SPE 108984
A New Method of Evaluating the Productivity Index for Nonlinear Flows
E. Aulisia and A. Ibragimov, Texas Tech. University, and P.P. Valkó, SPE, and J.R. Walton, Texas A&M University
Copyright 2007, Society of Petroleum Engineers
This paper was prepared for presentation at the 2007 SPE Annual Technical Conference and
Exhibition held in Anaheim, California, U.S.A., 11–14 November 2007.
This paper was selected for presentation by an SPE Program Committee following review of
information contained in an abstract submitted by the author(s). Contents of the paper, as
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
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
Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper
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.
Box 833836, Richardson, Texas 75083-3836 U.S.A., fax 01-972-952-9435.
Abstract
This paper addresses the effects of non-linear, non-Darcy
flows in reservoir on the value of the productivity index. The
productivity index (PI) of the well draining a reservoir with no
flux on the exterior boundaries is defined as the ratio between
the production rate and the pressure drawdown. Experience
shows that during the dynamical process of hydrocarbon
recovery this ratio stabilizes to some constant value, which, in
general, is a non-linear function of both the pressure
drawdown and the production rate.
The linear (Darcy) case is well understood, and excellent
approximate formulae are available to compute the PI for
various well-reservoir geometries. These formulae are
generally obtained through a semi-analytical solution of the
transient problem. To handle the more realistic non-linear
situation, the current practice is to solve first the nonlinear
problem many times for different values of production rate,
and then to add some ad-hoc corrective parameter(s) in the
linear formulae in order to reproduce the non-linear nature of
the flow.
Our approach, based on recent progress in the modeling of
transient Forchheimer flows, uses partly non-numerical
techniques to evaluate the productivity index. It provides, for a
wide class of reservoir geometries, an accurate enough
approximate algebraic relation between the PI for the non-
linear Darcy-Forchheimer flows and the PI for the Darcy
flows.
We show that the solution of the original problem can be
obtained by applying direct variational methods, which are
computationally less expensive than the grid based techniques.
In addition, by using the existing and newly developed
approaches in the theory of symmetrization, alternative "non-
analytical" algorithms are presented to assist optimal well
design, without repeatedly solving numerically the reservoir
simulation problem.
The approach will be demonstrated on practical well
optimization problems.
Introduction
To clarify the main results of this work we will review the
definition of the productivity index (PI) and underline some
assumptions about non-linear flows. Formally speaking, the
productivity index is one of the most basic characteristic of the
well performance and does not require any assumption about
the equations of the flow motion and the equation of state of
the fluid. The concept of productivity index expresses the
following fact: once the well production is, in some sense,
stabilized, then the ratio between the production rate and the
pressure drawdown (difference between the reservoir average
pressure and the well average pressure) is practically
independent from the production history or even from the
operating conditions. The higher is the values of the PI the
better are the performances of the reservoir.
Consider a bounded and isolated reservoir with no-flow on
outer boundary. A well producing with either constant rate or
constant pressure is characterized by the productivity index
defined in (Muskat, 1937; Larsen, 2001; Dietz, 1965; Dake,
1978 and Raghavan, 1991) as
Q(t ) .
PI (t ) =
ΔP(t )
Here Q (t ) is the production rate and ΔP(t ) = Pa (t ) − Pw (t ) is
the pressure drawdown, where Pa (t ) is the average reservoir
pressure and Pw (t ) is the flowing bottom-hole pressure. We
are particularly interested in the asymptotic (late time) value
of the PI. For constant production rate, Q (t ) = Q stabilization
means that the difference between the average and the well-
bore pressure (the denominator) becomes time invariant. This
flow regime is called pseudo-steady state (PSS). In the case of
constant wellbore pressure, both the numerator and
denominator change in time, but their ratio asymptotically
stabilizes to a constant value, leading to the flow regime called
boundary-dominated (BD) (see Muskat, 1937; Larsen, 2001;
Dietz, 1965; Dake, 1978; Raghavan, 1991; and references
therein).
In case of linear Darcy flow for slightly compressible
fluid, the most popular way to evaluate the productivity index
for constant thickness reservoir is based on the representation
of the PI as the following product PI = F0 J . Here
2π k h depends only on the fluid and on the porous
F0 =
Bμ
media properties together with the reservoir thickness. J is the
dimensionless productivity index which depends on the well-
reservoir geometry and on the type of flow regime. The PSS
2 SPE 108984
and the BD productivity indices are traditionally estimated by
the following equation
1 ,
J Dietz =
1 ⎛ 4A ⎞
ln⎜ ⎟+s
2 ⎜⎝ e γ C A rw2 ⎟⎠
where γ is the Euler’s constant, A is the drainage area, rw is the
wellbore radius and s is the skin factor. The shape factor CA
for computing the PSS PI can be obtained from Dietz (1965).
To compute the BD PI one can use shape factors in Helmy and
Wattenbarger (1998). In practice, for simple polygonal
domains with dimensionless radius RD > 1000 the same shape
