Vous êtes sur la page 1sur 14

SPE 104581

Transient Behavior of Multilateral Wells in Numerical Models:


A Hybrid Analytical-Numerical Approach
C. Aguilar, SPE, BP Alaska, and E. Ozkan, SPE, H. Kazemi, SPE, M. Al-Kobaisi, SPE, and B. Ramirez, SPE, Colorado
School of Mines

Copyright 2007, Society of Petroleum Engineers


transient period and flow convergence. However, a cursory
This paper was prepared for presentation at the 15th SPE Middle East Oil & Gas Show and grid refinement may not produce the desired accuracy; thus,
Conference held in Bahrain International Exhibition Centre, Kingdom of Bahrain, 11–14 March
2007. increasing the need for very fine grid that requires more
This paper was selected for presentation by an SPE Program Committee following review of
computational power and time. In addition, special grid
information contained in an abstract submitted by the author(s). Contents of the paper, as structures are often required to capture the details of flow
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to
correction by the author(s). The material, as presented, does not necessarily reflect any convergence around complex wellbores and to simulate the
position of the Society of Petroleum Engineers, its officers, or members. Papers presented at
SPE meetings are subject to publication review by Editorial Committees of the Society of
associated transient flow regimes, add to the overall
Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper complexity of the numerical computation.
for commercial purposes without the written consent of the Society of Petroleum Engineers is
prohibited. Permission to reproduce in print is restricted to an abstract of not more than
300 words; illustrations may not be copied. The abstract must contain conspicuous
acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O. Objective
Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435. The objective of this paper is to improve the representation of
single- and dual-lateral wells in numerical models for more
Abstract accurate and computationally efficient simulation of pressure-
This paper presents an extension of transient well index transient responses. Also, a practical approach will be
approach to simulate pressure transient behavior of presented to model the dual-lateral wells in uniformly
multilateral wells. This approach uses an analytical solution distributed, Cartesian grid. This approach involves proper
for the well index at early times and switches to the numerical accounting for the orientation, length, and friction-head loss of
well index at late times. The use of the transient well index the multi-lateral segment crossing a grid block. The proposed
eliminates the need for excessive grid refinement around the approach can be easily implemented in the conventional
well. In this paper, we have improved the accuracy of the reservoir simulators without compromising the computation
transient well index approach and have provided for a flexible time.
and easily implementable approach to place multilaterals in
conventional, Cartesian-grid reservoir models. Outline
After a summary of the pertinent literature and background,
Introduction we first present the improvement of well transmissibility for a
Pressure-transient responses of wells are conventionally single lateral. Then, we discuss the addition of a second lateral
analyzed and interpreted by using analytical solutions of to the model. We, finally, compare the accuracy of the
diffusion equation for relatively simpler reservoir numerical model against analytical solutions under various
architectures. For more complex reservoir situations, conditions of heterogeneity and skin effect.
involving multi-phase flow and reservoir heterogeneity,
numerical simulation is usually the only resort. Numerical Literature Review
simulators generally focus on the long-term performance of Blanc et al. (1999) applied an unsteady-state, radial flow
reservoirs. These simulators are not very sensitive to the short- equation for vertical wells, known as transient well index, to
term characteristics of flow in the near-wellbore region, which simulate pressure transients more accurately. They defined a
is the focus of the short-term, pressure-transient tests. For well-block radius that varied with time and was different from
example, the conventional transmissibility and well indices the steady- and unsteady-state definitions of the well-block
used in numerical simulation may adequately represent fluid radius proposed in the earlier studies by Peaceman (1977,
movement between relatively large grid blocks as well as the 1983) and Babu et al. (1991). This approach eliminated the
fluid withdrawal or injection at well blocks over relatively need for excessive grid refinement around the well.
large time steps when the transient radius of investigation of Another approach proposed for vertical wells involved the
the well is sufficiently large. Furthermore, these simulators use of an equivalent transmissibility in the near wellbore
fail to account for the flow convergence near the well region (Blanc et al., 1999, Ding et al., 1998, and Ding and
accurately at shorter times unless very small grid and time Renard, 1994). In conventional models, transmissibility terms
steps are used. for the well grids are assumed to have the same functional
Grid refinement around well has been used both to form as those in the grids without wells or well perforations.
improve numerical calculation of bottom-hole pressure during
2 SPE 104581

Al-Mohannadi et al. (2003, 2004) documented the use of Ding et al. (1998), proposed a method by modifying the
the conventional- and transient-well index concepts for transmissibility around the well grids to improve the flow
different grid structures and time step sequences for horizontal calculations. This approach, referred to as the corrected
wells. Their investigation used a three-dimensional, two-phase transmissibility, is applicable to off-center wells and to
flow simulator developed by Al-Mohannadi (2004). They flexible grids.
compared the simulated responses with the analytical Blanc et al. (1999) recognized that the pseudo-steady-state
pressure-transient solution for a horizontal well in a closed well index is not appropriate for simulating early transient
bounded reservoir given by Ozkan and Raghavan (1991). time behavior of well tests. To correct this problem, they
They concluded that the conventional well index should be derived a transient numerical productivity well index for
used with log-distributed grid to obtain a close match with the Cartesian grid using the following unsteady-state radial flow
analytical solution at early times. They also concluded that equation.
there was an upper limit to the well-grid coarseness for the
4π ( 0.006328 ) Δzk , (6)
conventional well index, beyond which the transient results TWI =
⎛ ro2 ⎞ ⎛ rw2 e −2 s ⎞
would loose numerical accuracy, but applying the transient Ei ⎜ − ⎟ − Ei ⎜ −
⎜ 4 × 2.637 × 10 kt ⎟
−4 ⎜ 4 × 2.637 × 10−4 kt ⎟⎟
well index in a coarse grid would yield a good match of the ⎝ ⎠ ⎝ ⎠
analytical solution. In addition to the above studies, the other They also concluded that the Peaceman approach could be
pertinent references include Goktas and Ertekin (1999), Ding used as long as the following relation is fulfilled and only if
and Jeannin (2001), Archer and Yildiz (2001), Ozkan et al. the well is located at the center of a grid block surrounded by a
(1997), and Ozkan et al. (1989). minimum of 5x5 uniform nodes:
Several specific references about numerical modeling of
wells and analytical models for single and dual-lateral wells φμ ct ΔL2
deserve particular attention because of their direct relevance to t≥ , or t DL ≥ 2.6 (7)
10−4 k
this work. Peaceman (1977) conducted one of the first studies
concerning well representation in numerical models. He For early times, they also concluded that their transient
concluded that the bottomhole flowing wellbore pressure at numerical productivity index correction yielded much better
radius rw could be related to the block pressure calculated at an results than the conventional Peaceman approach.
equivalent radius ro by the following relation: An improved extension of Blanc et al (1999), is used in
this paper for application to horizontal and multilateral wells.
q To validate the accuracy of the simulated results, we compare
pwf = po − , (1)
WI them with analytical solutions. Below, we briefly present the
analytical solutions used in the comparisons.
where, WI is the conventional well index derived based on the
steady-state, radial flow assumption and given by
Analytical Models for Single- and Dual-lateral Wells
2πΔzk For single horizontal wells, we compare our numerical
WI = . (2) simulation results with the analytical solution of Ozkan-
μ ln ⎡⎣ ro (rw e− s ) ⎤⎦
Raghavan (1991) for a closed, rectangular reservoir. For dual-
Peaceman (1977) showed that for a square grid, the lateral wells, we use Ozkan et al. (1997) pressure-transient
equivalent radius is ro = 0.2Δx. For unsteady-state flow, ro can solution.
be derived from: In our comparisons, we focus on specific characteristics of
pressure-transient responses for single- and dual-lateral wells.
1/ 2 The flow regimes can be divided in two groups: infinite-acting
ro = ΔL ⎡⎣ 4t D exp ( −γ − 4π pDb ) ⎤⎦ . (3)
and boundary-dominated. Infinite-acting period is important
for estimating reservoir parameters while boundary-dominated
2π kh flow is important for productivity estimation and production
Where, pDb = ( pi − pb ) and γ = 0.5772 .
qμ forecast. Therefore, infinite-acting flow regimes are of interest
to this study.
In a later work, Peaceman (1983) extended the equivalent In a horizontal-well test, three major flow regimes may be
radius definition to a rectangular grid block given by: identified during the infinite-acting period: early-time radial,
1 intermediate-time linear, and late-time pseudoradial flow
ro = 0.14 ⎡ Δx 2 + Δy 2 ⎤ 2
. (4) (Ozkan et al., 1989, and Ozkan, 2001). Fig. 1 shows the
⎣ ⎦ fundamental characteristics of these flow regimes.
The analysis was also extended to include anisotropic For dual-lateral wells, four flow regimes may be identified
permeability: during the infinite-acting flow period (Ozkan et al., 1997).
1
These flow regimes are shown in Fig. 2 and may be grouped
(
⎢ y x ( ) ) x y ( )⎥ (
⎡ k k 1 2 Δx 2 + k k 1 2 Δ y 2 ⎤ 2
) under two categories. The first category includes the flow
ro = 0.28 ⎣ ⎦ . (5) regimes during which the laterals do not have interference
from each other. These are the early-time radial flow in the
⎢⎣ y x (
⎡ k k 14 + k k 14⎤
x y ⎥⎦ ) ( ) vertical plane, intermediate-time linear flow in the horizontal
plane, and intermediate-time pseudo-radial flow in the
horizontal plane. The second category includes the late time
SPE 104581 3

