You are on page 1of 8

Well Test Interpretation of Vertically

Fractured Gas Wells


Robert A. Wattenbarger,* SPBAIME, StanfordU.
Henry J. Rarney, Jr., SPE-AIME,StanfordU.

Introduction
The trend in gas well testing has been to rely more ideal gas flow, including the effects of formation dam-
on the early-time flow data of drawdown and buildup age and turbulence. Two actual field cases were pre-
tests than on stabilized flow tests. The stabilized test- sented, both of which exhibited extensive fractures
ing methods often are not adequate for complicated and turbulent flow. For the pressure range of these
modern gas reservoirs because many of these reser- well tests, the use of ideal gas equations was accept-
voirs have extremeljj low permeabllities and the tran- able. These field cases were significant in being the
sient flow period of a well test can last for days or first published data showing the occurrence of turbu-
weeks. lence in fractured gas wells. We, also, observed such
It has become common to augment the flow capaci- cases in practice.
ty of gas wells by hydraulically fracturing their pro- Our purpose here is to extend the theory of frac-
ducing formations. In deep reservoirs the induced tured gas well testing to the flow of real gases. To
fractures are generally vertical and tend to follow a determine the effects of wellbore storage and turbu-
single plane of weakness. The presence of a vertical lence on well test interpretation, we developed a iinite-
fracture at the wellbore complicates the transient flow difference model to simulate well test conditions,
behavior of a low permeability gas well. The flow is Since gas wells usually are widely spaced and have
further complicated when turbulence occurs near the high compressibility, the emphasis has been put on
wellbore. the early transient behavior before the effects of the
Russell and TruitU published transient drawdown outer boundary are noticeable at the wellbore.
solutions for verticaUy-fractured Iiquid weIIs. They
developed methods of drawdown and buildup testing The Mathematical Model
utilizing these solutions, which were based on xm- The geometry of our mathematical model is similar
merical simulation. Clark* applied the basic Russell- to that used by Prats.4 The well is centered in a circu-
Truitt solutions to analy% fractured water injection lar uniform reservoir. A vertical fracture of infinite
wells by falloff tests. Field examples were given to flow capacity penetrates the formation and passes
substantiate the method of analysis. through the wellbore. The wellbore itself is not im-
The Russell-Truitt solutions and well testing meth- portant. Fig. 1 shows a sketch of the problem. Fig. 2
ods can be extended to gas well flow. MiUheim and shows the idealized model.
Clchowicz’ presented the drawdown equations for The equation for the flow of real gas in a porous
medium with uniform porosity, . . .permeability
. and
● Now with Scientific Software Corp., Denver. thickness iss-r

In the analysis of field data it is important to bear in mind the proper relationship between
the radial flow period and the linear flow period. The transition between these periods is
quite long and can be misinterpreted as being the linear flow period. Another factor that
can complicate analysis is turbulence.

MAY, 1%9 Reprinted from May, 1969, Iesue of JOURNAL OF PETROLEUM TECHNOLOGY 62S
the fracture. Eq, 5Cshows the well production rate to
V“($vm)=y* . . . . . . (1) be a sum of the gas flowing from the sand face plus
the gas that evolves from the wellbore storage effect.
where m is the real gas pseudo pressure, m(p), given Solution of this problem is particularly difficult
by because of the wellbore boundary condition for the
constant-rate case. Because of the complexity of the
problem and the nonlinearity of Eqs, 1 and SC, it is
m(p) = 2 P’ dp’ . . (2) necessary to resort to numerical techniques to solve
{ Z(p’)p(p’) the problem.
o
and 8 is the Darcy law “correction factor” which Computation Technique
accounts for non-Darcy (or turbulent) flow and is
Due to the symmetry of the problem, it is necessary
defined by to consider only one quadrant of the physical plane.
;=8 +L.. . . . (3) Before applying finite differences to Eq. 1, a coordi-
nate transformation was made. The transformation,
The Forchheimer equation gives an expression that similar to that of Prats,4 is illustrated in Fig. 2, The
accounts for inertial effects in flow through porous coordinate transformation is a conformal mapping
media.s For two-dimensional flow the Forchheimer that has certain computational advantages. The trans-
equation is generalized to formed Eq. 1 becomes

