Вы находитесь на странице: 1из 19

SPE 144560

A Three-Parameter Dual Porosity Model for Naturally Fractured Reservoirs


Hadi Jabbari, SPE, and Zhengwen Zeng, SPE, University of North Dakota

Copyright 2011, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Western North American Regional Meeting held in Anchorage, Alaska, USA, 711 May 2011.

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

Abstract

This study presents a new model to analyze the behavior of naturally fractured reservoirs (NFRs). The model considers
the geomechanical behavior of the fractured reservoirs and the corresponding effects on the fracture aperture, which may
be an important parameter for stress-sensitive NFRs. It includes the elasticity parameters, such as Poissons ratio and
Youngs modulus.

Pressure depletion in an NFR, inherent to production, can result in effective stress change that, in turn, may change
fracture permeability in NFRs. This may influence the behavior of NFRs, which has been studied thoroughly in the
literature; however, not analytically but only numerically. A rigorous mathematical model is developed in this study
which couples geomechanical and fluid flow aspects for characterizing a stress-sensitive NFR. In the model, we consider
a semi-infinite radial flow with flux boundary conditions at the wellbore (Neumann condition) and constant pressure at
infinity (Dirichlet condition). It is also assumed that the external forces acting to the reservoir are constant, that there is
interaction between the two regions, i.e. matrix and fracture system, via the change of pore pressure and effective stress.

Since the change in effective stress induced by reservoir compaction may affect the pressure buildup curves in NFRs,
new equations on buildup interpretation are developed to better understand this effect, and to precisely evaluate the
reservoir properties. An iterative algorithm is also developed which enables us not only to interpret the buildup data to
compute the fracture parameters, namely fracture porosity, fracture permeability, fracture storage capacity, and the
elasticity parameter, but also to evaluate them more accurately. Such a new technique improves NFR characterization
through the inclusion of the fourth dimension, time, into buildup interpretation. This new well test interpretation method
is a dynamic reservoir characterization, which is usually the mission of time-lapse (4-D) multicomponent seismology.

From this study, we found that knowing three parameters is sufficient to both distinguish a reservoir with dual porosity
from that of a homogenous one and to characterize a dual porosity reservoir. Two of the parameters represent the fracture
storage capacity and matrix-fracture interaction. A third one reflects the effect of both matrix geomechanics and fracture
aperture decline on the behavior of NFRs as time progresses. These parameters can be obtained by precisely interpreting
drawdown or buildup tests. This solution is applicable to any reservoir containing dual porosity rocks.

Introduction

Rocks in reservoirs are usually fractured to some extent. This results in forming widely spaced joints, due to mild
fracturing, or vugs with shattered rocks, due to violent rupturing. This type of formation is known as a dual porosity
reservoir. It is apparent that no matter how many wells are available, a detailed description of fracture geometry is
impossible. To evaluate such heterogeneous formations, it is essential to prepare a descriptive mathematical model which
includes a simplified or an idealized fracture system. However, since the usefulness of such a model depends on how
accurately the simplified fracture system resembles the actual heterogeneous system, any effort to narrow down those
assumptions is strikingly valuable.

Fluid flow in fractured reservoirs behaves differently from that in homogeneous ones. In homogeneous single porosity
reservoirs there is one single flow regime, whereas in NFRs there are two: fractures and matrix. Therefore, characterizing
a heterogeneous formation with two distinct media, matrix and fractures, requires more caution and a more sophisticated
mathematical model, such as dual porosity solution.
2 SPE 144560

The diagnosis of the difference between homogeneous and heterogeneous reservoirs is very important when interpreting
well test data of fractured reservoirs. When we load a fractured medium by production, early drawdown data will show
the fracture characteristics, and late data will reflect the combined fracture and matrix behavior of the system. Yet, the
data at high values of producing time reflects the impact of pore pressure, and thus effective stress change on the
behavior of fractured reservoirs.

Beginning in the 1960s, some research on dual porosity systems was conducted on uniform NFRs [1-3]. Much of the
studies from that era are still in use since they are fundamental. In dealing with NFRs, reservoir engineers prefer using
the well-known dual porosity model introduced by Warren and Root [2]. The concept of fractured reservoir modeling was
later giving more attention to model the fluid flow in NFRs [4-7].

Abdassah and Ershaghi [8] developed a triple-porosity model to analyze the well tests that shows anomalous slope change
during the transition period and where the behavior could not be explained by dual porosity models. Chen and Bai [9]
studied the stress-dependent permeability for anisotropic fractured rocks. They modeled the change in fracture
permeability as a function of the perturbation of either pore pressure or matrix skeleton stress under isothermal
conditions.

Chin et al. [10] presented a general numerical procedure for solving the coupled equations that govern isothermal single-
phase fluid flow in a deformable conventional reservoir. Using the proposed procedure, they assessed the effect of
permeability change induced by reservoir compaction on pressure buildup curves. Gutierrez et al. [11] extended the Biots
[12]
equations for multiphase fluid flow in deformable conventional porous media. According to this work, rock
deformation and fluid flow are fully coupled processes that should be accounted for simultaneously, and can only be
decoupled for predefined simple loading conditions. Work by Raghavan and Chin [13] presents correlations to evaluate
productivity losses in conventional reservoirs with stress-dependent permeability. For stress-sensitive conventional
reservoirs, they concluded that the change in permeability, caused by any change in effective stress during production,
may significantly reduce expected productivity.

Recently, there has been a growing interest on the importance of geomechanics in reservoir simulation and
characterization, particularly for naturally fractured reservoirs [14-17]. Oluyemi and Ola [18] used the Forchheimer [19] model
for turbulent flow coupled with a stress model to obtain a model for flow partitioning, from which they obtained an
enhanced understanding of flow partitioning and the possible effects of stress regime on the behavior of NFRs.

Several approaches have been proposed to implement geomechanical effects into reservoir simulation and
characterization. However, most proposed models are based on numerical analysis, and exact analytical solutions for
nonlinear geomechanics, fluid-flow problems are in general not available. It is the objective of this study to develop a
robust mathematical model, based on the dual porosity model of Warren and Root [2], for investigating the unsteady-state
behavior of naturally fractured reservoirs with the inclusion of changes in effective stress of the formation.