pseudo-radial flow regime in the horizontal plane that appears


G k
when the laterals feel the influence of each other. All of these v = − ( ∇p − γ ∇ D ) . (9)
flow regimes may not develop in every system. μ
The finite-difference form of Eq. (8) may be expressed as
⎡ Δ xTx ( Δ x p − γΔ x D ) ⎤
⎢ ⎥

( ⎥
V
⎢ +Δ yTy Δ y p − γΔ y D ⎥ − q = R Δ t (φ b ) .
Δt
) (10)
⎢⎣ +Δ z Tz ( Δ z p − γΔ z D ) ⎥⎦

A complete expansion of Eq. (10) is shown in the Appendix.

Time Steps
Logarithmic time steps were used to run the simulations. To
generate the time steps, the following algorithm was used:

tn = Δt1e(
n −1) Δτ
(11)
for n = 1, 2 ,….Nmax,
Δtn = tn − tn −1 (12)
Fig. 1 - Pressure transient flow regimes for a horizontal well.
and
1 ⎛t ⎞
Δτ = ln ⎜ max ⎟. (13)
N max − 1 ⎝ Δt1 ⎠

Well Representation
The objective of this project is to use a transient well index for
horizontal laterals. The flow rate from a well is defined as
I well max
qtn = ∑ i =1
q in, +j ,k1 −
5.6146C
Δt / 24
(
pwell ni , +j ,1k − pwell ni , j ,k (14) )
where, q in, +j ,k1 is the flow rate at the sand face defined by

(
q in, +j ,k1 = WI in, j ,k p in, +j ,k1 − pwell ni , +j ,1k , ) (15)

Fig. 2 - Pressure transient flow regimes for dual-lateral wells and C is the wellbore storage coefficient. Thus,
⎧ I well max n ⎫
Numerical Model to Simulate Pressure-Transient
Response of Multilateral Wells

⎪ i =1 ∑ (
WI i , j ,k p in, +j ,k1 − pwell ni , +j ,1k )⎪⎪
qt = ⎨
n
⎬ , (16)
The numerical simulator used in this paper is a modification of
the single-phase simulator developed in Al-Mohannadi (2004).
⎪ 5.6146C n +1
(
⎪− Δt / 24 pwell i , j ,k − pwell i , j ,k

n
) ⎪


Modifications include dual-lateral well representation,
improved transient well-block radius calculation, and transient and
transmissibility for the well blocks. Cartesian grid system is
used for all simulations. n +1 n +1 ⎢
(
⎡ Di , j , k − Di +1, j ,k )⎤⎥
pwell = p + γ n
i + 2 , j ,k ⎢
1 . (17)
i , j ,k welli +1, j ,k
− HFi n+ 1 , j ,k ⎥
Finite-Difference Formulation ⎣ 2 ⎦
The single-phase, mass-balance equation for fluid flow in
porous media is given, in Cartesian coordinates, by In Eq. (17), HF is the friction head given by
2
∂ ∂ ∂ ∂ Δxi + 1 , j ,k vwellt
− ( ρ vx ) −
∂x ∂y
( )
ρ v y − ( ρ vz ) − ρ qˆ = ( ρφ ) . (8)
∂z ∂t HFi n+ 1 , j ,k = fi n+ 1 , j ,k
2 i + 1 , j ,k
2
, (18)
2 2 2d pipe g
Where, vx, vy, and vz are the x, y, and z-components of Darcy
where,
velocity given by
4 SPE 104581