–Vp=
( ~+BPk~
P
The quantity in parentheses is non-directional and is
‘‘)~;. . k
(4)
. .
The coefficient 8 has the same value in the u-v plane
. . . . . . . . (6)

equal to the reciprocal of & as it does for the corresponding point in the x-y plane.
The constant-rate drawdown case will be consid- The Jacobian, 3(u, v), represents the ratios of ele-
ered for the geometry in Fig. 2. The boundary and mentaI areas of the two planes. For this transforma-
initial conditions are: tion,
?TZ(X, y,0)=)7’Zi , . . . . . . . (5a) 3(U, v) = (sinh’ u + sin’ v) L’ . . (7)
~m - The transformation of Eqs. 5 is straightforward
— = O on the outer boundary
an and need not be dkcussed. Some details of the nu-
m(x:O, t)=%,ll (t), –L<x<L, t>O, merical procedure am given in Appendix A.
. . . . . . . . . . (5b) This particular computing plane was chosen b-
cause the coordinate lines lie along the streamlines
and isopotential lines in the physical plane. This is
,++=kh~(,
:),=0 six strictly true only for certain incompressible flow
cases4 but is very nearly true for compressible flow
at times after the flow near the wellbore has stabilized.
— % Vw Cp dm
() —
dt well
. (5C) Aligning the computing coordinates in this manner
reduces the numerical truncation errors. This is ac-
These conditions do not allow for pressure drop along complished by locating the mesh points such that the

VERTIC$:A~:ACTURE
/

PHYSICAL (Z2) PLANE

:t ‘m
il I
i II I +%>

I II I 5?
COMPUTATION [W2) PLANE
vERTICAL “*GO*
FRACTURS
m!
-2!#
~ --x
!-+4

z’ SEALED LOWER
BOUNDARY
CIRCULAR EXTERIOR
BOUNDARY 1 @13313
/
,/su13HTLY
ExTERIOfl
ELLIPTICAL
EOUNOARY
NO SCALE
~$ 0’ --u
%

Fig. l—Schematic sketch of vertical fracture flow model. Fig. 2-Coordinate transformation of verticai
fracture modei.

626 JOURNALOF PETROLEUMTECHNOLOGY


mesh spacing is small where the higher-order deriv- TABLE 1-COMPARISON OF RESERVOIR PROPERTIES
ativesare relatively large, thus consistently keeping the WellA.1● Vi#&l;’ Well B
truncation error small. Orlglnsl Hu@hOtl”l
L
Another advantage of the form of Eq. 6 is that the ~;;~:t:~~?p::;::b ~’a —-x%i-
2684
190. 190
5000
195
computer program can solve a number of problems Porosity, fraotion 0.084 0084 0,10
Water saturation, fraction 0.:; O.:j 0
that conformably ,map into the computational plane Nat pay, ft
~a; gravity, fraction
with the same boundary conditions. To solve these 111,: 111.:
problems, it is necessary only to change the calcula- k: md 0.158 0,10 lZ
B (dlvlded by 0.92), W 5.87 X 10M 9.27 X 10~ —
tion of :he Jacobian. For example, the radial-flow *These data, along with the fiald data shown In Fig. 9, were taken
well problem can be solved by using the Jambian from Raf. 3.
● ●Tha adjusted data (k and p) were used to fit tha flald date with

the model.
J(u, v)=e’”rwz , . . . . . . (8)
which represents the usual logarithmic transforma- for the liquid case, we then raise the question of the
tion. For the purpose of checking the computer pro- applicability of this theory to the real gas caae. The
gram, Eq. 8 was used to solve radial-flow drawdown best way to answer this question is by analyzing the
problems. The results checked with analytical solu- numerical results. For the numerical solution, hypo-
tions and with one-dimensional numerical Solutions.s thetical Well B was used. The properties of Well B
In particular, this method of checking the program are given in Table 1.
was essential for turbulent flow since no published
solutions exist for the fracture case. Red Gas Solutions
The first solutions consider constant-rate drawdown
Results of Simulation
without the effects of turbulence or wellbore storage.
Liquid Case Analogy The ~damterm is taken to be zero throughout this
The basic solutions for early-time and intermediate- paper. Fig, 3 shows several cases with drawdown vs
time transient flow are well known, At early times the in t plotted (the usual drawdown plot). The dimen-
system behaves as though all flow were normal to the sionless time is based on the known wellbore radius,
face of the fracture. This is the linear flow period. If rw.
the nonlinear effects of Eq. 1 were not important the These plots show that the fractures can be inter-
solution of the linear flow period would be preted as negative skin effects for the various fracture
lengths. If desired, these plots can be normalized to
plot on a single curve by using a dimensionless time

