Sie sind auf Seite 1von 7

Modeling of Acid Fracturing

K.K. Lo, SPE, and R.H. Dean, SPE, Areo Oil & Gas Co.

Summary. The paper presents a theoretical framework for modeling acid fracturing stimulations. Starting from the fundamental equations
of fluid mechanics, fracture mechanics, convection, and diffusion, the paper outlines the steps necessary to derive simplified equations
for an acid fracturing model. Unlike some existing models, the coupled problem of fracture geometry, acid transport, and diffusion is
solved simultaneously in this paper. Although an infinite reaction rate is assumed in the solution of the problem, an empirical correlation
is used to account partially for finite reaction rates. Errors in the governing equations of some of the existing models are identified.
To assess the accuracy of the approximations used in the present model, exact solutions are used for comparison. Predictions from the
present model are compared with a model in the literature, and the results are found to be in reasonable agreement. As in all existing
acid fracturing models, not all the phenomena of the acid fracturing process have been incorporated into the present model. Nevertheless,
the present model is improved over existing models because it is derived from fundamental equations and thus forms a basis from which
further improvements can be made.
Introduction
Acid fracturing is a common well stimulation technique in the petroleum industry for limestone and dolomite formations. In an acid
fracturing treatment, an inert fluid (known as the pad) is injected
into a well under high pressure, creating a fracture in the formation.
As the fracture length is increased with continued fluid injection,
acid is injected into the formation, reacting with the formation on
the fracture surface. The acid is transported along the fracture by
convection during fracturing. At the same time, the acid is transferred to the reactive surface by diffusion and by fluid leakoff into
the formation. Once the acid reaches the fracture face, it reacts with
the formation. Because the acid fracturing process is complicated,
simplifying assumptions have to be made to make the problem of
modeling the process tractable. On the other hand, several important
features have to be retained to model the physics of the process
properly, including fracture geometry (fracture length, width, and
height), fluid leakoff rate, convection along the fracture, mass
transfer of the acid to the rock surface, and the acid reaction rate
on the surface. Of course, these processes occur simultaneously
during the acid fracturing treatment, so they are not independent
of one another. In limestone formations, the acidizing process is
limited by the rate of acid transport, not by the reaction rate. As
one of the simplifying assumptions in this paper, acid reaction rates
are ignored at the well and during the stimulation.
This paper describes a model of acid fracturing based on a twofluid generalization of the Perkins and Kern I model and a onedimensional (l0) approximation of the general two-dimensional
(20) diffusion-convection problem. Several authors 2-7 previously
presented acid fracturing models. The model described in this paper,
although in many ways similar to those described in previous publications, is derived directly from the 20 model, and the masstransfer rate comes directly from the analysis of the 20 diffusionconvection problem.
The model described here consists of two parts: a fracturing model
and an acid transport model. We first write the governing equations
for a Perkins-Kern fracture model derived by Nordgren. 8 We then
describe the Perkins-Kern approximation and generalize it to two
fluids. The two-fluid generalization of the Perkins-Kern approximation is the basis of the fracture model proposed in this paper.
For the acid transport model, the 20 convection-diffusion equation
is used as a starting point for the derivation of the 10 approximation
averaged over the fracture width. The mass-transfer rate obtained
from such an approximation is compared with the full 20 masstransfer rate obtained from solving the 20 equation.

Governing Equations
Fracture Model. In Nordgren's8 fracture model, the fracture
height, h, is assumed to be constant. The rate of fluid leakoff per
unit length into the formation at any point in the fracture can be
approximated by
qw=(2hCL )I.Jt-r, ................................ (1)

Copyright 1989 Society of Petroleum Engineers

194

where t=time and CL =leakoff coefficient that is usually measured


in a static filtration test or a mini fracture test. The factor 2 in Eq.
1 accounts for fluid leakoff rates from both fracture surfaces. In
general, the coefficient is a function of the reservoir properties,
fracturing-fluid properties, and filter-cake buildup. Note that r=r(x)
is the time it takes for the fracture to reach Point x.
The governing equations for Nordgren's fracture model consist
of the continuity equation and the fracture-width/pressure-elasticity
relationship:
(CJqICJx)+(7rhI4)(CJb max /CJt)+qw=O ................... (2)

and b max = [2(l-/t2)hlE]~ (x), ....................... (3)


where h=fracture height, q=fluid flux rate, qw=fluid-Ioss rate at
the fracture surface, b max = maximum opening at the center of the
fracture cross section, E=Young's modulus, /t=Poisson's ratio,
and ~[=p(x)-a]=net pressure acting on the fracture surface.
In addition, many fracturing fluids approximately obey a powerlaw relationship between the shear stress, s, and the shear strain
rate, -y,

s=K-yn =Klduldy In-1duldy, ........................ (4)


a form commonly assumed for non-Newtonian fluids. In Eq. 4,
u is the velocity down the fracture, nand K are fluid constants,
and y is in the direction normal to the fracture wall. The fracture
fluid is Newtonian if n= 1.
Guillot and Ounand 9 showed that fracture fluids can exhibit
Newtonian behavior (n= 1) at low shear rates and power-law behavior (n< 1) at high shear rates. They found that an Ellis model
produced a reasonable fit of their experimental data for a wide range
of shear rates (0.01 to 2,000 seconds-I). In addition, for the high
shear rates normally encountered in fracturing applications, they
found that the power-law model was a suitable approximation for
calculating fracture shapes. The power-law model in Eq. 4 is used
in the next section to calculate fracture shapes.
Perkins and Kern Approximation. As Nordgren observed, an approximate solution of Eqs. 1 through 4 with fluid loss can be obtained from the zero leakoff solution. That is, we first obtain the
zero leakoff solution by setting the time derivative and the fluidloss rate, qw' in Eq. 2 to zero and by integrating to obtain an expression for the fracture width, b max :
bmax(x) = [