Conceptual Flow Model

To develop a cogent model for a naturally fractured reservoir, it is essential to introduce a model which reflects the
response of a real reservoir. The model to be employed in this study is based on a mass balance model, assuming Darcys
flow in the fractures. The general assumptions in the modeling are the same as those in Warren and Roots model [2]:

a) The material containing the primary porosity is homogeneous and isotropic and is contained within a systematic
array of identical, rectangular parallelepipeds (matrix).

b) All of the secondary porosity is contained within an orthogonal system of continuous, uniform fractures which are
oriented so that each fracture is parallel to one of the principal axes of permeability (fractures).

c) The complex of primary and secondary porosities (matrix plus fractures) is homogeneous. Flow to the wellbore can
only occur from the fracture network; hence, flow through the primary-porosity (from matrix to wellbore) cannot
occur.

The idealized model reservoir is shown in Fig.1.


SPE 144560 3

[2]
Fig.1. Idealization of the heterogeneous system (After Warren and Root ).

For a uniform reservoir which is homogeneous and anisotropic, and for which the connate-water saturation in the
secondary porosity is negligible, the single phase flow of a slightly compressible liquid can be described by Greens
theorem to obtain the continuity equation and to derive the following (see Appendix A):

1 r kf r p p
.( p2 ) = m ctm 1 + f ctf 2 (1)
t t

The matrix-fracture interaction is described by a quasi-steady state pressure relation as [2]:


p1 k m
m ctm = ( p2 p1 ) (2)
t
where , shape factor, has the dimensions of reciprocal area which reflects the geometry of the matrix elements, and
controls the flow between matrix and fractures. The initial conditions and boundary conditions, in field units, are:

I .C. : (t = 0) p1 = p2 = pi

k f h p2
q= (3)
B.C .s : (t 0) 141.2 B ln r r = r
w

p2 ( r , t ) = pi

In fact, three different types of flow mechanics can be distinguished: a) transient or infinite acting behavior, b) steady-
state with outer boundary pressure, and c) pseudo-steady-state denoting a no-flow boundary condition. However, infinite
acting behavior has been employed in this study (conditions in (3)).

Considering the change in fracture permeability due to change in effective stress, we derived a differential system which
reflects the effect of such a change on the behavior of NFRs. In Eqs.4 and 5 we define a new term which represents this
phenomenon. Therefore, the principle differential system (Eqs.1 and 2) turns out to be as follows (see Appendix B).

r2 f 3 (1 2 ) p 2 p 2
2 2
p1 p2
p 2 + co + 3 1 + c f + + = m ctm + f ctf
9 E f x y k f t t
(4)
p1 k m
m ctm t = ( p 2 p1 )

The transformed equations in polar coordinate, initial condition and boundary conditions can be rewritten as follows (see
Appendix B):

2 2 1 2 2
2
1 2
2 + = (1 ) +

(5)
(1 ) 1 =
( 2 1)

4 SPE 144560

I .C : ( = 0) 1 = 2 = 0

2
B.Cs : ( 0) =1 = 1 (6)

2 ( , ) = 0
where the dimensionless groups, in field units, are defined as:
kf h
( , , ) = ( pi p ( , , ) ) (7)
141.2q B
x +y
2 2

= (8)
rw

= tan
1 y
(9)
x
4
2.64 10 k f t
= (10)
rw (m ctm + f ctf )
2

f ctf
= (11)
m ctm + f ctf
k m rw
2

= (12)
kf

141.2q B f 3 (1 2 )
= co + 3 1 + cf + (13)
kf h 9 E f
Two of the dimensionless groups, and , are very significant; they are the groups by which we can describe the
deviation of the behavior of an NFR with dual porosity from that of a homogenous porous medium [2].

(fracture storage capacity) is a measure of the fluid stored in fractures as compared with the total fluid present in the
reservoir, and it has a value between zero and unity. (inter-porosity flow coefficient) is a measure of the heterogeneity
scale of the system and quantifies the fluid transfer capacity from matrix to a fracture network. A third one, , reflects
the effect of geomechanics and the decline in fracture aperture on the behavior of NFRs as time progresses. This
parameter depends on the geomechanical properties, fluid compressibility, and fracture properties (Eq.13).

The differential system in Eq.5 is nonlinear and can be solved by perturbation method [20-22]. Using the Laplace
transformation and solving the nonlinear system subject to the corresponding boundary conditions (6), we obtained the
solution as follows (see Appendix C):

K 0 ( )

Ko ( sf ( s ) )
d

2 (1, ) = l
1
l
1
sf ( s )
(14)
s sf ( s ) K 1 ( sf ( s ) )
s sf ( s ) K 1 ( sf ( s ) )


where s is the Laplace transformation variable, K o ( sf ( s ) ) and K ( 1
sf ( s ) ) are modified Bessel functions of the
second kind, and of zero and first order, respectively, and f ( s ) is:
(1 ) s +
f ( s) = (15)
(1 ) s +

The asymptotic solution, which is valid for sf ( s ) 0.01 ( 100 , or, 100 if 1 ) [2]
, is obtained by the following
[23]
substitutions :
SPE 144560 5

sf ( s )
Ko ( )
sf ( s ) ln
2
(16)
K1 ( sf ( s ) ) 1
sf ( s )

where 0.57721 is Euler-Mascheroni constant (also called Eulers constant). Extending the formula of the indefinite
integral for the modified Bessel functions [24], we derived the following form for the small values of argument, i.e.
sf ( s ) 0.01 , by using the regression technique:


K 0 ( ) sf ( s )
1 sf ( s ) 2 2
d ln + ln + +
2
(17)
sf ( s )
2 2 2 2 24
Inverting the Laplace transformation in Eq.14 while also considering formation damage, gives us Eq.18 for drawdown
test interpretation. Note that formation damage (skin) causes an increase in the resistance to flow near the wellbore, and
can be caused by a variety of reasons, such as plugging or partial completion.

