You are on page 1of 10

Transient Analysis of Acoustically

Derived Pressure and Rate Data


c.s. Kablr*, SPE, Schlumberger Perforating & Testing Center
F.J. Kuchuk, SPE, Schiumberger-Doll Research
A.R. Hasan, SPE, U. of North Dakota

Summary. A pressure-buildup test conducted on a sucker-rod pumping well is often distorted by long-duration wellbore storage.
In fact, this distortion could be so severe that even a week's shut-in period may not allow a semilog analysis. A longer shut-in
period becomes economically discouraging because of lost production.
Low energy and low transmissivity in the reservoir, coupled with increased fluid compressibility, contribute to this long-duration
storage phenomenon. One way of reducing the storage effect clearly lies in the simultaneous analysis of downhole pressure and
flow rate, estimated from casinghead pressure and rising annular liquid-level measurement made by acoustic well sounding (AWS).
Ascertaining the quality of the indirectly measured pressure and rate data constitutes one of the objectives of this study.
Several methods exist to translate the AWS measurement to downhole pressure and rate data for the subsequent transient
analysis. We show that even an empirical hydrodynamic correlation provides satisfactory transient-pressurelflow-rate data for
convolution and deconvolution analyses for moderate pumping-liquid columns. When long annular liquid columns are encountered,
translating the AWS measurement with a mechanistically based hydrodynamic model appears to be a prudent approach.
Interpretation of several transient tests show that automated convolved-type-curve or history matching of field data is a powerful
tool for reservoir-parameter (total mobility, skin, fracture half-length, and storage coefficient) estimation. Use of downhole rate for
the convolved type-curve matching reduces the variable storage coefficient by several orders of magnitude, thereby enhancing the
quality of the estimates. Deconvolution of downhole pressure and rate data and the pressure-derivative approach significantly aided
in well/reservoir flow-model identification.
A simple algorithm for computing the Laplace transform of the wellbore pressure for an infinite-conductivity vertically fractured
well in an infinite reservoir is developed in this work for a rapid, iterative-type computation used in automated convolved-type-
curve analysis.

Introduction
Testing pumping wells presents an interesting challenge because column (gas percolation effect) and compression of the gas column
of the need to deal with indirect measurements of downhole pres- by the liquid column.
sure and flow rate. Because of economic considerations, direct pres- The varying wellbore-storage problem can be handled in two
sure measurements, after the tubular hardware is removed, is not ways. One approach is to analyze the transient-pressure and flow-
commonly practiced. Thus questions often arise regarding the va- rate data 8- IO simultaneously with convolution and deconvolution
lidity of indirect measurements and their subsequent interpretation- approaches. The alternative is to adapt the unconventional test
leading to estimation of permeability, skin, average pressure, frac- method II and to perform interpretation with either pressure and
ture half-length, etc. rate data simultaneously or pressure data alone.
There are two approaches for estimating downhole pressure and The other alternative approaches that we explored are deconvo-
flow rate. The first method involves continuous measurement of lution with constraints l2 and the automated type-curve analysis.
casinghead pressure together with tracking the movement of the Recently, the automated type-curve approach was used for
gaslliquid interface during a transient test, using an AWS homogeneous reservoirs 13 and naturally fractured reservoirs 14
device. 1-3 Translation of these measurements to downhole pres- without considering the downhole-rate data. In other words, down-
sure and rate is made with a wellbore hydrodynamic model/corre- hole rate is predicted by the constant-storage model. Subsequent-
lation. 1,4-6 The second approach uses the mass-balance principle 7 ly, a number of studies used the automated convolved-type-curve
to accomplish the same task. matching to interpret tests from homogeneous reservoirs, 15 fault-
Both methods have certain limitations. For example, the AWS ed reservoirs, 16 low-transmissivity reservoirs, II and layered reser-
method is not particularly suitable in a foaming annulus. The major voirs,17 involving the downhole flow-rate data.
apparent uncertainty in the AWS method stems from the use of The purpose of this study is two-fold: to search for a suitable
hydrodynamic model/correlation, however. On the other hand, any wellbore hydrodynamic model to generate the downhole pressure
leak in the system (e.g., a wellhead leak) or wellbore crossflow data and to search for an appropriate reservoir flow model (p D
during well shut-in can cause problems in the second method. model) for interpreting the test data. In addition, we wanted to ex-
In a typical pumping well, one is confronted with a long test be- plore the quality of indirectly measured flow-rate data through
cause of long-duration wellbore-storage distortion period. Partic- deconvolution analysis because the technology for direct measure-
ularly, low reservoir transmissivity and high system compressibility , ment oflow flow rates (less than 100 RB/D [16 m 3 /dD, prevalent
precipitated by high gas saturation in the reservoir and/or in the in pumping wells, is not readily available.
wellbore vicinity, contribute to this long-duration storage problem.
The problem is compounded further by low reservoir energy as- Mathematical Models
sociated with pumping wells. Continuously-Varying-Rate Case. The relationship between in-
In general, if the wellbore-storage coefficient remains constant, directly measured downhole transient-pressure and flow-rate data
type-curve analysis can be applied to pressure data directly. Un- in a pumping well can be given by the convolution integra1 18
fortunately, in pumping wells, the early-time data reflect a chang-
ing storage coefficient because of the changing density in the liquid tD
PwbD=Jr qb(T)[PD(tD-T)+S]dr, ................... (1)
• Now with Schlumberger Well Services.
o
Copyright 1988 Society of Petroleum Engineers provided that q D(O + ) = O.
SPE Formation Evaluation, September 1988 607
1000

900
8
...
~
800

.'l 700
fiI
600

0
G:
500
_ Rise in gas/o11 interface (At>O) '? 400

........
III
~
-I-f--- - Initial gas 1011 interface 300

~ PUMP i~
a.
200

100

0
10-' 10' 10'
Arbitrary reference Shut-in Time. Hours
-------~----------------
point (Pwb. qwb evaluated)

_ Pwb ' - - -

== '\. ==·Psf
Fig. 2-Pressurelflow-rate history for Well 1.

