Practical Solutions for Pressure-Transient

Responses of Fractured Horizontal Wells in

Unconventional Shale Reservoirs
M. Brown, SPE, and E. Ozkan, SPE, Colorado School of Mines; R. Raghavan, SPE, Phillips Petroleum (Retired);
and H. Kazemi, SPE, Colorado School of Mines

Summary of data. For pressure-transient-analysis purposes, analytical models

This paper presents an analytical trilinear-flow solution to simulate are preferred if they can represent sufficient details of fluid and
the pressure-transient and production behaviors of fractured hori- reservoir characteristics.
zontal wells in unconventional shale reservoirs (Ozkan et al. 2009). The trilinear-flow model couples linear flows in three contigu-
The model is simple, but versatile enough to incorporate the fun- ous flow regions, as sketched in Fig. 1: the outer reservoir beyond
damental petrophysical characteristics of shale reservoirs, includ- the tips of the hydraulic fractures (denoted by the subscript O), the
ing the intrinsic properties of the matrix and the natural fractures. inner reservoir between hydraulic fractures (denoted by the sub-
Special characteristics of fluid exchange among various reservoir script I), and the hydraulic fracture (denoted by the subscript F).
components may also be considered. Computational convenience Each region can have distinct properties. The inner reservoir may
of the trilinear-flow solution makes it a practical alternative to be homogeneous or naturally fractured, and the hydraulic fractures
more rigorous but computationally intensive and time-consuming may have finite conductivity. To derive a practical model, some
solutions. Another advantage of the trilinear-flow solution is the idealizations and simplifying assumptions are made. The model is
convenience in deriving asymptotic approximations that provide derived for single-phase flow of a constant-compressibility fluid.
insight about potential flow regimes and the conditions leading to Single-phase gas flow is handled through pseudopressure transfor-
these flow regimes. Though linear- and bilinear-flow regimes have mation. Flow from the reservoir into the horizontal well is only
been noted for fractured horizontal wells in the literature on the by virtue of the hydraulic fractures; that is, the production directly
basis of their diagnostic features, they have not been associated from the surface of the horizontal well is assumed to be negligible
with particular reservoir characteristics and flow relationships. The (because the surface area of the horizontal well is negligible
trilinear-flow solution also provides a suitable algorithm for the with respect to the total surface area of the hydraulic fractures).
regression analysis of pressure-transient tests in shale reservoirs. Hydraulic fractures are assumed to have identical properties and
are equally spaced by a distance, dF , along the horizontal well.
Introduction This assumption can be removed by an approach similar to that
Production of horizontal wells from unconventional shale reservoirs used by Raghavan et al. (1997); however, creating equally spaced
is a result of flow in shale matrix, in the reservoir fracture network hydraulic fractures of similar properties is a common field prac-
(referred to as natural fractures in this paper), and in hydraulic tice and, unless there is significant contrast in individual-fracture
fractures (assumed as biwing transverse fractures in this study). properties, the use of the average fracture properties should be
Although it is possible to develop detailed analytical (Chen and sufficiently accurate. Moreover, our ability to discern individual-
Raghavan 1997; Raghavan et al. 1997) and numerical (Medeiros fracture properties from pressure-transient analysis is moot.
et al. 2008) models to represent transient fluid flow toward a mul- The SRV around the horizontal well is envisioned like a natu-
tiple-fractured horizontal well in tight, unconventional reservoirs rally fractured porous medium and idealized as a dual-porosity
such as shale, the downside of these models includes the increased region. In this paper, we use two common models of dual-porosity
computational requirements, the implicit functional relationships of idealization. We refer to the model that considers pseudosteady
key parameters, and the inconvenience in their use in iterative appli- fluid transfer from matrix to fracture (Warren and Root 1963) as
cations. If, however, the natural-fracture network is a result of shear the pseudosteady model. The model that incorporates transient
slippages because of hydraulic fracturing and localized around the fluid transfer from matrix to fracture (Kazemi 1969; de Swaan O
horizontal well, then some simplifications are possible. 1976; Serra et al. 1983) is referred to as the transient model. The
For hydraulically fractured horizontal wells in shale reservoirs subscripts m and f are used to denote the matrix and fracture media
with matrix permeabilities in the range of microdarcies or below, of the dual-porosity reservoir. The choice of the particular dual-
the contribution of the reservoir beyond the stimulated reservoir porosity model does not affect the general solution for the trilinear-
volume (SRV) is usually negligible (Medeiros et al. 2008; Mayer- flow model, but it does affect the definition of key parameters.
hofer et al. 2010). Despite the complex interplay of flow among Each hydraulic fracture is considered to be a finite-conductiv-
matrix, natural fractures, and hydraulic fractures, the key character- ity porous medium. 1D linear flow, akin to flow in vertical-well
istics of flow convergence toward a multiply-fractured horizontal fractures, is assumed in the hydraulic fractures because wellbore
well within the SRV may be preserved in the trilinear-flow model storage normally masks the very-early-time radial-flow conver-
presented by Ozkan et al. (2009). The model is not applicable if gence toward the well within the hydraulic fractures (Soliman
the regions beyond the well tips dominate the well response. This et al. 1990; Mukherjee and Economides 1991; Larsen and Hegre
condition is not likely to exist, however, unless the shale-matrix 1991, 1994); however, the effect of radial-flow convergence at
permeability is well above the microdarcy range or the bottomhole the fracture/horizontal-well intersection is taken into account by
pressure or rate is unrealistically low. Other than that, the trilinear a flow-choking skin. The wellbore-storage effect is incorporated
model can be used whenever the use of an analytical model is into the model by convolution.
warranted by the complexity of the problem and the availability Additionally, flow in the inner reservoir between hydraulic
fractures and flow in the outer reservoir beyond the tips of the
hydraulic fractures are both assumed to be linear. Linear flow in
the inner reservoir is a result of assuming either a nonperforated
Copyright © 2011 Society of Petroleum Engineers
horizontal well between fractures or that the hydraulic fractures
This paper (SPE 125043) was accepted for presentation at the SPE Annual Technical dominate production. The bisector of the distance between two
Conference and Exhibition, New Orleans, 4–7 October 2009, and revised for publication.
Original manuscript received for review 2 July 2009. Revised manuscript received for
hydraulic fractures is a no-flow boundary because of the assump-
review 15 July 2011. Paper peer approved 8 August 2011. tion of identical hydraulic fractures (Fig. 1).

December 2011 SPE Reservoir Evaluation & Engineering 663

Fig. 1—Schematic of the trilinear-flow model representing three contiguous flow regions for a multiple-fractured horizontal well.

For most common applications in unconventional shale-gas the medium  = m (matrix) or f (fracture), then the bulk properties
reservoirs, the outer reservoir does not contribute to production of the medium  are defined by
significantly. If it contributes, however, its contribution is akin to
linear flow toward a finite-conductivity fracture. As discussed by   = r   . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (1)
Raghavan et al. (1997) and Chen and Raghavan (1997), at long
times, a multiple-fractured horizontal well behaves like a single In Eq. 1, r = V Vb , where V is the bulk volume of the medium
fracture between the two outermost fractures along the horizontal  and Vb is the bulk volume of the total system (Vb = Vm+Vf). It
well. Therefore, flow from the outer reservoir is mainly in the is customary to introduce the pseudosteady dual-porosity model
direction perpendicular to flow in the inner reservoir. Unless the (Warren and Root 1963) in terms of the bulk properties. Charac-
horizontal well is short, minimum allowable flow rates are reached teristics of the matrix and fracture media are incorporated into the
under this flow regime. pseudosteady dual-porosity medium by the storativity and flow-
The single-porosity results of the trilinear-flow model are the capacity ratios defined, respectively, by
same as the early-time solutions of Chen and Raghavan (1997).
We verify the trilinear-flow model by comparing the results with
( )  ct f
=
the semianalytical solution of Medeiros et al. (2008). A field ( ) ( )
 ct +  ct
f
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (2)
m
example demonstrates the versatility of the solution in investigat-
ing the effects of various parameters influencing pressure-transient and
responses.
k
 =  C 2 m . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (3)
Definitions kf
Under the conditions assumed in this paper, the pressure-transient
response of a horizontal well with nF identical transverse hydraulic In Eq. 3, C is a reference length that is chosen in this paper as the
fractures can be modeled by considering one of the fractures pro- half-length of the hydraulic fracture; that is, C = x F (in comparing
ducing from a rectangular reservoir section at a rate equal to qF = the results of this study with other models, adjustments should be
q/nF , where q is the total flow rate of the horizontal well (Fig. 2). made to the values of  to compensate for different choices of
As sketched in Figs. 1 and 2, the fracture is located centrally in the reference length).
closed rectangular drainage area of size 2xe × 2ye, which is equal to The definition of the matrix shape factor, , in Eq. 3 is an
1 / nF of the total drainage area of the horizontal well. The fracture important issue, but this discussion is outside the scope of the
has a half-length of xF and width of wF and penetrates the entire current paper. Kazemi et al. (1976) proposed the following first
thickness, h, of the formation. approximation for the shape factor of rectangular matrix blocks:
For convenience, we derive the trilinear-flow solution in terms
of dimensionless variables. Here, we first explain the definitions of ⎛ 1 1 1⎞
 = 4 ⎜ 2 + 2 + 2 ⎟ , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (4)
the parameters used in the pseudosteady and transient dual-poros- ⎝ L x L y Lz ⎠
ity models. We, then, present the definitions of the dimensionless
variables used in the trilinear-flow model. where Lx,Ly , and Lz are the dimensions of the matrix block. For
the purposes of this paper, we will assume square matrix blocks of
Dual-Porosity Parameters. In dual-porosity idealizations, a natu- side-length L and use  = 12/L2 to compute the shape factor.
rally fractured reservoir may be characterized in terms of intrinsic The transient dual-porosity model (Kazemi 1969; de Swaan
or bulk (equivalent) properties. If  denotes an intrinsic property of O 1976; Serra et al. 1983) may be introduced in terms of either
intrinsic or bulk properties of the fracture and matrix media. In
this paper, we will follow the Serra et al. (1983) derivation of the
transient dual-porosity model in terms of the intrinsic properties.
wF The transient dual-porosity model represents the naturally frac-
dF ye tured reservoir as a stack of alternating matrix and fracture slabs.
h
If hf and hm denote the thickness of each fracture and matrix slab,
Symmetry respectively, and nf and nm are the number of fracture and matrix
element
slabs, respectively, then the total thicknesses of fracture and matrix
slabs are hft = nfhf and hmt = nmhm, respectively. Thus, the formation
2xe
xF xe thickness h = hft + hmt.
xF Serra et al. (1983) define the storativity and flow-capacity ratios
for the transient dual-porosity model, respectively, by
2nF ye