The dimensionless variables are defined in the No-


menclature. The dimensionless time is given in terms
of rWbut could have been expressed in terms of L. WELLB
The skin effect, which is shown in the second term, q~ = 0.01
results from damage to the formation at the fracture 14 r~ / rW “ 3,0!X
face. It is assumed in Eq. 9 tkdt the damage region is
close enough to the fracture face that the flow has . L/r_ = 4
stabilized. 12 ● L/rW = 10
After a period of time the transient flow changes a L/rw = 20
from linear to radial. The flow around the wellbore
stabilizes and the fractured well behaves as a well v L/r_ = 100
that has a large effective radius, rto’.This effective 10
radius is approximately equal to L/2 for a well in
-.
which the exterior boundary is sufficiently far from
the fracture.’-’ For the radial fbv period, the sohl- ‘: 8
tion would be ~
1 /
tnD(tD) = 1.1~ 10g tDr’) + 0.4045 + Sf,.c + &f~~,
6 - .
. . . . . . . . . . (lo)
if nonlinear effects could be neglected. The damage
skin effect, ~dm, has the same value as in Eq. 9. The
4 –
sf,.. term is the “negative skin effect” resulting tkom
the presence of the fracture. Its value is approximately
.
sf~.s~ =ln* . . . . . ...(n) 2 -

At longer times the effects of the exterior boundary I I * 1,


will stabilize to pseudo steady-state flow. Open-flow 0 2 8 10
potential tests — in terms of M(P) — can be used for log to
this type of flow, but we will not discuss this here.
Fig. 3-Effect of fracture iength on constant-rate
Having described the linear and radial flow periods drawdown.

MAY, 1969 627


based on L or rW’.From the usual skin effect analysis, LO
I I
it is possible to fmd L from a drawdown test,* when
sd~~= O. If &j.~# O,its value can be estimated from
the linear flow data ( ~~ plot): at the t‘= Ointercept,
which is extrapolated from the straight line of the v~ =
=
0.1

plot, ~dam is equal to the dimensionless pressure drop. ED WELL B


This can be seen from Eq. 9. Then L can be found ● Q. “ 1.0

by applying Eqs. 10 and 11 to the log t plot for the ● q. . 0.01


radial flow period. 0.01-
Numerous computer runs were conducted to test
the nonlinearity of the radial flow period, Unlike the
slope for true radial flow,617the slope of the radial
flow period tends to be too low in a fractured well.
Fig. 4—Early solutions of constant-rate drawdown in a
The kh computed from a well test would tend to be fractured well,
correspondingly high. This effect is rate dependent
and becomes rather significant at very high rates.
Lo
Cases were simulated at practical flow rates in which I I I

this error reached almost 10 percent. That is, the Ah


that would have been calculated from the drawdown
.-
test would have been almost 10 percent higher than
the true value. So the plot maybe misleading, The ~; -0
=
0.1 -

plot will be used later in the paper and ordinarily E“


waL B
should be used to obtain quantitative test results. ● t . 10,LWI
\

Log-Log Plots 0.01 -

A plot of log of pressure drop vs log t can be a useful


tool for qualitatively analyzing transient behavior 10 10z 10’ ,34 ,0$
(see Fig. 4). The different flow regimes can some- Ii
times be recognized by the shape and slope of the Fig. S-Effect of wellbore storage on early solutions of
curve. A plot of flow that is dominated by wellbore constant.rate drawdown in a fractured wall,
storage is a straight line with a slope of 1.0 (see Fig.
5). A plot of flow that is dominated by linear flow
is a straight line with a slope of 0.5. A plot of the
radial flow period is a gradually curving line, Exam- WELL B