⎛ n ⎞ (Ozkan et al., 1989 and Clonts and Ramey, 1986) and is given
n
⎜ qwellti+1, j ,k Bi +1, j ,k ⎟ by
vwellt =⎜ ⎟ / 86400 , (19)
i + 1 , j ,k
2 ⎜⎜ π 2 ⎟⎟ ⎧ ⎡ ⎤⎫
d
⎝ 4 ⎠ ⎪ ⎢ ⎥⎪
qB μ ⎪ ⎢ ro2 ⎥⎪
Δp = − ⎨ Ei ⎢ − ⎬
64 4π (0.006328) Lh k y k z ⎪ ⎢ ⎛ 2.637 × 10−4 k z ⎞ ⎥⎥ ⎪
fi n+ 1 , j ,k = , for N RE ≤ 2000 (20)
2 n
N REi ⎪ ⎢ 4 ⎜⎜ φ ct μ
t⎟ ⎪
⎟⎥
+ 12 , j , k ⎩ ⎣ ⎝ ⎠ ⎦⎭
(28)
0.06157 ρin+ 1 , j , k ⎛ q n B n ⎞
In order to use this equation in the numerical model, it should
n
N REi = 2 ⎜ t ⎟, (21)
+ 12 , j , k
5.6146rw ⎜ μ ⎟ be adapted to the grid system used. This is done by replacing
⎝ i +1, j , k ⎠
Lh by the length of the well-grid Δxi. Using the Newton
Raphson method, we solve for ro as follows:
ρ ( p ) = ρb ⎡⎣1 + c ( p − pb ) ⎤⎦ , (22)
f (roi )
roi+1 = roi − , (29)
ρ f ′(roi )
γ= , (23)
144 where
And qt in stb/d in Eq. (21). ⎧ ⎡ ⎤⎫
⎪ ⎢ ⎥⎪
For horizontal wells, the steady-state well index is qB μ ⎪ ⎢ ro2i ⎥⎪
f (roi ) = − ⎨ Ei ⎢ − ⎥ ⎬ − Δp
expressed as 4π (0.006328)Δx k y k z ⎪ ⎢ ⎛ 2.637 × 10−4 k z ⎞ ⎥⎪
⎛ k ⎞ Δxi , j ,k
n
⎪ ⎢ 4 ⎜⎜ φ ct μ
t⎟ ⎪
⎟⎥
WI in, j , k = 2π ( 0.006328 ) ⎜ ⎟ ⎩ ⎣ ⎝ ⎠ ⎦⎭
⎝ μβ ⎠i , j ,k ⎡⎛ ⎛ r ⎞ 2 ⎞ r ⎤ (30)
⎢⎜1 − ⎜ w ⎟ ⎟ ln o + s ⎥
⎢⎜ ⎝ ro ⎠ ⎟ rw ⎥ In Eq. (29), roi+1 and roi are the new and old estimates of the
⎣⎝ ⎠ ⎦
(24) equivalent well-block radius and Δp = poi - po, where poi is the
and the transient well index is expressed as initial pressure and po is the pressure in the well-block.
The time-dependent well-block radius calculated by this
Δxi , j , k
n
⎛ k ⎞ approach is applicable for (Blanc et al, 1999):
TWI in, j ,k = 4π ( 0.006328 ) ⎜ ⎟
⎝ μβ ⎠i , j , k ⎧ ⎛ φ ct μ ro2 ⎞ ⎫
⎪ Ei ⎜⎜ − ⎟ ⎪ φ ct μ ⎧⎪ ( L 2 )2 z 2 ( z − h )2 ⎫⎪
⎪ ⎝ 4 × 2.634 × 10 k z t ⎟⎠ ⎪
−4
t≤
h w
min ⎨ ; w ; ⎬ (31)
⎪ ⎪ 10−4 ⎪⎩ k 5 k 5 kz ⎪

⎨ ⎛
( ) ⎞⎬
2 x z
−s
⎪ ⎜ φ c t μ r w e ⎟⎪
⎪− Ei ⎜ − −4 ⎟⎪ In Eq.(31), zw is the elevation of the horizontal well from the
⎪ ⎜ 4 × 2.634 × 10 k z t ⎟⎪
⎩ ⎝ ⎠⎭ bottom boundary of the formation (well eccentricity), and h is
(25) the formation thickness. After this time, ro becomes constant
given by Eq. (27).
where
k = k y kz . (26) Modifications for Dual-Lateral Wells
To simulate transient pressure responses of a dual-lateral well,
Eq. (25) is used in this paper to calculate the transient well two horizontal wells are introduced in the system. The
index with an equivalent well-block radius ro(t) that varies as a simplest approach to model dual-laterals in a Cartesian grid
function of time. This time-dependent equivalent well-block system is to place one of the laterals along the length of the
radius, ro(t), is different from the common steady-state flow grids and the other lateral along the diagonal of the grids as
definition of ro given by shown in Fig. 3.
The numerical representation of the lateral along a grid
⎛ Δy Δz ⎞ −0.5 axis is the same as the single-horizontal well case discussed
ro = ⎜ ⎟e . (27)
⎝ π ⎠ above. However, for the lateral in the diagonal direction, it is
necessary to make adjustments to the well index, friction head,
We explain the time-dependent well-block radius below. and equivalent well-block radius. For a single horizontal well,
these parameters are calculated based on the distance between
Calculation of the Transient Well Block Radius each perforation (Δx). If the well is in a diagonal direction, the
The equation used to calculate the well-block radius is distance between perforations is Δl given by
obtained by considering radial flow around the axis of
horizontal well in the vertical plane. This equation is the same Δl = Δx 2 + Δy 2 . (32)
as the exponential integral solution for early-time radial flow
SPE 104581 5

Δxi , j ,k Δyi , j ,k
n
⎛ k ⎞
Tzn = 0.006328 ⎜ z ⎟ (37)
iwell , jwell ,kwell + 1
2 ⎝ μ B ⎠i , j ,k + 12 Δzi , j ,k + 12 − ro ( t )

L h2

?l

Lo

?x

Lh1

Fig. 4 - Distance correction for transmissibility terms in well-grids

Fig. 3 - Dual-lateral wells in Cartesian grid system The transmissibility modification significantly influences
the variation of the well-block radius, ro(t). Fig. 5 shows the
Thus, for the lateral in the diagonal direction, Δx is replaced behavior of the well-block radius, ro(t), which includes the
by Δl. In addition, the horizontal permeability normal to the effect of the transmissibility modification compared with the
well axis, ky, is replaced by k ŷ , given by well-block radius used in the original model (Al-Mohannadi,
2004) (denoted by ro(t) original in the figure). Both well-block
k yˆ = kwi = k y cos 2 θi + k x sin 2 θi . (33) radii are calculated with the Newthon-Rapshon method
explained earlier. When the well-block radius increases with
time, the distance between adjacent grids used in the
Transmissibility Modification for Grids with Perforations transmissibility terms decreases (Fig. 4). This causes an
Transmissibility terms for the grids with perforations should increase in the transmissibility values. Also shown in Fig. 5 is
be calculated accordingly with the well-grid conditions. the behavior of roT(t) used in the transmissibility terms.
Transmissibility terms [Eqs. (A9) through (A14) in the 1.E+02 ro(t) roT(t)
Appendix] are calculated between the mid-points of adjacent ro(t) Original Derivative

grids. The grid dimensions, ∆x, ∆y, or ∆z in the denominator


of these equations account for the distance between the mid-
Early-Time Radial Flow
points of adjacent grids. During early times, the unsteady-state
behavior of flow makes the well-block radius variable as a 1.E+01

function of time. Thus, the distance between adjacent grids


with well perforations is also variable as a function of the
ro(t), ft

well-block radius. The same principle applies to the vertical


plane transmissibilities in the y- and z-directions (Fig. 4).
1.E+00
Therefore, the transmissibility terms are modified to account
for the time dependence of the well block radius as shown
below:
Δxi , j ,k Δyi , j ,k
n
⎛ k ⎞
Tzn = 0.006328 ⎜ z ⎟ (34) 1.E-01
iwell , jwell ,kwell − 1
2 ⎝ μ B ⎠i , j ,k − 12 Δzi , j ,k − 12 − ro ( t ) 1.E-04 1.E-03 1.E-02 1.E-01 1.E+00 1.E+01
Time, hr
1.E+02 1.E+03 1.E+04

