SPE 93053
Multilevel Fracture Network Modeling of Naturally Fractured Reservoirs
H. Kazemi, SPE, S. Atan, SPE, M. Al-Matrook, SPE, J. Dreier, SPE, and E. Ozkan, SPE, Colorado School of Mines
Copyright 2005, Society of Petroleum Engineers Inc.
This paper was prepared for presentation at the 2005 SPE Reservoir Simulation Symposium
held in Houston, Texas U.S.A., 31 January 2005 2 February 2005.
This paper was selected for presentation by an SPE Program Committee following review of
information contained in a proposal 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 a proposal of not more than 300
words; illustrations may not be copied. The proposal must contain conspicuous
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.
Abstract
This paper provides a review of common approaches for
simulation of naturally fractured reservoirs and a new model
formulation that is more amenable to the utilization of detailed
geologic information from deterministic models, multipoint
statistical simulations (MPS), and discrete fracture network
(DFN) models. Unlike the common dual-porosity and dual-
permeability models, the new model considers flow in several
sets of fractures (that is, micro, macro, and mega fracture
levels) and the wells can intercept any class of fractures.
Matrix flow is also included in the formulation. A dual-mesh-
computing algorithm is used to capture the major orientation
of fractures in the flow network. The algorithm consists of the
conventional five-point discretization for the coarse grid and a
special nine-point discretization scheme for the fine grid. The
size of the coefficient matrix for the discretization scheme can
be reduced because of the dominance of vertical flow drainage
in fractured reservoirs.
Introduction
Geologic reservoir characterization is the most crucial first
step in construction of a credible reservoir flow model for
naturally fractured reservoirs. The most notable approaches
for constructing a geologic model include the classical
deterministic methods, where geologists make the best
interpretation from existing data and build a model of the
reservoir. In the last several years, however, the deterministic
approaches have been complemented by quantitative
geostatistical approaches such as multipoint statistical
simulation (MPS) and discrete fracture network (DFN)
modeling.
The MPS specifically generates a depositional and
lithofacies model of a reservoir by incorporating the
geological architecture and properties of the rock fabric and is
touted as a very promising stepping stone in constructing
viable numerical reservoir models.1,2 For reservoir modeling
purposes, however, the stratigraphic flow units and fracture
flow paths and network must be identified and included in the
depositional model.3 Therefore, the DFN modeling has
become popular and is often used to model the fracture flow
path and connectivity.4 However, both the deterministic and
geostatistical techniques are highly subjective and, while such
fundamental geologic modeling techniques are often the
necessary starting points, any revisions toward the
construction of an ultimate reservoir model depend on
calibration against reservoir performance as well as carefully
designed flow tests. Thus, in the calibration process, one must
construct a flow model, which is also consistent with the
geologic model. This is the ultimate objective of this paper.
Below, we first present an overview of the tools, models,
and information used in the characterization and numerical
simulation of naturally fractured reservoirs. Then, we
introduce a new model formulation that can incorporate
detailed geologic information. The discussion of the results
follows.
Overview
Characterization of naturally fractured reservoirs and
simulation of flow in fractured porous media requires
information obtained and calibrated with different tools and
methods. Understanding the tools and methods involved in the
process is therefore essential for the correct interpretation of
the final characterization and simulation models.
Significance of Fractures
Fractures may affect the oil and gas production mechanism in
many ways. They can enhance production if utilized properly,
as in gas-induced gravity drainage of oil by careful monitoring
and steering of the gas-oil contact in the fractures; or they can
adversely affect oil production if the flow paths are misaligned
in injection-production well pattern in waterflooding and gas
injection applications.
To improve oil and gas recovery in naturally fractured
reservoirs, the dominant flow paths must be identified and
appropriate production schemes must be devised. Identifying,
characterizing, and mapping the fracture network in terms of
aperture, length, height, connectivity, conductivity, and
frequency distribution are crucial for optimal reservoir
management. Equally important is the selection and design of
appropriate experiments (as in special core analyses and
centrifuge-based measurements) to be conducted on the matrix
rock samples to determine the interdependence between the
matrix rock and fractures.
2 SPE 93053

