Vous êtes sur la page 1sur 16

,

SPE

fbdBtuof R3tramnm#n3are

SPE 19784
Interference Testing of Finite Conductivity
Fractured Wells

Hydraulically

D,N, Meehan, Union Pacific Resources/Stanford U,, and R.N. Home and H.J. Ramey Jr., Stanford U.
SPE Members

,
T

Copyright 1989, society of Petroleum Englnem, Inc.


Thla paper was prepared for presentationat the S4th Annual T. .+nlcal Conference and Exhlbilionof the Society of Petroleum En@nears held In San Antonio,TX, October S-1 1, 1SSS,
This paper was selected for presentation by an SPE Program Commlttea followingreview of informationcontained in an abetracf eubmltted by the author(a).Contents of the paper,
as presented, have not been reviewed by the Society of Petroleum Errginseraand are aubjecr to correction bythe
uthor(s), The material, es presented, does not nemsarlly refieol
any positionof the society of Petroleum Englneera, its offIcers, or membere, Papers presented at SPE meetlrrgsare subjectto publicationreview by EdltorlalCornmltreeeof the Society
of Patrokum Enginaera.PernrieelorItocrrQYla reebictedto an abstractof notmorethan 300 words.Illuatratlonsmay notbe copbd. The abstractshouldce+rteinconspicuouseck~
of where and by whom the paper la presented. Write Publication MunaWr, SPE, P.O. Box 83SS3S, Richardson, TX 7WSS4S3S. Telex, 7S0SS9 SPEDAL

This paper presents techniques for the design and analysis of interference tests when both the active and observation well are intercepted by hydraulic fractures.
These
techrdques are based on new mathematical
solutions. The
solution is general for any values of dimensionless fracture
conductivity including infinite conductivity, The solution is
presented in Laplace spac~ fracture skin, wellbore storage,
naturally fissured matrix behavior, etc. are readily included,
Any rate or pressure schedule at the active well can be analyzed,
Compass orientation of the wells hydraulic fractures can
be determined from such an interfer
ce test. Relative fracture lengths, conductivities,
and azi
% uth significantly affect W+ performance, Reservoir heterogeneities may significantly alter interference response,
The value of knowing
hydraulic fracture azimuth can also be determined.
Performance of a hydraulically fractured well near a large natural
fracture or another hydraulically fractured well is also forecast, Values of fracture conductivity
and fracture length
fer each fracture can not be determined uniquely from the
interference response and must be determined separately,
Historical
Approach
GXapplication
of interference tests is the determination of fracture orientation in hydraulically fractured wells
(i.e., wells intersecting a vertical fracture). Pierce et al, [11
described a method for determining fracture rizimuth and
fracture length using pulse testing, Ttis method requires
pulse tests before and after the fracturing opeation, and is
not applicable for very low permeability systems or finite
conductivity fractures,
Uraiet et al, [2] developed a technique for azimuth determination
using pressures rccordcd
at an unfractured observation well, This work comidercd
Refererrcea and illuatrationa

at end of paper,

137

only uniform flux fkacturea; an extension by Cinco-Ley and


S.arnaniego [3] considered finite conductivity fractures, None
of these methods work for hydraulically fractured ohservation wells. Abobiae and Tiab [4] and Ekie et al. [5] also
used uniform flux models for interference work. Unamb&uof fracture azimuth requires two obeerous determination
vation wells located at orientations other than 90 and 180
degrees from the active well.
Reservoir heterogeneities
complicate the determination
of fracture ~jmuth
fmm jnte~emnce tats,
Permeability
variationa in the drainage area of a well are averaged in a
manner that is not entirely understood.
No current model
exists to assess the impact of these heterogeneities on interference test determination
of fracture azimuth,
Unfortunately,
both high-resolution (high cost) pressure
tranaduceti and very long teets are required for these interference tests because of the extremely low permeability of
most candidate formations, Reservoir heterogeneities,
multiple layers, and surface interference can render azimuth detection by interference testing infeasible, Resolution with
respect to azimuth is also low at angles greater than 45,
Field tests used to evaluate fracture azimuth were reported
bly Frohne and Mercer [6], [7] and Sarda [8] . Elkins and
Skov [9] , Komar et a/, [10],[11] , Komar :and Shuck [12] ,
and Locke and Sawyer [13] described field experiments that
involved contouring pressures at offset wells during drawdowm, fracturing operations, and injection,
Historically, most work has used the line source approximation at the observation well, This is not without
reason, as numerous interference tests in fields with known
heterogeneities have resulted in tests that match the classical exponential integral solution, The uniform flux model is
only reasonable for short fracturea, and in these cases, the
observation well must be close to the active well to dMer-

,
2

Interference

SPE 19784

Teathu of Finite conductivity H~drauliC@ Fractured Wells

entiate between various fracture azimuths.


