Journal of Petroleum Science and Engineering 82–83 (2012) 174–186

Contents lists available at SciVerse ScienceDirect

Journal of Petroleum Science and Engineering

journal homepage: www.elsevier.com/locate/petrol

Review of well models and assessment of their impacts on numerical reservoir

simulation performance
Francis A. Dumkwu ⁎, Akand W. Islam, Eric S. Carlson
Department of Chemical & Biological Engineering, The University of Alabama, Tuscaloosa, AL, 35487-0203, USA

a r t i c l e i n f o a b s t r a c t

Article history: Reservoir simulators have long become primary reservoir engineering tools for simulating fluid flow in porous
Received 15 June 2011 media, recovery analyses and performance evaluation in the oil and gas industry. This review focuses on recent
Accepted 2 December 2011 advancement in the development and representation of conventional and advanced well models used in
Available online 13 December 2011
reservoir simulators for simulating fluid flow in porous media since the last review on the subject (Demetri
and Farouq, 1993). Well models have been developed to different degrees of sophistication resulting in signifi-
Keywords:
conventional well model
cant positive impact on numerical reservoir simulation performance. In spite of the complexities of these
advanced well model developed well models and the successes recorded with them in predicting well and reservoir performances,
sink some predictions made with them in some cases do not match actual field performances. The major factors
source contributing to these observed well models' limitations are emphasized and future perspectives on the de-
reservoir simulator velopment of next generation well models which are suitable for more efficient, comprehensive, faster and
highly scalable next generation reservoir simulators with enhanced computing capabilities are advised.
© 2011 Elsevier B.V. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
2. Representation of conventional wells in numerical reservoir simulators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
2.1. Default well model representation in a numerical simulator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
2.2. Non-default conventional well model representation in a numerical simulator . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
2.3. Limitation of conventional well models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
3. Representation of advanced wells in reservoir simulator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
3.1. The limitations of advanced well models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
4. The next generation well models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
5. Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
Nomenclature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185

1. Introduction
Williamson and Chappelear (1981) defined a well model as an an-
alytical model used in reservoir simulation system to represent fluid
flow within a grid block as it enters or leaves a well. In reservoir sim-
ulation, the basic objective of a well model is to supply source and
sink terms to the reservoir model (Holmes, 2001).
⁎ Corresponding author.
E-mail address: fadumkwu@crimson.ua.edu (F.A. Dumkwu).
It has long been noted by researchers that unless a well is centrally-
positioned in a cylindrical coordinate, it was unfeasible to represent
it with an internal boundary. This is due to the fact that the ratio of
wellbore radius to the expected grid-block length is large. As noted
by Settari and Aziz (1974), if the size of the grid block is far larger
than the size of the well, the pressure of the grid block computed
by the reservoir will not be a good approximation of the well pressure.
Williamson and Chappelear (1981) stated that even when it was
practicable to use very fine grids to represent a well as an internal
boundary, other factors, such as partial perforation, partial penetration
and skin which are relevant to local flow needed to be considered. In
light of this development, an alternative technique for representing
wells was developed. In this new technique, wells are approximately

0920-4105/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.petrol.2011.12.005
 F.A. Dumkwu et al. / Journal of Petroleum Science and Engineering 82–83 (2012) 174–186
represented as a source function in a simulator. The idea to represent a
well as a source in a simulator was first proposed by Schwabe and Brand
(1967). According to Williamson and Chappelear (1981), “The relation-
ship between the strength of this source, the wellbore flow, and the
flow in the surrounding grid blocks comprises an essential part of the
well model. The source representation of a well can be described as a
local, approximate, steady and singular solution of fluid flow equa-
tions”. Sandface pressure and flow rate are typical boundary conditions
for a well model.
A Numerical Reservoir Simulator has long become a primary res-
ervoir engineering tool used by Reservoir engineers to conduct reser-
voir simulations which are useful for making large capital decisions,
estimate reserves, and to diagnose and improve on the performance
of producing reservoirs in the oil and gas industry. The reservoir is
represented by gridblocks on the simulator. A numerical simulator
is deployed for the computation of pressure on a grid block which is
representative of the average pressure of the surrounding gridblocks.
A well model is a key component in any reservoir simulator as it sup-
plies a set of realistic boundary conditions for the computation of
pressure distribution in a reservoir. For the modeling of a well in a
reservoir simulator, a well model is required to relate the well flow
rate to the well pressure and the gridblock pressure containing the
well. Two types of well models are currently in use in numerical sim-
ulators for the simulation of petroleum reservoirs, namely conven-
tional and advanced well models.
Conventional well models treat the wellbore as a single unit with-
in the model. It is used to provide a detailed analysis of fluid flow in
standard vertical and deviated wells. Conventional well models are
developed to analyze well pressure responses based on certain funda-
mental concepts (Erdal Ozkan, 2001). A conventional well model does
not give acceptable accuracy when used to model advanced well
models since it does not accurately represent all physical processes
taking place in the wellbore (Holmes et al., 1998; Holmes, 2001;
Augustine, 2002).
The advent of horizontal and multilateral drilling technology as
well as complex wellbore configurations due to the implementation
of intelligent completion systems has necessitated the development
of increasing complex well models for use in reservoir simulation. Ad-
vanced wells, such as horizontal, multilateral and smart wells are
characterized by complex geometries, increased length of perforated
intervals and presence of control devices downhole in smart wells.
More so, advanced well models are formulated to correct the limita-
tions of conventional well models through a detailed representation
of wellbore composition, rate, and pressure distribution. Unfortunate-
ly, despite the sophistication of these advanced well models and the
successes recorded with them, some predictions made with them in
some cases do not match the actual performance. This limitation is
caused by the uncertainties inherent in the assumptions made during
their formulations (Aziz et al., 1999; Mullane et al., 1996).
In order for the reservoir simulator to provide accurate results,
both conventional and advanced well models must be able to predict
well performance accurately over the life time of the reservoir. This
review focuses on the major milestones recorded in the development
of conventional and advanced well models used in numerical simula-
tors for conducting reservoir simulation, the major limitations of
these models, assessment of their impacts on reservoir simulation
performance, and future perspectives on the development of next
generation well models which are suitable for more efficient, compre-
hensive, faster and highly scalable next generation reservoir simula-
tors with enhanced computing capabilities.
2. Representation of conventional wells in numerical reservoir
simulators
In numerical reservoir simulation, the reservoir is represented by
gridblocks on the simulator. Thus, the simulator is an invaluable tool
175
for the computation of pressure on a grid block which is representative
of the average pressure of the surrounding gridblocks. Well models are
used to relate the rate of flow of fluid to or from a well to the difference
between the wellblock pressure and the flowing wellbore pressure.
During the formulation of conventional well models certain assump-
tions are made based on certain fundamental concepts (Erdal Ozkan,
2001). Firstly, a conventional well is assumed to be a line-source and
an infinite conductivity wellbore. This assumption is justified on the
basis that the magnitude of the pressure drop in the wellbore is negligi-
bly small compared to the magnitude of the pressure drop in the reser-
voir. Secondly, a conventional well is assumed to contain a single lateral
withdrawing fluid along its entire length and thirdly, a well is assumed
to have a skin region that is uniformly distributed along the well.
2.1. Default well model representation in a numerical simulator
Peaceman (1978) showed that the grid pressures obtained in the
numerical simulation of a single-phase flow into a single well com-
pleted at the center of a square gridblock are equal to the actual flow-
ing pressure at a radius of 0.2 Δx. This radius is called the equivalent
wellblock radius and it is the radius at which the steady state flowing
pressure for the actual well is equal to the numerically calculated
pressure for the well block. In terms of the equivalent radius the
steady state pressure distribution around a well, po, and the well
flow rate, q, are given by the expression:
2πkh po −pwf
q¼
: ð1Þ
μ ln ro =
rw
This model assumes a radial flow around the well so that a loga-
rithmic relation is satisfied between the wellblock pressure po and
wellbore pressure pwf. Eq. (1) is Peaceman's well model and it is usu-
ally the default well model in a general reservoir simulator.
For wells with different geometric orientation in the gridblock, the
value of the equivalent radius changes from that of the default value
shown in Eq. (1). A new equivalent radius, ro could be derived and
substituted into Eq. (1) taking into consideration well geometries,
the location of the well within the grid block, the number of wells lo-
cated in a grid block, the degree of well penetration, the nature of the
permeability of the porous media in which the well is located, and
multiple wells flowing with arbitrary rates (Peaceman, 1982, 1983,
1987, 1990; Shiralkar, 1988, 1989).
2.2. Non-default conventional well model representation in a numerical
simulator
Non-default well model is any well model with a value of equiva-
lent radius different from Peaceman's default value given in Eq. (1).
Well models are represented as a source and sink term in a simulator
and the relationship between them can be characterized by an inflow-
performance relationship (Chappelear and Williamson, 1981;
Williamson and Chappelear, 1981). This inflow performance relation-
ship holds true for a well completed in more than one grid cell. A typical
conventional well model used in a simulator usually has a well rate re-
lationship (Fung et al., 2005) for any hydrocarbon specie i in mol/day
given by:


