You are on page 1of 9

Semi-Analytical Well Model of Horizontal

Wells With Multiple Hydraulic Fractures


J. Wan and K. Aziz, SPE, Stanford U.

Summary • Refine the grids, so that the true geometry of the fractures is
A well model is required to relate the well rate to the well pressure modeled. In this method, fractures are represented by very fine
and the wellblock pressure while modeling wells in reservoir simu- gridblocks; properties of the fractures, such as permeability, are
lators. The well index is calculated automatically by simulators assigned to those fracture blocks. Schulte10 evaluated the transient
for conventional wells, but it is generally calculated by the user pressure behavior of a fractured vertical well in this way. How-
and supplied in the input while modeling nonconventional wells ever, the size of the fracture blocks was so small that numerical
when the default procedure available in commercial simulators is instabilities were encountered. To overcome this problem, Roberts
not adequate. et al.11 increased the fracture grid width and decreased the frac-
In this paper we describe a new semi-analytical solution for ture permeability.
horizontal wells with multiple fractures. The fractures can be ro- • Modify the transmissibilities of the blocks that contain the
tated at any horizontal angle to the well, and they need not fully fractures. Work has been done in independently generating frac-
penetrate the formation in the vertical direction. This solution for ture transmissibilities with the transmissibilities of reservoir blocks
hydraulically fractured wells is obtained by applying Fourier adjacent to the fracture plane.12,13 Therefore, the productivity of
analysis to a 2D solution; therefore, this solution is easy to obtain the fracture is included while the negligible volume of the fracture
when the 2D solution is available. The analytical solution provides is excluded. These authors also apply various coupling methods to
the well pressure that can be combined with a numerically com- deal with the interfaces between the fracture model and the reser-
puted gridblock pressure to obtain the well index (WI). Results of voir model. Hegre14 computed correct transmissibilities for a
our rigorous solution are compared with empirical approaches cur- coarse grid, which gives the same long-term pressure behavior as
rently in use for calculating the well index and the productivity a fine-grid model. The transmissibility adjustments are based on
index. Examples are given when horizontal wells with fractures are results from a model, which is fine gridded in the near wellbore/
modeled using our new approach and conventional methods. These fracture area with explicit modeling of the fractures.
results demonstrate the need for computing the correct WI. • Modify the effective wellbore radius or well index. Cinco-
Ley and Samaniego15 showed that the effective wellbore radius is
Introduction one-fourth of the total fracture length for an infinite conductivity
The technology of fracturing horizontal wells is being widely used vertical fracture, so this effective wellbore radius can be used
by the petroleum industry. Therefore one will ask the questions: instead of rw in the productivity index formulas. Hegre and
Are the currently available models adequate for predicting perfor- Larsen16 showed that the effective wellbore radius for a horizontal
mance of horizontal wells with fractures? If not, how can we well intercepted by either a longitudinal fracture or multiple trans-
develop more appropriate models? verse fractures can be analytically determined. They used the rew
For the modeling of wells in reservoir simulators, a well model calculated in this manner in the well index formula (see Eq. 15)
is required to relate the well flow rate to the well pressure and the instead of rw.
wellblock pressure. Peaceman1,2 established a mathematical rela- The approach of modifying the effective wellbore radius is
tionship between the wellblock pressure and wellbore pressure for used in this study. A semi-analytical approach is used to calculate
a fully penetrating vertical well under the condition of 2D flow. the well index.6,8 First we develop the analytical solution for a
This relationship is valid for isolated wells under steady-state or horizontal well with multiple fractures in a brick-shaped drainage
pseudosteady conditions. Peaceman’s well model is usually the volume. With the wellbore pressure from this analytical solution
default in general reservoir simulators. Because of the variety of and the wellblock pressure from a single-phase numerical simula-
nonconventional wells used by industry, such as horizontal and tor, the effective wellbore radius is calculated.
fractured wells, and because of the variety of grid systems em- Several analytical solutions have been presented for transient
ployed, much research is being conducted on improving simulator flow in fractured wells. In 1973, Gringarten and Ramey17 used
accuracy. Some of the more important works are summarized here. source functions and the Newman product method for solving
Babu et al.3 extended Peaceman’s work to the case of a uniform transient flow problems. These transient flow solutions have been
flux horizontal well in a slab-like drainage area. Penmatcha et extended to predict the behavior of infinite-conductivity vertical
al.4,5 have extended Babu and Odeh’s model to the case of infinite fractures in homogeneous formations or in double-porosity me-
well conductivity. Jasti et al.6 present results to increase simulator dia.18–20 Guo et al.21 have developed methods for the performance
accuracy when solving the problem of a system of partially pen- prediction of horizontal wells with multiple fractures. Their work
etrating wells having arbitrary trajectories. Ding7 used the concept does not account for the effects of interference between fractures.
of layer potentials and applied transmissibility adjustments. Maiz- Horne and Temeng22 accounted for the interference between mul-
eret8 developed an analytical solution to investigate well indices tiple fractures by the superposition of influence functions. In 1991,
for nonconventional wells. In 1998, Wan et al.9 compared several Ozkan and Raghavan23,24 presented an extensive library of solu-
well models, including explicit well models, and showed the im- tions in terms of the Laplace-transform variable. They considered
portance of correct well indices. a wide variety of wellbore configurations, different bounded sys-
Our focus here is on horizontal wells with single or multiple tems, and homogeneous or double-porosity reservoirs. The paper
fractures. Basically, there are three ways to model hydraulic fractures: by Chen and Raghavan25 basically used the Ozkan-Raghavan so-
lutions in studying a multiply fractured horizontal well in a rect-
angular parallelepiped. Their method to deal with multiple frac-
tures is similar to that of Horne and Temeng, but Chen and Ragha-
Copyright © 2002 Society of Petroleum Engineers
van solved the problem in Laplace space and rewrote the
This paper (SPE 81190) was revised for publication from paper SPE 54627, first presented formulation in a more efficient way. Except for the work of Ozkan
at the 1999 SPE West Regional Meeting and Exhibition, Anchorage, Alaska, 26–27 May.
Original manuscript received for review 7 July 1999. Revised manuscript received 16 No-
and Raghavan,23,24 which includes the solutions of wells with 3D
vember 2001. Manuscript peer approved 20 August 2002. fractures in a horizontal infinite reservoir, the works mentioned

