Академический Документы
Профессиональный Документы
Культура Документы
Mathematical Preliminaries
The relationship between flow-rate and pressure signals across the
sandface (in the wellbore) can be described as a convolution
operation 1-4 :
Apw(t) = tqSjD(T)Ap ~j(t-T)dT, ..................... (I)
o
where .lpw=wellbore pressure drop and qsjD=normalized sandface flow rate, qsjlq" where qsj =sandface flow rate and q,=a
reference flow rate. For Eq. I, the initial pressure of the formation is assumed to be constant, uniform, and the same as the initial
pressure of the wellbore . .lp ~j(t) in Eq. I is defined as 5
Ap~j(t)=Apf(t)+Apl)(t),
......................... (2)
where o(t) is the Dirac delta function. Apj(t) and Aps are the pressure drops across the formation and the skin region, respectively,
for a constant flow rate q,. The Laplace transform of Eq. I can
be written as
.lP w(s) =sij sfD (s).lP sf (s). . ........................ (3)
For most well tests, the tool (including pressure gauge and flowmeter) is located just above the perforations. However, they could
also be located at any point in the wellbore, including the wellhead. Like the distinction between the surface and downhole flow
rates, a difference also exists between the sandface flow rate, qSj'
and the flow rate at the tool location (measured flow rate, qm) because of storage. This difference can be expressed as 4- 7
Copyright 1990 Society of Petroleum Engineers
.lPw(t) = J
.......... (5)
q dT
(S)[
fljisf(s)
], .............. (6)
1+ (CI q)s2 flji sf (s)
..................... (7)
o
For some well-test conditions, the relationship between the sandface and measured flow rates can be expressed as 89
qsj(t)=qm(t)[l-exp(-at), ........................ (8)
o
The Laplace transform of Eq. 9 can be written
.lP w(s)=s[ij mD(s) -ij mD (s+a)).lpsj(s). . ........... (10)
As Eq. 8 shows, if qm(t) is constant, Eq. 9 will become the solution for the exponential-wellbore-flow-rate case presented by van
Everdingen 8 and Hurst. 9 The Laplace transform of Eq. 9 for the
same case, qmD = I, can be written 10
.lp w(s)=a.lpsf (s)/(a+s) . ......................... (11)
4500
18000
4000
14400
3500
7000
10500 14000
12050
10BOO
e:.
III
oj
7200
drawdown (solid)
2500
2000
i!
3600
buildup (SYDlbols)
-,-_--==_......___.....,.__J..
12~00
-L--""F'=-_ _
0.01
10
0.1
tlme,hr
Fig. 1-Pressure and flow rate for Well A drawdown and buildup tests.
12150
Interpretation Methods
In this section, we briefly discuss the convolution, nonlinear-leastsquares-estimation, and deconvolution methods, which will be used
to analyze the well-test examples.
Convolution. Here, logarithmic and generalized rate convolutions,
as well as modified Homer methods, are discussed.
The conventional multirate ll - 14 (Ref. 14 gives more literature
on the subject) and logarithmic (sandface-rate) convolution 10,15,16
methods are the same if the Riemann sum is used for the integration of the convolution integral given by Eq. 1. For both methods,
one also can use other numerical integration techniques. For the
multirate case, however, it does not make any difference which
integration technique is used because the number of the measured
rate data is small for a large time span, making the integration
timestep large. On the other hand, for the sandface-rate convolution, the flow rate can be measured every second. Thus, a variety
of numerical methods lO ,14-20 can be used to integrate Eq. 1.
In terms of testing procedure, flow rates for a multirate test are
measured at the surface, while pressure is measured at the sandface. In other words, a multirate test basically consists of sequential constant-rate drawdowns during which only transient downhole
pressure is continuously measured and flow rates usually are measured intermediately. During each drawdown, the flow rate has to
become constant rapidly; otherwise, the wellbore storage will
strongly affect pressure measurements. Thus, if the flow rates fluctuate rapidly, the test cannot be analyzed with the multirate procedure. For this situation, one has to use a nonlinear least-squares
estimation (automated type curve) with the model given by Eq. 5
if the wellbore storage is constant. Pressure and flow-rate measurements in the same time span and at the same wellbore location
close to the sandface will minimize problems associated with multirate testing.