Use of infinite
conductivity
fracture models is unrealistic for low permeabilityy reservoirs where long fractures are createci to iiow
commercial quantities of hydrocarbons,
Fundamental
flow equations for Newtonian fluid flow
in homogeneous porous media and corresponding assumptions are well known [14] , Numerous solutions for different
boundary conditions have been published,
Many of these
were direct analogs to solutions of heat conduction problems due to the similarity between the diffusivity equations
in temperature
and in pressure [15] . Gringarten popularized the use of Source and Greens functions for solving these
problems [16] . This section derives the finite conductivity
fracture pressure and flux calculations,
Laplace space formulations for equations are used, simplifying the problem
and increasing flexibility of the solution,
Cinco and Samaniego [17] presented a mathematical model
that has become standard for evaluating @ite conductivity
hydraulic models, The basic procedure is a semi-analytic
one in which the hydraulic fracture is modeled with a large
number of elements (usually 20 to 40 per wing). Each element is modeled as having uniform flux; however, flux distribution is not known a priori. Reservoir and fracture flow
equations are equated along the fracture and the discretized
system is solved for wellbore pressure and flux distribution.
Numerous extensions of the kclmique have been published,
Cinco and Meng [18] and van Kruysdijk [19] recently
presented formulations in Laplace space for finite conductivity fractures
Solving the equations in Laplace space has
several advantages; viz,,

4. The pressure

derivative

5. Convolution to obtain variable


ciently rapid that computerized
regression is possible,

rate solutions is suffiautomatic type-curve

ofiacfure

flow equation8

Consider a fracture to LS a homogeneous slab of uniform porous media with height h, width b~ and half
length ZJ. Since fracture length is much longer than
fracture width, fluid intlux at the fracture ends m.y
be neglected, Fluid entem the fracture facea at a raw
g(z, t) per unit of fracture length. Unsteady-state
flow
in the fracture can be described as:

with the following initial and boundary


apj

-x .=O =

cond~tiorm

. %@
2bfkjh

(5)

(6)

2. Addition

and:
pf(Xjt=O)=pi>
By neglecting

Wellbore storage solutions are obtained by a simple


manipulation
of the constant rate Laplace solution
with skin (p, D(s)):

0< W<2!f

the fracture
LYpf

(1)

P.D=PD+S

an-

6. The Laplace space formulation allows immediate solution of transient pressure response for naturally fractured cases by substituting
the term s~(s) for s in
the wellbore pressure solution terms that result from
the reservoir flow model. Here, f(s) will be one of
the dual porosity models, typically either for transient
matrix flow or pseudo-steady state matrix flow. Cinco
and Meng [N presented a formulation that n~glected
compressible flow in the fracture (fracture linear flow).
demonstrated
the accuracy of this approxCinco [181
imation, The van Kruysdijk model [19] included consideration of compressible flow in the fracture.

1. This method is fast using the Stehfest algorithm~20]


for rapid inversion to real space. Previous techniques
required discretization in both time and space,
of wellbore skin effect and wellbore storage
are easily obtained. Inclusion of wellbore storage will
typically be importreal well testing problems,
Sandface dimensionless pressure (P,D) is simply conventional dimensionless pressure plus skin damage effect (S),

group pj term is obtained

alytically.

~-~

compressibility
/J !lf(~>tDzf)
bf h

Now, defining dimensionless

(7)
term
(8)

variable as:
(9)

~,D(s)
%D(g)

(2)

1 + cD~2F,D(~)

(lo)
3. By solving in Lapktce space, constant pressure
tions for q~ and QD are easily obtained.

solu(11)

?Wf)($)
= s [1 +

cD&#D($)]

(3)

Cumulative production for constant pressure production is simply dimensionless flow rate divided by the
Laplace space variable (s), as integrating with respect
to time is synonymous to division by 8,

noting that:
pfD(*,

t) =

-PjD

(12)

(13)

Substitution

and cancellation

leads to

changed from [-1 1] to [0 1] by utilizing this symmetry.


This does not usually appll* for interference problems.

This equation can be integrated twice to yield the pressure drop between the wellbore and any point in the
fracture:
PwD(~Dzf)

PjD(~D,

~D.f)

Discretization

and

matrix

formulation

The integral invo~ving the lfo terms can be integrated


for the discretized fluxes as described in Appendix A.
Discretization
into n equal length fracture segments
(on each fracture half-length) and the approximation
of uniform flux over eachsection reduc~- the double
integral of the fluxes to

(16)

fl~?,D(z@)dzd==

(19)

ZTfLli(3)
1
i-=
1

j-1

(A~)2

{xD;[D~q,D(stl,t~=,)d~tt~.t}

&(ZDj

iAz)

+(Ax)2

This integration uses the no flux boundary condition


at the tips, the known value of dimensionless pressure
at the origin, and the total flux condition to evaluate
the constants of integration.
oReservoir

~~fDj(~)

Subscripts for ZD imply that iocaticms ~Dj BIW mid~okts of the jth segment. Values for CD, and zDi+~
are at the beginning and end of the ith segment, respectively. This system of equations can be solved for
the single well problem and has been previously discussed for the real space solution,
In the following
section, the details of solution for interference with
two finite conductivity
hydraulically fkactured wells
is given.
.
Interference
Between
Two FYnite Cond ~ctivity
Hydraulically
Fractured
Wells

jlo w equations

Dimensionless pressure drop at any point in space due


to a plane source of height h, length 221 with flux
density g~(z, tD=J) is:

In this section, a solution is generated for the combined interference problem of finite conductivity y fractures intersecting both active and observation wells.
Formulation is for different length fractures with different values of (kjbj)D.

Equating these two equations for ~D = O and -1<


~D < 1 and taking the Laplace transform yields:

This solution can evaluate behavior of either a line


source well or finite conductivity
hydraulically fractvred well located near a natural fracture or another
hydraulically fractured well. This method would quantify such impacts and predict the behavior of a hydraulically fractured well in a naturally fractured system with large, widely spaced fractures. Primary use
of the solution is for the well teztirq
+.erference problem. Nomenclature
is illustrated in . .gure 1.

