Documente Academic
Documente Profesional
Documente Cultură
a r t i c l e i n f o a b s t r a c t
Article history: This paper presents a Green's function formulation for flow to a well in a composite system with a planar
Received 30 December 2008 interface. It is difficult to obtain solutions to the diffusion equation under such circumstances. Usually,
Accepted 20 November 2009 for this problem, solutions are obtained in terms of the Laplace transformation followed by a Fourier
transformation. But difficulties arise in reducing the solutions to a state that enables efficient numerical
Keywords: calculations. In this work, flow to a well represented by a continuous line-source near a partial hydrologic
composite systems
barrier (a composite system) is solved along the lines proposed by Sommerfeld (1909). His approach to
plane interface
Green's function
working the problem results in a scheme that is highly efficient from a computational perspective, an
sources and sinks essential requirement for processing the inverse problem. Results obtained by the new solution are
pressure behavior compared with those of Bixel et al. (1963).
© 2009 Elsevier B.V. All rights reserved.
0920-4105/$ – see front matter © 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.petrol.2009.11.015
230 R. Raghavan / Journal of Petroleum Science and Engineering 70 (2010) 229–234
1 pffiffiffiffiffiffiffiffiffiffiffiffi
γ1 ðx; yÞ = K
∞ −yq
s = η1 r1 + ∫0 f1 ðαÞe 1 cosðαxÞdα; ð2:10Þ
2πη1 0
and
∞ yq
γ2 ðx; yÞ = ∫0 f2 ðαÞe 2 cosðα xÞdα: ð2:11Þ
Fig. 1. Schematic of the system.
and (continuity in flux for all time, t), we may write Eq. (2.10) as
" #
∂p1 ðx; y; tÞ ∂p ðx; y; tÞ 1 e−q1 j y−y0 j
T1 = T2 2 ; y = 0: ð2:7Þ ∞
γ1 ðx; yÞ = ∫0
−q y
+ f1 ðαÞe 1 cosðαxÞdα: ð2:16Þ
∂y ∂y 2πη1 q1
where dS′ denotes the element of a line or surface through which fluid where
is withdrawn and γ̅(x, y, z) is the Laplace transformation of γ(x, y, z; t),
the fundamental solution. In our case, the source-strength is uniform 2 2
r2 = x + ðy + y0 Þ ;
2
ð2:19Þ
over time and space, thus Eq. (2.8) becomes
and
1 q̃
Δpðx; y; zÞ = ∫ γðx−x ′ ; y−y ′ ; z−z ′ ÞdS ′ : ð2:9Þ
s ðϕcÞ S̃ T2
Tr = : ð2:20Þ
T1
We shall now construct γ by the Sommerfeld (1909) procedure.
The fundamental solution is also discussed in Shendeleva (2004). If we consider the integrals in Eqs. (2.17) and (2.18), we see that
Mandelis et al. (2001) discuss the Sommerfeld (1909) procedure for the response at any point M is a function of ηr and Tr, where
point-sources. We first note that fluid is extracted by a well bore that
is assumed to be a line-source (2D flow) and is produced at a constant η1
ηr = : ð2:21Þ
rate, qw. As explained in Bellman et al. (1949), the Sommerfeld η2
R. Raghavan / Journal of Petroleum Science and Engineering 70 (2010) 229–234 231
The dominant role played by ηr may be understood if we note that expect this result to hold in situations where η1 ≠ η2, but no proof is
the integrals may be simplified considerably if ηr = 1 (see Section 2.2). available in the literature. We prove this assertion by inverting the