Ideally, we would like to know the sandface flow rate to interpret the measured wellbore pressure. If wellbore flow rate is not
measured, other indirect met"ods exist to determine the sandface
flow rate. The first method is to measure the movement of the
gas/liquid interface with an acoustic device. 21-23 The second approach is to apply the mass-balance principle to the wellbore
volume. 24 ,25 The third method is to determine the sandface flow
rate from the measured wellbore pressure 26- 28 with Eq. 4, provided that qm is constant or zero and that the wellbore storage remains
constant for the duration of the test.
The logarithmic convolution can be obtained from Eq. 1 by use
of the logarithmic approximation for t:.Pj as ll ,12 (oilfield units)
Jw(t)=!:.pw(t)/qmD(t)=m[jlct(t,qmD)+b], ............ (12)
where J w is the "reciprocal PI"29-31 or "rate-normalized pressure," 10,15,16 ftct(t,qmD)=[I/qmD (t)llM~(r) log (t-r)dr=log376
12200
12250
-3.2275+0.87S.
For radial flow, a linear plot of J w vs. hct should yield a straight
line with a slope m and an intercept mb from which permeability
and skin can be estimated.
The logarithmic convolution method is simple and easy to use
and is similar to semilog methods in many respects. It performs
reasonably well for a fully penetrated well in a homogeneous reservoir with negligible wellbore storage between the tool and sandface. Thus a diagnostic logarithmic convolution derivative 27 ,32
may help determine whether the use of a radial model is valid for
the convolution interpretation.
Other convolution techniques can be developed for different flow
geometries as a diagnostic tool. Next, we consider use of the generalized rate-convolution method to estimate the reservoir pressure
and to verify the model.
For convenience, let us assume that a well is produced at a normalized rate of qmD until shut-in (or another drawdown). At any
time after shut-in, Eq. 7 can be partitioned as
.r---.
~Pd
/
:'-E-
1000
::
Method
-Log convolution
Deconvolution
Nonlinear estimation
Horner
Superposition
~P
Test
Drawdown
Drawdown
Drawdown
Buildup
Buildup
kH
kv
(md)
82.7
96.3
110.0
83.8
91.7
(md)
64.5
9.8
10.6
NA
11.4
.2.L. -S
11.5 3.0
12.7
15.9
10.5
12.1
3.0
4.8
NA
3.0
oa
'r:"
<l
100
0.01
0.1
10
time.hr
o
where t=tp +~t and ~t=test time. For buildup tests, the afterflow
rate, qmD' becomes too small to be measured after some time. It
is important, therefore, to write the integral in Eq. 14 in terms of
l-qmD' Like the logarithmic convolution, Eq. 14 can be rewritten
pw(~t)=Pi-mfrct(~t,qmD)'
........................ (15)
where P w = wellbore pressure and fret = the generalized-rateconvolution time function, which depends strictly on the system
and its parameters:
-IJ1t[I_qmD(T)~ 'wf(~t-T)dTJ
.................. (16)
If qmD is constant (this may be typical for buildup tests) for the
time interval [O,tp )' Eq. 16 can then be rewritten
frct(~t,qmD)= ~ [ ~wf(~t+tp)
-IJ1t[l-qmD(T)~'wf(~t-T)dTJ
.................. (17)
................................... (18)
procedure has to be used to estimate the initial pressure, permeability, and skin from a modified Horner plot. The extrapolated pressure obtained from the modified Horner method may be more
accurate than what is obtained from the Horner method.IO,IS However, one has to be cautious when the modified Horner method is
applied to a system with a behavior that cannot be depicted by the
simple logarithmic approximation.
