Matrix-Fracture Transfer Shape Factor For Modeling Ow of A Compressible Uid in Dual-Porosity Media

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

Advances in Water Resources 34 (2011) 627–639

Contents lists available at ScienceDirect

Advances in Water Resources


journal homepage: www.elsevier.com/locate/advwatres

Matrix–fracture transfer shape factor for modeling flow of a compressible fluid


in dual-porosity media
Ehsan Ranjbar, Hassan Hassanzadeh ⇑
Department of Chemical and Petroleum Engineering, Schulich School of Engineering, University of Calgary, 2500 University Drive N.W., Calgary, Alberta, Canada T2N 1N4

a r t i c l e i n f o a b s t r a c t

Article history: The matrix–fracture transfer shape factor is one of the important parameters in the modeling of fluid flow
Received 21 September 2010 in fractured porous media using a dual-porosity concept. Warren and Root [36] introduced the dual-
Received in revised form 17 February 2011 porosity concept and suggested a relation for the shape factor. There is no general relationship for deter-
Accepted 22 February 2011
mining the shape factor for a single-phase flow of slightly compressible fluids. Therefore, different studies
Available online 17 March 2011
reported different values for this parameter, as an input into the flow models. Several investigations have
been reported on the shape factor for slightly compressible fluids. However, the case of compressible flu-
Keywords:
ids has not been investigated in the past. The focus of this study is, therefore, to find the shape factor for
Shape factor
Fractured porous media
the single-phase flow of compressible fluids (gases) in fractured porous media. In this study, a model for
Matrix–fracture transfer function the determination of the shape factor for compressible fluids is presented; and, the solution of nonlinear
Compressible fluids gas diffusivity equation is used to derive the shape factor. The integral method and the method of
Single-phase gas flow moments are used to solve the nonlinear governing equation by considering the pressure dependency
of the viscosity and isothermal compressibility of the fluid. The approximate semi-analytical model for
the shape factor presented in this study is verified using single-porosity, fine-grid, numerical simulations.
The dependency of the shape factor on the gas specific gravity, pressure and temperature are also inves-
tigated. The theoretical analysis presented improves our understanding of fluid flow in fractured porous
media. In addition, the developed matrix–fracture transfer shape factor can be used as an input for mod-
eling flow of compressible fluids in dual-porosity systems, such as naturally fractured gas reservoirs, coal-
bed methane reservoirs and fractured tight gas reservoirs.
Ó 2011 Elsevier Ltd. All rights reserved.

1. Introduction interaction and govern the mass transfer between the matrix
blocks and the fractures [4,31]. The rate of mass transferred from
A naturally fractured reservoir (NFR) is defined as a reservoir the matrix to the fracture is directly proportional to the shape fac-
that contains fractures created by natural processes, such as dias- tor. For modeling of naturally fractured reservoirs, an accurate va-
trophism or volume shrinkage. These natural fractures can have a lue of the shape factor is required to account for both the transient
positive or negative effect on fluid flow [2,26]. NFRs are usually and pseudo steady-state behavior of the matrix–fracture interac-
thought to consist of an interconnected fracture network, which tion and also the geometry of the matrix–fracture system.
provides the main flow paths (fractures have high permeability In a pseudo steady-state, the rate of mass transfer per unit bulk
and low storage volume), and the reservoir rock or matrix, which volume is proportional to the difference between the fracture pres-
acts as the main source of fluid storage (matrix blocks have low sure and the average matrix block pressure, as given by:
permeability and high storage volume) [5]. In such a system, in
km q
addition to the intrinsic properties of the matrix and fracture, the q¼ rðpm  pf Þ ð1Þ
interaction between the matrix and fractures should be modeled
l
accurately. where q is the rate of the fluid exchange between the matrix and the
Currently, the dual-porosity approach is one of the computa- fracture, km is the matrix permeability, l is the fluid viscosity, q is
tionally efficient and widely used methods to model fluid flow in the fluid density, pm represents the average matrix block pressure,
fractured reservoirs. In this approach, the matrix and fracture are pf is the fracture pressure, and r is called the matrix–fracture trans-
separated into two different media, each with its own properties. fer shape factor and has dimensions of L2. The matrix–fracture
A transfer function has been used to represent the matrix–fracture transfer term, q is related to the matrix pressure by:

⇑ Corresponding author. Tel.: +1 403 210 6645; fax: +1 403 284 4852. @pm
q ¼ /m cm q ð2Þ
E-mail address: hhassanz@ucalgary.ca (H. Hassanzadeh). @t

0309-1708/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.advwatres.2011.02.012
628 E. Ranjbar, H. Hassanzadeh / Advances in Water Resources 34 (2011) 627–639

Nomenclature

A cross-sectional area [L2] Greek letters


A(tD) first coefficient of the trial solution b correction factor
B(tD) second coefficient of the trial solution c gas specific gravity
Bg gas formation volume factor d penetration depth
C(tD) third coefficient of the trial solution gm matrix hydraulic diffusivity [L2/T]
cm matrix total compressibility [LT2/M] g average hydraulic diffusivity [L2/T]
D(tD) fourth coefficient of the trial solution gD dimensionless hydraulic diffusivity
hm = 2Lc matrix block length [L] gD1 dimensionless fracture hydraulic diffusivity
k dimensionless exponent of solution of gas diffusivity
km matrix permeability [L2]
equation using the moment method
Lc matrix block characteristic length [L]
l location in the matrix block with pressure equal to the
l fluid viscosity [M/LT]
q fluid density [M/L3]
average pressure [L]
r shape factor [1/L2]
MW gas molecular weight
/ porosity
pm matrix block pressure [M/LT2]
wm average matrix block pseudo pressure [M/LT3]
pm average matrix block pressure [M/LT2]
wf fracture pseudo pressure [M/LT3]
pf fracture pressure [M/LT2]
wi initial pseudo pressure [M/LT3]
q matrix–fracture fluid transfer [M/L3T]
wD dimensionless pseudo pressure
qm matrix–fracture fluid transfer [L3/T]
wD average dimensionless pseudo pressure
qacc accumulation rate [M/L3T]
Q cumulative fluid exchange [L3] Subscripts
R residual in the method of moments D dimensionless
t time [T] f fracture
tD dimensionless time g gas
T reservoir temperature [K] i initial condition
Vb matrix block volume [L3] m matrix
xD dimensionless distance R reservoir conditions
Z gas deviation factor SC standard conditions

where /m is the matrix porosity and cm is the total isothermal com- value of the shape factor for slightly compressible fluids cannot
pressibility (rock and fluid) [22]. be used for gases, or its applicability has not been verified in the
The pseudo steady-state model does not fully account for the literature. The same problem exists for coalbed methane reservoirs
pressure transient within the matrix. There is another exchange and most fractured tight gas reservoirs.
model for fluid flow between the matrix and fracture, which ac- In this paper we develop an approximate semi-analytical shape
counts for the pressure transient in the matrix block by solving factor model and investigate the gas specific gravity, pressure and
the diffusivity equation, which is given by: temperature dependency of the shape factor for the flow of a com-
  pressible fluid in fractured porous media. The determination of
km @pm such a shape factor has not been addressed in earlier works. To
r rpm ¼ /m cm ð3Þ
l @t determine the value of the shape factor, diffusivity equation for
gas flow, which is a nonlinear PDE, is solved using the combination
The matrix hydraulic diffusivity is defined as gm = km/lcm/m.
of the integral method and the method of moments and this
Since both viscosity and the isothermal compressibility are con-
solution is used to evaluate the shape factor. A fine-grid, single-
stant for a slightly compressible fluid, the hydraulic diffusivity is
porosity, numerical model is used to verify the approximate
also a constant and Eq. (3) is a linear partial differential equation
semi-analytical dual-porosity model presented in this study.
(PDE). This equation can be solved using ordinary methods, such
as separation of variables or the Laplace transform [8]. This type
of model is called a transient model and does not use the shape fac- 2. Previous works
tor directly [16]. In other words, the matrix diffusivity equation
(Eq. (3)) can be solved with the fracture pressure as a boundary There are two main categories for modeling fluid transfer be-
condition. The shape factor can then be calculated by a combina- tween a fracture and a matrix using dual-porosity concept. These
tion of Eqs. (1) and (2) and the analytical solution of the diffusivity include single-phase models, which are used in well test analysis,
equation, subject to fracture pressure boundary condition. and a multiphase approach, which is applied in multiphase flow
Different models have been presented for the evaluation of the simulation in fractured media [27]. In this section, a brief review
shape factor for NFRs for slightly compressible fluids. Most of these of the relevant literature dealing with the single-phase shape fac-
models use a pseudo steady-state assumption in the shape factor tor for slightly compressible fluids is presented.
derivation and obtain a constant for the shape factor. Several inves- Warren and Root [36] defined the dual-porosity model for the
tigators considered the transient state of the matrix–fracture inter- single phase flow of a slightly compressible fluid under a pseudo
action and obtained a time-dependent shape factor. steady-state condition. Warren and Root obtained dimensionless
All of the models that include a shape factor presented in the lit- shape factors of 12, 32 and 60 for one, two and three sets of frac-
erature are for slightly compressible fluids and assume constant tures, respectively, for uniformly distributed sets of parallel
viscosity and total isothermal compressibility. Due to the high- fractures.
pressure dependency of fluid isothermal compressibility and Application of the shape factors in numerical simulation of
viscosity for compressible fluids, such as gases, either the derived dual-porosity media was introduced by Kazemi et al. [20] to
E. Ranjbar, H. Hassanzadeh / Advances in Water Resources 34 (2011) 627–639 629

