Anda di halaman 1dari 6

44s8

Analysis of Pressure Buildup Tests in a Naturally


Fractured Reservoir
G. E. Crawford, SPE-AIME, Exxon Prmiuclimr Research Co.
A. R. Hagedorn, SPE-AIME, Exxon Production Research Co
A, E. pierce, SPE-AIME. Exxon Production Research Co.

Introduction
The press~re drawdown and buildup behavior of naturally fractured reservoirs has been investigated theoretical]y by several authors. 1-3Warren and Rootl depicted
the fractured reservoir as a system of identical, rectangular paralle Iopipeds separated by a regular network of
fractures and derived a pressure-response function for
this system. Two additional parameters were required to
describe the interflow between the granular matrix porosity and that of the interconnecting fracture network. The
standard semilog plot of the buildup-pressure response vs
shut-in time, neglecting wellbore storage effects, is
characterized by parallel straight lines at early and
late times, as shown in Fig. 1. The displacement and
end points of these straight lines are functio.ls of the two
new parameters introduced into the model.
Odehs mode12 for a naturally fractured reservoir and
his predicted pressure response are essentially the same
as those of Warren and Root. However, Odeh concluded
that the effects of fractures would be negligible and the
pressure response for a uniformly fractured reservoir
would be the same as that for a homogeneous reservoir.
Unfortunately. Odeh based this conclusion on calculations using properties of a particular naturally fractured
reservoir for which these effects were negligible. (The
results of these calculations are consistent with those
based on the Warren and Root model,) In addition, the
expression for wellbore-pressure response contained two
terms of opposite sign that were a result of the fracturematrix intertlow. Odeh concluded that these terms tended

to cancel; thus, the fractures could be neglected in cases


of homogeneous fracturing. These results, however, depend on the properties of the fracture-matrix system investigated and cannot be generalized to cover all
homogeneously fractured reservoirs, In a discussion of
Odehs paper, Warren and Root4 published limited data
that shcwed the effects of the fracture-matrix interflow
can be prevalent in the buildup-pressure response.
Kazemi3 approximated a naturally fractured reservoir
with a layered system composed of a thin, highly conductive layer, representing the fracture, adjacent to a thicker
layer with low conduct ivity and high storage capacity,
representing the matrix. Pressure responses were obtained using numerical integration and were compared
with those given by the W.amen and Root model. The
results agreed quite well, with some minor differences
attributed to the removal of the Warren and Root approximation of pseudosteady-state flow between matrix and
fractures. Kazemi concluded that the Warren and Root
model was valid in reservoirs with uniform fracture distribution and with large contrast between matrix and
fracture flow capacities,
Although there is general agreement among these
theories, only limited data have been published that can
be used for their verificat ion. Since most buildup tests are
run with surface shut- in, the characteristic response of a
n~turally fractured system (if it is present) is probably
often masked by wellbore-storage effects. In addition,
shut- in periods are often too short to permit detection of

A previously published model describing pressure-buildup behavior of naturally flaetured


reservoirs was combined }titha nonlinear, least-squares regression technique to analyze
buildup data. The model adequately described the buildup response and was useful for
obtaining effect ive formation permeability in the cases studied.
NOVEMBER.

1976

1295

k,= matrix permeability.

the late-time straight line. Also, even within the


framework of these models, homogeneous-type behavior
can occur. in this paper, field data are presented that are
best explained by assuming the presence of natural fractures. These data, uncomplicated by wellbore storage
effects, were matched by combining a least-squares regression technique with the Warren and Root model for
pressure res~nse. The excellent agreement of the data
with calculated results provides evidence to support the
validity of these theoretical models.

Flow through the reservoir to the wellbore, assumed to


occur only in the fracture network, is described by

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (4)

where
k21, kzu = permeabilities of the secondary porosity
in x andy directions, respectively, and
c-z= total compressibilityy of the secondary
porosity.