X
nly
Qi ¼ WIk λ̄o ρo xi þ λ̄g ρg yi ðpk −pbh −ψwell dzk Þ: ð2Þ
k¼1
Eq. (2) portrays the inflow rate for each fluid as proportional to the
drawdown, the fluid mobility , and well index or connection transmis-
sibility factor (WI). Well index accounts for well pressure losses within
the grid cell due to radial inflow into the well (Peaceman, 1983). Fur-
thermore, as noted by Wolfsteiner et al. (2001) well index accounts
176 F.A. Dumkwu et al. / Journal of Petroleum Science and Engineering 82–83 (2012) 174–186

for the geometry of the gridblock, location and orientation of the well for non-uniform grids for homogeneous, anisotropic reservoirs or
segment in the gridblock, and rock properties. The wellbore pressures completely heterogeneous reservoirs for a partially penetrating
at each grid-cell completion are related by an approximate hydrostatic vertical well.
pressure difference in the wellbore. Typically, the hydrostatic pressure Peaceman (1995) developed an equation that quantitatively
is calculated from the average density of the fluid mixture in the well- addressed the question of how the value of equivalent radius is influ-
bore or more precisely from the local density of the fluid mixture in enced by the interaction between wells completed in separate well-
each section of the wellbore between adjacent grid-cell completions. blocks flowing with arbitrary well rates. The equivalent wellblock
For this simple well model, a single variable analogous to the well's bot- radius of each well, m, in a reservoir with nw wells is given by the ex-
tomhole pressure indicates the state of the well and for any given set of pression:
reservoir conditions, the bottomhole pressure constraint equation,
!qk =
which exactly preserves the well's target rate, can be obtained easily. r eq;m
2¼
r eq;mk qm

If a well is constrained to operate at a given oil-rate target, the sum of ð1Þ

¼ ∏ ð1Þ
m; k ¼ 1; 2; …; nw : ð6Þ
r eq;m k≠m r eq;m
the oil inflow rates from each completed grid-cell must equal the target
oil rate (Holmes, 2001).
Since the last review on the subject (Demetre and Ali, 1994), ap- Eq. (6) takes into account the interaction among all the wells and
preciable progress has been made with respect to the development it is valid both for steady state and pseudosteady state conditions.
and representation of well models for reservoir simulation. Chen et Peaceman (2003) presented a general equation for calculating the
al. (1995) presented analytical models which can be coupled to a con- well index (WI) of each well, m, in a reservoir with nw wells flowing
ventional finite difference simulator for the accurate computation of with various rates. This model takes into account arbitrary wellblock
flowing bottomhole pressure for vertical, horizontal and slanted wells flow rates and the interaction between wellblocks. Thus an accurate
situated in an infinite slab, homogeneous or anisotropic uniform per- well index can be computed for each wellblock at each timestep, irre-
meability reservoir with a single phase fluid flowing through it. The spective of whether or not the well rates vary with time. For well m,
analytical expression for calculating flowing bottomhole pressure for a WI can be calculated from the equation:
well is given by: !
1 1 q 1 1

  ¼ þ ∑ k ð2¼Þ
− ð1Þ
m; k ¼ 1; 2; …; nw ð7Þ
ro WIm WI1m k≠m qm WI mk WIm
pwf Anal ¼ po −C:μ:B:q: ln r
=ðk:hÞ: ð3Þ
w

Where nw is the total number of wells in all the blocks, k is the

The conventional finite difference simulator numerically computes
flowing bottomhole pressure for a well from the expression:
2
  3 well index of well m in the presence of only one other well, well k,
ro
ln þs
pwf Numl ¼ po −C:μ:B:q:4
rw
5:
k:h
A pseudo skin factor, s, can be computed by substituting the ana-
lytical and numerical flowing bottomhole pressures obtained from
Eqs. (3) and (4), respectively, into Eq. (5).


s ¼ k:h: pwf Anal −pwf Num =ðC:μ:B:qÞ:
The pseudo skin factor is used to account for the well deviation
angle and the effect of partial penetration of wells.
Lin (1995) developed new models for partially penetrating verti-
cal wells with any number of perforated intervals in isotropic, aniso-
tropic, layered, or 3D, in heterogeneous reservoirs using non-uniform
grids. These models have capability for accurately simulating non-
uniform flux distributions along perforated intervals. As stated by
Lin, these models generate a constant well index for each perforated
layer based on the quasi-steady state conditions earlier established
for unsteady state flows. The well indices generated depend on the
permeability distribution in the neighborhood of the well, reservoir
thickness, wellbore radius, and the locations of perforated intervals.
For a heterogeneous reservoir, it was further observed that the calcu-
lated well indices and the rate distribution along the well were greatly
controlled by the subgrid permeability distribution near the wellbore.
These models were utilized in a coarse-grid simulation and the gener-
ated fluxes were agreeable (within 1% error) to that obtained using
index for all the other wells, qm is the rate of well m, qk is the rate of
well k, WIm is the well index of well m acting alone and WImk is the
ð4Þ producing at the same rate.
Peaceman (1990) and Su (1995) developed analytical models for
representing off-center wells. Agaev (1996) presented an improved
technique for modeling off-center well in multidimensional reservoir
simulation. With this improved technique, the exact position of off-
center well can be located in a gridblock. The model also accounts
for the influence of neighboring gridblocks. This new technique helps
to improve the accuracy of simulation results and reduces the com-
ð5Þ
puting time required for full field reservoir simulation using coarse
grids.
In reservoir simulation using finite difference scheme, numerical
productivity indices (PI) are used to relate wellblock, wellbore pres-
sures and the flowrate of a well. This technique was based on the
concept of equivalent wellblock radius and this model works fine
for uniform grids. However, for non-uniform grids the equivalent
wellblock radius may result in error when computing wellbore pres-
sure or oil flowrate. Ding and Renard (1994) developed a well model
based on finite volume approximation in which the analytical solu-
tion for near-well pressure is included via modified transmissibilities
between gridblocks so that flow around a vertical well is completely
described. This new technique does not require an equivalent well-
block radius but rely on accurate computation of new interblock
distances (Fig. 1.) which are helpful in improving modeling of flow in
proximity of the well. The interblock distance Leq, 1 is given by the
expression
Δx 1
=
y ln rw
2

fine-grid solutions. Dogru Ali (2010a) developed a new analytical ex- Leq;1 ¼ o : ð8Þ
Δy
2 arctg o
pression for calculating the equivalent well radius in a homogeneous, Δxo
anisotropic reservoir with a uniform square grid around the well path.
This new expression has a similar structure as that of Peaceman's For a square gridblock, the interblock distance is obtained from
equation, but with a new parameter introduced to account for partial the relation
penetration and for vertical flow. Furthermore, Dogru Ali (2010b)    
presented a new analytical expression and a simple numerical pro- 2 Δx
Leq;1 ¼ Δx ln : ð9Þ
cedure for calculating the equivalent well radius and well index π rw
 F.A. Dumkwu et al. / Journal of Petroleum Science and Engineering 82–83 (2012) 174–186 177

The wellblock pressure Po calculated from the equivalent trans-

missibilities equation corresponds to the equivalent wellblock radius
which is involved in transmissibility calculations. The relationship be-
tween the computed wellblock pressure and the wellbore pressure
can be established through the conventional productivity index equa-
tion given in Eq. (11). This method allows improvement to be made
on the accuracy of both well performance and the numerical scheme.
It is also applicable to any 3D well geometry, including multilateral
wells located in any type of grid.
Ding et al. (1998a, 1998b) improved on the well model developed
by Ding and Renard (1994) and Ding et al. (1998a) by modifying the
transmissibilities in order to make it suitable for modeling complex
wells with distorted gridblocks. He proposed a control volume meth-
od which can accurately handle both the well performance and the
flux calculation in proximity of the well.
For commonly used flexible grids in reservoir simulations, it has
Fig. 1. Wellblock pressure calculation.
been observed that the gridblocks in the vicinity of the well are usu-
ally significantly large compared to the wellbore radius, especially in
field applications. Furthermore, the gridblock sizes increases geomet-
For anisotropic systems, interblock distance is given by: rically in the radial direction outwards from the well (Aavatsmark et
al., 1998a, 1998b; Edwards and Rogers, 1994; Fung et al., 1991;
Δx 1
=2 Heinemann et al., 1989;). To improve on the standard linear approach
pffiffiffiffiffiffiffiffi
ffi
y qffiffiffiffiffiffiffiffiffiffi
 ln
r w ð1þ kx =ky Þ=2
Leq;1 ¼ o ky k : ð10Þ for the near-well modeling of vertical wells using flexible grids, Ding
1=2
2 x
arctg Δyo k1=2
x and Jeannin (2001) proposed control-volume technique based on
Δxo ky
logarithmic approach. He presented a multi-point and two-point
flux approximation schemes which can help reduce the flux trunca-
This new technique is particularly suited for use in both non- tion errors for the point source “singular” solution calculation, thus
uniform grids and non-isolated wells. Ding et al. (1998a) extended enhancing the accuracy of the fluid flow modeling in the near-well
the previous technique to include off-center wells and flexible grids. region.
The modified model allows the use of productivity index and equivalent The accurate prediction of well deliverability using appropriate well
wellblock radius. In a typical reservoir simulator, the PI is given by models in reservoir simulators strongly depends on how the near-well
flow restrictions, such as mechanical skin, partial completion, altered
q 2πkh 1
PI ¼ ¼
: ð11Þ formation caused by residual completion fluid or liquid dropout, and
po −pw μ ln ro =
rw non-Darcy flow effects are represented as well as how they are per-
mitted to interact with each other in the reservoir simulation model.
Wellblock transmissibilities in conventional models are given by The conventional method computes the effective total skin as a lump

t
sum of individual components without accounting for the effect of non-
Δyo linear skin interactions among near-well flow restrictions. The skin in-
T i ¼ kh ði ¼ 1; 3Þ teractions are nonlinear and often multiplicative (Hwang and Odeh,
Δx1=2
: ð12Þ 1995); therefore, the effective total skin is much larger than that com-
Δxo
T i ¼ kh ði ¼ 2; 4Þ puted through conventional method. Hwang (2000) proposed total
Δy1=2
skin equations for 1D and 2D reservoirs which accurately represent
nonlinear interactions among the near-well flow restrictions. These
In the modified well model it is possible to obtain the pressure po
equations are used to quantify the effect of different well models on
corresponding to the equivalent wellblock radius ro through the use
productivity index and the obtained results can be compared with
of equivalent wellblock transmissibilities which relates po to the pres-
that obtained from direct simulation. The proposed effective total skin
sures of adjacent blocks through equivalent interblock distanceLeq, i.
factor ST for a vertical wellbore in a 1D radial homogeneous reservoir

t
is given by the algebraic equation:
Δyo
T eq;i ¼ kh ði ¼ 1; 3Þ
Leq;i
ð13Þ ! ( ) !2
Δx XI
T eq;i ¼ kh o ði ¼ 2; 4Þ 1 h S h
Leq;i ST ¼ Sp þ Sa þ m þ 1 þ Dr ðiÞ Sd ð16Þ
γ hp krgð1Þ i¼1
hp