128 (n+ I)K(l-/t2)h [(2n+l)i

37r

---

In(L-x) }lI(2n+2) ,

nh

.................................... (5)

where i=injection rate, L=fracture length, and the fracture width


is required to be zero at the fracture tip. We then account for the
fluid loss by modifying the fracture volume. In deriving Eq. 5, we
made use of the velocity profile and its relationship to i to give an
expression for the pressure gradient down the fracture; the volumetric flow rate is equal to the velocity u integrated over the cross
section between the fracture surfaces. Because i is independent of
SPE Production Engineering, May 1989

,------_._---------------------,

rol'---L2 - -...,Oojl"'--- LlL2---~'I


FLUID 1

FLUID 2

w,
Fig. 1-Two fluid stages along one wing of the fracture.

Fig. 2-Acid concentration profiles along the plates.

x by assumption, the pressure gradient can be integrated and Eq.


3 used to eliminate Ap to yield Eq. 5 (see Ref. 1 for details). We

For Fluid 2, integrating Eqs. 1 through 4 from L2 to 0 and using


the Perkins-Kern approximation gives

can integrate Eq. 5 to obtain the fracture volume, V (both wings):

V=('II'/2)h \max dx ................................ (6)

128 (n2 + I)K2(1-J.'2)h

b z =[

3'11'

o
3'11'2
E
[nh
= 128 K(1-J.'2)(2n+3) (2n+ l)i

In o2n+3
b

, .......... (7)

where bo=maximum fracture width at the wellbore. A factor ('11'/2)


is included in Eq. 6 because the a~erage fracture width for an elliptical cross section is given by b='II'b max /4, where b max is the
maximum width of the cross section. The nonzero fluid loss is accounted for in the above approximation by equating the fracture
volume to the actual fluid volume injected minus the fluid loss; Le.,
V=it-2] L

rqw

dt dx .............................. (8)

=it-8CLh

.J t-T(X) dx. . ........................ (9)

(2n 2 + l)i2ln2
[

JII(2n2 +2)
L2 +brn2 + 2
, ........... (12)

n2h

where b2 = fracture width at the wellbore. For widths at any point


less than L 2 , Eq. 12 still applies, except L2 in Eq. 12 is replaced
by (L 2 -x). Integrating Eq. 12 from 0 to L2 gives the total fracture
volume from 0 to L 2 :

3'11'2
2

V = 128 K2(1-J.'2)(2n2+3)

ln 2

[n2h
(2n2+1)i2

X(bin2+3-brn2+3) . ............................ (13)

We now look at the fluid volumes to obtain a second set of


equations in terms of L I , L 2 , and t (time). The total volume of
Fluid 1 remaining in the fracture (both wings) is

The factor 2 in the second term of Eq. 8 represents the total fluid
loss in both wings of the fracture. From Eqs. 7 through 9, one can
form an implicit equation for L at a given t, which means that the
fracture length can be solved as a function of time given an injection
rate, i. Eqs. 5 through 9 are known as the Perkins-Kern I approximation. Physically, the approximation means that fluid leakoff has
very little effect on the fracture shape and fluid leakoff primarily
affects only the overall fracture volume. That is why the zero-leakoff
b max -L relationship in Eq. 5 can be used to approximate the
generalleakoff case, as long as the global fluid balance in Eq. 8
is maintained.
To generalize the Perkins-Kern approximation to two fluid stages
for acid fracturing treatments, one can apply a similar argument
to the two stages. In what follows, all quantities with the subscript
j (j = 1,2) are associated with the fluid at Stage j. Fig. 1 shows the
arrangement of the two stages. Fluid 2 extends from the wellbore
to a distance L2, while Fluid 1 is between L2 and the end of the
fracture at L I . b l is the fracture width at L 2 , the trailing edge of
Fluid 1, and b 2 is the fracture width at the wellbore. For Fluid 1,
evaluating Eq. 5 at L2 gives

b j128 (nl+l)K I (I-J.'2)h


l

C3'11'
X

(2n l
[

+ l)i2ln1

nih

(L I -L 2 )

JII(2n1 +2)
.

.........

