Professional Documents
Culture Documents
Water Influx PDF
Water Influx PDF
van Everdingen & Hurst model. The model presented in the or Equation (7):
work of van Everdingen & Hurst (1949) deals with two types
of aquifers: radial and linear. Applying the Laplace tD
We = U ∫ q D (t D − τ D ) Δp(τ D )dτ D (6)
transformation, van Everdingen & Hurst solved the diffusivity 0
equation of the reservoir-aquifer system considering as
boundary condition a constant pressure in the boundary. The
tD
flow supplied by the aquifer at the point of contact with the We = U ∫ W D′ (t D − τ D ) Δp(τ D )dτ D (7)
reservoir is given by the Darcy equation, Equation (1) 0
2πfkhΔp 0 ⌠ D
t
t dt
We ≡ ∫ qdt = ⎮ qD dt D (3)
0 μ ⌡0 dt D
Calling WD(tD) the integral of qD with regard to tD , Equation
(3) gains a simplified form.
We = UΔp0W D (t D ) (4)
(11). This implies recalculating all steps when the value of tDn
⎛ J pi ⎞
is changed. This happens when we wish to evaluate the q = J ( pi − p ) exp⎜⎜ − t ⎟⎟ (17)
accumulated influx at each new step in the behaviour ⎝ Wei ⎠
prediction. Note that the value of We must be evaluated for the
time and regime of the aquifer at that moment. Thus, the Equation (17) is the equation of the flow of water from the
equation that evaluates We changes with the change in flow aquifer into the reservoir as a function of time and the pressure
regime. This is an important point to be taken into account in drop at boundary, (pi − p). This equation is general and does
the numerical programming. not depend on aquifer geometry. By integrating it, the
following results:
Fetkovich approximate model. The Fetkovich model, 1971, t
⌠ ⎛ J pi ⎞
applies to finite aquifers and admits pseudosteady-state regime We ≡ ∫ qdt = J ( pi − p ) ⎮ exp⎜⎜ −
t
t ⎟⎟ dt (18)
flows. This model has a flaw in its mathematical conception, 0
⌡0 ⎝ Wei ⎠
which is why it is called an approximate model. However, it
has the advantage of being practical in terms of programming, Or
as it does not require the application of the superposition ⎡ ⎛ ⎞⎤
principle as in the van Everdingen & Hurst model, which We =
Wei
( pi − p ) ⎢1 − exp⎜⎜ − J pi t ⎟⎟⎥ (19)
requires all calculations to be done again at each time-step, pi ⎢⎣ ⎝ Wei ⎠⎦⎥
resulting in long processing times. Although processing
Although the Fetkovich model is not strictly correct in its
technology is very advanced nowadays, it is always desirable
representation of the superposition principle, it became quite
to use methods demanding short processing times. In
popular nonetheless for its ease of numerical programming.
developing the model, Fetkovich assumed a pseudosteady-
By discretizing Equation (19), Fetkovich allowed it to be used
state regime with water flow from the aquifer to the reservoir,
for the case of pressure variation at the boundary, resulting in
as in Equation (12).
Equation (20), by which it is possible to calculate the value of
the influx for any given time interval Δtn.
= J ( pa − p )
dWe
q= (12)
⎡ ⎤
( p a n−1 − pn ) ⎢1 − exp⎛⎜⎜ − J pi Δt n ⎞⎟⎟⎥
dt Wei
ΔWe n = (20)
where J is the aquifer productivity index, p a the aquifer mean pi ⎣⎢ ⎝ Wei ⎠⎥⎦
pressure, and p the pressure in the reservoir-aquifer boundary.
Including the material balance equation into Equation (12), the where:
following results: ⎛ 1 n −1 ⎞ ⎛ We n −1 ⎞
⎛ W ⎞
p a n −1 = pi ⎜1 −
⎜ Wei ∑ ΔWe j ⎟⎟ = pi ⎜⎜1 − Wei ⎟⎠
⎟ (21)
⎝ j =1 ⎠ ⎝
p a = pi ⎜⎜1 − e ⎟⎟ (13)
⎝ Wei ⎠ and
where Wei is the maximum influx that a sealed aquifer can
supply, corresponding to the expansion of water in the aquifer p n−1 + p n
pn = (22)
when the pressure drops from pi to zero, defined by Equation 2
(14).