ples of linear flow data on a log-log plot are shown


in Fig. 4.
WellboreStorage
14 I L=4. Oft
rw = 0.33 ft

One would not expect wellbore storage to be so im-


portant in fractured wells as it is in wells with low
flow capacities. The pressure drawdown responds
slower in fractured wells and the wellbore storage
/
effect tends to die out during the early flow.
Fig. 5 shows a simulation of an extreme case of /
wellbore storage in a fractured well. In this case the
wellbore storage effect obscures the linear flow period “ /
and the wellbore storage solution grades directly into 1’‘/
the radial solution. To detect this type of flow, the
log-log plot migk be helpful. Numerical results for
other cases show that the end of the wellbore storage
effect, at fDto’$
canbe approximated by
rw~ 2
tw’=6
()~ . . . . . (12)

when the wellbore storage soluticn grades into the


linear flow period.
TurbulentFlow
The most interesting effect studied was that of turbu-
lent flow. The effect of turbulence on either the linear “o 2 4 6 8 . 10
or radial flow periods is much too complicated for log t ~
analytical solutions. The numerical technique de- Fig; 6-Effect of turbulence on constant:rate drawdown
scribed in Appendix A was specially designed to ..
fcr a fractured well.

6%” JOURNAL OF PETROLEUM TECHiOIXX3Y ,


model accurately the turbulence in a fractured well. shown ra&ating from the origin in Fig. 7. These tur-
Some representative numerical solutions are shown bulence solutions plot as straight lines parallel with
in Fig. 6. The solutions reach a straight line (on the the laminar solution on a log-log plot. There does not
radial In t plot), after which the turbulence behaves seem to be a straightforward method for correlating
like a skin effect. The amount of the turbulent skin the slopes of Fig. 7. It is important to recognize, how-
effect, s,.,,, depends on the flcjwrate. There is also a ever, that the turbulence causes the slopes to be too
slight dependence of ~~.~bon the gas viscosity near high, so that when this slope is used to determine the
the wellbore. This viscosity effect causes the slope of fracture length, the calculated fracture length will be
the in t drawdown curve to be too high, just as in the too small.
truly radial case, and can lead to an erroneously low
computed value of kh. Discussion of Well Test Application
A method of correlating the value of s,.,~ with the A detailed description of a method of analyzing draw-
physical properties of the system is useful. From nu- down tests for fractured gas wells was given by Mill-
merical results, it was determined that an accurate heim and Cichowicz.s Their analysis was based on
approximation is ideal gas equations. The conversion to real gas flow,
in terms of m (p), is straightforward, Although we do
s~”~b= 0.92 D(p)g . . . . . (13) not recommend any fundamental departures from
The value of D(P) is exactly the same as the truly their method, there are several points to be brought
radial case except rw’ (equal to L./2) is substituted out that might possibly prevent misinterpretation of
for rw. This substitution was suggested by M.illheim field data.
and Clchowicz,s but the factor of 0.92 is an empirical
result of our numerical solutions. Duration of Linear and Radial HOW
Turbulence tends to prolong the time required to The same problems arise in choosing the proper
reach the straight-line portion of the drawdovm curve. straight-line portion of a ~~plot as arise in choosing
This time (the beginning of stabilized turbulence) can the proper straight-line portion of a In t plot. With
be approximated by t~~s 100 or by imperfect field data, one can sometimes fmd several
regions that seem to be straight lines. To locate the
t(hr) = 4 ~ lo, —.. +PCLZ .. (14) proper straight lines, it is helpful to know where the
k
~Tstraight line occurs with relation to the in t straight
Turbulence takes an interesting and rather unex- line.
pected form for the earlier linear flow period. The
Fig. 8 shows the ~; plot and the In t plot on the
turbulence never stabilizes, unlike in the radial case.
same graph. These data are taken from Russell and
This much can be deduced by integrating Eq. 4 for
steady-state flow. The numerical results of Fig. 7 Truitt,’ Notice that there is a rather long transition
show that the turbulence solutions are straight lines period between the linear flow and the radial flow.
For this basic case, the transition period occurs when
on the ~~ plot. The effect of turbulence keeps in-
creasing until the flow departs from the linear pattern.
The solutions for three different /3-factors are 0.065< TD’ <1.0 . . . . . . (15a)

log t ok
1.21 -d -2 0
.- 2 4
1 I 1 I / 1 I
WELLB t I I I I I I 1 I
L=4
2.0 -
q= 13.94 MSCF/D/FT
Lo -
+

+
0.8 -

e
= 0
n 0.6 =
E
: ].0

0.4“

0.2 -

o 1 2 3 4 5 6
o 0.7
0.1 0.2 0.3
6’ Jr0”4 ‘“5 0“6

Fig. 7—Turbulence during linear flow period for Fig. 8-Linear and radial flow periods for a fractured wall
various rates. (from Ref. 1).