(10)

Integrating b max as in Eq. 6 but from L2 to LI gives the same relation as Eq. 7, except bo, n, and K all have the subscript 1. i is
replaced by i 2, which is the current injection rate when two fluids
are present in the fracture:
VI =

3'11'2

128 K 1(1- J.'2)(2nl

[nih

+ 3)

(2nl

SPE Production Engineering, May 1989

+ l)i 2

ln 1

brn1 + 3 . . . (11)

VI =i It,- rI8CLlh.Jt-TI(X) dx
L2

r L2

-J
o

8CLl h.J T2(x)-T, (x) dx, .................... (14)

t,

where i, = injection rate during Stage I, = time when Stage 1


ended, and T2(x)=time when the leading edge of Fluid 2 passed
Point x. Values of T, (x) and T2(x) are retained at selected points
along the fracture to evaluate the integral expression in Eq. 14. The
corresponding equation for Stage 2 is

................................... (15)
Eq. 15 assumes that the fluid loss for the second stage obeys Eq.
1 where the leakoff coefficient is CL2 and T(X) is the time that the
fracture tip reached Point x. Eq. 1 is an adequate approximation
to the leakoff rate for the second stage when CLI and CL2 are of
the same magnitude. Eqs. 14 and 15 can be written as VI =
gl (L I .2,1) and V2 =g2(L I ,L2 ,t). Other forms of fluid leakoff rate
(other than Eq. I), such as a constant leakoffrate, can also be used
in Eq. 15 for the acid. In the case of a constant leakoff rate, CL2
and both square-root terms in Eq. 15 are replaced by V2(t-T2),
where V2 is the constant leakoff velocity .
Substituting Eqs. 10 and 12 into Eqs. 11 and 13 gives two
equations for the volumes VI and V2 in terms of the unknowns Ll
and L 2 These two equations can then be combined with Eqs. 14
and 15 to produce two nonlinear equations relating the three
unknowns L I, L 2 , and t. For a given t, the two nonlinear equations
are solved for L, and L2 with a Newton-Raphson technique with
residual-monitored damping. Thus, the fracture length Ll and fluid
interface location L2 can be calculated as functions of time. As ex195

pected, Eqs. 11 through 15 reduce to the one-fluid Perkins-Kern


equations in the literature when the two fluids are the same. Eqs.
11 through 15 are generalized to the case of multiple fluids in the
Appendix.

Acid Transport. The general acid transport model can be described


by the 2D continuity equation and the diffusion-convection
equation 2 :

ID Approximation. So far, other than the pa~allel-plate approximation in which the domain b is replaced by b, the "integrated"
Eq. 22 is an exact equivalent of its 2D counterpart, Eqs. 16 through
18. The ID approximation enters when the exact Nusselt number
is replaced by its averaged value down the plates; Le.,

I dx/

_[LaC
NNu..,NNu=-b J _
o <Jy y=bl2

rLCmdx .

.......... (25)

<JuIiJx+<Jvl<Jy=O .................................. (16)

and u(<JCICJx) + v(<JCI<Jy) =D (<J 2 CI<Jy2), ................ (17)


where u, v = lateral and transverse velocity components, C=acid concentration, and D=effective diffusivity. We have dropped the time
derivative and the diffusion term in the x direction in Eqs. 16 and
17. The effective diffusivity is much larger than the molecular diffusivity owing to mixing as the fluid moves down the fracture. 2
In using a 2D convection-diffusion equation as a starting point
of the derivation, we assume that the z variation in the equation
is negligible. Consistent with this approximation, we also assume
thl!t the dOl!!ain of the 2D equation is 2D and can_be defined by
-b12 <y<bI2, where there is no z dependence and b is the average
width across the elliptical cross section of the fracture. Because the
cross-sectional shape varies slowly with respect to z, we expect this
domain approximation to be valid in a region away from the crack
tips. That is, the solution near the center of the cross section for
the truly 3D problem can be approximated by that of a problem of a
pair of parallel plates with bbeing the spacing between the plates.
In all the following calculations, it is assumed that the reaction
rate at the walls is much faster than the mass-transfer rate to the
walls or through the porous walls. An infinite reaction rate will
be used in all calculations, and all acid reaching the fracture surface
will react immediately at that surface. This infinite reaction rate
causes the acid concentration at the fracture surface to be zero; i.e.,
c=o at y= -b12 and b12 . ......................... (18)

Define the average of any function f as

f='; Jbl2f dy .................................... (19)


b -b12

and the average weighted concentration as

Cm =

!_J

bl2 Cu dy.

.............................. (20)

bu -b12

We now derive a ID equation from Eq. 17 by "averaging" over


the width of the fracture-i.e., applying the averaging operator in
Eq. 20 to Eqs. 16 and 17:
bu(<JCmlox)=2[D(<JCloy)l y=bl2] +2Cmv w , ......... (21)
where Vw = leakoff velocity at the wall.
Using the boundary condition in Eq. 18 and combining Eqs. 16
and 17, we obtain
<JCml<Jx = (2lub )[vw-(Dlb )NNU]Cm, ................. (22)

where NNu, the Nusselt number, is defined as


NNu = -(bICm)(oCI<Jy)l y =bI2' ...................... (23)
NNu is related to the mass-transfer rate of the acid to the wall, w,
by
w=(Dlb )NNu(X)Cm(x) . ............................ (24)

Equations similar to Eq. 22 have been derived il! the literature 2 ,7


for the average concentration, but the term (Dlb )NNu(X) in Eq.
22 (known as the mass-transfer coefficient) has been treated as a
constant that can be determined experimentally. 7 Based on the
above derivation, the exact mass-transfer coefficient varies along
the fracture and depends on the solution of the problem. Hence,
care must be taken in applying an experimentally determined constant to Eq. 22 because, if the wrong value of the mass-transfer
coefficient is applied, the right side of Eq. 22 may become positive.
This leads to a physically incorrect, increasing concentration profIle
down the fracture.
196