Table 2 shows the equations for the aquifer productivity
Wei = ct Wi pi index (J), for radial and linear aquifers, and permanent and
(14)
pseudosteady-state flow regimes.
After separating the variables, the equation can be Carter&Tracy model. The Carter&Tracy model (1960),
integrated from t = 0 (quando We = 0 e p a = pi ) a t, similarly to Fetkovich, does not require the application of
Equation (15) effect superposition for the calculation of the influx. It is a
p comprehensive model, as it covers any flow geometry, as long
J pi t ⌠ a dp a as the solution for the dimensionless pressure as a function of
−
Wei ∫0 dt =⎮⌡ p p a − p (15)
time is known for the geometry of the given aquifer. This
i
comprehensiveness made it one of the most popular models,
The solution proposed by Fetkovich results in Equation (16). and it is easy to program numerically. In the Carter&Tracy
model, the value of the accumulated influx We is approximated
⎛ J pi ⎞
p a − p = ( pi − p ) exp⎜⎜ − t ⎟⎟ (16) by Equation (23).
⎝ Wei ⎠
At this step of development, the Fetkovich model does not We (t D j ) = We (t D j −1 ) + a j −1 (t D j − t D j −1 ) (23)
retain the precision of mathematical principles, as the pressure
p at the boundary is not constant, but depends on time.
where aj-1 is a constant. This equation assumes that in the
Replacing Equation (16) into Equation (12), the following
interval between tDj-1 e tDj the influx varies linearly with time.
results:
4 SPE 107265
The accumulated water influx can be expressed by the where δ∞ is the constant drainage radius under the
convolution integral, as in Equation (24): pseudosteady-state regime, and A is the area open to water
influx (cross section of the porous medium in the reservoir-
tD j
dW D (t D − τ)
We (t D j ) = U ⌠
⎮ Δp (τ) dτ (24) aquifer boundary) According to Leung in his original article,
⌡0 dτ the pseudosteady-state drainage radius, δ∞, depends on how
the boundary pressure varies with time: step variation “Step
where tD is the dimensionless time defined for each aquifer Interpolation Boundary Pressure”, SIBP, or linear variation
geometry, U is the influx constant, Δp(tD) = pi − p(tD) the “Linear Interpolation Boundary Pressure”, LIBP. For a linear
pressure drop at the boundary, WD (tD) the dimensionless aquifer, the dimensionless drainage radius (δ∞/L) has values of
accumulated water influx, τ a dummy integration variable, and 0,4053 and 0,3333 for SIBP and LIBP variation.
j refers to time discretization. Solving by the Laplace For a radial aquifer, in addition to the pressure condition in
transformation, the expression for the constant aj-1 is found, as the inner boundary (at contact), the drainage radius also depends
in Equation (25). on the aquifer size given by the parameter reD . Table 4 presents
U Δp(t D j ) − We (t D j −1 ) p ′D (t D j ) the drainage radii for values of reD between 1,1 and 240, both for
a j −1 = (25) step pressure variation (SIBP) and for linear pressure variation
p D (t D j ) − t D j −1 p ′D (t D j ) (LIBP). This table is similar to that published by Leung, 1986
with a few differences in accuracy.
Replacing constant aj-1 in Equation (23) results in: Starting from the material balance equation, Leung
U Δp(t D j ) −We (t D j −1 ) p′D (t D j ) established that the water influx flow is given by the derivative of
We (t D j ) = We (t D j −1 ) + (t D j − t D j −1 ) (26) the accumulated influx with respect to time, as shown in
p D (t D j ) − t D j −1 p′D (t D j )
Equation (29)
The expression of Equation (33) is better than that of Equation relative size of the aquifer-reservoir, disregarding this ratio
(32) because it is more convenient to evaluate an integral may result in an error in the productivity index calculation, as
involving pressures than pressure derivatives. The pressure will be seen in the results. Users of commercial flow
data at the boundary as a function of time, needed to calculate simulators that use Fetkovich's simplified formula are warned
the convolution integral, are usually expressed as discrete against the error incurred in the calculation of J. It is suggested
values in time. Thus, in order to calculate the integral, some that, for future revisions of the source code of these
form of data interpolation is necessary. Two simple simulators, the equation presented in this work be used, since
interpolation schemes have been suggested by Leung: it represents the derivation of the pseudosteady-state regime
equation without neglecting the ratio rD2/2reD2.