(ct )m
= . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (5)
Fig. 2—Multiple-fractured horizontal well and the symmetry ele-
ment used in the derivation of the trilinear-flow model.
(ct ) f

664 December 2011 SPE Reservoir Evaluation & Engineering

and Note that, regardless of the choice of the dual-porosity model, CFD
and CRD are defined on the basis of the bulk permeability of the inner
⎛ C ⎞⎛k h ⎞
2
⎛ C ⎞⎛k h ⎞ 2
reservoir, kI . We also define the following diffusivity ratios:
 = 12 ⎜ 2 ⎟ ⎜ m mt ⎟ = 12 ⎜ 2 ⎟ ⎜ m m ⎟ . . . . . . . . . . . . . . . . (6)
⎝ hm ⎠ ⎝ k f h ft ⎠ ⎝ hm ⎠ ⎝ k f h f ⎠
F
Note that choosing the matrix slab thickness, hm, in the transient FD = . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (15)
I
model equal to the side-length of the square matrix blocks, L, in
the pseudosteady model, yields the same value of the flow-capacity and
ratio, , for both dual-porosity models. We will use this equivalence
in the comparison of the results from both dual-porosity models. O
OD = , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (16)
I
Dimensionless Variables. We define the dimensionless pressure
and time, respectively, by where I is the diffusivity of the inner reservoir (Eq. 9) and F and
O are the diffusivities of the hydraulic fracture and outer reservoir
kI hI
pD =
141.2qF B
( pi − p) . . . . . . . . . . . . . . . . . . . . . . . . . . . (7) given, respectively, by

kF
and F = 2.637 × 10 −4 . . . . . . . . . . . . . . . . . . . . . . . . . . (17)
(ct )F 
I
tD = t, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (8) and
x F2
where kO
O = 2.637 × 10 −4 . . . . . . . . . . . . . . . . . . . . . . . . . (18)
kI −4
(ct )O 
I = 2.637 × 10 . . . . . . . . . . . . . . . . . . . . . . . . . . . (9)
(ct )I  For gas flow, F and O are evaluated at the highest pressure
For analysis of gas wells, we follow the usual approach and incor- during the test (Al-Hussainy et al. 1966). For horizontal wells
porate the liquid-flow analogy of Al-Hussainy et al. (1966) through producing from shale, as a first approximation, the outer-reservoir
the pseudopressure transformation: properties may be taken to be the same as the properties of the
inner-reservoir matrix.
p
p′
m ( p) = 2 ∫ dp′. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (10) Mathematical Model
pb
z
Our analytical derivation of the trilinear-flow model follows the
Time transformations such as those given in Agarwal (1979) or same lines as Cinco-Ley and Meng (1988) who presented the finite-
Scott (1979) may be used to analyze buildup tests. No such trans- conductivity-fracture solution in a dual-porosity reservoir. As noted
formations should be used for drawdown tests. in the Definitions section, we consider one-quarter of a hydraulic
In Eqs. 7 and 9, and in the rest of the definitions given here, fracture in a rectangular drainage region (Figs. 1 and 2). We derive
the subscript I refers to the property of the inner reservoir. For a the solutions for the outer reservoir, inner reservoir, and the hydraulic
homogeneous (single-porosity) inner reservoir, the definitions are fracture, and then couple the solutions by using the flux- and pres-
based on the homogeneous matrix properties. If the inner reservoir sure-continuity conditions on the interfaces between the regions. It
is naturally fractured and the transient dual-porosity model is used, is more convenient to derive the solution in the Laplace-transform
kI, hI, and (ct)I refer, respectively, to the intrinsic fracture perme- domain because we consider the possibility that the inner reservoir
ability (kI = kf), total fracture thickness (hI = hft = nfhf), and the may be naturally fractured. In this work, we used the algorithm pro-
intrinsic fracture storativity [(ct)I = (ct)f]. For the pseudosteady posed by Stehfest (1970) to numerically invert the results computed
dual-porosity model, kI is the bulk fracture permeability (kI = k f ), in the Laplace-transform domain into the time domain.
hI is the formation thickness (hI = h), and (ct)I is the total system
⎡ ( ) ( )
storativity ⎢⎣(ct )I =  ct f +  ct m ⎥⎦. It must be noted that the
⎤ Outer Reservoir Solution. Assuming 1D flow in the x-direction,
definition of the dimensionless pressure is the same for both dual- the diffusivity equation and the associated boundary conditions for
(
porosity models because k f h ft = k f h ft h h = k f h. ) the outer reservoir are given by
The dimensionless distances in the x- and y-direction are
defined, respectively, by ∂ 2 pOD s
− p = 0 , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (19)
∂x D2 OD OD
x
xD = . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (11)
xF
⎛ ∂pOD ⎞
and ⎜⎝ ∂x ⎟⎠ = 0, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (20)
D x D = xeD
y
yD = . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (12) and
xF
The dimensionless distances to the reservoir boundaries are given pOD = pID , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (21)
x D =1 x D =1
by xeD and yeD.
In our model, we use the following definitions of dimensionless
fracture and reservoir conductivities, respectively: where the overbar symbol indicates dimensionless pressure in the
Laplace-transform domain and s is the Laplace-transform param-
k F wF eter with respect to dimensionless time, tD. The solution of the
CFD = . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (13)
kI x F boundary-value problem in Eqs. 19 through 21 can be obtained in
the Laplace-transform domain as follows:
and
cosh ⎡⎣ s OD ( xeD − x D ) ⎤⎦
kI x F pOD = pID . . . . . . . . . . . . . . . (22)
C RD =
kO ye
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (14) x D =1
cosh ⎡⎣ s OD ( xeD − 1) ⎤⎦

December 2011 SPE Reservoir Evaluation & Engineering 665

In Eq. 22, the outer-reservoir solution, pOD, is given in terms of From Eq. 32, we can substitute
the inner-reservoir pressure at the interface of the inner and outer
reservoirs, pID x =1. ∂pID
= −
F pFD
D
. . . . . . . . . . . . . . . . . . . . . . . . (34)
Inner-Reservoir Solution. Flow in the inner reservoir is assumed ∂yD yD = w D 2
yD = w D 2

to be 1D and in the y-direction perpendicular to the hydraulic-

fracture plane. The associated partial-differential equation is given and write Eq. 33 as follows:
by
∂ 2 pFD
∂ 2 pID ⎛ 1 ⎞ ∂pOD −  F pFD = 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (35)
+⎜ − upID = 0, . . . . . . . . . . . . . . . (23) ∂x D2
∂yD2 ⎝ yeDC RD ⎟⎠ ∂x D x D =1
In Eqs. 34 and 35,
where
⎡ ⎛ w ⎞⎤
u = sf ( s ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (24)
F = O tanh ⎢ O ⎜ yeD − D ⎟ ⎥ . . . . . . . . . . . . . . . . . . . . (36)
⎣ ⎝ 2 ⎠⎦
and and
⎧
⎪1 for homogeneous inner region 2
F s
⎪ F = + . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (37)
⎪ s (1 −  ) +  CFD FD
f (s) = ⎨ for pseudosteady dual-porosity inneer reservoir
⎪ s (1−  ) + 
⎪ The boundary conditions of the hydraulic-fracture problem
⎪1 +  ( 3s ) tanh
⎩
( 3 s  ) for transient dual-porosity inner reservoir
are given by
. . . . . . . . . . . . . . . . . . . . . . . (25)
∂pFD
= 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (38)
Substituting from Eq. 22, ∂x D x D =1