Warren and Root Model


Warren and Root considered a dual-porosity system consisting of an intergranular primary porosity, 41, and
a superimposed secondary porosity, dz, that is due to
the presence of natural fractures. The irreg~,ar geometry
of the fractured reservoir was approximated with an array of un ifoml permeable blocks containing the primary
porosity. The orthogonal system of spaces separating the
blocks forms a regular network of interconnecting fractures containing the secondary porosity. Flow can occur
between the matrix blocks and the fracture system, but
flow through the reservoir to the well occurs only in the
fracture network. Th? porosit ies @i and @z, defined as
effective porosities, are related to the matrix porosity,
4,,,, and the fracture porosity, &as follows:

4,=(1 4,
&=of.

162.6 q@ jog~ + 0.351


P,,.f= pi -- ..=.. ___
{
kJl
+ 0.435

)4,,, (1.s,.(.),,,.
.........(1)

_q)j = _@J
d(
/4

(P2

p,)

Ei
[(

r
U(lu)

+ 0.87.s,, 0.87 s*,

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

where
~=.

..+25>
cp,cl + cj+(;

(3)

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

Ei(-(Id

. . . . . . . . . . . . . . . . . . . . . . . ...(2)

This assumes the connate water saturation in the fractures


is negligible.
Warren and Root defined two pressures, p, and PZ, as
being the average pressures in the primary and secondary
porosities, respectively. Flow between these porosities,
assllmcd to be pseudosteady state, is described by
~,(,,

Eq. 4 allows for an anisotropic fracture permeability


because of different fracture conduct ivities. However,
Warren and Root solved Eqs. 3 and 4 for the isotropic
case and included the effects of anisowopic fracture permeabilities as a skin term, s*. Their solution for the
wellbore pressure during drawdown of an infinite reser/oir is

A =

where

@k>!:I(:. dimensionless

parameter
governing interporosity flow

k,
c, = total compressibilityy of the matrix
system;
a = shape parameter, with dimensions of
reciprocal area, that depends on the
geometry of the matrix block, and

, dimensionless par;metcr
relating the storage of
the secondary porosity to that of
the combined system. and

Eq. 5 is an asympt~tic solution valid for


T>lOOwii
7>100

A << l, . . . . . . . . . . . . . . . . . ..(6aJ
.;

ifw<<l

. . . . . . . . . . . . . . ..(6b)

or
~>100forallcascs.

LATE - 71MERSJWY+T

LINE _

TRANSITION

., . . . . . . . . . . . . . . . ..(6c)

A plot ofplrj vs log -rfrom Eq, 5 is characterized by two


parallel straight-line segments. one at early times and one
at late times, connected by a transition pressure response.
For appropriate values of A and OJand for small values of
r [a condition of Eq. 6 must hold) such that Ei ( .Y) = In
( 1.78x), Eq. 5 maybe written as
Pwf = Pi

62f61~@-

rlO~ T +

(0.35 I)

:oo~
1000

EARLY-TIME

+log

STRAIGHT LINE
REGION

100

th+At

10

At
~g.

lBuildup

response predicted

by Warren and Root model,

.]. +0.87s],
(JJ

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

where
s =s,/ S*.

For later times such that the Ei terms are negligible, Eq. 5
JOURNAL

OF PETROLEUM

TECHNOLOGY

may be written as
AOf=l+ -

1626q@
~ ,

[log7 + (0.35]) + 0.87s] .

k2R
. . . . . . . . . . . . . . . . . . . . . . . ....

. . . . . . .

(8)

Eqs, 7 and 8 give parallel straight lines with vertical


separation Ap = m log ( l/oJ). Therefore, the value of o
controls the magnitude of the fractured-matrix behavior.
The locations in time of the end of the early-time straight
line and the beginning of the late-time straight line are
dependent on h, because this parameter appears in the
numerator as a multiplicative factor of 7 in both Ei
arguments.
A pressure-buildup equation can be constructed for the
naturally fractured system using the superposition principle in the same manner as for a homogeneous reservoir.
For a well shut in for time At after production at constant
rateq for time rP, the sand-face pressure, px,, is given by

Ei((i%
)Ei(A:::)%
............
)1}
Ei

A(7P-I-Ar;
OJ(l-(JJ)
(

(9)

