You are on page 1of 12

Well Pressure Behavior of a

Naturally Fractured Reservoir


Tatiana D. Streltsova, SPE. Exxon Production Research Co.

Abstract
The pressure response pattern of a naturally fractured the matrix flux contribution to fracture flow may be
reservoir is considered under the assumption allowing assumed proportional to either the pressure difference
matrix-to-fracture crossflow to result from a diffusion between matrix and fracture or to the averaged pressure
mechanism of fluid transfer through the matrix. The gradient throughout the matrix block. The former
transitional pressure during time-variant cross flow is assumption, introduced in fractured reservoir description
shown to develop on a semilog plot a linear segment with by Barenblatt and Zheltov I and Barenblatt et al. 2 and
a slope equal to one-half that of the early- and late-time employed by Warren and Root,3 has an advantage of
pressure segments. For a single well, this allows use of a simplifying the mathematical analysis of the flow prob-
conventional Homer-type analysis. lem and a disadvantage of not correctly representing
either the mechanism of pressure readjustment between
Introduction matrix and fracture by time-variant crossflow or the for-
A naturally fractured formation is generally represented mation pressure response during the transitional time.
by a tight matrix rock broken up by fractures of secon- According to this assumption, the matrix flux is indepen-
dary origin. The fractures are assumed continuous dent of spatial position, which can be true only when
throughout the formation and to represent the paths of pressure is linearly distributed in space-i .e., at a state of
principal permeability. The high diffusivity of a fracture pressure equilibrium or at a pseudosteady-state time.
results in a rapid response along the fracture to any This assumption, therefore, is often referred to as a
pressure change such as that caused by well production. "pseudosteady-state" or "lumped-parameter" flux
The rock matrix, having a lower permeability but a assumption. It neglects the matrix storage capacitance by
relatively higher primary porosity, has a "delayed" allowing an instantaneous pressure drop throughout the
response to pressure changes that occur in the surround- matrix as soon as fracture depletion occurs. The pressure
ing fractures. Such nonconcurrent responses cause response of a medium subject to this assumption has a
pressure depletion of the fracture relative to the matrix, characteristic S-shape transitional curve with an inflec-
which in tum induces matrix-to-fracture crossflow. This tion point. The curve connects the initial pressure seg-
period of transient crossflow takes place immediately ment (the early-time fracture response) to the final
after the fracture pressure response and before the matrix pressure segment, representative of the late-time
and the fracture pressures equilibrate, after which the pseudosteady-state flow of an equivalent uniform
formation acts as a uniform medium with composite medium that has fracture permeability and composite
properties. (the sum of fracture and matrix) storage.
The effect of assumptions made on the nature of By contrast, the averaged gradient assumption on
matrix and fracture interaction is manifested during this matrix-to-fracture crossflow, while somewhat com-
transitional period of matrix-to-fracture fluid transfer. plicating a mathematical analysis of the problem, has an
The flux of fluid released by the matrix depends on the advantage of more correctly describing the pressure
matrix size, porosity, permeability, and the matrix/frac- equilibration process that occurs during the transitional
ture pressure difference. At the matrix/fracture interface, period. Matrix fluxes arising from fluid expansion forces
are subject to Darcy flow and, thus, to diffusivity-type
0197-7520/83/0101-0782$00.25
flow constraints. As such, they are proportional to
Copyright 1983 Society of Petroleum Engineers of AIME pressure gradients which, in general, are not constant but
OCTOBER 1983 769
dependent on spatial position as long as the storage during the transitional stages of flow in naturally frac-
capacity of the medium influences the response time. It tured reservoirs because fluid expansion forces rather
is a diffusion-type process that allows the matrix to than gravitational forces dominate the tluid flow. Only
replenish the depleting fracture. Matrix pressure and, when the pressure becomes linearly distributed in space
consequently, flux are not linearly distributed along a does the gradient become a constant. In naturally frac-
direction normal to the fracture plane. The spatially tured reservoirs, this situation corresponds to a time
dependent flux within a matrix block is subject not only when matrix and fracture pressures are at equilibrium.
to matrix permeability but also to matrix storage. The The formation then responds in a pseudosteady-state
averaged gradient assumption is, therefore, a manner, acting as a uniform medium with composite
"distributed-parameter" flux assumption. Application properties.
of this assumption 4 leads to a transitional pressure In general. analysis presented in this paper can be ap-
response that is different from that of a lumped- plied to a two-component interacting continuum whose
parameter model in both the time of onset of the matrix intensive variables are governed by diffusion-type equa-
pressure support and the shape of the subsequent transi- tions. For example, in heat conduction, the distributed-
tional pressure curve prior to pseudosteady-state medium parameter formulation would embody Fourier's law
response. The distributed-parameter or "gradient" flow relating heat flux to temperature gradient, whereas the
model predicts transitional pressure which, when plotted lumped-parameter formulation would employ Newton's
vs. logarithm of time, gives a linear segment that con- law of cooling relating heat flux to temperature dif-
nects the early- and late-time pressure curves without an ference. In electrical fields, the distributed-parameter
inflection point. Likewise, no inflection point is noted in formulation would correspond to resistance-capacitance-
the calculated transient-pressure curves from models of type response of a medium while the lumped-parameter
Kazemi,5 deSwann,6 or Najurieta,7 where matrix-to- formulation would embody Ohm's law relating current
fracture influx is taken in a nonsteady-state form. In the to a potential difference between the points of the
latter works, however, the theory is insufficiently medium. The approach taken in the paper can be extend-
developed to utilize the transient-pressure segment quan- ed to such diverse fields of diffusional heat and mass
titatively in the data analysis. It is proved here analytical- transfer as pollutant dispersion, chemical mixtures,
ly that the slope of this transitional straight line on a plant-cell water transport, blood diffusion, etc.-i.e.,
Homer plot is equal to one-half the slope of the early- whenever two distinct flow pathways are present. The
and late-time straight line." This transitional segment appropriateness of a particular formulation (linear or dif-
thus allows application of a conventional Homer-type fusive type of component interaction) will depend, first,
semilog analysis, which is much simpler than the use of on the time region of interest (transitional or composite)
either type curves or nonlinear regression technique re- and second, on the disparity of the relevant component
quired for a curve-fitting procedure. parameters, which determines the duration of responses
This paper examines in detail the differences in ap- of each component and the time of their interaction.
plication of the two flux models and points out advan-
tages of the gradient model in describing and analyzing
Analysis of Assumptions on Pressure
actual pressure behavior in a naturally fractured forma-
Response Mechanism of Matrix Block
tion. The model is supported by examination of pub- A fractured reservoir is usually viewed as a double-
lished well-pressure response data,9-11 including the porosity medium where the transport properties of the
original example of Warren and Root. 12 The published rock are attributed almost entirely to the presence of frac-
data show, as a rule, a long linear transition between the tures, while the storage of transmitted fluid is associated
early- and late-time segments. The inflection point is not with intergranular primary or matrix porosity. Serving as
present. Examples of Crawford et al. 10 are reanalyzed permeability channels, fractures control the tluid-
here in light of the new model by application of a pressure distribution. Changing fracture pressure results
Homer-type analysis, which is much easier than the in pressure differentials across the matrix. This causes a
regression analysis used in Ref. 10. time-dependent matrix -to-fracture flow.
It is worthwhile to note, however, that the lumped- Two assumptions on response of matrix blocks to
parameter flux assumption does correctly represent the pressure changes at the fracture/matrix interface are
transitional behavior of slowly moving underground discussed here. The first assumption, used in the Warren
fluids undergoing gravitational drainage, where a fluid and Root model, determines matrix-to-fracture flow rate
density difference provides the gravitational force. Con- as proportional to the pressure difference between matrix
sequently, a gravitational replenishment will take place and fracture. The second assumption, employed in the
whenever the interface between immiscible tluids of dif- gradient-flow model, considers matrix -to- fracture flow
ferent density (air/water, gas/oil, oil/water) is moved governed by a linear-diffusion process. Pressure
from its equilibrium position. The pressure becomes distribution in the matrix block, the matrix-to-fracture
nearly hydrostatic and varies linearly with depth. The flow, and the characteristic pressure pattern of a natural-
flux is then directly proportional to the pressure dif- ly fractured formation that result from these two models
ference since the gradient becomes a constant, indepen- are analyzed next.
dent of depth. Fluid flow of this type occurs in uncon- Warren and Root Model
fined aquifers and in gas-cap or bottomwater-drive reser-
The boundary condition at the matrix/fracture interface,
voirs. In each case, the formation can be vertically
z=O, is
stratified. These constant pressure gradients do not occur
at::.p 111 (z) k", at::.pl1I(':)
*In a contemporaneous work, 8 Serra et at have also established the half-slope of a v",(z)=¢l1I c "' H - - - - .... (I)
transitional pressure response and desCribed its practical utility at J-t az
770 SOCIETY OF PETROLEUM ENGINEERS JOURNAL
where v(z) is a fluid volume per unit area per unit time where s is the Laplace parameter.
released by the matrix. The solution of Eq. 7 in Laplace space is:
If Interface Boundary Condition I is assumed to repre-
sent a cross flow proportional to the fracture/matrix I1p m(Z) =l1p[cosh(H - z),jS/'Y/ m]/cosh(H,jS/'Y/ m),
pressure difference, then integrating Eq. I along z from 0
to H to give an average pressure, I1p m, across the matrix ........................ (8)
block half-thickness, H, leads to the following lumped
crossflow relationship. where the following boundary conditions are used:

I1Pm =l1p at z=O

and

or (jI1Pm
- - = 0 at z=H.
(jz
kill
--[I1Pm(z)]
IH -CPmcm
_ H 2 (jI1Pm
--
~ 0 ~ Eq. 8 gives the pressure distribution across the matrix in
response to a pressure change at the matrix/fracture in-
terface, I1p.
Since I1p=l1p", at z=O, the preceding equation gives The matrix -to-fracture flow, Vm' at the block face
z=O, from Eq. 8 is
k 11/ I1p - I1p 11/
- - - - - - - - =ex(l1p-I1Pm), ... (2)
~CPmcmH H _ k m (jl1p", k m _ r-;- r-;-
Vm = - - - - =-l1pvs/'Y/m tanh(Hvs/'Y/m)·
~ (jz ~
where ex='Y/m/H2, 11m =kll)CPmcm~ . ............ (3) ........................ (9)

This boundary condition was used by Warren and


Root 3 (Eq. 9, Page 248), with parameter I/H2 as ex, to In the inverted form, these solutions (Eqs. 8 and 9)
represent the matrix-to-fracture flow rate. are, respectively,
Boundary Condition 2 implies an instantaneous adjust-
ment of flux to changes in the pressure difference be-
tween matrix and fracture. Only the resistance to flow
across the matrix, H/k m , attenuated by the matrix
storage, Sm =CPmcmH, is assumed to affect the flux
magnitude at any time.
A solution to the first-order linear Eq. 2, provided ex is
a constant, is the time-dependent pressure in a form
and
I1pm =!1p(l-e -al), ....................... (4) 00

which gives the interface flux, v m' as L:


11=1.3.5

................. (11)

The pressure distribution (Eq. 4) and the flow rate (Eq.


5) will be compared with those from the gradient-flow Comparing the matrix pressure distribution, I1p "I'
model. from the two models-i.e., comparing Eq. 4 with Eq.
lO-one can see that in the gradient model, the pro-
The Gradient-Flow Model gressive leveling of matrix pressure is dependent on the
The equation for the matrix pressure distribution, I1p 11/' relative coordinate, z' =z/ H, in addition to the time-
assuming only vertical flow to a fracture, dependent combination, ext, common to both models.
The dimensionless pressure distribution, I1p m(z')/ I1p,
1 (jl1p III calculated from Eq. 10 for various ext values, is plotted in
--- ...................... (6) Fig. 1 vs. the dimensionless coordinate z' from 0 to I.
'Y/m (jt This figure shows that the pressure change over the
matrix block thickness is substantially nonlinear. This
when subjected to the Laplace transform becomes does affect the flux at the matrix face. On the other hand,
the solution from the Warren and Root model (Eq. 4)
(j211Pm S
gives no change in pressure magnitude across the matrix
(jZ2 = ;:::I1Pm, ....................... (7) thickness, implying at any time an instantaneous
pressure adjustment throughout the matrix block
OCTOBER 1983 771
0.6
~

~
E
~
<l
0.4
...
<l

...e
<l

0.2

1.
Z' = z/H at

Fig.1-Dimensionless pressure distribution, t1Pm(z')/t1p, Fig.2-Dimensionless matrix pressure, t1Pm(z')/t1p, as a


over the matrix block thickness (Eq. 10). function of cd for a given relative coordinate, z' (Eq.
10).

thickness. To compare the pressure behavior in time for fracture pressure responses from both the Warren and
these two models, the dimensionless pressure from Eq. Root and the gradient models are considered.
10 is replotted in Fig. 2 as a function of at
for a given To assess the validity of the lumped-parameter
relative coordinate, z'. The solid-circles line in Fig. 2 is crossflow assumption, we compare the matrix flow rate
the dimensionless mean pressure value obtained by in- (Eq. 5) with that of Eq. II. If only the first term in the
tegrating Eq. 10 over z from 0 to H: series (Eq. II) is taken-i.e.,

