Sie sind auf Seite 1von 7

SPE 84879

Simulation of Non-Darcy Flow in Porous Media Including Viscous, Inertial and

Frictional Effects
H.A. Belhaj, SPE, K.R. Agha, S.D. Butt and M.R. Islam, SPE / Dalhousie University, Halifax, N.S., Canada

Copyright 2003, Society of Petroleum Engineers Inc.

This paper was prepared for presentation at the SPE International Improved Oil Recovery A number of physical and petrophysical properties are very
Conference in Asia Pacific held in Kuala Lumpur, Malaysia, 20–21 October 2003.
crucial to determine the economic value of any oil or gas
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
discovery. Among these properties, porosity, permeability and
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to reservoir pressure are the most critical parameters.
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 Consequently, a high degree of accuracy in estimating these
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
parameters is required.
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 The principle calculations of fluid flow in reservoirs is based
acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O.
Box 833836, Richardson, TX 75083-3836 U.S.A., fax 01-972-952-9435.
on the precise measurements and predictions of initial
reservoir pressure, current reservoir pressure and well bottom-
hole flowing pressure as they reflect the pattern of energy and
Abstract the potential of producing hydrocarbons from a certain
In this paper a new diffusivity flow equation has been derived reservoir. The transmissibility equation that is fundamentally
to describe fluid flow in porous media including both Darcian used in modern reservoir simulators to predict reservoir
and non-Darcian behaviors. This equation is based on the behavior depends mainly on pressure predictions at any point
fundamental Darcy’s equation, Forchheimer’s equation and of space and time during the life of the reservoir in addition to
Brinkman’s equation. The pressure gradient as predicted by the porosity and permeability of the pay zone. Properties of
the new diffusivity equation includes both viscous terms in reservoir fluids that are necessary for the calculations of
Darcy’s and Brinkman’s equations and the inertial forces term reserves, recovery factors and economic limits can be
in Forchheimer’s equation. Both Forchheimer’s inertial and determined with a great deal of accuracy once the reservoir
Brinkman’s viscous effects are expected to become significant has been discovered and fluid samples are obtained.
at high flow velocity due to the interactions between fluid
layers among themselves and with the media. A numerically The proposed method introduced herein is a new diffusivity
simulated model has been specifically developed based on the equation that not only considers Darcy’s1 equation but it
newly derived partial differential equation. The Crank- includes inertial and viscous forces terms of Forchheimer’s2
Nicholson approximation technique was successfully used to and Brinkman’s3 respectively. The proposed model is
model the newly derived diffusivity equation using suitable expected to represent fluid flow in reservoirs while both
boundary conditions. A wide range of fluid flow and porous Darcian and non-Darcian flow are taking place. The prime
media characteristics has been tested, and predictions of the target is addressing the matrix/fracture flow behavior in oil
numerical model showed very consistent results in all ranges. reservoirs while high flow rates in gas reservoirs are also
included. Keeping away from the elusion of modeling specific
An experimental laboratory program was designed to verify scheme of fractured porous medium system with imposed
the numerical model predictions. Comparison showed fractures frequency, orientation, spacing and aperture that do
excellent agreement between experimental data and the not reflect the real reservoir case, the proposed model is
numerical model predictions. The flow velocity versus general in its implementation and is applicable to all scenarios
pressure gradient profiles resulting from both the numerical that exist in oil and gas reservoirs.
and the experimental programs depicted a great deal of
compatibility. At the non-Darcy region “high velocity”, the It is understood, that the speed of flow (velocity) in porous
inertial forces tend to predict non-linear higher pressure drop media is the source of trouble to the currently used Darcy’s
than the Darcian linear prediction, while the frictional non model since its fundamental assumption is “limiting to low
linear effect holds the pressure gradient closer to the Darcian spatial velocity”4,5. Any variations of flow velocity in porous
trend. The effect of the shift upward from the Darcian linear media would be as a result of fluid properties, rock properties,
trend is much more significant than the shift downward caused pressure drop or any combinations of these factors change.
by the frictional effects. The findings of this study are Therefore, velocity has been used as an indicator of both fluid
expected to be applicable to both gas and oil reservoirs with and media characteristics variation.
no scale-up effort.
2 SPE 84879

