You are on page 1of 10

Pressure-Transient

Analysis for
Horizontal Wells
Flkrl J. Kuchuk, SPE, and Peter A. Goode, SPE,
Schiumberger-Doll Research; Bradley W. Brice, * SPE, and
David W. Sherrard, * * SPE, Standard Alaska Production Co.; and
R.K. Michael Thambynayagam, SPE, Schlumberger Well Services

Summary. In recent years, pres- Introduction where llPwf is the constant-flow-rate re-
sure-transient behavior of horizontal Interpretation of well tests from horizontal sponse of the system, including the effect of
wells has received considerable at- wells is much more difficult than interpre- the wellbore volume below the measuring
tation of those from vertical wells because point where the downhole flow rate, qm'
tention because of the increase in of a considerable wellbore storage effect, the and pressure, P w ' are acquired simultane-
horizontal drilling. This paper pre- 3D nature of the flow geometry and lack of ously. qmD=qm1q is the normalized flow
sents an interpretation method for radial symmetry, and strong correlations be- rate. The Laplace transform of llPwf is
tween certain parameters. Also, zonal var- given as 13 ,14
horizontal-well pressure-transient iations of vertical permeability and shale
testing that is applied to a buildup distribution complicate interpretation.
A well-defined flow period, comparable • llPsf(s)
test from a horizontal well in the Prud- llPwf(s) = ... (2)
hoe Bay field. The complex flow to that of the infinite-acting radial flow peri- 1+ [(Clq)s21lPsf(s)]
od (free from storage and boundary effects)
geometry associated with horizontal of a vertical well, is not apparent for
C is the wellbore storage coefficient re-
wells makes well-test interpretation horizontal wells, largely because most
sulting from the volume below the meas-
difficult. Unique determination of the horizontal wells exhibit partial penetration
uring point or the total volume of the well
effects even when they are fully perforat-
system parameters from pressure ed.I-1O Specific methods have been pro- if the downhole flow rate is not acquired and
data with a short testing time (typical posed to identify flow regimes and their IIp sf is given by
test times for vertical wells) and/or durations under ideal conditions. Although
Reiss II and Sherrard et al. 12 presented
production time is not possible. We
performance and production data from
must run drawdown and buildup several horizontal wells and mentioned in- where oCt) is the Dirac delta function and
tests and acquire the downhole flow terpretation of well-test data, they did not ilps is the pressure drop caused by skin.
rate with pressure to estimate the show how to extend the identification and llPf is the constant-flow-rate response of a
usage of flow regimes to the interpretation horizontal well (Fig. 1) completed in an in-
reservoir parameters accurately. of real pressure-transient tests. finite anisotropic medium bounded above
This paper presents a method for the in- and below by horizontal planes. The hori-
terpretation of well-test data from horizontal zontal and vertical permeabilities are denot-
wells and analyses of pseudosynthetic and ed by kH and kv, respectively. The flow of
real well-test data. a slightly compressible fluid of constant
compressibility and viscosity is assumed
Solutions With and throughout the medium. Gravity effects are
Without Gas Cap or Aquifer neglected. For the formation response, two
In a horizontal well, there is usually con- types of top and bottom boundary conditions
siderable wellbore volume (50 to 100 bbl) are considered. In the first case, both the top
below the measurement point, even if the and the bottom boundaries have no-flow
downhole flow rate is measured or a down- conditions. In the second case, one of the
hole shut-in device is used. The storage boundaries is at constant pressure (the sys-
effect with this additional volume typically tem has either a gas cap at the top boundary
lasts longer than that in a vertical well in the
or an active aquifer at the bottom bounda-
same formation because the anisotropy re-
ry), while the other is a no-flow boundary.
duces the effective permeability at early times
For convenience we call the first model the
to .JkHkV' For a horizontal well, the down-
hole pressure at any point in the wellbore "no-flow-boundary model" and the second
(from the surface to sandface) is given by model the "constant-pressure-boundary
model. " In this paper we assume that in the
Pw(t)=p;- j qmD (T)llPWf(t-T)dT,
t
latter model the constant-pressure boundary
is at the top (the gas-cap case), but the equa-
o
.................. (1) tions may be readily adapted for the case of
an aquifer at the bottom.
'Now at BP Exploration (Alaska) Inc.
.. Now at BP Exploration Co. Ltd. (Scotland) The Laplace transform of the constant-
Copyright 1990 Society of Petroleum Engineers flow-rate response of the system described

974 August 1990 • JPT


4400 10000

4200 8000

No-Flow
or
Constant Pressure Boundary
..
Oi
4000 6000
c
iii
i
." Ii
~
!.. 3800 4000
it
dr.wdown (SOlid) I i '\ ~
3600
buildup (oymbolo) I ~.. '" 2000

No-Flow Boundary
.~
~~--~~---r------.-~~~------+
'-,
0.01 0.1 10 100
time, hr

Fig. 1-Horizontal well model (after Ref. 9). Fig. 2-Pressure and flow rate for the drawdown test and Build-
up Test A.