2 (1, ) =
1 + 0.75 ln + 0.80908 + Ei Ei + 2S

2 (1 ) 1 1 + 0.75
* (18)
2 9 2 ( x )
+ dx
24 32 0+ 4 x
where S is skin factor, and ( x ) is the elasticity function which is defined as:

x x
( x ) = ln( x ) + Ei Ei
(1 ) 1
2 3 x x Ei x
x

+ e 1 ln E1

(19)
256(1 ) 1
x
(1 ) x x x
+e (1 )
+ E1 E1
ln (1 )
Ei and E1 are Exponential Integral functions, defined as:
u
e
Ei ( x ) = du (x 0) (20)
x
u
e 1
x u

E1 ( x ) = + ln( x ) + du ( x 0)
0
u (21)
= Ei ( x ) i

and the constant is:



= 4e2 (22)

Pressure Buildup Behavior

In the oil industry, most of the information from a well test comes from interpreting the pressure buildups. Interpreting a
drawdown test is limited by the flow rate fluctuations inherent to production. Such fluctuations cause pressure variations
that are greater near the end of the test than that at the beginning. The zero flow rate which corresponds to pressure
buildup does not have this problem.

*
The lower limit of integration in the integral of the elasticity function is equal to the time after which the solution is valid, depending upon the values of and
(Warren and Roots conditions [2]).
6 SPE 144560

The behavior of a shut-in well following a constant flow rate is considered. Buildup behavior can be described by
superimposing the drawdown equation (Eq.18). If producing time, t p , is fairly long and if certain conditions hold , then
buildup pressure is obtained as Eq.23, and is depicted in Fig.2 (Horners plot).

q B (1 + 0.75 )
p +
t p + t Ei 2 ( x)
p ws = pi ln + Ei
1



(1 ) 1 + 0.75
dx (23)
4 k f h t
4x
where p ws is shut-in pressure, t is the time since production stopped (shut-in time), and is the dimensionless shut-
in time.

Fig.2. Interpretation of buildup data.

For relatively small times the early straight line in the semi-log coordinates acts as though the reservoir is homogeneous.
This behavior, in field units, is described by the following relationship:

162.6q B (1 + 0.75 )
p
t p + t ( x)
p ws = pi t 1.151(1 + 0.75 ) + 4 x dx
log (24)
kfh 0
and for high values of shut-in time, i.e. the late straight line in the semi-log coordinates, Eq.23 turns into:

162.6 q B (1 + 0.75 ) t p + t
p ws = pi log (25)
kfh t
Hence, the fracture permeability ( k f ) can be determined from combining the slope of the late straight line (Eq.27) with
the elasticity parameter (Eq.13). This, in field units, gives us:

81.3q B 3 (1 2 )
kf = 1 + 1 + 2.61m co + 3 1 + f cf + (26)
mh 9 E f

where m defined as:


p 3 / and 100 if 1 , or , 100 1 / if 1 [2]
SPE 144560 7

162.6 q B (1 + 0.75 )
m= (27)
kf h
is the slope of the late semi-log straight line, in psi/cycle, to be obtained from the buildup data (Fig.2). Substituting the
calculated permeability (Eq.26) into Eq.27, and solving for the elasticity parameter, , gives us:

1 mk f h
= 1 (28)
0.75 162.6 q B
Nonetheless, we will obtain a higher value for initial pressure if we extrapolate the early portion of a buildup curve
to (t p + t ) / t = 1 . Such a difference can be readily related to the fracture storage capacity, , by combining Eqs.24 and
25. This yields:
p
( x)
+ dx

m 1.151 (1+ 0.75 ) + 4 x (29)

=10 0

where is the vertical displacement between the early and late semi-log straight lines for a buildup curve, when both are
extrapolated to (t p + t ) / t = 1 (Fig.2). Note, also, that since the elasticity may have effects on a buildup curve, is not
computed at any earlier time. Otherwise, we would obtain an erroneous value for . Also, the skin factor, S , can be
obtained from the superposition concept as follows:

p1hr p wf kf
S = 1.151(1 + 0.75 ) log + 3.227
rw ( m ctm + f ctf )
2
m
(30)
9
p
( x)
2 2

+ + dx
24 32 0
+ 4x
It is apparent, from Eqs.29 and 30, that producing time can have significant effect on buildup-test curves from stress-
sensitive NFRs, i.e. on the fracture storage capacity, fracture permeability, and the degree of damage. This is in
accordance with the work by Chin et al. [10], where by interpreting data obtained from simulation runs concluded that the
effect of producing time on the buildup-test curves can be significant.

Fracture Porosity Estimation by Well Test

Although there are methods for determining the secondary porosity of a reservoir, such methods are prone to error since
fracture porosity is very scale-dependent. One common way to estimate the fracture porosity is using well logging. The
porosity computed from neutron log represents the combination of both matrix and fracture porosities while sonic log
only measures the matrix porosity. Therefore, the fracture porosity can be obtained by subtracting the sonic porosity from
the neutron porosity. But, the radius of investigation in well logging is limited to a few feet around the wellbore.
Therefore, the value obtained from well logs does not represent an average value for the reservoir sector under study,
despite all efforts to make it accurate.

Moreover, when interpreting buildup data, by using Eqs.26-30, the value of the secondary porosity has such a major
impact on the properties of an NFR. The value of f can be 100% in a particular location of reservoir, but the value for
the whole reservoir is generally less than 1% [25]. According to Nelson [26], fracture porosity is always less than 2%; in
most reservoirs it is less than 1%, with a general value of less than 0.5%. An exception to this rule of thumb is vuggy
fractures where porosity can vary from 0 to a very large value.

Because of such an uncertainty for fracture porosity, a trial-and-error method is introduced here for reaching satisfactory
results from well test analysis. The method is to try out various values of the secondary porosity until error is sufficiently
reduced or eliminated. This process can use the value obtained from well logging, if available, as the first estimate. Using
well test data through an iterative procedure, we can achieve a much more accurate fracture porosity. This fracture
porosity can be considered as an average value for the part of the reservoir involved in the compressible zone, in which
pressure has been affected during the flow period (radius of investigation).