Note that if y = 0 was an impermeable or a constant-pressure integral in Eq. (2.17) or in Eq. (2.18) along the lines suggested by
boundary, then r2 would represent the location of the image well. Bellman et al. (1949), and articulated, more recently, in Shendeleva
There is no image source for region 2. The observation regarding (2004). Our procedure involves computing the logarithmic derivative,
image wells provides an opportunity to highlight an interesting tdp/dt. For example, multiplying both sides of Eq. (2.17) by the Laplace
physical explanation of the solution given in Eq. (2.17). The first term transform variable, s, applying the Inversion Theorem for the Laplace
is the Laplace transformation of the Theis (1935) solution. As the transformation and then multiplying the result by time, t, we have
second term represents the contribution of an image well at the point
2 3
r12 r2
j j
(0, y0) of opposite strength, this term represents the contribution of dp1 qw B 4 −4η t − 2 cosðαxÞ
−1 ∞ −q ðy Þ
dα 5;
+ y
1 −e 4η1 t + 4tL ∫0 e
the ‘barrier’ at (y = 0) had it been maintained at a constant pressure t = e 1 0
dt 4πT1 q1 +Tr q2
pi. The third term represents a correction to the second term and
reflects the properties of region 2 at any point M in region 1. ð2:28Þ
2.2. Some limiting forms where the symbol L− 1 represents the inverse transformation. To
obtain a conclusion similar to that derived from Eq. (2.25), we need to
We now show that the standard results in the literature may be obtain the long-time approximation of tdp/dt. That is, from Eq. (2.28)
extracted from Eq. (2.17) if we assume η1 = η2. If we assume that we have
Tr = 1, the second and third terms of Eq. (2.17) cancel, and on applying
the Inversion Theorem for the Laplace transformation, we obtain the
Theis (1935) solution (dropping the subscript 1 where appropriate)
j dpdt j
1
t→∞
=
qw B −1 ∞ −q1 ðy + y0 Þ cosðαxÞ
πT1
L ∫0 e
q1 +Tr q2
dαjs→0 : ð2:29Þ
!
2 We may show that (see Eq. (A.5))
qw Bμ r
pi −pðx; y; tÞ = − Ei − : ð2:22Þ
4πkh 4ηt
−1 ∞ −ðy + y0 Þq1 cosðαxÞ
L ∫0 e dαjs→0 ð2:30Þ
q1 +Tr q2
If, however, we were to assume that Tr = 0 (sealing boundary), pffiffiffiffiffi
then the second and third terms of Eq. (2.17) may be combined with T ηr 1 1 1
= r ∫0 3 1
du;
the help of Eq. (2.15), and 4t ½u+ Tr2 ηr ð1−uÞ2 ½u+ ηr ð1−uÞ2
" ! !#
2 2
qw Bμ r r where ηr = η1/η2. Performing the integration, we have
pi −pðx; y; tÞ = − Ei − 1 + Ei − 2 : ð2:23Þ
4πkh 4ηt 4ηt
−1 ∞ −ðy + y0 Þq1 cosðαxÞ 1 2
L ∫0 e dαj s→0 = ; ð2:31Þ
q1 + Tr q2 4t 1+ Tr
On the other hand, if Tr → ∞, then (pressure is constant at the
interface; p(0, 0; t) = pi)
that is,
" ! !#
r 21 r 22
jt dpdt j
qw Bμ qw B 2
pi −pðx; y; tÞ = − Ei − −Ei − : ð2:24Þ 1
= : ð2:32Þ
4πkh 4ηt 4ηt t→∞ 4πT1 1+ Tr
For the general case, we may write Thus, we may conclude that the slope of a plot of p1(x, y; t) vs. t on
" ! !# semilogarithmic coordinates would result in a straight line with slope,
2 2
qw Bμ r 1−Tr r m, proportional to 2/(T1 + T2) for all combinations of η1 and η2. This
pi −p1 ðx; y; tÞ = − Ei − 1 + Ei − 2 ð2:25Þ
4πk1 h 4ηt 1+ Tr 4ηt result may be derived in a similar manner for region 2.