The Nusselt number approaches a limiting value as we proceed


down the plates so that sufficiently far along the plates the mass
transfer is proportional to Cm. To calculate Eq. 25, we make use
of the solution of a similar problem in heat transfer obtained by
Terrill. to Terrill studied a 2D problem of a pair of porous parallel
plates held at a constant temperature. His heat transfer problem is
directly analogous to the convection-diffusion problem considered
here. Translated to the convection-diffusion setting, his problem
can be stated as follows: the fluid is flowing at steady-state conditions between the plates while the fluid is simultaneously leaking
off through the porous plates at a constant velocity, V. At x=O
(where the x axis is the centerline that runs between the plates),
the acid concentration is set to a constant value across the gap between the two plates, and a steady-state concentration profile forms
downstream. The acid concentration is constrained to be zero along
the plate boundaries. The maximum velocity at x=O is U. Fig. 2
shows schematically the evolution of the concentration profile down
the plates.
The above problem is more tractable than the acid transport
problem described by Eqs. 16 through 18 because the gap b~tween
the plates for the above problem is kept constant, whereas b in the
fracture problem varies withx. The fracture width is such a slowly
varying function of x, however, that wOe assume that the approximate Nusselt number derived from the simpler problem of the
constant-width parallel plates continues to apply for our acid
transport model. Terrill derived the velocity profile (u and v) between the porous parallel plates by solving the Navier-Stokes and
continuity equations. This velocity solution was then substituted
into the thermal conduction-convection equation, a direct analog
of Eq. 17, and the temperature distribution was solved by separation of variables and eigenfunction expansion. Making use of
Terrill's solution for the convection-diffusion problem considered
here, we can express the first three terms of the expansion fo.! C
and Cm. Eq. 25 was then integrated to give an expression fo~NNu
as a series expansion in terms of the Peclet number, N Pe = VbI2D:
NNU'" 4.10+ I. 26NPe +O.04Nfe' .................... (26)

which is valid for Peclet numbers of order 1. For large Peclet


numbers (>20), Eq. 26 would be replaced by NNu ",,2Npe , as derived in Ref. II.

Comparisons
Now we compare the solution of the 2D prob~m with that of the
10 equation (Eq. 22) with NNu replaced by NNu in Eq. 26. The
width between the plates is now b. All comparisons will be restricted
to a set of parallel plates with constant fluid leakoff velocity for
simplicity. One can show that the expression on the right side of
Eq. 22 with NNu replaced by NNu is always negative, so the concentration will always decrease as the acid moves down the fracture.
A program was developed to solve the 2D concentration equation
numerically while the ID equation can be solved by quadrature.
The following comparisons assume that the plates are 100 ft [30
m] in height with a gap of 0.1 in. [0.254 em]. The fluid has a viscosity of 100 cp [100 mPa 's], a density of 1 g/cm 3 , and a diffusivity of 0.0001 cm 2 /s. Fluid is moving down the plates at a rate
of 10 bbllmin [0.0265 m 3 /s] at x=O, and for the first comparison,
fluid is leaking off at a rate of 0.001 ft/min [5.08 x 10- 6 m/s]; the
second comparison has zero leakoff. For these physical parameters,
the Reynolds number down the plate, N Reu ( = UbI2v), is 4.35 and
the Schmidt number, N Se( = vID), is 10,000. The leakoff Reynolds
numbers, N Rev (= Vb12v) , are 6.45 x 10- 5 and 0.0 for the leakoff
case and the no-Ieakoff case, respectively. The Peelet number is
0.65 in the 1eakoff case.
SPE Production Engineering, May 1989

en-------------------------------~--------,
NReu

NSc

=
=

NRev =

4.35
10000.

NReu=

6.5E-5
4.35

10000

NSc

.
1 .
~
"
1
"'"
NRev =6.5-5:i
',-,---------------. -------,---------------- ----0.0 ------ --

..

LEGENO

..

".

Two Dimensional SoIn

o+-____
~~------~------~--------~------~
10
100

Dlstanc. (II)

One DImensIonal Soln

...~M------.~..

O+.------M~-----~~O------~~----~.~.O~~
DIstance (II)

Fig. 3-Nusselt numbers along the plates.

Fig. 4-Concentration profile along the plates with leakoff.

The variable Nusselt number in Eq. 23 was calculated by first


solving the 2D equations (Eqs. 16 through 18) numerically and then
evaluating the expression in Eq. 23 for the leakoff and no-leakoff
cases. The Nusselt numbers are shown in Fig. 3 as a function of
distance along the plates. In both cases, the Nusselt numbers start
out very large and quickly decrease to a limiting value. One can
show that the Nusselt number goes to infinity like x - y, as x approaches zero. After about 30 ft [9 m], the Nusselt numbers approach limiting values of 4.45 and 3.77 for the leakoff and the
no-leakoff cases, respectively. The corresponding Nusselt numbers
predicted by Terrill's analytical result, NNu =3.77 +NPe +0.087
NJe+ ... , are also 4.45 and 3.77, showing excellent agreement
between the asymptotic results and the 2D numerical solution.
Based on Eqs. 22 and 26, the ID concentration solution was computed for the leakoff problem. Its concentration profile is shown
in Fig. 4, along with the Cm profile computed from the 2D concentration solution. The concentrations have been scaled to one at
x=O. The ID solution is very similar to the 2D concentration profile
over the entire length of the plates.
The ID concentration is larger than the 2D concentration near
the entrance of the plates and smaller than the 2D concentration
farther down the plates because the ID solution uses an average
Nusselt number while the 2D solution has a variable Nusselt number
that begins at infinity atx=O and decreases to a limiting value farther
down the plates. The average Nusselt number for this case is 4.93;
the limiting Nusselt number is 4.45.
For the zero-leakoff case, the ID and 2D solutions are compared
in Fig. 5. Again, the agreement is very close over the entire length
of the plates. For this problem, we also performed the lD calculations using the limiting Nusselt number of 3.77 instead of the