Fig. 5 - Well-block radius as a function of time


n
⎛ ky ⎞ Δxi , j , k Δzi , j , k
Tyn = 0.006328 ⎜⎜ ⎟⎟ (35) Fig. 6 shows the behaviors of the modified
iwell , jwell − 1 ,kwell ,
2 ⎝ μ B ⎠i , j − 12 , k Δyi , j − 12 , k − ro ( t ) transmissibility, Tmod, which includes the effect of variable
roT(t), and the original transmissibility, T, which is constant.
⎛ ky ⎞
n
Δxi , j , k Δzi , j , k There is, however, a limit for the variation of the
Tyn = 0.006328 ⎜⎜ ⎟⎟ (36) transmissibility, Tmod, in time. It has been found that roT(t) used
iwell , jwell + 1 ,kwell
2 ⎝ μ B ⎠i , j + 12 , k i , j + 12 , k − ro ( t )
Δ y in the transmissibility terms should become constant at
and ( )
t = 7.4 × 10−8 kt φct μL2h where k is given by Eq. (26). At the
end of the early-time, radial flow period, roT(t) should decrease
in order to match the analytical solution (Fig. 5). Therefore, as
shown in Fig. 6, the transmissibility terms start decreasing at
6 SPE 104581

the end of the early-time flow period. Fig. 6 also shows the modification. This case was run using the grid and well
comparison between the steady-state well index (WI) and the structure described in Table 2. Fig. 7 shows the pressure drop
transient well index (TWI) using the well-block radius, ro(t). and its derivative for Case A1. Also shown, for reference, in
Note that the two well indices become close at the end of the Fig. 7 is the pressure response using the conventional, steady-
early-time flow period. In the following sections, it can be state well index denoted by NPI. Fig. 7 indicates that the use
seen that the transmissibility modification improves the match of both transient well indices yields better results than the
with the analytical solution at early times for both single- and conventional well index. Between the two transient well
dual-lateral wells. indices, the one with the transmissibility modification
improves the match with the analytical solution.

TABLE 1 - RESERVOIR AND WELL PROPERTIES USED IN THE


SIMULATIONS
Property Value
Porosity, φ, fraction 0.2
-1
Fluid Compressibility, c, psi 4 x 10-6
Formation Compressibility, cf, psi-1 1 x 10-5
Formation Volume Factor, B, bbl/stb 1.02
Viscosity, μ, cp 1.0
Formation Temperature, T, °F 120
Initial Pressure, p, psi 3000
x-direction Permeability, kx, md 10, 20, 50
y-direction Permeability, ky, md 10, 20, 50
Fig. 6 - Transmissibility and well index behavior as a function of z-direction Permeability, kz, md 5, 25
time Formation Thickness, h, ft 150, 220, 270
x-Direction Formation Size, xe, ft 2000,3600,6000
Validation of the Numerical Model
The objective of this section is to validate the numerical y-Direction Formation Size, ye, ft 2000,3600,6000
modeling approaches of this study for simulating pressure Horizontal and Laterals Well Length, Lh, ft 500,1400,2000
transient responses of single- and dual-lateral wells. Simulated Wellbore Radius, rw, ft 0.25
pressure-transient and derivative responses are compared with Constant Flow Rate, q, scf/d 2000
the responses obtained from the analytical models for different
sets of data.
The numerical model is tested for different conditions
TABLE 2 - GRID AND WELL DIMENSIONS FOR CASE A1
including three-dimensional permeability anisotropy, laterals
Imax 12 IWELL 5-8 Δx, ft 500
in different layers, laterals with different lengths, and uniform Jmax 13 JWELL 7 Δy, ft 20
and non-uniform skin. This section also documents the results Kmax 7 KWELL 4 Δz, ft 20
on the effect of different grid sizes, use of rectangular grids in Xe, ft 6000 Ye, ft 6000 Ze (h), ft 220
the horizontal plane to simulate laterals with arbitrary phasing, Lh, ft 2000
and on the comparison between steady-state well index and
transient well index. Case A2 – Effect of Permeability
To validate the transmissibility modification approach and For Case A2, three different sets of permeability data are
the well representation for dual-laterals wells, different cases used. Set 1 represents the base permeabilities defined in Table
were run and are shown in this section. The data for the 3. In Sets 2 and 3, the horizontal and three-dimensional
reservoir and well properties used in the simulations are equivalent permeability are increased, respectively. Table 3
shown in Table 1. shows the reservoir and well dimensions as well as the
permeability values used in this section. The purpose of this
Single Lateral Cases section is to test the model for different durations of various
In this section, three cases are considered: Case A1 flow regimes. By increasing the horizontal permeability, the
investigates the effect of the transmissibility modification on intermediate-time flow regime becomes shorter for the same
the transient well index, Case A2 shows the effect of reservoir dimensions because the lateral boundaries are felt
permeability and flow regimes, and Case A3 highlights the earlier than that in the base case (Fig. 8). By increasing the
effect of grid size. three-dimensional equivalent permeability, the early-time
radial flow ends earlier because of the larger vertical
Case A1 – Transient Well Index with and without permeability. For all three cases simulated in this section, the
Transmissibility Modification pressure and derivative responses match well with the
Case A1 compares the pressure-transient responses using analytical solution (Fig. 8).
transient well index with and without the transmissibility
SPE 104581 7

pressure drop is smaller and the early-time flow regime is


longer. Table 4 shows the reservoir and well dimensions used
in this case. For all the cases shown in Fig. 9, the match with
the analytical model is excellent. This indicates that the
modifications made to the numerical model in this study are
not sensitive to the grid dimensions.

TABLE 4 - DATA USED TO INVESTIGATE THE EFFECT OF THE


GRID SIZE FOR A SINGLE HORIZONTAL WELL (CASE A3)
Grid size, ft 90 Grid size, ft 50
Imax 40 Imax 40
Jmax 40 Jmax 40
Kmax 3 Kmax 3
xe, ft 3600 Xe, ft 2000
ye, ft 3600 Ye, ft 2000
ze (h), ft 270 ze (h), ft 150
IWELL 13-27 IWELL 16-25
Fig. 7 – Pressure-transient response with the transmissibility JWELL 20 JWELL 20
modification (Case A1) KWELL 2 KWELL 2
Lh, ft 1350 Lh, ft 500
TABLE 3 - RESERVOIR AND PERMEABILITY VALUES USED FOR
CASE A2
Imax 40 IWELL 13-27 Δx, ft 90
Jmax 40 JWELL 20 Δy, ft 90
Kmax 3 KWELL 2 Δz, ft 90
xe, ft 3600 ye, ft 3600 ze (h), ft 270
Lh, ft 1350
Set 1 (Base) Set 2 Set 3
kx md 10 kx, (khx2) md 20 kx, (kx5) md 50
ky, md 10 ky, (khx2) md 20 ky, (kx5) md 50
kz, md 5 kz, (khx2) md 5 kz, (kx5) md 25

Fig. 9 - Effect of the grid size on the pressure-transient response


of a horizontal well (Case A3)

Dual-Lateral Cases
In this section, the transmissibility modification approach is
tested for dual-lateral wells. In addition, the new approach
proposed in this paper to represent dual-laterals in Cartesian
grid is validated. For the cases presented in this section, the
dual-lateral wells are placed with a 45º angle in the same
horizontal plane. Next section, explores the extension of this
Fig. 8 - Effect of permeability on the pressure-transient response approach to arbitrary phasing of the laterals.
of a horizontal well (Case A2)