We can use the above rule of thumb for the first guess to input to the iterative algorithm (Fig.3) if other estimation is not
available. Also, we assume that is not as sensitive as fracture properties to elasticity, and it can be obtained from the
proper analysis of conventional type-curve match [27,28]. The flowchart in Fig.3 demonstrates the steps of this trial-and-
error algorithm.
8 SPE 144560

Fig.3. Algorithm to determine the average values of NFR properties.

Results and Discussion

Based on the solutions obtained in this study, we discern the need for a three-parameter model with which to describe
naturally fractured reservoirs. Comparisons of the new model to both the asymptotic homogeneous solution and that of
Warren and Root [2] showed that the new model gives more precise results. Also, the importance of the third parameter
(elasticity parameter) was investigated graphically and found to be significant, albeit at high values of producing time.

When accepting the derived equations, it is necessary to show that the limiting behavior is physically precise, i.e. when
is 0 or 1, or is infinity. As the fracture storage capacity ( ) tends to unity, or as the interporosity flow coefficient
( ) tends to infinity, both the Warren and Roots [2] solution and the one proposed in this study approach the asymptotic
solution at early times. This is a proper result since the primary porosity must vanish if we have the above conditions
for and (Fig.4). However, due to the elasticity of matrix itself, the new solution and that of Warren and Root [2]
demonstrate increasing differences at later times. Under this condition, the dimensionless pressure drop can be obtained
as (see Appendix D):

( 1 + 0.75
) ln + 0.80908 + 2 S 9
( )

2 2

2 (1, ) = + ln . ln(100 )
2

2 8 3
1 + 0.75 4 100 (31)
SPE 144560 9

Fig.4. Comparison of the new solution with Warren & Roots solution ( = 1 , S = 0 ) (drawdown).

Yet, we observe the effect of matrix elasticity at late times. This means that the combined effect of production rate and
geomechanical properties still dominate the behavior of the stress-sensitive NFRs, especially at high values of time.
Note, also, that since matrix elasticity causes the fracture aperture to diminish as time goes on, the pressure drop in the
fracture network, depending upon elasticity parameter, would be mitigated. Because of this, the new solution results in
lower values of pressure drop than conventional solutions do.

On the other hand, if approaches zero (negligible storage capacity in the secondary porosity), the results of such a
special case are shown in Fig.5 and Eq.32 (see Appendix D).

2 (1, ) = ( 1 + 0.75
2
) ln + 0.80908 Ei ( ) +
2S
1 + 0.75

x
2 3


ln( x ) Ei ( x )
ln E1 ( x )
(32)
9 256
2 2


4 6
+ + x
dx
8 0
+ x xe

Fig.5. Comparison of the new solution with Warren & Roots solution ( = 0 , S = 0 ) (drawdown).
10 SPE 144560

From this figure it is apparent that the fracture pressure drop ( 2 (1, ) ) increases when production begins, and it
asymptotically approaches the behavior of a homogeneous reservoir whose is quite large.

In Fig.6 different patterns associated with finite values of and are shown. In the study by Warren and Root [2], it
was shown that the two linear pieces of the curves are parallel to the asymptote but vertically displaced by an amount of
log(1 / ) . However, according to the new solution presented in Eq.18 and Fig.6, we discern two facts: (1) the difference
is not constant, but a function of time ( ), fracture storage capacity ( ), interporosity flow coefficient ( ), and
elasticity parameter ( ); and (2) the so-called straight lines are not exactly lines, but curves with very small curvatures.
From this and for the stress-sensitive NFRs, it goes without saying that the early portions and the late portions of the
curves would not necessarily be parallel to the asymptote.

Fig.6. New solutions obtained by asymptotic inversion ( = 0.01 ) (drawdown).

The transitional curves, which connect the two linear portions, represent the interaction between and , which is the
communication between the primary and secondary porosities. The most notable feature of Fig.6 is that as time goes on,
the solution will not approach that of the homogeneous asymptotic solution. This means that the influence of elasticity of
the matrix blocks would be significant at later times rather than early times. Fig.6 clearly shows that the effect of stress
sensitivity increases with increased producing time. Similar conclusion can be drawn when interpreting a buildup test
from a stress-sensitive NFR (Eqs.26-30). Although Fig.6 is in dimensionless form, a field data plot of pressure drop
versus log time would have similar appearance. The two straight lines can be either parallel or unparallel depending upon
matrix block elasticity as well as time.

The Warren & Roots [2] solution is valid if we assume that there is no change in fracture conductivity with respect to
change in pore pressure and effective stress. However, change in pressure due to production or injection will affect the
effective stress, and this change will impact fracture permeability, consequently.

In Fig.7 the behavior patterns for the different values of with = 0.1 are shown. It is obvious that when varies,
the behavior is dependent on the elasticity of formation. The behavior of an NFR with a higher value of elasticity
parameter shows more deflection from that of a homogeneous, single porosity reservoir.
SPE 144560 11

Fig.7. New solutions for different values of ( = 0.1 ) (drawdown).

Figure 8 is a semi-log graph of a constant rate well in an infinite acting reservoir with varying degrees of damage or
stimulation (negative skin). It is apparent from this figure that type curve matching damaged well data to estimate skin
factor will not yield an accurate result, due to the flatness of the curves. This parameter can be obtained from interpreting
buildup data (Eq.30). Also, it is clear that the degree of skin displaces the curves vertically downward (stimulated well)
or vertically upward (damaged well).

Fig.8. New solutions for different values of skin factor ( = 5 10 ).


3

Example

To illustrate the application of the ideas introduced here, a buildup example presented by Najurieta [29] is discussed. This
is a suitable example since the buildup test is run after a very long producing time (8610 hours). Hence, we are more
likely to observe the effect of elasticity in such a test. The reservoir and pressure buildup data of this test are shown in
Table 1:
12 SPE 144560

Table 1 Reservoir and pressure buildup data [29]

The log-log graph of p versus t (Fig.9) shows that the pressure data is not obviously influenced by wellbore storage
effects.