LJ Fig. 1-Definltions of various pressures In Eq. 2.


ly measured pressure with rate, search for the storage coefficient
is made nonetheless in the automated or computerized type-curve
matching, along with permeability and skin. Typically, the value
of the storage coefficient is orders of magnitude smaller than the
values normally obtained without using the rate. In this respect,
we observe that the automated type-curve matching with downhole
rate l5 ,16 is superior to logarithmic convolution8 and other
Eq. I implicitly assumes that no storage volume exists beneath methods 9 because the storage effect below the point of measure-
the point of rate and pressure measurement. This assumption, how- ment is considered in the automated convolved-type-curve approach.
ever, is not valid because of dead wellbore volume below the per- Unlike logarithmic convolution, where generally only a segment
forations, storage volume below the point of measurement, and of the data can be used, use of the entire data set is made in
storage in the fracture if the well intercepts an infinite-conductivity convolved-type-curve matching.
fracture. 10,19 Even though the measurement reference point could Alternatively, one could search for a PD model and then its pa-
be made at the perforations in a pumping well, strictly speaking, rameters, such as permeability and skin, by using the deconvolu-
the flow measurement is required within the perforation tunnels to tion method. 1O Thus, deconvolution aids identification of weIll
eliminate the storage effect in a nonfractured well. Thus we can reservoir flow geometry. For the reasons discussed above, how-
largely reduce the effects of storage on transient-pressure behavior ever, deconvolution does not yield completely storage-free data for
by either measuring or deducing downhole rate. Fig. I illustrates conventional interpretation; the storage effects are only reduced to
this point schematically. a large degree.
Assuming the system to be linear, one can write Eq. I to incor-
porate the storage below the point of measurement as follows: Constant-Wellbore-Storage Case. One can approximate down-
hole flow rate, assuming constant wellbore fluid density, as 18,20
tlpsj=f:..Pwb+flpwj, ................................ (2)
dPwD
where qD(tD)=I-CD--, ............................. (3)
flpsj = pressure drop at true sandface, within reservoir, dtD
f:..pwb = pressure drop at point of measurement, and
where CD=5.615C/21r4>p.cthr~. The Laplace transform of the
f:..Pwj = pressure drop between point of measurement and
dimensionless wellbore pressure solution for the constant wellbore
sandface. storage and skin effect can be written by combining Eqs. I and 3 as
The formation parameters of permeability and skin can be estimat- SPD(S)+S
ed by convolving a PD model, reflecting the wellireservoir flow PwD(S) = , ....................... (4)
geometry, with measured rate, qD' Although we convolve indirect- S+CDS 2 [SP D(s)+s]

1040

-
960

880
Slope = 643 psi/cycle

............ 0

0.
80(,
Intercept

"hi" ~
= - 185 psi

44 md-ft/cp
720
i
~
0.
640
5 = - 4.9
"
~ 560
"0

.. slope 0.97 ~ 480

~ 400

320 ",
10' 240 ............ &
"
160
10° 10 1 10- 2 10' 10'
SHUT -IN TIME, HOURS Shut-in Time. Hours

Fig. 3-Log-log analysis of buildup data, Well 1. Fig. 4-MDH graph shows late-time pseudoradlal period,
Well 1.

608 SPE Formation Evaluation, September 1988


B80~------------------------------------~,
BOO
Slope = 582 pSI Vi
720 a.
640
Intercept = - 200 psi
~
oj
kh/IL = 51 md-ft/cp
560 ~ 102
4BO
S = - 4.3 ~ t slope = 0.36

~ 400
'"
0.
;;
w
~ 320 "~
Z
o
z 240 U

160 10' HASAN ET Al DATA


.a. OBSERVE:O
BO o OECONVOlVEO
o 0

O+----r--~----r_--~--~--~----r_--~--~ 100 10'


-1.6 -1.2 -O.B -0.4 0.0 0.4 O.B 1.2 1.6 2.0 SHUT-IN TIME. HOURS
RATE-CONVOLVED TIME FUNCTION

Fig. 6-Deconvolved pressure shows fractured-well response,


Fig. 5-Sandface-rate convolution analysis, Well 1. Well 1.

where s=Laplace transform variable. One can solve Eq. 4 using rate data simultaneously. The logarithmic-convolution approach 8
the Stehfest algorithm,21 provided that the Laplace transform so- presents a problem because a different wellbore geometry (hydrau-
lution, PD, for the constant-rate case is available for the well/reser- lically induced fracture) was encountered in the field examples con-
voir system of our interest. sidered. Only the late-time data conforming to the pseudoradial flow
Solutions for PwD, as given in Eq. I, can also be obtained with period could be straightened on the logarithmic-convolution plot.
Eq. 3 and a PD model, discussed in the Appendix, from the fol- In automated type-curve matching, we consider matching model
lowing algorithm I9 ,22: responses with field pressure data with and without the downhole
rate. We call the latter case convolved-type-curve matching to in-
PwD(tDn+ 1) = [PD(tDn+ 1)+S+CDX +tsPwD(tDn)]!(l +ts ), dicate convolution of the model with rate data.
The objective of automated type-curve matching is to minimize
..................................... (5a)
the sum of squares between a model (radial, fractured, etc.) response
where and the observed field data. Thus, the reservoir parameters are deter-
mined by minimizing the objective function, J({f):
n-1
X= E fpwD(tm+1)-PwD(tm)]
=> 1 N =>
;=0 tDi+ 1 -tm J«(3 )=- E [ilp~b«(3 ,t;)-ilp'i:,b(ti)]2 . .............. (8)
2 ;=1

The Marquardt 25 algorithm is used to minimize Eq. 8.


and
Field Examples
Two field examples are discussed in detail to provide an insight
into data interpretation, particularly involving downhole pressure
Exponential-Wellbore-Rate-DecIine Case. For certain cases, the and flow rate. A third example is also presented to give an idea
wellbore flow rate can also be modeled by the following of using the deconvolution method as a tool to identify the com-
relationship 23 ,24 : patibility between pressure and rate data.