factors are used in the evaluation of both PSS and BD PIs.
However, it is known that the BD PI of a well is, in general,
different from the PSS PI. In particular, the empirical evidence
from numerical calculations is that the PSS PI is always
greater or equal than the BD PI (Helmy and Wattenbarger,
1998). In our paper Ibragimov at al., 2005 it was rigorously
proved that this is a general statement. Namely, for any
reservoir/well geometry the PSS PI is always grater than the
BD PI. Detailed review of the traditional methods for
computing the productivity index was presented in our
previous publication, Ibragimov at al. (2004). The traditional
methods and techniques for evaluating the productivity index
impose serious restrictions on the geometry of the reservoir.
In papers Ibragimov at al. (2005 and 2004) we have
presented an alternative method for evaluating both the PSS
and BD productivity indices of a well. It is based on the
solution of the corresponding time independent (steady-state)
boundary value problems in opposition to transient methods. It
was shown that in a reservoir of volume V the dimensionless
PI for BD regime is equal to J BD = λ0V . Here λ0 is the first
eigenvalue of a certain mixed BVP problem for elliptic
differential equations.
The dimensionless PI for PSS regime is equal to
−1
J PSS = V ⎛⎜ ∫ u dV ⎞⎟ . Here u is the solution of a certain mixed
⎝ ⎠
BVP problem for elliptic differential equations.
From a reservoir engineering point of view these new
relations between the solutions of transient and steady state
problems allow direct use of powerful finite element/volume
methodology for computing the PI. These methods become
particularly useful in evaluating the productivity indices for a
well of arbitrary configuration in a general three-dimensional
flow situation, since the mentioned steady-state problems are
solved in a bounded domain corresponding to the actual
reservoir shape. The productivity indices of the well (PSS and
BD) are expressed in terms of the solutions of the steady-state
problems, which can be obtained numerically using well
established finite volume or finite element software packages.
The above results were presented at the SPE annual
conference, 2004, and it was suggested an extension of the
developed methods to non-Darcy flows. The current work –
partly motivated by this suggestion – reports the progresses we
have made in the modeling of the one phase non-linear flows
in porous media and indicate directions for future research.
The purpose of this paper is to emphasize the impact of the
non-linear nature of the flow filtration on the value of the
productivity index. In this paper we introduced a generalized
non-linear Darcy equation, with non-linear permeability tensor
depending on the gradient of pressure. This allows to model
different types of flows within one mathematical framework.
In particular it has been shown that there exist non-linear
permeability matrices such that the flows subjected to both the
“power” and the “two term” Forchheimer laws can be
modeled by introducing a generalized non-linear potential
flow. For PSS regime an analytical formula to compute the
productivity index is presented for above mentioned flows. In
case of “two term” law in circular reservoir an analytical
formula for the productivity index has been explicitly
obtained. For other types of geometries different numerical
simulations were performed by solving just the steady state
problem. The data obtained by these simulations allow making
a detailed analysis of the dependence of the PI on the
geometry of the reservoir/well system and on the parameters
of the reservoir. Some important results are summarized in the
conclusion.
1. Background
Darcy's law predicts the following common observation in
reservoir engineering: the higher is the production rate Q the
higher is pressure drawdown ΔP . The observed data shows
that during the dynamical process of hydrocarbon recovery the
ratio between these two quantities stabilizes in time to a
certain value. This value is, in general, a non-linear function of
both ΔP and Q . There are different ways to classify the
observed data. In Aulisa et al., 2007 we present two of the
most popular approximations of the field data: the “two term”
law, AQ + BQ 2 = ΔP , and the “power” law, CQ n = ΔP . Both
equations were originally introduced by Forchheimer in 1901
and in 1930, respectively.
Nowadays it is commonly believed that the inertial terms
cause the deviation from the linear Darcy law. It is important
to mention that the actual nature of these inertial forces is not
fully understood. As it was pointed by Bear, 1972 (see also Li
and Engler, 2001; Whitaker, 1996; Tavera et al., 2004), “Most
of the experiments indicate that the actual turbulence occurs at
a Reynolds number at least one order of magnitude higher
than the Reynolds value at which the deviation from the Darcy
law is observed”. Detailed discussion of these types of non-
linear approximation for the inertial terms, and for the
conditions when those approximations are applicable can be
found in Bear (1972), see also Li and Engler (2001), Whitaker
(1966), Tavera et al. (2004) and Muskat (1937).
In this paper we approximate the “power” and the “two
term” laws by assuming the coefficients of the permeability
matrix to be function of the pressure gradient. This allows
reducing the original system of equations, given by the
momentum balance, the equation of state, and the continuity
equation, to one non-linear parabolic equation for pressure
only. Using this model we obtained a compact and easy to
apply formula for the evaluation of the productivity index
similar to that obtained in the linear case.
SPE 108984 3