Eq. 9 is subject to the asymptotic conditions (Eq, 6) on


both rP and AT. For large values of 7P, the last two Ei
functions are negligible, and Eq. 9 reduces to that found
in Warren and Root-s Appendix B. Under these conditions, the pressure buildup plots as two parallel straight
lines connected by a curved transition region in the same
manner as the drawdown (see Fig. 1). For shorter flow
times, all Ei terms may contribute. When significant,
these terms cause a slight curvature in the early-time
buildup data. In matching the data presented in this
paper, the contribution of the last term was negligible.
The following expression for the skin factor can be
obtained by solving Eq. 5 fors:
S= 1.151

Tc2tp
Pi P1.f _ ]Og
m
+2~2)/J~fr2
(4F
1
+
{
[

,3.*3+o.435~i(++)

sumption if the distribution of shape factors, a, can be


replaced by a single value for the reservoir volume influenced during a well test. Since this volume is generally
small compared with the total reservoir size, the approximation is probably reasonable. Such an assumption is
perhaps no more severe than assuming permeability and
porosity are constant in the vicinity of a well when using
the homogeneous reservoir model to interpret buildup
tests,

Analysis of Data
Data considered in this paper were obtained during a
series of short-term buildup tests of new completions in a
reservoir known from independent! sources to be naturally
fractured. Flow times ranged from 4 to 30 hours. Data
were taken using bottom-hole shut- in a~d high-precision
pressure gauges, Afterflow rates were calculated to
be less than 0.1 percent of the production rate for all
buildup-pressure points recorded.
Conventional Horner plots of pressure data for five
buildup tests are shown in Figs, 2 through 6. Additional
data ale given in Table 1. As these plots show, the
buildup curves appear to have multiple straight-line segments. At first glance, one might choose the second
straight-line region (the line of smallest slope) as being
representative of the formation permeability. However,
in the absence of appreciable afterflow, the early-time
straight line requires explanation. If these data are a result
of damage, the drilling damage would extend up to 100 ft
from the well, which is considered implausible. Severa!
alternative heterogeneous reservoir models were considered in an effort lo explain the three-slope behavior
observed in these tests. Included were radian y varying
formation properties (such as porosity and permeabilityy),
layered reservoirs, boundaries, partial penetmt ion of a
homogeneous reservoir, and a naturally fractured reservoir system. Among the possibilityies considered, the last
appeared to offer the most consistent explanation of all
available data, which included information from logs and
cores.
The Warren and Root model (Eq, 9) was used to
calculate results for comparison with field data, subst ituting the Homer time, 11,,for r,,. All terms were included in
matching these data, since flow times involved were
relatively short and the rariges of values for A and w were
unknown. For longer flow times, such that the last twc Ei
functions are negligible, analysis of the data can be accomplished graphically if the two straight lines are fully
developed in the data. Such a method is described by
Warren and Root. However, for short flow times, as
discussed previously, the early-time straight line may
have curvmme because of influence of the additional
terms in Eq. 9. If this is the case, the graphical analysis
would be inadequate and trial-and-error calcuIatiofl)s
would have to be made varying the four parameters pi, k2,

-%%)]}
(]0)
A small contribution to the skin factor arises from the first
Ei term if the flow times are short; however, the last term
almost always can be neglected. The two skin terms, Sd
and s*, were combined since a distinction between the
skin due to completion and the skin due to anisotropic
permeability cannot be made from a single- well test.
The Warren and Root model is a highly idealized
representation of the actual physical conditions existing
in a naturally fmctured reservoir. The use of a regular] y
shaped matrix block and fracture system is a mathematical convenience. However, this should be a valid asNOVEMBER

, 1976

lAtLElTEST
Test
A

:
D
E

(psi/7ycle)

45.6
233.4
22.7
17.6
72.0

DATA

(:P)

P,.j(At=O)
(psi)

G
8.0
6.1
4.2
2.5

3,430
3,255
3,225
3,537
2,936

(ft)

130
170
130
175
262
1297

A, and 0.3simultaneously. To simplify this task, a nonlinear least-squares regression program was used to vary
these parameters automatically and minimize

,
o

06SERVED
CALCULATE

--

~ = 5.3 hr.
Q = 1465 RBIO

E2 = 233 md
u = 102

P, = 3721 P31

#&f

2,, ~ ,.5

R=

//

/99

,x

3660 [111

I I

111111

1000

3600
o
--

TEST B
06SERVED
CALCULATED

111111

,/

~z = 20md
: = .127
- 1.2 x 104

35WI

?dW

10

g
2
;
&

of calculated with measured pressure buildups


for Test A.

lh = 23.8 h,
q = 61 RWD
P, = 3707 p,

33s0

100

Fig. 2Comparison

~/
,/

,6

300
?