∂pOD and
= −
O pID , . . . . . . . . . . . . . . . . . . . . . . . . . . . (26)
∂x D x D =1
x D =1
∂pFD 
=− . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (39)
where ∂x D xD =0
CFD s

Then, the dimensionless pressure solution for the hydraulic frac-

O = s OD tanh ⎡⎣ s OD ( xeD − 1) ⎤⎦, . . . . . . . . . . . . . . . . (27) ture is obtained as

and assuming pID ≠ f ( x D ), Eq. 23 becomes  cosh ⎡⎣  F (1 − x D ) ⎤⎦

pFD = . . . . . . . . . . . . . . . . . (40)
∂ 2 pID
CFD s  F sinh  F ( )
− O pID = 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (28)
∂yD2
Wellbore-Pressure Solution. The dimensionless wellbore pres-
with sure is obtained from Eq. 40 at xD = 0 and is given by

O 
O = + u . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (29) pwD = pFD ( x D = 0 ) = . . . . . . . . . . . . . (41)
C RD yeD CFD s  F tanh ( F )
The boundary conditions for the inner reservoir are given by In the preceding derivations, we have assumed 1D (linear) flow
within the hydraulic fracture; that is, we have ignored the radial
⎛ ∂pID ⎞ convergence of flow toward the wellbore within the hydraulic
⎜⎝ ∂y ⎟⎠ = 0, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (30) fracture. Mukherjee and Economides (1991) provided the follow-
D yD = yeD
ing equation to compute the skin factor caused by flow choking
and within the fracture:

pID = pFD . . . . . . . . . . . . . . . . . . . . . . . . . . . . (31) kI hI ⎡ ⎛ h ⎞ ⎤

yD = w D 2 yD = w D 2 sc = ⎢ ln ⎜ ⎟ − ⎥ . . . . . . . . . . . . . . . . . . . . . . . . . . (42)
k F wF ⎣ ⎝ 2rw ⎠ 2 ⎦
Thus, the solution of Eq. 28, with the boundary conditions given
by Eqs. 30 and 31, is Adding the choking skin to Eq. 41, we obtain the following solu-
tion, which is a good approximation for dimensionless wellbore
pressure after the end of the radial flow in the hydraulic fracture:
cosh ⎡⎣ O ( yeD − yD ) ⎤⎦
pID = pFD ( yD = w D 2 ) ⎡ ⎛ w ⎞⎤
, . . . . . . . . . . . . (32)
pwD =

+
sc
. . . . . . . . . . . . . . . . . . . . . (43)
cosh ⎢ O ⎜ yeD − D ⎟ ⎥
⎣ ⎝ 2 ⎠⎦ CFD s  tanh ( ) s

To develop a more practical solution, the effect of wellbore

which depends on the hydraulic-fracture solution, pFD .
yD = w D 2 storage should be taken into account at early times. The effect of
wellbore storage can be incorporated into the solution by substitut-
Hydraulic-Fracture Solution. The partial-differential equation
ing pwD from Eq. 43 into the following convolution expression in
describing 1D flow in the hydraulic fracture is given by
the Laplace domain:

∂ 2 pFD 2 ∂pID s pwD

+ − p = 0. . . . . . . . . . . . . . . . (33) pwD ,storage = , . . . . . . . . . . . . . . . . . . . . . . . . . . . (44)
∂x D2 CFD ∂yD yD = w D 2
FD FD 1 + C D s 2 pwD

666 December 2011 SPE Reservoir Evaluation & Engineering

 TABLE 1—DATA USED IN THE COMPARISON OF TRILINEAR- AND SEMIANALYTICAL-MODEL RESULTS
Well, Reservoir, and Fluid Data Model Parameters
(Intrinsic Properties) (Pseudosteady Dual-Porosity Model)
~
Formation thickness, h (ft) 100 Matrix bulk permeability, k m (md) 0.01
~
Wellbore radius, rw (ft) 0.3 Matrix bulk porosity, φm 0.1
~ −1
Reservoir size in x-direction, xe (ft) 1,325 Matrix bulk compressibility, c tm (psi ) 2.3 E–04
~
Reservoir size in y-direction, ye (ft) 608 Fracture bulk permeability, k f (md) 1
~
Viscosity, µ (cp) 0 .0 2 Fracture bulk porosity, φ f 4.5 E–04
~ −1
Matrix permeability, km (md) 0.01 Fracture bulk compressibility, c tf (psi ) 3 E–04
Matrix porosity, φm 0.1 Dimensionless fracture conductivity, CFD 0 .2
−1
Matrix compressibility, ctm (psi ) 2.03 E–04 Dimensionless reservoir conductivity, CRD 82.237
−2
Matrix-block thickness, hm (ft) 10 Outer-reservoir diffusivity, ηO (ft /hour) 2.46 E+04
Fracture permeability, kf (md) 1, 00 0 Inner-reservoir diffusivity, ηI (ft−2/hour) 2.45 E+06
Fracture porosity, φ f 0.45 Hydraulic-fracture diffusivity, ηF (ft−2/hour) 2.5 E+09
−1
Fracture compressibility, ctf (psi ) 3 E–01 Outer-reservoir diffusivity ratio, ηOD 1 . 0 2 E+03
Fracture thickness, hf (ft) 0. 0 1 Hydraulic-fracture diffusivity ratio, ηFD 1.01 E–02
Number of fractures, nf = number of matrix
10 Dimensionless hydraulic-fracture width, wFD 2 E–04
blocks, nm
Matrix-block dimensions, Lx = Ly = Lz (ft) 10 Dimensionless reservoir size in x-direction, xeD 2.65
Hydraulic-fracture permeability, kF (md) 1,000 Dimensionless reservoir size in y-direction, yeD 1.216
Hydraulic-fracture porosity, φF 0.1 Flow-choking skin, sc 3.545
−1 −2
Hydraulic-fracture compressibility, ctF (psi ) 2 E–04 Matrix shape factor, σ (ft ) 1.2 E–01
Hydraulic-fracture half-length, xF (ft) 500 Storativity ratio, ω 6.61 E–03
Hydraulic- fracture width, wF (ft) 0.1 Flow-capacity ratio, λ 3 E+02
Wellbore-storage coefficient in Fig. 3, C (bbl/psi) 0 Dimensionless wellbore-storage coefficient, CD (Fig. 3) 0
Wellbore-storage coefficient in Fig. 4, C (bbl/psi) 1.05 E–02 Dimensionless wellbore-storage coefficient, CD (Fig. 4) 1.84 E–05

where the dimensionless wellbore-storage coefficient, CD, is given solutions for fractured vertical wells in layered reservoirs. For a
by homogeneous reservoir, substituting f(s) = 1 and u = s in Eq. 43,
the trilinear-flow model yields the same asymptotic solutions at
5.615C early times as the methods of Chen and Raghavan (1997) and
CD = . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (45)
2 (cht )I x F2 Raghavan et al. (1997). This provides analytical verification of
our trilinear-flow model.
The definition of CD depends on the dual-porosity model selected We also verify the trilinear-flow solution by comparing the
to represent the inner reservoir and, for the same value of the results with the semianalytical solution of Medeiros et al. (2008).
wellbore-storage coefficient, C, a different dimensionless well- The semianalytical solution models hydraulic fractures as porous
bore-storage coefficient, CD, is obtained for the pseudosteady and media and considers radial-flow convergence toward the wellbore
transient dual-porosity models. within the hydraulic fractures. Therefore, the semianalytical solu-
We also use the solution given by Eq. 43 (or Eq. 44) to compute tion is expected to capture the characteristics of the early-time
dimensionless buildup pressures from the following superposition flow regimes in hydraulic fractures (fracture radial- and radial/
equation: linear-flow regimes). However, the accuracy of the computations
at very early times is hindered by gridding and discretization of
( ) (
L ⎡⎣ psD ( t D ) ⎤⎦ = L ⎡⎣ pwD t pD ⎤⎦ − L ⎡⎣ pwD t pD + t D ⎤⎦
,
) . . . . . (46)
the boundaries. Similarly, the trilinear-flow model incorporates
the effect of flow choking in hydraulic fractures through a skin
+ L ⎡⎣ pwD ( t D ) ⎤⎦ factor and, thus, becomes accurate after the end of the early-time
radial flow in hydraulic fractures. Considering the limitations of
where the trilinear and semianalytical models, differences are expected
in the results of the two models at early times. However, when the
psD =
kI hI
141.2qF B
( )
pws − pwf ,s . . . . . . . . . . . . . . . . . . . . . . . (47) effect of wellbore storage is taken into account, the early-time flow
regimes are masked and the limitations of the models at early times
become insignificant for practical purposes. In the following, we
In Eq. 46, tpD is the dimensionless producing time, ΔtD is the will first present the comparison of the results from the trilinear and
dimensionless shut-in time, and L[pD(tD)] denotes the Laplace semianalytical models without the effect of wellbore storage. Then,
transform of pD(tD) with respect to tD. In Eq. 47, pws is the bot- we will demonstrate the effect of wellbore storage on early-time
tomhole shut-in pressure and pwf,s is the bottomhole pressure at results. Table 1 presents the data used in the comparison. Inspec-
the instant of shut-in. tion of Table 1 should reveal the details of the physical system that
can be considered in the trilinear-flow model.
Model Verification Fig. 3 shows the results of the trilinear and semianalytical
Chen and Raghavan (1997) and Raghavan et al. (1997) devel- (Medeiros et al. 2008) models without the effect of wellbore storage
oped analytical pressure-transient solutions for multiple-fractured (CD = 0). Pseudosteady dual-porosity behavior has been assumed
horizontal wells in homogeneous reservoirs by using the ideas of for the inner reservoir. The data given in Table 1 have been used for
Bennett et al. (1985) and Camacho-V. et al. (1987), who developed both models, with the exception of an infinite-acting outer reservoir