December 2002 SPE Journal 437


above only deal with 2D fractures. Larsen and Hegre26 tried to infinite conductivity solutions on individual fractures. With the
solve the 3D fracture problem, but they replaced the pressure in the unform flux solution, the pressure at 0.732 of the fracture half-
fracture by the average pressure in the z-direction and neglected length is used as the approximate infinite conductivity solution.18
flow in the fracture parallel to the horizontal well. With the infinite conductivity solution, each fracture is divided
Our study on multiply-fractured horizontal wells follows Chen into segments, and the interference between the segments of all the
and Raghavan,25 but we extend it to build new analytical solutions fractures is calculated. The unknowns are n·qDj and pwD and the
for 3D fractures based on published 2D solutions. Also, we com- equations to solve are wellbore pressure drop by convolution and
pute the effective wellbore radius (rew), which can be used for the the sum of total dimensionless flow rate to be 1. Full details of the
calculation of well index as discussed later. solution procedure can be found in Chen and Raghavan25 and
Wan.32

Mathematical Model 3D Fractures in a Parallelepiped Region. Define zwD as the cen-


Fracture Reviews. Generally, a fracture the plane of which is ter of the fracture in the z-direction, hwD as the depth of the frac-
perpendicular to the well is called a transverse fracture, and a ture. The pressure diffusion equation is given by
fracture the plane of which is parallel to the well is called a lon-
⭸2 pD ⭸ pD hD
gitudinal fracture. In practice, neither of these conditions may be ⵜ2pD + = + q ␦, . . . . . . . . . . . . . . . . . . . . . . . . . (3)
valid, and the fracture plane may have an angle to the wellbore. All ⭸ zD2 ⭸ tD hwD wD
of these cases are considered here. where ␦ = ␦xy␦z, z1 = zwD − hwD Ⲑ 2 and z2 = zwD + hwD Ⲑ 2, with
A fracture is considered to be 2D if it fully penetrates the
formation in the z-direction. Consequently, 3D fracture only par-
tially penetrates the formation in the z-direction. 再
1, z1 ⱕ z ⱕ z2
␦z = 0, otherwise , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (4)

Model Description. We make the assumption that the flow from


the reservoir to the wellbore sections between fractures is negli-

1, 共x, y兲 is on the fracture
␦xy = 0, otherwise , . . . . . . . . . . . . . . . . . . . . . . . . (5)

gible compared with the flow from the reservoir to the fracture and
plane. Four fracture models are built in sequence. First, an infinite
reservoir with a 2D fracture in any direction; second, a box-shaped ⭸2 ⭸2
ⵜ2 = + , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (6)
reservoir with a 2D fracture in any direction; third, a box-shaped ⭸ xD2 ⭸ yD2
reservoir with multiple 2D fractures of any direction along a hori-
The initial and boundary conditions are
zontal well; and fourth, a box-shaped reservoir with multiple 3D
fractures along a horizontal well. pD = 0 at tD = 0, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (7)
⭸ pD
2D Fracture in an Infinite Reservoir. We follow the approach of = 0 at zD = 0, hD. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (8)
Ozkan and Raghavan23,24 and construct the solutions in Laplace ⭸ zD
space.27,28 In an infinite and horizontal reservoir with closed upper After applying Laplace transform on Eq. 3, we have
and lower boundaries, the 2D vertical-fracture solution with unit
flow rate can be written in the Laplace space as26,27 ⭸2 pD hD
ⵜ2 pD = = spD + q ␦. . . . . . . . . . . . . . . . . . . . . . . . . . . (9)
⭸ zD2 hwD wD