For any interpretation method, graphical or nonlinear least
squares, ensuring the validity of the assumed model is crucial because we cannot guarantee a priori that the selected model will be
valid for the interpretation of the system output. Use of diagnostic
tools, such as semilog and log-log derivative plots, may be required
to recognize the model by its subtle features. Semilog and log-log
plotting and the start of the semilog straight line have been used
as diagnostic tools for the last 30 years. In the early 1980's, loglog derivative plotting,33,34 including derivatives with respect to
superposition time 33 ,34 and logarithmic convolution time, 27 ,32
have become an important diagnostic tool. The derivatives with
respect to the Horner superposition {Iog[(tp +At)/~t]} and modified Horner times 10 UmHt) to determine the start of the semilogstraight-line period as a function of shut-in time, producing time,
skin, and wellbore storage have also been presented. 10 Unlike the
log-log derivative methods, these are postdiagnostic (after-nonlinearestimation) techniques.
Nonlinear Least-Squares Estimation. Since the early 1930's,
reservoir parameters have been estimated from transient well-test
data with graphical type-curve procedures that consist of matching
the measured pressure or flow-rate data with type curves derived
from analytical solutions. I,6,7,3S-40 Besides the inherent subjectivity present in the graphical methods, it is practically impossible to
extend the method to cases with more than three parameters. Thus,
it is desirable to use nonlinear least-squares estimation (also called
automated type-curve or history matching), which is normally distributed and unbiased, to estimate parameters from well-test data.
In solving the estimation problem, one seeks a model that fits a
given set of output data and knowledge of what features in that model
are acquired by the data. Evaluation of model features can be done
iteratively in the process of estimation and by the diagnostic tools.
If the uncertainties about the model are resolved by the diagnostic
tools, however, the estimation can be carried out with greater confidence at a minimal cost.
Application of nonlinear least-squares estimation to analyze welltest data has gained growing importance since the early
1970's. 14,40-43 More recently, there has been considerable new interest in the nonlinear least-squares estimation 20,44-49 with constant-wellbore-storage, variable-rate-superposition, or convolution
solutions for the response of the system.
For the nonlinear least-squares estimation, an objective function
(the residual sum of squares), J, is minimized to obtain estimates
by ensuring the best fit between the measured pressure (output) and
the model behavior, which includes the flow rate (input). Thus,
the nonlinear estimation refers to minimizing
I Nm
J(x)=-
i=1
377
7000
Q
......
5250
III
4i
l!
3500
..
1750
0.2
0.4
0.6
0.8
.--f
0
1000
penetration ratio. b
'
2000
3000
4000
5000
dp/dt, psilhr
tained from the measured downhole rate if the wellbore flow rate
varies exponentially.
(t:..Pw)n
-E;:/ (qmD)n-i(flpd)i ,
............. (20)
(qmDh
1 dilPw(t)
flpd(t)=-
ex
dt
rp
1/;N(p)=2-J
dp ....................... (22)
Pi Ph J.l.(p)z(p)
Field Example.
The objective of the interpretation of the following tests is not to
produce numbers from each analysis. Instead, we demonstrate certain salient features of each technique and compare them with converttional techniques. The well-test examples given are well-run field
experiments compared with well tests we usually encounter. In many
instances, the infinite-acting radial flow does not occur during a
well test. Cost or operational restrictions can make it impractical
to carry out a test of sufficient duration to attain radial flow. In
these circumstances, convolution and deconvolution techniques may
be the only approach available for the interpretation of short tests.
For example, well-test interpretation for saturated reservoirs is often
confounded by the presence of a gas cap, which often creates at
least two well-known interpretation problems: the allowance of a
large standoff to inhibit gas coning can lead to very low penetration ratios, and if a well is in direct communication with a gas cap,
the infinite-acting radial-flow period will never occur.
2500
llpw
1...,- 7.
1000
..
.iii
i i
'<",
let
'"0
........
..
\',
40
2000
dpw/dlntH
';~-,>'\.
wi
~
'.
100
"::~_:Pjtsup
1000
......... \'"
~.
1500
"!.
~
<l
.&;
Q,
;:
100
1000 <l
.. '~~
.t',::::;:>
SOD
measured pressure
dpidf,pt
0.01
0.1
dpidlntH
dp,jdlntH
1
10
time, hr
pressure and flow rate, the data acquisition was halted for 10 hours
because of operational problems.