December 2011 SPE Reservoir Evaluation & Engineering 667

 1 E+05

DIMENSIONLESS DERIVATIVE, dpwD /dlntD

1 E+04

DIMENSIONLESS PRESSURE, pwD

1 E+03

1 E+02

1 E+01

1 E+00
PRESSURE (Trilinear)
DERIVATIVE (Trilinear)
1 E–01 PRESSURE (Semianalytical)
DERIVATIVE (Semianalytical)
1 E–02
1 E–03 1 E–02 1 E–01 1 E+00 1 E+01 1 E+02 1 E+03 1 E+04 1 E+05
DIMENSIONLESS TIME, tD

Fig. 3—Comparison of the trilinear- and semianalytical-model (Medeiros et al. 2008) results without wellbore storage.

in the semianalytical model). Because the trilinear model assumes In Fig. 4, we consider the same case as in Fig. 3, but we add
that the reservoir is limited to the SRV, the trilinear-model responses the results for a wellbore-storage coefficient of C = 0.0105 bbl/psi.
in Fig. 3 show a unit-slope behavior caused by the closed outer As a result of this relatively small wellbore-storage effect, most of
boundary at late times. The semianalytical-model responses fall the hydraulic-fracture flow regimes are masked. This discussion,
slightly below the trilinear-model responses at late times because however, is intended for rigor and completeness; otherwise, for
of the weak support of the outer reservoir beyond the SRV. a reasonable wellbore-storage effect, the differences displayed
The differences in the results of the two models shown in Fig. between the early-time responses of the trilinear and semianalytical
3 at early times are also to be expected and may be explained as models should disappear within a few seconds for most practical
follows. The semianalytical model displays the characteristics of applications (for the particular case in Fig. 4, tD = 10−3 corresponds
radial/linear flow (Larsen and Hegre 1991, 1994; Al-Kobaisi et al. approximately to 30 seconds). We must also note that the apparent
2006) by a constant derivative trend at early times. The derivative wellbore-storage effect on the early-time behavior of the semiana-
responses of the trilinear-flow model, on the other hand, display lytical model is a numerical artifact.
the characteristics of linear flow in hydraulic fractures. Except at
very early times, the trilinear model matches the pressure responses Impact of Choice in Dual-Porosity Models
of the semianalytical model, indicating that the effect of the early- Although the transient and pseudosteady dual-porosity models are
time fracture radial flow is incorporated into the pressure responses known to have differences at intermediate times, the ramifications
of the trilinear-flow model accurately enough by assuming a chok- of the choice in dual-porosity models are sometimes overlooked
ing skin, as defined in Eq. 42. (especially when using standard numerical packages). Using the

1 E+05
PRESSURE (Trilinear)
DERIVATIVE (Trilinear)
DIMENSIONLESS DERIVATIVE, dpwD /dlntD

1 E+04 PRESSURE Wellbore Storage (Trilinear)

DIMENSIONLESS PRESSURE, pwD

DERIVATIVE Wellbore Storage (Trilinear)

PRESSURE (Semianalytical)
1 E+03 DERIVATIVE (Semianalytical)

1 E+02

1 E+01

1 E+00

1 E–01

1 E–02
1 E–06 1 E–04 1 E–02 1 E+00 1 E+02 1 E+04
DIMENSIONLESS TIME, t D

Fig. 4—Comparison of the trilinear- and semianalytical-model (Medeiros et al. 2008) results with wellbore storage.

668 December 2011 SPE Reservoir Evaluation & Engineering

 TABLE 2—DATA USED IN THE COMPARISON OF DUAL-POROSITY MODELS
Well, Reservoir, and Fluid Data
(Intrinsic Properties) Model Parameters
~
Formation thickness, h (ft) 250 Matrix bulk permeability, k m (md) 1 E–06
~
Wellbore radius, rw (ft) 0.3 Matrix bulk porosity, φ m 0.05
~ −1
Reservoir size in x-direction, xe (ft) 250 Matrix bulk compressibility, ctm (psi ) 1 E–06
~
Reservoir size in y-direction, ye (ft) 250 Fracture bulk permeability, k f (md) 0.16
~
Viscosity, µ (cp) 0.0184 Fracture bulk porosity, φ f 3.6 E–05
~ −1
Matrix permeability, km (md) 1 E–06 Fracture bulk compressibility, ctf (psi ) 8 E–11
Matrix porosity, φ m 0.05 Dimensionless fracture conductivity, CFD 2.5 E+02
−1
Matrix compressibility, ctm (psi ) 1 E–06 Dimensionless reservoir conductivity, CRD 1.6 E+05
12.5 Outer-reservoir diffusivity, ηO (ft /hour)
−2
Matrix-block thickness, hm (ft) 1. 09 E + 0 3
−2
Fracture permeability, kf (md) 2,000 Inner-reservoir diffusivity for pseudosteady fluid transfer, ηI (ft /hour) 1.74 E+08
Fracture porosity, φ f
−2
0.45 Inner-reservoir diffusivity for transient fluid transfer, ηI (ft /hour) 2.42 E+11
−1 −2
Fracture compressibility, ctf (psi ) 1 E–06 Hydraulic-fracture diffusivity, ηF (ft /hour) 1. 43 E + 14
Fracture thickness, hf (ft) 0.001 Outer-reservoir diffusivity ratio, ηOD (pseudosteady fluid transfer) 6.25 E–06
Number of fractures, nf = number of
20 Outer-reservoir diffusivity ratio, ηOD (transient fluid transfer) 4.5 E–09
matrix blocks, nm
Matrix-block dimensions, Lx = Ly = Lz (ft) 12.5 Hydraulic-fracture diffusivity ratio, ηFD (pseudosteady fluid transfer) 8.22 E+05
Hydraulic-fracture permeability, kF (md) 1 E+06 Hydraulic-fracture diffusivity ratio, ηFD (transient fluid transfer) 5.92 E+02
Hydraulic-fracture porosity, φ F 0.38 Dimensionless hydraulic-fracture width, wFD 4 E– 05
Hydraulic-fracture compressibility, ctF
−1 1 E–06 Dimensionless reservoir size in x-direction, xeD 1
(psi )
Hydraulic-fracture half-length, xF (ft) 250 Dimensionless reservoir size in y-direction, yeD 1
Hydraulic-fracture width, wF (ft) 0.01 Flow-choking skin, sc 0.018
Wellbore-storage coefficient in Fig. 5, C −2
0 Matrix shape factor, σ (ft ) 7.68 E–02
(bbl/psi)
Wellbore-storage coefficient in Fig. 6, C
0.0105 Storativity ratio, ω (pseudosteady fluid transfer) 5.76 E–08
(bbl/psi)
Storativity ratio, ω (transient fluid transfer) 1.39 E+03
Flow-capacity ratio, λ 3 E–02
Dimensionless wellbore-storage coefficient, CD (Fig. 5) 0
Dimensionless wellbore-storage coefficient, CD (pseudosteady model,
1.2 E–02
Fig. 6)
Dimensionless wellbore-storage coefficient, CD (transient model, Fig. 6) 1.67 E+01