MAY, 1969 629


and 2, The stabilized turbulence can be directly related
0.45< rnD(tD) <1.10 . . . . . (15b) to the effective wellbore radius, r,,’ of a fractured well.
The second relationship is important: the pressure 3. The proper relationship between the radial flow
drop is over twice as great at the beginning of the period and the linear flow period should be kept in
radial flow period as at the end of the linear flow mind in analyzing field data. The transition between
period. If this relationship is not kept in mind, the these periods is rather long and can be misinterpreted
transition period could easily be mistaken for the as being the linear flow period.
4, It is likely that turbulence within the fracture
linear flow period on the ~;plot, Fig. 8 demonstrates
itself is sometimes more important than turbulence
how straight t-histransition period appears. Once the within the formation,
proper straight line has been located on the In t plot,
5. It is sometimes useful to match complicated
this information can be used to locate the proper well performance with a computer model to deter-
straight line on the ~; plot, Appendix B shows an mine flow parameters.
example of a field case in which the interpretation of
the ~;plot is difficult. Nomenclature
isothermal compressibility, psi-l
Natureof Turbulence
So far we have referred to turbulence as though it v.
wellbore storage constant,
occurs entirely in the formation, If this is the case 2m#hr,,Z
the turbulence effect should be directly correlatable to turbulence term coefficient (Mscf/D) -i
the formation turbulence factor, ~. This factor is cor- formation ‘tickness, ft
related with k and + as shown in Ref. 8, If the /3 Jacobian Gicoordinate transformation
computed from a well test is drastically higher than permeability to gas, md
the k, + correlation indicates, one might suspect that one-half fracture length. See Fig. 2.
something other than the formation itself is contribu- real gas pseudo pressure, Eq. 2, psiz/cp
ting to the turbulence. dimensiordess real gas pseudo pressure
The model we have used does not include the flow
in the fracture itself. In actual cases there is a high 1.987 X 105~ kh ‘i;%
likelihood that turbulence will occur as the gas flows P,cT
down’ the fracture toward the wellbore. Extremely index number of exterior boundary
high velocities are attained in the fracture if the entire mesh point
production rate from a fractured well is “funnelled” pressure, psi
through a narrow fracture. The Reynolds numbers in flow rate, Mscf/D
the fracture can be thousands of times greater than in skin effect
the formation, The amount of turbulence that occurs time, hr
in the fracture depends on the width of the fracture
and the properties of the propping agent. The field 0,000264 h
dimensionless time,
examples given in Ref. 3 showed calculated ~’s that @cL2
were about 100 times as great as those given in the 0,000264 kc
correlation in Ref. 8. This indicates that most of the dimensionless time,
@crw2
reported turbulence might have been occurring in the
fracture rather than in the formation, formation temperature, ‘R
computation space coordinate
ComplicatedField Cases computation space coordinate
There are so many parameters involved in analyzing macroscopic gas velocity
a fractured gas well test, that interpretation can often wellbore volume, cu ft
be complicated. This is especially true if turbulence gas deviation factor
occurs and the duration of the testis not long enough turbulence parameter, ft-’. See Ref. 3
to give good data for the radial flow period. It is for equation.
helpful to use computer simulation to match the data turbulence coefficient, E@.3
in these cases. The parameters of the model are ad- gas viscosity, cp
justed until a good match is obtained. An example gas density
of this type of data-matching is given in Appendix B. formation porosity, fraction
This example case was extremely difiicult to analyze
without using the modeling technique. Acknowledgments
We should like to thank the National Science Foun-
Conclusions dation and Stanford U. for tinancial support during
The results of this study have led to the following this work.
inclusions:
1. The drawdown testing method for vertically References
fractured wells can readily be extended to the real gas 1. Russell, D. G, and Truitt, N. E.: ‘Transient Pressure Be-
havior in Vertically Fractured Reservoirs”, J, Pet. Tech.
case by using the m(p) function. At high rates, how- (Oct., 1964) 1159-1170. -
ever, the calculated IA will tend to be too high in the 2. Clark, K, K.: “Transient Pressure Testing of Fractured
absence of turbulence. Water Injection Wells”, J. Pet. Tech. (June, 1968) 639-643.