Figure 1: a) “Power” law CQ n = ΔP , b) “Two terms” law AQ + BQ 2 = ΔP

2. The generalized Darcy's Law In this article, we introduce a convenient framework by

There are different approaches for deriving the momentum
equation for the non-Darcy phenomena (Forchheimer, 1901;
Li and Engler, 2001; Bear, 1972 and references therein). The
equation of motion can be derived from the more general
hydrodynamic equation (see for example Pane and Straughan,
1999), or by using capillary models for describing the fluid
filtration through the porous media (see for example Dmitriev
et al., 2001; Whitaker, 1966; Li and Engler, 2001). In this
paper we will implement another more intuitive approach
based on the constitutive equation. Namely, we will build
independently two different vector equations for the “power”
and the “two term” Forchheimer laws. In the theoretical part
of the paper we will consider the general 3-D case because in
its generality. Let us denote by
⎛ k1,1 k1, 2 k13 ⎞
1⎜ ⎟ G
K = ⎜ k 2,1 k 2, 2 k 2, 3 ⎟
k⎜
⎜ k k k ⎟⎟
⎝ 3,1 3, 2, 3, 3 ⎠
the dimensionless, symmetrical, and positive definite
permeability tensor, here k is a normalization factor
characterizing the pivotal direction of the filtration. The
parameter k can be taken to be equal to some average value of
the permeability, (for example k = 3 k1,1k 2, 2 k3,3 , or
k = (k1,1 + k 2, 2 + k3,3 ) / 3 , etc).
G G
( )
Let Ab , b = ∑ ai , j bi b j be a bilinear form, where
i, j
G
b = (b1 , b2 , b3 ) is a generic vector and A = (ai , j ) is a matrix.
Let μ and ρ be the viscosity and the density of the fluid,
μ , and
α= β = β ρ , where β is the Forchheimer factor
k
Dake, (1978).
taking into account the anisotropies of the non-linear flow. It
is important to state that there are other possible ways to work
with anisotropy of the media. Some other techniques to derive
the constitutive equation for anisotropic porous media were
presented in Dmitriev et al, (2001), Whitaker (1996), and in
the references therein. Strictly speaking, even the linear
anisotropic Darcy equation can be derived in different ways.
In this article, following Aulisa et al., 2007, we use the two
following equations of motion:
G
( G G G
)
α v + β K −1 v , v v = − K ∇p , ..................................... (1)
for the “two term” law and
( G G
) G
c n K −1v , v n −1 v = − K ∇p , .......................................... (2)
for the “power” law.
Here v is the velocity of the filtration, p the reservoir
pressure, 1 ≤ n ≤ 2 , c is a constant depending on parameter n ,
and it is equal to α when n = 1 . ∇p = ( p x , p x , p x ) is the
1 2 3
pressure gradient, where the sub-indices ( x1 , x2 , x3 ) are used
to indicate the Euclidian coordinates.
Both equations share important features and can be
replaced by the generalized non-linear Darcy equation.
Namely, let us denote
1
η (∇p ) = ( K ∇p , ∇ p ) = ∑ k i, j p xi p x j . ........................ (3)
k i, j
We have showed that equation (1) is equivalent to equation
G
v = − M (η (∇ p ) K ∇ p , .................................................. (4)
where
4 SPE 108984
2 . ............................................(5)
M (η ) =
α + α + 4β η
2
We have also shown that equation (2) can be reduced into
G
v = − N (η (∇ p ) K ∇p , ...................................................(6)
where
( )
(1 − n ) / n
N (η ) = c −1 η .................................................... (7)
Clearly, equation (4) converges to the classical, anisotropic
Darcy equation as β → 0 . The same conclusion is obtained
by setting n = 1 into (7).
The above arguments allow introducing the following
Generalized Non-Linear Darcy Equation
G
v = −G(η (∇p) ) K ∇p ..................................................... (8)
Here the non-linear term G depends on the gradient of
pressure and is defined as G (η (∇ p )) = M (η (∇ p )) in case of
“two term” law, and G (η (∇p )) = N (η (∇ p )) in case of
“power” law.
3. Mathematical model for the productivity index of
the well
Assume the fluid to be “slightly compressible”,
ρ ′( p ) = γ ρ ( p ) , where the constant γ characterizes the
compressibility of the fluid. Using standard arguments about
slightly compressible liquids in the equation of continuity and
taking into account the generalized Darcy equation of the
motion we will arrive to the following non-linear diffusive
parabolic equation for pressure
∂p
γ = div (G (η (∇p ) K ∇p ) . ..........................................(9)
∂t
Note that equation (9) depends on the type of non-linearity
used for G (η (∇p )) but the construction of the forthcoming
formulation is formally the same for the two types of
Forchheimer flows.
Equation (9) characterizes the fluid (oil or gas) flow in the
reservoir generated by different regimes of well exploitation.
Several boundary conditions (see Raghavan, 1991; Ibragimov
et al., 2005 and 2004) can be used to simulate the
corresponding regimes of production. In our intended
application the exterior boundary of the reservoir is considered
impermeable. The boundary condition on the well, the non-
flux condition on the exterior boundary, and the initial
pressure distribution form the initial boundary value problem
(IBVP) for equation (9). At this moment we will not specify
any particular boundary conditions on the well.
Let U be the drainage region bounded by the well Γw and
by the exterior boundary Γe . Let p( x,0) = p0 ( x) be the given
initial pressure distribution in the reservoir U . Let us denote
with PU (t ) the average pressure in the reservoir U , and
with PW (t ) the average pressure on the well Γw .
Definition 1. (Productivity index). Let the pressure
G
function p ( x, t ) and vector velocity field v be subjected to the
generalized Darcy equation (8). Then the productivity index of
the well in the isolated drainage area U is defined as ratio
G Q . ................................................. (10)
J G ( p, v , t ) =
PU − Pw
Here Q is total flux over the well boundary Γw (i.e. Q is the
production rate).
Definition 2. (PSS regime) Let the production rate Q be time
independent. We will call the regime of production a pseudo-
steady state regime (PSS), if the corresponding pressure
drawdown (the difference between the pressure average in the
reservoir PU (t ) and the pressure average on the well PW (t ) ) is
constant.
Another important regime of production is the so called
boundary dominated (BD) regime. In this paper we will not
discuss the productivity index associated with this regime of
production referring to Ibragimov et al., 2005 and 2004, where
the BD problem was studied for the Darcy case. The focus of
this work is to study the time invariant (late time) productivity
index of the well for the non-linear Forchheimer flows. In the
case of BD regime the bottom hole pressure is fixed and the
rate of the production is vanishing in time. Therefore the
effects of the non-linearity in the generalized Darcy law
become negligible in time.
3.1 PSS Productivity Index
For short vertical and slanted wells the wellbore pressure can
be assumed to be uniformly distributed on the wellbore at any
time. If a horizontal well is assumed to have an infinite
conductivity, then the pressure is uniformly distributed on the
wellbore too. One should note that such assumption fails for
very long slanted wells. We will assume that at each moment
the pressure on Γw is uniformly distributed.
Let the well produce with constant rate Q , if the pseudo-
steady-state is reached then the value of the PSS PI is given by
G QV ...................................... (11)
J G ( p, v , t ) = J q ,G =
∫ q,G
U
w dV
Here wq ,G is the solution of the following steady-state
boundary value problem
div (G (∇wq ,G ) K ∇wq ,G ) = −
q ................................... (12)
,
V
wq ,G = 0 , .................................................................. (13)
Γw
∂wq ,G
G = 0 , ................................................................ (14)
∂n Γe
with non-flow boundary condition on the exterior boundary. In
the following we will refer to (12-14) as auxiliary boundary
SPE 108984 5