8 I v m =(2I:1pk mlH) e-(O.57r/H)'YJ,/, ............ (13)


1:1pm =l:1p [ 1 - -2
.6 -2
7r n= 1.3.5 n then the crossflow (Eq. 13) will be identical to that of
Eq. 5 provided a and Sm are each multiplied by a con-
stant such that
.exp [ - (:7r) 2 at]] . .............. (12)
a=O.257r2YJmIH2 ........................ (14)

The open-circles line plotted in Fig. 2 is the dimen- and


sionless pressure, I:1Pmll:1p, calculated from Eq. 4 of the
Warren and Root model. Sm =8¢mc mH17r2. . ....................... (15)
Examination of these two lines shows that the matrix
block pressure response predicted by the Warren and Qualitatively, the first -term approximation (Eq. 5)
Root model lags significantly behind that given by the compares favorably with Eq. II only when t is large.
gradient model. The effect of the matrix block on the The assumption of a lumped cross flow relationship is,
overall behavior of a fractured formation is thus con- therefore, valid only for the late-time periods. For exam-
siderably delayed. Accordingly, the transitional pressure ple, for at=O.5, the first term of Series 11 is
portion that depicts the effect of matrix would deviate exp( -O.257r 2 )=O.2912, while the second is exp
much later from the first pressure segment, the initial Ei- (-I.I257r 2 )=O.OOO0l5. However, for at=O.005, the
curve in the Warren and Root model, than it would from first term is exp( -O.OOI257r 2 )=O.9877, the second is
the gradient model. At late times, however, the two lines exp(-O.OII257r 2 )=O.8949, the third is
on Fig. 2 converge. Being delayed in response at early exp( -O.03757r 2 )=O.7346, and the term for n= 15 still
times, the matrix contribution from the Warren and Root gives a contribution of exp(-O.28l7r 2 )=O.0625.
model must subsequently change at a more rapid rate Therefore, for small values of time, when convergence is
than does that from the gradient model. This results in a slow, it is necessary to take several terms in Series II.
leveling and an inflection point in a transient -pressure The first-term approximation, which is the assumption of
curve from the Warren and Root model. By similar the Warren and Root model, will inevitably modify the
reasoning, the transient curves of the Warren and Root early-time pressure behavior.
model would less rapidly approach the third, final Ei- Quantitatively, the validity of the first-term approx-
type pressure segment than would transient curves from imation can be assessed with a given accuracy for a
the gradient model. This qualitative difference in rapidly convergent series. Assuming, for example, that
pressure pattern is illustrated in the next section where the second term in Series 11 should be no larger than
772 SOCIETY OF PETROLEUM ENGINEERS JOURNAL
10% of the first term, we obtain the following where s is the Laplace parameter, a is the Hankel
relationship. parameter, and an asterisk denotes a quantity subjected
to both Laplace and Hankel transformations.
exp( -9x):s 0.1 exp( - x), The quantity VIII * is found by applying the Hankel
transform to Eq. 9:
which is satisfied whenever
* kill * ~ ~
vm = -flp "s/1/ 11/ tanh(H" s/1/ 1/1)' .•..••••• (22)
/A
Thus, the first-term approximation is satisfactory when
at is sufficiently large that Substituting Eq. 22 into Eq. 21 and applying the in-
verse Hankel transform in the form
at=(1/ m/ H2 )t? 0.1167, ................... (16)

which may be expressed as H- 1 ( 2I 2) =Ko(sr),


a +s
................. (23)

(r/H) 2 ?0.1167r2/1/mt .................... (17)


where K 0 denotes the Bessel function of the second kind
or of zero order, we arrive at the general solutions for the
fracture pressure distribution, flp, and the matrix
H:s 2. 93.J1/ mt. ........................... (18) pressure distribution, flp 1/1' in dimensionless Laplace
space as, respectively, 4
Therefore, the first-term approximation is valid only
when the block size is sufficiently small to be contained
within the area of pressure depletion (the radius of in-
flp=-q-~Ko
27rT s
[r s+rD.J s¢mclII/¢t,ct(H/h)
' ,
vestigation) at any particular flow time. Eq. 18 means
that the matrix block has attained a pseudosteady-state
condition with an instantaneous adjustment of fluxes to
tanh(.J s¢ I/IC 111/¢ fCf/ r D)] '/2] ..... ,..... (24)
interface pressure changes. Only the resistance to flow
across the half-thickness H, klll/H, attenuated by the and
matrix storage, ¢mcmH, is affecting the flux magnitude
at any time. Such condition, as mentioned in the In- flp m =flp[tanh (H.J s/1/ 1/1 )]/ H.J s/1/ Ill' .. , ..... (25)
troduction, does correctly represent the transitional
behavior of gravitational replenishment. In a naturally where rD=rJkm/kt~/H, s=sr 2 /1/ is the dimensionless
fractured reservoir, this situation corresponds to a time Laplace parameter, and
when matrix and fracture pressures are at equilibrium. 11
The formation then responds in a pseudosteady-state
T= ~ (kfh//A); =2nk f h/ /A=kh 1/ /A.
manner acting as a uniform medium with composite ;=1
properties.
Eq. 24 derived for a single symmetric element consisting
Pressure Distribution in a Naturally of a single matrix block remains unchanged in form for
Fractured Reservoir the total thickness, hI, containing n identical blocks if q
The general differential equation for a radial fracture is interpreted as the total production and T as the effec-
flow augmented by a matrix flow contribution, V IIl , is tive transmissibility T=khl//A, since (q/kfh);=q/khl'
Although the inverted forms of these equations are
available (Ref. 4, Eqs. 14 and 18), Eqs. 24 and 25 in the
present study are in the most convenient form to
calculate values of flp(t) and flp III (t) with the numerical
Laplace transform inversion method presented by
where matrix-to-fracture flow, VIII' per unit area of the Stehfest. 13
matrix/fracture interface, z=O, per unit time is In the following sections, the fracture pressure
behavior (Eq. 24) is illustrated by application to single-
k m aflpm well and multi well tests. Then matrix pressure behavior
V m= - - - , z=O . ..................... (20) (Eq. 25) is considered. This would be the case applicable
/A az whenever a fracture does not intersect the wellbore.
Finally, the influence of a gas cap on the pressure
The equation for the matrix pressure distribution is behavior of a fractured reservoir is presented, Such in-
Eq.6. fluence is most pronounced in interference testing.
Eq. 19, subject first to the Laplace transform and then
to the Hankel transform of zero order, is Pressure Pattern for a Single-Well
Test Analysis
Dimensionless pressure values calculated from Eq. 24
for various r D and 4t D values are shown in Fig. 3 for a
matrix-to-fracture storage ratio, Sm/Sr= 10. At early
OCTOBER 1983 773
TABLE 1-DIMENSIONLESS PRESSURE VALUES, TABLE 2-2/lPa FUNCTION (Ea. 29) AND ITS
APm/Ap, FROM EXACT SOLUTION, Ea. 12, AND THREE TERMS, COMPUTED FOR ra =0.05
ITS APPROXIMATION, Ea. 27 AND VARIOUS 4t a = 4'1'tlr2 w VALUES
Eq. 12-Eq. 27 Term
Eq.12 4to 2/lpo 2 3
at Eq. 12 Eq. 27 (%) 1.0 2.823 -0.577 3.400 0.0
0.001 0.0316 0.0316 0.0 3.0 3.372 0.521 2.851 0.0
0.005 0.0798 0.0798 0.0 10.0 3.974 1.725 2.249 0.0
0.01 0.1128 0.1128 0.0 30.0 4.524 2.824 1.700 00
0.02 0.1593 0.1593 0.19 100.0 5.126 4.028 1.098 0.0
0.05 0.2518 0.2523 0.20 300.0 5.677 5.127 0.548 0.002
0.1 0.3558 0.3568 0.28 1000.0 6.424 6.331 - 0.536 0.147
0.2 0.5024 0.5046 0.44 3000.0 7.431 7.429 - 0.603 0.605
0.5 0.7643 0.7373 3.53 10000.0 8.633 8.633 -1.205 1.205
1.0 0.9314 1.1284 21.15 30000.0 9.732 9.732 - 1.754 1.754
1.5 0.9800 1.3820 41.02 100000.0 10.936 10.936 - 2.356 2.356