The equivalent interblock distance, Leq, 1 is related to the equiva-

n

o
lent wellblock radius through the equation: Sa ¼ ∑ 1=krg ðiÞ−1 ln r i =r i−1 ð17Þ

 i
Δx
ln r1=2
Leq;1 ¼ Δyo ð14Þ
o

2arctg ðΔyo =Δxo Þ Dr ðiÞ ¼ fIðiÞ=I−1gr w ð1=r i−1 −1=r i Þ ð18Þ

Where Δyo and Δxo are the wellblock dimensions. The new tech-
nique can be implemented easily in a conventional reservoir simula-
tor by multiplying the conventional wellblock transmissibilities
given in Eq. (12) by constant factors Ri given by the relation
T eq;i
Ri ¼ : ð15Þ
Ti
re = rw, and the effect of the converging flow stream on non-Darcy flow
.
is ignored. Moreover, γ indirectly accounts for the effect of kV ratio
kH
on ST and it is often ignored by assuming γ = 1; where γ is an adjusting
factor with γ ≥ 1, h is total formation thickness, Dr is relative non-Darcy
skin coefficient, S is skin factor, and krg is gas relative permeability.
178 F.A. Dumkwu et al. / Journal of Petroleum Science and Engineering 82–83 (2012) 174–186

For 2D multilayer cases with crossflow, the effective total skin equation remains the same even when the well operates under a va-
factor ST is computed from the equation: riety of possible constraints.
2 3 There is currently an increasing quest for faster, more efficient,
more comprehensive, and highly scalable next generation reservoir
6 ∑ ðkhÞi 7
ST ¼ Sp þ 6

 open 7 ð19Þ simulators with enhanced computing abilities. A high resolution res-
=
4 Sm ðiÞ
5
re
∑open ðkhÞi ln r −0:75 þ δSa ðiÞ þ k1 ðiÞ þ ωSd ðiÞ ervoir simulator is vital for better definition and description of fluid
w rg
flow in a reservoir for improved field development. Such a new scalable
  parallel reservoir simulator requires a comprehensively formulated
R
− ln þ 0:75 well model which, among others, can handle global phase-component
rw
partitioning. In such a system, provisions should be made for multiple
1 components and phases to exist as well as allowing any component to
where i run for all open layers, krg (i) is the r-direction gas permeabil-
ity in cell 1 of the layer. exist in any phase (DeBaun et al., 2005; Guyaguler and Ghorayeb,

 2006; Holmes et al., 2010).
1
δ¼ ∑ ðkhÞi ∑ ðkhÞj
γ all open = ð20Þ
2.3. Limitation of conventional well models
( )
2
X
I Before the development of advanced well models specific for
ω¼ 1þ
i¼1
Dr ðiÞ
all
=
∑ ðkhÞi ∑ ðkhÞj
open
ð21Þ modeling horizontal, multilateral and smart wells, conventional well
models were used to provide the source and sink functions for nu-
merical simulators. However, it was observed that these models did
I is the number of altered zones in all radial direction in the reser-
not give accurate simulation results in view of the fact that they were
voir, and (kh)i is the layer flow capacity. Moreover, the contribution of
not comprehensive enough to adequately represent the physics of the
each layer's mechanical skin Sm(i) is appreciably reduced via the har-
processes taking place in these wells (Holmes et al., 1998; Holmes,
monic averaging process.
2001; Augustine, 2002).
A number of techniques developed for solving pressure equation in
This inherent limitations suffered by conventional well models
oil reservoir models containing a slew of wells require computational
are caused by two major factors. First, the pressure gradient within
meshes that conform to the wells. Baumann et al. (1999) developed a
the wellbore at formation level is limited to an approximate hydro-
well model which accurately computes the pressure distributions void
static head computation, thus ignoring pressure loss due to friction.
of mesh clustering around the wells. This well model is based on a
In the same vein, the effect of slippage between the wellbore fluids
generalized finite element method which has no special basis func-
is neglected in the calculation of hydrostatic head; thus, the computed
tion. Thus, the new discretization technique takes into account all the
hydrostatic gradient based on an assumed homogeneous density of
important parameters that characterize the near-wellbore pressure
the fluid mixture in the wellbore is not accurate. The assumption of a
distribution. This model was validated through a pressure maintenance
homogeneous density for the fluid mixture in the wellbore is a wrong
study conducted on Kern River field, Texaco, containing more than 6000
one considering the fact that individual fluid that makes up the mixture
wells. The result obtained from this study using this proposed technique
flow with different velocities.
agreed with that obtained from analytical method.
Second, there is a high tendency for crossflow to occur when a well,
Reservoir simulation numerical scheme can be targeted for paral-
completed in two or more poorly communicating regions of the forma-
lelism to ensure excellent simulator scalability to a large number of
tion, has appreciably differing pressures. Cross flow is expected to occur
processors. In parallel computation utilizing many processors, grid
in this producing well in view of the fact that some fluid flows into the
cells are partitioned into various processing units, thus a single well
well from the high-pressure regions, and others flow back out into the
can traverse several grid cells accommodated in multiple processors
low-pressure regions. Despite the occurrence of crossflow, a single set
where information regarding well inflow must be assembled to con-
of variables is still used in conventional well models to represent
struct the well constraint equations. Therefore, it is essential that the
the entire contents of the wellbore within the formation. This prac-
modeling of wells for such parallel computing be done in an accurate
tice stems from the assumption of a completely mixed fluid in the
and robust manner. In mega-cell simulation, wells can be perforated
wellbore. This assumption is not accurate since the fluid mixture
in sundry layers referred to as well cells whose individual inflow contri-
flowing back into the formation does not represent the average con-
butions make up the total target rate. More so, the layers productivity
tent of the wellbore. To ensure accurate modeling of crossflow, the
indices can vary widely and the well cells communicate via the wellbore
simulator must calculate both the reinjection rate and the composi-
pressure conditions. For a time step to converge, the material balance
tion of the injected fluid mixture. The injected fluid composition
equations for all grid cells and the well constraint equations active at
should reflect that of the fluid mixture in the wellbore, and this
that specific time step must be satisfied. Fung et al. (2005) developed
would depend on the inflow into the producing completions and
a set of well constraint equations in the context of parallel mega cell res-
the rate of injection from the surface for injection wells.
ervoir simulation. He presented explicit well model, semi-implicit well
model as well as Fully-Implicit Fully-Coupled well model which is the
3. Representation of advanced wells in reservoir simulator
favorite well model used in majority of modern implicit reservoir
simulators.
The development of complex wellbore configurations due to the
Fully-Implicit Fully-Coupled well model for the oil rate constraint
implementation of intelligent completion systems coupled with the
is given by the expression:
advent of horizontal and multilateral drilling technology necessitated

 the development of increasing complex well models for use in reser-
X
nly
nþ1 nþ1 nþ1
Qo ¼ WI i λi;o pi −pbh −ψwell dzi : ð22Þ voir simulation. Advanced wells, such as horizontal, multilateral and
i¼1 smart wells are characterized by complex geometries, increased
length of perforated intervals and presence of control devices down-
This well model is solved simultaneously with other material bal- hole in smart wells. Furthermore, it is formulated to accommodate
ance equations to obtain well bottom-hole pressure, Pbh. The well gra- the incorporation of a detailed representation of wellbore composi-
dient ψwell is computed at the initial time step and is kept constant tion, rate, and pressure distribution. Complex wells often contain a
over the Newtonian iterations. The basic form of the well constraint mixture of complex devices, such as flow and inflow control, as well
 F.A. Dumkwu et al. / Journal of Petroleum Science and Engineering 82–83 (2012) 174–186 179

as pumps, which help to increase sweep efficiency of reservoir fluids, productivity with even greater formation exposure. Dittoe et al. (1996)
improve well cleanup and delay water or gas breakthrough into the gave very simple definitions of horizontal, vertical and multi-lateral
producing well. Complex well models are formulated to accomplish models as shown in Fig. 2 (a, b, c). Ouyang and Aziz (1998) first intro-
the following objectives: duced a popular arbitrary configuration and trajectory of multi-lateral
wells as shown in Fig. 2 (d) which later were used in modeling investiga-
(1) To accurately model multilaterals wells alongside their comin- tions (Wolfsteiner et al., 2000).
gling branch flows. Giger et al. (1984) presented the inflow performance of an open
(2) To provide an accurate description of multiphase flow in a hole horizontal well which was expanded by Joshi (1988). Joshi's so-
wellbore. lution is expressed as:
(3) To account for pressure drop due to friction, hydrostatic, and
2πkH h△p
acceleration as well as velocity difference between the phases qH ¼
 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
 ð23Þ
due to slip. aþ a2 −ðL=2Þ2 Iani h I ani h
μB ln L=2 þ L ln 2r
(4) To simulate flow control devices downhole of the well.
w