Case B1 – Transient Well Index with and without


Case A3 – Effect of Grid Size
Transmissibility Modification
Two different square grid sizes are used in Case A3: 50 ft
Similar to Case A1 for a single lateral discussed earlier,
and 90 ft. For both grid sizes, three layers are considered in
Case B1 compares the transient well index with and without
the vertical direction with the same grid dimensions.
the transmissibility modification with the conventional well
Therefore, each case has a different reservoir thickness. The
index. The results shown in Fig. 10 indicate that the
results are shown in Fig 9. This case is mainly to test the use
transmissibility modification yields responses that are closer to
of larger grids (smaller, 20ft-by-20ft grids were used in Case
the analytical model than those obtained with the conventional
A above). At the same time, it is possible to see the variation
transmissibility and conventional well index. Therefore, in the
of the pressure drop for different reservoir dimensions with the
rest of this paper, transmissibility modification is used to run
same permeability. For a thicker reservoir (h=270ft), the
8 SPE 104581

all cases for dual-lateral wells. Table 5 shows the data for the uniform skin, laterals with different length, and formation
laterals and reservoir simulated in this case. These data with damage.
Set 1 permeabilities are also used as the base case for the other
simulated dual-lateral well cases in this section.

TABLE 5 - DATA USED FOR DUAL-LATERAL WELL


SIMULATIONS (CASE B1)
Imax 40 Jmax 40 Kmax 3
Δx, Δy, Δz, ft 90 Xe, ft 3600 Ye, ft 3600
IWELL 13-27 JWELL 13 KWELL 2
IWELLD 14-24 JWELLD 14-24 KWELLD 2
ze (h), ft 270 Lh1, ft 1350 Lh2, ft 1400
θ, degrees 45
Set 1 (Base) Set 2 Set 3
kx, md 10 kx,(khx2) md 20 kx, (kx5) md 50
ky, md 10 ky, (khx2) md 20 ky, (kx5) md 50
kz, md 5 kz, (khx2) md 5 kz, (kx5) md 25

Fig. 11 - Effect of permeability on the pressure-transient


responses of dual-lateral wells (Case B2)

TABLE 6 - DATA USED TEST THE EFFECT OF THE GRID SIZE


FOR DUAL-LATERAL WELLS (CASE B3)
Imax 40 IWELL 16-25
Jmax 40 JWELL 17
Kmax 3 KWELL 2
Xe, ft 2000 IWELLD 17-23
Ye, ft 2000 JWELLD 18-24
Ze (h), ft 150 KWELLD 2
Lh1, ft 500 Lh2, ft 495
Δx, ft 50 θ, degrees 45
Δy, ft 50
Δz, ft 50

Fig. 10 - Pressure-transient response with the transmissibility


modification (Case B1)

Case B2 – Effect of Permeability


Case B2 is the same as Case A2 for a single lateral. The
data used for this case are shown in Table 5. In addition to the
base case permeabilities, two more cases are considered. In
one of the cases, the horizontal permeability is increased
whereas the other case corresponds to an increase for the
three-dimensional equivalent permeability. As shown in Fig.
11, for all permeability values considered, the simulated
pressure-transient responses yield a good match with the
analytical solution.
Case B3 – Effect of Grid Size
Two different square grid sizes were chosen (50ft and 90ft)
for Case B3. The well and reservoir dimensions for the 90ft x
90ft and 50ft x 50ft grid cases are shown, respectively, in
Table 5 and Table 6. It is shown in Fig. 12 that the simulated
Fig. 12 - Effect of grid size on pressure-transient responses of
responses match well with the analytical solution for both dual-lateral wells (Case B3)
cases.
Rectangular Grids
More Complex Dual-Lateral Cases Using rectangular grids allows placing laterals at any angle
Additional cases for dual-lateral wells are discussed in this by changing the dimensions of grids in the horizontal plane
section to validate the model proposed for more complex (∆x and ∆y). For example, the cases discussed in the previous
conditions. These cases include rectangular grids in the sections are limited to dual-laterals at 45º angle because of the
horizontal plane, laterals in different layers, uniform and non- use of square grids. In this section, laterals are placed at 60o
SPE 104581 9

angle. The desired angle is obtained by using different


dimensions for ∆x and ∆y as show in Table 7 and placing one TABLE 8 - GRID AND WELL DIMENSIONS USED TO SIMULATE
LATERALS IN DIFFERENT LAYERS
lateral along the x-axis and the other lateral along the diagonal
Δx, ft 50 Imax 30
direction of the grids, similar to that shown in Fig. 5. The
Δy, ft 50 Jmax 30
pressure-transient responses simulated with this rectangular
Δz, ft 50 Kmax 5
grid structure compare well with the analytical model as Xe, ft 1500 IWELL 11-20
shown in Fig. 13. Therefore, this approach is suitable to model Ye, ft 1500 JWELL 11
dual-laterals of arbitrary phasing (or single horizontal wells Ze (h), ft 250 KWELL 2
not aligned with the permeability directions). The advantage Lh1, ft 500 IWELLD 12-18
of the proposed approach is the use of the Cartesian grids Lh2 ft 495 JWELLD 12-18
while modeling arbitrary orientations of the wells. The use of θ, degrees 45 KWELLD 4
Cartesian grid system in the manner proposed in this paper
leads to a structured grid and a simplified matrix. This helps to
reduce computational problems.

TABLE 7 - GRID STRUCTURE USED FOR RECTANGULAR GRID


CASE
Δx, ft 52 Imax 40
Δy, ft 90 Jmax 40
Δz, ft 90 Kmax 3
Xe, ft 2080 IWELL 15-27
Ye, ft 3600 JWELL 13
Ze (h), ft 270 KWELL 2
Lh1, ft 780 IWELLD 14-26
Lh2, ft 1351 JWELLD 14-26
θ, degrees 60 KWELLD 2

Fig. 14 - Pressure-transient responses for dual-laterals placed in


different layers

Skin
In this section, the effect of skin factor on dual-lateral well
responses is discussed. The cases run include uniform and
equal skin for each lateral, uniform but different skin for each
lateral, and non-uniform skin along each lateral. Semi-log
analysis is also demonstrated to validate the results obtained.
The data used to run these cases are shown in Table 5 and
the skin values are shown in Table 9. Fig. 15 shows the
simulated pressure-transient responses for dual-lateral wells
with uniform and equal skin for each lateral. As expected, the
o
Fig. 13 - Pressure responses for dual-lateral wells with 60 angle pressure drop is higher than the undamaged well case and the
in rectangular grids simulated responses yield a good match with the analytical
model.
Laterals Placed in Different Layers It must be noted that the skin effect is modeled differently
In this case, a square grid with five layers is used as shown in the analytical and numerical models considered here. The
in Table 8. The laterals are placed in the top and bottom analytical model uses a skin value which is scaled with the
layers. The well location, zw is incorporated into the simulator ratio of the horizontal well length to the formation thickness.
with the following equation: Also, for anisotropic formations, the scaling includes the ratio
of directional permeabilities. The input skin values used to
Δz
zw = ( h − Δz × k well ) + , (38) compute the analytical and numerical results shown in Fig. 15
2 are related by the following expression:
and is also used in Eq. (31) to calculate the end of the early
kx k y kz h
time radial flow. In Eq. Error! Reference source not found., s( analytical ) = s( numerical ) . (39)
kwell is the layer number where the lateral is placed. Fig.14 k y k z Lh
shows that the pressure-transient responses for this case match
the analytical model very well.
10 SPE 104581