data given in Table 2 and assuming there is no wellbore storage, in instantaneous stabilization of flow in the matrix as soon as the
Fig. 5, we compare the trilinear-flow results for the pseudosteady fluid transfer from matrix to fracture begins. The transient dual-
and transient dual-porosity models. We plot the pseudosteady dual- porosity idealization, on the other hand, takes into account the
porosity responses as a function of t D and the transient dual-poros- transient-flow conditions before pseudosteady (stabilized) flow
ity responses as a function of t D (1+  ) to make the comparison is established in the matrix. The duration of the transient-flow
at the same real time (the definitions of dimensionless time for period is determined by the permeability, size, and the geometry
the pseudosteady and transient dual-porosity models use the total- of the matrix blocks. If the matrix is tight and the fracture density
system properties and intrinsic fracture properties, respectively). is small, then an extended period of transient fluid transfer from
Although the responses for t D ≤ 10 −4 are not of practical interest, matrix to fractures is possible. Under these conditions, the dip in
they are included in Fig. 5 to display the expected dual-porosity the derivative responses is either delayed (if pseudosteady state
characteristics at early, intermediate, and late times. is established in the matrix before pseudosteady state develops
Fig. 5 shows that the pseudosteady and transient dual-porosity in the total system) or masked by the reservoir boundary effects
models yield the same responses at early times, when the flow in the (in this case, the matrix and fracture systems reach pseudosteady
naturally fractured medium is dominated by the fractures, and at late state concurrently). For small matrix blocks and relatively high
time, when the responses are dominated by the total system. The inter- permeabilities, which are more common in conventional tight gas
mediate-time responses from the two models, however, are completely reservoirs, the transient-flow period in the matrix is shorter and the
different. The dip in the derivative responses, which is usually taken differences between the results of the two dual-porosity idealiza-
as an indication of a naturally fractured reservoir, is not displayed by tions may be insignificant for practical purposes.
the transient dual-porosity idealization. This result deserves scrutiny The differences between the two dual-porosity idealizations in
because the contribution of the matrix flow (or the lack of it) is an Fig. 5 indicate that instantaneous pseudosteady state in the matrix
important discussion in the development of shale plays. is not a good assumption for the dual-porosity shale reservoir con-
The appearance of a dip in the derivative responses of dual- sidered in this figure. In general, because of the very low matrix
porosity models requires two conditions: considerable flow from permeability, the transient dual-porosity model is more appropriate
matrix to fracture network and pseudosteady (stabilized) flow in for shale reservoirs. Under these conditions, the lack of the charac-
the matrix. The pseudosteady dual-porosity idealization assumes teristic dip in the pressure derivative responses should not be taken

December 2011 SPE Reservoir Evaluation & Engineering 669

 1 E+05

1 E+04

DIMENSIONLESS DERIVATIVE, dp wD/dlntD

DIMENSIONLESS PRESSURE, p wD
1 E+03

1 E+02

1 E+01

1 E+00

1 E–01
PRESSURE (Pseudosteady)
1 E–02
DERIVATIVE ((Pseudosteady)
PRESSURE (Transient)
1 E–03
DERIVATIVE (Transient)
1 E–04
1 E–08 1 E–06 1 E–04 1 E–02 1 E+00 1 E+02 1 E+04
DIMENSIONLESS TIME, tD (Pseudosteady) OR tD/(1+ω ) (Transient)

Fig. 5—Comparison of the pseudosteady and transient dual-porosity models for naturally fractured inner reservoir.

as an indication of negligible shale-matrix contribution. In Fig. 5, Asymptotic Approximations and Flow Regimes
the deviation of the transient dual-porosity responses from the early- Asymptotic approximations for the trilinear-flow solution may be
time, natural-fracture-dominated flow behavior is the only indication obtained by considering the limiting forms of Eq. 43 for large and
of the start of the matrix contribution. For the transient dual-porosity small values of the Laplace transform parameter s (the limiting forms
case considered in Fig. 5, the matrix does not reach pseudosteady for large and small values of s represent the early- and late-time
state before pseudosteady state is established in the total system behavior of the solution, respectively). Depending on the relative
(the unit-slope derivative behavior at late times). Therefore, the values of s, , and , approximate forms of f(s) given in Eq. 25 are
characteristic dip in the derivative responses is not shown (masked substituted into Eq. 43. Suitable approximations of f(s) for pseudos-
by the reservoir depletion under pseudosteady state). In Fig. 6, teady and transient dual-porosity models are given, respectively, by
we use the same data that were used in Fig. 5 (Table 2), but we
consider a wellbore-storage coefficient of C = 0.0105 bbl/psi. The ⎧1 as s → 0
responses for the pseudosteady and transient dual-porosity models ⎪
are considerably different. More importantly, the results shown in f ( s ) = ⎨ ⎡⎣ s (1 −  ) ⎤⎦ for intermediate values of s . . . . . . (48)
Fig. 6 indicate that some characteristics of the pressure and deriva- ⎪ as s → ∞
⎩
tive responses may be attributed to inaccurate flow regimes because
of the inappropriate choice of the dual-porosity model. and

1 E+05
PRESSURE Wellbore Storage (Transient)
DERIVATIVE Wellbore Storage (Transient)
DIMENSIONLESS DERIVATIVE, dp wD /dlntD

1 E+04 PRESSURE Wellbore Storage (Pseudosteady)

DERIVATIVE Wellbore Storage (Pseudosteady)
DIMENSIONLESS PRESSURE, p wD

PRESSURE (Transient)
1 E+03 DERIVATIVE (Transient)
PRESSURE (Pseudosteady)
DERIVATIVE (Pseudosteady)
1 E+02

1 E+01

1 E+00

1 E–01

1 E–02
1 E–02 1 E–01 1 E+00 1 E+01 1 E+02 1 E+03 1 E+04 1 E+05 1 E+06
DIMENSIONLESS TIME, tD/CD (Pseudosteady) OR t D /(1+ω )/CD (Transient)

Fig. 6—Effect of wellbore storage on pseudosteady and transient dual-porosity models for naturally fractured inner reservoir.

670 December 2011 SPE Reservoir Evaluation & Engineering

 TABLE 3A—ASYMPTOTIC SOLUTIONS FOR PRESSURE AND DERIVATIVE:
PSEUDOSTEADY DUAL-POROSITY MODEL

Slope of straight line on log pwD vs. log Asymptotic solution for Slope of straight line
Asymptotic solution for pwD ′ = dpwD d log t D ′ vs. log t D
tD pwD on log pwD
Early Time
tD tD
1 pwD = 1 ′ =
pwD 1
CD CD
Intermediate Time
π (1 − ω ) π yeD π
2 pwD = + + + sc 0 p′wD = 0 → −∞
2 yeD λ 6 3CFD

π ⎛1− ω ⎞ + s
14

3 pwD = ⎜ ⎟ c 0 ′ =0
pwD → −∞
2CFD ⎝ λ ⎠

π 1− ω π → −∞
4 p wD = + + sc 0 p′wD = 0
2 λ 3CFD
π t 1D 4 1 π t1D4 1
5 pwD = + sc → if sc → 0 ′ =
pwD
2CFD Γ (5 4 ) 4 4 2CFD Γ (5 4 ) 4
π t 1D 4 1 π t1D 4 1
6 pwD = + sc → if sc → 0 ′ =
pwD
2CFD Γ (5 4 ) ω1 4 4 4 2CFD Γ (5 4 ) ω1 4 4

π tD π 1 1 π tD 1
7 pwD = + + sc → if sc → 0 and CFD → ∞ ′ =
pwD
ω 3CFD 2 2 ω 2
π 1 1 1
8 pwD = π tD + + sc → if sc → 0 and CFD → ∞ ′ =
pwD π tD
3CFD 2 2 2
Late Time ( t AD = t D xF2 A )

π yeD π yeD 1 sc
9 pwD = 2π t AD + + + sc → 1 when t AD >> + + ′ = 2π t AD
pwD 1
6 3CFD 12 6CFD 2π

⎧1 +  as s → 0 of approximate solutions; the objective is merely to draw attention

⎪ to the multiple intermediate-time approximations that are possible
f ( s ) = ⎨  ( 3s ) for intermediate values of s . . . . . . . . . (49) with the same diagnostic slope values on log-log plots of pressure
⎪1 as s → ∞ and derivative responses. From Tables 3A and 3B, we see that
⎩
identical traces are produced by myriad causes at intermediate
Similarly, tanh ( ) may be replaced by the following approxima- times and it is not possible to attribute a characteristic to a par-
tions in deriving asymptotic approximations to Eq. 43: ticular cause. Then, the use of the analysis techniques based on
the slope characteristics of straight lines is not warranted because
⎧ as  → 0 not only the slope, but also the value, of the straight line should be
⎪ matched with an asymptotic relationship. Therefore, although non-
tanh ( ) = ⎨ −  3 3 for intermediate values of  . . . . . . . (50) uniqueness is an issue because of the large number of regression
⎪1 as  → ∞ parameters, matching the entire data span with a model response
⎩
(regression analysis) usually converges to a better solution (or
It is possible to derive several asymptotic approximations diverges altogether). After an initial match is obtained, the results
for the trilinear-flow solution by considering the relationships may be refined by straight-line analysis using the asymptotic solu-
in Eqs. 48 through 50. In Tables 3A and 3B, we present a list tions given in Tables 3A and 3B.
of the asymptotic approximations that can be derived for early,
intermediate, and late times. The existence and applicability of the Field Example
asymptotic relationships with respect to field data depend on the Here, we consider a field example that demonstrates the use of the
complex interplay of a large number of parameters controlling the results and discussions presented in the previous sections. In this
pressure-transient responses. The early- and late-time asymptotic example, we demonstrate the analysis of a common type of data
solutions in Tables 3A and 3B are typical of the wellbore-stor- in shale-gas wells, which consists of the daily record of pressure
age-dominated flow and boundary-dominated (pseudosteady-state) and production rates.
flow in naturally fractured reservoirs, respectively. Multiple inter- The data used in this example are from a multiple-fractured
mediate-time solutions in Tables 3A and 3B are a result of possible horizontal gas well in the Barnett shale and are taken from Ander-
combinations of  and  corresponding to different conditions of son et al. (2010). Fig. 7 shows the pressure and gas-production-
fluid transfer from matrix to natural fractures. (The conditions for rate data for more than 3 years. The other data that were available
fluid transfer from matrix to natural fractures are imposed not only for this example are given in Table 4. We used rate-normalized
by the contrast between the matrix and fracture properties, but also pseudopressures and material-balance time in the analysis. We
the geometrical aspects of the matrix and fracture media and the computed the material-balance time, tmb, as explained by Medeiros
general reservoir flow characteristics). et al. (2010).
The information in Tables 3A and 3B is only a partial list of the Fig. 8 shows the matching of the field data with the trilinear
asymptotic approximations that may be derived. Other approxima- model. Because of the low matrix permeability, we used the
tions may be derived analytically, graphically, or empirically. Our transient dual-porosity assumption in the analysis. The derivative
objective in this paper, however, is not to derive an exhaustive list responses display half- and unit-slope behaviors at early and late