[29]
Fig.9. Log-log graph of pressure data for Najurietas example .

The Horner plot in Fig.10 depicts a semi-log graph of the data, from which the slope of the late straight line ( m ) and the
vertical displacement ( ) are obtained as 32 psi/cycle and 70 psi, respectively. It is apparent that the test has not
distinctly shown the early straight line. Hence, this line is obtained from an extrapolation.

Fig.10. Semi-log graph of Najurietas buildup data.


SPE 144560 13

The reservoir data can be calculated by the usual type-curve match methods presented in the literature, (for e.g. refs.27
and 28). Applying the conventional methods to the data, we obtained the following results for the reservoir properties:
k f = 990 md , = 0.0004 , = 2 10 , and S = 3.67 (skin from Najurieta [29]).
6

8
After some trials and iterative calculations using the proposed algorithm (Fig.3), with a relative error of 10 for
convergence, we obtained the reservoir properties for different combinations of elasticity parameters as shown in Table
2. Since a first estimate for the fracture porosity was not available, we used the rule of thumb as the first guess for the
fracture porosity; that is 1% [25,26].

Table 2 Final results from buildup data interpretation using the introduced algorithm

IterationNo.
Runs kf,md f(%) Skin
untillconvergence

E1=4.5e6,1=0.26 980 0.06 0.00211 0.1080 6.6 24


E1=4.5e6,2=0.30 966 0.08 0.00273 0.0749 6.3 18
E1=4.5e6,3=0.35 956 0.10 0.00346 0.0500 6.1 13
E2=7.5e6,1=0.26 955 0.10 0.00354 0.0477 6.1 12
E2=7.5e6,2=0.30 952 0.11 0.00387 0.0395 6.0 11
E2=7.5e6,3=0.35 948 0.12 0.00427 0.0310 5.9 9
E3=9.5e6,1=0.26 951 0.12 0.00396 0.0375 6.0 10
E3=9.5e6,2=0.30 949 0.12 0.00421 0.0322 5.9 9
E3=9.5e6,3=0.35 946 0.13 0.00451 0.0264 5.8 8
[psi]
According to these results (Table 2), the elasticity parameters can affect the behavior of buildup curves and the behavior
of naturally fractured reservoirs. Comparing the values of fracture permeability calculated by using the proposed method
in this work (Table 2) and that obtained from conventional techniques [27,28] ( k f = 990 md ), we discern that the
conventional methods estimate a higher value for fracture permeability and a lower value for skin factor. These different
estimations can be a result of ignoring the geomechanics in the NFR modeling. Last but not least, the comparisons show
that the proposed method gives a value for the fracture storage capacity, , almost ten times higher than that obtained
from the conventional techniques. This discrepancy comes from the inclusion of elasticity in our calculations.

Application of such a new well test interpretation technique in NFR characterization has the potential use of monitoring
both the reservoir properties and their changes with time, which can be called dynamic reservoir characterization. This is
similar to using time-lapse (4-D) multicomponent seismology to image and monitor the reservoir changes with time [30].
Since storage capacity ( ) controls effective compliance in NFRs, and compliance itself is the bridge between seismic
anisotropy and storage capacity, well testing and time-lapse multicomponent seismology are the critical tools to integrate
in dynamic reservoir characterization [31]. Using both methods (well tests and time-lapse multicomponent seismology),
the elastic parameters of an NFR can be determined more completely; independently, they both yield non-unique
solutions.

We have proved that the behavior of naturally fractured reservoirs analyzed by either drawdown or buildup tests is
dependent not only on the fluid properties, but also on the matrix skeleton parameters and on the rate of withdrawal
(Eq.13). Hence, whereas characterizing a fractured reservoir with only two parameters, and , might be sufficient for
many NFRs, it might not be so for stress-sensitive reservoirs, especially those with intense elasticity parameter impact.

Conclusions

The material presented in this study is related to the interaction between fluid flow and rock deformation in both reservoir
simulation and well test interpretation. This type of interaction (coupling) is particularly significant in stress-sensitive
fractured reservoirs, where fracture permeability can be orders of magnitude greater than that of matrix.

For characterizing NFRs, we have extended the sugar-cube model of Warren & Root [2], which was developed for
intermediate naturally fractured porous systems. A new mathematical model for the reservoirs suspected of being
naturally fractured is established from which the following conclusions can be drawn:
14 SPE 144560

Field test transient pressure interpretation confirms the need for a three-parameter dual porosity model for
characterizing stress-sensitive NFRs. The new model presented in this work considers the effects of
geomechanics and rock compaction which account for the fracture permeability change in the modeling process.

For stress-sensitive NFRs, both the drawdown and buildup curves can behave differently from those curves taken
from NFRs with constant permeability. Interpreting such well tests seems to be challenging, requiring a new
interpretation method, as discussed in this work.

A high value of producing time amplifies the severity of reservoir stress sensitivity, which is evident on buildup
responses. Producing time affects buildup curves for stress-sensitive NFRs that exhibit permeability change,
caused by rock compaction.

Comparing the results from the conventional methods and the method proposed in this study, we observed
noticeable difference between them. This validates the effect of geomechanics on buildup data, which is clarified
by the new interpretation technique. In addition, since the new technique takes the effective stress change into
account, it can be considered as a dynamic reservoir charatcterization method, which is a new approach to 4-D
analysis in reservoir engineering.

The iterative algorithm developed in this work is a tool to evaluate the average values of the key parameters for a
naturally fractured reservoir. This method correlates the parameters obtained from well testing to those obtained
from other methods to mitigate the errors in our estimations.