qD=l-e-{3tD, .................................... (6) Weill. This example was first reported by Fetkovich and Vienot. 9
The test data come from a hydraulically induced fractured well
where production of all three phases occurs. Fig. 2 compares the
where (3 is a constant and is given as (3=acpp,c t r;/0.OO02637k.
two sources of pressure and flow-rate data derived from the
The Laplace transform of the dimensionless wellbore pressure
Gilbert 5 correlation and from the model of Hasan et al. 6 A rela-
for the exponential-rate case may be obtained by combining Eqs. 1
and 610: tively minor difference in the pressure response can be observed
with virtually no difference in the rate computation.
_ (3(sji D + S) The log-log graph of pressure-buildup data, obtained from the
PwD(s)= .............................. (7) Hasan et al. hydrodynamic model, 6 indicates masking of the early-
s«(3 +s) time period (up to 2 hours) by the wellbore-storage effect even in
the presence of a fracture (Fig. 3). Only the very-late-time data
Thus, solution of Eq. 1 depends on availability of the PD function (> 30 hours) are modeled by the pseudoradial flow period, as evi-
in the real space, while either Eq. 4 or 7 is solved when the PD denced by the flat derivative signature (Fig. 3), the Miller-Dyes-
function is available in the Laplace space. In other words, one can Hutchinson (MDH) plot (Fig. 4), and the logarithmic-convolution
compute the wellbore pressure, PwD' either from Eq. 1 using Eq. plot (Fig. 5). As earlier noted by Fetkovich and Vienot, 9 convolv-
6 or from Eq. 7. In the Appendix, we present the PD models used ing with logarithmic approximation of the exponential integral func-
in this study. tion (known as logarithmic-convolution) does not provide the
expected straight-line relationship for the entire data band. Simply
Interpretation Methodology stated, we need to convolve measured data with the correct PD
In this study, several interpretation methods are applied to obtain model to perform a total rather than a partial data set analysis.
reservoir parameters of transmissivity, skin, fracture half-length, To gain insight into the PD function, reflecting the near-wellbore
and storage coefficient in some cases. flow behavior, we generated the deconvolved pressure from the
Conventional type-curve and semilog analyses do not easily lead AWS-derived pressure and flow-rate data. Fig. 6 compares the de-
to unique estimations of reservoir parameters because of dominat- convolved pressure with those of measured response. Here, we
ing influence of the storage or afterflow effect. This storage dis- clearly observe the storage-free behavior with a maximum slope
tortion can be largely minimized by convolving the pressure and of 0.36, instead of the half-slope reported earlier in Ref. 9. Calcu-
SPE Fonnation Evaluation, September 1988 609
o OECONVOLVEO
• cor .. 0
•• COr .. 0003
_cor ..
_cor ..
•. cor ..
0'
001
003
..
•• ~!! ,t
_COF '" 03

Vi
0..
10'

..'.'
<i
<1
W
PPESSUPE '"v>::>
~
'"
0..
10'
PRESSURE DERIVATivE ~
UJ

"<
Z
I
U

10' • 00S<lM0
o DECON-GIt..8ERT
II OECQN-HKR
10' 10'
10'
SHUT -IN TIME. HOURS

Fig. 7-Matchlng deconvolved pressure with Inflnlte-


conductivity-fracture type curves, Well 1. Fig. 8-Comparlson of deconvolved data, Well 1.

lations indicate that the linear flow period, characterized by a slope ly, the derivative signature corroborates the fractured-well behavior,
of 0.5 on the log-log plot, had ended at 0.22 hours. This observa- inferred earlier from deconvolution analysis.
tion implies that most of the early-time data are in transition from Although an early-time mismatch is noted, the quality of the over-
linear to radial flow regime. Thus, the well response may perhaps all fit is quite good. When the downhole rate is estimated by the
be modeled better by the infinite-conductivity vertical fracture rather constant-storage model, the quality of the fit suffers both at early-
than by a uniform-flux fracture. The rationale for this contention and middle-time regions even though the results are essentially the
is that the half-slope period ends a cycle earlier for the infinite- same (Fig. 11). Note that the results from the two approaches are
conductivity model than the uniform-flux fracture solution. not expected to be equal because the storage coefficient, C, is not
When the deconvolved pressure data are superposed on the type the same in two cases. Indeed, by convolving with BHF, we es-
curves generated for infinite-conductivity vertical fracture with sentially obtain a zero value for C. In contrast, when the rates are
storage, Fig. 7 shows that the storage effects are largely reduced. estimated from the constant-storage model, a much higher value
Furthermore, the corresponding flat derivative response (not per- of C (0.063 bbl/psi [1.45 x 10- 3 m3 /kPa]) is noted. This optimized
formed on data) is indicative of a fractured well's characteristic C value is largely substantiated by Fig. 12.
behavior in the pseudoradial flow period. The C values in Fig. 12 are computed from the constant-storage
A similar fractured-well response is obtained when we decon- model; i.e., C=qs/(24 dpldt). This figure also helps explain the
volve pressure with the rate data, obtained from Gilbert's history mismatch obtained (Fig. 11) until about 1.5 hours of shut-
hydrodynamic correlation, 5 as Fig. 8 indicates. These deconvolved in is reached. Before this time, C varied by as much as a factor
data can be matched with either the infinite-conductivity vertical of 2. For all practical purposes, a constant C value of 0.053 bbl/psi
fracture type curve or the Gringarten et al. 26 type curve for aci- [1.22 x 10- 3 m 3 /kPa] is attained beyond 1.5 hours: the data scat-
dized well, or may be used for MDH analysis. As expected, only ter reflects the lack of desired-quality computed bottomhole pres-
the late-time data are straightened on the deconvolved MDH plot, sure (BHP), necessary for a derivative operation. Obviously, by
Fig. 9. incorporating BHF into the transient analysis we could improve our
An automated history-matching algorithm was used for interpre- parameter estimation. Thus, in principle, one should attempt to in-
tation with a radial model. Two approaches with respect to the in- corporate BHF into transient analysis whenever the quality of such
ner boundary condition are taken for interpretation. In the first data permits.
method, history matching of pressure data is achieved by using the By repeating exactly the same procedure as described earlier, we
predicted bottomhole flow rate (BHF), while in the second method generated Figs. 13 and 14 for the Hasan et al. data. In Fig. 13,
the BHF is predicted by the constant-storage model. we observe an improved match at early times compared with the
Fig. 10 shows the history match of pressure and its derivative, Gilbert data (Fig. 10); however, the middle-time period lacks the
convolved with BHF, for the Gilbert data set. The derivative desired quality of match. As Fig. 14 suggests, a similar trend per-
response is obtained by taking the derivative of pressure, !J.p, with sists when the history match is sought with the constant-storage
respect to the logarithm of the superposition time function. Clear- model. .

880.---------------------------------------~
GILBERT DATA
A Y(A,SUREO 6p
800 ~ 10} o Y[ASURED ll.p'
•• CQlJPUT[O top
o
=
_CQlJPUTEOll.p
4.'II11.I..·~""III."'''~
"'">
Slope 585 psi/cycle
720
w kh!!" = 52 md-ft/cp ~ ..... ",

kh/!" ~ 51 md-ft/cp co
~ 640 :; 5 = - 4 35
s ~ - 4.3
.' [5
~ 560 o c ~ 0 063 bbl/ps, dJ 00°
o w
~ 480 '" :::J J = 1 195 PSI
2 .' 0

~
6u400
~ 10'
g: ••
UJ
0320 '"
o .'
........
240

160
......
10" 10' 10° 10' 10'
SHUT -IN TIME, HOURS SHUT -IN TIME. HOURS

Fig. 9-Semllog analysis of deconvolved pressure, Well 1. Fig. 10-Automated type-curve matching with storage model,
Well 1.