Many experimental studies and field evidences show that Non- Substitute Equation (2) into Equation (5):
Darcy behavior would take place in both gas and liquid
reservoirs under practical conditions. It is no secret that the ρ  ∂P µ  ∂V ∂ρ ∂ 2 ρ ∂V ∂ρ ∂2 ρ
empirical Darcy law is no longer accepted to fully describe  + V + ρβV 2  + +V 2 + = −φ (6)
µ ′  ∂X K  ∂X ∂X ∂X ∂X ∂X ∂t∂X
fluid flow in porous media. This led the investigation of
modifying Darcy’s equation and deriving better alternative
governing equation that may represent fluid flow in Re-arrange:
porous media.
ρ ∂P ρµ ρ 2 β 2 ∂V ∂ρ ∂2 ρ
In this investigation, a practical and reliable model is + V+ V + +V +
presented to describe Darcian and non-Darcian fluid flow
µ ′ ∂X µ ′K µ′ ∂X ∂X ∂X 2
behavior in porous media. The model has been numerically ∂V ∂ρ ∂2 ρ
simulated and tested with very encouraging results. Although, = −φ
the model has been designed for single phase linear flow, ∂X ∂X ∂t∂X
modification of this model to cover multi-phase and both
radial and spherical flow is expected to be achieved with the ∂2ρ
Using chain rule for: we get:
same accuracy and less effort. ∂X 2
Governing Equation
∂2ρ ∂  ∂ρ  ∂  ∂ρ ∂P 
The pressure gradient in the suggested alternative form of =  =  =
equation is affected by the very well known viscous term in ∂X 2
∂X  ∂X  ∂X  ∂P ∂X 
Darcy’s equation, the inertial term in Forchheimer’s equation (8)
∂  ∂P  ∂2P
and the second viscous term in Brinkman’s equation that  ρC = ρC
describes frictional forces among the layers of flowing fluid ∂X  ∂X  ∂X 2
and between fluid and the media. The new diffusivity equation
is derived from the simplest form of this equation for single
phase and linear flow system. The pressure gradient in the X- As well for: we get:
direction can be expressed by the following equation: ∂t∂X

∂P µV ∂ 2V ∂2ρ ∂  ∂ρ  ∂  ∂ρ ∂P 
=− − ρβ V 2
+ µ′ (1) =   =   =
∂X K ∂X 2 ∂ t∂ X ∂t  ∂X  ∂t  ∂P ∂X  (9)
∂  ∂P  ∂2P
Re-arrange:  ρC  = ρC
∂t  ∂X  ∂ t∂ X
∂ 2V 1  ∂P µ 2 
=  + V + ρβ V  (2) ∂ρ
∂X 2 µ ′  ∂X K  And for: we get:

We have: ∂ρ ∂ρ ∂P ∂P
= = ρC (10)
∂X ∂P ∂X ∂X
∂ (Vρ ) ∂ρ
= −φ (3)
∂X ∂t 1 ∂ρ
C= (11)
ρ ∂P
∂V ∂ρ ∂ρ
ρ +V = −φ (4)
∂X ∂X ∂t
∂2ρ ∂2ρ ∂ρ
Substitute for , and from Equations (8), (9)
Take the derivative of both sides of Equation (4) with respect ∂X 2
∂t∂X ∂X
to X: and (10) respectively into Equation (7):

∂ 2V ∂V ∂ρ ∂ 2 ρ ∂ρ ∂V ∂2ρ ρ ∂P ρµ ρ 2β 2 ∂V ∂P
ρ + +V + = −φ (5) + V + V + 2 ρC +
∂X 2 ∂X ∂X ∂X 2 ∂X ∂X ∂t∂X µ ′ ∂X µ ′K µ′ ∂X ∂X (12)
∂ P 2
∂ P 2
VρC = − φρ C
∂X 2
∂ t∂ X
SPE 84879 3