account for the flow between the matrix and the fracture. Kazemi Sarda et al. [30] presented a model to derive the single-phase
et al. [20] obtained the values of shape factor as 4, 8 and 12 for one, shape factor. They used numerical simulations of a single-phase
two and three sets of fracture, respectively. fluid flow on discrete fracture models and derived the dimension-
Thomas et al. [32] studied various fine-grid, single-porosity and less shape factors of 8, 24 and 48 for 1D, 2D and 3D flows, respec-
single-block, dual-porosity models. For a three-dimensional oil– tively. In their model, the shape factor was spatially dependent on
water model, they obtained a good match with a dimensionless the local properties of the matrix blocks and the surrounding
shape factor of 25. Coats [11] derived a shape factor exactly twice fractures.
the value of Kazemi et al. [20]. Coats solved the diffusivity equation Hassanzadeh and Pooladi-Darvish [16] used the Laplace domain
for slightly compressible fluids using the Fourier finite sine trans- solutions of the diffusivity equation for different geometries and
form to obtain dimensionless shape factors of 12, 28.45 and boundary conditions to show that the shape factor depends on
49.58 for the one-dimensional (1D), two-dimensional (2D) and the pressure depletion regime in the fracture. In their model, a dif-
three-dimensional (3D) cubic rock matrices, respectively. Udea fusivity equation was solved for a slightly compressible fluid in the
et al. [33] introduced a multiplication factor to the Kazemi et al. Laplace domain for Cartesian, cylindrical and spherical coordinates.
[20] shape factor. They concluded that Kazemi’s shape factor Hassanzadeh et al. [17] extended their previous analysis and
needed to be multiplied by factors of two and three for 1D and used infinite acting radial and linear dual-porosity models, where
2D flows, respectively. the boundary condition is chosen at the wellbore, as opposed to
Zimmerman et al. [38] developed a new dual-porosity model the matrix boundary. They solved coupled equations for the frac-
for single-phase fluid flow in fractured porous media. Their model ture and matrix, taking into account the transient exchange be-
used a nonlinear ordinary differential equation to describe the tween the matrix and the fracture. They showed that, for a well
process and calculate the matrix–fracture interaction term, which producing at a constant rate from a naturally fractured reservoir,
more precisely models the flux during early and late periods of the appropriate value of stabilized dimensionless shape factors
the interaction. This equation was shown to be more accurate are 12, 25.13 and 38.9 for one, two and three sets of fractures,
than the linear Warren and Root equation [36]. Their model respectively. The pseudo steady-state values of dimensionless
was capable to perform accurate dual-porosity simulations, using shape factors for 1D, 2D and 3D flows are shown in Table 1.
a substantially smaller number of grid blocks than would be In general, two approaches have been used to determine the
needed in a fully discretized simulation. Zimmerman et al. [39] shape factor including non-coupled [9,16] and coupled [17]. In
also improved their model to simulate unsaturated flow pro- the non-coupled case fracture acts as a boundary condition to
cesses in the dual-porosity media. Chang [9], Kazemi and Gilman the matrix block. However, in the coupled case the boundary con-
[19] and Lim and Aziz [22] also derived matrix–fracture shape dition is considered at the wellbore and the shape factor is derived
factors for single-phase flow of a slightly compressible fluid for by coupling the pressure diffusion in both matrix and fracture. In
various parallelepiped geometries of the matrix blocks. Chang most of these studies for a slightly compressible fluid the time
[9] showed that, for 1D, 2D and 3D diffusive flows, the shape for stabilization of the shape factor constant was reported to be
factor is a function of time. Shape factor values are high at the about 0.1. In addition, it was shown that the shape factor constant
initial depletion stage during unsteady-state transfer; and, as is inversely proportional to the square root of dimensionless time
time passes, the shape factors converge to a stabilized value. during the transient period [17].
The stabilized values of the shape factors for 1D, 2D and 3D flows Penuela et al. [27] developed an interporosity flow equation for
are shown in Table 1. naturally fractured gas condensate reservoirs. They derived a time-
Quintard and Whitaker [29] used the volume-averaging tech- dependent shape factor using pseudo functions and accounting for
nique to derive the shape factors for single-phase flow of slightly the multiphase effect of gas condensate systems. Lu and Connel
compressible fluids. They derived the value of dimensionless [23] derived a dual-porosity formulation for fluid flow in gas reser-
shape factors of 12, 28.4 and 49.6 for 1D, 2D and 3D flows, voirs that couples large-scale fractures and micropores within
respectively. dual-porosity media by considering the adsorption behavior.
Bourbiaux et al. [6] used a numerical technique and derived a There have also been recent attempts to find the shape factors
dimensionless shape factor for 2D matrix–fracture transfer, based for two-phase flow and thermal processes, which are beyond the
on a pseudo steady-state assumption. The shape factor demon- scope of this work and are not reviewed here [7,10,34]. As an
strated a transient behavior and converged to a constant value of example, Civan and Rasmussen [10] developed an analytical model
20.1 at a later time. for rectangular shape matrix blocks. They investigated capillary
imbibition process in fractured porous media. van Heel et al. [34]
used a dual-permeability approach to derive the shape factor for
Table 1
different recovery mechanisms, particularly for thermal processes.
Pseudo steady-state values of single-phase dimensionless shape factors.
van Heel et al. showed that the recovery mechanism has an impor-
Investigator(s) 1D flow 2D flow 3D flow tant effect on the shape factor value [34,35].
Warren and Root [36] 12 32 60 All models discussed above have been developed for single-
Kazemi et al. [20] 4 8 12 phase flow of a slightly compressible fluid, where the viscosity
Coats [11] 12 28.45 49.58 and isothermal compressibility were assumed to be constant.
Zimmerman et al. [38] 9.87(p2) 19.74(2p2) 29.61(3p2)
Kazemi and Gilman [19] 9.87(p2) 19.74(2p2) 29.61(3p2)
However, due to the variability of the isothermal compressibility
Lim and Aziz [22] 9.87 or 18.17a or 25.67b or and the viscosity of a compressible fluid (gas) with gas specific
p2 2p2 3p2 gravity, temperature and pressure, these derived shape factors can-
Chang [9] p2 2p2 3p2 not be used for flow of a compressible fluid in a fractured porous
Hassanzadeh and Pooladi-Darvish [16] 9.87 18.2a 25.65b
media, or this application needs further verification.
(constant fracture pressure)
Hassanzadeh et al. [17] (constant rate) 12 25.13a 38.9b In this paper, we solve gas diffusivity PDE that is nonlinear
Quintard and Whitaker [29] 12 28.4 49.6 equation, due to the high-pressure dependency of the viscosity
Bourbiaux et al. [6] – 20.1 – and isothermal compressibility of gases, and derive a shape factor
Sarda et al. [30] 8 24 48 for gas flow in dual-porosity systems that accounts for both tran-
a
Cylindrical approximation. sient and pseudo steady-state fluid exchange between the matrix
b
Spherical approximation. and the fracture. We then validate our results with fine-grid
630 E. Ranjbar, H. Hassanzadeh / Advances in Water Resources 34 (2011) 627–639