610 SPE Formation Evaluation, September 1988


HASAN ET AL. DATA
.. I,I(ASVRED lop
o t.ltASUREO .6p'
._CO!oIPUfEO IIp
_COMPUTED !:>p'

kh/ll = 48 md-ft/cp

- 4.48
10'
C = ° o·••
o
J = 1200 pS;2 kh/ll = 50 md-ft/cp
10'
o• s = - 4.4

C = 0

o . J = 3069 psi 2
10'

10° 10' 10' laC to' 10'


SHUT -IN TIME. HOURS SHUT -IN TIME. HOURS

Fig. 11-Automated convolved-type-curve matching, Well 1. Fig. 13-Automated convolve~-tvpe-curve matching, Well 1.

-10 -2

10.2

Q.
9.6
v; '"
o
Q.

'J
9.0 '"w
'-'
<D >
<D
8.4
"'l'
~0 7.8 ffi
u ~ 10
2

w <r
7.2
~
g 6.6
~
a:
kh/ll = 45 md-ft/cp
Q.
<J> s=-4.55
w
a: 6.0 '"ow
0

~ .. _.'" '"~
C = 0.052 bbl/ps;

'"
5.4

4.8
- - - - '- 6 ... _ ..

' ...
......
~
g: 10'
J = 1 140 PSl2

4.2
10-' 10C 10' 10° 10'
SHUT -!N TIME. HOURS SHUT -IN TIME. HOURS

Fig. 12-Behavlor of storage coefficient during buildup, Fig. 14-Automated type-curve matching with storage model,
Well 1. We!11.

A match was also sought with an infinite-conductivity-fracture


model by using the automated history-match algorithm. Fig. 15 dis- INFINITE-CONDUCTIVITY MODEL

plays the match: rel~tively large residuals suggest an inferior match


kh/ll = 41 md-ft/cp
rather than a radial model. Earlier, we explored the question sur-
rounding this match by simulating the fracture with a composite ~ x, = 66 ft
c>
model. 27 Results indicated that a radial model gives a better match <l - 4.67
w'
than the fracture model. ~ 10' 7160 psi 2
Reservoir parameters obtained from these interpretations are es-
sentially the same, suggesting that any hydrmlynamic correlation
oit
;;
will yield valid results for this specific example. In other words, w
co
this example is not too demanding to test the robustness of any ?i
I
hydrodynamic correlation. We caution, however, that certain well U

conditions may severely test the limits of a hydrodynamic correla-


10'
tion. These conditions include wells with a long initial liquid column,
contributing more than 95% of the BHP and/or experiencing a high
GOR production. Even in relatively short liquid columns (less than 10" 10° 10'
SHUT -IN T!ME. HOURS
1,000 ft [300 m]), a high gas rate can conceivably lead to a situa-
tion comparable to the long columns. Fig. 15-"utomated convolved-type-curve matchlrg, Hasan-
Table 1 summarizes the results obtained from various analyses. Kaplr data, Well 1.
Although good agreement in the skin values is obtained when com-
pared with those of Fetkovich and Vienot, 9 reservoir transmissivity
values differ because initial flow rates are quite different. This dis- The liquid-level data are first translated to BHP and BHF using
crepancy in flow rates is probably explained by the way the raw both the Gilbert correlation and the Hasan et al. hydrodynamic
data are handled in the two studies. model. Fig. 16 compares the results. The log-log graph of pres-
sure and its derivative are shown in Fig. 17. Tl).e c:arly-time data
Well 2. Test data from this well, initially reported by Browns- are masked by storage as the unit-slope period suggests. However,
combe, 2 were subsequently considered by others. 3 ,9,28 Interpre- the derivative signature indicates the presence of a frac~re. MDH
tation efforts made by Fetkovich and Vienot9 and Thompson and analysis indicates (Fig. 18) that only the very-late-time period may
Reynolds 28 indicate that the data are from a fractured well. Here, have a pseudoradial flow regime, mpch like the previous example.
we re-examine the test to shed more light through other interpreta- Deconvolution of pressure and rate data produces a slope of 0.40
tion methods. on the log-log graph (Fig. 19), suggesting a fractured well response.
SPE Formation Evaluation, September 1988 611
TABLE 1-SUMMARY OF ANALYSIS RESULTS, WELL 1

Parameter
khlp,
Analysis Method _ _R_e_s_e_rv_o_ir_M_od_e_I_ _ Model for BHP/BHF (md-ft/cp) Sk.in
MDH (measured pressure) Radial HKR*/storage 44 -4.9
MDH (deconvolved pressure) Radial HKR/HKR 51 -4.3
Sandface-rate convolution Radial HKR/HKR 51 -4.3
Automated type curve Radial HKRlstorage 45 - 4.6
Automated type curve Radial HKR/HKR 50 -4.4
Automated type curve Radial Gilbert/Gilbert 48 - 4.5
Automated type curve Radial Gilbert/storage 52 - 4.4
Automated type curve Infinite-conductivity fracture HKR/HKR 41 -5.3

'Hasan et a/. 6 hydrodynamic model.

1200
1200
_8HP, HASAN ET At
1100
1100 _814'. HASAN ET AL.
A 81-!P, GILBERT
o 8HF. GILBERT 1000
Slope = 772 psi/cycle
'" 1000 Intercept = - 356 psi
"-
."a.
0 9(0

.
CD
900
"'- Kh I,u. = 127 md-ft/cp

~
80"
800
&I = - 4.9

0
0:
700
.~
a.
700

600
s

.
.
?
600
500 "E 500
.9
.
'"
~ 400 '0 400
.
CD

i 300 300
.
. . . .. . . .. .
~ 200
a. ~. 200
"-Q
100 "-'Q,.
"'0 100
0 10- 2 10" 10' 10'
10- 2
10" 10' 10' Shut-in Time, Hours
Shut-in rime, Hours

Fig. 18-MDH graph shows late-time pseudoradlal period,


Fig. 16-Pressure/flow-rate history, Well 2. Well 2.