After 10 hours, the recording of pressure and rate resumed again
and the production rate was found to be stable at 13,256 BID. After '" 1 hour of production, a flow profile survey was conducted,
which is shown in Fig. 2 (this profile is slightly different from the
earlier profile 57 because of reinterpretation of the data). As Fig.
2 shows, a few of the bottom perforations were not contributing
to flow. This was not surprising because this formation, in particuiar the bottom zone, has had scaling problems throughout the fi~ld.
After the survey, the well was shut in for a buildup test.
Drawdown Test. As pointed out by Gringarten et al. ,39 the system identification is the first step of the interpretation. It is already
known from the geophysical information (openho1e logs, cores,
well-to-well correlations, etc.) that the formation is mildly layered
with discontinuous shale streaks embedded throughout. From the
production profile survey (Fig. 2) we know that Well A is a partially penetrated well with the penetration ratio, b=hw/h, of 0.49.
Moreover, the initial production profile survey, which was taken
after well completion, indicated that the lower perforations were
contributing to flow. With the exception of the proftle survey, which
clearly provides the zones that are in direct communication with
the wellbore (open intervals), most of the information mentioned
above will be treated as a priori input for the determination of the
well/reservoir model. The fluid and formation properties for this
well are: <1>=0.21, !L=0.86 cp, ct =0.OOOO21 psi -1, rw=0.355 ft,
h=187.5 ft, producing perforations, hw=91.5 ft, production time,
tp=19.0 hours, initial pressure, pi=4,495.0 psi, pressure, Pw at
tp =3, 160.0 psi, and production rate, q at tp = 13,256 BID.
Without any further assumptions for the system, let us evaluate
the features of the model with diagnostic tools. As Fig. 1 shows,
the pressure data do not exhibit any recognizable features of the
system because of the flow rate variations in the drawdown period. Fig. 3 presents the wellbore pressure change, t:.Pw, deconvolved pressure, !:.pd, and its derivatives with respect to In(t) and
fspt (spherical time,jspt = 1/.Jt), and the derivative of ratenormalized pressure, J w (from the logarithmic convolution), with
respect to flet on a log-log plot. The spherical derivative is also included for this case because of a possible hemispherical flow caused
by partial penetration. Of course, this is not a unique set of diagnostic plots. Depending on a priori information about the system,
many different sets of pressure and/or derivative plots can be used.
Note that the derivative of the measured pressure is excluded in
Fig. 3 because it was very noisy (including negative values) after
0.1 hours because of fluctuations in the flow rate.
The log-log plots of t:.Pd and t:.Pw in Fig. 3 do not show any interesting features, with the exception that a unit slope period is apparent on the t:.Pw curve at early times. On the other hand, the
deconvolution and logarithmic convolution derivatives exhibit a possible hemispherical flow period between 0.08 and 0.3 hours, possibly two radial (the second one is the pseudoradial) flow periods
(as shown in the logarithmic convolution derivative plot) between
SPE Fonnation Evaluation, December 1990
0.01
10
0.1
tlme,hr
0.002 and 0.08 (this one is not definite) and 2.4 and 7 hours, and
a long transition period between 0.9 and 7 hours during which the
spherical derivative also indicates that at least the system is changing from a hemispherically dominated flow to a radially dominated flow. The second radial (pseudoradial) flow period from the
logarithmic convolution is not verified by the deconvolution derivative. Nevertheless, these features coincide with the certain characteristics of a partially penetrated well in a homogeneous radial
reservoir, except for the disagreement between the late-time deconvolution and logarithmic convolution derivatives.
The a priori information from other sources and the profile survey have influenced our decision to select certain types of plots
and to observe certain features of the system. We continue to perform the interpretation steps with the assumption that we will modify
our model as analysis dictates. A priori information assists us in
making the system identification decisions and saves time, but it
also deters us from examining all possible diagnostic and plotting
tools (although they could be inexhaustible).