1 1
K0关公s公共xD − ␣兲2 + yD2 兴 d␣, . . . (1)
+LfD The boundary condition given by Eq. 8 requires the application
pD共xD,yD;s兲 =
s 2LfD −LfD of the infinite Fourier cosine series in the z-direction, which can be
expressed as
where K0⳱the first-order modified Bessel function, and s⳱the ⬁
n␲zD
Laplace variable. In order to avoid difficulties as yD→ 0, we recast
the integral as
pD 共rD,zD;s兲 = qwD 兺 f 共r ,s兲cos h
n=0
n D , . . . . . . . . . . . . . . . . . . . (10)
D

where f refers to a specific function which will be defined next


and rD = 公xD2 + yD2 . By some derivations,36 the final forms of

1 公u共1ⳲxD兲LfD
pD共xDⱍⱕ1, yD = 0;s兲 = K0共␣兲d␣, . . . . (2) fn are
2sLfD公s 0

ⵜ2 f0 = sf0 + ␦xy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (11)


and use some special series expansion to compute 兰0x K0共␣兲d␣.27 and for n ⳱ 0. For n > 0, we have

冉 冊
All solutions in Laplace space are transformed back into real space
using the Stehfest algorithm.29 n2␲2 hD n␲ zwD n␲ hwD
ⵜ2 fn = s+ fn + 4␦xy cos sin . . . (12)
hD2 hwDn␲ hD 2hD
2D Fracture Within Closed Boundaries. Because we are con-
We notice that f0 is exactly the Green’s function of a 2D uni-
sidering fractures in random directions, we use the method of images
form-flux fracture. Similarly, fn also have the form
as proposed by Larsen30 to determine the coordinates of sufficient
numbers of image wells to compute the pressure distribution.
The equations presented so far are for fractures under the con-
dition of uniform flux. To solve the infinite conductivity case, the
fn共rD,s兲 =
4hD
n␲ hwD 冉n2␲2
f0 rD,s + 2 cos
hD
n␲ zwD n␲ hwD
hD
sin
2hD 冊
, . . . (13)

fracture is divided into several uniform flux segments; then, by Now the way to get 3D solutions is simplified; as long as we
using superposition, the desired solution is obtained. Lu31 dis- have 2D solutions, we can calculate f0 and fn, and then use Eq. 10
cussed the sensitivity of pressure results on segment numbers. We to form the 3D solutions. One thing needs to be mentioned here is
normally use five segments to represent each fracture. that the way we construct the 3D solutions limits the 3D fracture
plane to be vertical to the horizontal plane, but the fracture plane
Multiple 2D Fractures in Closed Boundaries. Interference be- can be rotated by any horizontal angle.
tween multiple fractures is accounted for by the superposition of There are some other computational issues, such as how to
influence functions.25 For the case of multiple fractures, we as- speed up the calculations and how to get infinite conductivity
sume infinite conductivity for the horizontal wellbore; therefore, solutions for 3D fractures, which are fully explained in Ref. 32.
the wellbore pressures are identical at each point at which the At pseudosteady state, the difference between 2D and 3D so-
wellbore intersects a fracture. We apply both uniform flux and lutions is a constant, which can be considered to be a pseudoskin:

438 December 2002 SPE Journal


Fig. 2—Effects of fracture angles for a square reservoir problem.
Fig. 1—Flow regimes obtained with a single 2D fracture.
and thus, both solutions are compared for three cases. For the first
pwD,3D = pwD,2D + Spseudo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (14) two cases, reservoir is square with a centered well. In one case, the
fracture length (xf) is one-eleventh of the reservoir length. In the
By the form of Eq. 10, this pseudoskin can be expressed very other case, the fracture length is nine-elevenths of the reservoir
easily.32 When t is large enough (s is very small), the pseudosteady length. As we can see from Figs. 4 and 5, there are some differ-
state is reached. ences between uniform flux and infinite conductivity solutions,