value problem. Note that p ( x, t ) = B − Q t + wq ,G ( x) is wq ,β

Γw
= 0 , .................................................................. (18)
γV
solution of equation (9), where the constant B is uniquely ∂wq ,β
G = 0 .................................................................. (19)
defined by the value of the initial reserves. ∂n Γe
The auxiliary BVP (11-14) is time independent (steady-
state), therefore the calculation of the PSS PI in formula (11) Formula (11) for the PSS PI evaluation is formally equal
has the advantage of a reduced CPU time if compared to the
G qV . .................................... (20)
original transient problem. J q ,β ( p, v , t ) = J q , β =
Equation (11) gives the value of the PSS PI, i.e., it is ∫ wq,β dV
explicitly assumed that the pseudo-steady state is reached U

from the beginning of production at time t = 0 . Once again to compute the PSS PI one should first solve the
auxiliary BVP (16-17). For most of the cases it can be done
3.2 Uniqueness of PSS PI only numerically, but for the axial symmetric case the PI can
The value of the PI provided in formula (11) is not the only be evaluated explicitly using an analytical formula.
possible value for the productivity index in a pseudo-steady
state. If the pressure is not uniform over the wellbore, then the 4.1. PSS PI for circular drainage area.
value of the pseudo-steady state productivity index can be This paragraph is dedicated to the classical reservoir
different from the one defined in formula (11). However, if the engineering problem: fluid filtration in a cylindrical reservoir
pressure is uniform on the wellbore, J q ,G is the unique value with fully penetrated vertical well located in its center, Figure
2.
of the PSS PI and formula (11) is its analytical representation.
To find the analytical formula for PSS PI for axial-
In many applications it is reasonable to assume that initially
symmetric “two” term Forchheimer flow we first derive an
(at time t = 0 ) the pressure is uniformly distributed in the
alternative formula for the PI in term of the radial velocity
reservoir. Under this assumption, on the admissible non-
linearity for the generalized Darcy law, we believe that the v β ,q ( r ) generated by PSS pressure distribution p ( r , t ) .
value of the PSS PI, as provided by equation (11), is the only qV 1 ........ (21)
attainable value of the productivity index in a pseudo-steady J q ,β = =
∫ wq,β dV
re
state regime. This result was proved for linear case, and was 2π ∫ α (vq , β ) 2 + Qβ (vq ,β ) 3 rdr
U
observed on all simulated data. In future we are planning to rw
proof this property for the transient non-linear solution,
rigorously.
The theoretical framework for PSS PI evaluation does not
depend on the type of non-linearity in generalized Darcy
equation: the “two term” or the “power” Forchheimer laws.
However the computational results and analytical solutions are rw
different for each of the two cases. In this paper we will re
present further analytical and numerical studies, only for
“two” term Forchheimer law.

4. The “two term” Forchheimer law for

homogeneous, isotropic reservoir
In case of homogeneous, isotropic reservoir the generalized Figure 2. 2-D scheme of the fully penetrated vertical well in a
Darcy equation takes a simpler form. To distinguish this case cylindrical reservoir.
we denote
2
G (∇p ) = f β ( ∇p ) = , .......................(15)
α + α + 4β ∇p
2
Next, we rewrote the auxiliary BVP (17-19) in polar
coordinates for the radial component of the velocity vector
2 2 2
here ∇p = px + p y + pz . Then the generalized Darcy
equation for “two term” Forchheimer law takes the following r −1
d
(rvq,β ) = − 2Q 2 , rw < r < re , ................... (22)
dr π (rw − re )
form
G vq , β ( re ) = 0.
v = − f β ( ∇p )∇p ..........................................................(16)
Correspondingly, the auxiliary BVP becomes The function

div ( f β (∇wq ,β ) ∇ wq ,β ) = −
Q .....................................(17) 2
Q ( re − r 2 ) ................................................ (23)
, vq ,β (r ) =
V 2π ( re − rw )
2 2
6 SPE 108984

solves the IVP (22). This expression can be used to derive the What is different and dependent from the parameter β is
productivity index in term of velocity. Namely the radial the pressure distribution. For the PSS regime the steady state
velocity is plugged into the alternative formula for J q , β in part of the pressure distribution in case of Darcy-Forchheimer
(21). After some analytical work and with the help of flow (solution of the BVP (17-18) is given by
Mathematica one can obtain the following explicit expression 2 2
Q(r 2 − rw − 2re ln(r / rw ) 2 )
for the PI of the fully penetrated vertical well in the cylindrical pq, β (r ) = −α +
4π (rw − re )
2 2
reservoir ....... (25)
2 3 2 4
Q (r rw + r rw + rrw − 6rrw re + 3re )
2 3 2
20πrw ( rw − re )(re + rw ) 2 (re − rw )α . .................(24) + (r − rw ) β
2 2
J q ,β = 12π 2 rrw (rw − re ) 2
2 2
G
Here The expression of the pressure explicitly depends on β and
Q , and it is obviously different for the Darcy and for the “two
G = 2(rw − re ) 3 (rw + 4rw re + 5re )Qβ +
2 2

term” Forchheimer cases. Difference in the pressure

5πrw ( rw − re )(rw + re ) 2 ( rw − 3re )α +
2 2
distribution can be very big for a quite large range of the
20πrw re (rw + re )α ln(rw / re )
4 parameters, Figure 3.