an empirical correlation is used to translate the liquid-level data


10'lrr====~~------------'----'
to BHP and BHF.
To attain compatibility between pressure and rate data, we de-
convolved different pressures with the rate. Fig. 25 shows that both
the Gilbert and the Podio et at. pressures indicate a failure (horizon-
tal line) in deconvolution, meaning incompatibility in pressure and
rate data. In contrast, pressure data derived from the mechanistic
Godbey-Dimon! and Hasan et at. 6 models show successful decon-
volution; i.e., positive slopes are indicated. Only the data points
up to the beginning of the phase-redistribution phenomenon are con-
sidered; thus, no attempt is made to interpret seven data points to
evaluate formation parameters.
We point out, however, that this well had some 5,300 ft [1615
m] of liquid level at the beginning of the test, which is very demand-
10'~--~~~r-~~~-r--~~~r-~~~-r--~ ing on the hydrodynamic model/correlation for pressure computa-
10-2 10' 10' 10'
Shut-In Time, Hours
tion. In other words, any error in the gas void-fraction estimation
seriously affects pressure prediction. This point was discussed in
detail in Refs. 3 and 6.
Fig. 17-Log-log analysis of buildup data, Well 2.
The purpose of this example is just to highlight the potential prob-
lem that may exist when an empirical hydrodynamic correlation
is used for estimating BHP.
As expected, the deconvolved MDH graph, Fig. 20, produces a
trace very similar to the logarithmic-convolution plot; thus con-
firming the presence of a fracture. Discussion
We attempted convolution or automated type-curve matching with The purpose of this study was to analyze pressure and flow-rate
fracture as well as radial models. Figs. 21 and 22 show the match data gathered from pumping wells. The underlying thought was to
quality obtained with infinite-conductivity and radial models, re- seek the consistent hydrodynamic model/correlation used to trans-
spectively. Except for the early-time data, the superior match with late the liquid-level measurements to downhole pressure and rate.
the radial model is quite obvious. Similar results are obtained when Interpretation of simultaneous downhole pressure and rate datil
Gilbert's pressure data are used with flow rate, as displayed in Figs. gives better estimates of reservoir parameters when convolution,
23 and 24. All results are summarized in Table 2. history matching, or automated type-curve matching is used. This
contention is borne out by the fact that wellbore-storage coefficient,
Well 3. This well test, exhibiting wellbore-phase redistribution, along with other reservoir parameters, is sought during the course
was first reported by Brownscombe 2 and subsequently considered of nonlinear least-squares analysis. Field examples show that the
by Hasan and Kabir 3 and Hasan et at. 6 The purpose of reconsider- storage coefficient is reduced to a large degree when the downhole
ing this example is to demonstrate problems that may surface when rate is incorporated into the analysis; nonetheless, it is seldom re-

612 SPE Fonnation Evaluation, September 1988


10'<r----------------------------------~,.----_,
10'
HlIMflel.a1. Data RADIAL MODEL
.&. I,IEASUREO
_CALCULATEO

'm
G.
0.
g <J

I
10'
£ Slope 0.40"
kh/I" ~
102 md-ft/cp
a:
.'"
c

c
a
s

C
~

~
- 6

0.0097 bbl/psi
"
u

10' J ~
595 pSil

Shut-in Time, Hours 10-' 10' 10' 10'


Shut-in Time, Hours

Fig. 19-Deconvolved pressure shows fractured-well re-


sponse, Well 2. Fig. 22-Automated convolved-type-curve matching, Well 2.

1000
GILBERT DATA INFINITE-CONDUCTIVITY MODEL
.!. UEASURED
900 10' _CALCULATED

800
Slope = 762 psi/cycle

Intercept = - 448 psi


700
'm
0-
kh/I" ~ 129 md-ft/cp
0. 600
<I kh/I" ~ 203 md-ft/cp
s ~ - 5.0
'"~ 500
5 = - 6
~
~
0
400

300
.'"
c
a
C ~ 0.076 bbl/psi
J = 5031 psi 2
"
U la'
200

100 ..
0
10" 10' 10' 10' 10 2 10'
Shut-in Time, Hours Shut-in Time, Hours

Fig. 20-Semllog analysis of deconvolved data, Well 2. Fig. 23-Automated convolved-type-curve matching,
GilbertS data, Well 2.

GILBERT Dl'TA
RADIAL MODEL
10' -r;:::::::::::::::::::::;-~IN:;;:F:;;:IN;;";IT7E=-C;:;O;7,N~DU:;-;C::;'-T;;;IV;7-1Ty~---::7-----' 10'
6 I,lEASURED
_CAlCULATED

MODEL

'm
0-
0.
<110 2

i kh/I" ~ 120 md-ft/cp

G.
~
kh/I" ~ 152 md-ft/cp

x, = 131 ft
. c
o'"
s = - 6

C ~ 0.016 bbl/psi
C ~ 0.054 bbl/psi
"
U

J = 538 psi l
J = 8244 psi" 10'

10' 10' 10'


Shut-in Time, Hours
10' 10' 10'
Shut-in Time. Hours

Fig. 24-Automatic convolved-type-curve matching, GilbertS


Fig. 21-Automated convolved-type-curve matching, Well 2. data, Well 2.

moved completely. This observation is consistent with the physi- data significantly aid reservoir-model identification, leading to the
cal system in place-Le., the presence of dead volume below use of an appropriate model for automated type-curve analysis, for
perforations and volume of the fracture, if applicable. example. In this respect, the first two examples stress the value
Thus, automated type-curve matching generally leads to a better of pressure~derivative analysis as a model-identification tool. Thus,
reservoir-parameter estimation compared with a convolution scheme use of both pressure-derivative and deconvolution methods provides
that does not consider the storage effect. Logarithmic convolution powerful tools for reservoir-model identification.
of Meunier et al. 8 and convolution with other PD models, such as We point out that deconvolution may be used as a diagnostic tool
that given by Fetkovich and Vienot,9 fall into the latter category. for checking the consistency between the pressure and flow-rate
For the same reason, deconvolved pressure data are also not com- data. Lack of a perfect harmony between the two measured/inferred
pletely free from storage. In most cases, however, the deconvolved entities is a reflection of unreliability of a data set. In this respect,

SPE Fonnation Evaluation, September 1988 613


TABLE 2-SUMMARY OF ANALYSIS RESULTS, WELL 2

Parameter
khlf-t
Analysis Method ___ Re_s_e_rv_o_ir_M_od_e_I_ _ Model for BHP/BHF (md-ftlcp) Skin
MDH (meaSured pressure) Radial HKR/- 127 -4.9
MDH (deconvolved pressure) Radial HKR/HKR 129 -5.0
Autortlated type curve Radial HKR/HKR 102 -6.0
Automated type curve Infinite-conductivity fracture HKR/HKR 152 -4.4
Automated type curve Radial Gilbert/Gilbert 120 - 6.0
Automated type curve Infinite-conductivity fracture Gilbert/Gilbert 203 - 6.0

not to preserve the reservoir mass balance. Chu et al. 31 and


Raghavan 32 showed that as long as no significant saturation gra-
dient exists in the wellbore vicinity, total mobility adequately ap-
proximates the single-phase flow theory. Thus, all the field

I
~
..: . .
~ 0
.0
interpretation presented here is inherently subject to this approxi-
mation .