冋 ⬁ especially when the fracture is longer. As shown in Fig. 5, the


n␲ zD
pD共s → 0兲 ≅ qwD f0共rD,s兲 + 兺 f 共r ,0兲 cos
n=1
n D
hD
boundary effects are seen earlier for the infinite conductivity so-
lution than for the uniform flux solution. The reason is that instead
of a uniform flux distribution, the infinite conductivity solution has
= qwD f0共rD,s兲 = qwD Spseudo. . . . . . . . . . . . . . . . . . . . (15) the maximum flow rates at the two ends. Thus, the boundaries
Because the first term is the 2D solution, the second term will affect the well response early for infinite conductivity fractures.
be the pseudoskin. We have observed that the uniform flux solution becomes unac-
ceptable when the fracture is more than one-half the reservoir
Results and Discussion of Pressure Response length. Both uniform-flux and infinite-conductivity solutions are
based on legitimate assumptions, but the (0.732 of fracture half-
We are concerned with the pressure-transient behaviors of a single length) rule gives mismatch results. For the third case, the reser-
fracture or multiple fractures in a brick-shaped reservoir. Both voir is nonsquare, xe/xf=40, ye/xf=5, xw/xe=0.5, yw/ye=0.1. The
uniform flux and infinite conductivity conditions on the fractures differences are significant as shown in Fig. 6.
are considered.

Single 2D Fracture. In this case, we consider a fractured well in


a 1:4 rectangle, and the fracture is located at the center of the
rectangle. Fig. 1 shows the dimensionless pressure and dimension-
less pressure derivative vs. dimensionless time. At early time, the
linear flow regime is obtained where the derivative curve follows
a half slope line. Following linear flow, the flow becomes radial,
which is a horizontal line on the derivative curve, and the value of
the radial flow line should be 0.5. Because the aspect ratio of this
rectangle is 1:4, we notice a channel-type flow regime in which the
closer boundary becomes dominant first. Finally, when the farther
boundary begins to affect the well response, we will see a unit
slope line showing pseudosteady state.
The effect of the angle of a single 2D fracture depends on the
fracture length and the reservoir’s aspect ratio. In a square reser-
voir, this effect will be noticeable only when the fracture length is
large. Fig. 2 shows the effect on results by rotating a fracture to
45°. As expected, the effect of boundaries is felt earlier for the 0°
(longitudinal) fracture because it is closer to the boundaries. Be-
cause the reservoir is square, a longitudinal and a transverse frac-
ture of the same length have the same effects. When the reservoir
is nonsquare, a fracture that is parallel to the short boundary will
feel the boundary effect later than a fracture that is parallel to the
long boundary. This is shown in Fig. 3.
In this paper, most solutions are for infinite conductivity frac-
tures; however, in some cases, we also may be interested in uni- Fig. 3—Effects of fracture orientation for a nonsquare (3:1) res-
form flux solutions (pressure at 0.732 of fracture half-length, Lf), ervoir problem.

December 2002 SPE Journal 439


Fig. 4—Uniform flux and infinite conductivity solutions, xf/xe=1/11. Fig. 5—Uniform flux and infinite conductivity solutions, xf/xe=9/11.

• Pseudoradial flow, in which flow across the tips of the hori-


We validated our solution with that of Chen and Raghavan.25 zontal well becomes dominant.
With the reservoir of xe/xf=40, ye/xf=5, we almost get a perfect In the case of a reservoir with closed boundaries, the flow will
match for the centered fracture case (Fig. 7). eventually reach pseudosteady state. Fig. 8 presents the pressure
response of a horizontal well with five transverse fractures in a
Multiple 2D Fractures. Chen and Raghavan25 have shown the rectangular drainage region. The position of each fracture and its
different flow regimes for the multiple-fracture system. For infinite spacing is fixed, as shown is Fig. 9.
conductivity fractures in an unbounded reservoir, there are four Three cases included in Fig. 8 include an infinite reservoir case
main regimes. and two bounded reservoir cases. The bounded reservoir solution
• Linear flow, in which the flow into the fracture is normal to follows the infinite reservoir solution at early times. The smaller
the fracture planes. Each fracture behaves independently. reservoir with a drainage area of 100×80 reaches pseudosteady
• Early-radial flow. While flow is into the fracture tips, there is state after early radial flow. For the 420×200 reservoir, the area is
still no communication between fractures. This period depends larger, so the compound radial flow period appears. As soon as the
primarily on the fracture length and spacing, so this period may not closer boundaries become dominant, the reservoir enters the chan-
show up. nel-type flow regime, and with the farther boundary it enters the
• Compound-linear flow, in which fractures interact and the pseudosteady state flow regime.
flow becomes normal to the vertical plane that contains the hori-
zontal well. Single 3D Fracture. When the fracture is not fully penetrating the
formation in the z-direction, the problem is 3D. In the following
cases, the reservoir dimensions are 2200×2200×100 ft3. Two

Fig. 6—Uniform flux and infinite conductivity solutions, fracture


close to one boundary. Fig. 7—Validation of 2D fracture solution.