Acknowledgements

The authors would like to thank the following sponsors for financial support: US DOE through contract of DE-FC26-
08NT0005643 (Bakken Geomechanics), North Dakota Industry Commision together with five industrial sponsors
(Denbury Resources Inc., Hess Corporation, Marathon Oil Company, St. Mary Land & Exploration Company, and
Whiting Petroleum Corporation) under contract NDIC-G015-031, and North Dakota Department of Commerce through
UNDs Petroleum Research, Education and Entrepreneurship Center of Excellence (PREEC). We also wish to express
appreciation to Kathleen Vacek for her assistance on preparing this paper.

Nomenclature

B = formation volume factor, Rbbl/STB t = time, T (hr in Eq.10)


-1 2
c1 = total compressibility of primary porosity, psi , (LT /M) tp = producing time, hr (T)
c2 = total compressibility of secondary porosity, psi-1 , (LT2/M) t = shut-in time, hr (T)
cf = fracture compressibility, psi-1 (LT2/M) = shape factor, 1/L2
-1
co = flowing liquid compressibility, psi (LT /M)2 0.57721, Eulers constant
cp = pore compressibility, psi-1 (LT2/M) = interporosity flow coefficient, dimensionless
-1 2
cr = matrix skeleton compressibility, psi (LT /M) = fracture storage capacity, dimensionless
E = Youngs modulus of elasticity, psi (M/LT2) = elasticity parameter, dimensionless
h = thickness of formation, ft (L) (x) = elasticity function, dimensionless
kf = fracture permeability, md (L2) = Poissons ratio, dimensionless
2
km = matrix permeability, md (L ) = radial coordinate, dimensionless
p1 = pressure in primary porosity, psi (M/LT2) = fluid density, M/L3
p2 = pressure in secondary porosity, psi (M/LT2) 1 = pressure change in primary porosity, dimensionless
2
pi = initial pressure, psi (M/LT ) 2 = pressure change in secondary porosity, dimensionless
pws = shut-in pressure, psi (M/LT2) = time, dimensionless
3
q = flow rate, STBD ( L /T) p = producing time, dimensionless
rw = wellbore radius, ft (L) = shut-in time, dimensionless
S = skin factor, dimensionless m = primary porosity, fraction
Swc = connate-water saturation, dimensionless f = fracture porosity, fraction
SPE 144560 15

References
1. Barenblatt, G.I., I.P. Zheltov and I.N. Kochina. 1960. Basic concepts in the theory of seepage of homogeneous liquids in fissured
rocks. PMM, Vol.24, No.5.
2. Warren J. E. and P. J. Root. 1963. The behavior of naturally fractured reservoirs. Soc. Pet. Eng. J. 245-55, Trans. AIME, 228,
Sept.
3. Kazemi, H. 1969. Pressure transient analysis of naturally fractured reservoirs with uniform fracture distribution. Soc. Pet. Eng. J.
451-62, Trans. AIME, 246.
4. Kazemi, H., L.S. Merril, K.L. Porterfield and P.R. Zeman. 1976. Numerical simulation of water-oil flow in naturally fractured
reservoir. SPE 5719, presented at the Fourth Symposium on Numerical Simulation of Reservoir Performance, Los Angeles,
California, Dec.
5. Guilman, J.R. and H. Kazemi. 1983. Improvements in simulation of naturally fractured reservoirs. Soc. Pet. Engr. J.695-707,
Aug.
6. Thomas, J.R., T.N. Dixon and N.G. Pierson. 1983. Fractured reservoir simulation. SPE 9305, presented at the SPE Annual
Technical Conference, Dallas, Texas, Feb.
7. Saidi, A. 1987. Reservoir engineering of fractured reservoirs-fundamental and practical aspects. TOTAL Edition press.
8. Abdassah, D. and I. Ershaghi. 1986. Triple-porosity systems for representing naturally fractured reservoirs. SPE Form. Eval.
April, pp. 113127.
9. Chen, M. and M. Bai. 1998. Modeling stress dependent permeability for anisotropic fractured porous rocks. Int. J. of Mech. Min.
Sci. Vol. 35, No. 8, pp.1113-1119.
10. Chin L.Y., R. Raghavan and L.K. Thomas. 2000. Fully coupled geomechanics and fluid flow analysis of wells with stress-
dependent permeability. SPE 58986. SPE J. (5), March.
11. Gutierrez M., R.W. Lewis and I. Masters. 2001. Petroleum reservoir simulation coupling fluid flow and geomechanics. SPE
72095. SPE Res. Eva. & Eng. June.
12. Biot, M.A. 1941. General theory of three dimensional consolidation. J. Appl. Phys. 12, 155.
13. Raghavan, R. and L.Y. Chin. 2002. Productivity changes in reservoirs with stress-dependent permeability. SPE 77535, presented
at the SPE Annual Technical Conference held in San Antonio, Texas, 29 Sep.-2 Oct.
14. Min, K.B., J. Rutqvist, C. Tsang and L. Jing. 2004. Stress-dependent permeability of fractured rock masses: a numerical study.
Int. J. of Rock Mech. & Min. Sci. 41, 1191-1210.
15. Dautriat, J., N. Gland, E. Rosenberg and S. Bekri. 2007. Stress-dependent permeabilities of sandstones and carbonates. SPE
110455, presented at the SPE Annual Technical Conference held in Anaheim, California, 11-14 Nov.
16. Abbasi, J., M. Sharifzadeh and B. Ferdowsi. 2008. Three dimensional numerical modeling of stress-dependent permeability
considering aperture distribution through fractured rock masses: case study, Roodbar dam formation. ISRM Int. Symp., 5th Asian
Rock Mechanics Symposium (ARMS 5), 24-26 Nov., Tehran, Iran.
17. Tao, Q., C.A. Ehlig-Economides and A. Ghassemi. 2009. Modeling variation of stress and permeability in naturally fractured
reservoirs using displacement discontinuity method. ARMA 09-047, presented at the 43rd US rock Mech. Symp. held in Ashville,
NC, June 28th-July 1st.
18. Oluyemi, G.F., and O. Ola. 2010. Mathematical modeling of the effects of in-situ stress regime on matrix-fracture flow
partitioning in fractured reservoirs. SPE 136975, presented at the SPE Annual Technical Conference, Tinapa-Calaber, Nigeria,
31 July-7, Aug.
19. Forchheimer, P. Wasserbewegung Durch Den Boden. Z. Ver Deutsch. 1901. Ing. 45, pp.1782-1788.
20. He, J.H. 1999. Homotopy perturbation technique. Comp. Math. Appl. Mech. Eng. 178(3-4) (1999), 257262.
21. He, J.H. 2000. A coupling method of homotopy technique and perturbation technique for nonlinear problems. Int. J. of Non-linear
Mech. 35(1), 3743.
22. Korenev, B.G. 2002. Bessel functions and their applications. Chapman & Hall/CRC, 2002.
23. Van Everdingen, A.F. and W. Hurst. 1949. The application of the Laplace transforms to flow problems in reservoirs. Trans.,
AIME , 186, 305-324B.
24. Abramowitz M. and I.A. Stegun. 1964. Handbook of mathematical functions with formulas, graphs, and mathematical tables.
Chap. 11. Integrals of Bessel functions, edited by Yudell L.Luke, 11.1.22, p. 421.
25. Garcia, L.E.P. 2005. Integration of well test analysis into a naturally fractured reservoir simulation. MS Thesis, Texas A&M
University.
26. Nelson, R.A. 2001. Geologic analysis of naturally fractured reservoirs. Gulf Professional Publishing, Woburn, MA.
27. Stewart, G. and F. Ascharsobbi. 1988. Well test interpretation for naturally fractured reservoirs. SPE 18173, presented at the SPE
Annual Technical Conference, Houston, TX, 2-5 Oct.
28. Onur, M., A. Satman and A.C. Reynolds. 1993. New type curves for analyzing the transition time data from naturally fractured
reservoirs. SPE 25873, presented at the SPE Rocky Mountain Regional/Low Permeability Reservoirs Symposium, Denver, Co,
12-14 April.
29. Najurieta, H.L. 1980. A theory for pressure transient analysis in naturally fractured reservoirs. J. Pet. Tech., 1241- 1250, Trans.,
AIME, 269.
30. Davis, T.L. 2006. Multicomponent 4-D seismic reservoir characterization of tight gas sands, Rulison Field, Colorado. SEG
Annual Meeting Expanded Abstracts.
31. Cardona, R. 2002. Fluid substitution theories and multicomponent seismic characterization of fractured reservoirs. PhD Thesis,
Colorado School of Mines.
32. Bear, J. 1972. Dynamics of fluids in porous media. American Elsevier Publishing Co., New York.
33. Aguilera, R. 1999. Recovery factors and reserves in naturally fractured reservoirs. J. of Canadian Pet. Tech., Distinguished
Authors Series, July, volume 38, no. 7, p. 15-18.
34. Jabbari, H., Z. Zeng and M. Ostadhassan. 2011. Impact of in-situ stress change on fracture conductivity in naturally fractured
reservoirs. ARMA 11-239, accepted for presentation at the US Rock Mechanics/Geomechanics Symposium held in San
Francisco, CA, June 2629.
16 SPE 144560