TABLE 9 - THE RESULTS OF THE SEMI-LOG ANALYSIS FOR solution used in this study does not include the option for non-
SKIN ESTIMATES
uniform skin distribution.
Numerical Input Semi-log Analytical Input Semi-log
Δpw 3.7703 Δpw 11.19088604
t 1.43E-04 t 3.50E-02
S1=1, S2=1 8.63E-01 S1=0.22, S2=0.21 1.01E+00
mer 2.922776 mer 2.942913
keff 7.206281156 keff 7.156971315

Figure 16 - Semi-log plot for skin analysis – uniform skin case

TABLE 10 - VALUES USED FOR THE NON-UNIFORM SKIN CASE


Lateral 1 Lateral 2
Well-node Skin Well-node Skin
1 4 1 4
Fig. 15 - Pressure-transient responses for dual-lateral wells with 2 4 2 4
uniform skin
3 4 3 3
4 3 4 3
Fig. 16 is the semi-log plot of the pressure responses in 5 3 5 2
Fig. 15 for the uniform skin case. From this plot, it is possible 6 3 6 2
to calculate the slope of the early-time semi-log straight line, 7 2 7 2
mer, and estimate the skin factor by using the following 8 2 8 3
9 2 9 3
equation:
10 3 10 4
⎛ Δpwf keff ⎞ 11 3 11 4
s = 1.151⎜ − log t − log + 3.23 ⎟ , (40) 12 3
⎜ mer 2
φ ct μ rw ⎟ 13 4
⎝ ⎠
14 4
where 15 4

162.2qB μ The semi-log plot of the early-time responses in Fig. 17 is


keff = (41)
mer [ Lh1ϕ (θ w1 ) + Lh 2ϕ (θ w2 )] shown in Fig. 18. The skin estimates from the semi-log
analysis for this case are given in Table 11.
and
TABLE 11 - ESTIMATES OF THE NON-UNIFORM SKIN VALUES
ϕ (θ wi ) = k y k z cos 2 θ wi + k x k z sin 2 θ wi . FROM SEMI-LOG ANALYSIS

Eq. (40) is the skin equation for a vertical well with keff Different skin
Different skin
per well Semi-log Semi-log
calculated from the slope of the early-time semi-log straight per perf.
line for a dual-lateral well. Eq. (40) assumes that each Δpw 6.0107 Δpw 8.8073
perforation is a small vertical well in a slab of thickness ∆x, t 1.43E-04 t 1.43E-04
which is the grid size along the horizontal well axis. The skin Skin S1=1 S2=5 1.54E+00 Diff. Skin per perf 2.83E+00
factors calculated for the numerical and analytical pressure mer 2.922776 mer 2.942913
responses are shown in Table 9. The agreement between the keff 6.394832969 keff 7.158351609
input and calculated values are in reasonable agreement for the
simulated responses. It is possible to run an additional case using the skin value
Fig. 17 shows the results for cases in two categories: The calculated with the semi-log analysis as a uniform skin value
first category is for uniform but different skin values (1 and 5) to check if the semi-log estimated values may be interpreted as
in each lateral. The other category considers non-uniform (U- an average skin for the system. Figure 19 shows that for non-
shape) distributions of skin factor along each lateral as shown uniform skin along each lateral, the skin calculated from the
in Table 10. The results for the non-uniform skin case are not semi-log analysis is close to an average skin for the system.
compared with the analytical solution because the analytical On the other hand, for uniform but different skin values for
SPE 104581 11

each lateral, the skin calculated with semi-log analysis and consider different permeability ratios for the layers including
used as uniform skin for the system does not match the the laterals. In the first case, the layers with the laterals have a
pressure drop in the original case. Therefore, this value cannot smaller permeability contrast than the second case (Table 12).
be interpreted as an average value for the system.

Figure 19 - Pressure-transient responses simulated with skin


Figure 17 - Pressure-transient responses of dual-lateral wells with values from semi-log analysis
uniform and non-uniform skin
One of the main advantages of numerical models over
analytical models is the possibility to represent heterogeneous
systems. For the example considered in this section, analytical
models that use an average permeability to represent the
layered system may only work at late times. The two cases in
this section are compared with the analytical solution using an
average permeability calculated from the permeability values
used in the numerical model. These average permeability
values are shown in Table 12.

Figure 18 - Semi-log plot for skin analysis – non-uniform skin


case

Laterals with Different Length


In this section, the model is tested for laterals with
different length (810-ft and 1400-ft) using the data in Table 5.
Additionally, uniform and equal skin is also considered in this
case. The results are shown in Fig. 20. The simulated
responses match well with the analytical solution. It can be
seen that if one of the laterals has a shorter length, the pressure
drop is higher than that in the case with equal lengths because Figure 20 - Pressure-transient responses for different lateral
of the smaller contact area with the reservoir. As expected, the lengths with and without skin
case with skin shows a higher pressure drop.
It can be seen in Fig. 21 that for Case 1 with small
Heterogeneity permeability contrast between layers, the analytical pressure
In this case, a heterogeneous reservoir is considered using response is close to the numerical one. On the other hand, in
five layers and each lateral is placed in a different layer. The Case 2 for a larger permeability contrast between laterals, the
permeabilities are anisotropic and different for the layers analytical pressure response does not match the numerical
where the laterals are placed. Two cases were run (Fig. 21) to pressure response.
12 SPE 104581

TABLE 12 - PERMEABILITY DATA USED TO SIMULATE A Nomenclature


LAYERED RESERVOIR
b reciprocal formation volume factor,stb/rb..
General data
B formation volume factor,rb/stb..
Δx, ft 50 Imax 30 ct total compressibility, psi-1
Δy, ft 50 Jmax 30 c fluid compressibility, psi-1
Δz, ft 50 Kmax 5
cø porosity compressibility, psi-1
Xe, ft 1500 Lh1, ft 500
Ye, ft 1500 Lh2 ft 920
C wellbore storage coefficient, rb/psia.
Ze (h), ft 250 θ, degrees 45 D depth, ft.
Permeability data – numerical f friction factor.
Case 1 Case 2
h reservoir thickness, ft.
kx (Layer 1-5) 25, 25, 15, 10, 10 kx (Layer 1-5) 90, 90, 15, 10, 10 HF friction head, ft.
ky (Layer 1-5) 15, 15, 12, 10, 10 ky (Layer 1-5) 70, 70, 12, 10, 10 k permeability, md.
kz (Layer 1-5) 3, 3, 4, 5, 5 kz (Layer 1-5) 3, 3, 4, 5, 5 kw horizontal permeability in the normal direction of the
Permeability data - analytical well axis, md.
Case 1 Case 2 L reference length in the system, ft.
kx 17 kx 43 Lh horizontal well length, ft.
ky 12.4 ky 34.4 p pressure, psia.
kz 3.79 kz 3.79 pwf flowing bottom-hole pressure, psia.
q surface rate, stcf/d or stb/d
ro equivalent well-block radius, ft.
rw wellbore radius, ft.
s skin.
t time, hour.
T temperature, oF.
v velocity, ft/sec.
VR Rock volume, ft3
WI conventional well index, rcf/d/psi.
TWI transient well index, rcf/d/psi.
x distance in x-direction, ft.
y distance in y-direction, ft.
z distance in z-direction, ft.
zw horizontal well location, ft.