Below, we present a comparison between the analytical and 4.2. Properties and alternative formulae for the PSS PI for
the numerical results for the radial “two term” Forchheimer the “two term” Forchheimer law.
law for some particular values of the parameters. The results in Table 1 clearly indicate that for β Q = constant
The geometrical parameters are: the PSS PI has the same value. It appears to be a common
- the radius of the reservoir: re = 100 m ; property for the “two term” Forchheimer flows. Assume the
- the reservoir thickness: H = 10 m ; most general 3-D flow in the reservoir and consider the
- the well radius: rw = 0.3 m . following dimensionless variables and parameters
The fixed hydrodynamic parameters are: α
x = Lx′, wq ,β ( x ) = w′( x ′) , with L = V . ................ (26)
3

- α −1 = k / μ = 1.01× 1010 m 2 / (Pa ⋅ s ) , L

- γ = 0.658 × 10 −9 Pa −1 . The auxiliary problem (17-19) can be rewritten in its
The varying hydrodynamic parameters are: dimensionless form as
β = 2.4318×1011 , 2.4318×1012 , 2.4318 ×1013 Pa ⋅ s 2 /m3
Q / H = 10−4 , 10 −3 , 10−2 m 2 /s .

Table 1. Productivity index for the cylindrical reservoir, and fully penetrated vertical well.

Q / H = 10 −4 Q / H =10 −3 Q / H =10 −2
Analytic I Analytic I Analytic I
β = 2.4318×1011 0.1976 0.1976 0.1971 0.1971 0.1928 0.1928
β = 2.4318 ×1012 0.1971 0.1971 0.1928 0.1928 0.1582 0.1582
β = 2.4318 ×1013 0.1928 0.1928 0.1582 0.1582 0.05659 0.05659

Table 1 reports the numerical values of the well productivity div′( f ( ∇′wδ′ )∇′wδ′ ) = −1 , on U ' ................................ (27)
index versus the varying hydrodynamic parameters β and Q .
Column I contains the values of the PI obtained using the wδ′ Γ′w
= 0 , .................................................................... (28)
numerical solution of the BVP (17-18) in formula (20).
Column “Analytic” contains the PI values obtained with the ∂wδ′
analytical formula (24). G = 0 . ................................................................. (29)
∂n Γ′e
For each β and Q the difference is negligible. It is also
worth to mention that in the radial case the velocity fields for
the Darcy and the two terms Forchheimer flows are the same
and independent from the parameter β . Therefore by knowing
β and Q , the values of the PI for the two flows can be
obtained from each other just using algebraic relation.
SPE 108984
Figure 3. Pressure distribution in the radial reservoir:
Here div′ , ∇′ , U ′ , Γw′ and Γ′e denote the divergence, the
gradient, the drainage volume, the well boundary and the
exterior boundary in dimensionless coordinates. In (27)
f ( ∇′w′ ) =
2 , with δ = β Q ................(30)
1 + 1 + 4δ ∇′w′ α L2
Then the PSS PI can be rewritten in the following way
L ..........................................................(31)
J q ,β =
α ∫ wδ′ dV ′
U′
For a given reservoir and given parameter α = μ formula
k
(30) clearly states that the PSS PI depends only on the
parameter s = β Q .
Property 1. For a given constant value s = β Q , the PSS PI
depends only on α = μ , and on the geometry of the drainage
k the PI requires the solution of the auxiliary problem for each
area.
This result was originally obtained through numerical
simulation and has a clear physical interpretation: in the
Forchheimer equation the coefficient β and the rate of
production Q have the same impact on the value of the well
PSS PI. We also believe that it can be applied for the solution
of the following inverse problem: obtaining the hydrodynamic
characteristics of the reservoir by analyzing the production
data.
To highlight another important feature of the PSS PI for “two
term” Forchheimer law let us remind the PSS PI formula for
the corresponding Darcy case (Ibragimov et al., 2005)
7
A) Darcy Case ( β = 0 ),
B) Darcy-Forchheimer Case β = 2.4318 × 10 13 Pa ⋅ s 2 /m 3 .
V ,......................................................... (32)
J Darcy =
∫ 1,0
w dV
U
where w1, 0 is the solution of the corresponding linear auxiliary
boundary value problem
(
div α −1∇w1,0 = − ) 1 in U ......................................... (33)
V
,
w1, 0 Γ = 0 .................................................................... (34)
w
∂w1, 0
= 0 ................................................................... (35)
G
∂n Γe
Property 1 does not remove the existing gap between the
formula for the evaluation of the PSS PI for the Darcy (32)
and for “two term” Forchheimer (20) cases. As one can see in
(32), the PI does not depend on the production rate Q , and can
be computed by solving the auxiliary problem (25-26) only
once for a given U . In the non-linear case, formula (20) for
given production rate, Q . This is unfortunate, but it reflects
the non-linear nature of the fast Forchheimer flows.
In the next section the relation between J q , β for
different β and Q under some constraints about the flow are
analyzed. Note that in terms of β and Q the Darcy PSS PI is
equal to J Darcy = J 1, 0 .
4.3 Definition of the uniform flows, and an approximate
formula for the PSS PI
Henceforth we will use the dimensional formulation of the
well PI. The aim of this section consists in finding some
constraints on the flow that will overcome the obstacle
observed at the end of the previous section, and will allow to
obtain an alternative approximate formula for the PI. For this
8 SPE 108984