average Nusselt number. The results are shown in Fig. 6, along


with the 2D results. As expected, the ID solution in this case is
always larger than the 2D solution asymptoting to the correct value
down the plates.
Eq. 22 models acid transport for a power-law fluid. The expression for the average Nusselt number in Eq. 26, however, is
strictly valid only for a Newtonian fluid lO (Le., n=1 in Eq. 4).
Bird et al. 12 presented solutions to a variety of convectiondiffusion problems involving power-law fluids. On the basis of the
limiting Nusselt numbers presented there for a tube with a circular
cross section, Eq. 26 will underestimate the average Nusselt number
for a typical power-law fluid of exponent 0.25 by about 15%.

NRev

NReu

NSc=

The Mode.
For the acid transport model, we have from Eqs. 22, 25, al!d 26,
with V replaced by the leakoff velocity, v w, and N Pe =vwbIW,
ilCmlilx=(2/iib)[ -4.1(D Ib)+0.37v w -0.Olvw2 bID )Cm,
. ................................... (27)
where, unlike the parallel-plate problem, u, b, and Vw are assumed
functions of x. Except when the Peclet number is of order 1, the
third term on the right side of Eq. 27 normally can be ignored in
most applications. The inclusion of the term, however, guarantees
that the right side will always be negative.
Eq. 27, together with Eqs. 10 through 15, constitutes the acid
fracturing model. For each timestep, the fracture length and width
are calculated from Eqs. 10 through 15 for a given injection rate.
One can then calculate the fluid velocities and integrate Eq. 27 to
calculate the velocity-weighted average concentration, Cm. The

0.0
4.35
10000.

NRev
NReu
NSc

=
=

0.0
4.35
10000.

..' .......... .

..
N

..

LEGEND

"-

On, Dimensional Sol"

Two Dimensional Sol"

O~-----T---o
.0

, , , ,.

Two Dimensionaf SoIn

__~----~------~----~------~
'00

LEGEND

On. DImensIonal Soln

1~O

Dlstonc. (II)

200

no

Fig. 5-Concentratlon profile along the plates without leakoff.

SPE Production Engineering, May 1989

...

O+O------~M------'~OO------~~----~.~"------~ZM------~
Dlstanc. (II)

Fig. 6-Concentratlon profile using the limiting Nusselt


number.
197

rate of acid transport to any point on the fracture surface is multiplied by the timestep size to calculate the amount of acid reaching
that point during that timestep. These amounts are then summed
for the entire fracturing job to calculate the total amount of acid
reaching any point along the fracture surface.
The diffusivity, D, in Eq. 27 will be larger than the molecular
diffusivity because of surface roughness at the fracture walls and
the circulation induced by density differences caused by the acid
reaction. The model calculates an effective diffusivity from the
Reynolds number for flow parallel to the fracture from correlations
based on the work of Roberts and Guin 7 and Nierode and
Williams. 3 For finite reaction rates, the model uses the correlation
of Nierode et al. 4 to calculate an effective diffusivity, although,
as noted by Williams et al., 2 the correlation has little theoretical
basis.
During pumping, we calculate the amount of acid transported to
the fracture surface, but unspent acid still remains in the fracture
when the pumps are shut down. This unspent acid will also react
with the fracture surface to give a productivity enhancement. We
assume that the unspent acid present in a gridblock at the end of
pumping will react with the fracture surface at that gridblock. This
unspent acid is added to the acid that accumulated at that point during
the fracturing job to arrive at the total amount of acid that reacts
at that point along the fracture surface. Note that Cm must be converted to an average concentration, C, during calculation of the
amount of unspent acid remaining in the fracture.