earlier (formation), I::1Pf, is given by 9 the well and if the influence of that boundary The identification of flow regimes for a
is noticed before any significant effect froIl). transient test from a horizontal well can be
{3[ 00
the ends ofthe well, a second period (hemi- invaluable when used with deconvolved
I::1Pf(s)=--; F(a)+2 j~1 radial flow) will develop. After this period, -pressure (the system influence function).
if the well is very long compared to the for- The transition and duration of each flow
j7rZ j7rZw ] mation thickness, with reasonable vertical period can disclose the fine details of the sys-
XF(~j )cos-cos-- , ........ (4) permeability, a linear flow period may de- tem, particularly if used with proper display
h h velop in the case of two no-flow boundaries. techniques (e.g., log-log, semilog, and pres-
If one of the boundaries is a constant-pres- sure derivative), each of which may exhibit
where ~j=a+[j(7rLv,lh).,jkvIkH p, for sure boundary, the linear flow period will particular features of the system.
the no-flow-boundary case, and by not be observed and a wellbore steady-state Display techniques can be a powerful
pressure controlled by the pressure at the identification tool, particularly if the well-
2{3 [ (j - Vz )7rZ
boundary will be achieved. In the case of bore-storage effect is minimized by meas-
I::1Pf(s)=- F(~j)cos---
a h an infinite reservoir bounded by two no-flow uring downhole flow rate or by using a
planes (parallel to the x-y plane), a third (fi- downhole shut-in tool. The interpretation of
(j- Vz)7rZ w ] nal) radial flow period similar to that ob- a given horizontal well test can be enhanced
Xcos , ............. (5) considerably if geological and geophysical
h served for vertical wells may develop.
methods can independently confirm the pres-
where ~j=a+[(j-Vz)(7rL'hlh).,jkvlkH)2, Interpretation ence and location of the boundaries.
for the constant-pressure-boundary case. In For identification and estimation, the con-
Although analytic solutions and their char- ventional, convolution, deconvolution, and
Eqs. 4 and 5 the pressure is evaluated at acteristic behavior for the constant-flow-rate
z=zw+(rwl2)(1 +.,jkv1kH)' and nonlinear least-squares estimation methods
case are available in the literature, 1-9 it is are used. Because these methods have been
F(v)=J
r 00 du sin 2 u
' .......... (6)
well known that in any real well, the con- presented in detail in the literature (e.g.,
stant-flow-rate condition is difficult to Ref. 15), we will not discuss them here.
o u 2 .,j u2 +v achieve. At early times, wellbore storage Their application to horizontal well-test ex-
(even with a shut-in device or downhole-rate amples (with the solutions presented earli-
where (3= 141.2qp,/(k Hh), a=L'ti,r/>p,c t measurements) masks the well-defined char- er), however, is demonstrated in the next
-;- (0.000263k Hs) , and s is the Laplace acteristics of the pressure response. At late section.
transform variable. times, constructing a complete model for a
Because we are using the line-source ap- given horizontal well while the transient
proximation, Eqs. 4 and 5 are valid only Simulated Example
travels away from the wellbore is difficult
when tD/r~D>25 (see Ref. 9 for the defi- In this section, we analyze transient tests
because of the mUltiple boundaries. More- from a hypothetical well where we used Eq.
nitions). This is an important limitation be-
over, a slight decline of the flow rate as a 1 to generate the pressure response. The for-
cause several minutes may be required to
function of time smoothes the sharp features mation and fluid properties and production
satisfy this condition for large anisotropy.
of the pressure behavior, which usually car- data for this example are kH= 100 md;
Eq. 1 is used with a given downhole flow
rate, qm' to compute downhole-pressure ries information about the system. kv=1O md; S=10; Ly,=500 ft; r/>=0.20;
data with skin and wellbore storage (re- Interpretation of well-test data from a p,=0.80 cp; ct =2.0x 10- 5 psi-I; rw=
sulting from the volume below the rate horizontal well presents two problems that 0.35 ft; h= 100 ft; zw=20 ft; Pi =4,400 psi;
measuring point) for a horizontal well with must be solved sequentially and iteratively. q=lO,OOO BID (before buildup); tp=7,053
and without the gas-cap effect (Eqs. 4 and The first is definition (diagnosis) of the hours; and C=O.OI bbl/psi (for the horizon-
5). Eq. 2 should be used if flow-rate meas- system-i.e., identification of boundaries. tal section of the well).
urements are not available. Geological and geophysical methods may The reservoir is assumed to be of infinite
Distinct flow periods may occur during provide clues about the presence and loca- extent but is bounded by no-flow or
a transient-pressure test from horizontal tion of the boundaries. These sources, how- constant-pressure boundaries at the top and
wells (see the Appendix). The first period ever, may not reveal the interference bottom. Although a synthetic model (the sys-
is characterized by radial flow around the boundaries between the wells. The second tem given by Fig. 1 with known parameters)
well, where there is negligible effect from problem is the estimation of the distance to is used, the measured flow rate, which came
the ends of the well and no effect from the the boundaries from the wellbore and other from an actual well, is used to compute the
boundaries. If a vertical barrier (the bottom unknown reservoir parameters (e.g., perme- downhole pressure, Pw, from Eq. 1. Well-
or top no-flow boundary) to flow exists near abilities and skin). bore storage is also included in the model
JPT· August 1990 975
.. 100
Q.
.Q.
100
,;
~
.
,;
>
"iii "iii
.2: >
...:;; , . ...~
.
'

10 100
0.01 0.1 10 100
0.01 0.1
time, hr lime, hr

Fig. 3-Comparison of derivatives for the drawdown test. Fig. 4-Comparison of derivatives for Buildup Test A.

TABLE 1-CROSS CORRELATION


to simulate the storage effect below the rate the derivative of the wellbore pressure with
OF PARAMETERS measuring point (tool). Although the the wellbore storage.
wellbore pressure is simulated, it is treated Fig. 3 shows that both the logarithmic
kH kv C as if it were real for interpretation purposes. convolution and deconvolution derivatives
(md) (md) S (bbl/psi) Thus, the simulated pressure, Pw' is called become slightly flat between 0.6 and 2
kH 1.000 the measured pressure. hours. It also shows that a long transition
kv -0.668 1.000 period exists between 2 and 50 hours, with
S -0.266 0.619 1.000 Drawdown Test an increasing sl~pe and finally another flat-
C -0.771 0.989 0.800 1.000
Fig. 2 shows the well bore pressure and ac- tening period toward the end of the test.
tual flow rate for a 72-hour pseudosynthetic These features agree with certain character-
drawdown test. Note that the drawdown istics of a horizontal well in a homogeneous
pressure exhibits a semilog straight-line peri- reservoir. As discussed earlier, a horizon-
od between 1.5 and 10 hours. As Fig. 3 tal well can experience a few infinite-acting
shows, however, the derivative of Pw does radial flow periods (a flat derivative on a
not confirm the existence of an infinite- log-log graph), depending on the geometry
acting radial flow period. Consider the first, and the parameters of the system. The ques-
and perhaps most basic, question: how can tions now are whether these two flattening
we identify this system from the wellbore periods result from infinite-acting radial
pressure with and without flow rate? The flow regimes and if they do, which regimes
pressure and pressure derivative unfor- (first and second, first and third, or second
tunately do not show any recognizable fea- and third) cause them.
tures of the system, with the exception that Although the early-time deconvolved
an infinite-acting radial flow period ob- pressure might be affected by the additional
served from the semilog plot is not con- wellbore storage (below the measurement
firmed by the pressure derivative. Next, point), its derivative shows another transi-
"Display techniques system identification is continued with tion period between 0.25 and 0.6 hours with
can be a powerful wellbore-pressure and flow-rate measure- an increasing slope. This suggests that the
identification tool, ments. 15· 1? Although in this paper we use first radial flow period is masked by the
the identification methods described in Ref. wellbore storage and that the two flattening
particularly if the 15, the same task can be performed with the periods are the result of the second and third
wellbore.storage effect slightly different techniques discussed in radial flow.periods. The long transition peri-
is minimized by Refs. 16 and 17. od between these two radial flow periods
Fig. 3 also presents the derivatives of the also supports the same assumption (obser-
measuring downhole deconvolved pressure (dpd1d In t and dPdl vation). Visual examination of derivatives
flow rate or by using a dfsp,r) with respect to In t and !spt (Le., will not answer any further questions about
11-Ji ) and the derivative of the normalized these flow periods, but the derivatives can
downhole shut·in tool." pressure (d! w1dfiet) with respect to fiel' The be used to estimate certain parameters that
convolution and deconvolution enable us to can help to identify essential features of the
look at the behavior of the system without well/reservoir system.
the effects of flow-rate variations. Exami- The slopes for the second and third radial
nation of the spherical derivative is impor- flow periods obtained from the derivative
tant for horizontal wells because they usually of the deconvolved pressure (Fig. 3) are ap-
exhibit partial-penetration effects. This is not proximately 55 and 115 psi/cycle, respec-
a unique set of identification (diagnostic) tively. Note that the slope of the third radial
plots. Depending on a priori information flow period is almost twice that of the second
about the system, different sets of pressure one. This does not violate the assumptions
and/or derivative plots generally can be we made about the flow regimes. It may,
used. The deconvolution and logarithmic however, imply another possible interpreta-
convolution derivatives in Fig. 3 become tion-i.e., the flattenings could be results of
almost identical after I hour. Because of the the first and second radial flow periods, be-
additional wellbore storage, they look like cause the slope of the first period is half the
976 August 1990 • JPT
100
4300