To demonstrate the broad variation of fracture

permeabilities that could exist in petroleum reservoirs, we
resort to Fig. 1, taken from Ref. 5, which shows tracer
breakthrough times in four wells in a confined 15-acre five-
spot pattern in a naturally fractured carbonate reservoir. The
breakthrough times are from seven to thirty days in the four
producers for a travel distance of about 540 feet between the
injector and producer well pairs. The simulation results using
an areal 15 x 15 x 1 grid, with x = y = 54.4 feet, resulted in
directional fracture permeability pairs, k fx and k fy , of 500 and
2500, 34000 and 28000, 3500 and 500, and 17000 and 10000,
in milidarcy, for the drainage areas surrounding Wells A, B,
C, and D, respectively. The fracture and matrix porosities
were 0.01 and 0.18; matrix permeability was 30 md, and water
and oil viscosities were 0.9 cp and 8 cp. These results also 7
Fig. 3 An illustration of natural fractures in Midale reservoir.
illustrate the importance of dynamic flow tests in calibration
of flow models. Fig. 2, taken from Ref. 6, presents another example of
diversity of the effective directional permeability magnitudes
and ratios in a Rocky Mountain field. For this field, the
effective principal permeability ratios varied from five to
twenty-seven in three separate clusters of wells. In these types
of reservoirs, pressure-transient interference tests are useful to
estimate the effective directional permeabilities.
Fig. 3 shows a schematic of fractures from Midale
carbonate reservoir7 and Fig. 4 displays a DFN representation
of multilevel fracture sets in a quartzite reservoir.8 These
examples demonstrate the complexity of fracture morphology
in real reservoirs and hence highlight the difficulty of
constructing viable numerical models for naturally fractured
reservoirs.

Fig. 1 Tracer performance in a four well confined five-spot

5
pattern in a naturally fractured reservoir.

8
Fig. 4 Multilevel fractures in a quartzite reservoir.

Reservoir Characterization
Reservoir characterization usually starts with analog
reservoirs, outcrops, and seismic data. The next step is the use
Fig. 2 Pressure interference results in a waterflood project.
6 of well logs, core cuttings, sidewall cores, and wellbore
images to develop a preliminary description of the reservoir.
For naturally fractured reservoirs, the discrete fracture
network (DFN) modeling has become a desirable starting
SPE 93053
point, but calibration of such models against short- and long-
term pressure transient tests is most crucial. In fact, field
experience has shown that the ultimate accuracy of the
reservoir characterization will depend on the iterative process
of calibration against dynamic data from several wells as new
data become available.
Pressure-Transient Testing
Pressure-time data from pressure transient tests (both pressure
buildup and interference) are very useful in characterizing
naturally fractured reservoirs. These tests are based on the
solution of the pressure diffusion equation satisfying
appropriate boundary conditions. The time domain of interest
is from few seconds to few hours to few days only. In
comparison, the time domain of interest for reservoir
performance modeling is from few days to few months to
several years. The pressure diffusion equation represents
single-phase (or equivalent single-phase) compressible flow.
The gravity and capillary force components of flow do not
play negligible roles in the propagation of the pressure pulse.
In contrast, the gravity and capillary forces play a very
prominent role in long-term reservoir performance behavior.
Thus, equations used both for transient testing and
performance forecasting of naturally fractured reservoirs will
be given later.
Tracer Testing
Tracer testing is based on the arrival time of the tracer in the
injected fluid; thus, it is used as a quantitative guide to
determine preferential flow paths usually caused by the
presence of fracture channels. In naturally fractured reservoirs,
a tracer injected in a given well flows preferentially in the
fracture channels but part of it will be retained by flow into the
matrix, which causes a chromatographic delay in the
breakthrough time of the tracer in the production wells. Thus,
the tracer transport must account both for the fracture and
matrix flows. Furthermore, the transport equations for the
tracer must preserve the integrity of the physical dispersion as
compared to the grid-dependent numerical dispersion. This
issue can be properly handled with the use of the fine grids
proposed in this paper.
Reservoir Modeling
The classical reservoir modeling approach uses two distinct
continua to describe a naturally fracture reservoir, leading to
the customary dual-porosity/dual-permeability models.9
However, the fracture width and length may have fractal
distributions as evidenced by outcrop measurements. Thus, a
good model should utilize a multilevel approach for fracture-
matrix distribution.10,11 This model should have a nested
arrangement of varying size fractures and matrix blocks,
where micro, macro, and mega size fractures are imbedded in
the matrix rock and are separated by matrix blocks.
Furthermore, the wellbore (horizontal, vertical, or slanted)
should be considered as an extension of the fracture network
and the reservoir model should correctly connect the wellbore
to the appropriate intercepting fracture sets.
Historically, the dual-porosity model (using sugar-cube
approximation for the matrix) has been a very good
educational first step for modeling naturally fractured
3
reservoirs; however, it has not been very practical. Specialized
dual-porosity/dual-permeability models have provided better
representations of the petroleum reservoirsespecially in
multiphase flow situations.9 The flow models constructed
around the DFN reservoir characterization are natural
extensions of the classical dual-porosity approach, but the
fracture characterization demand for such models is extensive
and requires calibration against flow measurements in the
field.8,10,11 Finally, there is a strong interest in an appropriate
level of upscaling reservoir data from the fine scale to a
computationally accurate coarse scale. This can be
accomplished effectively with the use of dual-mesh and
streamline-based reservoir models.12,13 These models use very
fine computational grid, similar to the mesh size used in
geocellular reservoir characterization models.
Reservoir Flow Model
In this section, we present the underlying observations and
ideas, which lead to the two distinct flow models of fractured
reservoirs presented in this paper. For both models, we start
with the observations of the fracture morphology in outcrops,
assuming outcrop can be a realistic analog of fractured
petroleum reservoirs. The first significant finding from
outcrop measurements along a scan line is that fracture width
(aperture) has a power-law fractal distribution (Fig. 5). It is
also known that the wider fractures are both longer and have
larger spacing, which can lead to additional fractal
distributions.
Fig. 5 An idealization of cumulative frequency vs. width/spacing
of a fracture system.
Flow Model I
Because we cannot account for all fractures, we divide a given
fractal distribution into three classes (ensembles) as in Fig. 5.
For each class, we determine the average width w f and,
consequently, the average separation ls . As an example, we
choose ls1 , ls 2 , and ls3 to be 20 ft, 1.0 ft, and 0.05 ft,
respectively, where subscripts 1, 2, and 3 refer to mega,
macro, and micro fractures, respectively. Then, based on
physical considerations, we assume that micro fractures are
connected to matrix and macro fractures, macro fractures are
connected to matrix, micro fractures, and mega fractures, and
mega fractures are connected to matrix and macro fractures.
4 SPE 93053