Temperature Dependence
This paper has not addressed the effects of temperature variations
along the fracture. Temperature variations could influence the masstransfer rate, the acid reaction rate, and the fluid rheology. Temperature calculations may be included in the model by incorporating
an analytical expression for the temperature distribution or by
solving numerically for the temperature distribution in much the
same way as one solves for the acid distribution. Sinclair 13 discussed the procedure for incorporating an analytical expression of
the temperature distribution, while Lee and Roberts 14 solve numerically for the temperature distribution.
Productivity-Increase Calculations
For productivity-increase calculations, the model determines the
amount of rock dissolved within each gridblock. The acidized
fracture width for a gridblock is the dissolved fracture volume for
the gridblock divided by the fracture height and gridblock length.
These idealized widths are then used as input for calculating the
corresponding conductivity through each gridblock.
In the actual physical acidizing process, the acid will not dissolve
a uniform fracture width that acts as a single parallel plate for production. Instead, the acid will create many ridges and valleys along
the fracture face, and it is precisely this heterogeneous distribution
of peaks and valleys that allows portions of the acidized surfaces
to remain separated after the fluid has drained off and the fracture
has tried to close. Because portions of the acidized surface are held
open only by means of irregularities along the fracture surfaces,
however, one would expect that the degree of propping should be
a function of the formation softness, the in-situ closure stresses,
and perhaps several other formation parameters. For example, if
a formation is too soft or the in-situ stresses are too large, the ridges
propping the fracture open become very compressed and very little
of the fracture will remain open after the acidizing treatment.
Because the relationship between the fracture conductivity and
acidized fracture width will be a function of reservoir properties,
one ideally should determine the fracture-conductivity-to-acidizedwidth relationship experimentally by flowing acid between slabs
of core to create fracture widths. One would then press the two
slabs of core together under in-situ stress conditions and flow fluid
between the slabs to determine the corresponding conductivities.
In this way, fracture conductivity as a function of acidized width
for the formation of interest could be determined.
In many applications, one will usually not undergo the time and
expense involved in experimentally determining the relationship between the acidized width and fracture conductivity. In such cases,
correlations that express fracture conductivity as a function of
198

fracture width can be used. The acid transport model uses the
Nierode and Kruk 15 correlations, which were determined for a
variety of core samples under different load levels.
After the fracture conductivities have been determined for each
gridblock, productivity improvements can be calculated. The steadystate productivity calculations based on Raymond and Binder's16
paper were used to calculate the productivity increase after acid
fracturing. In the calculations, radial flow around the wellbore is
assumed and the steady-state radial-flow equations are integrated
out from the wellbore. At any distance r from the wellbore, they
assume that the radial conductivity is given by the sum of211"rtimes
the reservoir permeability and two times the local fracture conductivity. In this way, they can account for fmite-conductivity fractures;
however, because of the radial-flow assumption, their results may
not be accurate for fractures extending over a large portion of the
drainage area.

Comparison With Models In the Literature


The acid fracturing model described in this paper is based on a twofluid generalization of the Perkins-Kern model and uses Eq. 27 to
model acid transport. The fluid velocities and fracture widths in
Eq. 27 are allowed to vary along the fracture length.
The acid transport model (Eq. 27) is a generalization of the model
presented by Nierode and Williams, 3 who use Terrill's 10 solution
(for parallel plates) in graphical form to predict the acid penetration
distance. In this paper, Terrill's solution is incorporated into a
fracture model to predict the fracture length and the acid penetration
distance simultaneously. Eq. 27 is similar to the model presented
by Roberts and Guin. 7 For the case of infinite reaction rate at the
walls, Roberts and Guin's equation becomes
iJCliJx=(2/iib)(v w -Kg )C, ......................... (28)

where Kg =mass-transfer coefficient. Eq. 28 was also derived in


Refs. 14 and 17. In Refs. 7, 14, and 17, however, no distinction
was made between averaged quantities as defined in Eq. 19 and
velocity-averaged qulU!tities as defined in Eq. 20. For Eq. 28 to
be rigorously correct, C should be replaced by Cm . Comparing Eq.
28 with Eq. 22, one can see that the equations are very similar,
except that Eq. 28 is expressed in terms of a mass-transfer coefficient and Eq. 22 uses the Nusselt number.
As shown in Ref. 17, a foam acid system is governed by the same
equation as Eq. 28, except that the diffusion coefficient, D (and
hence Kg) is replaced by DI(I- r) (r is foam quality as defined
in Ref. 17). Hence, foam fracture acidizing can be modeled by the
same equations in this paper with a modified mass-transfer
coefficient.
Using Eq. 22 with Eq. 26 has three advantages over using Eq.
28. First, Eqs. 22 and 26 show explicitly how the mass-transfer
rate depends on the fracture width and fluid leakoff, while these
terms are embedded in Kg in Eq. 28. D in Eq. 22 will be a function
primarily of the Reynolds number down the plate, while Kg will
be a function of the Reynolds and Schmidt numbers. The second
advantage is that use of Eqs. 22 and 26 allows the mass-tran&fer
rate to vary along the fracture. And third, the right side of Eq. 22
will always be negative when used with Eq. 26, while the right
side of Eq. 28 may become positive if the leakoff velocity is much
larger than the measured leakoff velocity, which occurred during
experimental determination of Kg.

Comparison With Published Results


Williams et al. 2 presented a set of acid fracturing calculations for
a well completed in a limestone formation at a depth of 7,SOO ft
[2286 m]. The formation has a SO-ft [lS-m] -thick oil zone with
a permeability of O.S md and a viscosity of O.S cp [O.S mPa s].
Additional reservoir properties are listed in Table I.
The reservoir is stimulated with two fluid stages. A pad fluid
of a given volume is injected followed by a suitable acid volume.
The stimulation is performed with pad volumes of ISO, 300, 4S0,
and 600 bbl [23.8, 47.7, 71.5, and 9S.4 m3], followed by acid
volumes of 78,97, 121, and IS4 bbl [12.4, IS.4, 19.2 and 24.5
m 3]. Additional fracturing fluid properties are listed in Table 2.
Using the fracture and the acid transport model presented earlier,
the simulations took less than S seconds on a Macintosh II. The
SPE Production Engineering, May 1989