Note that for the Laplace transformation,


the following properties are used (reference equation numbers
from Reference [21] are shown in bold face),
29.2.1

Semi-Analytic

Solution

Fracture flow equations and reservoir equations are


written and coupled for each well as in previous derivations, However, in this solution, fluxes from both fractures must be specified. In Laplace space, the dimensionless pressure drop at the active well is:
(20)

+-(:(?,dz,d~d==
+
For a single well in an infinite reservoir, the fracture is
symmetric, ~~D(XD, .$) = T~D(-zD,
s); the integration
limits of Equation 18 for the reservoir term may be

,c~D+l/A@)

~ ~mD-l/A,Ao,
129

jDo(x

a)KO (fi~~)

dz

I
SimiIarly, for the observation
.,
pwD(.)

well,

-1 JmD+AAo
qfDo(z, S)lic,(l

x~ - x

I #i)fzz

2 =n8D-1/~AO]

+-fl(~,D(.!$)d.d.=
11
~ * 7J,~A(z, 8)Ko(fi&
L

(21)

Zy + djj)dz

terms are [~.D - l/AIAO] ZmD + l/AIAOl].


At early times the cross terms (B C) are negligible,
The result is infinite acting behavior at the active well
and zero pressure drop at the observation well, At late
times, the magnitude of the cross terms approaches
that of the main diagonal. In practice, the order of
the matrix (Table 1) can be reduced, in this case from
(m+ 2) (m+ 2) tom, m. This reduction speeds the
matrix inversion and reduces storage. Matrix inversion is typically repeated eight times in each time step
in the Stehfest algorithm numerical Laplace transform
invemion. Details of this reduction are presented in
Appendix B. Other methods for accelerating the solution, both for filling the matrix and inversions, are
also discussed in Appendix B.

The equation for the observation well only holds for


equal fracture conductivities.
Modification
for different conductivitiea and permeability anisotropy will be
discussed. If n matrix blocks are used for each fracture
half length, the above equations constitute 4n equations for 4n + 2 unknowns.
For the remaining two
equations, the flux conditions at each well are incorporated,
Letting m = 272for the totaI number of fracture blocks,
for the active weli:
(22)
for the observation

well:
klo
i=l

For the simplified case of equal fracture lengths and


equal fracture conductivities,
the matrix formulation
is given in Table 1, Aij terms and Dij terms arise
from the pressure drops at the fractured active and
observation wells due to the fluxes at each of those
wells respectively,
For the special ctie of ~fA = z~o
and (kjbf)A = (k@~)o, these two terms will be identical.
The Bij and Cij terms are the crow ter~s,
which contribute to the effects of the active well fluxes
on the wellbore pressure at the observation well, and
vice versa, Equations 20 and 21 are discretized using
the remdts of Equation 19. The counters for the ij
terms must be kept consistent with the direction of
discretization.
Modifications
Conductivities

for Different

Fracture

and the relative

fracture

permeability

width product

I
I

as:

(25)
The ratio of the values of (klbj) q for the two wells is

Effects

of Azimuth

and

Spacing

Typically, a group of curves is displayed showing varying fracture azimuths which range from 15-90. As active well responses show small variations due to fracture azimuth, only the 15 and 90 degree cases are displayed. C)bservation well figures show responses for
each 15 degree increment, except for cases when all
of the responses are spaced very close together.
For
the graphsof pD at the active well, the infinite acting

Lengths,

Distances require scaling when fracture lengths or conductivities are not equal. The mtio of the active wells
hydraulic fracture length to that of the observation
well is given as:
zf,4
(24)
AIAo] = ~

In this section, a series of figures are used to summarize solution results. A series of &mea will give the
dimensionless pressure (pD) and dimensionless p&ssure derivatives (p~) at the active and observation
wells. Dimemionless
pressure derivative groups are
useful for evaluating more subtle characteristics
and
as a diagnostic tool. These derivativare calculated
from the Laplace space solutions directly and do not
require numerical differentiation.

(23)

Because the origin of the observation well is displaced


ZmD along the x-direction and dD in the y-direction,
additional changes to the formulation are also required.
Solution consistency requires the active and observation wells to be formulated in their specific coordinate
systems. For example, the integration limits for the
Aij terms are [-1 1], Correapondlng limits for the Dij

solution is also displayed.


Active well solutions are
plotted as a function of tD.jo Invariably, the infinite
acting solution overlays the data at early times. At
late times, the active well interference solutions show
varying levels of negative skin; During a transition
period of varying length, active well pressures and the
pressure derivative groups fall below the corresponding values for the irdlnite acting well. During thh time,
interference is most pronounced.
For the interference wells, the exponential integral is
plotted, along with ~t) and pi. observation
well soFor relhltions are plotted as a function of tD.f/F~.
atively small valuea of TD, and for low values of azimuth,
it is clear that the line source solution is a poor approximation,
As TDincreases to valuea of four and
above, all of the azimuth solutions collapse to the line

D. N. Meehan,

5PE 19784

iarge vaiues of @due to the ditliculty in differentiating


between these angles. Initiai estimates of fracture azimuth (from other techniques) shouid be used to avoid
attempting
tests large angles.

source solution, Figuree 2-9 present type curves for


(kfbf)D = r for vahes of r~ varying from 0.5-4.0.
Fracture fluxes at active and observation wells were
different for both high and low conductivity
cases.
Figures 10-14 illustrate early and late time fluxes at
the active and observation wells for both Klgh and
low fracture conductivity csaea and fracture azimuths
varying from 15-90.
oPermeabili@

At late times, the influence of 8 on observation weil


response decr~ses
and becomes negligible for most
cases at about tr=f/rD2 >10. values of ?IJ >4.0 aiso
show minimal r~sponse to 6 at ail times, The effect of
finite conductivity
at the observation well is minimal,
except at extremely early times. Observation well responses show negligible differences for varying values
of [k~b~]Ao, It is therefore impassible to determine the
value of FCD for either well from interference testing.
The active and observation wells v -ue of FCD and Z!
must be determined by independent, active tests.