4120

"0
Q. ..
a. 3940
,;
III
~
iii
..f€
::I

3760
~
t a.
Horner:
".
'C
p' = 4376 psi
3580 '.
dpst'dlnt
0 observed
gen. rate convolution: .....
computed
p' = 4400 psi
10 3400

0.01 0.1 10 102 104


100
time, hr time functions

Fig. 5-Comparison of derivatives of the deconvolved and com· Fig. 6-Horner and generalized rate convolution for Buildup
puted pressures for the drawdown test. Test A.

slope of the second period. For now let us the buildup pressure. Fig. 4 also presents .nonlinear estimation is to determine a
assume that they are the second and third the derivatives of the logarithmic convolu- posteriori cross correlation of the estimates
flow periods. With Eq. A-3, the geometric tion (dJ w/dAt) and deconvolution [dp dl (see Ref. 18 for a full sensitivity analysis).
mean of the horizontal and vertical permea- d In t and abs(dpd1d!spt)]. The basic char- The cross-correlation analysis indicates,
bilities (.JkHkV) from the second radial acteristics of these derIvatives are similar to though not absolutely, what can be deter-
flow period is 47 md. kH from the third ra- the derivatives of the drawdown at early mined from a nonlinear estimation.
dial flow period is 113 md (with Eq. A-7). times. Note that the logarithmic convolution Table 1 presents the cross-correlation
The formation thickness is used for this per- and deconvolution derivatives are not avail- matrix for the parameters, which are esti-
meability calculation because the flow re- able after 4 hours because the afterflow be- mated without measured errors. Note that
gime is assumed to be the third radial flow comes zero at that time. Note also that the kv and wellbore storage are strongly cor-
period. With these estimates, kv is 19.8 md. derivative of the shut-in pressure is smooth related. The skin is correlated to both kH
As a result of many different flow periods, compared to the drawdown pressure. These and kv; for large measurement errors (not
the total skin obtained from the analysis of derivatives confirm the findings from the shown here), however, it is correlated to
each flow period is different, as the Appen- drawdown test. kH more than to kv. If the wellbore storage
dix shows, while the damage skin is the resulting from the horizontal section of the
same. The Appendix also shows that the Final Interpretation well is constant, then the storage coefficient
horizontal and vertical permeabilities must For the nonlinear least-squares estimation, can be estimated accurately. On the other
be known to compute the damage skin. De- the measured pressure and its deriyative, the hand, if the wellbore storage is not constant,
pending on the flow period, the standoff dis- deconvolved pressure and its derivative, and it affects the estimation of reservoir param-
tance, well length, and formation thickness other measured and processed data can be eters, particularly kv (the parameter most
should also be known. Using the estimated used as observed system behaviors to match sensitive to measurement errors).
permeabilities in Eqs. A-4 and A-8, we es- the computed response. To avoid perform- Fig, 5 displays a good match between the
timated the damage skins from the second ing convolution during the first estimation deconvolved (drawdown) and computed
and third flow periods to be 12.9 and IS, step, we use the deconvolved pressure as an pressure derivatives. The estimates obtained
respectively. The damage skin obtained observed system behavior because the effect from this match are kH= 100.0 md, kv=
from different flow periods should be the of the flow-rate variation is eliminated. In 9.0 md, S=6.7, and C=0.012 bbllpsi.
same if the analysis is correct. this case the reservoir model (the system These estimates compare well with the ac-
In the absence of a buildup test or any response) is given by Eq. 2. If we were to tual (model-input) values presented earlier.
other test preceding the drawdown test, the use the measured wellbore pressure and its When these estimates are compared with the
next interpretation step is naturally the ap- derivative as an observed behavior and Eq. 1 estimates from the flow regimes (Fig. 3),
plication of the nonlinear least-squares es- for the computed response, we would have which are used as initial guesses, we observe
timation to investigate various aspects of the to perform a number of convolution opera- that the vertical permeability and skin have
model, to improve the above estimates, and tions at each estimation step (iteration), changed the most.
to obtain additional parameters. Thus, the depending on the number of unknown pa- In the following nonlinear estimation, the
buildup test will be analyzed next. rameters. measured drawdown pressure and its deriva-
We now examine various aspects of the tive are used as observed system behaviors.
Buildup Test A model with the nonlinear least-squares es- The model pressure for this case is computed
A 72-hour pseudosynthetic buildup test is timation to ensure that the model satisfies with Eq. I, which is the convolution of the
created after 7,053 hours of production from the estimated parameters and all the assump- measured flow rate and the pressure com-
the· same well. Fig. 2 also presents the tions. Earlier identification indicates that the puted with Eq. 2. Using estimates from the
pressure and afterflow rate for the buildup reservoir model is bounded at the top and deconvolved pressure as initial guesses, we
test. Because of the large total well volume, bottom by two no-flow planes. We assume are able to estimate the actual parameters
the measurable afterflow rate period is that formation thickness, standoff (zw), and (model-input values) after a few iterations.
approximately 4 hours. Fig. 4 presents the well half-length (L'h) are known. The ver- Although the final match is not presented,
derivatives with respect to Homer time tical and horizontal penneabilities, skin, and it was, as expected, a perfect match. Note
[dPw/d In thoro where thor =(tp +.:It)/.:lt] wellbore storage-i.e., the additional that for each parameter, at least one convo-
and multirate superposition time (dPwl storage below the measurement point, be- lution and evaluation of Eq. 2 with Eq. 4
d In t sup ), which eliminates the effect of cause the deconvolved pressure will be must be performed per iteration. The func-
flow-rate variation during the drawdown on used-are estimated. One important step in tion evaluation from Eq. 4 is not trivial.

JPT • August 1990 977


4400

~
o measured
+
_ _ .... (Caool)

computed without rMe Ie.. 2)


, . ..I
"~
computed with rite
, ,
' ............
4200

..... .
iii
a. Ii '
a. .,
..
100
vi .; ,.~"
~
.' ;; 4000 ft·······. U... 'wMllofl.
OJ II)
II)
>
l'! Horner:
...~ a. p'
m
= 4382 psi
= 89.5 psl/eyl. \,
3800
gen. rate convolution
p' = 4402 psi
\..0 '00000
",",
m = 130 psl/eyle
10 3600
1 2
0.01 0.1 10 10 10 103 104 105
time, hr time functions

Fig. 7-Comparison of measured and computed derivatives for Fig. a-Horner and generalized rate convolution for Test B.
Test B.