Conclllsions
10'-
1. In the absence of wellbore crossflow or other wellbore anomaly
(e.g., foam), indirect downhole pressure and flow-rate measure-
ments made by AWS provide interpretable data from buildup test-
ing in pumping wells.
2. Automated type-curve matching is a powerful interpretation
tool for estimating reservoir parameters because a variety of reser-
~"DGodN,-DlmOD.1
1IiIo1II...ured Preuure
11o..o i)e(!oD",ol"d
DeeoDoroh.4 Padlo .tel.
De<:otlyol~dGII"rt
voir models can be used and the wellbore-storage effect below the
• ~oayohed H_D el a1. point of measurement is accounted for .
10'
3. Both deconvolution and pressure-derivative methods signifi-
1~" 1~' cantly aid identification of well/reservoir models for subsequent
SHUT-IN TIME. HRS
conventional and other analyses.
Fig. 25-Comparison of deconvolved pressure, Well 3_ 4. Long-duration wellbore-storage-distortion period-a typical
scenario in a pumping well-can be minimized significantly by per-
forming simultaneous analysis of pressure and flow-rate data (decon-
deconvolution is a superior tool because convolution (superposi- volution, automated type-curve matching, etc.).
tion) is merely a smoothing process; thus, successful convolution 5. For moderate pumping Jiquid-column height ( < 1,000 ft [ < 305
may not necessarily tell tis anything about the data quality. m]), use of any of the four hydrodynamic correlations/models ap-
Our experience indicates that, except for long liquid columns, pears to be satisfactory for analysis of simultaneous pressure and
successful deconvolution can be performed even when Gilbert's em- flow-rate data. However, hydrodynamic models for oil/water and
pirical hydrodynamic correlation is used for computing downhole gas/oil/water flow are needed to improve pressure and flow-rate
pressure. Presumably, short liquid columns do not create enough computation for transient testing of pumping wells.
inconsistency in pressure calculation to present any problem while
deconvolving with the rate data. Thus, it appears that for a vast NOinenclature
majority of the field tests, any of the four correlations/models 1,4-6
B = FVF, RB/STB [res m 3 /stock-tank m 3 ]
available for translating the acoustic liquid-level data for the pur-
c I = total system compressibility, psi -1 [kPa -1 ]
pose of transient analysis could be used. We caution that all these
hydrodynamic correlations are designed to handle only the gaslliq- C = wellbore storage coefficient, bbl/psi [m 3 /kPa]
uidlflow, however, and that no provision is made to account for h = formation thickness, ft [m]
oil/water liquid/liquid flow. Thus, further research is needed to J = objective function, Eq. 8
describe the hydrodynamics of liquidlliquid flow and, more appro- k = permeability, md
priately, of three-phase mixture in the wellbore. kf = fracture permeability, md
This study shows that the pressure and flow-rate data derived Ko = modified Bessel function of second kind and zero
from acoustic measurements are useful for· obtaining a reservoir order
model and estimating its parameters even though direct validation K 1 = modified Bessel function of second kind and first
of BHP measurement is not provided. Fot example, the results of
order
the late-time MDH analysis of the pseudoradial period for both Wells
1 and 2 agree very well with those of type-curve matching involv- N = number of measured data points
ing the entire data set. The implication is that we ought to have PD = dimensionless pressure for constant rate
a consistent data set for such an agreement to be possible. We point Pwb = measured wellbore pressure, psi [kPa]
out that the late-time data are essentially independent of any PwbD = dimensionless measured wellbore pressure
hydrodynamic correlation because the gas percolation has ceased Pwf = flowing wellbore pressure, psi [kPa]
for all practical purposes. In other words, BHP is directly depend- q = flow rate. STBID [stock-tank m 3 /d]
ent on the absolute accuracy of the acoustic device gathering the qD = qwb/Bqr, normalized bottomhole flow rate
data at late times. On the other hand, the early-time data are sub- qt = total bottomhole flow rate, BID [m 3 /d]
ject to error in modeling the two-phase flow behavior in the well- qwb = downhole flow rate, B/D [m 3 /d] .
bore, in addition to the tool resolution. Thus, an agreement of results
r = radial distance, ft [m]
obtained frorli transient analysis of a segment and the entire data
rD = dimensionless radial distance
set provides a clue about the consistency of a data set, especially
when BHF data are incorporated into the analysis. rw = wellbore radius, ft [m]
For transient interpretatiort, we used the total mobility concept s = Laplace transform variable
of P!!rrine 29 and Martin 30 to handle the three-phase flow in the S = skin factor, dimensionless
reservoir. The same approach has been used by Fetkovich and Vie- t = time, hours