times, the dimensionless well pressure. 2flp D = we obtain an alternative solution to Eq. 24 for short and
47rTflplq (2flpD =flpTI70.6q, in oilfield units), when intermediate times in the form 15
plotted in semilogarithmic coordinates, forms an initial 00
straight-line segment representing response of com-
flplII(z.')=flp ~ (-1),,+1
pressible fractures to the well production. This first
,,=1
pressure segment is described by the Ei-curve associated
with the fracture parameters, - Ei( - r 2 /4y/t). y/ = (2n - I) - z.' (2n - I) + z.' ]
ktf p.,cP fC f= TIS being the fracture ~iffusivity (y/ =0.00633 [ erfc + erfc ,J;;z . . ...... (26)
2.Jm 2 at
ktfp.,CPfcf=0.00633 TIS, sq ftlD, In customary Units). ~s
time progresses, the matrix block, whose response IS At small values of O'.t. all the ternlS of the series are in-
delayed because of the low matrix block permeability, finitesimal except for the first. since the function erfc(a)
provides flow to the fracture with the result that the frac- quickly decreases with an increase in the argument. a;
ture pressure changes at a slower rate than shown by the for example, erfc(I)=O.I77. erfc(1.5)=0.036.
first segment, forming thus a transitional region where erfc(2) =0.005. and erfc(2.7)=0.0001. The mean matrix
pressure is affected by the matrix block flow contribu- block pressure (the closed-circles line of Fig. 2) for
tion. This transitional pressure curve is of linear small values of O'.t by integrating the first tenn in Series
character over most of its duration. A feature of par- 26 over z' value from 0 to I is:
ticular interest is that the slope of this transitional line is
one-half that of the first segment. As the matrix block flp 111 =2flp .JO'.tl7r. . ....................... (27)
and fracture pressures equalize, the matrix block in-
Eq. 27 very closely approximates the exact solution
fluence dies out and the transitional pressure curve
(Eq. 12) up to the time O'.t=0.5. as shown in Table 1,
develops into a final, ~hird pressure. se~me~t, which is
where the dimensionless matrix block pressures are
described by the El-curve. - El(r-/4y/ t), where
calculated for a given parameter O'.t. During this time. the
Y/* =ktfp.,(CPmcm +CPfcf)· The effective storage capaci~y matrix pressure and. consequently, the matrix block flow
of the formation is then equal to the sum of the matnx
contribution. is proportional to.J t . The slope of the frac-
block and the fracture storages. while the permeability of
ture pressure, which for a zero matrix contribution is
the formatiun is equal to that of the fractures. With the
proportional to In t, is then changed during this timeto
pressure in equilibrium within the radius of investiga-
half its value. Table I, in fact. shows that the in-
tion, the fluid replenishment comes from depletion of
termediate pressure curve is quite uniform. maintaining
matrix blocks laterally more distant from the well-i.e.,
this one-half slope during almost the entire transient
beyond the radius of investigation.
process as the matrix and fracture pressures conver~e.
To show the validity of the one-half quantitative rela-
For O'.t= 1, the deviation from one-half the slope in-
tionship between the slopes of the transient and limiting
dicative of the true fracture permeability is only 21 %.
pressure curves, let us return to the transform (Eq. 8).
but O'.t= I corresponds to the time when matrix pressure
which is more convenient for analysis of pressure at
is almost equal to the fracture pressure (see Fig. 2). If
small values of time when matrix flow exerts its max-
one assumes that the equalization in pressures is com-
imum influence. Expanding IIcosh(a) in Eq. 8 into a
plete at O'.t=2.5, when flplIIl flp < 1%. then the transient-
series 14
pressure segment merges with the final, the third.
00
pressure segment at time t,
I/cosh(a)=2 ~ (-I)"+l exp[-(2n-l)a],
11=1 t=2.5H 2 /Y/III' ............................ (28)
and using the inversion formula
where t is in hours if the matrix diffusivity is
L -I [~e-cJ";J =erfc~,
s 2.Jt Y/m=2.637x 10-4k,,/p.,CPlllcll/' sq ft/hr.

774 SOCIETY OF PETROLEUM ENGINEERS JOURNAL


12

12
10

Sm/S, " 100

rf! 6
<I
N ~ 6
<I
N

10' 10' 10' Ill' 10' 10' 10' 10' 10' 10' 10' 10'
4tD "4~t/r' CtD "C~t/r'

Fig. 3-Dimensionless fracture drawdown, 2t.po = 47rTt.p/q Fig. 4-Comparison of pressure patterns from the Warren
(2t.p 0 = t.p Tl70.6q), in a naturally fractured reservoir, and Root and the gradient flow models,
Sm/SI = 10 (Eq. 24). (Sm/SI = 100).