440 December 2002 SPE Journal


Fig. 9—Five 2D fractures in a rectangular drainage region.

Fig. 8—Five 2D fractures in different drainage areas.


When kz is small, we observe the following flow patterns:
• Early linear flow, in which the well feels the upper and lower
cases are included in Fig. 10. If we define hfD as the ratio of the boundaries only at late times, so this period is elongated.
fracture depth and the total formation thickness, hfD is from 0.4 to • Spherical type flow, in which the boundaries haven’t affected
0.7. The difference between them is the vertical permeability, kz. We the well response. Because this is a 3D problem, a spherical type
notice in Fig. 10 that the two cases have different derivative curves, flow shows up, as indicated by the negative pressure derivative slope.
except at late times. Therefore, we will discuss them separately. • Late radial flow. With the effects of upper and lower bound-
For the larger kz case, the main flow patterns are shown in Fig. aries, flow across the tips of the horizontal well becomes dominant.
11. Only the fracture is shown in Fig. 11; no horizontal well is Finally, both cases enter the pseudosteady-state period.
shown because it is neglected in our calculation. Figs. 12 and 13 compare uniform flux and infinite conductivity
• Early linear flow, in which the flow into the fracture is nor- 3D solutions. Because the reservoir’s horizontal length is much
mal to the fracture planes. Flow across the fracture edges is neg- larger than the reservoir depth, the fracture length has a greater
ligibly small. effect than the fracture depth on the differences between these two
• Transitional flow, in which the presence of the upper and solutions. This is similar to the case of multiple 2D fractures in that
lower boundaries is felt at the wellbore earlier, causing the pres- when the fracture is longer, the differences become larger.
sure derivative to decrease; at this time the derivative line is flat-
tened. There is flow across the fracture edges. Pseudoskin. Fig. 14 shows a comparison of 2D and 3D fracture
• Late linear flow, in which the flow pattern is normal to the solutions. During pseudosteady state, the difference between them
whole vertical plane that contains the fracture. is a constant value, which is the pseudoskin. We define hD as the
• Late radial flow, in which the flow across the tips of the ratio of the depth and the formation thickness. For the 3D fracture
horizontal well becomes dominant. the hD of which is from 0.4 to 0.7, the pseudoskin is 2.18; for the
3D fracture the hD of which is from 0.2 to 0.8, the pseudoskin
is 0.662.

Fig. 10—Flow regimes obtained with a single 3D fracture. Fig. 11—Multiple 2D fractures in a rectangular drainage region.

December 2002 SPE Journal 441


Fig. 12—Comparision between uniform flux and infinite conduc- Fig. 13—Comparision between uniform flux and infinite conduc-
tivity solutions for a single 3D fracture, kz=1 md. tivity solutions for a single 3D fracture, kz=50 md.

Multiple 3D Fractures. Consider a drainage volume that is square Fig. 16 shows the PI vs. dimensionless fracture half-length (2D
in the x-y plane. We use two examples to show 3D solution for fracture) and the PI vs. dimensionless fracture depth (3D fracture).
multiple equally spaced fractures. In Fig. 15, we do not see late For PI vs. LfD, we have compared our results with those of Kuppe
radial flow, because it is substituted by channel type flow and and Settari.14 The shape of the curves is similar, but the two
pseudosteady-state flow. solutions for 2D fractures shown in Fig. 16 are different. These
differences need further investigation.
Productivity Index and Well Model of
Fractured Well Well Model. We can relate well pressure and wellblock pressure
through the radial flow equation
The model properties for the following problems are shown in
Table 1.
krp
qp = WI 共p − pw 兲, . . . . . . . . . . . . . . . . . . . . . . . . . . . . (17)
Effects of Fracture Length on the Productivity Index (PI). PI is ␮pBp p,block
defined as
␪ kh

冋 册
WI = , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (18)
bbl q ln共r0 Ⲑ rw兲 + S
PI = . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (16)
Day ⭈ psi p − pw
where ␪⳱the angle open to flow and S⳱the skin factor.

Fig. 14—Comparision between 2D and 3D fracture solutions. Fig. 15—Three 3D fractures.