614 SPE Formation Evaluation, September 1988


tD = dimensionless time based on wellbore radius 19. Kucuk, F.: "Generalized Transient Pressure Solutions With Wellbore
txjD = dimensionless time based on fracture h~f-Iength Storage," paper SPE 15671 available at SPE, Richardson, TX.
20. Agarwal, R.G., AI-Hussainy, R., and Ramey, H.I. Jr.: "An Investi-
w = fracture width, ft [m] gation of Wellbore Storage and Skin Effect in Unsteady Liquid Flow:
X = summation in Eq. 5b 1. Analytical Treatment," SPEJ (Sept. 1970) 279-90; Trans., AIME,
ex = system constant 249.
(3 = excJ>f.Lctr~/O.OOO2637k 21. Stehfest, H.: "Algorithm 368, Numerical Inversion of Laplace Trans-
11 = parameter vector forms," D-5 Communications oftheACM (Jan. 1970) 13, No.1, 47-49.
22. Cinco-Ley, H. and Samaniego-V., F.: "Effect of Wellbore Storage and
f.L = reservoir fluid viscosity, cp [Pa' s] Damage on the Transient Behavior of Vertically Fractured Wells," paper
T = dummy integration variable SPE 6752 presented at the 1977 SPE Annual Technical Conference and
cJ> = formation porosity, fraction Exhibition, Denver, Oct. 9-12.
23. van Everdingen, A.F.: "The Skin Effect and Its Influence on the Produc-
Subscripts tive Capacity ofa Well," JPT(June 1953) 171-76; Trans., AlME, 198.
24. Hurst, W.: "Establishment of the Skin Effect and Its Impediment to
sf = sandface Fluid Flow into a Wellbore," Pet. Eng. (Oct. 1953) B-6-B-16.
w = wellbore 25. Marquardt, D.W.: "An Algorithm for Least-Squares Estimation of Non-
wf = flowing wellbore linear Parameters," J. Soc. Ind. App. Math. (June 1963) 11, No.2.
26. Gringarten A.C. et al.: "A Comparison Between Different Skin and
Superscripts Storage Type-Curves for Early-Time Transient Analysis," paper SPE
c = calculated 8205 presented at the 1979 SPE Annual Technical Conference and Ex-
hibition, Las Vegas, Sept. 23-26.
m = measured 27. Kabir, C.S., Kuchuk, F.J., and Hasan, A.R.: "Transient Analysis of
, = derivative with respect to time Acoustically Derived Pressure and Flow Rate Data," paper SPE 15481
- = Laplace transform presented at the 1986 SPE Annual Technical Conference and Exhibi-
tion, New Orleans, Oct. 5-8.
Acknowledgments 28. Thompson, L.G. and Reynolds, A.C.: "Analysis of Variable-Rate Well-
Test Pressure Data Using Duhamel's Principle," SPEFE (Oct. 1986)
We are thankful to M. Fetkovich of Philips Petroleum Co. for 453-69.
providing additional information on Well I and to Schlumberger 29. Perrine, R.L.: "Analysis of Pressure Buildup Curves," Drill & Prod.
management for permission to publish this work. Prac., API (1956) 482-509.
30. Martin, J.C.: "Simplified Equations of Flow in Gas Drive Reservoirs
and the Theoretical Foundation of Multiphase Pressure Buildup Anal-
References yses," JPT (Oct. 1959) 309-11; Trans., AIME, 216.
1. Godbey, I.K and Dimon, C.A.: "The Automatic Liquid Level Moni- 31. Chu, W.C., Raghavan, R., and Reynolds, A.C.: "Pressure Transient
tor for Pumping Wells," JPT(Aug. 1977) 1019-24. Analysis of Two-Phase Flow Problems," SPEFE (April 1986) 151-64.
2. Brownscombe, E.R.: "Afterflows and Buildup Interpretation on Pump- 32. Raghavan, R.: "Well Test Analysis for Multiphase Flow," paper SPE
ing Wells," JPT (Feb. 1982) 397-412. 14098 presented at the 1986 SPE IntI. Meeting on Petroleum Engineer-
3. Hasan, A.R. and Kabir, C.S.: "Determining Bottornhole Pressures in ing, Beijing, March 17-20.
Pumping Wells," SPEJ (Dec. 1985) 823-38. 33. Gringarten, A.C., Ramey, H.I. Jr., and Raghavan, R.: "Unsteady-
4. Podio, A.L., Tarrillion, M.J., and Roberts, E.T.: "Laboratory Work State Pressure Distributions Created by a Well With a Single Infinite-
Improves Calculations," Oil & Gas J. (Aug. 25, 1980) 137-46. Conductivity Vertical Fracture," SPEJ (Aug. 1974) 347-60; Trans.,
5. Gipson, F.W. and Swaim, H.W.: "Designed Beam Pumping," Proc., AIME,257.
19th Annual Southwestern Petroleum Short Course, Lubbock, TX (April 34. Earlougher, R. C. Jr.: Advances in Well Test Analysis, Monograph Ser-
1972) 95. ies, SPE, Richardson, TX (1977) 5.
6. Hasan, A.R., Kabir, C.S., and Rahman, R.: "Predicting Liquid Gra- 35. Hale, B.W. and Evers, J.F.: "Elliptical Flow Equations for Vertically
dient in a Pumping Well Annulus," SPEPE (Feb. 1988) 113-20. Fractured Gas Wells," JPT (Dec. 1981) 2489-98.
7. Kabir, C.S. and Hasan, A.R.: "Application of Mass Balance in Pumping 36. Lee, S.T. and Brockenbrough, J.R.: "A New Approximate Analytic
Well Analysis," JPT(May 1982) 1002-10. Solution for Finite-Conductivity Vertical Fractures," SPEFE (Feb. 1986)
8. Meunier, D., Wittmann, MJ., and Stewart, G.: "Interpretation of Pres- 75-88.
sure Buildup Test Using In-Situ Measurement of Afterflow," JPT(Jan. 37. Kucuk, F. and Brigham, W.E.: "Transient Flow in Elliptical Systems,"
1985) 143-52. SPEJ (Dec. 1979) 401-10; Trans., AIME, 267.
9. Fetkovich, M.J. and Vienot, M.E.: "Rate Normalization of Buildup
Pressure Using Afterflow Data," JPT (Dec. 1984) 2211-24.
10. Kucuk, F. and Ayestaran, L.: "Analysis of Simultaneously Measured Appendix-Reservoir Models (PD functions)
Pressure and Sandface Flow Rate in Transient Well Testing," JPT(Feb. For the interpretation of field data, we consider the radial and
1985) 323-34. infinite-conductivity vertical-fracture models.
11. Kabir, C.S. and Kucuk, F.: "Well Testing in Low-Transmissivity Oil
Reservoirs," paper SPE 13666 presented at the 1985 SPE California Radial Model. The Laplace transform of PD for a constant flow
Regional Meeting, Bakersfield, March 27-29.
12. Kucuk, F., Carter, R.G., and Ayestaran, L.: "Numerical Deconvolu-
rate for radial flow in an infinite-acting reservoir is given as 18
tion of Wellbore Pressure and Flow Rate," paper SPE 13960 available
at SPE, Richardson, TX. KO(-J;)
13. Rosa, A.J. and Horne, R.N.: "Automated Type-Curve Matching in PD=-- for rD=1. ................... (A-I)
Well Test Analysis by Using Laplace Space Determination of Parame-
ter Gradients," paper SPE 12131 presented at the 1983 SPE Annual s-J; KI(-J;)
Technical Conference and Exhibition, San Francisco, Oct. 5-8.
14. Barua, 1., Kucuk, F., and Gomez-Angulo, J.: "Application of Com- Infinite-Conductivity Vertical Fracture. The dimensionless pres-
puters in the Analysis of Well Tests From Fractured Reservoirs," paper sure for an infinite-conductivity vertically fractured well produc-
SPE 13662 presented at the 1985 SPE California Regional Meeting, ing at a constant rate in an infinite-acting reservoir is given by 33,34
Bakersfield, March 27-29.
15. Guillot, A.Y. and Home, R.N.: "Using Simultaneous Flow Rate and