numerical simulations. The derived shape factor may be used for km T sc A wm  wf


dual-porosity modeling of conventional gas reservoirs, coalbed qsc ¼ ð8Þ
Tpsc l 2
methane and fractured tight gas reservoirs. We also investigate
the effect of gas specific gravity, temperature and pressure on In the case of a matrix–fracture system (as illustrated in Fig. 1),
the shape factor. Eq. (8) is multiplied and divided by the bulk volume of the matrix
The following section describes the method that we used to de- block, in order to determine the transfer function. The ratio of A/
rive the single-phase matrix–fracture transfer shape factor for [l(Vb/2)] is called the shape factor (since the flow occurs in the half
dual-porosity systems. plane of the matrix, the total volume of the matrix block, Vb, is di-
vided by two in this equation) and has the dimension of the reci-
procal of the area; and, the final formula for the matrix–fracture
3. Derivation of the shape factor for compressible fluids
transfer function for compressible fluids (e.g. gases) is given by:
Due to the high-pressure dependency of the isothermal com- T sc V b km r
qsc ¼ ðwm  wf Þ ð9Þ
pressibility and viscosity of a compressible fluid and consideration 4psc T
of the non-ideality of the gas phase, there is no guarantee that the
where wm is the average matrix block pseudo pressure and wf is the
shape factors calculated for isothermal flow of a slightly compress-
fracture pseudo pressure.
ible fluid will continue to be applicable to the gas flow, as they
In a dual-porosity system accumulation term in the matrix
were obtained based on a constant isothermal compressibility
block is expressed based on the following equation:
and viscosity. In this section, a model for a shape factor of single-
 
phase flow of a compressible gas is presented. @ðqm /m Þ @q @/
qacc ¼  ¼  /m m þ qm m ð10Þ
We start with Darcy’s law and derive the shape factor for flow of @t @t @t
a compressible fluid in a dual-porosity system. Darcy’s law for gas
flow in porous media is given by: Using the pressure-dependent density equation for compress-
ible fluid and the definition of real gas pseudo pressure (Eq. (7))
km A dp leads to the following equation for accumulation term in the ma-
qgsc ¼  ð4Þ
lBg dx trix for a compressible fluid:
where A is the cross-sectional area and Bg is the gas formation vol- MW @w
ume factor, which is defined as follows: qacc ¼  l/m cm m ð11Þ
2RT @t
ðV g ÞR qgsc psc ZT psc ZT where R is the universal gas constant. Eq. (11) is divided by density
Bg ¼ ¼ ¼ ¼ ð5Þ
ðV g Þsc qgR pT sc T sc p and multiplied by half of the matrix block volume to change its unit
from kg/(m3 s) to m3/s as follows:
where T is the absolute temperature; Z is the gas deviation factor
and accounts for gas non-ideality; and, subscripts g, sc, and R denote Z @w
qR ¼  l/m cm V b m ð12Þ
gas, standard condition, and reservoir condition, respectively. Com- 4pm @t
bining Eqs. (4) and (5) and integrating along half of the matrix block
Definition of the gas formation volume factor (Eq. (5)) is used to
length (Fig. 1), we obtain:
convert the gas rate into the standard conditions as given by:
Z Lc Z pf
km A T sc p
dx ¼ dp ð6Þ T sc V b lcm /m @wm
Lc l Tqsc psc pm lZ qsc ¼  ð13Þ
4psc T @t
In the above equation l is a time-dependent length where the
Combining Eqs. (9) and (13) leads to the definition of the single-
matrix pressure is equal to its average pressure. In the literature
phase shape factor for compressible fluids, as given by:
for gas flow, real gas pseudo pressure is defined as follows [18]:
Z p lcm /m @wm
p r¼   ð14Þ
w¼2 dp ð7Þ km wm  wf @t
pb lZ
Inserting Eq. (7) into Eq. (6), the following equation for Darcy’s law It should be noted that we cannot use a single value for the vis-
is obtained: cosity–compressibility product, because this product is strongly
pressure-dependent for a compressible fluid. In Eq. (14), the aver-
age pseudo pressure and its time derivative obtained from solution
hm of the gas diffusivity equation is used to find the shape factor.

4. Solution of gas diffusivity equation


Matrix
The gas diffusivity equation for Cartesian flow can be expressed
as:

@ 2 w lcm /m @w
Fracture ¼ ð15Þ
@x2 km @t
Eq. (15) is a nonlinear PDE, due to the pressure dependency of the
viscosity and isothermal compressibility, which cannot be solved
Lc by ordinary methods. Using the definition of the matrix hydraulic
diffusivity, gm, which is defined as km/lcm/m in Eq. (15), leads to
the following equation:
x @w @2w
¼ gm ðpÞ 2 ð16Þ
@t @x
Fig. 1. Schematic of the matrix–fracture model.
E. Ranjbar, H. Hassanzadeh / Advances in Water Resources 34 (2011) 627–639 631