442 December 2002 SPE Journal


Peaceman1,2 derived the following expression for r0 (for a well
parallel to the x-axis)
关共kz Ⲑ ky兲1 Ⲑ 2⌬y2 + 共ky Ⲑ kz兲1 Ⲑ 2 ⌬z2兴1 Ⲑ 2
r0 = 0.28 . . . . . . . . . . . . . . . . (19)
ln共r0 Ⲑ rw兲 + S
For fractured well, r0/rw can be substituted by r0/rew. Cinco-
Ley and Samaniego15 and Hegre and Larsen16 have derived ap-
propriate expressions for rew. They have shown that rew is approxi-
mately half the fracture half-length. In our case, we calculate r0/rew Fig. 16—Productivity index of a single 2D or 3D fracture.
from our analytical solutions and our numerical simulator, then
apply Peaceman’s r0 to get rew. On the other hand, as long as we
have the correct well index, we can calculate the correct wellbore than the central fracture; therefore, their effective wellbore radius
pressure. In our study, we only considered the case in which a is larger. The effective wellbore radius changes with time until
fracture is completely contained in a single gridblock. Comparison pseudosteady state. The solution for three 3D transverse fractures
between our solutions with Eclipse33 solutions is shown in Fig. 17. (z/h⳱0.4 to 0.7) is also calculated for comparison and shown
The dotted line is our analytical single-fracture wellbore pressure; in Fig. 20.
the dashed line is the wellbore pressure calculated by Eclipse for Conclusions
a fully penetrating vertical well without any fracture. Because both General solutions are developed for horizontal wells with multiple
wells (vertical and horizontal with fracture) are completed in the 2D or 3D fractures. These solutions can be used to predict the PI
same single gridblock and they produce at constant rate, numerical and the WI for such wells.
simulations will yield identical gridblock pressures. The well pres-
sures, however, will be different, and they will depend on the WI. Nomenclature
It is obvious that if we can give correct well index, we can get the ct ⳱ total isothermal compressibility, Lt2/m, psi-1
same wellbore pressure as that from the analytical solution. The G ⳱ Green’s function
calculated effective wellbore radius, shown in Fig. 18, is more
h ⳱ formation thickness, L, ft
accurate than that of one-fourth of total fracture-length rule, which
only gives an approximation. hD ⳱ dimensionless formation thickness
In the case shown by Fig. 19, there are three 2D transverse hwD ⳱ dimensionless 3D fracture thickness
fractures along the horizontal well; each fracture is located in one kr ⳱ relative permeability
block. We use a 5×5 grid, so one fracture is located in each of the
three central blocks. The outer fractures have higher flow rates

Fig. 17—Model validation. Fig. 18—Validation of the effective wellbore radius rew.

December 2002 SPE Journal 443


Fig. 19—The correct effective wellbore radius rew, 2D fractures. Fig. 20—The correct equivalent wellbore radius rew, 3D fractures.