Although not of direct interest to the present study, it should be
for region 1, and the analog of Eq. (2.25) for the region y b 0 is noted that the expressions for G1 and G2 given in the Appendix A may
! be used to express the pressure distributions in the time-domain. The
qw Bμ 2 r2 convolution integral that results would have to be evaluated
pi −p2 ðx; y; tÞ = − Ei − 1 : ð2:26Þ
4πk1 h 1+Tr 4ηt numerically, and the issues involved in taking that approach are
discussed in Bixel et al. (1963). The time-domain solutions do not
Eqs. (2.25) and (2.26) are identical to the results given in Bixel serve our purpose of having a scheme that is numerically efficient,
et al. (1963) and Nind (1965). Eqs. (2.23)–(2.25) may be viewed as a particularly if we are to solve complex problems such as horizontal
special case of the well known proposition that the effect of a linear wells and fractured wells. In fact, none of the problems noted in
barrier on transients created by a source, S, in porous media may be Section 4 are readily tractable with a time-domain solution.
handled with the help of an image source, S′, that is symmetrical to
the actual source about the interface. In general, we may write 3. Results
pi −pðx; y; tÞ = Π⊗S + Π⊗S ′ ; ð2:27Þ The principal goal of this section, simply, is to demonstrate that
correct results may be obtained from the new formulation given here
where Π is the solution for a source placed in an infinite porous by comparing results with the Bixel et al. (1963) model.
medium having the properties of the region where S is located, and ⊗ All results given below were obtained by numerically inverting the
is the convolution product that applies both to space and time. Such integrals in the solutions for Δpj ðx; yÞ with the aid of the Stehfest
examples for a large number of situations with complex well (1970a,b) algorithm. The code was verified in the usual fashion by
configurations may be found in Raghavan and Ozkan (1994). comparing results with the limiting forms given above. In addition,
Eqs. (2.25) and (2.26) lead to the result that the slope of a plot of derivative responses were compared with the results obtained by
p(x, y; t) vs. t on semilogarithmic coordinates would result in a computing the expressions in Shendeleva (2004) after appropriate
straight line with slope, m, proportional to 2/(T1 + T2). Intuitively, we modifications at several locations (x, y). We now present a few
232 R. Raghavan / Journal of Petroleum Science and Engineering 70 (2010) 229–234
k1 t
tD = ; ð3:1Þ
ϕ1 c1 μ ℓ2
4πk1 h
pD ðxD ; yD ; tD Þ = ½p −pðx; y; tÞ: ð3:2Þ
qw Bμ i Fig. 3. Derivatives of the pressure with respect to the natural logarithm of time at the
well are given in adimensional form. The solutions are the counterpart of the results in
Fig. 2. At early times the derivative values are equal to 1 and if times are long enough
The unbroken lines in Fig. 2 are the solutions obtained by inverting asymptotically approach 2/(1 + Tr).
the right-hand side of Eq. (2.17) at the producing well. Inversion of
the first two terms, as is well known, leads to exponential integrals.
The third term was inverted numerically by the Stehfest (1970a) and if t is large enough, then from Eq. (2.32) we have
algorithm. All solutions are identical at early times as they follow the
response predicted by the Theis (1935) solution: 2
j tD dp1D =dtD j tD →∞ = : ð3:5Þ
! 1+Tr
qw Bμ r2
pi −p1 ðx; y; tÞ = − Ei − 1 : ð3:3Þ
4πk1 h 4η1 t As shown in Fig. 3, the solutions obtained from Eq. (2.17) are in
accord with the predictions given in Eqs. (3.4) and (3.5).
The slope of the line in Fig. 2 during this period is 2.303. At later
times, the solutions diverge as the influence of the interface becomes 4. Some additional results
dominant, and ultimately the slopes of the lines follow the trend
predicted in Eq. (2.32). The circles shown in Fig. 2 are the solutions The results given in §3 show conclusively that we are able to
extracted from Bixel et al. (1963) by reading their charts. Agreement reproduce the Bixel et al. (1963) solutions. The charm of the Bellman
between the two solutions is quite good. The comparison in Fig. 2 et al. (1949) study is its ability to obtain solutions by the Laplace
accomplishes the principal goal of this section. transformation in terms of very simple functions. The solutions given
A better physical understanding of the characteristics of the in Eqs. (2.17) and (2.18) enable us to obtain a suite of solutions that
solutions in Fig. 2 may be obtained if we consider the corresponding are essential for practical implementation. For completeness we
logarithmic derivative, tDdpD/dtD; see Fig. 3. From Eq. (3.3), at early merely highlight some expansions of the results that may be readily
times we obtain derived. Specifically, we consider methods to handle variable-rate
problems as well as reservoirs containing natural fractures.