 3
It should be noted that gm is a strong function of pressure; and, 1  xD
since pressure is a function of distance and time, the matrix wD ¼ 1  ð26Þ
1  dðt D Þ
hydraulic diffusivity is a function of these two parameters as well.
For solving this PDE, we use the same assumption as was made by Using this trial solution in the integral form of Eq. (24) leads to the
Agarwal [1] for real gas pseudo time method and assume that the following solution for early time pseudo pressure:
compressibility-viscosity product is a function of time and not pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 3 !3
24bgD1 t D  1
xD þ 1  xD 1
space. To consider the effect of space, we multiply the hydraulic wD ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ 1  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi tD <
diffusivity by a correction factor, b. This correction factor is then 24bgD1 tD 24bgD1 tD 24bgD1 t D 24bgD1
obtained using fine-grid numerical simulations. ð27Þ
Another common assumption is that we neglect the rock com-
pressibility as compared to the gas compressibility and assume where gD1 is the fracture dimensionless hydraulic diffusivity and is
that total isothermal compressibility is equal to the fluid isother- defined as follows:
mal compressibility [1,18]. Therefore, for shape factor determina- gm ðxD ¼ 1Þ km =lf cf /m
tion, the following PDE (Eq. (17)) with the initial and boundary gD1 ¼ ¼ ð28Þ
g g
conditions listed below (Eqs. (18a)–(18c)) should be solved
  Integrating of Eq. (27) over bulk volume of the matrix block, the
@w @ @w
¼ bgm ðtÞ ð17Þ following equation for the early time average dimensionless pseu-
@t @x @x do pressure is obtained:
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
t ¼ 0 ! w ¼ wi ð18aÞ 24bgD1 t D 1
wD ¼ ; tD < ð29Þ
@w 4 24bgD1
x¼0! ¼0 ð18bÞ
@x Derivation of Eqs. (27) and (29) is shown in Appendix A.1 in more
x ¼ Lc ! w ¼ wf ð18cÞ detail.
By changing Eq. (14) into a dimensionless form using the
It should be noted that Lc is the matrix block characteristic
dimensionless variables of Eqs. (20)–(23), the following equation
length and is half of the matrix block thickness (hm). A schematic
for the dimensionless shape factor is obtained:
representation of the matrix–fracture system is illustrated in Fig. 1.  
For solving Eq. (17), we define the average hydraulic diffusivity 4 1 @wD
rh2m ¼  ð30Þ
over the matrix–fracture pressure drawdown as follows: gD wD  1 @t D
Z pf Z pf
1 km km 1 dp In Eq. (30), gD is not a constant and is a function of pressure,
g¼ dp ¼ ð19Þ
pf  pi pi lcm /m /m pf  pi pi lcm specific gravity and temperature since the viscosity–compressibil-
ity product is dependent on these parameters. Dimensionless
It should be noted that the average hydraulic diffusivity for any hydraulic diffusivity is defined as follows:
specified matrix–fracture system can be calculated using common    
numerical integration methods, such as trapezoidal or Simpson’s 1 km km 1 km 1
gD ¼ ¼ and gD1 ¼ ð31Þ
rules. For any specified naturally fractured gas reservoir, the aver- g lcm /m g/m lcm g/m lf c f
age hydraulic diffusivity is a function of the matrix porosity and
Substituting Eq. (29) and its derivative into Eq. (30) leads to the fol-
permeability, temperature, matrix–fracture pressure drawdown
lowing equation for the early time dimensionless shape factor, as
and gas specific gravity.
given by:
Eq. (17) and its initial and boundary conditions (Eqs. (18a)–
pffiffiffiffiffiffiffiffiffiffiffiffiffi !
(18c)) can be made dimensionless by defining the following 4 6bgD1 1 1 1
2
dimensionless variables r hm ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffi ; tD < ð32Þ
gD 4  24bgD1 t D tD 24bgD1
w  wi
wD ¼ ð20Þ The method of moments [3,9,12] is used to find the late time
wf  wi
solution of Eq. (24) with the suggestion of a third-order polynomial
x
xD ¼ ð21Þ trial solution (Eq. (33)) and the residual R (Eq. (34)) as follows:
Lc
g ðtÞ 1
gD ðtÞ ¼ m ð22Þ wD ðxD ; tD Þ ¼ AðtD Þ þ Bðt D ÞxD þ Cðt D Þx2D þ Dðt D Þx3D ; tD P
g 24bgD1
gt ð33Þ
tD ¼ ð23Þ  
L2c @wD @ @w
R¼  bgD D ð34Þ
@t D @xD @xD
By combining Eqs. (20)–(23) into Eqs. (17)–(18), the following
dimensionless PDE with initial and boundary conditions (Eqs. In Eq. (33), the unknown coefficients A, B, C and D are found
(25a)–(25c)) can be obtained using boundary conditions and making the zero and first moments
  of R (Eqs. (35) and (36), respectively) vanish by enforcing the fol-
@wD @ @w
¼ bgD ðtÞ D ð24Þ lowing conditions [3]:
@tD @xD @xD
Z 1  Z 1 
@wD @ @w
RdxD ¼ 0 !  bgD ðtÞ D dxD ¼ 0 ð35Þ
t D ¼ 0 ! wD ¼ 0 ð25aÞ 0 0 @tD @xD @xD
@w Z 1 Z 1   
xD ¼ 0 ! D ¼ 0 ð25bÞ @wD @ @w
@xD xD RdxD ¼ 0 ! xD  bgD D dxD ¼ 0 ð36Þ
0 0 @tD @xD @xD
xD ¼ 1 ! wD ¼ 1 ð25cÞ
After enforcing the boundary conditions of Eqs. (25b) and (25c) on
We use integral method [13,15,37] to find the early time solu- the trial solution (Eq. (33)) and inserting the trial solution into the
tion of this nonlinear diffusion equation by defining the penetra- integral equations of Eqs. (35) and (36) and using the initial condi-
tion depth (1  d(tD)) and the following third order trial solution: tion from the early time solution, the following equation for the
632 E. Ranjbar, H. Hassanzadeh / Advances in Water Resources 34 (2011) 627–639

dimensionless pseudo pressure at the late time is obtained: verify the developed model. Tables 2 and 3 summarize the data
used for the numerical simulations and the approximate semi-
wD ðxD ; t D Þ ¼ ð1  1:252 expðbk1 tD Þ þ 0:489 expðbk2 t D ÞÞ
analytical model of the base case (gas specific gravity, c = 0.7,
þ ð1:793 expðbk1 tD Þ  6:175 expðbk2 t D ÞÞx2D T = 93.3 °C and pressure ranges from 45 to 22.5 MPa).
þ ð0:541 expðbk1 tD Þ þ 5:686 expðbk2 t D ÞÞx3D ; It should be noted that, in Table 3, the gas viscosity and the gas
1 deviation factor were calculated using the PVTi module of Eclipse;
tD P ð37Þ
24bgD1 the viscosity was calculated using the Lohrenz, Bray and Clark
(LBC) correlation; and, three parameters of the Peng-Robinson
equation of state were used to calculate the gas deviation factor
Integrating over bulk volume of the matrix block leads to the fol- in PVTi [14]. The gas isothermal compressibility (cg) was calculated
lowing equation for the late time average dimensionless pseudo using the correlation of Mattar et al. [24] and McCain [25].
pressure: For the base case (c = 0.7, T = 93.3 °C and pressure drawdown
from 45 to 22.5 MPa), the calculated values of the average hydrau-
1 lic diffusivity and the dimensionless fracture hydraulic diffusivity,
wD ¼ 1  0:790 expðbk1 tD Þ  0:148 expðbk2 tD Þ; tD P
24bgD1 gD1, using Eqs. (19) and (28) are 0.03457 and 0.3691, respectively.
ð38Þ The following equation gives the cumulative fluid exchange from
the matrix block for the base case, which has been obtained using
where k1 and k2 are functions of the fracture dimensionless hydrau- the developed approximate semi-analytical model and the data
lic diffusivity, gD1, and are defined as follows: provided in Tables 2 and 3
k1 ¼ 2:486gD1 ð39Þ
k2 ¼ 32:181gD1 ð40Þ 78:9256
Q¼ ð2:1404 expð0:00793btÞ  0:4010
gD
The derivation of Eqs. (37)–(40) is shown in Appendix A.2 in more
detail.  expð0:10266btÞ þ 2:5414Þ ð42Þ
Substituting Eq. (38) and its derivative in Eq. (30) leads to the
following equation for the late time dimensionless shape factor,
where Q is the cumulative fluid exchange from the matrix in stan-
as given by:
dard cubic meters and t is the real time in seconds. The derivation of
4bgD1 1:964 expðbk1 t D Þ þ 4:763 expðbk2 t D Þ 1 this equation is shown in Appendix B. The obtained values for the
rh2m ¼ ; tD P correction factor (b) and the matching parameter (gD) for the base
gD 0:790 expðbk1 tD Þ þ 0:148 expðbk2 tD Þ 24bgD1
case are 0.730 and 0.3127, respectively.
ð41Þ
In Eq. (41), all parameters, except b and gD, can be calculated for
any specified gas, in order to determine the late time dimension- Table 2
less shape factor at any value of dimensionless time. Parameter b Data used for the approximate semi-analytical and numerical models.
is the correction factor that considers the effect of the variability
Data for fine-grid model
of the hydraulic diffusivity with distance. The two unknown Grid dimension: 22  1  1
parameters, b and gD, are used as matching parameters. To do so, Grid spacing: Dx (matrix) = 0.5, 2  0.25, 4  0.125, 0.09, 2  0.0625,
the cumulative fluid exchange from the matrix to the fracture is 3  0.06, 3  0.03, 0.01, 0.005 m
calculated using numerical simulations [14] and then used to Dx (fracture) = 20, 100, 1000, 10,000 m
Dy = Dz = 4 m
determine b and gD. The matching parameter, b, controls the early Fracture porosity = 1
behavior of the late time solution, since it is included to account for Fracture permeability = 4000 mD
the pressure gradient in the matrix. These pressure gradients are Common data for the approximate semi-analytical and numerical models
important in the early behavior of the late time solution; therefore, Gas specific gravity = 0.7
we choose to match the early part of the late time solution using b. Matrix permeability = 1 mD = 9.869233  1016 m2
On the other hand, the dimensionless diffusivity coefficient (gD) Matrix porosity = 0.1
Initial pressure = 45 MPa
controls the pseudo steady-state fluid exchange between the ma-
Fracture pressure = 22.5 MPa
trix and the fracture. Since the pseudo steady state forms the basis Reservoir temperature = 93.3 °C
for the stabilized fluid exchange between the matrix and the frac- hm = 4 m
ture, as shown by Eq. (30), we use gD to match the pseudo steady-
state part of the solution.
In the following section verification of the developed model is
presented. Table 3
Main data of gas for comparison of the approximate semi-analytical and simulation
results (base case).
5. Model verification
Pressure Gas deviation Viscosity  108 cg  105
A fine-grid, single-porosity numerical model (Eclipse 100) was (MPa) factor (kPa s) (kPa1)
used to verify the matrix–fracture shape factor derived in this pf 22.5 0.7825 2.35 3.2920
study. For one set of fractures, the fine-grid model is considered 25 0.7766 2.52 2.5536
as a half plane with the thickness of Lc equal to 2 m. To maintain 27.5 0.7744 2.70 1.9660
30 0.7752 2.89 1.5061
a constant fracture pressure, four large-grid blocks, with porosity 32.5 0.7788 3.08 1.1541
of one, permeability of 4 Darcy and pressure equal to the fracture 35 0.7847 3.27 0.8884
pressure, were defined. The matrix was defined as region one 37.5 0.7925 3.46 0.6897
and the fracture as region two; and, the total fluid exchange from 40 0.8019 3.65 0.5417
42.5 0.8127 3.83 0.4312
region one to region two in standard cubic meters from the simu-
pi 45 0.8245 4.01 0.3479
lator were used to determine the correction factor (b) and gD and to
E. Ranjbar, H. Hassanzadeh / Advances in Water Resources 34 (2011) 627–639 633