December 2011 SPE Reservoir Evaluation & Engineering 671

 TABLE 3B—ASYMPTOTIC SOLUTIONS FOR PRESSURE AND DERIVATIVE:
TRANSIENT DUAL-POROSITY MODEL

Slope of straight line on log pwD vs. Asymptotic solution for Slope of straight
Asymptotic solution for pwD ′ = dpwD d log t D
log tD pwD line on
′ vs. log t D
log pwD
Early Time
tD tD
1 pwD = 1 ′ =
pwD 1
CD CD
Intermediate Time
π π
18 18

= ⎛⎜ 9 9 ⎞⎟ ′ = ⎛⎜ 9 9 ⎞⎟
3 t 1D8 1 3 t D1 8 1
2 pwD + sc → if sc → 0 pwD
⎝λ ω ⎠ 2CFD Γ (9 8 ) 8 ⎝ λ ω ⎠ 8 2CFD Γ (9 8) 8

π t 1D 4 1 π t1D4 1
3 pwD = + sc → if sc → 0 ′ =
pwD
2CFD Γ (5 4 ) 4 4 2CFD Γ (5 4 ) 4
π t1D 4 1 π t 1D 4 1
4 pwD = + sc → if sc → 0 ′ =
pwD
2CFD Γ (5 4 )(1 + ω ) 4 2CFD Γ (5 4 )(1 + ω ) c
14 14
4 4

π ⎛ 3t D ⎞ + π + s π
14 14
1
′ = ⎛ 3t D ⎞ 1
5 pwD = ⎜ ⎟ → if sc → 0 and CFD → ∞ pwD ⎜ ⎟
2Γ (5 4 ) ⎝ λω ⎠ 8Γ (5 4 ) ⎝ λω ⎠
c
3CFD 4 4

π 1 1 1
6 pwD = π tD + + sc → if sc → 0 and CFD → ∞ ′ =
pwD π tD
3CFD 2 2 2
1 3π tD π yeD π 1 3π t D 1
7 pwD = + + + sc ′ =
pwD
yeD λω 6 3CFD 2 yeD λω 2

π tD π 1 1 π tD 1
8 p wD = + + sc → if sc → 0 and CFD → ∞ pwD =
1 + ω 3CFD 2 2 1+ ω 2
Late Time ( t AD = t D xF2 A )

2π t AD π yeD π t AD yeD 1 sc 2π t AD
9 pwD = + + + sc → 1 when >> + + pwD = 1
1+ ω 6 3CFD 1+ ω 12 6CFD 2π 1+ ω

times, respectively. Pseudopressure responses, on the other hand, permeability of 10−4 md. The dual-porosity representation of the
do not display any characteristic flow behavior (they remain flat) SRV used in the trilinear model yields a matrix permeability of
at early times. This is a common behavior in shale-gas wells and is 10−6 md and an intrinsic natural-fracture permeability of 2×104 md
usually attributed to liquid loading in the wellbore. To account for (corresponding to a bulk natural-fracture permeability of 1.2 md).
this effect, we used a constant skin factor in the analysis. Overall, Although it is not possible to quantitatively compare the perme-
the match is satisfactory, and Table 5 shows the results of the anal- ability estimates from the single- and dual-porosity idealizations of
ysis. Anderson et al. (2010) analyzed these data by representing the the SRV, a qualitative evaluation is consistent with the expectation
SRV around the well as a single-porosity medium with an assumed that the matrix permeability in the dual-porosity model should be

3,500 10000

Pressure 9000
3,000
Flow Rate
8000

2,500 7000
Flow Rate, Mscf/d
Pressure, psia

6000
2,000
5000
1,500
4000

1,000 3000

2000
500
1000

0 0
0 200 400 600 800 1,000 1,200 1,400
Time, d

Fig. 7—Pressure and production data.

672 December 2011 SPE Reservoir Evaluation & Engineering

 TABLE 4—INPUT DATA Mscf) and the data in Table 5 yields an estimate of the stimulated
reservoir area as 61 acres. This is higher than the area of the
Initial pressure, pi (psia) 3,109 SRV estimated in Fig. 8 (41 acres). A similar difference appears
Formation thickness, h (ft) 300 between the stimulated-reservoir-area estimates from the late-time
Formation temperature, T (°F) 106
flowing-material-balance (40 acres) and early-time linear-flow (71
acres) analyses reported by Anderson et al. (2010). On the basis
Wellbore radius, rw (ft) 0.23 of our discussions in this paper, however, such differences should
Matrix porosity, φm (fraction) 0.04 be expected.
Specific gravity, γ 0.588
Horizontal-well length, L (ft) 3,250 Conclusions
Fracture half-length, xF (ft) 275
In this paper, we have presented a practical analytical model and
discussed the analysis of pressure-transient responses of multiple-
Number of hydraulic fractures, nF 19 fractured horizontal wells in unconventional shale reservoirs. The
Gas saturation, Sg (fraction) 0.9 following conclusions are warranted from the work presented in
Water saturation, Sw (fraction) 0.1 this paper:
1. The key characteristics of flow convergence toward a multiple-
fractured horizontal well in a shale reservoir are preserved in a
smaller than that in the single-porosity approach of Anderson et al. relatively simple, trilinear-flow model. The trilinear-flow model
(2010). If the fractured medium is evaluated as a homogeneous is a practical alternative to rigorous numerical or semianalytical
system, then the existence of a natural-fracture network provides models.
a higher effective permeability for the total system. The area of 2. The dual-porosity model with transient fluid transfer from the
the SRV (L × 2xf = 1.79 × 106 ft2 = 41 acres) used for the match matrix to the natural fractures is more appropriate for low matrix
in Fig. 8 is consistent with the estimate of Anderson et al. (2010) permeabilities encountered in unconventional shale reservoirs.
from flowing-material-balance analysis (40 acres). Pressure-derivative responses of shale wells may not display the
Because the early- and late-time straight lines on the derivative dip that is usually taken as an indication of a naturally fractured
responses in Fig. 8 indicate the existence of linear- and pseudoste- reservoir. This, however, does not indicate that the contribution
ady-flow behaviors, respectively, we have also analyzed the data by of shale matrix is negligible.
using the straight-line technique. We first plotted the pseudopres- 3. It is possible to derive asymptotic approximations that indicate
sure responses as a function of the square root of material-balance multiple straight-line relationships for the pressure and deriva-
time in Fig. 9. At early times (between 5 and 15 days1/2), a straight tive responses of multiple-fractured horizontal wells. However,
line with slope 7,733 psi2-D1/2/cp-Mscf can be drawn through the there are multiple possibilities for the diagnostic slope relation-
data. From the expressions given in Table 3B, there are three pos- ships commonly identified on pressure and derivative responses
sibilities for a linear-flow behavior (Expressions 6, 7, and 8). The of shale wells, such as one half or one fourth. The same
use of the results in Table 5 yields consistent results only with diagnostic behavior may be a result of different fluid-transfer
Expression 6. [We used the slope relationship for Expression 6 mechanisms between matrix and natural fractures, and thus may
to calculate the fracture half-length. The calculated fracture half- correspond to different straight-line relationships for the same
length was xF = 225 ft. Anderson et al. (2010) compute the fracture time ranges.
half-length as 275 ft from their linear-flow analysis.] 4. Because multiple straight-line relationships may display the
As also indicated by the unit-slope derivative behavior in Fig. 8, same diagnostic features in the same time interval, not only the
the Cartesian plot of the pseudopressure vs. material-balance time slope but also the value of the straight line should be matched
shown in Fig. 10 displays a straight line at late times, indicating with an asymptotic relationship to analyze the well-test data.
pseudosteady-state behavior. The use of Expression 9 in Table Therefore, techniques based on the slopes of characteristic
3B with the slope of the straight line in Fig. 10 (210 psi2/cp- straight lines may not be warranted. Although nonuniqueness is

1 E+07

Pseudopressure Drop (Data)

Logarithmic Derivative (Data)
Pseudopressure Drop (Model)
Logarithmic Derivative (Model)
Δm(pwf)/qsc, psi 2-d/cp-Mscf

1 E+06
Δm(pwf)/qsc)/dln tmb

1 E+05
Half-slope line
d(Δ

Unit-slope line
1 E+04

1 E+03
1 E+00 1 E+01 1 E+02 1 E+03 1 E+04
tmb , d

Fig. 8—Matching of the field data with the trilinear model.