Nevertheless, this last estimation is impor- than 1 psi after 0.3 hours. It is important buildup test are not precisely accounted for.
tant because in the previous estimation the to look at and compare the impulse response The first reason is particularly true for
deconvolved pressure, which is processed with the deconvolved pressure to determine horizontal wells that experience many differ-
data, is used as an observed system be- whether the postulated (identified) model ent flow regimes. 4 These problems can be
havior. If the model is correct and the initial satisfies the subtle features of the real reser- minimized if we use the generalized
guesses are close to their actual values, we voir. If the deconvolved pressure is unavail- rate-convolution technique IS to estimate
should be able to obtain the final estimates able, the logarithmic convolution or reservoir pressure.
after a few iterations (convolutions). For an Gladfelter 17 deconvolution pressure may Fig. 6 presents the shut-in pressure as a
actual well test, it is certainly true that the also be used if a certain set of conditions is function of Horner time and generalized
estimates may be completely different from met. 15 ,17 The start of each flow period rate-convolution (lotrcr) time. As discussed
the values obtained from the deconvolved given in Refs. 8 and 9 can also be used to in Ref. 15, such a graphic representation of
pressure. A substantial improvement can determine whether the estimated parameters the Homer and generalized rate convolution
and probably will be made in the estimates. satisfy the time conditions. is useful to compare and understand the sys-
We may even have a complete mismatch. We can also apply the same estimation tem's behavior. It is well known that the
In that case, we should return to the system- procedure to the buildup data. As stated Horner method is not useful if the storage-
identification step, possibly through exten- earlier, the behaviors of the drawdown and free, infinite-acting radial flow period is not
sive use of other information such as geo- buildup tests are quite similar, as Figs. 3 and apparent. On the other hand, the general-
logical and geophysical data, which may 4 show. Thus, a nonlinear estimation for the ized rate-convolution method can yield the
provide additional features about the system. buildup test is not performed. In the absence extrapolated pressure at any time, provided
In this iterative process, the geological and of the downhole flow-rate measurements, that the production period is sufficiently
geophysical data used for the model con- we have to perform a nonlinear estimation long. This plot magnifies the early-time
struction may also be altered, with justifi- for the buildup test using Eq. 2. Although data, as does a conventional Horner plot.
cation, to satisfy the observed system it may not be desirable, we may also have Thus, it is often beneficial to display the late-
behavior. to compute the downhole flow rate using the time portion ofthe same graph (upper right
If the estimates obtained from the final es- constant wellbore storage model to perform corner of Fig. 6). As this figure shows, the
timation using the wellbore pressure and the above analyses. Horner curve bends upward even at late
flow rate are satisfactory, the impulse We have now constructed a model with times, which implies that the semilog
response-the derivative of the formation . its parameters, excluding the reservoir pres- straight line is not fully developed. Also, the
pressure, !:..pJ, from Eq. 4 without storage sure. To complete the test interpretation, we extrapolated pressure, p*, obtained from the
and skin effects-of the system should be also must estimate the reservoir pressure Horner plot is 4,376 psi, which is 24 psi
the same as the derivative of the decon- (extrapolated or initial). For the above anal- lower than the initial pressure before the
volved pressure after the wellbore-storage ysis, the absolute pressure change drawdown test (i.e., the model-input pres-
effect becomes negligible. This comparison [!:..Pw=Pw(t)-Pw(t=O)] is used for iden- sure). For most vertical wells, knowing the
is shown in Fig. 5. Note that after the tification and estimation. The initial well- initial or average pressures may help to ob-
wellbore-storage period, the impulse bore pressure, [Pw(t=O)], measured just tain the correct semilog straight line. For
response is almost the same as the deriva- before the test (drawdown or buildup), can horizontal wells, however, this knowledge
tive of the deconvolved pressure. The im- be different from the initial or average reser- may not be helpful unless the flow period
pulse response indicates that the second voir pressure for producing wells. Thus, an is precisely known.
radial flow period, identified from the additional correction to the estimated reser- The pressure obtained from the general-
derivative of the deconvolved pressure, is voir parameters may be required. ized rate convolution is exact. The general-
almost invisible because of the effect of the The Horner or other semilog methods ized rate convolution provides a uniform and
upper boundary. However, the deconvolved have been used extensively for the determi- consistent analysis for verification of the
derivative exaggerates the actual flattening. nation of reservoir pressure. As Ref. 15 model and its parameters using the transient
It also indicates that the first radial flow peri- shows, however, these methods may not well test and past production data.
od is entirely masked by the wellbore work well because the behavior of many
storage and the third radial flow period is wellireservoir systems cannot be simulated Buildup Test B
accurately identified. Fig. 5 shows that the by the simple logarithmic approximation of To show some of the interpretation problems
maximum deviation between the impulse the exponential integral solution and because associated with horizontal wells, let us sup-
response and deconvolved derivative is less flow-rate variations before and during the pose that we run a 24-hour buildup test,

978 August 1990 • JPT


20000

8400
9 5/8" casing 8590' lSooo
.., 8600 i - - - * ' I r - - - - T O P SADLEROCHIT 8631'
Q
iii
~
15.
<l>
GAS c
"0
(ij
8800 F----:------"M:-----GOe 8825'
7" LINER 8960'
f 10000
()
',e 41/ 2" SLOTIED LINER OIL ;I
>
<l>
9000 HOT904S'
-----------. ~
owe 9090'
WATER 5000
9200

2200 3000 3800 4600 5400


Departure, It
0
0 300 600 900 1200 1500
8P. pSi

Fig. 9-Location and completion of Well JX·2. Fig. 10-PI for Well JX-2.