Effective viscosity “ µ ′ ” is defined as: 1  ∂P ∂P  2 2 ρβ 2

−  + − V − V −
C µ  ∂ X ∂ Y  CK Cµ (19)
µ′ = , substitute into Equation (12): V  ∂2P ∂2P  ∂  ∂P ∂P 
φ  +  =  + 
φ  ∂ X 2 ∂ Y 2  ∂t  ∂X ∂Y 
ρφ ∂ P ρφ ρ 2 βφ ∂V ∂P
+ V + V 2
+ 2 ρC + Numerical Modeling
µ ∂X K µ ∂X ∂X (13) The mathematical solution for the partial differential equation
∂2P ∂2P (Eq. 19) is very difficult to achieve. Solving this equation
VρC = − φρ C numerically is an obvious alternative. First, Equation (19) has
∂X 2 ∂ t∂ X to be put in a dimensionless form. It contains four partial
derivatives of pressure with respect to X and Y coordinates,
Divide both sides of Equation (13) by − φρ C : two of first order and two of second order. It also contains
another two second derivatives of pressure with respect to
1 ∂P 1 ρβ 2 2 ∂V ∂P both time and X and Y coordinates alternatively. Other
− − V− V − − parameters are considered constants at a particular set of
Cµ ∂X CK Cµ φ ∂X ∂X experiment scheme.
V ∂2P ∂2P
= The six mentioned partial derivatives have been transformed
φ ∂X 2 ∂t∂X to finite differences. The fully implicit numerical solution of
Carnck-Necoloson6,7 method has been employed to solve
∂V Equation 19 using suitable boundary and initial conditions as
Again, using chain rule for we get: shown in Figure 1. The solution is considered only for a
steady-state condition. A grid system of 11 x 14 has
∂V ∂V ∂P been used.
= (15)
∂X ∂P ∂X Results and Discussions
The definition that has been adopted in this study is that any
Substitute in Equation (14): behavior that does not follow Darcy’s linear trend is defined as
non-Darcy regardless of the cause of that non-linearity. The
1 ∂P 1 ρβ 2 2 ∂V  ∂P  non-Darcian behavior as such can be caused by a single effect
− − V− V −   − or a combination of effects. Forchheimer2 added to Darcy’s
Cµ ∂X CK Cµ φ ∂P  ∂X  viscous term an inertial term that he believed is responsible for
V ∂2P ∂2P the non-linearity behavior associated with fluid flow in porous
= media at high flow rates. Later, his model has been used in gas
φ ∂X 2 ∂t∂X reservoirs. Brinkman3 was no different, he depicted that
Darcy’s law is valid if another viscous term is added to it. In
∂P  ∂P  this study, the authors believe that all terms should be included
is a small quantity and   is even smaller, therefore: since the effect of Darcy’s term is crucial at low spatial
∂X  ∂X 
velocity but, as flow velocity increases the effect of the other
 ∂P  two terms become significant. As shown in Equation (1), the
  ≈ 0 , substitute in Equation (16): pressure drop is affected by the three terms of Darcy,
 ∂X 
Forchheimer and Brinkman. Darcy’s and Forchheimer’s tend
to cause more pressure drop and Brinkman’s pushes to the
1 ∂P 1 ρβ 2 V ∂ 2 P ∂ 2 P opposite effect. Figure 2 shows the role of each of the three
− − V− V − = (17)
Cµ ∂X CK Cµ φ ∂X 2 ∂t∂X terms included in the proposed model.

Equation (17) is the diffusivity equation in the X-direction, for Figure 3 shows a comparison between the predictions of
the Y-direction we get: Darcy’s diffusivity equation model versus the suggested
model in Equation (19). It is very clear that at low velocity
both models prediction is almost identical, but as the flow
1 ∂P 1 ρβ 2 V ∂ 2 P ∂ 2 P
− − V− V − = (18) velocity increases the proposed model predicts higher pressure
Cµ ∂Y CK Cµ φ ∂Y 2 ∂t∂Y gradient than Darcy’s model. This fact suggests that the
proposed model would be favorable in representing flow
The summation of Equations (17) and (18) gives the behaviors in gas reservoirs, highly fractured reservoirs and in
diffusivity equation in two dimensions: the vicinity of the wellbores where high flow rates
are expected.
4 SPE 84879

Further verification of the proposed model can be inferred by direction of the other two terms (lower pressure gradient) and
comparing its predictions with another comprehensive model that this term expresses the frictions between fluid layers and
based on the Navier-Stokes equation that has been derived and fluid and media makes it very important when multi-phase flow
numerically modeled to represent fluid flow in porous is encountered. The authors believe that having more than one
media8,9. This model includes all of the terms mentioned in the fluid flowing in the reservoir simultaneously, which is the case
basic equation (Eq. 1) that our diffusivity equation (Eq. 19) in most reservoirs or during reservoir injection process, the
has been derived from in addition to another convective term frictional term (Brinkman’s term) role becomes more
drawn from the original Navier-Stokes equation. The significant. The authors also believe that the phenomenon of
comprehensive model predicts flow velocity in the X-direction viscous fingering and channelling that take place during
at any time using the following equation: miscible and immiscible displacement in oil reservoirs can be
explained and understood by implementation of the frictional
∂U µ ∂ 2U µφ viscous term of Brinkman.
= − U − βφ U 2