kx ⳱ permeability in the x-direction, L2, md Acknowledgments


ky ⳱ permeability in the y-direction, L2, md The authors would like to thank the Stanford U. Petroleum
kz ⳱ permeability in the z-direction, L2, md Research Inst.’s Reservoir Simulation Industrial Affiliates Pro-
l ⳱ segment length, L, ft gram (SUPRI-B) and the Stanford Project on the Productivity and
Lf ⳱ fracture half-length, L, ft Injectivity of Horizontal Wells (SUPRI-HW) for providing the
LfD ⳱ dimensionless half fracture length financial support for this study. This work was also partially sup-
Lrcf ⳱ reference fracture length ported by the U.S. DOE. Our special thanks also go to Pengbo Lu,
p ⳱ pressure, m/ Lt2, psi who contributed some sparkling thoughts on the analytical solu-
tions and offered considerable help on computational problems.
pD ⳱ dimensionless pressure
We also benefited from the comments given by Dr. Michael Prats.
q ⳱ flow rate, L3/t, STB/D
qwD ⳱ dimensionless well flow rate
rew ⳱ effective wellbore radius, L, ft References
ro ⳱ equivalent wellbore radius, L, ft 1. Peaceman, D.W.: “Interpretation of Wellblock Pressures in Numerical
rw ⳱ wellbore radius, L, ft Reservoir Simulation With Nonsquare Gridblocks and Anisotropic Per-
s ⳱ parameter of Laplace transformation meability,” SPEJ (June 1983) 531.
S ⳱ skin factor 2. Peaceman, D.W.: “Representation of a Horizontal Well in Numerical
t ⳱ time, t, day Reservoir Simulation,” paper SPE 21217 presented at the 1993 SPE
xD ⳱ dimensionless x Symposium on Reservoir Simulation, Anaheim, California, 17–20 Feb-
ruary.
xe ⳱ reservoir length in the x-direction, L, ft
3. Babu, D.K. et al.: “The Relation Between Wellblock and Wellbore
xf ⳱ fracture length, L, ft
Pressures in Numerical Simulation of Horizontal Wells,” SPERE (Au-
xw ⳱ x coordinate of fracture center, L, ft gust 1991) 324.
xwD ⳱ dimensionless x coordinate of fracture center 4. Penmatcha, V.R. and Aziz, K.: “A Comprehensive Reservoir/Wellbore
x1, x2 ⳱ dimensionless x coordinates of fracture tips Model for Horizontal Wells,” paper SPE 39521 presented at the 1998
yD ⳱ dimensionless y SPE India Oil and Gas Conference and Exhibition, New Delhi, India,
ye ⳱ reservoir length in the y-direction, L, ft 17–19 February.
yw ⳱ y coordinate of fracture center, L, ft 5. Penmatcha, V.R.: “Modeling of Horizontal Wells with Pressure Drop
ywD ⳱ dimensionless y coordinate of fracture center in the Well,” PhD dissertation, Stanford U., Palo Alto, California
y1, y2 ⳱ dimensionless y coordinates of fracture tips (1997).
zwD ⳱ dimensionless depth of fracture center 6. Jasti, J.K., Penmatcha, V.R., and Babu, D.K.: “Use of Analytical So-
zfD ⳱ the ratio of fracture depth to h lutions in Improvement of Simulator Accuracy,” paper SPE 38887
␦ ⳱ Dirac delta function presented at the 1997 SPE Annual Technical Conference and Exhibi-
tion, San Antonio, Texas, 5–8 October.
␾ ⳱ porosity
7. Ding, Y.: “A Generalized 3D Well Model for Reservoir Simulation,”
␮ ⳱ fluid viscosity, m/Lt, cp
SPEJ (December 1996) 437.
Subscripts 8. Maizeret, P.D.: “Well Indices for Non-Conventional Wells,” Master
D ⳱ dimensionless Report, Stanford U., Palo Alto, California (1996).
i ⳱ initial 9. Wan, J., et al.: “Effects of Grid Systems on Predicting Horizontal Well
n ⳱ integer indices Productivity,” SPEJ (September 2000) 309.
w ⳱ well 10. Schulte, W.M.: “Production From a Fractured Well With Well Inflow
Limited to Part of the Fracture Height,” SPEPE (September 1986) 333.
Superscripts 11. Roberts, B.E., van Engen, H., and van Kruysdijk, C.P.J.W.: “Produc-
- ⳱ Laplace transformed variable tivity of Multiply Fractured Horizontal Wells in Tight Gas Reservoirs,”

444 December 2002 SPE Journal


paper SPE 23113 presented at the 1991 Offshore European Conference, 28. van Everdingen, A.F. and Hurst, W.: “The Application of the Laplace
Aberdeen, 3–6 September. Transformation to Flow Problems in Reservoirs,” Trans., AIME (1949)
12. Settari, A., Puchyr, P.J., and Bachman, R.C.: “Partially Decoupled 186.
Modeling of Hydraulic Fracturing Processes,” SPEPE (February 1990) 29. Stehfest, H.: “Numerical Inversion of Laplace Transforms,” Commu-
37. nications ACM (1970) 13, No. 1, 47.
13. Kuppe, F. and Settari, A.: “A Practical Method for Theoretically De- 30. Larsen, L.: “A Simple Approach to Pressure Distributions in Geometric
termining the Productivity of Multi-fractured Horizontal Wells,” JCPT Shapes,” SPEJ (February 1985) 113.
(October 1998) 68. 31. Lu, P.B.: “Horizontal and Slanted Wells in Layered Reservoirs with
14. Hegre, T.M.: “Hydraulically Fractured Horizontal Well Simulation,” Crossflow” Master Report, Stanford U., Palo Alto, California (1997).
paper SPE 35506 presented at the 1996 SPE European 3D Reservoir 32. Wan, J.: “Models for Hydraulically Fractured Horizontal Wells,”
Modeling Conference, Stavanger, 16–17 April. Master Report, Stanford U., Palo Alto, California (1999).
15. Cinco-Ley, Heber, and Samaniego-V., F.: “Transient Pressure Analysis 33. Eclipse Reference Manual, 96A Release, GeoQuest, Schlumberger.
for Fractured Wells,” JPT (September 1981) 1749.
Appendix—Definition of Dimensionless Variables
16. Hegre, T.M. and Larsen, L.: “Productivity of Multifractured Horizontal
z
zD = 公k Ⲑ kz
Wells,” paper SPE 28845 presented at the 1994 SPE European Petro-
, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-1)
leum Conference, London, 25–27 October. Lref
17. Gringarten, A.C. and Ramey, H.J. Jr.: “The Use of Source and Green’s
hw
Functions in Solving Unsteady-Flow Problems in Reservoirs,” paper hwD = 公k Ⲑ kz , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-2)
SPE 3818; Trans., AIME, 255. Lref
18. Gringarten, A.C., Ramey, H.J. Jr., and Raghavan, R.: “Unsteady-State x
Pressure Distributions Created by a Well With a Single Infinite- xD = , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-3)
Lref
Conductivity Vertical Fracture,” SPEJ (August 1974) 347; Trans.,
AIME, 257. y
yD = , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-4)
19. Houze, O.P., Horne, R.N., and Ramey, H.J. Jr.: “Pressure-Transient Lref
Response of an Infinite-Conductivity Vertical Fracture in a Reservoir 0.0002637kt
With Double-Porosity Behavior,” SPEFE (September 1988) 510; tD = , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-5)
Trans., AIME, 285. V␾ct␮Lref
2