Table 5 shows the results of the calculation of J using
A – Linear Interpolation of Boundary Pressure, or LIBP.
equation (38) by Fetkovich, and Equation (39) suggested in
In this case, the discrete pressure data are linearly interpolated:
the present work.
⎛ p − pn ⎞ 2 π f kh
p LI (t ) = ⎜ n +1 ⎟ (t − t n ) + p n , t n ≤ t ≤ t n +1 (34) J =
⎝ Δt ⎠ ⎡ ⎛ re ⎞ 3 ⎤ (38)
μ ⎢ ln ⎜⎜ ⎟⎟ − ⎥
⎣ ⎝ o ⎠r 4 ⎦
B – Step Interpolation of Boundary Pressure, or SIBP. In
this case, pressures interpolated between t n and t n +1 are 2π f kh
J=
given by: ⎡⎛ r 2 ⎞ 2 1 ⎛ 3r 2 − 1 ⎞⎤ (39)
μ ⎢⎜⎜ 2 eD ⎟⎟ ln(reD ) − ⎜⎜ 2eD ⎟⎟⎥
⎢⎣⎝ reD − 1 ⎠ 4 ⎝ reD − 1 ⎠⎥
⎦
⎛ p + p n +1 ⎞
p SI (t ) = ⎜ n ⎟ , t n < t ≤ t n +1 (35)
⎝ 2 ⎠ Figure 2 shows the relative difference {Eq (38) - Eq (39)}/Eq
(39) versus reD.
By combining Equations (33) and (34) the expression for the 70,0%
calculation of the aquifer mean pressure in time t n +1 is
60,0%
obtained for the LIPB scheme:
difference relative (%)
50,0%
30,0%
And for the SIBP scheme, Equations (32) e (35) are combined,
resulting in: 20,0%
( )
10,0%
p n + p n +1
p a n +1 = p a n e −αΔt + 1 − e −αΔt (37)
2 0,0%
1,0 10,0 r e /r o 100,0
In his original work, Leung presented a table with model Fig. 2 - Relative difference of J for a finite radial aquifer using Equation
parameters for application in radial and linear aquifers. Table (38) e Equation (39)
3 is similar to the original Leung table, with the addition of an As shown in the graph of Fig. 2, for an aquifer smaller
equation for the case of a SIBP circular aquifer. A remark may than 10 times the reservoir, the difference in value is not
be made here about Leung's original work. In order to negligible.
reproduce the values of δ∞/ro presented in Table 4 for the case Furthermore, Table 4 shows that for values of reD where
of a circular SIBP aquifer, the equation suggested by this work ln(reD) < 0.75reD the value of J is inconsistent when
was used. The equation suggested by Leung in the original calculated by Equation (38). In spite of that, the accumulated
work does not reflect accurately the published data, probably influx volume calculated by the equation originally proposed
due to data input mistakes. Table 3 shows the equation that by Fetkovich is a good approximation compared to the van
reproduces the data published by Leung (1986). This may be Everdingen & Hurst model. In the cases presented below,
of interest to those willing to reproduce the data in the table values and errors with regard to the base model (van
for use in numerical programming. Everdingen & Hurst) using both equations will be shown.
Comparative study of models. Three cases were considered for comparing results of the
Before presenting comparative results for the influx from each accumulated influx calculation. At the end of this work, results
model, it is important to note that in the equation for the of Case 1 (Tables 7 to 12) are shown. The aim of showing
productivity index J of the aquifer, Fetkovich disregarded the these tables is to offer a source of reference to readers willing
ratio rD2/2reD2 that normally appears in solutions for the to reproduce the classical models in a discretized way. The
pressure drop in pseudosteady-state regime. When the solution values of WD for the van Everdingen & Hurst model were
is applied to the pressure drop in the well, it is reasonable to obtained by numerical inversion using the Stehfest algorithm
disregard the ratio rD2/2reD2 in the equation of the (1970).
pseudosteady-state regime. However, depending on the
6 SPE 107265
The variable for the comparison of the accumulated influx Case 2 - Aquifer radius 16 times larger than reservoir radius.
volume calculation is the relative aquifer-reservoir size, reD.