As Fig. 3 shows, the derivative of the logarithmic convolution
looks like a derivative of wellbore pressure with the wellbore-storage
effect. This is expected because of the large wellbore volume (a
rathole with a volume of '" 70 bbl) below the tool. Permeability
and skin estimates from the derivative plot of the logarithmic convolution, with the assumption that the flattening is caused by the
second radial-flow period, are given in Table 1. Note that the estimated skin should be the total skin (St =S/b+Sp ) of the system,
where S=damage skin and Sp=skin from partial penetration.
kH.../k; is estimated from the hemispherical flow period (Fig.
3) as 259 md, where kH and kv are the horizontal and vertical permeabilities, respectively. 58 kH is estimated from both radial-flow
regimes (the first caused by the open interval and the second by
the whole formation) of the derivative of the logarithmic convolution as 39 and 82.3 md. The formation thickness of 187.5 ft (see
Fig. 2) obtained from logs is used to estimate permeability. kv
would be 9.8 md from kH.../k; =259 md and k H=82.3 md.
kv also can be estimated with tDsI/h;'D from Fig. 4 and the expression (see Appendix A of Ref. 57 for the derivation):
4500
4300
~.
?
1.
.,
"
ill
"'"
"""
'" 4100
..
:l.
.;
::I
III
3900
'"
dp./dlnt
3700
Horner
modified Horner
generalized rate convolution
3500
0.01
0.1
10
100
tbne,hr
100
10
108
12
10
time,hr
Buildup Test. Fig. 1 also presents the buildup pressure and aftertlow rate during the buildup test, which was started after about
19 hours of production. As Fig. 1 shows, the measurable afterflow rate period is short ( "" 40 minutes). The missing sandface rate
data could be computed with Eq. 3 as discussed above. As shown
in Fig. 5, however, the wellbore-storage coefficient, C, which is
from the whole wellbore-storage volume and represents the slope
of the linear plot of the sandface rate vs. dp/dt (see Eq. 4), is not
constant. For buildup tests, when the wellbore-storage coefficient
becomes constant, a plot of qm(t) vs. dp/dt should yield a straight
line passing through the origin. Fig. 5 shows that the common
method of obtaining C for the sandface-flow-rate estimation from
the wellbore volume and the compressibility of the wellbore fluid
would not be reliable for this test because of changing wellbore
storage.
The log-log plots of the derivatives of the wellbore pressure with
respect to the Homer superposition time, dpw/d In(tH), and the
multirate superposition time, dpw/dtsup' (with the flow rate measured during the drawdown test) shown in Fig. 6, indicate that after
the wellbore-storage effect, the system slowly approaches a possible radial-flow period. The plot at the upper right shows that the
Homer semilog straight line is not fully developed. This could be
a result of the effect of the short producing time because the multirate superposition indicates a radial-flow period. As explained
above, the time for the start of the radial-flow period from the
derivative of the superposition plot and Eq. 23 can be used to estimate kv= 11.4 md. This value compares favorably with the kv obtained from the spherical derivative plot of the drawdown
deconvolved pressure. The horizontal permeability and skin computed from the same plot are given in Table 1.
The convolution, dJ w/d!lcp and deconvolution, dpd1d In(tH),
derivatives do not show any diagnostic features (Fig. 6). On the
other hand, as in the drawdown case, the derivative of the deconvolution pressure with respect to the spherical time function,
dpd1d!spt, indicates a short hemispherical flow period. The system,
at least, is changing from a hemispherically dominated flow to a
radially dominated flow. Thus the buildup behavior of the system
is similar to the drawdown behavior.
Final Interpretotion and Discussion. So far, we have been concerned mainly with the system-identification problem. At this point,
we have observed from both tests (1) changing wellbore storage,
(2) partial penetration effects, (3) no apparent outer-boundary effects, and (4) a fully developed radial-flow period owing to the entire formation. Moreover, the buildup test without the drawdown
flow-rate measurements (for the superposition) could have given
a misleading interpretation. For this buildup test, the parameters
obtained from the superposition derivative (Fig. 6) are assumed to
be more accurate than those from other techniques because the
radial-flow period is well-defined and the vertical permeability compares well with that from the drawdown deconvolved pressure.