Fig. 2 shows the cumulative fluid exchange in standard cubic


meters versus real time in seconds. The cumulative fluid exchange
is calculated by using the following equation:

Q ¼ 252:4004ð2:1404 expð0:005789tÞ  0:4010


 expð0:07494tÞ þ 2:5414Þ ð43Þ
As illustrated in Fig. 2, there is a reasonable match between the
developed model and the fine-grid numerical simulation.
Fig. 3 shows the log–log plot of the shape factor constant as a
function of dimensionless time for the base case, based on Eqs.
(32) and (41). Results demonstrate that the dimensionless shape
factor derived here shows a transient behavior and then converges
to a constant value at a late time. For the base case, the dimension-
less shape factor stabilizes at a dimensionless time of about 0.7,
which corresponds to a real time of about 81 s. The stabilized value
of the dimensionless shape factor is shown to be 8.57. As illus-
trated in the figure and according to Eq. (32) dimensionless shape Fig. 4. Comparison of the cumulative fluid exchanges between the matrix and the
fracture obtained from the approximate semi-analytical solution and the numerical
factor at the early times varies inversely proportional with the
model (c = 0.7, T = 71.1 °C and pressure drawdown of 45–22.5 MPa).
square root of time.
Fig. 4 shows comparisons between the approximate semi-ana-
lytical dual-porosity model derived in this study and the fine-grid, show an acceptable accuracy. Similar accuracy was found for two
single-porosity, numerical model for a gas specific gravity of 0.7, a other cases (c = 0.55, drawdown of 45–22.5 MPa and c = 0.70,
temperature of 71.1 °C, and a drawdown of 45–22.5 MPa. Results drawdown of 36–20 MPa) at a temperature of 93.3 °C. For all cases,
the permeability and porosity of the matrix and fracture and their
grid block sizes were the same; and, the factors that have an effect
on gas properties, including specific gravity, temperature and pres-
sure, were changed.
The developed model with b = gD1 = gD = 1 should recover the
shape factor for a slightly compressible fluid. To further verify
the model the shape factor developed here is compared with the
shape factor of the slightly compressible fluid. Results show that
the model could recover the slightly compressible fluid shape fac-
tor for the entire period of time. Detail of this comparison is shown
in Appendix C.

6. Effect of different parameters on the shape factor

6.1. Gas specific gravity effect

To study the effect of gas specific gravity on the shape factor, we


use a matrix–fracture pressure drawdown of 45–22.5 MPa and a
Fig. 2. Comparison of the cumulative fluid exchanges between the matrix and the temperature of 93.3 °C with the same rock properties applied in
fracture obtained from the approximate semi-analytical solution and the numerical the base case. Eqs. (32) and (41) are used to calculate the dimen-
model (base case).
sionless shape factor for different values of gas specific gravity. Ta-
ble 4 summarizes the values of the different parameters to be used
in Eqs. (32) and (41) for calculation of the shape factor at different
gas specific gravities.
In Table 4, g; gD1 ; k1 ; k2 and (wi  wf) are calculated using fol-
lowing equations for different gas specific gravities
Z pf Z pf
1 km km 1 dp
g¼ dp ¼ ð44Þ
pf  pi pi lcm /m /m pf  pi pi lc m
gm ðxD ¼ 1Þ km =lf cf /m
gD1 ¼ ¼ ð45Þ
g g
k1 ¼ 2:486gD1 ð46Þ
k2 ¼ 32:181gD1 ð47Þ
Z pi
p
wi  wf ¼ 2 dp ð48Þ
pf lz
Fig. 5 compares the values of the dimensionless shape factor for
different gas specific gravities. It should be noted that, for all of
these specific gravities, an acceptable match between the approx-
imate semi-analytical and numerical models, similar to Figs. 2 and
Fig. 3. Dimensionless shape factor for the base case. 4, was obtained. As illustrated in this figure, the transient values of
634 E. Ranjbar, H. Hassanzadeh / Advances in Water Resources 34 (2011) 627–639

Table 4
Different parameters used in the shape factor calculation for different gas specific gravities (c).

Gas specific gravity g gD1 gD (matching parameter) b (correction factor) k1 k2 (wi  wf)  1016
0.55 0.01971 0.5860 0.6950 0.97 1.45680 18.85807 6.0838875
0.60 0.02117 0.5465 0.6180 0.94 1.35860 17.58692 6.1667100
0.65 0.02492 0.4757 0.4790 0.86 1.18259 15.30850 6.0524500
0.70 0.03457 0.3691 0.3127 0.73 0.91758 11.87801 6.021300
0.75 0.05920 0.2685 0.1725 0.56 0.66749 8.64060 5.865268
0.80 0.11500 0.2739 0.0974 0.30 0.68092 8.81438 5.521590

Fig. 5. Effect of gas specific gravity on the dimensionless shape factor (T = 93.3 °C, Fig. 6. Cumulative fluid exchange for different gas specific gravities as a function of
pressure drawdown of 45–22.5 MPa). time.

the dimensionless shape factor increase as the gas specific gravity isothermal compressibility and the gas deviation factor, we study
increases. The stabilized values of the dimensionless shape factor the effect of this parameter on the shape factor. We investigate
for different gas specific gravities are very close to each other the shape factor for the base case temperature and three
and range from 8.13 to 8.67. Results show that lower gas specific other common reservoir temperatures, including 71.1, 82.2 and
gravities demonstrate a shorter transient period: for example, for 104.4 °C. Fig. 7 is a comparison of the shape factors for different
c = 0.55, the pseudo steady-state time is about tD = 0.3, but for temperatures for a gas specific gravity of 0.7 and a pressure draw-
c = 0.8, the pseudo steady-state fluid exchange occurs after down of 45–22.5 MPa.
tD = 2.6. Table 5 shows the stabilized values of the shape factor As illustrated in Fig. 7, there is a decrease in the transient values
for different gas specific gravities. of the shape factor as the temperature increases. There is also an
As illustrated in Table 5, as the gas specific gravity increases increase in onset of the pseudo steady-state by decreasing the
from 0.55 to 0.75, the stabilized value of the dimensionless shape
factor increases from 8.13 to 8.67; and, there is a decrease from
8.67 to 8.39, when specific gravity changes from 0.75 to 0.8.
Fig. 6 compares the cumulative fluid exchange versus time for
different gas specific gravities. Results show that c = 0.75 has the
highest cumulative fluid exchange from the matrix, and a gas with
c = 0.55 has the lowest cumulative fluid exchange from the matrix.
These observations are consistent with the stabilized values of the
shape factor shown in Fig. 5 and Table 5.

6.2. Temperature effect

Since temperature is an important parameter that changes


the properties of a compressible fluid, including the viscosity,

Table 5
Stabilized values of the dimensionless shape factor for different gas specific gravities.