Anisotropy

PermeabWy
anisotropy can be handled in the semianaiytic solution by appropriate
geometric substitutions. Approximations
for the dimensionless fracture
conductivity y are only approximate for very low values
of conductivity y and very early times. However, substitutions
of the values of ~ and x) into the definitions
of pD and tD=f are, for essentially exact. Solutions for
closed boundaries and interference tests require only
the adjusted geometries. By resealing axes for ~ ancl
x!, identicai wellbore dimensionless pressures are obtained,
Effects
of Different
ductivities

Jkacture

Lengths

and Con-

ranging from 0.5-4.0,

AIAo]

azimuth

dimensionless
loom

angles of 15 90, and


conductivities

Interference

With

Two

Active

Wells

By altering the previous matrix formulation, the effects simultaneow


production from both weils can be
denmnstrated.
Of interest here is the delineation of
when the fracture interference between the welis is of
importance.
Figures 15 and 16 compare the performance of two wells (0 = 15,90, and rD = 1) with low
dimensioniese fracture
conductivities with that of two
iine source weiis. By using the efktive wellbore radius
of the pseudodeady
state behavior at the iine source
weii, similar results are obtained. Therefore, solutions
for the finite conductivity
hydrauikaily fractured well
in a closed rectangular reservoir require oniy superpe
sition of modified iine source image welis.

The effect of the fracture length ratio AIAO1is investigated by keeping the active well at unit length and
varying the length of the observation well. Predicted
responses for

R. N. Horne, and H. J, Ramey, Jr.

Interference
Well

ranging from O.Irr to

With

a Constant

Pressure

Active

Most interference tests are designed for constant rate


behavior at the active weii, Low permeabHity reservoirs may require weeks or months to obtain the desired reservoir information.
In practice, maintaining
constant flow rates for low permeddit y wells during
that length of time is difficult, It is much easier to
maintain constant surface flowing pressure, If pressure
drops are large (as is the case for many low permeability weiis), bottomhole premum drops wiii vary in time
for constant values of surface flowing pressures, These
variations are t ypicaUy small for moderate flow rates,

were evaluated. For each of these examples, the relative fracture conductivities [kfbt]~o are held constant;
this implies changing values for the fracture permeabllity width product. However, this makea no practical
difference over the range of interest because sensitivity to [k@J]Ao is negligible. Values of AIAo] for O = O
and rLI 5 1 + A(AOIwere not considered because the
two fractures would physically overlay each other, Low
fracture angles in which the two fractures were in close
proximity often required increased numbers of fracture
blocks for stabiiity. Uniform flux over a given fracture
block is assumed; fracture blocks were given a maximum size equal to one tenth the distance to the other
well, dD.

Verification

Comparisons,

Example

Problem

A series of verifaction runs were used to ensure the applicability o! these new solutions over a wide range of
values. Comparisons with published results for simpler cases and with simulation respomes were used,
Numericai simulations cordlrmed the semi-anaiytic solution presented here.

Similarities in responses makes it clear that fracture


lengths and conductivities for the two wells cannot be
determined by a single interference test, and must be
determined independently,
This requires tests of sufficiently short duration to avoid interference.
However, the effect of interference at the active well is
not generally large except for small values of 6 and
rIJ. Comparing the curves for varying vahms of AIAO1
shows minimal sensitivity for angles greater than 45.
Caution shouid be exercised in analyzing tests with

Figures 17-19 illustrate the automated solutions using non-linear regression,


The non-iinear regression
was used to find a beat fit to the modeis provided
in the form of tables digitized from the results, The
input data was simuiated Using ~OD = r, @= 45, and

141

hhrferencc

T~ting

of Fifite

= 1. Figure 17 M the result of matchingthe simulated interference data to the line source solution. This
is the typical method of analyzing interference well
tests. The data match reasonably well at late times,
but do not match well at early times. Estimated reservoir parameters are in error by 10~o for permeability
and 18% for the qtcJJ product. No information is obtained about fracture azimuth from this type of analysis. In Mousli et td, u [221solution, the active well is
a uniform flux fracture, while the observation well has
an infinite conductivity hydraulic fiwture, For a similar analysis with Mousli et td. Usolution, the estimated
error associated with the non-linear regression is small
( < 5% for k,d, and q$cth); however, these parameters
are in error by 470, 9Y0, and 21 respectively. The estimates for permeability and dc~h are fairly good. The
fracture azimuth estimate by this techrique is poor.
Varying the initial estimated fracture azimuth did not
alter the non-linear digression estimate.
Figures 18
and 19 illustrate the results for the model developed
in this dissertation.
Agreement is good, with the estimated values of k, dc~h, and O in error by S1% for all
three caaea. Estimated fracture azimuth was 44.6 and
was independent of the initial parameter estimatea.