630 JOURNAL OF PETROLEUM TECHNOLOGY


3. Millheim, Keith K. and Cichowicz, Leo: “Testing and any trial solution, the integral in Eq. 5Cwas evaluated
Analyzing Low-Permeability Fractured (3ss Wells”, J, Pet. by
Tech, (Feb., 1968) 193-198.
4, Prats, M,: “The Effects of Vertical Fractures on Reservoir
Behavior — Incompressible Fluid Case”, Sot, Pet, Eng,J.
(June, 1961 ) 10S-118.
5. Wattenbarger, Robert A. and RameY,H. J., Jr.: “Gaswell
Testing With Turbulence, Damage and Wellbore Storage”, 3 t?%,
j (m,, j–m,, j) – &,,, j(rt2, j–ml, j)
J. Pet, Tech, (Aug., 1968) 877-887. 2Au
6. A1-Huwainy, R,, Ramey, H, J., Jr,, and Crawford, P. B.: . . . . . . . . . . (A-3)
“The Flow of Real Gases Through Porous Media”, 1. Pet.
Tech. (May, 1966) 624-636.
7, Wattenbargerj Robert A.: “Effects of Turbulence, Well-
bore Damage, Wellbore Storage? and Vertical Fractures The coefficients A ~ are selected to give a high-order
on Gas Well Testing”, PhD Dmertation, Stanford U,, integration according to Bode’s rule.7
Stanford, Calif. (1967). The computation procedure for each time step was
8, Katz, D. L,, Cornell, D., Kobayashi, R., Pcettmann, F. H., as follows: ( 1) guess a well pressure that applies to
Vary, J. A., Elenbaas, J, R. and Weinaug, C. F.: Hand- the entire interior boundary, (2) solve the resulting
bmk oj Nafuraf Gas Engineering, McGraw-Hill Book Co.,
Inc., New York (1959). matrix equation, (3) calculate the corresponding rate
according to Eqs. SC and A-3. If the computed rate
APPENDIX A is within 1 percent of the required rate, the solution
The finite-difference equations were developed from is acceptable; otherwise, the process must be repeated
the viewpoint of discretizing a continuous function, with another guess of well pressure. The nonlinear
m (u, v, t), rather than from the “cell” approach. coefficients were re-evaluated at the n+ 1 time Ievel
Often both derivations lead to the same result. In this each time the process was repeated.
case they do not, The present derivation allowed the Several techniques were tried for solving the matrix
model to have second-order accuracy throughout the equation.7 The alternating direction iterative proce-
u, v space, including the boundary conditions. It was dure (ADIP) and line successive over-relaxation
particularly important to have a high-order derivative (LSOR) were the most successful. LSOR, which was
approximation at the wellbore during the early flow preferred because of its simplicity in usage, was satis-
period and during turbulent flow. factory only when the lines of simultaneous solution
Eq, 6 was put in an implicit finite-difference form. were taken in the u-direction. When these lines were
Using the usual notation, the finite-difference eciua- taken in the v-direction, the solutions were very er-
tion can be written for the point i, j as roneous and tended to drift away from the correct
solutions, in spite of the fact that a reasonable con-
vergence criterion was met.
A peculiar problem was encountered with early-
time solution. Due to the ditliculty in approximating
the gradient of the linear-flow solution with our net-
work of “imbedded parabolas” (inherent in the finite-
differer.m method), the first few time steps would
give pressure drops that were too high. This errone-
ous solution would continue at almost constant pres-
sure until this pressure coincided with the proper
solution; then the remaining time steps gave proper
To compute the nordine~r coefficient, 8, a simplify- solutions. This erroneous early solution was found to
ing assumption was made. It was assumed that the be dependent on u-direction mesh spacing. Correct
magnitude of the velocity vector in Eq. 4 was equal earlier solutions could be obtained by a su5cient
to the magnitude of the velocity component in the reduction in mesh spacing. This same sort of problem
direction for which 8 was being computed. This as- was experienced by Russell and Truitt.l They resorted
sumption is not physically correct because 8 is not to the analytical solutions of these early times, We
directional. However, since the u, v coordinates were used a linear finite-difference model for the results
selected so the stabilized flow is along the u coordi- shown in Fig. 7.
nate, the assumption is nearly correct. The u-com-
ponent velocity is very nearly equal to the total APPENDIX B
velocity vector, making the calculation of the u-direc- Millheim and CichowiczSpresented field data for two
tion gradient accurate. There is very little flow along gas well tests. These data included plots of several
the v dmection, making the vdirection gradient rela- drawdown tests for each well. The drawdown tests
tively unimportant. were conducted at different rates to evaluate the tur-
The no-flow boundary conditions were approxi- bulence effects. From these data, they used their meth-
mated with reflection equations of the type od of analysis to calculate the flow parameters, /3, L,
and k. Fig. 9 shows their teat data for the highest flow
MN+I,J =mw-l, j . . . . .. . (A-2) rate on Well A-1, in terms of m(p). Table 1 gives
The gradient at the wellbore was approximated by some data that they reported for Well A-1.
a three-point formula of second-order accuracy. For We used the computer model to simulate the be-
MAY, 1%9 631
log I - days ‘ adjusted still more in other parameters to match the
4 -3 -2 -1
I 1 f f
0
I
I
I
2
data. However, the closeness of the k = 0.10 md run
?. 5 - b ● Fueld OatA Rd. 3 seems to indicate clearly that the well’s permeability
a
—— hkjjel win Or,$nal Parameters i Table II
‘1
is lower than originally calculated.
z
— h!adel wlh Adjusted P8raMelel$ [Table I) A study of Fig. 9 gives a feeling for the complexity
: ?.0
z‘t q , 2955 hE.CF I D
i of analyzing this type of test data. Although the
-- !
method of analysis is valid, it seems that the original
-.
=
.
1.5
I [ARI.Y