Greek Symbols
εD relative roughness of the well surface.
Figure 21 - Pressure-transient responses of dual-lateral wells in a γ density gradient, psi/ft.
layered reservoir φ porosity, fraction.
μ viscosity, cp.
Conclusions ρ density, lbm/ cf.
A numerical well model has been developed to accurately Δl diagonal horizontal length, ft.
simulate the pressure transient behavior of single- and dual-
Δx gridblock dimension in x-direction, ft.
lateral wells. The model equations can be easily implemented
in conventional simulators. The main conclusions are:
Δy gridblock dimension in y-direction, ft.
• The numerical well model is practical and produces
Δz gridblock dimension in z-direction, ft.
accurate pressure-transient responses of single- and Δt time step, hr.
dual-lateral wells without the need for grid
refinement around the well. Subscripts
D dimensionless.
• The representation of dual-lateral wells proposed in
eq equivalent.
this study is simple for easy implementation in
i x-direction.
different simulators for different applications.
j y-direction.
• The combination of analytical and numerical models
k z-direction.
used to calculate the well-block radius can be used as
n old time.
the foundation for more complex situations, such as
n+1 current time.
detailed well completions, and multilaterals with
t total.
branches.
x x-direction.
• The transmissibility modification presented in this y y-direction.
paper improves both the simulation of early-time z z –direction
pressure-transient responses, and the calculation of
the flow convergence around the well..
SPE 104581 13

Acknowledgment Ozkan, E.: “Analysis of Horizontal Well-Responses: Contemporary


This research was conducted at the Marathon Center of vs. Conventional,” SPE Reservoir Evaluation and Engineering
Excellence for Reservoir Studies (MCERS) at Colorado (Aug. 2001) 260.
Peaceman, D.W.: ”Interpretation of Well-Block Pressures in
School of Mines. We acknowledge the funding by Marathon
Numerical Reservoir Simulation,” Paper SPE 6893 presented at
Oil Company, Saudi Aramco and Repsol-YPF. the SPE-AIME 52nd Annual Fall Technical Conference and
Exhibition in Denver, October 9-12, 1977.
References Peaceman, D. W.: “Interpretation of Wellblock Pressures in
Al-Mohannadi, N., Ozkan, E., and Kazemi, H.: “Pressure-Transient Numerical Reservoir Simulation with Nonsquare Grid Blocks
Responses of Horizontal and Curved Wells in Anticlines and and Anisotropic Permeability,” SPE Journal, June 1983, 531.
Domes,” Paper SPE 84378 presented at the SPE Annual
Technical Conference and Exhibition, Denver, CO, Oct. 5 – 8,
Appendix
2003.
Al-Mohannadi, N., Ozkan, E., and Kazemi, H.: “Grid System
Requirements in Numerical Modeling of Pressure-Transient Finite-Difference Discretization
Tests in Horizontal Wells,” Paper SPE 92041 presented at the The continuity equation given by Eq. (8) for Cartesian
2004 SPE International Petroleum Conference in Mexico held in coordinates may be written by using finite-difference
Puebla, Mexico, Nov. 8–9 , 2004. discretization as:
Al-Mohannadi, N.: Simulation of Horizontal Well Tests, PhD
Dissertation, Colorado School of Mines, Golden, CO, 2004. ⎡ Δ xTx ( Δ x p − γΔ x D ) ⎤
⎢ ⎥
( )
Archer, R. and Yildiz, T.: “Transient Well Index for Numerical Well V
Test Analysis,” Paper SPE 71572 presented at the SPE Annual ⎢ +Δ yTy Δ y p − γΔ y D ⎥ − q = R Δ t (φ b ) , (A1)
⎢ ⎥ Δt
Technical Conference and Exhibition in New Orleans,
⎢⎣ +Δ z Tz ( Δ z p − γΔ z D ) ⎥⎦
Louisiana, 30 September-3 October, 2001.
Babu, D.K., Odeh, A.S. and Al-Khalifa,A.J.: “Numerical Simulatoin
of Horizontal Wells,” Paper SPE 21425 presented at SPE where:

( )
Middle East Oil Show in Bahrain, November 16-19, 1991. ⎡ p n +1 − p n +1 ⎤
Besson, J.: “Performance of Slanted and Horizontal Wells on an ⎢ i+1, j ,k ⎥
Δ xTx ( Δ x p − γΔ x D ) = Tx
i , j ,k
Anisotropic Medium,” Paper SPE 20965 presented at ⎢ ⎥
EUROPEC 90, The Hague, Netherlands, October 22-24, 1990.
i + 1 , j ,k
2
(
⎢⎣ −γ Di +1, j ,k − Di , j ,k ) ⎥⎦
Blanc, G., Ding, D.Y., Ene, A., Estebenet, T. and Rahon, D.:
“Transient Productivity Index for Numerical Well Test
Simulations,” in R. Schatzinger and J. Jordan, eds., Reservoir −Tx
(
⎡ p n +1 − p n +1
⎢ i , j ,k )
i −1, j ,k


⎢ ⎥
Characterization-Recent Advances, AAPG Memoir 71, 1999, p
163.
i − 1 , j ,k
2
⎣⎢ (
−γ Di , j , k − Di −1, j ,k ) ⎦⎥
Clonts, M. D. and Ramey, H. J., Jr.: “Pressure Transient Analysis for (A2)
Wells with Horizontal Drainholes,” Paper SPE 15116 presented
at the 56th California Regional Meeting, Oakland, CA, April 2-
4, 1986.
( )
Δ yTy Δ y p − γΔ y D = Ty
(
⎡ p n +1 − p n +1
⎢ i , j +1,k )
i , j ,k


Ding, Y. and Renard, G.: “A New Representation of Wells in ⎢ ⎥
Numerical Reservoir Simulation,” SPE Reservoir Engineering,
i , j + 1 ,k
2
⎣⎢ (
−γ Di , j +1,k − Di , j ,k ) ⎦⎥
May 1994, p 140.
Ding, Y., Renard, G., and Weill, L.: “Representation of Wells in
Numerical Reservoir Simulation,” Paper SPE 29123 presented −Ty
(
⎡ p n +1 − p n +1
⎢ i , j ,k )
i , j −1,k


⎢ ⎥
at the 13th SPE Symposium on Reservoir Simulation held in San
Antonio, February 12-15, 1998.
i , j − 1 ,k
2
⎣⎢ (
−γ Di , j , k − Di , j −1,k ) ⎦⎥
Ding, Y. and Jeannin, L.: “New Numerical Schemes for the Near- (A3)
Well Modelling with Discretization Around the Wellbore
Boundary Using Flexible Grids,” Paper SPE 66360 presented at
the SPE Reservoir Simulation Symposium in Houston, TX,
Δ z Tz ( Δ z p − γΔ z D ) = Tz
(
⎡ p n +1 − p n +1
⎢ i , j ,k +1 )
i , j ,k