TABLE 1-RESERVOIR PROPERTIES USED


IN THE MODEL CALCULATIONS

Fracture gradient, psi/ft


Fluid density, Ibm/ft 3
Fluid compressibility, psi- 1
Porosity
Reservoir temperature, of
Reservoir pressure, psi
Well spacing, acres
Well radius, ft
Young's modulus, psi
Poisson's ratio

0.7
52
0.0001
0.1
200
2,500

40
0.5
6.45 x 10 6

0.25

fluids were injected at a rate of 15 bbllmin [0.0398 m 3 /s], as opposed to the 10 bbllmin [0.0265 m 3 /s] used in Ref. 2, because 10
bbllmin [0.0265 m3 /s] would not support a growing fracture. The
Perkins-Kern fracture model used in this paper predicts longer, narrower fractures than the Geertsma-de Klerk model used in Ref. 2,
which necessitates higher injection rates for P-K simulations.
The model presented here predicts productivity increases of3.4,
3.6, 3.8, and 4.0 for the four stimulations. Given the differences
between the current model and the model in Ref. 2, this compares
well with Williams et al. 's predictions of 3.2, 3.6, 4.0, and 4.4.
The current model predicts productivity variations of 0.2 between
stimulations; Ref. 2 predicts variations of 0.4.

Conclusions
We have shown that the acid fracturing model proposed is derived
rigorously from a consistent approximation of fundamental
equations. The model includes a 10 approximation (averaged over
the fracture width) of the 20 acid transport model, and a two-fluid
Perkins-Kern approximation of the 20 fracture model. They were
shown to be excellent approximations of the corresponding 20
equations. While the present model is an extension of existing acid
fracturing models, it differs from existing models in that diffusion
and fluid leakoff are explicitly accounted for in the acid transport
equation. The concentration profile and the productivity calculations
based on the model presented here compare well with published
results. Further generalizations of the model are possible. For example, the model has not addressed the temperature variation from
the wellbore to the fracture tips, which could affect the mass-transfer
rate, the acid reaction rate, and the fluid rheology. If the temperature variation within the fracture is known, the equations can be
modified to model the phenomenon. The model can also be extended
to include more than two fluids. Because the model is based on
fundamental equations, it provides a sound theoretical framework
from which improvements in modeling the acid fracturing process
can be made.
Nomenclature
b = fracture width, L
Ii = average fracture width, L
b max = maximum width of fracture cross section, L
b o = maximum fracture width at the wellbore, L
b 1 ,b z = fracture width at L and at wellbore
C = acid concentration, m/L3
C = average acid concentration, mIL 3
Cj.CL\,CLZ = leakoffcoefficients for Fluids i, 1, and 2, Lltl-Z
CL = leakoff coefficient, Lltl-Z
Cm = velocity-averaged acid concentration, m/L 3
D = diffusion coefficient or mixing constant, LZ/t
E = Young's modulus, m/U z
fi,g,g\>gz = known functions, dimensionless
h = fracture height, L
i = injection rate, L3 It
ii, i l,iz = injection rate for Fluids i, 1, and 2, L 3/t
K.Ki.K\>Kz = fluid constant for fracture fluid and Fluids i, 1,
and 2, m/UZ-n
.
Kg = mass-transfer coefficient, LIt
L = fracture length, L
SPE Production Engineering, May 1989

TABLE 2-FLUID PROPERTIES USED


IN THE MODEL CALCULATIONS

Pad fluid
Average viscosity, cp
Fluid-loss C L1 , ftlmin 'h
Acid
Average viscosity, cp
Acid density, %
Fluid-loss eLl' ftlmin'l2

60
0.002
1.2
15
0.002

L j = distance from wellbore to interface between Fluids


i-I and i
LI = distance from wellbore to fracture tip, L
L z = distance from wellbore to fluid interface, L
n,nj,n\>nz = fluid constant for fracture fluid and Fluids i,
1, and 2
NNu = Nusselt number, dimensionless
NNu = average Nusselt number, dimensionless
N Pe = Peclet number, dimensionless
NRe.NReu'
N Rev = Reynolds numbers, dimensionless

NSc = Schmidt number, dimensionless


p = fracture pressure, m/U2
qw = totalleakoffrate (both faces) of one wing per unit
length, LZ/t
r = distance from wellbore, L
s = shear stress, m/Lt Z
t = time, t
t 1 = time of injection for Fluid 1, t
u = lateral velocity component, LIt
Ii = average lateral velocity component, LIt
U = velocity down the plate
v = transverse velocity component, LIt
vw = leakoff velocity at fracture surface, LIt
V2 = constant leakoff velocity, LIt
V = fracture volume, L3
Vi,vI,vZ = fracture volume for Fluids i, 1, and 2, L3
w = mass-transfer rate
x = distance from the wellbore, L
y = transverse distance from the centerline or the
channel or fracture, L
.y = shear rate, t- I
r = foam quality
p. = Poisson's ratio, dimensionless
" = viscosity
u = in-situ stress, m/LtZ
T = time for acid to reach a particular point in the
fracture, t
T i, T 1 ,TZ = time for Fluids i, 1, and 2 to reach a point in the
fracture, t