‘Low
RAOIAL

‘fR’Y . / /
/

4
straight lines for both the linear and radial flow per-
iods were in error. The linear flow period ends at
about 4 hours. The rest of the ~~ plot is a transition
period (although it appears to be a straight line, just
as in Fig. 8). The slope of the adjusted model is the

“’0”5L2KzcE5!E_l
o 1 2 3

./i -
4- 5 6 7 8
same as the field data for this linear period. For radial
flow, the flow period shown on the log t plot is before
the proper radial straight Iine occurs. Since it was
hours
clearly established that significant turbulence was oc-
Fig. 9-Matching field data with model to determhe curring in this well, the beginning of the straight line
proper system parameters, is given by Eq. 14. This indicates that the proper
straight line does not begin until long after the test is
havior of Well A-1. First, the model parameters were over. The “straight line”, which occurs at log ts O,
taken from Ref. 3 to see if these parameters described is in a period that is effeeted by transient turbulent
the pressure data that wore observed. Fig. 9 shows a behavior (see Fig. 6) and has a slope that does not
serious disagreement bet ween the model data and the properly reflect the formation kh.
field data. This disagreement was taken to imply that It seems that almost all of the data shown in Fig. 9
the parameters used did not truly describe the well. are in a transition period between linear and radial
The model was then adjusted to use k = 0.10 md flow. This behavior makes accurate well-test analysis
rather than k = 0.158. To make the computation difficult. Simulating well tests with the computer
consistent, /3 was adjusted inversely to k. ‘Thisadjust- model in this manner seems to be a practical method
ment yields the same turbulence ~s the original of checking interpretation and adjusting parameters
parameters. to fit the observed data. The computation costs are
Fig. 9 shows that the model simulates the field data vexy small. iTPT
fairly well after the permeability is lowered. Through-
out the duration of the test, the model solution lies Original manuscript received Aug. 6, 1968. Revised manuscript
below the actual data by a wmstant amount. This receivad Feb. 20, 1969. Paper (SPE 2165) was presented at SPE
43rd Annual Fall Meeting held in Houston, Tex., SePt. 29.Ott. 2,
difference is due to a value of ~crs~that was present 196* and at SPE 39th Annual Cellfomla Regional Fall Meeting
held In Bakersfield, CalIf., Nov. 7% 196& @ Copyright 1969 AmerL
in the well but that was not included in the model. can Institute ef Mlrdng, Metallurgical, and Petroleum Engineer%
Otherwisee the model fits the data fairly well through- Inc.
This paper wIII ba printed In Transactions volume 246, which
out the test. We point out. that the model might be will cover 1969.

632 JOURNALOF PETROLEUMTECHNOLOGY

You might also like