rLI

Early

Behavior

Observation well responses have been plotted as a function of tD=f/r~ where rD is the dktance between the,
wellbores of the active and observation well scaled by
the active well fracture length. Other plot ting functions were investigated for the time and preesure axea.
None of them were completely succesful in collapsing
all of the responsea. Approximate reductions are obtained for either early or late times,
Figure 20 is the observation well response for r~ = 0.8
for 15<8
~ 90 with the dimensionless pressure (pD)
and dimensionless pressure derivative group (p~) plotted as a function of tD.f/r&
The curves tend to collapse at late times; none of the curves matches the Ei
solution exactly. All derivativea reach a value of 0,5
(corresponding
with pseudo radial flow) at approximately the same value of tD=f/r&
Figure 21 shows
shifted and scaled pressure and time data.
Dimensionless pressure is resealed as:

where dD is the vertical distance (normal to the fracture direction) as defined in Figure 1. Dimensionless
time tDr/ is divided by dD rather than by rD,
This resealing and reshifting also works for different
values of ( kjbi )0 and AIAO1. Unfortunately,
it is impractical for test analysis since the desired parameter,
fracture azimuth (4) appears in both axes, However,
Figure 21 illustrates how this shift is relevant for test
design and understanding,
If an estimated azimuth
can be obtained a priori by another method, the type

Cmductivity

HYd~ulically

Fnactured

Wew

of plot given & Figure 21 indkates a minimum


design to dWerentiate fracture behavior.

(a)

SPE 197S4
teat

Vertical finite conductivity


hydrauUc fracturea
Significantly effect interference test response, Most
Si@ficant r@uh& are for rD <4.

(b) 1[t is pozsible to simultaneously solve for flux distributions at the active and observation wells to
generate both active and observation well respo-.
The active wel pressure response is not sensitive
to the presence of the observation well for rD 22.
At late times, the presence of either a large natural fracture (or joint ), or hydraulically fractured
well can be approximated
by an additional negative skin at the active well,
(c)

Hydraulic fracture conductivity


at the observation well is not as important to observation well
response as is the fracture conductivity at the active well. Significant errors in azimuth estimation
and test interpretation
occur when the active well
conductivity is neglected.

(d) Nothing in the shape of the observation well response indicates that <ither the active weU or observation well is hydraulically fractured.
Thus,
fracture length and conductivity can not be determined by interference teats. Since both fracture lengths are r&-@ad for design or analysis,
both wells must be tested independently.
(e) Different fracture lengths at observation wells and
active wells alter theduration
and magnitude of
fracture interference. Limiting responses for large
and small values of A[~ol can be obtained from
the line source, finite conductivity fracture pairs
as active and observation wells.
(f) For a fixed value of rD, observation well re9ponsebecomes insensitive tQ azimuth for @ 2 50.
Sensitivity to fracture azimuth is independent of
fracture conductivity,
(g) Constant pressure responses at the observation
well can be approximated
by neglecting the effect of the observation well fracture on the active
well. The Laplace space formulation of the solution presented simplifies the required calculations.

D. N, Meehnn, R, N, Home, ~d

SPE 19784

Integrating

and noting
dz=-%

~iKo(fi,.Dj-.,)=.=
@bD,-d

fi=D,

Bessel

ZD~+] -

[I

Di+l

]/

(31)

~o(U)du

77

l-=iti%d

The second integral is simpler as the term to be integrated is ] ~~j + z I V and the relative ~itiom
of the xDj ad ~Di ~ not critical to the fornndation.
So, the final expression is:
(32)
kDj+rDi+ll_

. ~
i

J
o

K@)du

Function
Handling cases for the relative positions of xDi and
z Dj was simplified as more general expressions were
introduced.
hversion

to be integrated

Pr ocedure

Unknowns are solved by inverting the left hand side


mat rix (Equation 21), This matrix is represented as
A-z = b where A is the known (n+l) by (n+l) coefficient mat rix, x is the unknown flux dist ribut ion vector,
and b is the right-hand side vector. Matrix inversion
is accomplished by computing the LU factorization of
the coefficient matrix, Factorization fails if the upper
triangular part of the factorization has a zero diagonal
element, Iterative refinement is performed on the solution vector to improve accuracy. The IMSL subroutine
DLSARGwas used for matrix inversions when iterative
refinement was included. The Stehfest algorithm [M]
is used to invert Laplace transformed
variables into
real space.

11*1.9

,S.+Ko(filzDj-Z1l)dz,

@dZ=

bj-~Di+ll*

xDj

~Djl

-[,:::::;,]

20 requires evaluating the integral of W I


$Dj
- Z I and & I xDj
+ d I frOm xDi tO ~Di+l with
respect to z, Abramowitz and Stegun [21] provide
a closed form infinite series for a similar integral as
follows:

form:

du

~i+O(I Dj-@l

Equation

However, the first equation

-Ko(u)

When this integral is combined for both upper and


lower vahwa (~~i and ~Di+\ ), the fouowing expression
is obtained, after some algebra

-J

Modified

(30)

l/s

the integration limits can be altered accordingly. Regardless of relative magnit udea of @Dj and ZDi, the
resultant integral can be expressed as follows:

the

(29)

u=fiI~Dj-~l

Acknowledgements

Appendix

which can immediately separated into two integrals


and expressed as .f~i+l - ~~i of the same function.
Substituting
for the argument of the integral:

..-. --- --.-fracture width


c
compressibility y
dimensionless wellbore storage coefficient
CD
d
interwell distance (normal to fracture direction)
k
permeability y
K.
modified Bessel function
pressure
P
flow rate
pressure derivative with respect to h of tD#f
;D
s
Laplace transform variable
t~=j dimensionless time, scaled to St
fracture half-length
x!.
Xm
interwell distance (parallel to fracture direction)
L
Laplace transform operation
skin effect
s
Eulers constant
7
integration
variables
u> (
ratio
of
active
to observation well fracture length
AIAOI
viscoait
y
P
porosity
4
Diacritical
dimensionless
fracture
;
dkxetiaation
counters
i} j
skin
8
t
total
w
wellbore
if
Laplace transform of a
.
x, x substitutions
for x in integration
correlating Substitution for pD
PI)

This work represents a portion of the first authors


Ph. D. research which was supported by the Stanford
Center for Reservoir Forecasting and Union Pacific Resources Corporation.