Eq. 28 can also be obtained from Fig. 3 by picking the sign, counteracts the second term (see Table 2). Thus,
times at which the transitional curve merges with the for later times, only the first term describes the pressure
final Ei-segment. According to Eq. 28, the larger the behavior that gives a slope, m, representing the forma-
matrix block dimension or the tighter the matrix, the tion permeability. This pressure curve will correspond to
longer the transitional, one-half slope region. the final pressure segment.
The one-half slope relationship between the in- The one-half relationship between the slopes of the
termediate transient r D curve and the final pressure seg- transient-pressure curve and the limiting pressure
ment, the Ei-curve indicative of formation permeability, segments provides an additional criterion for a single-
may also be independently shown from a simplified solu- well test interpretation. This may be particularly useful if
tion 16 developed for the time periods when fracture the early-time pressure segment is, as is often the case,
compressibility is no longer affecting pressure distribu- distorted by afterflow while the late-time pressure seg-
tion in a fractured formation-i.e., cf>fcf=O. This solu- ment is not yet developed owing to a limited flow period.
tion, written in the notation used here, is Then, the formation permeability, k, may be found from
the transitional curve slope, m', as (in field units):

162.6 qBJ-i
k= , .......................... (30)
2m'h l
411't H
LlpD=ln +In + from a plot of the pressure data in semilog coordinates.
1.78lr\~ (1.78111111t) ,/, As has been mentioned, the basic difference between
00
1 nH the Warren and Root and the gradient-flow models lies in
2 ~ -erfc , .................. (29) the shape of the transitional pressure segment and in the
11=1.3,5 n (11l11t)'/' time at which this transient segment begins to develop. It
has been shown that the gradient-flow model has a tran-
where sitional pressure curve of a linear character in semilog
coordinates. Furthermore, the slope of this transitional
pressure-drawdown portion is equal to one-half the
limiting pressure segment's slope. The time at which the
For small times, the third term of Eq. 29 approaches transient pressure deviates from the initial - Ei( - r 2 /411t)
zero because the arguments of the erfc functions are segment, according to the Warren and Root model,
large. During this time, therefore, the pressure response depends only on the fracture/matrix permeability ratio,
is described by the first two terms of Eq. 29 only. As one k/k ln , for a given well dimension. However, from the
can see, the first term alone would give in a semilog plot gradient-flow model it follows that such time is a strong
a straight line with a slope, m, inversely proportional to function of the matrix/fracture storage ratio, SIII/St, as
fractured-formation permeability, k. Addition of the sec- well. Fig. 4 shows the basic difference in drawdown pat-
ond term with a negative half-slope contribution would terns for these two models. The greater the storage
result, however, in a straight line having a slope one-half capacity of matrix relative to fracture, the earlier the time
that representing the formation permeability-i.e., at which the transient-pressure segment deviates from
m' =O.Sm. This slope will represent the transitional the initial - Ei( - r 2 /411t) curve. Since the earlier time
pressure response. As time increases, the arguments of mostly affects interference-test analysis, the details of its
the erfc functions became smaller with the results that behavior are discussed in a separate section. Before that,
the erfc functions themselves are nonzero. Furthermore, however, the field buildup data from single-well tests are
their sum-i.e., the third term in Eq. 29-approaches the examined in light of behavior predicted by the gradient
magnitude of the second term and, having the opposite model.
OCTOBER 1983 775
Application of the Gradient Model
TESTA
3720 to Field Data
Pi = 3715 psi (measured)
q~l_RB/D Actual field data are now used to demonstrate that not
3710 h/~ ~ 'D.7 ft/cp
only does the linear transitional response exist, it does
K ~ 162.6 x 1486 = 33 md
26 x 'D.7 indeed have a slope that is one-half that of the limiting
pressure segments. These data have been published
previously. 10
Drillstem test buildup data from Tests A, B, and C
from Ref. 10 are shown in Figs. 5,6, and 7. Three linear
segments of pressure curve are well identified for each
test, with the third, pseudosteady-state flow, segment
extrapolating to the measured initial formation pressure,
Fig. 5-Pressure buildup for Test A. Pi' The effective permeability, k, determined by a con-
ventional Homer analysis (Eg. 30) applied to <either of
the segment slopes, is shown on the figures. The
horizontal displacement of the earJy- and late-time
pressure segments corresponds to ratios of the Homer
37110
TESTB abscissa of 2.7, 3.8, and 2.1 for Tests A, B, and C,
3&60 f-
PI ~ 3707 pai (measured) respectively. The data, thus, suggest a fractured medium
q=61 RB/D
f- h/~ -21.3 ft/cp
with comparable matrix and fracture storages, the former
~ 36OOf- being two to three times the latter. This implies a high-
..; f- density fracture system within the radius of investigation
! 3Ii&O f-
of each well. This conclusion is supported by observed
ffi f-
f3600 core fracture densities of one to three fractures per foot.
f- Matrix permeability measured on cores was generally in
3460f-
the range 10 to 50 Jld.
0-
34000-
Matrix block dimensions can be calculated from the
buildup data if the intersection point of the transitional
1000 100 10
segment with the final pressure segment is present. In
Figs. 5,6, and 7 the Homer time ratios at the intersec-
Fig. 6-Pressure buildup for Test B.
tion points are 3.3, 5, and 4.8, which give shut-in times
at intersection, iltn of 2.3, 2.0, and 2.5 hours, respec-
tively. Formula 28 written in oilfield units for the in-
tersection time, ilt x , with matrix dimension L=2H
TESTe becomes, in hours,
Pi =3515 psi {measured) a"
...
~ ","'-
P'~3615psi
q~Gi RB/D ~/
h/~ ~21.3 ft/cp