∂t ρ ∂Y 2
ρK (20) An experimental program has been run in the laboratory
1  ∂U ∂U  φ ∂P specifically to compare its results with the proposed model
U +V − predictions. Real cut samples from two outcrops of prospective
φ  ∂X ∂Y  ρ ∂X
formations in the Windsor area, Nova Scotia, Canada have been
tested. Figures 5, 6, 7 and 8 show how the experimental results
In the Y-direction, the flow velocity at any time can be
compared with numerical model predictions in terms of flow
predicted by this model:
velocity versus pressure drop. It can be easily seen from these
plots that the numerical model proposed in this study matches
∂V µ ∂ 2V µφ fairly well with the experimental results suggesting serious
= − V − βφ V 2

∂t ρ ∂X 2
ρK (21) consideration of field implementation.
1  ∂V ∂V  φ ∂P
U + V  − Conclusions
φ  ∂X ∂Y  ρ ∂Y Lots of criticism has been reported against the use of Darcy’s
equation to represent all aspects of porous media flow. In this
If we drop the convective term  ∂U ∂U  and study rather than addressing particular case, a general solution
U +V 
 ∂ X ∂Y  has been introduced through the use of a single model capable
 ∂V ∂V  from the comprehensive model expressed by of representing both Darcian and non-Darcian flow in porous
U +V 
 ∂X ∂Y  media. The proposed numerical model includes the
Equation (20) and Equation (21) respectively, then the resultant conventional Darcy’s viscous term in addition to
model should be equivalent and gives the same solution of the Forchheiemr’s inertial term and Brinkman’s frictional term.
diffusivity model introduced by Equation (19). Having done
that, Equations (20) and (21) become: The proposed model has been compared to Darcy’s model and
another comprehensive model based on Navier-Stokes
equations. Predictions of the proposed model match Darcy’s
∂U µ ∂ 2U µφ φ ∂P
= − U − βφ U 2
− (22) model predictions at low spatial velocity but in high flow
∂t ρ ∂Y 2
ρK ρ ∂X velocity it predicted higher pressure gradient than Darcy’s to
account for the non-Darcian flow effect. Meanwhile, almost
identical prediction has been noticed when compared with the
∂V µ ∂ 2V µφ φ ∂P comprehensive model. The proposed model has been tested
= − V − βφ V 2 − (23)
∂t ρ ∂X 2
ρK ρ ∂Y against experimental data and shows excellent agreement.

The proposed model is ready for use in single phase cases

If we prove that both models are equivalent then we will have although it should be modified to suite multi-phase flow in
more faith in our model described by Equation (19).
reservoirs. It is believed that replacing Darcy’s model with the
suggested model in this study will enhance the flow behavior
Figure 4 shows a comparison between predictions of the predictions specially, when non-Darcy flow is expected.
proposed model (Eq. 19) and the remaining of the
comprehensive model represented by Equations (22) and (23).
A good agreement can be noticed clearly. This proves that the Financial support of this research is funded collectively by the
proposed comprehensive model introduced in this study (Eq.
Natural Science and Engineering Research Council of Canada,
19) is a better alternative to represent Darcian and non-Darcian the Atlantic Innovation Fund and the Killam Trustees.
flow in porous media including both matrix and fractures
contributions to flow simultaneously.

The frictional term introduced in this equation has the least

effect compared to other terms in the equation. But, the fact that
this term is driving pressure gradient prediction to the opposite
SPE 84879 5

