous stress of 27 200 kPa, at the
situation is common in injection processes. Fracture-deformation reference depth of 1196.6 m with a gradient of 22.6 kPa/m, is used
equations used here are developed for positive effective stresses. If to initialize all the stresses.
one attempts to use them for negative effective stresses, caution is Three sets of fractures with different deformation characteris-
required because fracture deformation is very sensitive to change tics are used to populate all reservoir grids. Under normal stress,
in stress and this makes the problem highly nonlinear. However, in these fracture sets deform according to Fig. 17. Table 2 shows the
this case the joint theory is likely not valid and induced (propa- deformation data of these fracture sets, which were chosen such
gating) fractures will form, starting from existing natural fractures. that the fracture permeability at initial stress is isotropic. Fractures
Modeling of such problems that combine fractured media and can close to a maximum of 90% of their aperture at zero effective
induced fractures is the subject of future work. stress.
Dynamic modeling of fracture permeability also improves the An injection well that injects water at the rate of 31.8 m3/d is
run times. Fig. 15 shows the run times for all four cases explained located at gridblock (1, 1, 4) of the stress model and a production
above. well at the opposite corner of the reservoir produces oil at the rate
of 33.4 m3/d, until the reservoir pressure is depleted to 3447.4 kPa.
Example 2. In the previous example, we focused on the effect of Pressure distribution inside the reservoir will be shown for early
dynamic fracture permeability on well productivity or injectivity. times when average reservoir pressures are almost the same for the
This example demonstrates its effect on pressure distribution and static and dynamic cases because of approximately equal amounts
creation of anisotropy because of differences in properties of the of fluid injections and withdrawals. The difference is caused by
fracture sets. The data are based on the classical example of Ka- slight differences in water production (water cut).
zemi et al. (1976). The effects of gravity are not present (the Fig. 18 depicts the pressure distribution at three different times
along the diagonal that connects the injector and producer. At early
Fig. 16—Structure of stress and the reservoir model (Example 2). Fig. 17—Deformation of fracture sets under normal stress.
Fig. 20—Comparison of pressure distribution at the reservoir Fig. 21—Comparison of pressure distribution at the reservoir
edges near injector. edges near producer.