The choice of this variable is justified by its importance as a Aquifer radius 40000 ft
parameter for the configuration of commercial flow Reservoir radius 2500 Ft
simulators.
Keeping the same reservoir properties as in Case 1, the
Case 1 - Aquifer radius 8 times larger than that of reservoir. behaviour of the accumulated influx and the relative error for
all models are shown in Fig. 4.
Aquifer radius 20000 ft
Reservoir radius 2500 ft Accumulated influx (bbl std x 103)
Thickness 60 ft 1600
Fetkovich (J Equation 38)
Porosity 22 % 1400 Carte&Tracy
Permeability 100 md Leung PSS LIBP
Accumulated Influx
1200 Leung PSS SIBP
Water viscosity 0.30 cp Leung MPSS LIBP
1000 Leung MPSS SIBP
Formation compressibility 4x10-6 (psi)-1 Fetkovich (J Equation 39)
800
Water compressibility 3x10-6 (psi)-1
Initial pressure 3501 psia 600
400
400
Fig. 4 - Results of accumulated influx - Case 2
200
For larger aquifers (16 times the reservoir in radius), both
days the Carter & Tracy and the Leung MPSS models displayed a
0
0 100 200 300 400 500 600
small error. Thus, the Leung model applies both for medium
and for large aquifers. The result is satisfactory compared to
Model We (bbl std) Error (%) that of van Everdingen & Hurst.
van Everdinghen&Hurst 1079950 -
Fetkovich (J Equation 38) 1122198 3.91%
Fetkovich (J Equation 39) 1101956 2.04% Case 3 - Aquifer radius 20 times larger that reservoir size.
Carte&Tracy 1111224 2.90%
Leung PSS LIBP 1084709 0.44%
Leung PSS SIBP 1065042 -1.33% Aquifer radius 50000 ft
Leung MPSS LIBP 1072863 -0.63% Reservoir radius 2500 ft
Leung MPSS SIBP 1080090 0.01%
Accumulated influx (bbl std x 103) Accumulated influx (bbl std x 103)
1600 400
Fetkovich (J Equation 38)
1400 Carte&Tracy 350
Leung PSS LIBP
Accumulated Influx
Accumulated Influx
Model We (bbl std) Error (%) Model We (bbl std) Error (%)
van Everdinghen&Hurst 1490563 - van Everdinghen&Hurst 325903 -
Fetkovich (J Equation 38) 1351149 -9.35% Fetkovich (J Equation 38) 327048 0.35%
Fetkovich (J Equation 39) 1347405 -9.60% Fetkovich (J Equation 39) 304720 -6.50%
Carte&Tracy 1478500 -0.81% Carte&Tracy 347795 6.72%
Leung PSS LIBP 1337603 -10.26% Leung PSS LIBP 326051 0.05%
Leung PSS SIBP 1306029 -12.38% Leung PSS SIBP 325828 -0.02%
Leung MPSS LIBP 1494593 0.27% Leung MPSS LIBP 325843 -0.02%
Leung MPSS SIBP 1494757 0.28% Leung MPSS SIBP 325903 0.00%
Fig. 5 - Results of the accumulated influx - Case 3 Fig. 6 - Results of accumulated influx for very small aquifers.
The graph in Fig. 5 shows that it is indifferent whether The following figures record the error curves versus reD.