which happens to be conventional wisdom, where ct = reservoir compressibility arid of the same plot is shown in the upper right
after 72 hours of production for the same Vw=wellbore volume. For production corner. The Horner curve bends continuous-
well. As will be shown later, the field ex- wells, ct should be equal to or less than the ly upward at late time. Unfortunately,
ample has similar flowing and buildup time compressibility of the wellbore fluid because neither the Horner (without enlargement)
periods. Fig. 7 presents the derivative of the of the pressure loss and presence of gas in nor its derivative (Fig. 7) is able to show
shut-in pressure with respect to the Horner the production string. Of course, Eq. 7 is the late-time curvature of the Horner plot.
time (measured curve). Note that the deriva- exact if c t is replaced by the compressibility The Horner semilog straight line yields a
tive becomes flat after 6 hours, resulting in of the wellbore fluid, if known. slope of 89.5 psi/cycle, which is 30% less
a horizontal permeability of 166 md (assum- With a lower limit of C=0.08 bbl/psi, ob- than the actual slope of the third radial flow
ing that the flattening of the derivative is a tained with Eq. 7, another nonlinear estima- period and is slightly larger than the slope
result of the third radial flow period). The tion is carried out, as Fig. 7 shows (denoted of the second radial flow period. The sys-
damage skin cannot be obtained from Fig. by "computed without rate, Case 2"). Al- tem's impulse response (Fig. 5), however,
7 with Eq. A-8 because the vertical perme- though the residuals are more or less the indicates that the semilog straight-line period
ability is unknown. same for Cases 1 and 2, Fig. 7 shows that observed in Figs. 7 and 8 is a transition peri-
The change in shut-in pressure and its the match is improved at late times. Note od between the second and third radial flow
derivative with respect to the Horner time that at early times, the model with a low ver- periods. Fig. 8 shows thatp* obtained from
is used for the nonlinear estimation to im- tical permeability and small storage and the the Horner plot is 4,382 psi, 18 psi lower
prove the estimates from the earlier analy- model with a high vertical permeability and than the initial pressure. For most real well
sis for Test B. For this estimation, the model large storage exhibit almost the same be- tests, a 10- to 20-psi error in the pressure
pressure is computed by Eq. 2 with a con- havior. estimate is considered small. It is not unrea-
The estimates for Case 2 are k H =182 sonable to think that the conventional in-
stant wellbore storage (total well volume).
md, kv=1.6 md, S=4.5, and C=0.15 bbl/ terpretation performed for Test B is correct
We assume that the formation thickness,
psi. Although these estimates are better than because the derivative indicates a radial flow
standoff, and other reservoir parameters are
the previous ones, they are still not close to period and because the Horner semilog
known, as in the previous case. As Fig. 7
the true values. Note that the horizontal per- straight-line plot produces the almost-correct
shows, the match between the measured and initial pressure (if known). Fig. 8 also shows
meability is almost the same as the one com-
computed pressures and their derivatives puted from the derivative (166 md). Unfor- that the generalized rate convolution yields
(denoted by "computed without rate, Case tunately, that is a misleading confirmation. the correct initial pressure (actually 2 psi
I ") is not satisfactory and the residuals are Let us assume that during this buildup test higher). Note that the slope increases slightly
large. However, note that the computed the afterflow rate is measured with the well- when the generalized rate-convolution time
pressure is able to capture the system's bore pressure. The measured and computed becomes greater than 80 (less than 1 hour).
global behavior. Nevertheless, it is usually derivatives obtained with the afterflow
difficult to simulate the downhole pressure measurements are also shown in Fig. 7 Discussion
with a constant storage model, particularly ("computed-with-rate" curve). The match It is natural to question whether the stand-
at early times. As shown later, we experi- is excellent, and the computed derivative off (zw or h-z w) and the formation thick-
ence the same difficulty in the interpretation traces the measured data well. The estimates ness can be estimated simultaneously with
of the field example. are kH= 133 md, kv=6.6 md, S=9.2, and other parameters. In principle, it is possible
As stated earlier, vertical permeability is C= 0.01 bbl/psi. These estimated values are to identify no-flow or constant-pressure
very difficult to estimate without wellbore not significantly different from the actual boundaries and to estimate the standoff and
flow-rate measurements because it is strong- values and are considerably improved by formation thickness uniquely, provided that
ly correlated with wellbore storage. The es- just using the afterflow measurements. The the downhole flow rate is measured ac-
timates obtained from this match ("com- identification steps, however, suffer exten- curately or the wellbore storage is constant
puted without rate, Case 1") are kH= 1,962 sively without drawdown-pressure and flow- and known. Drawdown tests are preferable
md, kv=0.002 md, S=O, and C=O rate data. In other words, it is difficult to for the estimation of the standoff and for-
bbl/psi. These estimates are not acceptable identify the system and flow regimes from mation thickness because the wellbore flow
because the wellbore storage cannot be zero, the buildup data because the duration of the rate can be measured accurately. On the
given the finite volume of the well. For most measurable afterflow rate is short compared other hand, the measurable afterflow during
naturally producing wells, at least a lower with that of the wellbore-storage effect. buildup tests could be too short for some
bound for the wellbore storage is given by Fig. 8 shows the shut-in pressure as a cases to minimize the weIIbore-storage ef-
function of the Horner time and generalized
C=ctVw , . . . . . . . . . . . . . . . . . . . . . (7) rate-convolution time. The late-time portion (To Page 1028)

JPT • August 1990 979


3800

' == 3804.4 psi

3600 ~m
-.
-123.8 psi/cycle

'iii :::: ..'.~

." "'.,
Co

~3400

''\. :::I
11/
11/
co
ii.
3700

~'\
3200 36g0
...
o with respect to Horner time 3680-+------------,.--- '.
10
'.".
+ with respect to logarithmic superposition time
3000+-~~TTr---"--~-~~~~--~-~~'~~
0.1 1 10 10 100
0.01 (.p+<'>.) 1<'>.
time, hr

Fig. ii-Pressure derivatives for the Well JX-2 buildup test. Fig. i2-Horner plot for the Well JX-2 buildup test.