purpose formula (20) for PSS PI is rewritten in term of the All the simulations have been performed using Comsol
velocity vector. Assume once more that the production rate is Multiphysics 3.2. Time step and grid size have been refined
given and is equal to Q . until convergence in time and space has been reached.
Performed numerical simulations once more highlight the
Definition of the velocity field generated by the pressure: introduced constrains, and estimate the impact on value of the
G PI of the three leading parameters of interest: the rate of
We will call vq , β ( x, ) the velocity vector field generated by the
production Q , the Forchheimer coefficient β , and the
PSS pressure distribution p( x, t ) = B − Q t + wq , β ( x) if geometry.
γU Note that in all the computations the productivity index is
G
vector vβ , q and the pressure p are related by the equation: dimensionless. To convert the computed dimensional PIs to
the dimensionless ones (in Dake, 1978) we multiplied the
G G G
− ∇wq , β = αvq , β + β vq , β vq , β ........................................(36) dimensional PIs by the factor μ /( 2π k ) for the 2-D cases and
by the factor μ /( 2π kH ) for the 3-D cases, respectively.
From (36) it follows
The transient PI obtained by solving the original
G G 2 G 3 Forchheimer system converges and stabilizes to the value of
− ∇wq , β ⋅ vq , β = α vq , β + β vq , β . ................................(37)
the PI obtained by using the PI PSS regime formula for all
The LHS in (37) is the scalar product between two vectors. cases and for wide variety of the initial data.
Recollect equation (16) and that wq , β solves the auxiliary
problem (17-19). It can be proved that the following integral
equality is satisfied 5.1 Rectangular reservoir and fully penetrated vertical
well (2D-case)
q G 2 G 3 In this section we compare the results obtained with the
V U∫
wq , β dV = ∫ (α vq , β + β vq , β ) dV ...........................(38)
U
original transient Forchheimer system and those obtained with
both the steady state auxiliary BVP (17-19) and the steady
This relation allows rewriting the formula for the PI PSS in state auxiliary BVP (33-35) for the Darcy case.
term of the velocity vector field The reservoir domain is modeled as a 3-D rectangular box.
A fully penetrated vertical well is considered. Because of the
QV Q2 . .............(39)
J q,β = = G G 3 boundary conditions on the well and on exterior boundary, the
∫ wq,β dV
2

U U
∫ (α vq ,β + β v q , β ) dV problem reduces to the 2-D geometry, sketched in Figure 4.

G
The velocity field vq , β depends on Q but can be
approximated under some constraints.
G
Assume that the vector velocity field v q , β ( x, y , z , t )
uniformly depends on the rate of the production in the
L x 2 = 4000 cm

D=500
following sense:
G G
vq , β ( x, t ) = Q v1, 0 ( x, t ) ....................................................(40)
Under this assumption the following approximate formula can
be derived for PSS PI in term of velocity field for Darcy case
L x 1 = 8000 cm
1 ....................................(41)
J β ,q = G 2 G 3
∫ (α v + Qβ v1,0 ) dV Figure 4. 2-D Scheme of the fully penetrated vertical well in
1, 0 rectangular reservoir.
U

5. Numerical analyses and simulation

In this part we evaluate and compare the PI in “standard”
reservoir/well geometries for the following cases:
A. the transient productivity index, formula (10), is
computed solving the original Forchheimer system of
equations;
B. the time invariant productivity index, formula (20) for
PSS regime, is computed solving the steady state
auxiliary BVP (17-19);
C. the time productivity index, formula (41) for PSS regime,
is computed solving the steady state auxiliary BVP (33-
35) for the Darcy case.
In the numerical simulations we study the impact of the
hydrodynamic parameters on the value of the PI. Results in
Table 2 are organized as following. In column T the PI is
computed using the original transient equation after pseudo
steady state is reached. In Column I the PI is obtained using
the numerical solution of the BVP (17-19) in formula (20). In
column II the PI is obtained using the numerical solution of
the BVP (33-35) in the approximate formula (41).
The fixed geometric parameters are:
- the midpoint of the well;
SPE 108984
- the radius of the well rw = 0.3 m ;
- the height of the box H = 10 m ;
- Lx1 = 80 m , Lx2 = 40 m , Lw = 38 m and D = 5 m .
The fixed hydrodynamic parameters are:
- α −1 = k / μ = 1.01× 1010 m 2 / (Pa ⋅ s ) ,
- γ = 0.658 × 10 −9 Pa −1 .
The varying hydrodynamic parameters are:
- β = 2.4318 ×1011 , 2.4318 ×1012 , 2.4318 ×1013 Pa ⋅ s 2 /m 3 ,
- Q / H = 10 −4 , 10 −3 , 10 −2 m 2 /s .
The values of the PI (column T) for the solution of the
transient Forchheimer system after stabilization become
almost identical to values of the PI (column I) for PSS. The
numerical computation of the F-Transient values takes long
time, requires fine grid and adjustment of the initial data.
These justify the use of the Darcy-Forchheimer
approximation.
The deviations of the PI value between columns I and II is
very small for given values of the production rate and
Forchheimer coefficient β . And this once more highlights the
fact that for simple flow geometry around vertical well the
impact of Forchheimer phenomena at list at the level of the PI
can be handled algebraically from the solution of the
corresponding problem for linear Darcy case.
5.2 Two dimensional fracture in rectangular reservoir (2-D
case).
In this section we consider a horizontal well in a 3-D box,
which is modeled as a vertical fully penetrated fracture. The
vertical flow in z direction is ignored. This leads to a classical
geometry presented in Figure 5. Here the midpoint of the well
coincides with the midpoint of the reservoir. In the numerical
computation we compare the impact of the geometry and of
the hydrodynamic parameters on the value of productivity
index.
θ
Lx2 Lw
Lx1
Figure 5. Scheme of the two dimensional fully penetrated
fracture in rectangular reservoir
The fixed geometric parameters are:
- the midpoint of the well;
- the radius of the well rw = 0.3 m ;
- the height of the box H = 10 m ;
- Lx1 = 80 m , Lx2 = 40 m , Lw = 38 m .
The varying geometric parameter is:
9
- the angle ϑ = 0 o , 45o , 90o between the well and the vertical
axis.
The fixed hydrodynamic parameters are:
- α −1 = k / μ = 1.01× 1010 m 2 / (Pa ⋅ s ) ,
- γ = 0.658 × 10 −9 Pa −1 .
The varying hydrodynamic parameters are:
- β = 2.4318 × 1011 , 2.4318 × 1012 , 2.4318 × 1013 Pa ⋅ s 2 /m 3 ,
- Q / H = 10 −4 , 10 −3 , 10 −2 m 2 /s .
Results of computation of the well productivity index for
different values of the angle ϑ , for different values of the
Forchheimer coefficient β , and different values of production
rate Q are reported in Table 3. The PI values in column I are
obtained using formula (20), while the values in column II are
obtained using the approximate formula (41).
Results in columns I and II are very close for small values
of production rate Q and Forchheimer coefficient β . As Q
and β increase, the difference increases as well. The
maximum difference (about 5%) can be observed for the
horizontal well ( ϑ = 90 o ). It reflects the fact, that in this
configuration, vaguely speaking, the maximum violation of
the uniform flow constraint occurs.
5.3 Cylindrical reservoir and vertically deviated (slant)
well (3-D case).
In the last section we present numerical results for the study of
a 3-D cylindrical reservoir with well deviated from the vertical
axis (the so called slant well) Figure 6.
Figure 6. 3-D Scheme of the cylindrical reservoir with vertical
deviated (slant) well.
The fixed geometric parameters are:
- the radius of the reservoir re = 100 m ;
- the reservoir thickness H = 10 m ;
- the radius of the well rw = 0.3 m ;
- the length of the well Lw = 8 m .
The varying geometric parameter is:
- the angle ϑ = 0 o , 15o , 30 o , 45o , 60 o , 75o between the well
and the vertical axis.
The fixed hydrodynamic parameters are:
- α −1 = k / μ = 1.01× 1010 m 2 / (Pa ⋅ s ) ,
- γ = 0.658 × 10 −9 Pa −1 .
10 SPE 108984