H. J. Rarney, Jr.

is of the
(28)

aDl

143

Interference

Appendix

Improvements

Improving

Fractured

Wells

Solution

The tinal value of the flux is solved by substitution


into Equation 35.

Performance

Seve@ methods can be used to decrease the number of


computations
for solving the matrix given in Table 1.
These include:
(a) Methods

to speed the matrix

(c) Methods to eliminate matrix calculations


Reducing

Matrix

Reducing

re-

q~,
qz .
qs
p

$Dl
%Dz

(33)

L?D3

Many of the terms which are added to the main


Aij terms are independent
of s. These can be
calculated once and added again for each Stehfest
step. This requirea a small increase in the total
storage requirements.

Integral evaluation for the integrals of the form


& Ko(u)du are replaced by 7/2 for values of x
greater than 20.

The n x n matrix components which require integral evaluation actually have only 2n -1 different
integrals. The tlret column and row contain all
the required values.

Values of [kjbf]~o
..- = 1 or AIAol = 1 reduce the
required number of calculatio-oa because many interference terms becomee identical.

For this example case, j = i = 3, and the


are used to solve the case of

equations

(34)

$WD(s) + ~ Aij ~Di


i=l

for i = 1,3. The fourth equation


a flow constraint that

arises directly

:*,=1

from

(35)

i=l

For clarity, these terms are always separated in the


main text. In practice, the matrix order is reduced by
including the constraint of Equation 35 into the matrix. This haa the combined advantages of decreasing
storage requirements,
decreasing computation
time,
and removing zero (0) terms from the main diagonal.
Rewriting Equation 35 for the case of j = 3,
9(3] = 1 -~g(i)

= 1 q(1)

Requirements

have a form which look similar


Ala 1
A23 1
A33 1
0

Building

entirely.

Order

Most of the matrices


tcx
All A12
A21 A22
A31 A32
111

Matrix

Depending of the value ofs, the time required to calculate all of the matrix components in a matrix such
as ~ven in Table 1. may take 20-40 Yo of the total
computation
time if done in a brute force manner.
Numerous simplifications accelerate matrix Ming, including

solution.

(b) Methods to reduce the number of calculatio~s


quired to fill the matrix.

Hydraulically

Calculation

Teeting of Finite Conductivity

q(2)

Early

and

Late

Time

Approximation

At very early times, interference terms are negligible, For the interference case with Ilnite conductivity
wells, the matrix of Table 1 may be reduced in size by
75%. At such early times the active and observation
wells behave independently
and infinite acting solutions apply. Duration of this time may be observed
from Figures 3-9 to depend on@ and rD. Typicai vrdues for the end of the infde
acting period range from
tDzj= 0.01-0.1.

(36)

i= 1

Substituting
into the equations
neous nomenclature,
Allql

+ A12q2+

A13(1 gl

and removing

92)

At late times, flux distributions


for both wells stabilize at dimensionless times on the order of tD=f/r& >
10. These flux distributions
can be retained to calculate wellbore pressures at the active and observation
wells directly. Thh may be done either in the Laplace
or real space formulations.
At late times, the active
well pressure response may be closely approximated
by a line source well with an effective wellbore radius.
The skin associated with the effective wellbore radius
arises both from the value of ( ki bj ) D and the distance
and angle to the interference well.

extra-

+ P = 01

(37)

A2191 + A2292 + A33(1 - ql - q2) + P = x2


A31ql + A32q2 + A33(1 - gl - g2) + P = Z3
Gathering
(All

like flux terms,

- A13)ql + (A12 - A13)g2 + P = c1 - A13(3S)

(A21 - A23)ql + (A22 - 213)q2 + P = Z2 - A23


(A31 - A33)ql + (A32 - A33)q2 + P = Z3 - A33
This can now be expressed in the reduced matrix form
as
(All - A13)
(A21 - A23)
(A31 - A33)

(A12 - A13)
(A22 - A23)
(A32 - 433)

1
1
1

q,
q,
P

~ol - A13
Z~2 A13
ZD3 - A13
(39 1
144
!

References
[1] A. E. Pierce, S, Vela, and K, T. Koonce, Determination of the Compass Orientation and Length
of Hydraulic Fractures by Pulse Testing. Jowmal
of Petroleum Techfiologg, 1433-1438, December
1975,

[12] C. A. Komar and L. Z. Shuck.


Pressure Responses ,YromInduccid Hydraulic Fracturea in Adjacent Wells Within a Petroleum tiervoi~
An
Experiment.
Journal of Petroleum Technology,
951-53, August 1975.

[2] A. Uraiet, R. Raghavan, and G. W, Thomas. Determination of the Compass Orientation of a Vertical Fracture by Interference Tests. Journal of
Petroleum Technology, 73-80, January 1977.