Horizontal and multi-lateral wells are introduced in an ever in- for L > Ianih.
creasing numbers into modern reservoir development. The applica- Economides et al. (1991) came up with a more appropriate ex-
tion of horizontal technology becomes more varied and widespread pression for horizontal-well inflow performance based on Muskat's
as the cost of well configurations declines. In some cases horizontal original work. The derived equation is given by:
technology is the preferred solution for poor-reservoir-quality forma-
tions (e.g. tight-gas reservoirs). Horizontal wells increase well pro- 2πkH h△p
qH ¼
 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 : ð24Þ
ductivity by increasing the reservoir-to-wellbore exposure. In the same aþ a2 −ðL=2Þ2
μB ln þ IaniL h ln r Iani h
vein, multi-lateral holes from the same vertical wellbore accentuate this L=2 w ðI ani þ1Þ

Fig. 2. (a) Horizontal well, (b) vertical well, (c) multi-lateral well, (d) arbitrary configuration of multi-lateral well.
180 F.A. Dumkwu et al. / Journal of Petroleum Science and Engineering 82–83 (2012) 174–186

They also presented a more appropriate analytical solution with a where p0 is the gridblock pressure and pw is the wellbore pressure in
constant rate at the inner boundary. The exact solution for well inflow the grid block.
is given by: Landman (1994) extended Dikken's (1990) horizontal well model.
It was assumed that the pressure–inflow relationship at the well bore
qH ¼ 2πkH hp=μBF ð25Þ could be represented by a local law as shown by the relation:


where, F ¼ ln pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi − 2b
−1 ðL=2Þ
ln r2L −1 þ 1−γ−V ðh=LÞ. qs ðxÞ ¼ J s ðxÞ½pe −pw ðxÞ ð31Þ
b −L
2h h
ðL=2Þ þb
2 2 L tan w

Here b is the small axis of the drainage ellipse given by where qs is the specific inflow into the well (flow rate per unit
 rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

4ffi
length), pw is the well pressure, pe is the pressure at an assumed
b ¼ e −1=2 þ 1=4 þ reeH , e is the eccentricity given as
constant-pressure boundary parallel to the well (or the equilibrium
pffiffiffiffiffiffiffiffi pressure at infinity of a large reservoir), and Js is the specific produc-
e ¼ 2=3ðL=2Þ, and γ is Euler's constant (0.5772). tivity index (PI). He proposed the following model for the productiv-
Babu et al. (1991) derived a general accurate analytical equation ity of a cased and perforated completion to calculate specific PI.
for calculating well block radius, r0, in flow-rate equation (Peaceman,
" #
1978) which is valid for vertical and horizontal wells, for any well loca- 4πk 1 1
tion, as well as for isotropic and anisotropic formations. The wellblock Js ¼ þ ð32Þ
μ σ r peq J mD
radius r0 can be obtained from the equation:
sffiffiffiffiffi!   σ = σ(x) is the perforation density which may vary along the well.

r  πa kz kx
πz 
0
ln ¼ þ 0:25 ln − ln sin w −1:84−BE −ðSxz Þ The adjustable constants rpeq is the equivalent radius of a single perfora-
h 6hn2x kx kz h
tion, and JmD is dimensionless “maximal specific productivity index”.

pffiffiffiffiffi
pffiffiffiffiffi
a= kx ≥0:75 h= kz h
 i−1
L L
rpeq can be calculated from the co-relation r peq ¼ 2p ln rpp þ St ,
ð26Þ h i−1
and JmD from J mD ¼ 12 ln Lp þr
L
w
where Lp and rp are the perforation
BE is the boundary term which can be considered as negligible. The length and radius respectively, and st is the appropriately normalized
term Sxz is given by total skin due to the wellbore and formation damage. For a partially
penetrating well the following co-relation was developed to predict ap-
  nX
−1  
   πn qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

π z 2 πnλ v−2n −2n proximate specific PI.
Sxz ¼ cos 1 þ xn x = sin 1 þ α 2n 1−xn x
nz n¼1 2nz 2nz
2 0 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi13−1
ð27Þ 4πk 6 BL−x þ ðL−xÞ þ r w C7
2 2
J s ðxÞ ¼ 4 ln@ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi A5 ð33Þ
μ
and well location given by: −x þ x2 þ r2w

ðxw ; zw Þ≡1=2ðvΔx; λΔzÞ Palagi and Aziz (1994) discussed extensively on the treatment of
wells in a flexible voronoi grid. They presented new, exact and simpli-
where,
fied well models for homogeneous reservoirs for calculating Peace-
man's r0 given by:
α n ≡ α sinðπn=2nz Þ;
 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2


xn ≡ α n þ 1 þ α 2n ; w 0
ln r eq =r w ¼ ðθ=2πÞ ξanal −ξnum ð34Þ
v ≡ ð2iw þ 1Þ; iw ¼ 0; 1; …ðnx −1Þ:
and
They also presented graphical methods with approximate curve-
fitting techniques with an easy-to-use equation for calculating r0: θkh 2πkh
Iw ¼ ¼ : ð35Þ
! lnðr 0 =r w Þ ξw 0
anal −ξnum

1=4 Δx2 Δz2 1 þ expð2:215−3:88nλ nz =α Þ
r 0 ¼ 0:14ðkx kz Þ þ : ð28Þ
kλ kz 1 þ 0:533ðα=nz Þ The analytical solution for the bottom-hole flowing pressure (BHFP)
was given by:
This equation is suitable for wells located near the center of the
θkh θ w
drainage area. For α ≤ (0.2nz) and isotropic cases, this equation ap- ðp −pw Þ ¼ ξ
proximates the Peaceman formula to within 10% error. In the range q Bμ 2π anal
36
0.2nz ≤ α ≤ nλnz and for nλ ≤ 20, this equation yields wellblock radii q q
ξw
anal ¼ ∑ m pDm −2π t t DA :
to within 13% of the correct analytically calculated values. For m q q
nλ > 20, the graphical solution is not suitable.
Brekke and Jonansen (1993) developed a well model based on where qm is the rate of mth well; pDm is the dimensionless pressure at
Peaceman's formulae for modeling local inflow performance: the location of the well of interest caused by the mth well and all of its

1=2 images; qt is the net production rate from the reservoir which can be
aΔx kx ky krα positive or negative; tDA = kt/φμctA is the dimensionless time; and A is
qα ¼
 ðpo −pw Þ ð29Þ 0
the reservoir area. Similarly, a numerical factor, ξnum , was defined
ro
Bα μ α log r þ S such that
w



1=2 1=2 θkh θ 0

ky
Δz 2
þ kz
Δy2 ðp −p0 Þ ¼ ξ
kz ky q Bμ 2π num
r o ¼ 0:28
1=4
1=4 ð30Þ
ky
þ kkz
k z y where ξ0num ¼ 2πkh qt
q Bμ ðpini −p0 Þ−2π q t DA :
 F.A. Dumkwu et al. / Journal of Petroleum Science and Engineering 82–83 (2012) 174–186 181

To ensure easy implementation of the homogeneous well model Settari (1996) developed empirical relationships to determine
code in the simulator, the authors provided solution for req as follows: productivity enhancement from fractures in horizontal wells. He de-
! veloped a linear flow equation given by:
0
 1
ln r eq ¼ ∑ T ij ln Lij −θkh = ∑ T ij : ð37Þ 0:001127ð2kH Þ 2xf h C
j j
q=ΔP ¼ @ þ PIAD ð41Þ
μBxnf
For exact well modeling of heterogeneous reservoirs they pro-
posed a series of techniques to obtain req. Furthermore, for a simpli- D ¼ η þ n1 xfd þ n2 x2fd þ n3 x3fd ;
fied heterogeneous well model, the following equation was proposed:

where, η ¼ η−1
0:05 =0:05 xfd þ 1 for xfd ≤ 0:05;
! and η¼η0:05 for xfd > 0:05:
ln r eq ¼ 
∑ T ij ln Lij −θk conn h = ∑ T ij ð38Þ
j j Helmy and Wattenbarger (1998) presented simple and easy-to-
use correlations to calculate the productivity of horizontal wells pro-
where θk conn ¼ ∑j θj k j , θj = angle of connection ij, k j = harmonic ducing at constant flowing bottomhole pressure or constant rate from
average of the permeability values evaluated at points logarithmi- bounded reservoirs. They used the productivity equation of a vertical
cally distributed between req * and grid point j (as shown in Fig. 3), well to describe the productivity of a horizontal well producing at a
* = estimated value of req used only to determine the position of
and req constant rate:
the points where permeability values are evaluated at each connection.
Su (1995) proposed the following well equation for multiple-phase kLeq
J CR ¼
 ð42Þ
2D models: 4Aeq
γr 0 − 2 ln C A þ sP
1 1
141:2Bμ 2 ln

X
4  

2 
q ¼ 2π F i ðkr kh=μBÞi ½ðpi −pw Þ= lnðr i =r w Þ : ð39Þ x x
where ln C A ¼ 4:485− 4:187−12:56 t eq þ 12:56 t eq þ 2:0 ln
i¼1 eq eq