The varying hydrodynamic parameters are: difference (about 5%), is observed for ϑ = 75 o . Vaguely
- s = β Q, with speaking, in this configuration the maximum violation of the
- β = 0, 2.4318 × 1012 , 2.4318 × 1013 Pa ⋅ s 2 /m 3 , uniform flow constraint occurs.
- Q = 8 × 10 − 2 m 3 /s

Table 2. Productivity index for the rectangular reservoir, and fully penetrated vertical well

Q / H = 10 −4 Q / H =10 −3 Q / H =10 −2
ϑ = 0D I II I II I II
β = 2.4318 ×1011 0.9537 0.9537 0.9535 0.9535 0.9516 0.9516
β = 2.4318 ×1012 0.9535 0.9535 0.9516 0.9516 0.9327 0.9327
β = 2.4318×1013 0.9516 0.9516 0.9327 0.9327 0.7779 0.7778
ϑ = 45 D I II I II I II
β = 2.4318 ×1011 1.2106 1.2106 1.2104 1.2104 1.2078 1.2078
β = 2.4318 ×1012 1.2104 1.2104 1.2078 1.2078 1.1826 1.1824
β = 2.4318×1013 1.2078 1.2078 1.1826 1.1824 0.9856 0.9768
ϑ = 90 D I II I II I II
β = 2.4318 ×1011 1.4940 1.4940 1.4937 1.4937 1.4900 1.4900
β = 2.4318 ×1012 1.4937 1.4937 1.4900 1.4900 1.4550 1.4542
β = 2.4318×1013 1.4900 1.4900 1.4550 1.4542 1.1970 1.1723

Table 3. Productivity index for the two dimensional fully penetrated fracture in rectangular reservoir

Q / H = 10 −4 Q / H =10 −3 Q / H =10 −2
T I II T I II T I II
β = 2.4318 × 10 11
0.1480 0.1483 0.1483 0.1477 0.1480 0.1480 0.1450 0.1452 0.1452
β = 2.4318 × 1012 0.1477 0.1480 0.1480 0.1450 0.1452 0.1452 0.1224 0.1225 0.1224
β = 2.4318 × 1013 0.1450 0.1452 0.1452 0.1224 0.1225 0.1224 0.0481 0.0480 0.0476

Table 4. Productivity index for the cylindrical reservoir with vertical deviated (slant) well.

ϑ = 0D ϑ = 15 D ϑ = 30 D
βQ I II Analytic I II I II
0 0.1988 0.1988 0.1976 0.2015 0.2015 0.2017 0.2017
8 × 2.4318 × 1010 0.1941 0.1941 0.1928 0.1967 0.1967 0.1969 0.1969
8 × 2.4318 × 1011 0.0593 0.0585 0.0600 0.0590 0.0697 0.0581
ϑ = 45 D ϑ = 60 D ϑ = 75 D
βQ I II I II I II
0 0.1974 0.1974 0.1878 0.1878 0.1693 0.1693
8 × 2.4318 × 1010 0.1927 0.1926 0.1834 0.1833 0.1655 0.1654
8 × 2.4318 × 1011 0.0590 0.0571 0.0573 0.0550 0.0531 0.0506