Equation (38) or (39) is used for the calculation of J in the Fig. 7 presents the error curves of the Fetkovich model using
Fetkovich model for large aquifers. Another point reported in Equations (38) and (39).
the original Leung work is the error increase in the PSS
model, which led the author to develop a simplified model, the Relative Error %
modified pseudosteady-state regime model, or MPSS. Note (base model - van Everdinghen&Hurst)
that the MPSS model error is practically the same as the van 12,0%
Everdingen & Hurst model. The graphs of Cases 1, 2 and 3 Fetkovich - Equation 38
show that as the aquifer size increases, two groups of models Fetkovich - Equation 39
10,0%
display distinct behaviours. The first group, composed of the
models Fetkovich (Equation 38 and 39) and Leung PSS, LIBP,
and SIBP, shows practically identical results, while the other
Relative Error %
8,0%
group, comprising the models MPSS, LIBP and SIBP, and
Carter & Tracy, also exhibits nearly identical behaviour, but
with results better than the first group compared to the base 6,0%
model (van Everdingen & Hurst).
For very small aquifers, such as those 4 times larger than 4,0%
the reservoir in radius, the models of Fetkovich (Equation 39)
and Carter & Tracy are less precise than those of Leung and
2,0%
Fetkovich (Equation 38), as shown in Fig. 6.
0,0%
0 5 10 r eD 15 20 25
period. In Fig. 8 the error curves of the Carter&Tracy and Other behaviour analyses of aquifer models can be done
Leung MPSS models are shown. to complement the present work.
7. It is important to point out that the conclusions and results
Relative Error % reported in the present work are applicable to aquifers
(base model - van Everdinghen&Hurst) with radial flow in limited reservoirs (pseudosteady-state
8,0%
regime).
Carte&Tracy
7,0%
Leung MPSS LIBP
Nomenclature
Leung MPSS SIBP
6,0% D Subscript - dimensionless variable
h Formation height
Relative Error %
Table 1: Classical equations of the van Everdingen & Hurst model for aquifers with radial and linear geometries
Aquifer Geometry
flow regime Radial Linear
Transient
K1 ( u ) tD
WD (u ) = WD (t D ) = 2
32
u K0 ( u ) π
Pseudosteady-state
I1 (reD u ) K1 ( u ) − I1 ( u ) K1 (reD u ) 1 − exp(−2 u )
WD (u ) = WD (u ) =
[
u 3 2 I 0 ( u ) K1 (reD u ) + I1 (reD u ) K 0 ( u ) ] [
u 3 2 1 + exp(−2 u ) ]
Steady state
I 0 ( reD u ) K1 ( u ) + I1 ( u ) K 0 ( reD u ) 1 + exp(−2 u )
W D (u ) = WD (u ) =
[
u 3 2 I 0 ( reD u ) K 0 ( u ) − I 0 ( u ) K 0 ( reD u ) ] [
u 3 2 1 − exp(−2 u ) ]
Table 2: Aquifer productivity index for radial and linear flow.
Aquifer Geometry
flow regime Radial aquifer Linear aquifer
Pseudosteady-state 2π f kh
J=
⎡⎛ r 2 ⎞ 2
1 ⎛ 3reD2 − 1 ⎞⎤
μ ⎢⎜⎜ eD
⎟
⎟ ln (r ) − ⎜ ⎟⎥
⎢⎣⎝ reD − 1 ⎠
2
4 ⎜⎝ reD2 − 1 ⎟⎠⎥
eD
⎦
3khw
(Equation proposed by the present work) J=
2π f kh μL
J =
⎡ ⎛r ⎞ 3⎤
μ ⎢ ln ⎜⎜ e ⎟⎟ − ⎥
⎣ ⎝ ro ⎠ 4 ⎦
(Equation of original Fetkovich work)
Steadystate 2π f kh
J= khw
⎛r ⎞ J=
μ ln⎜⎜ e ⎟⎟ μL
⎝ ro ⎠
kAL 2π f kh
, where f = θ 2π
μ (δ ∞ L ) μ (δ ∞ ro )
J
k η L2 k 2 η ro2 2
α = =
φμc t δ ∞ L (δ ∞ L ) φμc t δ ∞ ro reD
2
−1 (δ ) 2
∞ o reD − 1
r
⎡⎛ 2 ⎞ 2 ⎤
1 ⎛ 3r 2 − 1 ⎞⎟⎥
ro ⎢⎜ 2 eD ⎟ ln (reD ) − ⎜ 2eD
L r
LIBP, δ ∞
3 ⎢⎜⎝ reD − 1 ⎟⎠ 4 ⎜⎝ reD − 1 ⎟⎠⎥
⎣ ⎦
⎡ 2 1 ⎤
ro ⎢ 2 2 ⎥
⎣⎢ a1 reD reD − 1 ⎦⎥
4L
SIBP, δ ∞ ⎡2 1 ⎤
π2 Leung's equation ro ⎢ ⎥
⎣ a1 reD −1⎦
2 2
kt ηt kt ηt
tD 2
= 2
=
φμc t L L φμc t ro2 ro2
Constant a1 is the first root of Bessel's equation: J1 (am reD )Y0 (a m ) − J 0 (a m )Y1 (a m reD ) = 0 , where J0 and J1 are the Bessel first kind functions, and Y0 and Y1
the Bessel second kind functions.