K ~ 1~.~ ~::s ~ 332 md


from which the block size is calculated in feet as

(th~6t)~4.8
L- illr. km ) '/2
3600
(
2,3704>m c mJl

Matrix diffusivity (kl4>cJl) /1/' which varies from 1,390


md'psi/cf (k m =0.01 md, 4>/1/ =0.12, Jl=6 cp,
c m =1O- pSi-I) to 6,950 md·psi/cp(km=0.05 md),
Fig. 7-Pressure buildup for Test C. along with an average intersection time iltr =2.2 hours
gives block dimensions that vary from 1.13 to 2.5 ft.
This agrees with the core data measurements.
Data from Test D of Ref. 10, replotted here in Fig. 8,
are a good illustration of the short flow time effect on
TEST 0
3625 p·~3625 psi pressure-buildup pattern. Whenever the flow time prior
Pi ~ 3625
pai (measured)
q~l290 RB/D
to the well shut-in is insufficient, the ensuing buildup
~3620 h/~ ~ 37.5 ft/cp will be of a shape not customarily recognized as typical
..; K 162.6 x 1290 1.14 d o
of a fractured formation. 8,17 If flowed for a short time, a
rr: 4.9 x 37.5
;:)

ffi 3615 fractured formation exhibits a response that on a Homer


o
rr:
Q.
o plot can be either an apparent straight line extrapolating
o
3610
o to the initial pressure or a two-segment curve that con-
o caves slightly downward (see Fig. II of Ref. 17). The
characteristic three-segment pressure pattern analogous
to that of the drawdown will be exhibited by buildup data
only if the flow time prior to shut-in is long enough to
Fig. 8-Pressure buildup for Test D. develop the third drawdown segment over one log cycle.
776 SOCIETY OF PETROLEUM ENGINEERS JOURNAL
The slightly concaved downward segment can easily be
interpreted as a straight line subject to Homer analysis.
The slope of this pseudotransitional curve is less than
that of the fracture response early-time segment because
fracture replenishment modifying the slope has started,
but greater than the half slope of the characteristic transi-
tional segment, which had not been developed on the
drawdown. For this situation, a conventional Homer
analysis applied to the apparent straight line that ex- 3220

trapolates to the initial pressure will overestimate the ef-


fective formation permeability by a factor of 1.5 to 2 (see
Ref. 17). Consequently, the permeability of 1,140 md
determined by the analysis shown in Fig. 8, which uses
the extrapolated-to-initial-pressure segment, is over- Fig. 9-Long flow pressure buildup.
estimated and should fall in a range 570 to 760 md.
Fig. 9 shows another example of buildup data from
tests in the same field that gave all the previous examples
(Figs. 5 through 8). These buildup data are recorded 0'.----,------,-----,---,----==-- ......_-:::::;=
after a longer production period, til = 100 hours. Shut-in
time at intersection point, !::.t r =33.3 hours, along with
the core matrix diffusivity used previously, (k/¢CJl,)111 10'
=6,950 md· psi/cp, gives the matrix size dimension of
10 ft. The increased fracture spacing obtained for this
well agrees with the fracture density change observed
~10·1
throughout the field. N

Pressure Pattern for


Interference-Test Analysis 10·'

Dimensionless pressure-drawdown values, 2!::.p D


=47f-T!::.p/q (2!::.p D =!::.pTI70.6q, in customary units),
calculated from Eq. 24 for various rD and 4tD values, lOi'o!::;·,.....JI..L-y..-'---;Ic!;;O'-----.;IO.,-.--'Ic!;;O'-----.;IO'u-----.lIO.
are plotted in logarithmic coordinates in Figs. 10 and II 4to" 4~t/r'

for the ratio of matrix/fracture storages, Sill/Sf, equal to Fig.10-Dimensionless fracture drawdown, 2flpo = 47fTflplq
10 to 100, respectively. The dependence of pressure (2flp 0 = flp TI70.6q), plotted in logarithmic coordi-
responses on the Sm/Sf ratio is obvious from these nates for Sm/S, = 10 (Eq. 24).
figures. The r D-type curves are positioned between two
limiting Ei-curves, the left being the Ei-curve associated
with the fracture parameters, - Ei( - r2/411t), and the
right being the Ei-curve that is based on cumulative 10',-----,-----.----r---,.-----==-~--,

storage equal to the sum of the matrix block and fracture


storages, - Ei( - r2/411 * t). As such, the r D curves are
"compressed" when the ratio of storage decreases, or 10'
"expanded" when the ratio of storage increases. The
time each r D curve deviates from the initial Ei-curve,
and, consequently, its shape, are thus both dependent on
the storage capacity ratio, Sill/St. In order then to match
interference-test pressure data to the appropriate r D
curve uniquely, one should have an independent estimate 10·'
of the matrix/fracture storage ratio, Sill/Sf. Such an
estimate can be made from interference data if the initial-
and the late-time pressure segments (two limiting Ei-type
curves) are present. If the test duration does not allow 10i'0~·,~L....<~.--...L....--:;-!;;--'---~10:;-,----;I*'O'--~IO';;---~IO.
410" 4~l/r'
development of the late-time pressure curve, the estimate
of S,..JSf can be made from results of single-well tests. Fig. 11-Dimensionless fracture drawdown, 2flpo = 47fTflplq
If, however, the test time is relatively short, which is (2flpo =flpT/70.6q), plotted in logarithmic coordi-
often the case, an approximation to the solution (Eq. 24) nates for Sm/S, = 100 (Eq. 24).
can be applied. For small values of time (or large r
values), the pressure distribution may be found as 18

q 100 e -x .J(3114t D
!::.p=-- --erfc ,
47fT1/4fDx .Jx(X-1I4tD)

where
OCTOBER 1983 777
lO'r------,----.----,------.------. r (kcpc)/II
iJ=0.25----...:.....
H (kcpc)t

10"f------1f--------.~~~
=0.25rD..JCP/llc/II/CPfCt. . ................. (32)

The storage ratio. thus. appears explicitly in Eq. 31


through the parameter iJ.
Dimensionless pressure-drawdown values. 2!:1p D =
47rT!:1p/q (2!:1p D =!:1pTI70.6q. in customary units) for
various iJ and 4f D values are shown in Fig. 12. The
parameter (3. which according to Eq. 32 depends on both
the fracture/matrix permeability ratio and the
matrix/fracture storage ratio. is a characteristic of the
matrix block's influence on the overall pressure
response. The greater the matrix block's contributions
Fig. 12-Dimensionless fracture drawdown, 2flp 0 = 47r Tflp/q
(the greater the (3 value). the smaller is the pressure
(2flp 0 = flp Tl70.6q), plotted for various (3 and 4t 0
values from early-time approximation (Eq. 31). drawdown of the fracture. This is shown by the greater
deviation of the (3 curve from the limiting Ei-curve.
- Ei( - r2 /41/t). which is associated with just the fracture
flow parameters. Therefore. if one uses a conventional
10',----,.-----.-----,,----.-----.-----,
analysis based on the Ei-curve. which does not take into
account the pressure support offered by matrix blocks on
drawdown measurements. then the calculated fomlation
permeability will be overestimated.
10'
The shape of the (3-type curves is not as sensitive to the
ratio of storages. especially at early time. as is the shape
of the r D curves. Therefore. the application of the (3
relO"'
N
<I curves to matching field data can be made. An estimate
of the ratio of storages. however. is still required if one
wants to determine the ratio of permeabilities. klk/ll'
lit' from a given (3 value and an assumed matrix block
size. H.