380
4500
/~
3800
!1
1,;
~ 3100
2400
pseudopressure
o
measured rate
-computed rate
1Il
2500
2000
11000
1500
i
=
1000 ~
1:10
1700
500
l000+-------.-------r-----~~----~---L
0.001
0.01
0.1
10
tlme,hr
oI!I 100
<]
0.01
0.1
10
time, br
there is a single point in this curve that would have the correct
hemispherical slope. In other words, the length of the open interval is too large compared with the distance to the lower no-flow
boundary to have a well-defined hemispherical flow period. The
curve of I(dp4s /dfspt ) I in Fig. 8 (the spherical derivative of the
wellbore pressure including the effects of CD =48 and S=4.8) has
a minimum; this is also true for the curves of I(dp25 /dfspt) I (with
CD =25 and S=4.8) and I(dp lOo/dfspt) I (with CD = 100 and S=4.8)
at different times. The spherical derivative of the deconvolved pressure probably becomes flat for a short time period because of
wellbore-storage effect. It must then be by coincidence that the
derivative at this flattening period becomes approximately equal
to the hemispherical slope. A low or high value of the wellbore
storage would yield an inaccurate hemispherical slope. In general,
the hemispherical slope obtained from this flattening period will
be inaccurate. Nevertheless, the spherical derivative of the deconvolved pressure exhibits the true characteristics of a partially
penetrated well.
Fig. 9 presents the Homer, modified Homer, and generalizedrate-convolution plots where time functions are defined as
(t p +l1t)/l1t for the Homer, 10/mHt (Eq. 18) for the modified
Homer, and 10Irct (Eq. 16) for the generalized rate convolution.
As can be seen from Eq. 18,fmHt is a function of the skin, St, and
diffusivity constant, .". We therefore use the final estimates with
a total skin of 15.9 (St=Slb+Sp )' where Sp=6.0 (from Ref. 14
for b=0.49 and h wD =830). Strictly speaking, the application of
the modified Homer method is not valid because the well is partially penetrated. The generalized-rate-convolution time,frct, is obtained from PD given by Eq. A-I of Ref. 57 and the final estimates
of C, S, kH' and kv. The plots given in Fig. 9 are a convenient
way to display and compare the Homer, modified Homer, and generalized rate convolution together. The generalized-rate-convolution
plot, which is a semilog plot of Pw vs. frct yields a straight line
with a slope m (although it was known) and an intercept p* (the
initial or extrapolated pressure). The slope slightly increases after
fret = 100 ( < 1 hour) possibly because the partially penetrated well
model may not be not exact because all perforated zones are combined as a single-zone model and the afterflow rate could not be
measured at late times during the buildup.
Fig. 9 exaggerates the early-time data; in fact, the time interval
between 0 and 1 hour is about 14 log cycles, and it is onl)' two
log cycles for the time interval between 1 and 24 hours. Like other
semilog plots, it is unfortunate that this type of display relies on
the plotting scale. Of course, we could have looked at the derivatives of these plots, as we did for the Homer plot. They may not
be useful for the determination of the initial pressure, however,
which is the main objective of this type of plotting. Fig. 9 also
presents the late-time enlargement. The extrapolated pressure, p*,
obtained from the generalized rate-convolution curve, is 4,496 psi,
which is 1 psi higher than the initial pressure before the drawdown
test. Note that both the Homer and modified Homer methods depend on the existence of a storage-free, infinite-acting radial flow
4500
')
.~
.-
.............
3700
\~
Ul
s::Io
.;
~
III
0.01
0.1
10
"
'\"
2100
10~~~~~~r--r-r~~",--~~~~~~~
.....
..
,.
"...
'\
III
es::Io
measured (symbols)
computed (solid)
"".
".
....
2900
.~
.......
Horner..\.