πz a
sin h eq þ ln heq and
eq eq

V
The weighting factor, Fi, quantifies the influence of neighboring
cells on well production and is calculated by:
!0:858 ! !
beq heq
F 1 ¼ ð1−xD Þð1−yD Þ; sp ¼ −1  −0:025 þ 0:022 ln C A −3:781 ln
F 2 ¼ ð1−xD ÞyD ; Leq t eq
F 3 ¼ xD yD
and F 4 ¼ xD ð1−yD Þ:

t
2 ! ! ! !3
y yeq Leq Leq
Mochizuki (1995) extended Peaceman's equation for a well areal- 61:289−4:751 eq þ 4:652 þ 1:654 −1:718 7
6 t eq t eq beq beq 7
ly and vertically inclined at arbitrary angles with respect to grid lines 6 7
þ6 ! !1:472 7 :
6 heq t eq 7
in anisotropic reservoirs. The flow rate equation for an inclined well 4 5
in a grid block is expressed as: t eq t eq

k L ðθ; ωÞ

′ ′
′ Dogulu (1998) developed an improved numerical model which
Q ¼ 0:001127WI ðθ; ωÞ pgrid −pbhp ð40Þ
Bμ estimates the productivity of a perforated completion as a function of
perforating parameters like shot density, penetration depth and phas-
where WI ′ ðθ; ωÞ ¼
′ 2π 
rb ðθ;ωÞ
ing angle. Wolfsteiner et al. (1999) presented a semi-analytical method
ln ′ ðθ;ωÞ
rw
þs for the approximate modeling of the productivity of advanced wells in
θ is the angle measured from the vertical axis and ω is measured heterogeneous reservoirs. Their approach was based on Green's func-
from the x-axis. tions and was referred to as the s-k* approach, modeling permeability
Well length L′ in the transformed solid can be obtained from the heterogeneity in terms of an effective skin s that varied along the well
expression trajectory and a constant background permeability k*. Ibragimov et al.
(1997, 1999) developed highly precise flow rate formulae for horizontal
n


 o1=2 wells located in the middle of a bounded 3-D reservoir following the
L′ ðθ; ωÞ ¼ L k′ =kx ð sinθ cosωÞ2 þ k′ =ky ð sinθ sinωÞ2 þ k′ =kz ð cosθÞ2
n o1=2 procedures prescribed in literature (Borisov et al., 1964; Charny, 1948;
r ′w ðθ; ωÞ ¼ r ′w2 ðxÞð sinθ cosωÞ2 þ r ′w2 ðyÞð sinθ sinωÞ2 þ r ′z 2 ðzÞð cosθÞ2 Joshi, 1991; Kuchuk and Goode, 1988).
n o1=2
r′b ðθ; ωÞ ¼ r′b 2 ðxÞð sinθ cosωÞ2 þ r ′b ðyÞð sinθ sinωÞ2 þ r ′b 2 ðzÞð cosθÞ
k 2πL
q¼ ΔP when Rb =h < 20 ð43Þ
μ ln Lr þ 0:1 R−h
h
ln Rh

For slanted wells:

1 LðθÞ
qðθÞ ¼ C:F ðθÞ þ qH ðLmax ; 2hÞ
2 Lmax
1
 h ð44Þ
qV − qH L max; 2h
C¼ 2 L max
F ð0Þ

qH(Lmax, 2h) is the productivity of a horizontal well of length Lmax

Fig. 3. Evaluation of average permeability at connection ij. located in the middle of the reservoir of thickness 2h. When L(θ) =
182 F.A. Dumkwu et al. / Journal of Petroleum Science and Engineering 82–83 (2012) 174–186

h = Lmax = const (case of a perfect vertical well inclined at an angle α to The required influx into the first boundary block was given in
the vertical) the function F(θ) takes the form: terms of radial flow solution. The pressure profile was written as:
8  
> 1 1 υþ1 π w w  ln r− ln r w
> pðr Þ ¼ p þ pb −p : ð47Þ
< r −ð1− cosθÞ ; 0 ≤ θ<
ln r b − ln r w
ðλ þ 1Þ cosθ 2 3
F ðθÞ ¼ υ
>
> ð cosθÞ π π
: ; ≤θ≤
λþ1 3 2 The hole well model obliges the correct influx by setting the pres-
sure at the hole boundary rh to ph, where ph is given as:
where the well length is chosen subject to condition L(θ) = min{h/
cos α ; Lmax}; Lmax is the length limited by technological restrictions. w  w  ln r h − ln r w
ph ¼ p þ pb −p :
F(θ) is expressed as: ln rb − ln r w
8
>
> 1 h The hole well model was written in terms of well transmissibility
>
> ; when 0 ≤ α < arccos
>
> ðλ þ 1Þ2υ L max
>
<    T w as:
h 1 L h h
F ðθÞ ¼ − 1− max cosα ; when arccos ≤ θ < arccos :
>
> L max ðλ þ 1Þ cosθ 2υ h L max 2L max
w w  w w 
>
>  
>
> ð cosθÞυ L max h π
>
: ; when arccos < θ≤ q¼T p −pb ¼ T h p −ph ð48Þ
λþ1 h 2L max 2

w θkh w θkh
where T ¼ r and T h ¼ r :
If the well is located in the middle of the reservoir parallel to the lnr b lnr h
w w

top and bottom, the dependence of flow rate on the pressure drop
can fairly accurately be described by the following formula:
!   showed that linear solution technique was scalable with respect to
kϕ L pffiffiffi Lr h the number of blocks, which was an indication that the method will
q¼ π L þ π L pffiffi erf pffiffiffiffiffiffiffi ð45Þ
μ ln R ln R t ϖt
where ϖ ¼ μβk  ¼ piezoconductivitycoefficient: Yu and Daviess (2000)
illustrated the advanced well modeling methodology by quantifying
the benefits of installing an “intelligent” completion. They modeled
several “intelligent” completions in order to obtain an optimum pro-
duction performance. Some important attributes of “intelligent” well
configurations include:
(1) a multi-lateral or multi-zone well producing from one or more
reservoirs, e. g. using produced gas to accelerate oil production
as a form of artificial lift
(2) a well with the ability of choking the production flow, both
down-hole and at the surface
(3) a well with the ability to shut off water/gas producing zone at
the wellbore
(4) a well with the ability to separate water/gas and oil in the well-
bore and re-injecting the separated water into a separate zone
(5) a well with down-hole compressor installed for produced gas
re-injection into the reservoir.
Jenny et al. (2002) developed a hexahedral multiblock grid (MBG)
formulation for the modeling of two-phase reservoir flows. They also
parallelize efficiently. Chen et al. (2002) presented an application of
control volume function approximation (CVFA) method to the
modeling and simulation of 2D and 3D horizontal wells in petroleum
reservoirs. The base grid for their method was based on a voronoi
grid. Serve (2002) presented an enhanced semi-analytical framework
for modeling complex configurations, such as multi-lateral wells with
“intelligent” completions. His tool can handle wells operating under
rate or bottomhole pressure control and multiple wells with switch-
ing constraints can also be handled. Wan and Aziz (2002) described
a semi-analytical solution for horizontal wells with multiple fractures.
Their analytical solution provides the well pressure that can be com-
bined with a numerically computed gridblock pressure to obtain well
index. They related well pressure and wellblock pressure through the
radial flow equation:
krp
q ¼ WI ðp−pw Þ ð49Þ
μB
where WI ¼ θkh
r
 :
ln 0
rw þS

Fitzpatrick and Ponting (2001) described a model in which the

wellbore flow in terms of pressure, velocity and saturation was solved
concurrently with the reservoir flow equations. In that model, con-
straints at branches, items of well equipment, well target conditions,
the effects of friction, gravity, and fluid acceleration were included in
the formulation. Ouyang and Aziz (2001) developed a computer
package to perform a series of sensitivity studies that led to several
important and interesting observations. Their coupled model was ap-
plied to determine well productivity, well index for multiphase flow
simulation, wellbore pressure profile, and wellbore inflow or outflow
distribution for any time in the well's production/injection life.
Wolfsteiner and Durlofsky (2002) developed a near-well upscaling
procedure suitable for use with multiblock grids. Their technique is
applicable to grids that are radial (or nearly radial) in the near-well
region rather than rectangular. The method was well-suited for models
in which the general well trajectory was resolved by the simulation grid
and was capable of modeling reservoir heterogeneity down to the scale
of the wellbore itself. The influx q into a well in a homogeneous system
is constant in the radial direction (Fig. 4) and is given by

pw −pb p−pb
q ¼ θkh ¼ θkh : ð46Þ
ln rrb ln rrb Fig. 4. Hole well model for multiblock grid.
w
 F.A. Dumkwu et al. / Journal of Petroleum Science and Engineering 82–83 (2012) 174–186
Pedrosa and Aziz (1986) proposed a numerical scheme for the
near-well flow modeling, but their method is limited to radial grid-
blocks. Ding and Jeannin's (2004) approach is more general and can
be applied to other kinds of grids. Ding et al. (2004) presented bound-
ary integral formulation for well modeling in a 3D stratified petroleum
reservoir with arbitrary geometry. Fokker et al. (2005) proposed a semi-
analytic technique to solve the inflow equations in an approximate but
reliable manner. They provided a solution to 3D problems of single-
phase flow into a horizontal well, taking into account friction in the
wellbore. They also extended the method to describe fluid flow when
the well intercepts one or more fractures. Hasan and Kabir (2005)
presented a detailed mechanistic model to estimate film thickness
before computing frictional pressure-drop as gas flows past the wavy
liquid film surrounding the pipe wall. They also developed a simple
model for annular two-phase flow in wellbore.
Vicente and Ertekin (2006) developed a fully implicit numerical
model coupling reservoir and multifractured horizontal well flow dy-
namics to investigate flow behavior and predict the productivity of
such systems. The authors obtained a one-dimensional expression
describing single-phase flow in the wellbore domain that take into
account friction losses and radial influx along the wellbore:
183
Johansen and Khoriakov (2007) described an approach in the
modeling of multi-phase flow in advanced well completions and in
the near well region of the reservoir. In this approach, a base solver
is used to calculate pressure distribution, flow rates and phase frac-
tions in all components of the well completion using a simple bound-
ary condition and assuming homogenized flow. Kankom (2007)
modified Furui et al.'s (2003) model which described the relationship
between flow rate and pressure drawdown.