10·;0!;c.·,-L.-"--!:!-'L----cI~0',.-----:-;10~'---,1:-:,0';-----:10'~-~101
Matrix Block Pressure Behavior
4tD "4~t/" If a well is completed in a matrix block so that its per-
forated part is not intersected by a fracture. the buildup
Fig. 13-Dimensionless malrix block drawdown, 2flpo =
47rTflp/q (2flpo =flpT/70.6q), calculated for
pattem of such a well will be different from that dis-
Sm/Sf = 10 (Eq. 25). cussed previously.
The matrix block average pressure distribution. !:1 jJ Ill'
in Laplace space is given by Eq. 25. Dimensionless
pressure drawdown values. 2!:1p D =47rT!:1p/q (2!:1p [)
=!:1pTI70.6q. in customary units). calculated from Eq.
10'r---,.----,----,---..,.----,-------,
25 with the Stehfest 13 numerical Laplace transform in-
version method. are shown plotted to Figs. 13 and 14 for
ratios of storages. S/II/St. equal to 10 and 100. respec-
10' tively. As one can see. the type curves. plotted in
logarithmic coordinates. have a characteristic half slope.
The time when this half-slope is reached is a weak func-
,e 10. 1
tion of the ratio of matrix and fracture storages. The
<I
N
greater the storage capacity of matrix relative to fracture.
the earlier is the time at which one observes a half-slope
10" in buildup data. When. however. the matrix storage is
comparable to that of the fracture. the initial pressure
changes are rapid. and the shapes of buildup curves ap-
proach that of the Ei-curve for fracture flow.
10' 10' 10' 10'
4tD "4~t/,'
The matrix-type response for the pressure averaged
throughout the matrix block (Figs. 13 and 14) has an en-
Fig. 14-Dimensionless matrix block drawdown, 2flpo tirely different buildup pattem compared with that of the
= 47r Tflp/q (2flp 0 = flp T/70.6q), calculated for
Sm/Sf = 100 (Eq. 25). fracture (Figs. 5 and 6) yet such diversity is
characteristic of a naturally fractured reservoir. The half-
slope pattem cannot be matched at all to the conventional
Ei-curve. The physical reason for such a half-slope lies
778 SOCIETY OF PETROLEUM ENGINEERS JOURNAL
10' 10'

-' 0.318
10' 1. 10'

~ 10-1 re 10-'
N ~

10" 10"

10' 10'

Fig. 15-Dimensionless fracture drawdown affected by a gas Fig. 16-Dimensionless matrix drawdown affected by a gas
cap. cap.