~50Lw~,

111,11

low

1111111

1
1

10

100
~+At

Fig.3-

Comr3rison

of calculated
wit~tmeasured
for Test B,

pressure

3510

;
i
:
3500

~1111111

3490
1( )0

100

Fik4-Comt)arisonof

-th

Discussion of Results

10

+ At

where N is the total number of pressure measurements


considered, and Wj is a weighting function to guide the
searching program tow_arda match.
Starting values for /c2and pi were obtained from the
early-time slope and late-time straight-line extrapolation,
respectively, Starting values for h and w were somewhat
more difficult to determine; but after several matches
were obtained this did not present a problem. Poor starting values for these parameters usually resulted in
straight-line fits to the data. Attempts were made to find
ambiguities in values of the four parameters that would
produce fits of equal quality, Continued searching improved the fit in some cases, but POnew set of values
result@. Values for A and o varied significantly more
than k2 and pi as improvements were obtained. Best fi~s
were based on judgment decisions; however, values of k2
and pf were deemed more important and more reliable
than values of Aand w. Once a good match was found,
several attempts were made to improve it; then, the fit
with the smallest value of R was accepted as the best
match.

Values ofpl, 12, A, and u were obtained for more than 20


buildup tests by matching pressure-buildup data with the
Warren and Root model. The best matches of field data
with calculated results for the five tests illustrated in Figs.
2 through 6 are shown as the solid lines. The calculated
pressure respcmses follow the field data very well.
Matches of other tesls were of similar qualit y.
Calculated pressure responses for Tests A and C (Figs.
2 and 4, respective]) ) do not fit the first data point of each
test. For these tests, as well as some others, improved
matches were obtained by eliminating the initial data
point. Wellbore damage, which is accounted for in the
model only throug~. the skin factor, is thought to be the

buildups

3520

Wi[P,,..(Ar,)
P(Ar,)12.
...........(11)

3440

At

calculated
with measured
for Test C.

rxessure

buildurx

~,

7EST E
0
--

OwEfWED
CALCULATE<>

TEST D
o OS3ERVE!J
-CALCULATED
th = 10.9 hr.

J1l

I 1

%07

111111

11! 111

3360

I
1

10

100

I I 1 1

12(I

~+At

of calculated

with

for Test
1298

measured
D.

[,,

1
1

At

At
Fig. 5--Comparison

1,
10
~+At

pressure

buildups

Fig. S--comparison

of calculated

with

measured

pressure

buildups

-..

cause for these discrepancies. For tests that exhibited this


behavior, the differences between final flowing pressure
and the initial pressure reading (At = 4 minutes) was very
large compared with the pressure differences between
subsequent readings, indicating probable damage, For
exampie, for Test A, the pressure difference between the
final flowing pressure and initial pressure recorded at 4
minutes was 230 psi, which is an order of magnitude
greater than the pressure differences between subsequent
data points. The improvement in the over-all match
brought about by eliminating the first data point for Test
A is shown in Fig. 7. The solid line represents the match
including all points, and the dashed line represents the
best match with the first data point eliminated. The
former match does not fit the data.as well as the latter and
leads to a value of the parameter pi that is significantly
higher than the measured value for initial shut-in
pressure.
The sensitivity of the calculated effective-permeability
value to early-time data is also shown in Fig. 7. Elimination of the first data point reduces the early-time slope
and, thus, increases the calculated effective permeabilityy
from 101 to 233 md. This points out the need for very
accurate early-time data and illustrates the interpretation
problem caused by formation damage,
The only comparison of independently measured values with model results that could be made was that of the
initial shut-in pressure with pi calculated by the model,
With a few explainable exceptions, the calculated values
ofp, and measured initial shut-in pressures agreed within
test accuracy, A comparison of the model-derived permeabilities with core data is considered to be of little
value since core permeabilities obtained in fractured reservoirs are not representative of the effective reservoir
permeabilityy. Measured core permeabilities are, of
course, lower than the model-derived values for these
tests.
I.ittle interpretative use was made of the values for the
parameters A and OJ.Calculated values for fracture spacing based on A are strongly dependent on the assumed
geometry. Such calculations can be. made conveniently
only by assuming the regular block structure described by
Warren and Root. Therefore, calculated values of spac-