Pressure-Transient Fig. 10 presents the inflow performance the second radial. p*, calculated from the
obtained from the pressure and surface-flow- Horner semilog straight line shown in Fig.
Analysis for rate data of the flow tests. Although the data 12, is 3,804.4 psi, which is 40.6 psi higher
Horizontal Wells scatter at high pressure drops and rates, pos- than the initial average pressure before the
(From Page 979) sibly because of a severe gas evolution in the multirate test. We may think that a 40.6-psi
wellbore and reservoir, a linear relationship pressure difference is reasonable because the
fect. The linearized sensitivity analysis in- between pressure drop and flow rate is indi- extrapolated pressure of developed systems
dicates that min[zw' (h-z w)] and hare cated. The straight line does not go through is usually equal to or greater than the aver-
strongly correlated with each other, kv, and the origin because the system most likely did age reservoir pressure. As in Test B, even
C. Furthermore, identification and estima- not reach a steady- or pseudosteady-state though the Horner semilog straight line in-
tion may become a problem if the standoff period during a 12-hour flowing period- dicates an acceptable extrapolated pressure
is too small « 10 ft). i.e., the gas-cap effect might not have com- (it is not far from the initial average pres-
pletely dominated the wellbore pressure be- sure), the calculated permeability and skin
Field Example havior. Nevertheless, the productivity index could be incorrect.
This well-test example for Well IX-2 in the for Well IX-2 from Fig. 10 is 12 bbl/ As stated earlier, interpretation of well-
Prudhoe Bay field was discussed in detail by '(psi-D). test data from horizontal wells is generally
Fig. 11 presents the derivatives of the much more difficult than interpretation of
Sherrard et al. 12 The well was drilled in an
shut-in pressure with respect to the Horner those from vertical wells. Our difficulty in
oil zone of the Ivishak formation and is ap-
time and logarithmic-multirate time (because interpreting the buildup-test data from Well
proximately 120 to 134 ft (the exact location
the multirate test precedes the buildup). It JX-2 is compounded by a prolonged well-
is not known) below a gas cap and 5 to 10 ft
shows that the pressure derivative declines
above a continuous shale zone (Fig. 9). The bore-storage-dominated buildup period and
rapidly after 8 hours. This normally indi-
horizontal section of the well is 1,576 ft long ambiguity about the location of the gas cap
cates a gas cap. The late-time decline of the
(Ly, =788 ft) and is completed with a 4.5- and the well length. Additionally, the pro-
pressure derivative as a function of time for
in. slotted liner. The fluid and formation ducing time was short and disturbed by the
horizontal wells can be somewhat mislead-
properties-and production data for this well six flow tests, during which the flow rates
ing (unless it becomes exactly zero), how-
are </>=0.21; /L=1.3 cp; ct =4.0x 10- 5 ever, because some combinations of well!
were not exactly constant and were meas-
psi-I; rw=0.35 ft; jf=3,746 psi; q= reservoir parameters can give a similar
ured only at the surface.
17,037 BID; and tp =85.5 hours. . appearance-e.g., Fig. 7 in Ref. 9. The General diagnostic steps did not really
The average reservoir pressure at the time derivatives in Fig. 11 do not indicate any provide any definite clues about the exis-
of the test was below the bubblepoint pres- radial flow period. tence or location of the boundaries. Thus,
sure. The producing GOR, however, was In the Horner plot (Fig. 12), a straight line the reservoir model for the nonlinear esti-
the same as the solution GOR. Thus, a is drawn through the shut-in pressure be- mation is essentially based on geological,
reduction of oil permeability caused by liber- tween 9 and 24 hours, but the derivative plot geophysical, and well-log information. The
ated solution gas is to be expected in the res- (Fig. 11) does not indicate a well-defined reservoir is therefore assumed to be infinite
ervoir. Moreover, the effect of permeability radial flow period. Fig. 12 also shows the in the x and y directions and bounded by
reduction, which results from liberated so- late-time section of the Horner plot, which constant-pressure and no-flow boundaries at
lution gas near the wellbore, may appear as indicates that the shut-in pressure bends the top and bottom, respectively. Horizontal
positive skin. downward as time increases. In any case, and vertical permeabilities, damage skin,
According to Sherrard et al. , 12 Well IX- the Horner semilog straight-line period, if wellbore storage, gas-cap location (h-z w ),
2 was shut in for a 24-hour buildup test after it exists, must result from the second radial and completed horizontal length of the well
six 12-hour flow tests, during which the flow period. It cannot result from the third are estimated. As we know from the analy-
pressure and surface flow rates were meas- one because of the gas cap, and it cannot sis of the synthetic data, the estimation of
ured. The pressure was below the bubble- result from the first one because the bottom these parameters may not be unique and may
point pressure of the reservoir during the boundary is near the well. The permeability be strongly correlated.
flow periods and subsequent buildup test. and skin calculated from the Horner plot Before performing a nonlinear estimation,
Unfortunately, each flow test cannot be an- (with Eqs. A-3 and A-4) are 37.2 md and we defined upper and lower bounds for these
alyzed individually because the downhole 4.0, respectively. Note that for the perme- parameters on the basis of a priori informa-
pressure and surface flow rate measurements ability calculation, the well length is used tion (e.g., cores, well tests, and logs) about
were sparse. because it is assumed that the flow period is the reservoir because we are dealing with

1028 August 1990 • JPT


1000

1/1
c..
Q' " .. a drawdown test
o
Q "
Q'
followed by a buildup
test will produce
satisfactory results for
measured
100 computed with L1I2 =746 ft
the estimation of
computed with L1I2=650 ft reservoir parameters
computed with h-z w =134 ft for horizontal wells if
downhole flow rate
0.01 0.1 1 10 and pressure are
time, hr
measured."
Fig. 13-Comparison of measured and computed derivatives for the Well JX-2 build-
up test.

an almost ill-posed estimation problem. The The above matches are least sensitive to Conclusions
bounds are 50<k H<300 md, 0.01 <kv< the length of the well because the shut-in In this paper we applied interpretation tech-
40 md, 250<L v2 <788 ft, 20<h-z w < pressure is dominated by the well bore niques to pseudosynthetic and real well tests
134 ft, and om < C < 0.3 bbl/psi. The dis- storage for a long time. Fig. 13 also shows from horizontal wells, showing that the
tance to the bottom no-flow boundary from the sensitivity of the match when L y, is measurement of downhole flow rate is cru-
the wellbore, zW' is assumed to be 10 ft. changed from 746 to 650 ft while all other cial for system identification as well as pa-
Note that bounds for the skin are not speci- parameters remain the same (as in the case rameter estimation. Use of only pressure
fied because we have no idea what they are. for Lv, =746 ft). Fig. 13 shows that the data, particularly for wells with constant-
Furthermore, the estimated parameters new match has not changed appreciably. pressure boundary andlor wellbore storage,
should yield a PI of -12 bbl/(psi-D), which Note that the computed derivatives are the has major drawbacks for the identification
is obtained from the production data. same at early times for all three cases. The of the reservoir system and subsequent es-
With these bounds, a nonlinear estimation cases for Lv, =746 ft (with h-z w =25 ft) timation of its parameters. Although the
is performed with Eq. 2, the above reservoir and h-z w = 134 ft (with L'/2 =746 ft) be- third radial flow period is essential in es-
model, and the surface flow-rate data from have exactly the same until 1 hour, and then
timating the horizontal permeability in the
the flow tests. In other words, the produc- they diverge at late times. The PI for the
absence of the measured downhole flow
tion history before buildup is partially (be- case of L '/2 =650 ft becomes 12.4 bbl/(psi-
rate, the lack of its evolution in a reason-
cause it is the surface flow rate) accounted D), which is almost the same as the observed
able testing time makes interpretation diffi-
for by superimposing the rate data with Eq. PI. The uniqueness problem is now that the
cult. Large anisotropy ratio and the existence
2. Fig. 13 presents the match of the deriva- case of Lv, =746 ft is a better match and
has a smaller sum of the squares of residuals of multiple boundaries with unknown dis-
tive of the 24-hour measured buildup pres- tances to the wellbore increase the complex-
sure with the model response. Note that the than the case of Lv, =650 ft. On the other
hand, the PI for Lv, =650 ft is much closer ity of the interpretation problems.
match is reasonable except for the mismatch The deconvolved pressure and its deriva-
at late times. The estimates from this match to the observed value than that in the other
case. tive computed from measured downhole
are k H=88.3 md, kv=2.7 md, S=2.9, pressure and flow rate are effective system-
We tried other combinations of parame-
Llh =746 ft (the wellbore half-length), identification tools that can also provide ini-
ters to improve the first match. 19 As in
(h -zw) =25 ft, and C=0.15 bbl/psi. The tial estimates for the nonlinear estimation.
Test B, the shut-in pressure does not have
estimated standoff (h-z w ), the distance to Relying solely on nonlinear estimation
any sensitivity to such parameters as well
the gas cap from the wellbore, is small when length, vertical permeability, and distances without diagnostics may lead to an erroneous
compared with the 135 ft obtained from to the top and bottom boundaries when model and estimates.
openhole logs. Given the history of gas masked by the wellbore storage. The esti- Drawdown and buildup data tend to pro-
underrunning in the Prudhoe Bay field, how- mates from the first match (L '/2 =746 ft) are vide similar reservoir information at early
ever, a short distance to the constant-pres- taken as a best possible set of the estimates. times. At late times, they may provide com-
sure boundary could be justified. Fig. 13 However, they could be nonunique because plementary information about the system.
also presents the match with the same data of the storage effect at early times and mis- The generalized rate convolution, which
and model with a fixed value of h-z w = match at late times. The mismatch may also combines the measured drawdown and
134 ft. This match is very poor compared suggest that the model used for estimation buildup flow rates with the shut-in pressure
with ,the previous one. The new estimates is not accurate. The measure of the unique- for the estimation of the reservoir pressure,
are kH=50.0 md, kv=17.2 md, S=8.1, ness of the estimates from the match cannot works better than the Horner method. Long
Ly, =746 ft, and C=0.17 bbl/psi. The PI's be judged simply by inspecting the sum of buildup tests provide useful information for
obtained from the estimates of the first and the squares of the residuals and the plot of the onset of the third radial flow period,
second matches are 14 and 14.5 bbl/(psi-D), measured data with the model response un- while the same information may not easily
respectively. Although 14 is closer to the ob- less the uniqueness of each parameter is in- be obtained from drawdown tests because
served value, both PI's are practically the vestigated beforehand. As Shah et al. 18 of the small flow-rate fluctuations at late
same, given that the observed PI is obtained point out, a linearized sensitivity analysis times. Nevertheless, a drawdown test fol-
from the transient tests. may indicate uniqueness of the estimates. lowed by a buildup test will produce salis-