December 2011 SPE Reservoir Evaluation & Engineering 673

 TABLE 5—ANALYSIS RESULTS FOR EXAMPLE 2 (MATCHING WITH THE TRILINEAR MODEL)
Well, Reservoir, and Fluid Data Model Parameters
(Intrinsic Properties) (Transient Fluid Transfer From Matrix to Fractures)
~
Formation thickness, h (ft) 300 Matrix bulk permeability, km (md) 1 E–06
~
Wellbore radius, rw (ft) 0.23 Matrix bulk porosity, φm 0.036
~ –1
Distance to SRV boundary in x-direction, xe (ft) 275 Matrix bulk compressibility, c tm (psi ) 3.17 E–04
~
Half-distance between hydraulic fractures, ye (ft) 90.3 Fracture bulk permeability, k f (md) 1.2
~
Viscosity, (cp) 0.02 Fracture bulk porosity, φ f 4.8 E–06
~ –1
Matrix permeability, km (md) 1 E–06 Fracture bulk compressibility, c tf (psi ) 1.9 E–8
Matrix porosity, φm 0.036 Dimensionless fracture conductivity, CFD 3.03
–1
Matrix compressibility, ctm (psi ) 3.17 E–04 Dimensionless reservoir conductivity, CRD 3.66 E+06
–1 –2
Outer-reservoir compressibility, ctO (psi ) 3.17 E–04 Outer-reservoir diffusivity, O (ft /hour) 1.39 E–05
–2
Fracture permeability, kf (md) 2 E+04 Inner-reservoir diffusivity, I (ft /hour) 1.04 E+07
–2
Fracture porosity, φf 0.08 Hydraulic-fracture diffusivity, F (ft /hour) 1.09 E+07
–1
Fracture compressibility, ctf (psi ) 3.17 E–04 Outer-reservoir diffusivity ratio, OD 1.33 E–02
Fracture thickness, hf (ft) 0.0005 Hydraulic-fracture diffusivity ratio, FD 1.05
Number of fractures, nf = number of matrix
36 Dimensionless hydraulic fracture width, wFD 3.64 E–05
blocks, nm
Matrix-block dimensions, Lx = Ly = hm (ft) 8.33 Dimensionless reservoir size in x-direction, xeD 1
Hydraulic-fracture permeability, kF (md) 1 E+05 Dimensionless reservoir size in y-direction, yeD 0.33
Hydraulic-fracture porosity, φF 0.38 Flow-choking skin, sc 1.77
–1
Hydraulic-fracture compressibility, ctF (psi ) 3.17 E–04 Storativity ratio, ω 6.25 E+03
Hydraulic-fracture half-length, xF (ft) 275 Flow-capacity ratio, λ 1.09 E–02
Hydraulic-fracture width, wF (ft) 0.01
Wellbore-storage coefficient, C (bbl/psi) 0.0187
Skin due to liquid loading, s 300

an issue, matching the entire data with a model response usually ct = total compressibility, psi−1
converges to a better solution. After an initial match is obtained, ctf = total natural-fracture compressibility, psi−1
the results may be refined by straight-line analysis. ctF = total hydraulic compressibility, psi−1
ctm = total matrix compressibility, psi−1
Nomenclature ctI = total compressibility of the inner reservoir, psi−1
A = drainage area, ft2 ctO = total compressibility of the outer reservoir, psi−1
B = oil formation volume factor, bbl/STB C = wellbore-storage coefficient, bbl/psi

9 E+05

8 E+05

7 E+05
Δm(pwf)/qsc, psi 2-d/cp-Mscf

6 E+05

5 E+05

4 E+05

3 E+05

2 E+05 Slope = 7733 psi 2-d 1/2/cp-Mscf

1 E+05

0 E+00
0 10 20 30 40 50 60
1/2 , d 1/2
t mb

Fig. 9—Linear-flow analysis.

674 December 2011 SPE Reservoir Evaluation & Engineering

 9 E+05

8 E+05

7 E+05

Δm(pwf )/qsc, psi 2-d/cp-Mscf

6 E+05

5 E+05 Slope = 210 psi 2/cp-Mscf

4 E+05

3 E+05

2 E+05

1 E+05

0 E+00
0 500 1000 1500 2000 2500 3000 3500 4000
tmb, d

Fig. 10—Pseudosteady-flow analysis.

CD = dimensionless wellbore-storage coefficient wFD = dimensionless hydraulic-fracture width

CFD = dimensionless hydraulic-fracture conductivity x = x-coordinate, ft
CRD = dimensionless reservoir conductivity xF = hydraulic-fracture half-length, ft
d = distance between two fractures, ft xe = reservoir size in x-direction, ft
D = distance between outermost fractures, ft xeD = dimensionless reservoir size in x-direction
f = dual-porosity transfer function y = y-coordinate, ft
h = reservoir thickness, ft ye = reservoir size in y-direction, ft
hf = thickness of natural fractures, ft yeD = dimensionless reservoir size in y-direction
hft = total thickness of natural fractures, ft F = hydraulic-fracture parameter used in trilinear-flow model
hmt = total thickness of matrix slabs, ft O = outer-reservoir parameter used in trilinear-flow model
hm = thickness of matrix slabs, ft
F = hydraulic-fracture parameter used in trilinear-flow model
kf = natural-fracture intrinsic permeability, md
O = outer-reservoir parameter used in trilinear-flow model
k f = natural-fracture bulk permeability, md f = natural-fracture diffusivity, ft2/hr
kF = hydraulic-fracture permeability, md F = hydraulic-fracture diffusivity, ft2/hr
kI = permeability of the inner reservoir, md FD = hydraulic-fracture-diffusivity ratio, ft2/hr
kO = permeability of the outer reservoir, md I = inner-reservoir diffusivity, ft2/hr
km = matrix intrinsic permeability, md O = outer-reservoir diffusivity, ft2/hr
km = matrix bulk permeability, md OD = outer-reservoir diffusivity ratio, ft2/hr
C = characteristic length in the system, ft  = flow-capacity ratio
L = horizontal-well length, ft  = fluid viscosity, cp
Li = matrix-block dimension in i-direction, ft  = matrix shape factor, ft−2
m = pseudopressure, psi2/cp f = natural-fracture intrinsic porosity, fraction
nf = number of natural fractures  f = natural-fracture bulk porosity, fraction
nF = number of hydraulic fractures F = hydraulic-fracture porosity, fraction
nm = number of matrix blocks I = inner-reservoir porosity, fraction
pb = base pressure, psi m = matrix intrinsic porosity, fraction
pi = initial reservoir pressure, psi  m = matrix bulk porosity, fraction
pwf = flowing wellbore pressure, psi O = outer-reservoir porosity, fraction
q = production rate, STB/D, Mscf/D  = storativity ratio
rw = wellbore radius, ft
sc = horizontal-well flow-choking skin factor
t = time, hours, days Acknowledgments
tD = dimensionless time This research has been conducted within the Marathon Center of
tmb = material-balance time, hours, days Excellence for Reservoir Studies (MCERS). Parts of this work
have been completed to fulfill the MS degree requirements of
tp = producing time, hours, days
Margaret Brown at Colorado School of Mines. Financial support
tpD = dimensionless producing time for Brown’s graduate studies has been provided by MCERS and
T = reservoir temperature, °R the Petroleum Engineering Department at Colorado School of
u = sf(s) Mines. We also thank Fekete Associates for sharing the data used
wF = hydraulic-fracture width, ft in the Example.