Table 5: Productivity index of a radial infinite aquifer -- Equation (38) and Equation (39).
reD Equation (39) Equation (38) difference relative
1.1 300.3 -15.3 105.09%
1.5 61.1 -29.0 147.52%
2.0 31.7 -175.9 655.11%
3.0 17.3 28.7 65.78%
4.0 12.6 15.7 24.78%
5.5 9.5 10.5 10.63%
7.0 7.9 8.4 5.98%
7.6 7.5 7.8 4.95%
8.2 7.1 7.4 4.17%
.8.8 6.8 7.0 3.56%
9.4 6.5 6.7 3.08%
10.0 6.3 6.4 2.69%
14.5 5.1 5.2 1.21%
SPE 107265 11
Table 6: Drawdown pressure behavior of the aquifer-reservoir boundary used to Cases 1, 2 and 3.
t (d) 0 50 100 150 200 250 300 350 400 450 500
p (psia) 3501 3485 3474 3464 3455 3448 3442 3438 3435 3434 3434
Table 10: Total influx calculation -- Leung PSS SIBP model - Case 1.
pan+1
j t (d) p (psia) Wen+1 ΔWe
mean
0 0 3501 3501 0 0
5 50 3485 3499 39466 12217
10 100 3474 3495 124049 19629
15 150 3464 3489 237013 24238
20 200 3455 3483 366182 26637
25 250 3448 3476 501962 27297
30 300 3442 3470 636797 26586
35 350 3438 3463 764734 24795
40 400 3435 3458 881086 22154
45 450 3434 3453 982159 18843
50 500 3434 3449 1065042 15004
Table 11: Total influx calculation – Leung MPSS LIBP model- Caso 1.
pa, pss n+1 pa, mpss n+1
j t (d) p (psia) pLI(t) Wen+1 ΔWe
mean mean
0,0 0,0 3501 3501 3501 3501 0 0
5,0 50,0 3485 3484 3499 3498 47122 82849
10,0 100,0 3474 3473 3495 3494 135433 127346
15,0 150,0 3464 3463 3489 3488 250659 154658
20,0 200,0 3455 3455 3483 3482 380934 168432
25,0 250,0 3448 3447 3476 3475 516911 171542
30,0 300,0 3442 3442 3470 3469 651226 166252
35,0 350,0 3438 3438 3463 3463 778079 154345
40,0 400,0 3435 3435 3458 3457 892901 137225
45,0 450,0 3434 3434 3453 3452 992094 116000
50,0 500,0 3434 3434 3449 3448 1072863 91796
Table 12: Total influx calculation -- Leung MPSS SIBP model - Case 1.
pa, pss n+1 pa, mpss n+1
J t (d) p (psia) pSI(t) Wen+1 ΔWe
mean mean
0 0 3501 - 3501 3501 0,0 0
5 50 3485 3487 3499 3498 51719 14105
10 100 3474 3475 3495 3494 143736 20870
15 150 3464 3465 3489 3488 261322 24968
20 200 3455 3456 3483 3486 392897 26965
25 250 3448 3449 3476 3475 529338 27308
30 300 3442 3443 3470 3468 663460 26348
35 350 3438 3438 3463 3462 789602 24361
40 400 3435 3435 3458 3456 903305 21565
45 450 3434 3434 3453 3452 1001056 18132
50 500 3434 3433 3449 3448 1080090 14197