Gas specific gravity Stabilized value of the shape factor


0.55 8.13
0.60 8.27
0.65 8.49
0.70 8.57
0.75 8.67
0.80 8.39 Fig. 7. Effect of temperature on the dimensionless shape factor (c = 0.7, pressure
drawdown of 45–22.5 MPa).
E. Ranjbar, H. Hassanzadeh / Advances in Water Resources 34 (2011) 627–639 635

Table 6 present, the stabilized values of the shape factor is a weak function
Effect of temperature on stabilized value of the dimensionless shape factor. of the gas specific gravity and temperature, with this value varying
Temperature (°C) Stabilized value of shape factor between 8.13 and 8.69 for different gas specific gravities and dif-
71.1 8.69 ferent reservoir temperatures. The shape factor is stabilized at
82.2 8.62 dimensionless times ranging from 0.3 (c = 0.55) to 2.6 (c = 0.8),
93.3 8.57 which correspond to the real times of 60 and 90 s, respectively.
104.4 8.52 Chang [9] reported the dimensionless pseudo steady-state time
of 0.1, which corresponds to less than one day for commonly frac-
tured rock in the case of a slightly compressible fluid and a con-
temperature. Table 6 shows the stabilized values of the shape fac- stant fracture pressure.
tor for different temperatures. The stabilized values of the shape
factor are very close to each other, and there is a slight decrease 7. Second-order trial solution
in the stabilized value of the shape factor when increasing the
temperature. In Section 4, we assumed a third-order trial function for the
early and late time solution. In this section, a second-order trial
6.3. Pressure effect solution and its effect on the shape factor are considered. If we
use a second-order trial solution for Eq. (24) with the conditions
In the previous sections, the pressure drawdown range was 45– of Eq. (25) following equations for the early and late time solution
22.5 MPa. In this section, the effect of three other different pressure are obtained, respectively:
drawdowns on the shape factor is considered. Fig. 8 compares the !2
values of the shape factor for different matrix–fracture pressure 1  xD
wD ðxD ; tD Þ ¼ 1  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; t D < 12bgD1 ð49Þ
drawdowns. Fig. 8 show that the transient values of the shape fac- 12bgD1 t D
tor increase as matrix–fracture pressure draw-down increases.
This figure shows that the stabilized values of the shape factor wD ðxD ; tD Þ ¼ 1  expð0:25  3bgD1 tD Þ
for different cases are very close to each other.
þ ðexpð0:25  3bgD1 tD ÞÞx2D ; t D P 12bgD1 ð50Þ
Table 7 shows the stabilized values of the shape factor for dif-
ferent matrix–fracture pressure drawdowns. This table shows that Integration over the bulk volume of the matrix block, the equations
the variability of the stabilized values of the shape factor with for the early and late time average dimensionless pseudo pressure
pressure is negligible; and, in general, one may ignore the pressure are obtained as follows:
effect on the stabilized value of the shape factor. pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Results of this study show that, in a dual-porosity system sub- 12bgD1 t D
wD ¼ ; t D < 12bgD1 ð51Þ
ject to a constant fracture pressure when one set of fractures is 3
wD ¼ 1  ð2=3Þ expð0:25  3bgD1 t D Þ; tD P 12bgD1 ð52Þ
By substituting wD and its derivative into the shape factor for-
mula (Eq. (30)), the obtained equations for the early and late time
shape factor are as follows:
pffiffiffiffiffiffiffiffiffiffiffiffiffi !
2 4 3bgD1 1 1 1
r hm ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffi ; tD < ð53Þ
gD 3  12bgD1 t D tD 12bgD1
12bgD1 1
rh2m ¼ ; tD P ð54Þ
gD 12bgD1
where the fracture dimensionless hydraulic diffusivity (gD1), the
correction factor (b) and gD should be calculated as described in
the previous section.
Fig. 9 compares the derived shape factor for the second and
third order trial solution for the base case. Using the second-order
trial solution, the obtained value of the stabilized dimensionless
shape factor is 10.69, compared to a stabilized value of 8.57 for
the third-order trial solution. For the base case, the best match is
obtained for a correction factor of b = 0.70 and gD = 0.2900.
Fig. 10 is a comparison of the cumulative fluid exchanges versus
time for the numerical solution and the approximate semi-
analytical second- and third-order trial solutions for the base case.
Fig. 8. Effect of pressure drawdown on the dimensionless shape factor (c = 0.7,
T = 93.3 °C). As illustrated in this figure, the second- and third-order trial
solutions act in the same manner. It should be noted that, the
third-order trial solution is closer to the numerical solution and
is more accurate, since the second-order trial solution cannot accu-
Table 7 rately model the early and late transient fluid exchange between
Effect of pressure on the stabilized values of the dimensionless shape factor.
the matrix and the fracture.
Pressure drawdown (MPa) Stabilized value of shape factor
34–4 8.51 8. Summary and conclusions
42–18 8.54
36–20 8.53
In this study, a theoretical analysis of the shape factor for the
45–22.5 8.57
flow of a compressible fluid in fractured porous media is presented.
636 E. Ranjbar, H. Hassanzadeh / Advances in Water Resources 34 (2011) 627–639

second-order trial solution leads to a pseudo steady-state shape


factor, which is less accurate than the third-order trial solution in
predictions of the cumulative fluid exchange. In the both cases
the shape factor is obtained as a function of time. Results show that
the third-order model leads to a better match with the fine-grid
numerical simulations. And finally, it is worth pointing out that
the solutions provided here are approximate solutions.

Acknowledgments

The authors would like to thank Mehran Pooladi-Darvish


(Fekete Associates Inc. and The University of Calgary) for fruitful
discussions. We also acknowledge the financial support of the
National Iranian Oil Company (NIOC) Research and Development
during the course of this study.

Fig. 9. Dimensionless shape factor comparison for second and third order trial Appendix A. Solution of nonlinear diffusion equation
solution (base case).
Integral methods can be used to derive a closed form approxi-
mate solution to nonlinear diffusion equations [13,15,37]. In this
study heat integral method (early time solution) and method of
moments (late time solution) are used to solve the diffusion equa-
tion for compressible fluids in the fractured porous media.

A.1. Early time solution

Integral method is used to derive the early time solution of the


nonlinear diffusion equation as follows:
 
@wD @ @w
¼ bgD ðtÞ D ðA1:1Þ
@t D @xD @xD
tD ¼ 0 ! wD ¼ 0 ðA1:2Þ
xD ¼ 1 ! wD ¼ 1 ðA1:3Þ
@wD @ 2 wD
xD ¼ dðt D Þ ! wD ¼ 0; ¼ 0; ¼0 ðA1:4Þ
@xD @x2D

We use the following trial solution in the integral method:

Fig. 10. Comparison of the cumulative fluid exchanges between the matrix and the  3
1  xD
fracture obtained by approximate semi-analytical (2nd- and 3rd order) and wD ¼ 1  ðA1:5Þ
numerical solutions (base case). 1  dðt D Þ

Using this trial solution in the integral form of Eq. (A1.1) gives:
Z 1  3 ! Z 1  
Combination of the integral method and the method of moments d 1  xD d @w
1 dxD ¼ bgD ðtÞ D dxD
are used to solve the nonlinear diffusion equation resulting from d dtD 1  dðt D Þ d dxD @xD
the flow of a compressible gas in a dual-porosity system. The
ðA1:6Þ
approximate semi-analytical solution is validated with fine-grid
numerical simulations. The matrix–fracture shape factor is then Integrating Eq. (A1.6) and some simplification leads to the following
back calculated using the developed solution. ordinary differential equation:
Similar to the flow of a slightly compressible fluid, the shape
factor demonstrates a transient behavior and then converges to a ð1  dÞdð1  dÞ ¼ 12bgD1 dt D ; t D ¼ 0 ! dð0Þ ¼ 1 ðA1:7Þ
constant value during the pseudo steady-state period for a com-
pressible fluid. The average value of the stabilized dimensionless Solving this ordinary differential equation leads to the following
shape factor for compressible gas flow for one set of fractures un- equation for d:
der a constant fracture boundary condition was found to be 8.5 as pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
compared to p2 for a slightly compressible fluid. Results reveal that d¼1 24bgD1 tD ðA1:8Þ
the stabilized value of the matrix–fracture transfer shape factor for
a compressible fluid is a weak function of the temperature and the Since d 6 1 we have following equation for d:
gas specific gravity. On the other hand, pressure variation shows a pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
d¼1 24bgD1 tD ðA1:9Þ
negligible effect on the stabilized value of the shape factor.
Two approximate semi-analytical models for the shape factor Early time solution is valid till d = 0, so
calculation are presented by suggesting second- and third-order pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1
trial solutions and using the integral method and the method of 1 24bgD1 t D ¼ 0 ! t D ¼ ðA1:10Þ
24bgD1
moments. Both approximate semi-analytical models were verified
using a fine-grid, single-porosity, numerical simulations. The So the final solution for the early time has the following form:
E. Ranjbar, H. Hassanzadeh / Advances in Water Resources 34 (2011) 627–639 637