As one can see the results are very close for small values 6. Conclusions
of s . As s increases, the difference slightly raises as well. In this paper we studied fast compressible Forchheimer
The difference between the 2 columns also is going up as the (quadratic) flow in bounded porous media. The aim was to
angle ϑ increases, within given value of s . The maximum identify constraints on the flow, on the boundary, and on the
SPE 108984 11

initial data, ensuring the existence of the steady state invariant B = Coefficient, Pa·s2 / m6
for the dynamic solution of IBVP. Introduced assumptions for C = Coefficient, Pa· sn / m3n
G
the Forchheimer system, and the slight compressibility of the v = Velocity vector, m/s
fluids, reduce the original problem to the IBVP for scalar c = coefficient
quasi-linear parabolic equation for the pressure only. We M,N,G = Nonlinear functions
studied IBVP models of the fluid flow generated by the well U = Reservoir domain
and controlled by a given rate of the production. It has been x = Space variable
shown that the PI serves as a time invariant for a class of w = Solution of auxiliary problem
G
solutions specified on the boundary, and it can be explicitly n = Outward normal to the boundary of domain
evaluated by solving the auxiliary steady state problem. t = Time
The hereby obtained results reveal some interesting L = Distance, m
observations. We list some of them: s = Skin factor
- The velocity vector field in the Forchheimer equation can
be expressed as non-linear potential flow, as in equation Greek Symbols
(36).
α = Coefficient, Pa·s / m3
- There exists an initial pressure distribution such that the
β = Coefficient, Pa·s2 / m4
productivity index of the well is time invariant for the
corresponding solution of the IBVP. This result can be β = Forchheimer Coefficient, 1 / m
effectively utilized in reservoir engineering to compute γ = Coefficient, 1 / Pa
productivity index by solving the auxiliary steady state Γ = Reservoir boundary
BVP. φ = Porosity, fraction
- If the product between the Forchheimer coefficient β λ = eigenvalue
η = parameter, (Pa/m)2
and the production rate Q is constant then the well
μ = Viscosity, Pa·s
Productivity Index for the PSS regime is invariant with ρ = Density, kg/m3
respect to any variation of the production rate and the δ = Dimensionless parameter
Forchheimer coefficient.
- For a wide class of reservoir/well geometries and a wide Subscripts
range of hydrodynamical parameters there exists an
accurate enough approximate algebraic relation between V = Reservoir volume
the PI for the nonlinear Darcy-Forchheimer flows and the W = Well surface
PI for the Darcy flows. The approximation of uniform e = Exterior boundary
flow becomes less realistic when the product
s = βQ → ∞.
- Numerical results clearly indicate that the PI decreases as
References
the product s = β Q increases.
Aulisa, E., Cakmak, A., Ibragimov, A. and Solynin, A. 2007.
Variational Principle and Steady State Invariants for Non-Linear
Hydrodynamic Interactions in Porous Media. To appear in
Nomenclature Dynamics of Continuous, Discrete and Impulsive Systems
(Series A).
Latin letters Bear, J. 1972. Dynamics of Fluids in Porous Media. Dover
Publications, Inc., New York.
Q = Production rate, m3/s Dake, L.P. 1978. Fundamental in reservoir engineering, Elsevier,
ΔP = Drawdown, Pa Amsterdam.
P,p = Pressure, Pa Dietz, D.N., 1965. Determination of average reservoir pressure from
PI = Productivity Index build-up Surveys. JPT 17 (4): 955-959.
k = Permeability, md Dmitriev, M.N., Dmitriev N.M and Maksimov V.M. 2005.
h ,H = Thickness of the reservoir, m Representation of the Functions of the Relative Phase
J = Dimensionless PI Permeabilities for Anisotropic Porous Media. Journal Fluid
BD = Boundary dominated Dynamics 40(3).
PSS = Pseudo-Steady state Forchheimer, P. 1901. Wasserbewegung durch Boden Zeit. Ver.
R, r = Radius of the reservoir Deut. Ing. 45.
V = Drainage volume, m3 Helmy, W. and Wattenbarger, R.A. 1998. New shape factors for
K = Dimensionless permeability tensor wells produced at constant pressure. SPE-9970.
B = Formation volume factor, - Ibragimov, A.I., Khalmanova, D., Valko, P.P. and Walton, J.R. 2005.
A = Drainage area, m2 On a mathematical model of the productivity index of a well
A = Coefficient, Pa· s / m3 from reservoir engineering. SIAM J. Appl. Math., 65: 1952
A = Matrix of bilinear form, 1980.
12 SPE 108984

Ibragimov, A.I Khalmanova, D. Valko, P.P. and Walton, J.R. 2004.

Analytical Method of Evaluating Productivity Index for
Constant Production Rate or Constant Wellbore Pressure. SPE-
89935.
Larsen, L. 2001. General productivity models for wells in
homogeneous and layered Reservoirs. SPE-71613.
Li, D. and Engler, T.W. 2001. Literature Review on Correlations of
the Non-Darcy Coefficient. SPE-70015.
Muskat, M. 1937. The flow of homogeneous fluids through porous
media. McGraw-Hill Book Company.
Pane, L.E. and Straughan, B. 1999. Convergence and Continuous
Dependence for the Brinkman- Forchheimer Equations. Studies
in Applied Mathematics 102: 419 439.
Raghavan, R., 1991. Well Test Analysis, Prentice Hall, New York.
Tavera, C.A.P., Kazemi, H., and Ozkan E. 2004. Combine effect of
Non-Darcy Flow and Formation Damage on Gas Well
Performance of Dual-Porosity and Dual Permeability
Reservoirs. SPE-90623.
Whitaker S. 1996. The Forchheimer Equation: A Theoretical
Development. Transport in Porous Media 25: 27 61.