JPT • August 1990 1029


factory results for the estimation of reser- 2. Giger, F.M.: "Horizontal Wells Production Appendix-Flow Periods
voir parameters for horizontal wells if down- Techniques in Heterogeneous Reservoirs,"
This Appendix provides the definitions,
hole flow rate and pressure are measured. paper SPE 13710 presented at the 1985 SPE
Middle East Oil Technical Conference and slopes, and skin equations that can be used
Exhibition, Bahrain, March 11-14. to obtain reservoir parameters from the var-
Nomenclature
3. Daviau, F. et at.: "Pressure Analysis for ious flow regimes associated with horizontal
c = system total compressibility, psi - I Horizontal Wells," SPEFE (Dec. 1988) wells. All the slopes and skins relate to cases
C = wellbore storage constant, bbl/psi 716-24. where flow rate is constant. The durations
f = time function 4. Goode, P.A. and Thambynayagam, R.K.M.: of flow regimes are discussed in detail in
h = formation thickness, ft "Pressure Drawdown and Buildup Analysis Refs. 3 through 5, 8, and 9.
J = reciprocal productivity index or for Horizontal Wells in Anisotropic Media, "
SPEFE (Dec. 1987) 683-97; Trans., AIME,
rate-normalized pressure 283.
Two No-Flow Boundaries. First Radial
k = permeability, md 5. Clonts, M.D. and Ramey, H.J. Jr.: "Pres- Flow Period. This radial flow around the
. L = length sure-Transient Analysis for Wells With Hori- well may continue until the effect of the
m = slope zontal Drainholes," paper SPE 15116 nearest boundary is felt at the wellbore. It
p = pressure, psi presented at the 1986 SPE California Regional may not develop because of large anisotro-
q = flow rate, BID Meeting, Oakland, April 2-4. py. The behavior of the period is very simi-
6. Joshi, S.D.: "Augmentation of Well Produc- lar to the early-time behavior of partially
r = radius, ft tivity With Slant and Horizontal Wells," JPT
s = Laplace image space variable penetrated wells. The semilog straight-line
(June 1988) 729-39; Trans., AIME, 285.
S = damage skin slope is
7. Karcher, BJ., Giger, F.M., and Combe, J.:
t = time, hours "Some Practical Formulas To Predict 162.6qp,
o = Dirac delta function Horizontal Well Behavior," paper SPE 15430
mrl = 2.JkH V y, , ..•.••..•. (A-I)
11 = pressure diffusivity, ft 2 /hr presented at the 1986 SPE Annual Technical k L
p, = oil viscosity, cp
Conference and Exhibition, New Orleans,
Oct. 5-8. and damage skin is
T = dummy integration variable
8. Ozkan, E., Raghavan, R., and Joshi, S.D.:
tP = system porosity "Horizontal-Well Pressure Analysis,"
SPEFE (Dec. 1989) 567-75; Trans., AIME, S=1.151 1~Plhr +3.2275
Subscripts 287. L mrl
d = deconvolved 9. Kuchuk, F.J. et at.: "Pressure-Transient Be-
D = dimensionless or normalized havior for Horizontal Wells With Gas Cap
hor = Horner time or Aquifer," paper SPE 17413 presented at
the 1988 SPE California Regional Meeting,
H = horizontal
Long Beach, March 23-25.
i = initial 10. King, G.R. and Ertekin, T.: "Comparative
I = linear
lct = logarithmic convolution time
Evaluation of Vertical and Horizontal
Drainage Wells for the Degasification of Coal ~
-IOg( tPp,c )l, .......... (A-2)
m = measured Seams," SPERE (May 1988) 720-34. trw

p = production 11. Reiss, L.H.: "Production From Horizontal


Wells After 5 Years," JPT (Nov. 1987)
where ~Plhr=Pw(t=O)-Pw(t=l hour) for
r = radial drawdown tests and ~lhr=Pw(~t=l
1411-16; Trans., AIME, 283.
rct = rate convolution time 12. Sherrard, D.W., Brice, B.W., and Mac- hour)-pw(~t=O) for buildup tests, and Pw
s = skin Donald, D.G.: "Application of Horizontal at I hour for both tests is obtained from the
sf = sand face Wells at Prudhoe Bay," JPT (Nov. 1987) first radial flow period on the semilog,
sl = semilog 1417-25. Horner, or derivative plots.
spt = spherical time 13. van Everdingen, A.F. and Hurst, W.: "Ap- Second Radial Flow Period. This hemi-
ss = steady-state plication of the Laplace Transformation to cylindrical flow period follows the first radi-
Flow Problems," Trans., AIME (1949) 198, al flow and may occur when the well is close
sup = superposition time 171-76.
t = total 14. Agarwal, R., AI-Hussainy, R., and Ramey,
to one ofthe no-flow boundaries. The slope
V = vertical H.J. Jr.: "An Investigation of Wellbore obtained from this flow regime is twice that
w = well or wellbore Storage and Skin Effect in Unsteady Liquid obtained from the first period.
wf = well bore flowing Flow: I. Analytical Treatment," SPEl (Sept.
1970) 279-90; Trans., AIME, 249. I 62.6qp,
1/2 = half mr2 =2mrl =,,---;- , ...... (A-3)
15. Kuchuk, F.J.: "Application of Convolution
I = derivative -vkHkVLv,
and Deconvolution To Transient Well Tests,"
* = extrapolated paper SPE 16394 to be published in the Dec. and
= average 1990 SPEFE.
= Laplace transform 16. Joseph, J., Ehlig-Economides, C., and
~Plhr
A