Acknowledgment
We thank R.S. Schechter for helpful discussions in connection with
this work.
References
1. Perkins, T.K. and Kern, L.R.: "Widths of Hydraulic Fractures," JPT
(Sept. 1961) 937-49; Trans., AIME, 222.
2. Williams, B.B., Gidley, J.L., and Schechter, R.S.: Acidizing Fundamentals, Monograph Series, SPE, Richardson. TX (1979) 6.
3. Nierode, D.E. and Williams, B.B.: "Characteristics of Acid Reactions
in Limestone Formations," SPEI (Dec. 1971) 406--18; Trans., AIME,
251.
4. Nierode, D.E . Williams, B.B., and Bombardieri. C.C.: "Predictions
of Simulation From Fracturing Treatments," J. Cdn. Pet. Tech. (Oct.Dec. 1972) 31-41.
199

5. van Domselaar, H.R., Schols, R.S., and Visser, W.: "An Analysis
of the Acidizing Process in Acid Fracturing," SPEJ (Aug. 1973) 239-50;
Trans., AIME, 255.
6. Roberts, L.D. and Guin, J.A.: "The Effect of Surface Kinetics in
Fracture Acidizing," SPEJ (Aug. 1974) 385-95; Trans., AIME, 257.
7. Roberts, L.D. and Guin, J.A.: "New Method for Predicting Acid
Penetration Distance," SPEJ (Aug. 1975) 277-86.
8. Nordgren, R.P.: "Propagation of Vertical Hydraulic Fracture," SPEJ
(Aug. 1972) 306-14.
9. Guillot, D. and Dunand, A.: "Rheological Characterization of Fracturing Fluids by Using Laser Anemometry, " SPEJ (Feb. 1985) 39-45.
10. Terrill, R.M.: "Heat Transfer in Laminar Flow Between Parallel Porous
Plates," Inti. J. Heat Transfer (1965) 8, 1491-97.
11. Terrill, R.M. and Walker, G.: "Heat and Mass Transfer in Laminar
Flow Between Parallel Porous Plates," Appl. Sci. Res. (1967) 18,
193-220.
12. Bird, R.B., Armstrong, R.C., and Hassager, 0.: "Dynamics of Polymeric Liquids, " Fluid Mechanics, John Wiley & Sons, New York City
(1977) I, 470.
13. Sinclair, A.R.: "Heat Transfer Effects in Deep Well Fracturing," SPEJ
(Dec. 1971) 1484-92; Trans., AIME, 251.
14. Lee, M.H. and Roberts, L.D.: "Effect of Heat of Reaction on Temperature Distribution and Acid Penetration in a Fracture," JPT (Dec.
1980) 501-07.
15. Nierode', D.E. and Kruk, K.F.: "An Evaluation of Acid Fluid Loss
Additives, Retarded Acids, and Acidized Fracture Conductivity, " paper
SPE 4549 presented at the 1973 SPE Annual Meeting, Las Vegas, Sept.
30-Oct.3.
16. Raymond, L.R. and Binder, G.G. Jr.: "Productivity of Wells in Vertically Fractured, Damaged Formations," JPT (Jan. 1967) 120-30;
Trans., AlME, 240.
17. Ford, W.G.F. and Roberts, L.D.: "The Effect of Foam on Surface
Kinetics in Fracture Acidizing," SPEJ (Jan. 1985) 89-97.

Appendix-Generalization of the Fracture Model


to More Than Two Fluids
Let the number of stages be m. Set Lm+ I =0, b o =0, to =0. The

width relation becomes


blni + Z -bi~l+Z =f(ni,Ki,im)(L i -Li+I), i=2,3 . . . m, .. (A-I)
where i corresponds to Fluid i at Stage i, and f is a known function
of the injection rate at Stage m and the fluid constants. The lengths
Li are defined analogously as LI and Lz (see Fig. 1).

200

The volume of Fluid i, Vi, is


Vi=g(ni,Ki,im)(blni+3 -bi~1+3), .................. (A-2)

where g is a known function of the fluid constants and the injection


rate.
The leakoff relations for Fluids m and i (i= 1,2 ... m-I) corresponding to Eqs. 14 and 15 are

L'+I

J
o

8CUh[.JTi+1 -TI(X)-.JTi(X)-TI(X) ]dx .. (A-3a)

and Vm=im(t-tn-I)- Lm8CLmh[.Jt-TI(X)

o
-.J Tm(X)-TI (x) ]dx, ........................... (A-3b)

where Vi is the volume occupied by Fluid i. Equating Eqs. A-3


and A-2 gives m equations, which, coupled with the m equations
obtained from Eq. A-I, give 2m equations for the 2 m unknowns
L i , and b i for a given t.

51 Metric Conversion Factors


acres x 4.046 873
E+03
cp x 1.0*
E-03
ft x 3.048*
E-Ol
ft3 x 2.831 685
E-02
OF (OF-32)/1.8
Ibm x 4.535 924
E-Ol
md x 9.869233
E-04
psi x 6.894 757
E+OO
Conversion factor is exact.

mZ
Pa's

m
m3
C

kg
J.l.mZ

kPa

SPEPE

Original SPE manuscript received for review Feb. 12. 1988. Paper (SPE 17110) accepted
for publication Nov. 16. 1988. Revised manuscript received Oct. 14. 1988.

SPE Production Engineering, May 1989

Das könnte Ihnen auch gefallen