 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 3 !3
xD þ 24bgD1 t D  1 dC
1  xD 1 ¼ 12bgD1 C þ 48bgD1 D ðA2:13Þ
wD ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ 1  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; tD < dt D
24bgD1 tD 24bgD1 tD 24bgD1 t D 24bgD1
dD 40 140
ðA1:11Þ ¼  bgD1 C  bgD1 D ðA2:14Þ
dt D 3 3
The average dimensionless pseudo-pressure is calculated as Eigenvalues of this system of ODEs is obtained by making the fol-
follows: lowing determinant equal to zero [21,28]:
Z 1 Z 1
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 24bgD1 t D 1 12bgD1  k 48bgD1
wD ¼ wD dV ¼ wD dxD ¼ ; tD < ¼ 0 ! k2 þ 34:667bgD1 k þ 80b2 g2D1 ¼ 0
V d d 4 24bgD1  40 bgD1  140 bgD1  k
3 3
ðA1:12Þ ðA2:15Þ
Eq. (A2.15) has the following roots:
A.2. Late time solution
2:486gD1 b and  32:181gD1 b ðA2:16Þ
The late time solution of the nonlinear partial differential equa-
For the sake of simplicity, we take:
tion (PDE) using the method of moments is explained in more de-
tail in this part k1 ¼ 2:486gD1 ðA2:17Þ
  k2 ¼ 32:181gD1 ðA2:18Þ
@wD @ @w
¼ bgD ðtÞ D ðA2:1Þ
@t D @xD @xD The corresponding eigenvectors are then obtained; and, finally, the
1 following equations for unknown coefficients of C and D are
tD ¼ ! wD ¼ x3D ðA2:2Þ
24bgD1 derived.
@w
xD ¼ 0 ! D ¼ 0 ðA2:3Þ C ¼ 3:314m1 expðbk1 t D Þ  1:086m2 expðbk2 t D Þ ðA2:19Þ
@xD
D ¼ m1 expðbk1 t D Þ þ m2 expðbk2 t D Þ ðA2:20Þ
xD ¼ 1 ! wD ¼ 1 ðA2:4Þ
At the initial condition of tD ¼ 24b1g we have wD ¼ x3D . Therefore m1
The method of moments is used to find the late time solution of D1
and m2 are obtained by solving the following system of equations:
this PDE by suggesting a third order trial solution and the residual
(R) as follows: 0 ¼ 3:314m1 expð0:104Þ  1:086m2 expð1:341Þ
ðA2:21Þ
1 1 ¼ m1 expð0:104Þ þ m2 expð1:341Þ
wD ðxD ; tD Þ ¼ Aðt D Þ þ Bðt D ÞxD þ CðtD Þx2D þ DðtD Þx3D ; tD >
24bgD1 So we have the following values for m1 and m2:
ðA2:5Þ
m1 ¼ 0:541
  ðA2:22Þ
@w @ @w m2 ¼ 5:686
R¼ D bgD D ðA2:6Þ
@tD @xD @xD Therefore, the trial solution of the nonlinear PDE for the late time
Unknown coefficients (A, B, C and D) in Eq. (A2.5) are found using behavior is obtained as follows:
boundary conditions and making the zero and first moments of R wD ðxD ; tD Þ ¼ ð1  1:252 expðbk1 t D Þ þ 0:489 expðbk2 t D ÞÞ
vanish by enforcing the following conditions:
Z 1  Z 1   þ ð1:793 expðbk1 t D Þ  6:175 expðbk2 t D ÞÞx2D
@wD @ @w
RdxD ¼ 0 !  bgD ðtÞ D dxD ¼ 0 ðA2:7Þ þ ð0:541 expðbk1 tD Þ þ 5:686 expðbk2 t D ÞÞx3D ;
0 0 @tD @xD @xD
Z 1 Z 1    1
@wD @ @w tD P ðA2:23Þ
xD RdxD ¼ 0 ! xD  bgD D dxD ¼ 0 ðA2:8Þ 24bgD1
0 0 @t D @x D @xD
From the first boundary conditions, we can conclude that B = 0; and, Eq. (A2.23) is integrated over the matrix block volume to obtain the
the second boundary condition (Eq. (A2.4)) leads to: average matrix block pseudo pressure as follows:

A¼1CD ðA2:9Þ 1
wD ¼ 1  0:790 expðbk1 tD Þ  0:148 expðbk2 t D Þ; tD P
24bgD1
Solving Eqs. (A2.7) and (A2.8) and combining the results with
ðA2:24Þ
Eq. (A2.9) leads to a system of ordinary differential equations as
follows:

2 dC 3 dD Appendix B. Derivation of the cumulative fluid exchange


  ¼ 2bgD1 C þ 3bgD1 D ðA2:10Þ equation
3 dt D 4 dt D

1 dC 3 dD In this appendix, the derivation of Eq. (42), which was used to


  ¼ bgD1 C þ 2bgD1 D ðA2:11Þ compare the approximate semi-analytical model with the numer-
4 dt D 10 dt D
ical model, is presented in more detail. The equation for the ma-
It should be noted that gD1 is the dimensionless hydraulic diffusiv- trix–fracture transfer function discussed in Section 3 is derived
ity of the fracture at the outer boundary condition, which is defined as follows (Eq. (9)):
as follows:
T sc V b km r
gðx ¼ 1Þ 1 k qsc ¼ ðwm  wf Þ ðB:1Þ
gD1 ¼ gD @ xD ¼ 1; gD1 ¼ D ¼ ðA2:12Þ 4psc T
g g lf c f u
In Eq. (B.1), r is the shape factor in m2, T is the reservoir tem-
After some simplification, we reach to the following system of or- perature in K, (wm  wf Þ is the difference between the fracture
dinary differential equations (ODEs): pseudo pressure and the average matrix block pseudo pressure in
638 E. Ranjbar, H. Hassanzadeh / Advances in Water Resources 34 (2011) 627–639

kPa/s, and Tsc and psc are the standard condition temperature and 100
pressure in K and kPa, respectively.
Slightly Compressible Model

Dimensionless Shape Factor (σhm2)


The dimensionless shape factor and the pseudo pressure differ-
Compressible Model
ence as a function of dimensionless time are given as follows,
respectively:

4bgD1 1:964 expðbk1 t D Þ þ 4:763 expðbk2 t D Þ


rh2m ¼ ðB:2Þ
gD 0:790 expðbk1 tD Þ þ 0:148 expðbk2 tD Þ
10

wm  wi
wD ¼ ¼ 1  0:790 expðbk1 tD Þ  0:148 expðbk2 t D Þ ðB:3Þ
wf  wi

wm  wf ¼ ðwf  wi ÞðwD  1Þ
¼ ðwf  wi Þð0:790 expðbk1 tD Þ  0:148 expðbk2 t D ÞÞ ðB:4Þ
1
After substituting Eqs. (B.2) and (B.4) into Eq. (B.1), the follow- 0.001 0.01 0.1 1 10
ing equation for the matrix–fracture transfer function can be Dimensionless Time
obtained:
Fig. C1. Compressible shape factor model applied to a slightly compressible fluid as
T sc km V b bgD1 compared with the literature data [9,16].
qsc ¼ ðwi  wf Þð1:964 expðbk1 t D Þ þ 4:763 expðbk2 tD ÞÞ
psc Th2m gD
ðB:5Þ
For a slightly compressible fluid (b = gD1 = gD = 1) one should re-
Changing dimensionless time to real time using the definition of the cover the late time shape factor (p2). Therefore, in Eq. (41) when tD
dimensionless time in Eq. (B.5) leads to: goes to infinity the resulting value of the shape factor for a slightly
! !! compressible fluid is 9.9443, which is close to p2(9.8696).
T sc km V b bgD1 k1 g k2 g
qsc ¼ ðw i  w f Þ 1:964 exp bt þ 4:763 exp bt
psc Th2m gD L2c L2c
ðB:6Þ References
The cumulative fluid exchange can be calculated by the integration
[1] Agarwal RG. Real gas pseudo-time – a new function for pressure build-up
of Eq. (B.6) from zero to t as follows: analysis of MHF gas wells. SPE paper 8279; 1979.
Z t
[2] Aguilera R. Naturally fractured reservoirs. Tulsa, Oklahoma: PennWell Press;
1995.
Q¼ qðsÞds ðB:7Þ [3] Ames WF. Nonlinear partial differential equations in engineering. New
0
York: Academic Press; 1965.
8h i h i9 [4] Barenblatt GE, Zheltov IP, Kochina IN. Basic concepts in the theory of seepage
>
< 1:964 exp kL12g bt þ 4:763
exp k2 g
bt >
= of homogeneous liquids in fissured rocks (strata). J Appl Math Mech
T sc km V b gD1 k1 k2 L2c
1960;20:852–64.
Q SC ¼ ðwi  wf Þ h ic
4psc T ggD >
:  1:964 þ 4:763 >
; [5] Beckner BL. Improved modeling of imbibition matrix/fracture fluid transfer in
k1 k2
double porosity simulators. PhD dissertation, Stanford University; 1990.
ðB:8Þ [6] Bourbiaux B, Granet S, Landereau P, Noetinger B, Sarda S, Sabathier JC. Scaling
up matrix–fracture transfer in dual-porosity models: theory and application.
where Q is the total fluid exchange from the matrix in standard cu- SPE paper 56557; 1999.
[7] Bourbiaux B, Basquet R, Cacas MC, Daniel JM, Sarda S. An integrated workflow
bic meters and t is the real time in seconds.
to account for multi-scale fractures in reservoir simulation models:
implementation and benefits. SPE paper 78489; 2002.
[8] Carslaw HS, Jaeger JC. Conduction of heat in solids. London: Oxford University
Appendix C. Comparison of shape factors for compressible and Press; 1959.
slightly compressible fluids [9] Chang MM. Analytical solution to single and two-phase flow problems of
naturally fractured reservoirs: theoretical shape factor and transfer functions.
PhD dissertation, University of Tulsa; 1995.
The objective of this Appendix is to show that the developed [10] Civan F, Rasmussen ML. Analytical hindered-matrix–fracture transfer models
model recovers the shape factor for a slightly compressible fluid. for naturally fractured petroleum reservoirs. Paper SPE 74364; 2002.
[11] Coats KH. Implicit compositional simulation of single-porosity and dual-
Shape factor equations of a compressible fluid are given by Eqs.
porosity reservoirs. SPE paper 18427; 1989.
(32) and (41). Applying these two equations for a slightly com- [12] Crank J. The mathematics of diffusion. Oxford: Clarendon Press; 1975.
pressible fluid (b = gD1 = gD = 1) leads to the following equations [13] Finlayson BA. The method of weighted residuals and variational
principles. New York: Academic Press; 1972.
for the early and late time shape factors:
[14] Geo-Quest. Eclipse 100 technical descriptions 2009.1. Geo-Quest,
pffiffiffi   Schlumberger; 2009.
2 4 6 1 1 1
r hm ¼ pffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffi ; tD < ¼ 0:042 ðC:1Þ [15] Goodman TR. Application of integral methods to transient nonlinear heat
1 4  24tD tD 24 transfer. Advances in heat transfer, vol. 1. San Diego, CA: Academic; 1964. p.
51–122.
[16] Hassanzadeh H, Pooladi-Darvish M. Effects of fracture boundary conditions on
1:964 expð2:486tD Þ þ 4:763 expð32:181tD Þ
rh2m ¼ 4 ; matrix–fracture transfer shape factor. Transp Porous Med 2006;64:51–71.
0:790 expð2:486t D Þ þ 0:148 expð32:181tD Þ [17] Hassanzadeh H, Pooladi-Darvish M, Atabay S. Shape factor in the drawdown
solution for well testing of dual-porosity systems. Adv Water Res
1 2009;32:1652–63.
tD P ¼ 0:042 ðC:2Þ
24 [18] Ikoku CU. Natural gas reservoir engineering. Florida: Krieger Press; 1992.
[19] Kazemi H, Gilman JR. Multiphase flow in fractured petroleum reservoirs. In:
Fig. C1 is a comparison of the dimensionless shape factor for a Bear J, Tsang CF, de Marsily G, editors. Flow and contaminant transport in
slightly compressible fluid based on our model and other models fractured rock. San Diego: Academic Press; 1993. p. 267–323.
in the literature [9,16]. Results show that the developed model [20] Kazemi H, Merrill LS, Porterfield KL, Zeman PR. Numerical simulation of
water–oil flow in naturally fractured reservoirs. Soc Pet Eng J 1976:317–26.
could recover the behavior of the shape factor of a slightly com- [21] Kreyszig E. Advanced engineering mathematics. United States: John Wiley &
pressible fluid for the entire period of time. Sons Press; 1999.
E. Ranjbar, H. Hassanzadeh / Advances in Water Resources 34 (2011) 627–639 639

[22] Lim KT, Aziz K. Matrix–fracture transfer shape factors for dual-porosity [31] Sarma P. New transfer functions for simulation of naturally fractured
simulators. J Pet Sci Eng 1995;13:169–78. reservoirs with dual-porosity models. MSc thesis, Stanford University; 2003.
[23] Lu M, Connel LD. A dual-porosity model for gas reservoir flow incorporating [32] Thomas LK, Dixon TN, Pierson RG. Fractured reservoir simulation. Soc Pet Eng J
adsorption behavior – Part I. Theoretical development and asymptotic 1983:42–54.
analysis. Transp Porous Med 2007;68:153–73. [33] Ueda Y, Murata S, Watanabe Y, Fanatsu K. Investigation of the shape factor
[24] Mattar L, Brar GS, Aziz K. Compressibility of natural gases. J Can Pet Technol used in the dual-porosity reservoir simulator. SPE paper 19469; 1989.
1975;14:77–80. [34] van Heel APG, Boerrigter PM, van Drop JJ. Thermal and hydraulic matrix–
[25] McCain WD. The properties of petroleum fluids. Tulsa, Oklahoma: PennWell fracture interaction in dual permeability simulation. Soc Pet Eng J
Press; 1989. 2008:735–49.
[26] Ordonez A, Penuela G, Idrobo EA, Medina CE. Recent advances in naturally [35] van Heel APG, van Drop JJ, Boerrigter PM. Heavy oil recovery by steam
fractured reservoir modeling. CT&F Ciencia, Tecnologia y Futuro injection in fractured reservoirs. SPE paper 113461; 2008.
2001;2:51–64. [36] Warren JE, Root PJ. The behavior of naturally fractured reservoirs. Soc Pet Eng J
[27] Penuela G, Civan F, Hughes RG, Wiggins ML. Time-dependent shape factors for 1963:245–55.
interporosity flow in naturally fractured gas-condensate reservoirs. SPE paper [37] Zimmerman RW, Bodvarsson GS. An approximate solution for one-
75524; 2002. dimensional absorption in unsaturated porous media. Water Resour Res
[28] Polyanin AD. Handbook of linear partial differential equations for engineers 1989;25(6):1422–8.
and scientists. United States: CRC Press; 2001. [38] Zimmerman RW, Chen G, Hadgu T, Bodvarsson GS. A numerical dual-porosity
[29] Quintard M, Whitaker S. Transport in chemically and mechanically model with semi-analytical treatment of fracture/matrix flow. Water Resour
heterogeneous porous media. I: Theoretical development of region-averaged Res 1993;29(7):2127–37.
equations for slightly compressible single-phase flow. Adv Water Res [39] Zimmerman RW, Hadgu T, Bodvarsson GS. A new lumped-parameter model
1996;19(1):29–47. for flow in unsaturated dual-porosity media. Adv Water Res
[30] Sarda S, Jeannin L, Basquet R, Bourbiaux B. Hydraulic characterization of 1996;19(5):317–27.
fractured reservoirs: Simulation on discrete fracture models. Soc Pet Eng J
2002:154–62.

You might also like