[13] G. H. Locke and W. K. Sawyer, Constant Pressure Injection Test in a Fractured


ReservoirHistory Match Using Numerical Simulation and
Type Curve Analysis. SPE 55Q4, presented at the
50th Annual Meeting of the Society of Petroleum
Engineers held in Dallas, TX. 1975.

[3] H. Cinco-L. and F. Samaniego-V, Determination


of the Orientation of a Finite Conductivity Vertical Fracture by Transient Pressure Analysis. SPE
6750, presented at the 52nd Annual Meeting of
the Society of Petroleum Engineers held in Denver, CO. October, 1977.

The Flow of Homogeneous Fluids


[14] M, Muakat.
Through Porous Media. J. W. Edwards, Inc., Ann

Arbor, Michigan, 1946.


[15] H. C. Ca@aw

and J. C. Jaeger. Conduction of


Heat in Solid$. Clarendon Press, Oxford, 1959.

[4] E, O. Abobise and D. Tiab. Determining Fracture Orientation


and Formation
Permeability
From Pulse Testing. SPE 11027, presented at the
57th Annual Meeting of the Society of Petroleum
Engineers held in New Orleans, LA. September,
1982,

[16] A, C, Gringarten and H, J. Rarney Jr. The Use of


the Point Source and Greens Functions in Solving Unsteady-Flow
Problems in Reservoirs.
Society of Petroleum Engineers Journal, October
1973.

[5] S. Eldest N. Hednoto, and R. Raghavan, PulseTesting of Vertically Fractured Wells!. SPE 6751,
presented at the 52nd Annual Meeting of the Society of Petroleum Engineers held in Denver, CO.
1977,

[17] H. Cinco-L. and F. Samaniego-V, Transient Pressure Behavior for a Well with a Finite Conductivity Vertical Fracture. Society of Petroleum Engi.
neera Journal, 253-264, August 1978.

[61 K. H. Frohne and C. A. Komar. Offiet Mell Test:


An Engineering Study of Devonian Shale Production Characteristics.
Intersociet y Energy Conversion Engineering Conference held at Los Angeles,
CA, August 1982, 829177.

[18] H. Cinco-Ley and H, Z. Meng. Pressure Transient Analysis of Wells WMh Finite Conductivity
Vertical Fractures in Double Porosity Reservoirs.
SPE 18172, presented at the 63rd Annual Meeting of the Society of Petroleum Engineem held in
Houston, TX. October, 1988.

[7] K. H, Frohne and J. C. Mercer,


Fractured
Shale Gas Reservoir Performance
Study - An
Offset Well Interference Field Test. Journal of
Petroleum Technology, 36:2(SPE 11224):291-300,
February 1984.

[19] C. P. J. W. van Kr~ysdijk. Sernianalytical Modeling of Pressure Tra@~ents in Fractured Reservoirs. SPE 18169, presented at the 63rd Annual
Meeting of the Society of Petroleum Engineera
held in Houston, TX. October, 1988.

(8] J. P. Sarda. WeU-Linking by Hydraulic Fracturing Problem of Hydraulic Fracture Orientation,


Bull. SOc, Geel, (France), 26:5:827-831, 1984,

[20] H, Stehfeat. Algorithm 368: Numerical Inversion


of the
of the Laplace Transform, Communication
A, C, M,, 1(13), August 1970.

[91 L. F. Elkins and A. M, Skov. Determination


of
Fracture Orientation from Pressure Interference,
Transactions of the AIME, 219:301-304, 1960,

[21] M, Abramowitz
and L A. Stegun.
Handbook
of Mathematical Functions. Dover Publications,
ninth edition, 1970.

[10] C. A, Komar, W, K, Overbey Jr,, R, L, Rough,

[22] N, A, Mouali, R. Raghavan, H, C!inco-L., and F,


Samaniego-V, The Influence of Vertical Fractures
Intercepting Active and Observation Wells on Interference Tests. Society of Petroleum Engineers
Journal, 933-944, December 1980,

and W, G, Lambert.
Factors that Predict Fracture Orientation
in a Gas Storage Reservoir.
57(5):546-550,
Journal ofPetroleun@chnology,
1971,
[11] C. A. Komar et al, Delineating a Subsurface
Fracture System in a Petroleum ReservoirAn
Journal of Petroleum Technology,
Experiment.
531-37, May 1973,

. .Mu

Matrix formulation

AII
A21

A~,l
Ax
Cll
C21

cm,l
000

A,z
Azz

0,.
...

A~;2
Ax
C12

, .,
Ax
, ,,

C22

,.,

Cm,z ,..

Table 1
for interference

Al,m 1 B,,
/.@m 1 Bz]

B,z
Bzz

::
,,
::
,.

,.,
,,,

B1,~
B2,m

solution,

O
O

TjDAl

x/)1

?fDA2

Bm,m
o

~jDAm

Xl)n

00

JWDA($)

D2

,,

1 Bm,l
m,m
Ax
00
C,,m 0 DII
Cz,m O Dzl

L+,m
D2,m

.:: .
::
. .
c m,m 1 D~,l
o
01

1
1

~jDol
t7jD02

D;

?jDOm

10

~wDo($

m,m

1
0
0

0
0

...