l(
Pressure Measurements to Improve Analysis of Well Tests," SPEFE
(June 1986) 217-28. 0.134)
16. Kabir, C.S., Kucuk, F., and Gomez-Angulo, J.: "Well Test Interpre-
PD=1f2..J'lrtXjD erf - - - +erf(0866)]
-'--
tation in Faulted Reservoirs," Proc. sixth SPE Offshore South East ..J tXjD ..J txfD
Asia Conference, Singapore (Jan. 28-31, 1986) 152-61.
17. Kucuk, F., Karakas, M., and Ayestaran, L.: "Well Test Analysis of
Commingled Zones Without Crossflow," SPEFE (Aug. 1986) 324-54.
18. van Everdingen, A.F. and Hurst, W.: "The Application of the Laplace 0.018) ( 0.750)
-0.067Ei ( - - - -0.433Ei - - - . .. ..... (A-2)
Transfortuation to Flow Problems in Reservoirs," Trans., AlME (1949)
tXjD tXjD
186, 305-24.
SPE Formation Evaluation, September 1988 615
TABLE A·1-COMPARISON OF APPROXIMATE SOLUTIONS WITH ANALYTICAL
SOLUTION OF GRINGARTEN et al. 33

Po
t xfO Gringarten et al. Eq. A·5 Eq. A-7 Eq. A-4 Eq. A-8
0.10000x 10- 2 0.0560 0.0560 0.0554 0.0560 0.0560
0.12500 x 10- 2 0.0627 0.0626 0.0618 0.0627 0.0626
0.15000 x 10- 2 0.0686 0.0686 0.0676 0.0686 0.0686
0.20000 x 10- 2 0.0793 0.0792 0.0778 0.0793 0.0792
0.30000 x 10- 2 0.0971 0.0969 0.0949 0.0971 0.0969
0.40000 x 10- 2 0.1121 0.1119 0.1092 0.1121 0.1119
0.85000 x 10 - 2 0.1629 0.1627 0.1572 0.1634 0.1627
0.90000 x 10- 2 0.1676 0.1674 0.1615 0.1681 0.1674
0.10000 x 10- 1 0.1764 0.1763 0.1699 0.1772 0.1763
0.20000 x 10- 1 0.2455 0.2481 0.2362 0.2507 0.2481
0.30000 x 10- 1 0.2954 0.3024 0.2856 0.3070 0.3024
0.50000 x 10- 1 0.3696 0.3866 0.3616 0.3963 0.3866
0.60000 x 10- 1 0.3995 0.4216 0.3929 0.4342 0.4216
0.70000 x 10- 1 0.4263 0.4533 0.4212 0.4689 0.4533
0.80000 x 10- 1 0.4508 0.4824 0.4473 0.5013 0.4824
0.10000 0.4944 0.5347 0.4940 0.5605 0.5347
0.15000 0.5832 0.6415 0.5899 0.6865 0.6415
0.40000 0.8602 0.9644 0.8897 1.1210 0.9644
0.70000 1.0608 1.1853 1.1097 1.4829 1.1853
0.80000 1.1128 1.2412 1.1680 1.5853 1.2412
0.90000 1.1599 1.2914 1.2213 1.6815 1.2914
c:l.95000 1.1819 1.3147 1.2464 1.7276 1.3147
1.0000 1.2029 1.3370 1.2706 1.7725 1.3310
1.2500 1.2965 1.4354 1.3798 1.9817 1.4354
1.5000 1.3753 1.5175 1.4740 2.1708 1.5175

Because the Laplace transform of Eq. A-2 is not readily available, a = 2IFD , and
we resorted to piecewise approximation for PD functions to obtain FD = kf·w/k·xf·
their Laplace transforms for subsequent use in either Eq. 4 or 6
ofthe text. For the sake of convenience, four flow periods are de-
fined here: early-time linear, elliptic, elliptic/pseudoradial transi- Pseudoradial Flow Period. The Laplace transform of PD during
tion, and the pseudoradial flow period. In the following, we present the pseudoradial period is given by 37
the approximate solutions for each flow period, which are good
for all values of txJD as demonstrated later.
0.5772-2 In 2 1 )
PD = - ( + -In s , 1.25 <j In 21s < 00.
Linear Flow Period. The linear flow period is represented by 33.34 s 2s .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-8)
PD =.J 7rtxJD' ................................... (A-3)
In Eqs. A-4, A-6, A-7, and A-8,j represents the summation index
for function evaluation in the Stehfest algorithm. 21 In our compu-
Taking the Laplace transform of PD for the linear flow period tations, j= 1,2 ... 12 was used while the time bounds for each of
yields those equations were computed. Solutions for PwD in Eq. 4 or 7
can be obtained together withpD models given by either Eq. A-I
7r or Eqs. A-4, A-6, A-7, and A-8 using the Stehfest algorithm. Dur-
PD= - - , j In 2Is<0.OO8 . .................... (A-4) ing the course of this investigation, we observed that the piecewise
2sVs approach for solving the infinite-conductivity vertical fracture prob-
lem, as outlined here, is a far less computationally demanding op-
Elliptic Flow Period. The elliptic flow period is given by 35 eration than the Laplace transform of exact analytical solution in
terms of Mathieu functions. 37
PD =In(.J 7rtxJD +.J l+7rtxJD)' ..................... (A-5) Computations were made with the approximate infinite-
conductivity vertical fracture solutions for each flow period, and
Taking the Laplace transform of Eq. A-5 yields subse~ently compared with the analytical solution of Gringarten
et al. We observed that within the limits of txjD' the approxi-
mate solutions are in good agreement with the analytical solution
PD=~e(S/21r)KO(~)' 0.008<j In(2/s) <0.037 ..... (A-6) of Gringarten et aI., with a maximum error of 6.7 % at late times.
2s 27r Table A-I presents the results.

Elliptic Pseudoradial Transition Period. The Laplace transform SI Metric Conversion Factors
of the PD function for this transition period 36 is given by bbl x 1.589 873 E-OI m3
cp x 1.0* E-03 Pa's
b ft x 3.048* E-Ol m
PD= , 0.037<j In 2/s<1.25, ............ (A-7) psi x 6.894 757 E+OO kPa
21/; tanh 1/;
* Conversion factor is exact. SPEFE
where
b = -7rIFD ,
Original SPE manuscript received for review Oct. 5. 1986. Paper accepted for publication
June 15. 1987. Revised manuscript received Dec. 11. 1987. Paper (SPE 15481) first present·
1/; = .Ja.Js+£ ed at the 1986 SPE Annual Technical Conference and Exhibition held in New Orleans.
Oct. 5-8.

616 SPE Fonnation Evaluation. September 1988

You might also like