keq b p−pwf
q¼
h i  ð53Þ
141:2Bμ ln r ðhII aniþ1Þ þ hIπyb −1:224 þ S þ SPR
w ani ani
Here, SPR is partial penetrating skin factor and b is the reservoir
pffiffiffiffiffiffiffiffiffiffiffi
drainage dimension along horizontal well direction, and keq ¼ kH kV .
The model can be applied to a variety of wellbore trajectories including
horizontal wells, slanted wells, undulating wells, and multilateral wells.
Moreover, Kankom (2007) formulated the analytical inflow perfor-
mance to estimate the well performance. The closed form model of in-
flow performance for undulating wells is:
kH hΔp
 q¼ 1   
: ð53Þ
∂ Ax keq ∂Φ ∂p 141:2Bμ 2 lnð4AÞ− lnðC A Þ− ln γr2w þ Sθ þ SP þ S
Δx þ qsc ¼ Vϕc 50
∂x μB ∂x ∂t
Here A, ln(CA), Sθ, SP, S represent drainage area, shape factor,
2
Where Ax = πrw and porosity is 100% (ϕ = 1.00). Yang and Deo slanted skin factor, partial penetrated skin factor and damage skin
(2006) extended multiphase flow problem shown in Fig. 5 as follows: factor, respectively.
Lu and Tiab (2007) presented a simple productivity equation for a
kr θk horizontal well in pseudo-steady state in a closed anisotropic box-
q¼ ðp−pw Þ ð51Þ
Bμ lnðr=r w Þ shaped reservoir, using a uniform line skin model. The equation to
compute productivity is given by:
where θ is the span of the well in this particular triangle, r is the dis-
1=2

tance from the well to the triangle center. P is the pressure at the cen- 0:001127 ky kz L pavg −pw =ðμBÞ
ter of triangle, Pw is the well-bore pressure. K is the permeability of q¼ ð54Þ
Λ þS
the triangle, and rw is the well-bore radius.
The same interpolation function used in the triangular elements where Λ is box-shaped drainage volume given by,
was used in tetrahedral elements in three-dimensional space. The     n
Lx;D 1
 h io
multiphase flow equation is given by: Λ¼ − ln 4 sin πzw;D =hD sin πRw;D =ð2hD Þ
6hD 2π
kr θhk
q¼ ðp−pw Þ: ð52Þ
Bμ lnðr=r w Þ where S is the pseudo skin factor due to partial penetration. Jenny and
Lunati (2009) devised an extension of the multi-scale finite-volume
In this case, angle θ is the angle between the two triangles that method which allows the simulating of flow and transport processes
share the line well; h is the length of the well in the tetrahedral in reservoirs with complex well configurations. They considered an
block. Krogstad and Durlofsky (2007) presented a coupled wellbore- incompressible two-phase system in which the evolution of the
reservoir flow model that is based on multiscale mixed finite element phase saturations Sα = (α ∈ {1, 2}) and flow rate was described by:
formulation for reservoir flow linked to a drift-flux wellbore flow " !#
representation. ∂Sα ∂ krp ∂pα ∂z
ϕ − k −gρα ¼ −qα : ð55Þ
∂t ∂xi ij μ α ∂xj ∂xj

Multisegment well model has proven to be very effective in the

modeling of advanced wells. In this model, the wellbore is divided
into an arbitrary number of segments and the degree of accuracy
expected is a function of the number of segments used (Holmes et
al., 1998). Hasan and Kabir (2010) presented a robust model for a
two-phase flow in geothermal wells using the drift-flux approach.

3.1. The limitations of advanced well models

Both advanced analytical and numerical well models have been

Fig. 5. Illustration of the well model with well placed in the corner.
developed for predicting the actual performance of hydrocarbon res-
ervoirs. For reservoir performance to be accurately predicted, these
well models are expected to simulate transient multiphase flows in
systems comprising one or more connected hydrocarbon reservoirs
and several wells that are produced under specified constraints of
pressures or rates imposed on individual wells or groups of wells.
184 F.A. Dumkwu et al. / Journal of Petroleum Science and Engineering 82–83 (2012) 174–186
Unfortunately, despite the sophistication of these advanced well
models and the successes recorded with them, some predictions
made with them in some cases do not match actual field performance.
This position is buttressed by the result of the study conducted by
Mullane et al. (1996) in which the field performance of 29 horizontal
wells was compared with predictions. The authors observed that
while the estimated oil rate was matched, the predictions of approxi-
mately a half of the wells used for the study were within ± 50% of actual
performance and an average absolute error of 47% was recorded. While
this study was conducted for horizontal wells, a similar trend could be
obtained for all advanced wells models (Mullane et al., 1996).
The inability of these well models to accurately predict the actual
performance stems from the assumptions made during their formula-
tion. The sources of uncertainty in these well models as noted by Aziz
et al. (1999) are stated below:
1. The assumed well model saturation gradients cause the effective
phase permeabilities for the well region to differ from the corre-
sponding values for the well grid block.
2. The values of the effective absolute permeabilities for the well re-
gion may be at variance with average values obtained for the grid
block.
3. The default procedures available in reservoir simulators are devel-
oped based on assumptions that are most suited for modeling ver-
tical (conventional) wells than for horizontal (advanced) wells;
consequently, the equivalent grid block radius obtained may be
in error.
4. The effective skin may be an unknown quantity.
The afore-stated assumptions affect the accuracy of computed
well index (WI). Oil production rate, gas oil ratio (GOR), and water oil
ratio (WOR) are all sensitive to WI, hence the utilization of a wrong
value of WI during simulation affects the ability of these advanced
well models to predict accurately.
4. The next generation well models
Reservoir simulators have long become a primary reservoir engi-
neering tool for recovery analyses and performance evaluation in
the oil and gas industry. The main incentive for reservoir simulation
is increased profitability through better reservoir management. A simu-
lator requires a set of realistic boundary conditions for the computation
of pressure distribution in a reservoir which are supplied by well
models.
As multi-scale multi-domain data, geological and geophysical
models become more integrated with increasing resolution and com-
plexity over time, the quest for faster, more efficient, and more com-
prehensive next generation simulation engines becomes inevitable.
Therefore, using the present crop of available well models in such
complex next generation reservoir simulators will no doubt raise
questions among the worldwide industry practitioners regarding
their representativeness and the reliability of their predictions.
Recognizing this daunting challenge, it is imperative that all the
factors responsible for the inherent limitations of both conventional
and advanced well models be addressed. Effort should be geared to-
ward the elimination or minimization of errors and uncertainties im-
plicit in the formulation of next generation well models for modeling
conventional, complex multilateral, horizontal, fractured and smart
wells. More so, pressure losses due to friction, slippages and cross
flows should be adequately accounted for.
Furthermore, for increased flexibility, reliability and robustness of
next generation well models, adequate care should be taken to cover
a wide range of physical processes taking place in the well irrespec-
tive of their geometries to enable engineers evaluate many scenarios.
The well models should be comprehensive enough to provide opera-
tional flexibility to minimize risks in reservoir management. Among
others, such a comprehensively formulated well model should be
capable of handling global phase-component partitioning, allow mul-
tiple components and phases to exist as well as accommodate opera-
tional scenarios where any component can be made to exist in any
phase.
5. Concluding remarks
In this paper, the development and representation of conventional
and advanced well models in reservoir simulators were covered with
emphasis on analytical and numerical well models since the last re-
view on the subject in 1993. It is clear that a simulator requires a
set of realistic boundary conditions for the computation of pressure
distribution in a reservoir and these boundary conditions are supplied
by well models. Conventional well models are sufficient for modeling
vertical and slightly deviated wells. However, the development of
complex wellbore configurations due to the implementation of intel-
ligent completion systems coupled with the advent of horizontal and
multilateral drilling technology necessitated the development of in-
creasing complex well models for use in reservoir simulation for
modeling advanced wells, such as horizontal, multilateral and smart
wells. This is because conventional well models do not adequately
represent the physics of the process, such as pressure loss due to fric-
tion, effect of slippage between the wellbore fluid and crossflow. De-
spite the sophistication of these advance well models and the
successes recorded with them, some predictions made with them in
some cases do not match actual field performance. For this reason,
the development of well models that give accurate simulation results
that match actual field performances, though a challenging task,
should be given high priority.
The very first step in the development of accurate conventional
and advanced well models is the recognition of the sources of uncer-
tainty which affect the accuracy of well index during their formula-
tions. Next, all uncertainties implicit in their formulations should be
eliminated or minimized and pressure losses due to friction, slippages
and cross flows adequately accounted for. Furthermore, proper care
should be taken to cover a wide range of physical processes taking
place in the well irrespective of their geometries to enable engineers
evaluate a wide range of operational scenarios. All these require
painstaking optimization studies which can sometimes be a formida-
ble exercise.
As multi-scale multi-domain data, geological and geophysical
models become more integrated with increasing resolution and com-
plexity over time, the quest for faster, more efficient, and more com-
prehensive next generation simulation engines becomes inevitable. A
high resolution reservoir simulation is vital for better definition and
description of fluid flow in a reservoir which will ultimately lead to
improved field development. Among others, such a comprehensively
formulated well model should be capable of handling global phase-
component partitioning, allow multiple components and phases to
exist as well as accommodate operational scenarios where any com-
ponent can be made to exist in any phase. Therefore, well model
developments, especially advanced well models, require a clear
identification of the ultimate objective.
Nomenclature
B formation volume factor, RB/STB
h reservoir thickness, ft
rw radius of wellbore
a large axis of drainage ellipse, ft
A area, acres
(Cm, C, Dm) constants
h reservoir thickness, ft pffiffiffiffiffiffiffiffiffiffiffiffiffi
Iani permeability anisotropy variable, kH =kV
k permeability, md
k′ equivalent homogeneous permeability, md
L horizontal well length, ft
 F.A. Dumkwu et al. / Journal of Petroleum Science and Engineering 82–83 (2012) 174–186 185