Appendix A

Considering the continuity equation with isotropy in fracture conductivity [2], namely k fx = k fy = k fz , we have:
r kf r ( )1 ( ) 2
.( p2 ) = + (A.1)
t t
where the subscripts 1 and 2 denote primary and secondary regions, respectively. Considering the change in control
volume due to the change in effective stress (deformable CV), we have the following relations for the rate of change of
mass in matrix [32]:
1 1 Vb 1 1
Vs = Vb (1 mat ) = Vb 1 = const = (A.2)
1 2 Vb t (1 2 )(1 S wc ) m t
1 1 2 Vb e p
= (1 S wc ) m = cr (1 2 )(1 S wc ) m 1 (A.3)
t Vb e t t
where Vs is solid phase volume, Vb is bulk volume of matrix, mat is matrix porosity, e ( = t p1 ) is effective stress,
and cr is compressibility of matrix skeleton, which is defined as:
1 Vb
cr = (A.4)
Vb e
Assuming that there is interaction between the two regions (unlike Warren and Root [2]), and considering the definition of
total compressibility of the two regions, we obtain:
1 r kf r p p2
.( p2 ) = S m 1 + S f (A.5)
t t
where S m and S f , the matrix and fracture storage, are defined as:
S m = c11 + cr (1 2 )(1 S wc ) m
(A.6)
S f = (c2 + c f ) 2
When assuming that the pressure change in one region is independent of that in the other one, the total compressibilities
of the primary and secondary regions ( c1 and c2 ), are approximately given as [2]:
c p + S wc c w
c1 co +
1 S wc (A.7)
c 2 co
where co is oil compressibility, c w is water compressibility, S wc is connate-water saturation, and c p , pore
compressibility in matrix, is given as:
1 Vp
cp = (A.8)
V p p1
Fracture compressibility ( c f ) is an elusive parameter. It is recommended that we determine this parameter in the
laboratory using rocks (cores) from the reservoir. If these are not available (and they are not available most of the time)
we have to resort to empirical correlations [33]. However, a reasonable estimate of fracture compressibility can be
obtained if the connate-water saturation in the secondary porosity is assumed to be negligible [2]. Therefore,
1
cf = co (A.9)
p2
Thus,
ctm m = c11 + cr (1 2 )(1 S wc ) m
(A.10)
ctf f = 2co 2
Finally,
1 r kf r p p2
.( p 2 ) = ctm m 1 + ctf f (A.11)
t t
where ctm m and ctf f are matrix and fracture storage, respectively.
SPE 144560 17

Appendix B

Eq.1 gives the following form upon taking the derivatives:


1 r kf r kf r 2 1 r r kf r r
.( p2 ) = p 2 + k f . p 2 + . p 2 (B.1)