ing based on Warren and Roots formula describe an


equivalent block-structured reservoir, which may have
little relationship with existing physical spacing. Fractures that were not interconnected would affect the value
of the shape factor, a, but not in the same manner as those
forming block boundaries. Such fractures could not be
differentiated from connect ing fractures by examining
cores. Calculfited fract ure spacings based on the idealized
model and interpreted values for A are somewhat larger
than those believed to exist in the reservoir.
From the definition of OJ,it is theoretically possible to
calculate a value for fracture porosity, However, in practice, it is difticuit to obtain a value for C2, the fracture
compressibility. Porosity values calculated using maximum and minimum possible values of fracture compressibility, however, bracket porosities estimated from core
data.

Comments on Uniqueness
For a buildup test in which both early- and late-time
strajght lines are well defined, a unique set of values for
pi, k2, A, and w can be determined. However, in an actual
buildup test, complete development of both straight lines
seldom occurs without special test design. In such cases,
more than one set of values may exist. In the analyses
described in this paper, some attempts were made to find
more than one set of values that would give equivalent
matches of the same data. Although no equivalent
matches were found, it is felt that they probably exist.
The match shown in Fig. 3 for Test B suggests such a
possibility. This match yields a value ofp~ that is about 60
psi higher than the measured initial shut- in pressure.
Although the initial shut- in pressure for this test is sonlewhat questionable, it is possible that the transition region
between the two straight lines extends farther to the right
than predicted by the match (in the region where no data
are available). If th is is the case, a different set of parameter values might exist that would also match the data and
result in better agreement with the initial shut-in pressure.
It is_believcd that if other sets of values do exist, the value
of k2 would be nearly the same for all sets since the
early-time straight-line data should be sufficient to determine its value; however, w, A, and pi might be changed
significantly.

Suggested Test Procedure


0

063 ERVED
CALCULATE
TIME 1$[
DATA PT.
h
IHRSI
(WI
.06
3734
-_,
i3
3721

TQ

(md

101 .032
.102

233

2.6 x 10+
2,1 x 10.6

/
/

LLI!I

11, !,,

11111

10

100

~+bt

At

Fig, 7Comparison

NOVEMBER,

1976

of matches for Test A with and without first


data point.

In reservoirs where natural fracwres are prevalent, the


early- and late-time data are most important. Misinterpretation can occur, for example, if the early-time straight
line is masked by wellbore-storage effects, In such a
case, the slope may be mistakenly taken from the
transition-pressure region between the straight lines, and
the computed permeability and skin damage will be
higher than actual. If insufficient data are available to
indicate the existence of the late-time straight line, the
transition-pressure region could be mistakenly extrapolated to give a much lower reservoir pressure than that
provided by the initial shut-in measurement, thus giving a
false indication of depletion. To avoid these pitfalls, the
following procedures are suggested for tests of naturally
fractured reservoirs:
1. Use bottom-hole shut-in to minimize wellborestorage effects and thus avoid masking the early-time
straight-line development.
1299

s = ~d-s*, total skin factor


$d = skin due to completion
S* = skin due to anisotropy

2. Use sufficient shut-in time to define the late-time


straight line. This time has been found to be independent
of flow time arsd must be determined empirically for each
reservoir. For the first test in a reservoir suspected of
being natumlly fractured, a minimum shut-in time of 30
to 36 hours is recommended.
3. Use precision pressure gauges to insure that effects
resulting from natural fractures are detected.
Problems in analysis may arise when boundaries influence the late-time buildup data. If possible, initial tests in
a reservoir should be conducted in wells removed from
known or suspected reservoir limits.