L′ well length in transformed plane, ft d non-Darcy flow effect

n number of grid blocks p partially completed
p pressure, psi m mechanically damaged
q flow rate, STB/D A based on area
R radius of the external reservoir boundary (no-flow barrier), B well block or boundary cell
ft D dimensionless
r radial distance, ft H horizontal
r′ equivalent radial distance in transformed plane, ft h hole
t reservoir width, ft V vertical
S skin factor x, y, z direction in x, y, z axis respectively
Sα phase saturation w well bore
T transmissibility coefficient, dimensionless ij connection between gridblocks i and j
V bulk volume, ft 3 avg average
(Δx, Δz) grid size (ft, ft) bhp bottom-hole pressure
(xw, zw) well coordinates (ft, ft) bh bore hole
WI′ well index or well
pffiffiffiffiffiffiffiffiffiffiffi
ffi connection factor in transformed plane eq equivalent
α ðΔx=ΔzÞ kx =kz ≥1, scaled grid aspect ratio fd dimensionless fracture half distance
β = 1/α ≤ 1 scaled grid aspect ratio grid gridblock pressure
β* cubical elasticity ratio rp relative permeability
ρ density sc standard condition
μ viscosity, cp nf no flow boundary distance
θ well angle with respect to vertical coordinate.
re radius of drainage or external boundary
ro equivalent radius Superscript
nw number of wells, n * well of interest
I number of layered zones in the r direction w well bore
krg gas relative permeability
ω dimensionless term for non-Darcy flow effect
γ correlation factor for extra pressure drop around a wellbore
δ fraction of flow capacity for the open interval References
I non-Darcy flow coefficient
μ viscosity Aavatsmark, I., Barkve, T., Boe, O., Mannseth, T., 1998a. Discretization of unstructured
grids for inhomogeneous, anisotropic media. Part I: derivation of the methods.
q production rate of well. SIAM J. Sci. Comput. 19 (5), 1700.
pe Pressure at drainage radius, re Aavatsmark, I., Barkve, T., Boe, O., Mannseth, T., 1998b. Discretization of unstructured
pwf Bottom-hole pressure, flowing grids for inhomogeneous, anisotropic media. Part II: discussion and numerical
methods. SIAM J. Sci. Comput. 19 (5), 1717.
s pseudo skin factor Agaev, G.S., 1996. Improved technique for off-center well modeling in multidimen-
(λ, v) integer sional reservoir simulation. Paper SPE 35532.
R radius, ft Augustine, J.R., 2002. An investigation of the economic benefit of inflow control devices
on horizontal well completions using a reservoir-wellbore coupled model. Paper
ϕ porosity SPE 78293, Aberdeen, United Kingdom, October 29–31.
p0 initial pressure Aziz, K., Arbabi, S., Deutsch, C.V., 1999. Why is it so difficult to predict the performance
p dimensionless pressure of horizontal wells? J. Can. Pet. Technol. 38 (10), 37–45.
Babu, D.K., Odeh, A.S., Al-Khalifa, A.J., McCann, R.C., 1991. The relation between well-
Dr relative non-Darcy skin coefficient
block and wellbore pressure in numerical simulation of horizontal wells. SPERE,
H total formation thickness pp. 324–328.
ψwell wellbore gradient Baumann, C.E., Price, H.S., Reddy, M.P., Thuren, J.B., 1999. Full field pressure simulations
using a very accurate, yet inexpensive well model. Paper SPE 56620.
dz depth gradient
Borisov, J.P., Pilatovsky, V.P., Tabakov, V.P., 1964. Oil production using horizontal and
ρ molar density multiple deviation wells, Nedra, Moscow. (in Russian).
x reservoir variable or oil mole fraction Brekke, K., Jonansen, T.E., 1993. A new modular approach to comprehensive simulation
y gas mole fraction of horizontal wells. Paper SPE-26518, pp. 109–123.
Chappelear, J.E., Williamson, A.S., 1981. Representing wells in numerical reservoir sim-
Q production or Injection rate ulation: Part 2 – Implementation. Paper SPE 9770.
pbh bottom-hole-pressure at the reference depth Charny, I.A., 1948. Underground hydromechanics. (in Russian).
pk well cell pressure Chen, G., Tehrani, D.H., Peden, J.M., 1995. Calculation of well productivity in a reservoir
(1) simulator (I). Paper SPE 29121 Presented at the SPE Reservoir Simulation Sympo-
req, m equivalent wellblock radius sium held in San Antonio, Texas, February, pp. 12–15.
2=
req, mk equivalent wellblock radius of well m in the presence of one Chen, Z., Huan, G., Baoyan, Li, 2002. Modeling 2D and 3D horizontal wells using CVFA.
other well, well k, both producing at the same rate. Commun. Math. Sci. 1, 30–43.
DeBaun, D., Byer, T., Childs, P., Chen, J., Saaf, F., Wells, M., Liu, J., Cao, H., Pianelo, L.,
Tilakraj, V., Crumpton, P., Walsh, D., Yardumian, H., Zorzynski, R., Lim, K.T., Schrader,
M., Zapata, V., Nolen, J., Tchelepi, H., 2005. An extensible architecture for next gener-
Subscripts ation scalable parallel reservoir simulation. Paper SPE 93274.
Demetre, G.P., Ali, Farouq S.M., 1994. A Review of Analytical Well Models Used in Res-
i index of number of perforated intervals in a well
ervoir Simulation. Paper SPE-28179.
n total number of blocks with perforation intervals in a well Dikken, B.J., 1990. Pressure drop in horizontal wells and its effect on production perfor-
o oil mance. J. Pet. Technol. 42, 1426–1433.
Ding, Y., Jeannin, L., 2001. New numerical scheme for the near-well modeling with dis-
g gas
cretization around the wellbore boundary using flexible grids. Paper SPE 66360.
k, m well numbers Ding, Y., Jeannin, L., 2004. New numerical schemes for near-well modeling using flex-
w water ible grids. SPE J. 109–121.
k layer count Ding, Y., Renard, G., 1994. Representation of Wells in Numerical Reservoir Simulation.
SPERE, pp. 140–144.
∗ wellblock Ding, Y., Renard, G., Weill, L., 1998a. Representation of wells in numerical reservoir
a altered simulation. SPERE, pp. 18–23.
186 F.A. Dumkwu et al. / Journal of Petroleum Science and Engineering 82–83 (2012) 174–186
Ding, Yu., Renard, G., Simulation. SPE Reservoir Eng. 18–23. Weill, L., 1998b. Represen-
tation of Wells in Numerical Reservoir.
Ding, Y., Ha-Duong, T., Giroire, J., Moumas, V., 2004. Modeling of single-phase flow for
horizontal wells in a stratified medium. Comput. Fluids 33, 715–727.
Dittoe, S.R., Retnanto, A., Economides, M.J., 1996. An analysis of reservoir enhancement
in petroleum reservoirs with horizontal and multilateral wells. Paper SPE-37037,
pp. 115–123.
Dogru, A.H., 2010a. Equivalent wellblock radius for partially perforated vertical wells—
part I: anisotropic reservoirs with uniform grids. SPE J. 15, 1028–1037.
Dogru, A.H., 2010b. Equivalent wellblock radius for partially perforated vertical wells—
part II: anisotropic heterogeneous reservoirs with nonuniform grids. SPE J. 15,
1038–1046.
Dogulu, Y.S., 1998. Modeling of well productivity in perforated completions. Paper SPE-
51048, pp. 109–118.
Economides, M.J., Deimbacher, F.X., Brand, C.W., Heinemann, Z.E., 1991. Comprehen-
sive simulation of horizontal-well performance. SPE Formation Eval, pp. 418–426.
Edwards, M.G., Rogers, C.F., 1994. A flux continuous scheme for the full tensor pressure
equation. Proceedings, 4th European Conference on Mathematics of Oil Recovery,
Roros, Norway.
Erdal, O., 2001. Analysis of horizontal-well responses: contemporary vs. Conventional.
SPE Reserv. Eval. Eng. 4 (4), 260–269.
Fitzpatrick, A.J., Ponting, D.K., 2001. Modeling complex wells in detailed geologies.
Paper SPE-66370.
Fokker, P.A., Verga, F., Egberts, Paul J.P., 2005. New semianalytic technique to deter-
mine horizontal well productivity index in fractured reservoirs. SPE Reserv. Eval.
Eng. 123–131.
Fung, L.S., Hiebert, L.D., Nghiem, L.X., 1991. Reservoir simulation with a control-volume
finite element method. Paper SPE 21224.
Fung, Larry S.K., Su, H., Tan, C.T., Hemanthkumar, K., Pita, J.A., 2005. A fully-implicit
fully-coupled well model for parallel mega-cell reservoir. Paper SPE-106331.