Kuchuk, F.J.: "The Role of Downhole Flow S=2.302) +3.2275


Acknowledgments Measurements in Reservoir Testing," paper
SPE 18379 presented at the 1988 SPE Euro-
l mr 2
We are grateful to Schlumberger-Doll Re- pec Conference, London, Oct. 17-19.
search and Standard Alaska Production Co. 17. Ayestaran, L., Nawaz, H., and Kuchuk, F.J.:
for permission to publish this paper. We are "The Use of Convolution Type Curves for
also grateful to D.1. Wilkinson for his valu- the Analysis of Drawdown and Buildup
able comments. The interpretation of the Tests," paper SPE 18535 presented at the

~ )J, ..........
Prudhoe Bay example is that of the authors 1988 SPE Eastern Regional Meeting,
and does not necessarily reflect the opinion Charleston, WV, Nov. 2-4.
18. Shah, P.C. et at.: "Estimation of the Perme-
-IOg( tPp,ctrw (A-4)
of any Prudhoe Bay owner.
abilities and Skin Factors in Layered Reser-
voirs Using Downhole Rate and Pressure where ~p Ihr has to be obtained from this
References Data," SPEFE (Sept. 1988) 555-66.
1. Giger, F.M., Reiss, L.H., and Jourdan, A.P.: 19. Kuchuk, F.J. et at.: "Pressure Transient
flow period's plots.
"Reservoir Engineering Aspects of Horizon- Analysis and Inflow Performance for Hori- Intermediate-Time Linear Flow Period.
tal Drilling," paper SPE 13024 presented at zontal Wells," paper SPE 18300 presented This flow regime may develop after the ef-
the 1984 SPE Annual Technical Conference at the 1988 SPE Annual Technical Conference fects of the upper and lower boundaries are
and Exhibition, Houston, Sept. 16-19. and Exhibition, Houston, Oct. 2-5. felt at the wellbore if the horizontal well is
1030 August 1990 • JPT
much longer than the formation thickness. Authors
The slope of the linear straight line (pres-
sure vs. the square root of time) is given by

mil = 8.128q
2L'j,h
J J.t , ....... (A-5)
kH<PCt
and the skin is given by

2L V,VkHk v
S= .:lPOhr +2.303
141.2qJ.t
Kuchuk Goode Brice Sherrard

xlogL
rh7rrw (
1+ ~
[4
k;; ) sin (h7rZ w )l' Flkrl J. Kuchuk is a program leader at Schiumberger-Doll Re-
search Corp. in Ridgefield, CT. He previously worked for
Schlumberger in Houston and for Sohio Petroleum Co. in San
............... (A-6) Francisco. His interests include well testing, reservoir model-
ing, and bottom hole-flowing-pressure measurements. He
where .:lPOhr is the intercept. served on the 1983-85 SPE Editorial Review Committee, a 1986
Third (Intennediate) Radiid Flow Period. SPE Annual Meeting technical committee, and the 1986 SPE
After the effects of the top and bottom Well Testing Forum Series Committee. He holds an MS degree
boundaries are felt at the wellbore, a third from the Technic,al U. of Istanbul and MS and PhD degrees from
Stanford U., all in petroleum engineering. Peter A. Goode
radial flow pattern develops in the x-y plane. is a member of the professional staff at Schiumberger-Doll Re-
This period does not exist for wells with a Thambynayagam search Corp. He previously was a senior reservoir engineer
gas cap or aquifer. The semilog straight-line at Sohio Petroleum Co. and at Santos Ltd. in South Australia.
slope is He holds a BS degree in mathematics from the U. of Adelaide. Bradley W. Brice
is a staff reservoir engineer with BP Exploration (Alaska) Inc. in Anchorage. His
162.6qJ.t responsibilities include technical coordination of horizontal drilling efforts at Prudhoe
mr3 = , .............. (A-7) Bay. Recognized as an expert at implementing nonconventional wells, he was
kHh responsible for the design and implementation of the Northwest Fault Block Water-
flood Tracer Program and primary development of the Eileen West End. Brice was
and the skin is the Alaska Section's 1988 Engineer ofthe Year and a 1989 Distinguished Lecturer.
Brice was a reservoir engineer at Shell Oil Co. before 1974. He holds a degree from
S=2.303 [4 Lv, r.:lPlhr the U. of Missouri at Rolla. David W. Sherrard works at BP Exploration Co. Ltd.
~k;; h L m r 3 in Scotland. He previously was a senior reservoir engineer at Standard Alaska Pro-
duction Co. and a reservoir and production engineer in the North Sea and Alaska.
He holds a BA degree in engineering from Cambridge U. and an MS degree in pe-
troleum engineering from Heriot-Watt U. R.K. Michael Thambynayagam is a
manager of interpretation engineering at Schlumberger Well Services in Houston.
He also has worked for Sohio in San Francisco and for BP Petroleum Co. in London.
............... (A-8) Thambynayagam is a researcher, consultant, and coauthor of papers on process,
petroleum, and reservoir engineering. He holds a PhD degree in chemical engineer-
where ing from the U. of Manchester, U.K.

7rrw (r
Sz = -2.303 10gL-h- 1+ ~
(4)
k;; Constant-Pressure Boundary. If one sure at the boundary. Eq. A-lO is valid only
boundary is a constant-pressure boundary when VkH1kv (hILv,)<2.5.

x sin( 7rZ w
h
)l- rL~-;;;Lv,
(k;; ~
and a steady-state pressure is achieved at the
wellbore, then the skin is
51 Metric Conversion Factors
bbl
ep
x 1.589 873
x 1.0*
E-OI = m 3
E+OO = mPa·s
ft x 3.048' E-Ol = m
in. x 2.54* E+OO = em
psi x 6.894 757 E+OO = kPa
psi -1 x 1.450377 E-Ol = kPa- 1
'" Conversion factor is exact.
where .:lPlhr has to be obtained from this
Provenance
flow period's plots.
Eq. A-9 is valid only when VkH1kv(hl
L'/,)<5.
x log
l 8h
7rrw(1+Vk v lk H )
cot (7rZw)
--
2h
Original SPE manuscript, Pressure-Trans-
ient Analysis for Horizontal Wells, re-
ceived for review Oct. 2, 1988. Paper
Other flow regimes may also develop,
depending on the outer boundaries in the x
and y directions (see Ref. 4) and the well + (h-z w )
Lv,
(k;;
~-;;;
l, ......... (A-lO)
accepted for publication May 29, 1990. Re-
vised manuscript received April 26, 1990.
Paper (SPE 18300) first presented at the
geometry. For example, we may have a
spherical flow period if a horizontal well is 1988 SPE Annual Technical Conference and
much shorter in length than the formation where .:lPss = pressure drop difference be- Exhibition held in Houston, Oct. 2-5.
thickness. tween the well pressure and constant pres- JPT

JPT • August 1990 1031

You might also like