...
modified Homer
generalized rate convolution
1300
10
1000
lOS
10
10"
lOll
1013
time, hr
tlme.hr
where the exponential constant a=5.3 is determined from the measured flow rate. The constant 2,450 BID is the flow rate before shutin. Fig. 10 also presents the computed (from Eq. 24) flow rates.
Fig. 10 shows that the exponential decline given in Eq. 24 approximates the measured flow rate well up to 1 hour. The flow rate
computed from Eq. 24 is much smaller than the actual values because the flow rate declines very slowly after 1 hour (Fig. 10).
Fig. 11 presents the derivatives of the pseudopressure and normalized pseudopressure with respect to different time functions.
These derivatives indicate that the wellbore pseudopressure is heavily dominated by the wellbore storage and that the system is possibly becoming an infinite-acting radial flow after 10 hours (first
diagnostic observation). The convolution and deconvolution derivatives may not be accurate after 1 hour because the flow rate measurem!!nts or their extrapolation is unreliable. In general, when the
flow rate is undermeasured (less than its true value) or underestimated, its effect will appear as a wellbore storage provided that
the surface flow rate does not fluctuate rapidly. This is apparent
in convolution and deconvolution plots in Fig. 11. Thus, these
derivatives do not indicate any feature of the system earlier than
the Homer derivative. The semilog slope of an infinite-acting radial
flow period from Fig. 11 is 228 psi/cycle, which gives k=0.25 md
and S=ll.l.
The derivative of the deconvolution pseudopressure, with respect
to the spherical time function, dl/ld1dfspt, is also included in Fig.
11 to show whether the pseudopressure might be affected by lost
or plugged perforations. It is known that this well is fully perforated. The spherical derivative also indicates the pronounced effect
of the wellbore storage and possibly the beginning of an infiniteacting radial flow period.
As Fig. 11 shows, with the exception of very few data points
at the beginning, the deconvolved pseudopressures from the constrained deconvolution 19 and {3-deconvolution (Eq. 21) methods
give identical results. The advantage of the {3-deconvolution method
is that it is easy to compute. It can be continued even after the flowmeter data become unreliable below the flowmeter threshold value,
with the assumption that the downhole flow rate declines exponentially during the test. As stated above, this assumption did not work
for this test.
Fig. 12 shows the match of the derivative of the deconvolved
pseudopressure (the constant-rate behavior of the system including the effect of the additional volume) with the constant-welibOrestorage type curves for a fully penetrated well in an infinite reservoir. The parameters obtained from derivatives are used as initial
guesses for this matching. The estimates obtained from this typecurve matching are k=0.26 md, S=11.8, and C=O.OI bbl/psi
( CD = 16). This computed C value is slightly higher than that obtained from the 180-ft wellbore volume below the tool. These parameters compare well with those from derivatives.
Another nonlinear estimation is performed with a fully penetrated well in an infinite radial reservoir for the verification and im-
382
Conclusions
In this paper we applied convolution and deconvolution interpretation methods to two well tests. It is clear from the interpretation
SPE Formation Evaluation, December 1990
wi =
Superscripts
e = model or computed
m = measured
- = Laplace transform of
I = derivative
* = interpreted
Acknowledgments
I thank Schlumberger-Doll Research for permission to publish this
paper and Christine Ehlig-Econornides of Schlumberger Well Services for providing helpful discussions.
References
d = deconvolved
D = dimensionless
f = formation
H = horizontal
H = Horner time
i = initial
let = logarithmic convolution time
mHt = modified Horner time
N = normalized
p = perforated
r = reference
ret = rate convolution time
s = skin
sf = sandface
sl = sernilog
spt = spherical time
383
Nomenclature
b =
e =
C =
f =
h =
J =
Jw =
k =
m =
N =
p =
q=
r =
s =
S =
t =
W =
x=
=
[, =
0/
1/ =
J.I. =
T =
c/> =
1/; =
Subscripts
Author
FIIot .I. Kuchuk is a senior scientist and
program leader at Schiumberger-Doll Research Center In Ridgefield, CT. He
researches fluid dynamics in porous me
dla and performs reservoir testing.