jtD dp1D =dtD j tD →0 = 1; ð3:4Þ Variable-rate problems may be conveniently handled by Duha-
mel's formula. If pD(xD, yD; tD) is the dimensionless pressure at a point
(xD, yD) at time tD for the variable-rate problem, and if we denote the
solution given in either Eq. (2.17) or Eq. (2.18) by pDu, then Duhamel's
formula yields
pD ðxD ; yD Þ = s
qsf
pDu ðxD ; yD Þ; ð4:1Þ
where qsf(t) is the variable rate at the well (sandface rate) expressed
as a fraction of the reference rate, and q ̅sf is Laplace transformation
of qsf((tD). The symbol, s, is now the Laplace variable with respect to
tD given in Eq. (3.1).
The adimensional distances, xD, and, yD, are defined, respectively,
by
x y
xD = ; yD = : ð4:2Þ
ℓ ℓ
Here, the well bore storage constant, CD, is defined by noted. A gratifying feature of Sommerfeld's (1909) procedure is
that pressure distributions for systems with sealing or constant-
C pressure boundaries may be solved without resorting to the method
CD = : ð4:4Þ
4πϕhcℓ2 of images.
Combining Eq. (4.3) after taking the Laplace transformation with Nomenclature
Eq. (4.1), we obtain the well known result B formation volume factor [L3/L3]
C unit storage factor [L4T2/M]
s
pwDu +S
pwD = ; ð4:5Þ c total compressibility [L T2]
s + CD s2 ðs
pwDu + SÞ −Ei(−x) Exponential Integral
h thickness [L]
where S is the skin factor. Well storage at an observation well
k permeability [L*L]
(Tongpenyai and Raghavan, 1981) may also be handled in a similar
ℓ reference length [L]
way.
L distance between the interface and the well [L]
The producing rate for the constant-terminal-pressure solution
p pressure [M/L/T2]
may also be obtained by Duhamel's formula. In this case, Eq. (4.1) may
p′ logarithmic derivative [M/L/T2]
be expressed as follows (see van Everdingen and Hurst (1949))
qw rate; constant [L*L*L/T]
1 qsf(t) variable rate; fraction of reference rate
pwDu
qD = 2 ; ð4:6Þ rw radius of well bore [L]
s
r distance from well; radius [L]
where q ̅D is the Laplace transformation of qD(tD), the variable S skin factor
producing rate in adimensional form, given by s Laplace transform variable [1/T]
t time [T]
qðtÞBμ tD adimensional time
qD ðtD Þ = : ð4:7Þ
4πk1 hðpi −pw Þ T transmissivity [*L]
x(x,y,z) coordinates [L4T/M]
The symbol pw in Eq. (4.7) is the pressure at the well which in this η diffusivity [L*L/T]
case is constant. The solution for the pressure at any point in the flow μ viscosity [M/L/T]
domain corresponding to this situation may now be obtained from the λ Warren–Root parameter
following relation ω Warren–Root parameter
ϕ porosity [L3/L3]
qD
pD ðxD ; yD Þ = s pDu ðxD ; yD Þ; ð4:8Þ
where pD(xD, yD; tD) = [pi − p(x, y; t)]/(pi − pw), p D̅ u(xD,yD) is the
result derived in this work (see Eqs. (2.17) and (2.18)), and q D ̅ is Subscripts
given by Eq. (4.6). Solutions for the Warren and Root (1963) model D adimensional
may be derived by a simple change in nomenclature as noted in many w well bore
places; see, for example, Raghavan and Ozkan (1994). If we were to 1 region where well is located
simply replace s in the kernels of the Bessel functions and the 2 region on the other side of barrier
integrals of Eq. (2.17) and Eq. (2.18) by sf(s), where