in the linear (one-dimensional) diffusion mechanism that Note that for r D ~ 1, the pressure changes are very
governs the matrix-to-fracture flow. If, however, the small. Therefore, for multiple-well tests conducted
pressure is measured at a point within the matrix block, under a gas-cap support, one would require high-
the behavior will be different from that shown on Figs. precision pressure gauges to detect and to measure a
13 and 14 and will, in fact, be similar in shape to the Ei- recognizable interference effect.
curve but displaced horizontally as r D decreases and as
the ratio ¢mcm/¢Icf increases. Conclusions
The gradient-flow model for naturally fractured forma-
Effect of Gas Cap on Pressure Behavior tions, not being limited by the restrictive assumptions of
in a Naturally Fractured Reservoir the Warren and Root model, displays pressure-
drawdown patterns of somewhat different shapes and of
The fracture pressure distribution, t:.p, in a naturally greater diversity of shapes than does the Warren and
fractured reservoir with a gas cap at distance H above the Root model. The differences appear in both the shape of
observation point, in the Laplace space is the transitional pressure-drawdown segment that reflects
matrix block influence or pressure behavior and in the
time at which this transitional pressure segment begins to
develop.
The gradient-flow model shows that the transitional
pressure segment in a plot of drawdown vs. log time is,
cotanh(.J s¢ me 1I.1¢/cII r D)] '/2 J. . ............ (33) in fact, a straight line. Furthermore, the slope of this
straight line is equal to one-half the limiting, Ei-type,
segment slope. This half-slope relationship allows one to
The matrix block pressure distribution, t:. Pm' for the determine reservoir parameters from single-well test data
same case is without having the early- and late-time pressure
segments required by the Warren and Root model. This
t:.p m = t:.p tanh (H.J s/TJ m )/(H.J s/TJ m). . ...... (34) result is particularly significant for exploration well
testing.
Although the inverted forms of these two equations are The time at which the transitional pressure deviates
available (Ref. 19, Page 529, Eqs. 21 and 33, respec- from the initial - Ei( - r2 /4TJt) segment, according to the
tively), Forms 33 and 34 are most convenient for gradient-flow model, depends not only on the frac-
calculating the pressure, t:.p(t), with the Stehfest 13 ture/matrix permeability ratio, k/k"., as in the Warren
numerical Laplace transform inversion method. and Root model, but also on the ratio of matrix to frac-
Dimensionless pressure values, 2t:.p = 41f"Tt:.p/q, ture storages, Sin / SI' The greater the storage capacity of
calculated from Eqs. 33 and 34 are plotted in Figs. 15 matrix relative to fracture, the earlier is the time of the
and 16, respectively. The performance of a gas-cap- transitional curve deviation. This time of deviation and
dominated well is affected in such a manner that a the initial shape of the drawdown curve basically affect
transient-buildup (drawdown) plot shows a continual interference-test interpretation. If the test is long enough
flattening of the pressure-time curve as a result of for the data to match an entire r D curve so that the ratio
pressure support from the gas cap. As time goes on, of storages, Sin / SI' can be established from the displace-
pressure curves from both the fracture and the matrix ment of the limiting Ei-curves, the reservoir parameters
block responses approach a constant pressure value can be determined uniquely from such a match. If the
described by test time is too limited to determine the ratio S 11/ / SI from
these interference data, then this ratio should be
estimated independently to match the data to the ap-
q
t:.p=t:.Pm=-Ko(rD)' .................. (35) propriate set of r D type curves. An estimate of the S m / Sf
21f"T parameter can be made from results of single-well tests.
OCTOBER 1983 779
For the cases when the test time is relatively short, the (3- Stehfest numerical technique and the program for the
type curves are presented to alleviate the nonunique Laplace transform inversion method to me while discuss-
problem. They can provide an independent analysis of ing particulars of a naturally fractured reservoir
fracture flow in a naturally fractured reservoir. behavior. I thank the Western Production Div. of Exxon
Co. U.S.A. for providing the field data used in this
Nomenclature paper.
Cf = total compressibility of fracture, psi -I (Pa - 1 )
Cm = total compressibility of matrix, psi - 1 (Pa - I) References
h = half fracture thickness, ft (m) I. Barenblatt, G.E. and Zheltov, Y.P.: "Basic Flow Equations of
hI = formation thickness, ft (m) Homogeneous Fluids in Fissured Rocks," Dokl. Akad. Nauk
SSSR, (March 1960) 132, 545-48.
H = half-thickness of matrix block, ft (m) 2. Barenblatt, G.E., Zheltov, Y.P., and Kochina, I.N : "Basic Con-
k effective permeability of fractured reservoir, cepts in the Theory of Seepage of Homogeneous Liquids in
md Fissured Rocks," 1. Appl. Math. Mech. (May 1960) 24,
1286-1303.
kf = khI2nh=kHlh, fracture permeability, md 3. Warren, J.E. and Root, P.J.: "The Behavior of Naturally Frac-
km matrix permeability, md tured Reservoirs," Soc. Pet. Eng. 1. (Sept. 1963) 245-55; Trans.,
Ko = Bessel function of the second kind of zero AIME,228.
4. Boulton, N.S. and Streltsova, T.D.: "Unsteady Flow to a Pumped
order Well in a Fissured Water-Bearing F0n11ation," J. Hwlrolog.l'
n = h1/2H, number of matrix blocks over for- (1977) 35, 257-69,
mation thickness 5. Kazemi, H.: "Pressure Transient Analysis of Naturally Fractured
Reservoirs With Uniform Fracture Distributions," Soc. Pel. Eng.
t:.p = Pws -P"f' pressure drawdown in fracture, J. (Dec. 1969) 451-62; Trans., AIME, 246.
psi (kPa); P ws is static well pressure and 6. deSwaan, O.A.: "Analytical Solutions for Determining Naturally
P Hf is flowing well pressure Fractured Reservoir Properties by Well Testing," Soc. Pel. Eng.
J. (June 1976) 117-22.
t:.p m pressure drawdown in matrix block, psi 7. Najurieta, H.L.: "A Theory for Pressure Transient Analysis in
(kPa) Naturally Fractured Reservoirs," 1. Pet. Tech. (July 1980)
t:. p Laplace transform of t:.p 1241-50.
8. Serra, K.V., Reynolds, A.C., and Raghavan, R ..: "New Pressure
2t:.p D = 47rTt:.plq=t:.pTI70.6q, dimensionless Transient Analysis Methods for Naturally Fractured Reservoirs,"
pressure change J. Pel. Tech. (Oct. 1983) 1903-14.
q = production rate, RBID (res m 3 /d) 9. Strobel, c.J., Gulati, M.S., and Ramey, H.J. Jr.: "Reservoir
Limit Tests in a Naturally Fractured Reservoir-A Field Case
r = radial distance from producing well, ft (m) Study Using Type Curves," J. Pel. Tech. (Sept. 1976)
rw wellbore radius, ft (m) 1097-1106.
rD rlkmkflH, dimensionless radial distance 10. Crawford, G.E., Hagedorn, A.R., and Pierce, A.E.: "Analysis of
Pressure Buildup Tests in a Naturally Fractured Reservoir," J.
s - Laplace parameter Pet. Tech. (Nov. 1976), 1295-1300.
S = sr21YJ =dimensionless Laplace parameter II. Adams, A.R., Ramey, H.J. Jr., and Burgess, R.J.: "Gas Well
S = 2nh¢fcf, effective fracture storage Testing in a Fractured Carbonate Reservoir," J. Pel. Tech. (Oct.
1968) 1187-94; Trans., AIME, 243.
Sf = ¢fcf, specific fracture storage 12. Warren, J.E. and Root, P.J.: "Discussion of Unsteady-State
Sm ¢ mC In' specific matrix storage Behavior of Naturally Fractured Reservoirs," Soc. Pel. Eng. J.
t = time from start of production, hours (March 1965) 64-65.
13. Stehfest, H.: "Algorithm 358-Numerical Inversion of Laplace
t D = YJtl r2, dimensionless Transforms," Communications of the ACM (Jan. 1970) 13, No. I,
T = kh II p" fractured reservoir effective 47-49.
transmissibility, md-ft/cp (md-m/Pa' s) 14. Gradshtein, I.S. and Ryzhik, I.M.: Table of Integrals, Series and
Products, fourth edition, Academic Press Inc., New York City
vm matrix -to-fracture flow, per unit area per (1965) 1.232(2) 23.
unit time 15. Luikov, A.V.: Analytical Heat DiffUsion Theory, Academic Press
z vertical coordinate, ft (m) Inc., London (1968).
16. Dontsov, K.M. and Boyrchuk, B.T.: "Effect of Characteristics of
Z = zI H, dimensionless coordinate
I
Fractured Media on Pressure Buildup Behavior," /z.!'esria VUZ,
0.00633 kip,¢jCf=0.00633 TIS, fracture Oil and Gas (1971) Nt, 42-46 (in Russian).
hydraulic diffusivity, sq ft/D 17. Streltsova, T.D. and McKinley, R.M.: "Effect of Flow Time
Duration on Buildup Pattern for Reservoirs with Heterogeneous
(=kip,¢fcf=TlS, mm 2 /s) Properties, " paper SPE 11140 presented at the 1982 SPE Annual
YJm 0.00633 kmlp,¢mclIl' matrix hydraulic dif- Technical Conference and Exhibition, New Orleans, Sept. 26-29.
fusivity, sq ft/D (=kmlp,¢mcm, mm 2 /s) 18. Hantush, M.S.: "Modification of the Theory of Leaky Aquifers,"
J. Geophl's. Res., (1960) 65, No. II, 3713-25.
0.00633 kiP,¢mcm( =kiP,¢mcm)
YJ * 0.00633 [kiP,(¢mcm +¢jCf)] SI Metric Conversion Factors
[=ki p,(¢ mCm +¢fcf)] bbl x 1.589873 E-Ol m3
viscosity, cp (Pa' s) cp x 1.0* E-03 Pa's
fracture porosity, fraction ft x 3.048* E-Ol m
matrix porosity, fraction psi x 6.894757 E+OO kPa

Acknowledgments * Conversion factor is exact. SPEJ


I thank R.M. McKinley for valuable comments in the Original manuscnpt receIved in Society of Petroleum Engineers office Jan. 29, 1982.
course of preparing this work. I also thank H.1. Ramey Paper accepted for publication Feb. 11, 1983. Revised manuscript received May 25,
1983. Paper (SPE 10782) first presented at the 1982 California Regionat Meeting held
Jr. and M.1. Mavor of Stanford U., who presented the in San Francisco, March 24-26.

780 SOCIETY OF PETROLEUM ENGINEERS JOURNAL

You might also like