December 2011 SPE Reservoir Evaluation & Engineering 675

References Unconventional Reservoirs. Paper SPE 121290 presented at the SPE
Agarwal, R.A. 1979. “Real Gas Pseudo-Time”—A New Function for Pres- Western Regional Meeting, San Jose, California, USA, 24–26 March.
sure Buildup Analysis of MHF Gas Wells. Paper SPE 8279 presented http://dx.doi.org/10.2118/121290-MS.
at the SPE Annual Technical Conference and Exhibition, Las Vegas, Raghavan, R.S., Chen, C.C., and Agarwal, B. 1997. An Analysis of Hori-
Nevada, USA, 23–26 September. http://dx.doi.org/10.2118/8279-MS. zontal Wells Intercepted by Multiple Fractures. SPE J. 2 (3): 235–245.
Al-Hussainy, R., Ramey, H.J. Jr., and Crawford, P.B. 1966. The Flow of SPE-27652-PA. http://dx.doi.org/10.2118/27652-PA.
Real Gases Through Porous Media. J Pet Technol 18 (5): 624–636. Scott, J.O. 1979. Application of a New Method for Determining Flow
SPE-1243-A-PA. http://dx.doi.org/10.2118/1243-A-PA. Characteristics of Fractured Gas Wells in Tight Sands. Paper SPE 7931
Al-Kobaisi, M., Ozkan, E., and Kazemi, H. 2006. A Hybrid Numerical/ presented at the Symposium on Low Permeability Gas Reservoirs,
Analytical Model of a Finite-Conductivity Vertical Fracture Intercepted Denver, 20–22 May http://dx.doi.org/10.2118/7931-MS.
by a Horizontal Well. SPE Res Eval & Eng 24 (10): 345–355. SPE- Serra, K.V., Reynolds, A.C., and Raghavan, R. 1983. New Pressure Tran-
92040-PA. http://dx.doi.org/10.2118/92040-PA. sient Analysis Method for Naturally Fractured Reservoirs. J Pet Technol
Anderson, D.M., Nobakht, M., Moghadam, S., and Mattar, L. 2010. 35 (12): 2271–2283. SPE-10780-PA. http://dx.doi.org/10.2118/10780-
Analysis of Production Data From Fractured Shale Gas Wells. Paper PA.
SPE 131787 presented at the SPE Unconventional Gas Confer- Soliman, M.Y., Hunt, J.L., and El Rabaa, A.M. 1990. Fracturing Aspects
ence, Pittsburgh, Pennsylvania, USA, 23–25 February. http://dx.doi. of Horizontal Wells. J Pet Technol 42 (8): 966–973. SPE-18542-PA.
org/10.2118/131787-MS. http://dx.doi.org/10.2118/18542-PA.
Bennett, C.O., Camacho-V., R.G., Reynolds, A.C., and Raghavan, R. Stehfest, H. 1970. Algorithm 368: Numerical inversion of Laplace transforms.
1985. Approximate Solutions for Fractured Wells Producing Layered Commun. ACM 13 (1): 47–49. http://dx.doi.org/10.1145/361953.361969.
Reservoirs. SPE J. 25 (5): 729–742. SPE-11599-PA. http://dx.doi. Warren, J.E. and Root, P.J. 1963. The Behavior of Naturally Fractured
org/10.2118/11599-PA. Reservoirs. SPE J. 3 (3): 245–255. SPE-426-PA. http://dx.doi.org/
Camacho-V., R.G., Raghavan, R., and Reynolds, A.C. 1987. Response of Wells 10.2118/426-PA.
Producing Layered Reservoirs: Unequal Fracture Length. SPE Form Eval
Margaret Brown is a reservoir engineer working for Marathon Oil
2 (1): 9–28. SPE-12844-PA. http://dx.doi.org/10.2118/12844-PA.
Company in Houston. E-mail: mbrown6@marathonoil.com. She
Chen, C.-C. and Raghavan, R. 1997. A Multiply-Fractured Horizontal Well holds an MSc degree in petroleum engineering from Colorado
in a Rectangular Drainage Region. SPE J. 2 (4): 455–465. SPE-37072- School of Mines. Erdal Ozkan is a professor of petroleum engi-
PA. http://dx.doi.org/10.2118/37072-PA. neering and co-director of Marathon Center of Excellence for
Cinco-Ley, H. and Meng, H.-Z. 1988. Pressure Transient Analysis of Wells Reservoir Studies (MCERS) at Colorado School of Mines. E-mail:
With Finite Conductivity Vertical Fractures in Double Porosity Reser- eozkan@mines.edu. He holds BS and MS degrees from Istanbul
voirs. Paper SPE 18172 presented at the SPE Annual Technical Confer- Technical University and a PhD degree from the University of
ence and Exhibition, Houston. http://dx.doi.org/10.2118/18172-MS. Tulsa, all in petroleum engineering. Ozkan’s main research
de Swaan O., A. 1976. Analytical Solutions for Determining Naturally interests are pressure-transient analysis, modeling fluid flow
in porous media, horizontal and multilateral well technology,
Fractured Reservoir Properties by Well Testing. SPE J. 16 (3): 117–122.
and unconventional reservoirs. He has served as the Executive
SPE-5346-PA. http://dx.doi.org/10.2118/5346-PA. Editor of SPE Res Eval & Eng and chaired several SPE forums,
Kazemi, H. 1969. Pressure Transient Analysis of Naturally Fractured workshops, and conferences. Ozkan is a member of the SPE
Reservoir with Uniform Fracture Distribution. SPE J. 9 (4): 451–462. Reservoir Description and Dynamics Advisory Committee. He
SPE-2156-PA. http://dx.doi.org/10.2118/2156-PA. is the recipient of the 2007 SPE Formation Evaluation Award,
Kazemi, H., Merrill, J.R., Porterfield, K.L., and Zeman, P.R. 1976. Numeri- an SPE Distinguished Lecturer for the 2011–2012 season, and a
cal Simulation of Water-Oil Flow in Naturally Fractured Reservoirs. SPE Distinguished Member of SPE. Rajagopal Raghavan retired from
J. 16 (6): 317–326. SPE-5719-PA. http://dx.doi.org/10.2118/5719-PA. Phillips Petroleum Company in 2002. At Phillips, he was manager
Larsen, L. and Hegre, T.M. 1991. Pressure-Transient Behavior of Hori- of the Formation Evaluation Group. His primary interests are in
unsteady flow of fluids in porous media with emphasis on com-
zontal Wells With Finite-Conductivity Vertical Fractures. Paper SPE
putational techniques and in the description of oil and gas res-
22076 presented at the International Arctic Technology Conference, ervoirs. Raghavan holds a PhD degree from Stanford University,
Anchorage, 29–31 May. http://dx.doi.org/10.2118/22076-MS. a Diploma in petroleum production engineering from the
Larsen, L. and Hegre, T.M. 1994. Pressure Transient Analysis of Multifrac- University of Birmingham, England, and a BSc degree in electri-
tured Horizontal Wells. Paper SPE 28389 presented at the SPE Annual cal engineering from the Birla Institute of Technology, Ranchi
Technical Conference and Exhibition, New Orleans, 25–28 September. in India. He was awarded a ‘doctor honoris causa’ (honorary
http://dx.doi.org/10.2118/28389-MS. doctorate) by Stavanger University in Norway for contributions
Mayerhofer, M.J., Lolon, E.P., Warpinski, N.R., Cipolla, C.L., Walser, D., to graduate program at Stavanger University and technical
and Rightmire, C.M. 2010. What Is Stimulated Reservoir Volume? innovations in petroleum engineering. Raghavan is a member
of the National Academy of Engineering (NAE) and Honorary
SPE Prod & Oper 25 (1): 89–98. SPE-119890-PA. http://dx.doi.
Member of both AIME and SPE. He has served as the senior
org/10.2118/119890-PA. technical editor of SPE publications. Hossein Kazemi is the
Medeiros, F., Kurtoglu, B., Ozkan, E., and Kazemi, H. 2010. Analysis of Chesebro’ Distinguished Professor of Petroleum Engineering at
Production Data From Hydraulically Fractured Horizontal Wells in Colorado School of Mines and co-director of MCERS. E-mail:
Shale Reservoirs. SPE Res Eval & Eng 13 (3): 559–568. SPE-110848- hkazemi@mines.edu. He holds BS (1961) and PhD (1963)
PA. http://dx.doi.org/10.2118/110848-PA. degrees in petroleum engineering from the University of Texas
Medeiros, F., Ozkan, E., and Kazemi, H. 2008. Productivity and Drain- at Austin. Kazemi is a member of NAE, a Distinguished and
age Area of Fracture Horizontal Wells in Tight Gas Reservoirs. SPE an Honorary Member of SPE. He has served as distinguished
Res Eval & Eng 11 (5): 902–911. SPE-108110-PA. http://dx.doi. author and speaker for the SPE. Kazemi retired from Marathon
Oil Company in 2001 after serving as senior technical con-
org/10.2118/108110-PA.
sultant, director of production research, manager of reservoir
Mukherjee, H. and Economides, M.J. 1991. A Parametric Comparison technology, and executive technical fellow at Marathon in
of Horizontal and Vertical Well Performance. SPE Form Eval 6 (2): Littleton, Colorado, USA. At Colorado School of Mines, he has
209–216. SPE-18303-PA. http://dx.doi.org/10.2118/18303-PA. taught graduate courses in petroleum engineering, and super-
Ozkan, E., Brown, M.L., Raghavan, R.S., and Kazemi, H. 2009. Com- vised graduate research in reservoir modeling, well testing, and
parison of Fractured Horizontal-Well Performance in Conventional and improved oil and gas recovery.

676 December 2011 SPE Reservoir Evaluation & Engineering