⎢ ⎥
February 11-14, 2001.
Goktas, B. and Ertekin, T.: “Implementation of a Local Grid
i , j ,k + 1
2
⎣⎢ (
−γ Di , j ,k +1 − Di , j , k ) ⎦⎥

( )
Refinement Technique in Modeling Slanted, Undulating
Horizontal and Multi-Lateral Wells,” Paper SPE 56624
⎡ p n +1 − p n +1 ⎤
−Ty ⎢ i , j ,k i , j ,k −1 ⎥
presented at the SPE Annual Technical Conference and ⎢ ⎥
Exhibition in Houston, Texas, Octuber 3-6 1999.
i , j ,k − 1
2
⎣⎢ (
−γ Di , j ,k − Di , j ,k −1 ) ⎦⎥
Ozkan, E., Raghavan, R., and Joshi, S. D.: “Horizontal Well Pressure
Analysis,” SPE Formation Evaluation December, 1989, p567. (A4)
Ozkan, E. and Raghavan, R.: “New Solutions for Well-Test-Analysis The right hand side is defined as:
Problems: Part 2 - Computational Considerations and
Applications,” SPE Formation Evaluation September 1991, p
369.
Ozkan, E., Yildiz, T., and Kuchuk, F. J.: “Transient Pressure
Behavior of Dual-lateral Wells,” Paper SPE 38670 presented at
SPE Annual Technical Conference and Exhibition in San
Antonio, TX, October 5-8, 1997
14 SPE 104581

⎡ n⎛ Δb⎞ ⎤ ⎛ k ⎞
n
Δxi , j ,k Δyi , j ,k
⎢φ ⎜ t ⎟ ⎥ Li , j , k = Tzn = 0.006328 ⎜ z ⎟
⎢ ⎜⎝ Δ t p ⎟⎠ ⎥ i , j ,k + 1
2 ⎝ μ B ⎠i , j ,k + 12 Δzi , j ,k + 12
Δ t (φ b) = φ n Δ t b + b n +1Δ tφ = ⎢ ⎥ Δ t pi , j ,k
⎢ n +1 ⎛ Δt φ ⎞ ⎥ (A14)
⎢ +b ⎜⎜ Δ p ⎟⎟ ⎥
⎢⎣ ⎝ t ⎠ ⎥⎦ i, j ,k
{
Ei , j ,k = − Ai , j ,k + Bi , j , k + Di , j , k + Fi , j , k + H i , j , k + Li , j , k + WI in, j , k

(
 φ bct Δ t pi , j , k  φ bct p in, +j ,k1 − p in, j ,k ) +24
VRi , j , k ⎡ n ⎛ Δ t b ⎞ n +1 ⎛ Δ tφ ⎞ ⎤
⎢φ ⎜ ⎟ +b ⎜ ⎟ ⎥



(A5) Δt ⎢⎣ ⎜⎝ Δt p ⎟⎠ ⎜ Δt p ⎟ ⎥
⎝ ⎠ ⎦ i , j , k ⎪⎭
And the rock volume is: (A15)
VR i , j ,k = Δx i , j ,k Δy i , j ,k Δz i , j ,k (A6) And the right hand side is given by:

The flow rate is given by: ⎧ V ⎫


⎪ Ri , j ,k ⎡ n ⎛ Δ t b ⎞ ⎛ Δ φ ⎞⎤ ⎪ n
n +1
q i , j ,k = WI in, j ,k (p n +1
i , j ,k
n +1
− pwell i , j ,k ) (A7) Ri , j ,k = − ⎨24
⎪⎩ Δ t
⎢φ ⎜
⎜ Δ
⎣⎢ ⎝ t ⎠ p


+ b n +1 ⎜ t ⎟ ⎥
⎜ Δ ⎟ ⎬ pi , j ,k
⎝ t ⎠ ⎦⎥ i , j ,k ⎭⎪
p

where WI is the well index and has two expressions for steady-
state and transient period. The corresponding equations are
⎡Txn γ n 1 (
D
⎢ i+ 12 , j ,k i + 2 , j , k i +1, j , k
− Di , j ,k ) ⎤

given in the well representation section. +⎢ ⎥
Eq. (A1) can be rearranged in terms of transmissibility ⎣ 2
(
⎢ −Txi− 1 , j ,k γ i − 12 , j ,k Di , j ,k − Di −1, j , k
n n
) ⎥

coefficients and pressure, known as the node pressure
equation: ⎡Tyn γn 1 D (
⎢ i , j+ 12 ,k i , j + 2 ,k i , j +1,k
− Di , j ,k ) ⎤

+⎢ ⎥
Ai , j ,k pin, +j ,1k −1 + Bi , j ,k pin, +j −11, k + Di , j , k pin−+1,1 j , k

n n
(
⎢ −Tyi , j − 1 ,k γ i , j − 12 , k Di , j ,k − Di , j −1, k ) ⎥

+ Ei , j ,k pin, +j ,1k
2

+ Fi , j ,k pin++1,1 j ,k + H i , j ,k pin, +j +11, k + Li , j , k pin, +j ,1k +1


(A8)
⎢ i , j ,k + 12 i , j ,k + 2
(
⎡Tzn γ n 1 Di , j ,k − Di , j ,k +1 ) ⎤

+⎢ ⎥
+WI i , j , k pwelli , j ,k = Ri , j ,k
⎣ 2
n
(
⎢ −Tzi , j ,k − 1 γ i , j ,k − 12 Di , j , k − Di , j , k −1
n
) ⎥

where the transmissibility coefficients are expressed as:
2 ⎣ ( )
+WI in, j ,k γ in+ 1 , j ,k ⎡ Di , j , k − Di +1, j , k − HFi + 1 , j ,k ⎤ .
2 ⎦
Δxi , j ,k Δyi , j , k
n
⎛ k ⎞ (A16)
Ai , j ,k = Tzn = 0.006328 ⎜ z ⎟
i , j ,k − 1
2 ⎝ μ B ⎠i , j , k − 12 Δzi , j , k − 12
(A9)

n
⎛ ky ⎞ Δxi , j ,k Δzi , j , k
Bi , j ,k = Tyn = 0.006328 ⎜⎜ ⎟⎟
i , j − 1 ,k
2 ⎝ μ B ⎠i , j − 12 ,k Δyi , j − 12 ,k
(A10)

⎛ k ⎞ Δyi , j ,k Δzi , j , k
Di , j ,k = Txn 1 = 0.006328 ⎜ x ⎟
i − , j ,k
2 ⎝ μ B ⎠i − 12 , j , k Δxi − 12 , j , k
(A11)

Δyi , j ,k Δzi , j , k
n
⎛ k ⎞
Fi , j ,k = Txn 1 = 0.006328 ⎜ x ⎟
i + , j ,k
2 ⎝ μ B ⎠i + 12 , j ,k Δxi + 12 , j ,k
(A12)

n
⎛ ky ⎞ Δxi , j ,k Δzi , j , k
H i , j ,k = Tyn = 0.006328 ⎜⎜ ⎟⎟
i , j + 1 ,k
2 ⎝ μ B ⎠i , j + 12 , k Δyi , j + 12 , k
(A13)

Vous aimerez peut-être aussi