Figure 1: Illustration of nomenclature for interference testing of hydraulically


fractured wells

Active Well, rD =1,0, Fcd = pi, Theta = 15,90

10

0,1

Anslylic Soluti4in
1S Degrees

. . . . . . . 90 Degrees

0.01

().(f)l

0.01

0,1

10

100

DimensionlessTime, tDxf

Figure 2: Active well solutions for rD = 1 and FCD = n


.

t
146

1000

Observation Well, rD = 1.0, Fcd - pi, Theta=

0,001

0.01

JI

0.1

Figure 3: observation

Dimrosionk

15,,,90

10

100

1000

TMC, tDxf/rfY2

well solutions

for

rD

1 and JcD = n

Active Well, rD =2,0, Fcd = pi, Theta = 15,90

10

Analytic SoIuli{n

lst)ef?ecs

. . . . . . . 90 Dcglce$

0,01
0,001

0.01

0,1

10

100

1000

DimensionlessTime, tOxf

Figure 4: .4ctive well solutions for r~ = ~ and ~GD = T

Observation M, rD = 2,0, Fcd = pi, Theta= 15,90

10

0,01

0,001

0,01

0,1

10

100

fxmaluionless
Tree, lDxf/rfY2

Figure 5: observation well solutions for rD = 2 and Fc~ = r

1000

*E

19784

Active Well, rD =4.0, Fcd = pi, Theta = 15,90

10
j
L!
31
j
j

0,1 ,

Analytic Sohuk n

15 Dcgrccs

. . . . . . .

0,01
0,001

0,01

0,1

90 Degrees

10

100

DimensionlessTmc, tDxf

1000

Figure 6: Active well solutions for ?D= 4 and FCD = z

Observation

10 [

Well, rD = 4,0, Fcd = pi, Theta= 15,90


I
i
I

0,1

100

1000

Lme Some

&

0,01
O.o111

0,01

0.1

10

DimensionlessTime, tDxf/rlY2

Figure 7:Observation WCIIsolutions for rD = 4 ancl Fcu = rr

Active Well, rD =0S. Gcd= pi, Theta= 15,90

10

001
*

. . . . . . .

0,01
0,001

0,01

0,1

Sohuif n
Analytic
IS Degrees
90 Degrees

10

100

DimensionlessTime, lDxf

Figurr 8: Active well solutions for rn = 0,5 anti FcD = m

148

1000

.,,

SPE 19?84

Obswation Well, rD = 0.5, Fcd = pi, Theta= 15,90


I

10 I

0,1

n ni
U$U

0.001

0,01

0.1

10

1000

100

Dimcnsionk.w Tmc, IDxVrfX2

Figure 9: Observation well solutions for rk) = 0.5 and FCD = r

Active, Observation Well Fluxes, rD = 1,0, FcD = pi, Theta= 15


I

.-

. . . ..-

IDxf=0!1
ActiveWeU

. . .

Ots6ervNionWell I

-2

. ..-

. . . . . . . .

. .

II
$

-1

1
Dimensionless Distance, XD= x/xf

I
2,

= 0.1,FcI.I= n, 6 = 15,
Figure lo: Ac.tivc and obscrvatim well fluxes at tlj,,.j
iill(l t~ s 1

4ctive, Observation Well Fluxes, rD = 1.0, FcD = pi, Theta= 45


~~

2
~
*. . . .
. . . . . . . . . .. w..- . . . . . .

_
. . .

-2
.1

lDXI = 0,1
Active Well
Obwwlion Well I

Dimensionless Distance, XD= x/xf


Figure 11: Active uncl observation WC1lfluxes at t~tf = O,1,Fal) = n, @ = 45,
Imd ?D = 1

140

spE

lvl~~
.

;
iVellFluxes, rD = 1.0, FcD = pi, Theta= 90

Active, Obser@on

2
~

J
.

. . . . ----

L
_

-2

--

-1

-. ,

. . . . . . . . . . . . . . . . . . . . . ------

tDxf= 10.0
Active Well
Obsewntion Well

TLJ
0
-0.5
0.5
Dimensionless Distance, XD= x/xf

Figure 12: Active and observation well fluxes at tDsj = 10,Fc~ = z, O = !30,
and ?D= 1
Active, Observation Well Fluxes, rD = 1.0, FcD = pi, Theta=

.2A . ....
0

-2

. ..- . . . .
-..

lDxf= 10.0
Aclive Well
Obsefvmion Well 1

-1

...

Dimemionless Distance, XD= x/xf


Figure 13: Active and observation well fluxes at tnr, = lo,~~D = m, 6 = 15,
and ?D= 1

Active. Observation Well Fluxes, rD = 1,0. FcD = 100r)i.Them= 15

2
t~

L_.J
.. .

()
_

..,,

-2
-1

tl)xf = 10,0
Active Well
observation Wctl

-.. .

..

... . . . . . . . . .
. . .

0
1
Dimensionless Distance, XD= x/xf

= 10,Fc~j = 100~, 6 = 15,


Figmv 14: Activo aIMl observation WC1lfluxes at t/J..j
[ml

rl) = 1

150

1000 i

Observation Well, rD = 0.8, Fcd = 2000, Theta= 15...90

10

0.1
I

10

!__._.

-------

.--..J._..

100

10

. . ..

1=10

1000

n ill
..

Delt.st, hrs

Figllre IQ: I-cnfication

semi-analytic

plot

Of nonlinear

regression

0.01
match

of simukttcd

test

0.1

10
1
Dimensionless Time, tDxfhDA2

Figure 20: Observation well solution for r~ = 0.S and

t~

100

1000

FCD = T

solution

sN

10

Observation Well, rD =0.8,

:cd = 2000, A

Ie=l 5--90

15 Degree:

10 Degrees

&
1

0.1

h Line Source

(W?44U8
lDxurtY2)

0.01
0.1

10

103

Dimensionless The, tDxfldD**2

1000
1

Vous aimerez peut-être aussi