20. Warren, J.E. and Root, P.J.: “The Behavior of Naturally Fractured kh
Reservoir Properties by Well Testing,” SPEJ (June 1976) 117; Trans., pD = 共p − p兲, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-6)
141.2q␮ i
AIME, 261.
21. Guo, Genliang, Evans, R.D., and Chang, M.M.: “Pressure-Transient q
qwD = . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-7)
Behavior for a Horizontal Well Intersecting Multiple Random Discrete qref
Fractures,” paper SPE 28390 presented at the 1994 SPE Annual Tech-
nical Conference and Exhibition, New Orleans, 25–28 September. Jing Wan is a PhD student in petroleum engineering at Stanford
22. Horne, R.N. and Temeng, K.O.: “Relative Productivities and Pressure U. e-mail: jingw@pangea.stanford.edu. She holds a BS degree
Transient Modeling of Horizontal Wells With Multiple Fractures,” pa- in applied mechanics and mechanical engineering from the
U. of Science and Technology of China and an MS degree in
per SPE 29891 presented at the 1995 SPE Middle East Oil Show,
petroleum engineering from Stanford U. Khalid Aziz is the Otto
Bahrain, 11–14 March. N. Miller Professor of Earth Sciences and a professor of petro-
23. Ozkan, E. and Raghavan, R.: “New Solutions for Well-Test Analysis leum engineering at Stanford U. e-mail: aziz@pangea.stanford.
Problems: Part 1—Analytical Considerations,” SPEFE (September edu. Before joining the Stanford U. faculty, Aziz was a professor
1991) 359. of chemical and petroleum engineering at the U. of Calgary
24. Ozkan, E. and Raghavan, R.: “New Solutions for Well-Test Analysis and manager of the Computer Modelling Group. His research
Problems: Part 2—Computational Considerations,” SPEFE (September interests include reservoir simulation techniques, multiphase
1991) 369. and single-phase flow of oil, gas and their mixtures in pipes,
and horizontal and nonconventional well technology. He
25. Chen, C.C. and Raghavan, R.: “A Multiply Fractured Horizontal Well holds a BS degree in mechanical engineering from the U. of
in a Rectangular Drainage Region,” paper SPE 37072 presented at the Michigan, an MS degree in petroleum engineering from the U.
1996 SPE International Conference on Horizontal Well Technology, of Alberta, and a PhD degree in chemical engineering from
Calgary, 18–20 November. Rice U. An honorary member since 1996 and a distinguished
26. Larsen, L. and Hegre, T.M.: “Pressure-Transient Behavior of Horizon- member since 1983, Aziz served as a director in SPE during
tal Wells With Finite-Conductivity Vertical Fractures,” paper SPE 1997–2000, and as a board member on the committee for
22076 presented at the 1991 SPE International Arctic Technology Con- technical publications during 1997–2000. He also served as ex-
ference, Anchorage, 29–31 May. ecutive editor of SPE Journal during 1995–97, as a distinguished
lecturer during 1987–90, and on various other societywide and
27. Ozkan, E.: “Some Strategies to Exploit the Laplace Transformation for section committees. He is the recipient of the 1990 Distin-
a Tabulated Set of Numbers,” paper SPE 30552 presented at the 1995 guished Achievement Award for petroleum engineering fac-
SPE Annual Technical Conference and Exhibition, Dallas, 22–25 Oc- ulty, the 1988 Lester C. Uren Award, the 1987 Reservoir Engi-
tober. neering Award, and the 1979 Cedric K. Ferguson Medal.

December 2002 SPE Journal 445

You might also like