To determine the gradient terms we consider planar variation of functions for which we obtain:
p p p 2

k f p
2 2

kf r kf x x x kf p
2

p
2

. x
r
. p 2 = . = 2
= x
2
+
2
(B.2)
p p2 p p2 y
y
2 2

y y p
2
y
and,
k p k p2 p

f

1 p
f 2 2

1 x x x 1 k f p p
2 2
1 r
. x
r
k f . p 2 = . = = x +
2 2 2
(B.3)
k p k p2 p p2 y
y
f 2 2

f

y y p
2
y
Substituting the change in oil density into Eq.B.2 yields:
kf r k f co p p
2 2
r
. p 2 =
2
+
2
(B.4)
x y
Hence,
kf r 2 1 k f p2 p2
2 2
1 r kf r
.( p2 ) = p 2 + co + + (B.5)
k f p2 x y
The change in fracture permeability, as a function of fracture pressure, is obtained as [34]:
1 dk f f 3 (1 2 )
= 3 1 + c + (B.6)
9
f
k f dp2 E f

Substituting the above relationships into Eq.B.1, gives us the following nonlinear system:
r c 3 (1 2 ) p 2
2

p =
2
p1 p2
p +c + 3 1 + + + ctmm + ctf f
2 f 2

9 y
2 o f

E f x kf t t
(B.7)
c p1 km
tm m = (p p1 )
t
2

To transform the equations above to act in polar coordinate we use the dimensionless groups presented in Eqs.7-13. The
following partial differential terms can be obtained by transformation of variables in polar coordinates:

p2 2 sin 2
x = r cos . .


w
(B.8)
p2 = sin . 2 + cos . 2
y r w

p2
2
2 sin 2
2 2 2
sin 2 2 sin 2 2 sin 2
2 2

x = r cos . + . + +
2
. . .

2 2 2 2 2 2


w
(B.9)
p2 = sin . 2 + cos . 2 sin 2 2 sin 2 2 cos 2
2 2 2 2 2 2

+ +
2

y . . .
r
2 2 2 2 2 2

p1 1
= .
t
(B.10)
p2 2
= .
t
where and are defined as:
18 SPE 144560

q B
= 2 k h
f
(B.11)
= r ( c
2


w
tm
m + ctf f )
kf

Substituting the terms above into Eq.B.7 and considering symmetry in angular direction, we obtain a simpler nonlinear
differential system in polar coordinates:

2 2 1 2 2
2
1 2
2 + = (1 ) +

(B.12)
(1 ) 1 =
( 2 1)

Appendix C

Applying the perturbation theory in this case, we have the following for the approximation to the full solution, a series in
the small parameter, :
1 = 10 + 11 + 12 + ....
2

(C.1)
2 = 20 + 21 + 22 + ....
2

where 20 is the known solution to the exactly solvable initial problem (solution to the Warren and Roots model [2]) and
21 , 22 represent the higher-order terms which can be found iteratively by some systematic procedure. For small
these higher order terms in the series become successively smaller. Substituting the terms in C.1 for 2 (fractures) and
1 (matrix) into Eq.5 (or B.12) yields:
+ 1
(
+ 22 + ( ) 20 + 21 + 22 20 + 21 + 22
2 2 2
) ( )
2
=
20 21


(C.2)
( ) (
(1 ) 10 + 11 + 12 + 20 + 21 + 22
2 2
)
(1 ) ( 10
+ 11 + 12
2
) = ( 20
10 ) + ( 20 10 ) + ( 20 10 )
2

Collecting coefficients of like powers of , an approximate "perturbation solution" can be obtained by truncating the
series, namely by keeping only the first two terms. Hence, the following systems are readily obtained:
For :
0

1
2

+ = (1 ) +
20 20 10 20

(C.3)
(1 ) = ( )
10


20 10

I .C : ( = 0) = ( = 0) = 0
20 10


B.Cs : = 1 (C.4)
20
=1,


( , ) = 0
20

This system was solved by Warren and Root [2] and the solution is:

20 ( , ) = ln + 0.80908 + Ei
1
2

Ei
(1 )
2 ( )

1
(C.5)

For :
1

1
2
11 21
2

+
21 21 20
= (1 ) +

2

(C.6)
(1 ) = ( )
11


21 11
SPE 144560 19

I .C : 21 ( = 0) = 11 ( = 0) = 0
21
B.Cs : =1,
=0 (C.7)

21 ( , ) = 0
20
Plugging the term , calculated from C.5, into C.6, gives us:

1 21
2
1
+ = (1 ) +
21 21 11


2 2

(C.8)
(1 ) = ( )
11


21 11

with conditions in C.7 and using the Laplace transformation, the desired solution is obtained as:
F1 1 F1
2
1
+ sf ( s ) F1 = (C.9)
s
2 2

I .C : F1 ( , 0 ) = 0
F1
B.Cs : ( = 1, s ) = 0 (C.10)

F1 ( , s ) = 0
C.9, with conditions in C.10, is an inhomogeneous Bessel equation of the second kind, and the solution at the wellbore,
i.e. = 1 , is obtained as:
K 0 ( )


d
(C.11)
sf ( s )
F1 (1, s ) =
s sf ( s ) K1 ( sf ( s ) )
Hence,

K ( )
0 d
2 (1, ) 20 (1, ) l 1
sf ( s )
(C.12)
s sf ( s ) K1 ( sf ( s ) )


Finally, according to the relationships in Eqs.16 and 17, and inverting the Laplace transformation of C.12, we derived Eq.18.
Appendix D
If 1 :
( x ) = ln( x ) (D.1)
From Warren and Roots solution [2], we have:
i = 100 = 100 (D.2)
This gives:

( )

( x) ln( x )

1 1 1
dx = dx = ln ( x )
2


= ( ln ( ) ln (100 ) ) =
2 2
ln (
. ln 100
2
) (D.3)
i 4x 100
4x 8 100 8 8 100

If 0 :

x
2 3

( x ) = ln( x ) Ei ( x ) + e E1 ( x ) (D.4)
x

ln
256
This yields:

x
2 3

ln( x ) Ei ( x )
E1 ( x ) ln
( x) 256

(D.5)
dx = + x
dx
0
+ 4x 0
+ 4x 4 xe

Вам также может понравиться