Kuchuk holds an MS degree from the
Technical U. of Istanbul and MS and PhD
degrees from Stanford U., all In petroleum engineering.
384
43. Welty, D.H. and Miller, W.C.: "Automated History Matching of Well
Tests," paper SPE 7695 presented at the 1979 SPE Symposium on
Reservoir Simulation, Denver, Jan. 31-Feb. 2.
44. Rosa, A.J. and Home, R.N.: "Automated Type-Curve Matching in
Well Test Analysis by Using Laplace Space Determination ofParameter Gradients," paper SPE 12131 presented at the 1983 SPE Annual
Technical Conference and Exhibition, San Francisco, Sept. 5-8.
45. Barua, J., Kucuk, F., and Gomez-Angulo, J.: "Application of Computers in the Analysis of Well Tests From Fractured Reservoirs, " paper
SPE 13662 presented at the 1985 SPE California Regional Meeting,
Bakersfield, March 27-29.
46. Barua, J. et al.: "Improved Estimation Algorithms for Automated TypeCurve Analysis of Well Tests," SPEFE (March 1988) 186-96; Trans.,
AIME,285.
47. Guillot, A.Y. and Home, R.N.: "Using Simultaneous Downhole FlowRate and Pressure Measurements To Improve Analysis of Well Tests,"
SPEFE (June 1986) 217-26.
48. Kucuk,F., Karakas, M., and Ayestaran, L.: "Well Testing and Analysis Techniques for Layered Reservoirs," SPEFE (Aug. 1986) 342-54.
49. Shah, P.C. et al.: "Estimation of the Permeabilities and Skin Factors
in Layered Reservoirs With Downhole Rate and Pressure Data," SPEFE
(Sept. 1988) 555-66.
50. Coats, K.H. et af.: "Determination of Aquifer Influence Functions From
Field Data," JPT (Dec. 1964) 1417-24; Trans., AIME, 231.
51. Jargon, J.R. and van Poollen, H.K.: "Unit Response Function From
Varying-Rate Data," JPT (Aug. 1965) 965-69; Trans., AIME, 234.
52. Bostic, J.N., Agarwal, R.G., and Carter, R.D.: "Combined Analysis
of Postfracturing Performance and Pressure Buildup Data for Evaluating an MHF Gas Well," JPT(Oct. 1980) 1711-19.
53. Pascal, H.: "Advances in Evaluating Gas Well Deliverability Using
Variable Rate Tests Under Non-Darcy Flow, " paper SPE 9841 presented
at the 1981 SPE/DOE Low Permeability Gas Recovery Symposium,
Denver, May 27-29.
54. Al-Hussainy, R., Ramey, H.J. Jr., and Crawford, P.B.: "The Flow
of Real Gases Through Porous Media," JPT (May 1966) 624-36;
Trans., AIME, 237.
55. Meunier, D.F., Kabir, C.S., and Wittmann, M.J.: "Gas Well Test Analysis: Use of Normalized Pseudovariables," SPEFE (Dec. 1987) 629-36.
56. Lee, W.J. and Holditch, S.A.: "Application of Pseudotime to Buildup
Test Analysis of Low-Permeability Gas Wells With Long-Duration Wellbore Storage Distortion," JPT(Dec. 1982) 2877-87.
57. Kuchuk, F.J.: "New Methods for Estimating Parameters of Low Permeability Reservoirs," paper SPE 16394 presented at the 1987
SPEIDOE Low Permeability Reservoirs Symposium, Denver, May
18-19.
58. Raghavan, R. and Clark, K.K.: "Vertical Permeability From Limited
Entry Flow Tests in Thick Formations," SPEJ (Feb. 1975) 65-73;
Trans., AIME, 259.
m3
Pa's
m
p.m 2
kPa
kPa- 1
SPEFE
Original SPE manuscript received for review May 18, 1987. Paper accepted for publica
tion March 28, 1990. Revised manuscript received Jan. 19, 1990. Paper (SPE 16394) first
presented at the 1987 SPEIDOE Low Permeability Reservoirs Symposium held in Denver,
May 18-19.