+1
(~*.~

Conclusions

The following conclusions are based on these analyses of


buildup tests in a reservoir known. to be naturally
fractured.
1. The multislope behavior of the pressure-buildup
curves presented in this paper results from the presence of
natural fractures.
2. The Warren and Root model for naturally fractured
reservoirs adequately describes the buildup response and
is useful for obtaining initial formation pressure and
effective formation permeability.
3. The parameters A and w derived from the model
have only limited value for calculating fracture spacing
and porosity.
4. Use of bottom-hole shut-in, precision pressure
gauges, and extended shut-in periods are beneficial in
obtaining pressure-buildup data for interpretation of
naturally fractured reservoi~s.

Nomenclature
Z1= formation volume factor, RIYSTB

whereK=S

)
2fi
SW,= connate water saturation
f = time, hours
rh = Homer time = <cumulative
production/last rate) x 24, hours
tp = length of flow period, hours
WJ= weighting function forjth term in R
a = geometric parameter (shape factor)
describing flow between matrix
blocks and fractures ( l/sq ft)
A = ~@2
= interporosity flow parameter
K = viscosity, cp
7=2.637

12(

x 10-4
(4IC

1 +

4w2)P~u.2

dimensionless time
7P = dimensionless production time
shut-in
time
AT= 7 TP, dimensionless
O],

4P=

Primav

and

rjf = fracture

secondary

porosity

porosity

@,,, = matri;isrosity
.0 =

parameter
@,c-,

ox

storage

relating

of secondary

porosity

to total

storage

Units apply to all equations in text except Eqs. 3 and 4,


where Darcy units are applicable.

Acknowledgment

Cl,cz = compressibilities of primary and


We wish to thank the management of Exxon Production
secondary porosities, l/psi
Research Co. for permission to publish this paper.
h = formation thickness, ft
k ~= matrix permeabilityy, md
.
References
ZZ= average effec@e formation
1. Warren.J. E.and Root, P, J.: The Behavior of Naturally Fractured
permeability (k2 = a
for
Reservoirs. Sot. Pet. Eng. J. (Sept. 1963) 245-255; Trans..
AIME. 228.
an isotropic ~ase), md
2. Odeh, A. S.: Ursteady-S!a[e
Behavior of Naturally Fractured
m = 162.6 q~B/k# = slope of straight lines,
Reservoirs. Sot-. Pet. Eng. J. (March 1965) 60-66; Tram., AIME,
psilcycle
234,
pi = initial reservoir pressure, psi
3. Kazeml, H.: Pressure Transierr~ Analysis of Naturally Fractured
Reservoirs Wi[h Uniform Fracture Dismibu[ion, Sot. Pet. Eng. J.
p(Arj) = measured wellbore pressure at time A(J,
(12ec. Igrjg) 451-462; Tram., AIME, 246.
psi
4, Warren, J. E. and Root, P. J.: Disclmsion on Unsteady-State
p,,., = flowing wellbore pressure, psi
Behavior of Naturally Fractured Reservoirs. Sot. Pef. E@. J.
pw, = shut-in wellbore pressure, psi
LJPT
(March 1965) 64-65; Tram.. AIME, 234.
P,~,(Ar~)= shut-in w~llbore pressure calculated at
Atj from Eq. 9, psi
Or,gwd manmcr?il recewed nnSociety of Pet folevm Engmeem of f$ce Avg 2.1973
q = flow rate, STB/D
Paper acceptecl for Dubllcatlon May 24. ]974
Revised manuscr,pl recewed Aug 31
1976 Paper (SPE 4558 I was fIrsl Presented at the SPE.AIME 481h Annual Fall Meeting
rlc = wel!bore radius, ft
held m Las Vegas, Sepl 30.Ott
3. 1373, G5 Copyr,ght 1976 Amerrcan Inst,tute of
Mmmg. Metallurglca.
and Petroleum Engmews. Inc
R = sum of squares objective function

1300

JOURNAL

OF PETROLEUM

TECHNOLOGY

Anda mungkin juga menyukai