Nomenclature through Porous Media,” Trans. AIME (1953),

P Pressure 198, 79.
x Space Coordinate in flow direction 6. Patankar, S.V.: Numerical Heat Transfer and Fluid
y Space Coordinate perpendicular to flow direction Flow, McGraw-Hill, (1980).
β Non-Darcy coefficient 7. Kobayashi, M.H. and Pereira, J.C.F.: “A Comparison
t Time of Second Order Convection Discretization Schemes
µ Viscosity of flowing fluid for Incompressible Fluid Flow,” Communication in
ρ Fluid density Numerical Methods, 12, (7), 395-411, July 1996.
U Volume-Averaged flow velocity in the x-direction 8. Belhaj, H.A., Agha, K.R., Butt, S.D. and Islam, M.R.:
V Volume-Averaged flow velocity in the y-direction “'A Comprehensive Numerical Simulation Model for
β Non-Darcy coefficient Non-Darcy Flow Including Viscous, Inertial and
µ ′ Effective viscosity of flowing fluid Convective Contributions',” paper SPE – I, proc.,
C Isothermal compressibility factor presented at the 27th Annual SPE International
Technical Conference and Exhibition, Abuja,
K Permeability of the porous system
Nigeria, August 4-6, 2003.
φ Porosity
9. Belhaj, H.A., Agha, K.R., Nouri, A.M., Butt, S.D.,
Vaziri, H.F. and Islam, M.R.: “Numerical Modeling
References of Forchheimer’s Equation to Describe Darcy and
1. H.P.G. Darcy,: “Les Fontaines Publiques de villle de Non-Darcy Flow in Porous Media,” paper SPE
Dijon, Exposition et Application des Pricipes a 80440, proc., presented at the SPE Asia Pacific Oil
Suivre et des Furmules a Emplyer dans les Questions and Gas Conference and Exhibition, Jakarta,
de Distribution d’Eau,” Victor Dalmont, Paris, 1856. Indonesia, April 15-17, 2003.
2. Forchheimer, P.: “Wasserbewegung durch Boden,”
ZVDI, 45, pp. 1781, 1901. SI Metric Conversion Factors
3. Brinkman, H.C.: “A calculation of the viscous force To Convert from To Multiply by
exerted by a flowing fluid on a dense swarm of Ft m 0.3048
particles,” Appl. Sci. Res. A, 1, pp. 27-34, 1947. md m2 9.869E-16
4. Vafai, K. and Tien, C.L.: “Boundary and inertia psi kPa 6.894757
effects on flow and heat transfer in porous media,” lbm/ft3 kg/m3 16.01845
Int. J. Heat Mass Transfer, 24, pp. 195-203, 1981. lbm/ft-sec kg/m-sec 1.488163
5. Bruce, G.H., Peacceman, D.W., Rachford, H.H. and
Rice, J.D.: “Calculations of Unsteady-State Gas Flow

Y=H ∂Y

1,m ∆Y n,m

P = Pi 1,j+1 ∆ P = Po
1,j i,j
1,j -1

1,2 X
1,1 2,1 ∂P i-1,1 i,1 i+1,1 n,1 X=L

Fig. 1 Schematic of the system considered with boundary conditions

(The numerical grid used in this study: n =11 and m= 14)
6 SPE 84879

Darcy's viscous term

Pressure Gradient, MPa/m

Forchheimer's inertial term
Brinkman's frictional term



0 200 400 600 800 1000 1200 1400

V, cm/sec

Fig. 2 Contribution of the three terms included in the proposed model

ρ = 1 gr/cc
µ = 0.01 gr/cm sec
k = 100 md
φ = 0.236
Pressure Gradient, MPa/M

40000 β = 485747.85 m-1


0 50 100
V, cm/Sec

Proposed Model Darcy Model

Fig. 3 Comparison between the predictions of Darcy’s model against the proposed model

r = 1 gr/cc
Pressure Gradient, MPa/m

1.E+07 m = 0.01 gr/cm sec

k = 100 md
f = 0.236
6.E+06 b = 485747.85 m-1



0 10 20 30 40 50 60 70
V, cm/sec

Equivalent model -- Equations 22 & 23 Proposed Model -- Equation 19

Fig. 4 Comparison between the proposed model and the model represented by Eqs. 22 & 23
SPE 84879 7

1.E+04 12000
Sample 1 Sample 2

Pressure Gradient, kPa/m

Pressure Gradient, kPa/m


5.E+03 6000



0.E+00 0
0 500 1000 1500 2000 2500 0 500 1000 1500 2000 2500 3000
V, cm/sec V, cm/sec

Numerical Model Experimental Data Numerical Model Experimental Data

Fig. 5 Numerical predictions versus experimental Fig. 6 Numerical predictions versus experimental
results, Sample # 1. results, Sample # 2.

18000 8000
Sam ple 3 Sample 4
16000 7000
Pressure Gradient, kPa/m

Pressure Gradient, kPa/m


4000 2000

2000 1000

0 0
0 5000 10000 15000 20000 25000 0 50 100 150 200 250 300 350
V, cm/sec
V, cm/sec

Numerical Model Experimental Data Numerical Model Experimental Data

Fig. 7 Numerical predictions versus experimental Fig. 8 Numerical predictions versus experimental
results, Sample # 3. results, Sample # 4.