Furui, K., Zhu, D., Hill, A.D., 2003. A rigorous formation damage skin factor and reservoir
inflow model for a horizontal well. SPE Prof Facil, p. 151.
Giger, F.M., Reiss, L.H., Jourdan, A.P., 1984. The reservoir engineering aspects of hori-
zontal drilling. Paper SPE-13024.
Guyaguler, B., Ghorayeb, K., 2006. Integrated optimization of field development, plan-
ning, and operation. Paper SPE 102557.
Hasan, A.R., Kabir, C.S., 2005. A simple model for annular two-phase flow in wellbores.
Paper SPE-95523.
Hasan, A.R., Kabir, C.S., 2010. Modeling two-phase fluid and heat flows in geothermal
wells. J. Pet. Sci. Eng. 71, 77–86.
Heinemann, Z., Brandt, C., Munka, M., Chen, Y., 1989. Modeling reservoir geometry
with irregular grid. Paper SPE 18412 Presented at the SPE Reservoir Simulation
Symposium held in Houston, Texas, February 11–14.
Helmy, M.W., Wattenbarger, R.A., 1998. Simplified productivity equations for horizontal
wells producing at constant rate and constant pressure. Paper SPE-49090, pp. 379–388.
Holmes, J.A., 2001. Modeling advanced wells in reservoir simulation. Paper SPE 72493.
Distinguished Author Series: Journal of Petroleum Technology.
Holmes, J.A., Barkve, T., Lund, O., 1998. Application of a multisegment well model to
simulate flow in advanced wells. Paper SPE 50646.
Holmes, J.A., Byer, T., Edwards, D.A., Stones, T.W., Pallister, I., Shaw, G., Walsh, D., 2010.
A unified wellbore model for reservoir simulation. SPE 134928.
Hwang, M.K., 2000. Modeling nonlinear interactions among near-well flow restrictions
in well-deliverability prediction and simulation. SPE Reserv. Eval. Eng. 3 (4),
360–368.
Hwang, M.K., Odeh, A.S., 1995. Estimation of condensate dropout effects on well pro-
ductivity as skin change with multiplicative interactions among skin components.
Paper SPE 29894.
Ibragimov, A.I., Baganova, M.N., Nekrasov, A.A., 1997. Stationary and non-stationary
flow of fluid in a limited reservoir and inside horizontal well. The Petroleum Soci-
ety. Paper 97-132.
Ibragimov, A.I., Lukoil, M.N., Nekrasov, A.A., 1999. Domain decomposition methods and
their application to modeling of the performance of horizontal and slanting wells
in the anisotropic reservoirs. Paper SPE-51906.
Jenny, P., Lunati, I., 2009. Modeling complex wells with the multi-scale finite-volume
method. J. Comput. Phys. 228, 687–702.
Jenny, P., Wolfsteiner, C., Lee, S.H., Durlofsky, L.J., 2002. Reservoirs using hexahedral
multiblock grids. SPE J. 149–157.
Johansen, T.E., Khoriakov, V., 2007. Iterative techniques in modeling of multi-phase
flow in advanced wells and the near well region. J. Pet. Sci. Eng. 58, 49–67.
Joshi, S.D., 1988. Augmentation of Well Productivity With Slant and Horizontal Well.
JPT 729–739. Trans, AIME, 285.
Joshi, S.D., 1991. Horizontal well technology. PenWell Publishing Company, Tulsa.
Kankom, R., 2007. Modeling Performance of Horizontal, Undulating, and Multilateral
Wells. PhD Thesis. Texas A&M University.
Krogstad, S., Durlofsky, L.J., 2007. Multiscale mixed-finite element modeling of coupled
wellbore/near-well flow. Paper SPE-106179.
Kuchuk, F.J., Goode, P.A., 1988. Pressure transient analysis and inflow performance for
horizontal well. Paper SPE-18300.
Landman, M.J., 1994. Analytic modeling of selectively perforated horizontal wells. J.
Pet. Sci. Eng. 10, 179–188.
Lin, C.Y., 1995. New well models for partially penetrating wells in heterogeneous res-
ervoirs using non-uniform grids. Paper SPE 29122.
Lu, J., Tiab, D., 2007. A simple productivity equation for a horizontal well in pseudos-
teady state in a closed anisotropic box-shaped reservoir. Paper SPE-106970.
Mochizuki, S., 1995. Well productivity for arbitrarily inclined well. Paper SPE-029133,
pp. 397–405.
Mullane, T.J., Churcher, P.L., Tottrup, P., Edmonds, A.C., 1996. Actual vs. predicted well
performance, Weyburn unit, S.E. Saskatchewan. J. Can. Pet. Technol. 35 (3), 24–30.
Ouyang, L-Biao, Aziz, K., 1998. A simplified approach to couple wellbore flow and res-
ervoir inflow for arbitrary well configurations. Paper SPE-48936.
Ouyang, L-Biao, Aziz, K., 2001. A general single-phase wellbore/reservoir coupling
model for multilateral wells. SPE Reserv. Eval. Eng. 327–335.
Palagi, C.L., Aziz, K., 1994. Modeling vertical and horizontal wells with Voronoi grid. SPE
Reservoir Eng. 15–21.
Peaceman, D.W., 1978. Interpretation of well-block pressures in numerical reservoir
simulation. Soc. Pet. Eng. J. 183–194.
Peaceman, D.W., 1982. Interpretation of well-block pressure in numerical reservoir
simulation with nonsquare grid blocks anisotropic permeability. SPE 10528.
Peaceman, D.W., 1983. Interpretation of well-block pressure in numerical reservoir
simulation with nonsquare grid blocks anisotropic permeability. SPE J. 9,
531–543 (June 1983).
Peaceman, D.W., 1987. Interpretation of well-block pressures in numerical reservoir
simulation—part 3: some additional well geometries. SPE 28670, pp. 183–194.
Peaceman, D.W., 1990. Interpretation of well-block pressures in numerical reservoir
simulation: part 3—off center and multiple wells within a wellblock. SPE Reservoir
Eng. J. 227–232 (May 1990).
Peaceman, D.W., 1995. A New Method for Representing Multiple Wells with Arbitrary
Rates in Numerical Reservoir Simulation. Paper SPE 31066, pp. 1–22.
Peaceman, D.W., 2003. A New method for calculating well indexes for multiple
wellblocks with arbitrary rates in numerical reservoir simulation. Paper SPE
79687.
Pedrosa Jr., O.A., Aziz, K., 1986. Use of hybrid grid in reservoir simulation. SPERE, Trans.
AIME 281, 611–621.
Schwabe, K., Brand, J., 1967. Prediction of reservoir behavior using numerical simula-
tors. Paper SPE 1857.
Serve, J., 2002. An Enhanced Framework for Modeling Complex Well Configurations.
MS Thesis, Stanford University.
Settari, F.A., 1996. A practical method for determining the productivity of multi-
fractured horizontal wells. The Petroleum Society, Paper 96-22.
Settari, A., Aziz, K., 1974. A computer model for two-phase coning simulation. SPE J.
221–236 (June 1974).
Shiralkar, G.S., 1988. Calculation of flowing well pressure in reservoir simulation using
nine-point differencing. Petroleum Society of CIM 88-39-01.
Shiralkar, G.S., 1989. Calculation of flowing well pressure in reservoir simulation using
nine-point differencing. J. Can. Pet. Technol. 28 (6), 73–82.
Su, Ho-Jeen, 1995. Modeling off-center wells in reservoir simulation. SPE Reservoir
Eng. 47–51.
Vicente, R., Ertekin, T., 2006. Modeling of coupled reservoir and multifractured hori-
zontal well flow dynamics. Paper SPE 101929.
Wan, J., Aziz, K., 2002. Semi-analytical well model of horizontal wells with multiple
hydraulic fractures. SPE J. 437–445.
Williamson, S.A., Chappelear, J.E., 1981. Representing wells in numerical reservoir sim-
ulation: Part 1 – Theory. SPE 7697.
Wolfsteiner, C., Durlofsky, L.J., 2002. Near-well radial upscaling for the accurate model-
ing of nonconventional wells. Paper SPE-76779.
Wolfsteiner, C., Durlofsky, L.J., Aziz, K., 1999. An approximate model for the productiv-
ity of non-conventional wells in heterogeneous reservoirs. Paper SPE-56754, pp.
1–11.
Wolfsteiner, C., Durlofsky, L.J., Aziz, K., 2000. Approximate models for productivity of
nonconventional wells in heterogenous reservoirs. SPE J. 218–226.
Wolfsteiner, C., Aziz, K., Durlofsky, L.J., 2001. Modeling conventional and nonconven-
tional wells. Presented at 6th International Forum on Reservoir Simulation, held
in Hof/Salzburg, Austria, September 3–7.
Yang, Y., Deo, M.D., 2006. Modeling of multilateral and maximum reservoir-contact
wells in heterogeneous porous media. Paper SPE-99715.
Yu, S., Davies, D.R., 2000. The modeling of advanced “intelligent” well—an application.
Paper SPE-62950.