This connectivity hierarchy can be changed to other physically
realistic scenarios.
Flow Model I treats each fracture set and the rock matrix
as separate continuum. While this model can be used for
single- and multi-phase problems, it is more amenable to
single-phase applications. Nonetheless, to realistically account
for the various fracture orientations and connectivity, a special
perpendicular bisector (PBI) grid system (Fig. 6) was
constructed, which leads to a computationally efficient
structured coefficient matrix for the pressure solution (Fig. 7).
multiphase flow problems in naturally fractured systems. The
approach we chose is very similar to the technique used in the
streamline simulation approach but without the use of
streamlines.12 In essence, a sequential technique is used to
solve for the pressure surface using fluid total mobility. Then,
the pressure surface is used to compute the velocity field to be
used in the convection flow calculations of various phases and
components.
Model I Equations
For this formulation, the three-level fracture system chosen
earlier leads to the following equations for a single-phase
slightly compressible fluid.

Single-Phase Case: The single-phase flow equations are

given as follows.

Mega fractures:
kf1 pf1
i ( pf1 D ) m / f1 f 2 / f1 + qf1 = ( ct )f1 ........... (1)
t
Macro fractures:
kf 2 pf 2
i ( pf 2 D ) m / f 2 f 2 / f1 f 3 / f 2 + qf 2 = ( ct )f 2 (2)
t
Micro fractures:
kf 3 pf 3
i ( pf 3 D ) m / f 3 f 3 / f 2 + qf 3 = ( ct )f 3 ......... (3)
t
Matrix:
km pm
Fig. 6 Perpendicular bisector grid system for simulation of i ( pm D ) + ( m / f 3 + m / f 2 + m / f1 ) + qm = ( ct )m (4)
natural fracture reservoirs.
t
Where,
kf 2
f 2 / f1 = f 2 / f1 ( pf1 pf 2 ) ................................................. (5)

kf 3
f 3/ f 2 = f 3/ f 2 ( pf 2 pf 3 ) ................................................ (6)

km
m / f1 = m / f1 ( pf1 pm ) ................................................... (7)

km
m/f 2 = m/f 2 ( pf 2 pm ) .................................................. (8)

km
m/f 3 = m/f3 ( pf 3 pm ) .................................................. (9)

Two-Phase Case: For two-phase water-oil flow, the

equations are as follows.
Fig. 7 Matrix equation for pressure solution in a three-level
fracture system.
Mega fractures:
S wf1
Flow Model II ikf1wf1 ( pwf1 w D ) wm / f1 wf 2 / f1 + q wf1 = f1 (10)
This model treats the entire fracture system and the rock t
matrix as two separate interacting continua. The fracture
system is a compilation of all fractures contained in various The equations for the macro fractures, micro fractures, and
grid cells. The multi-level nature of the fracture system will be rock matrix are analogous to Eq. 10 and follow the functional
taken into account by the magnitude of the fracture forms of Eqs. 2, 3, and 4. The matrix/mega fracture transfer
transmissibility coefficients in various directions. This flow function has the form given in Eq. 11 and the others follow the
model is suitable for multi-phase problems. same format in parallel to Eqs. 5-9:
In addition to the special PBI grid introduced earlier to
capture various fracture orientations, another notable feature wm / f1 = m kmwf1/ m [ ( pwf1 pwm ) + w (hwf1 hwm )] ......... (11)
of this paper is that fully implicit solution is not used for
SPE 93053 5

Model II Equations
In this formulation, we treat the fracture system as a set of
-if of v t f + iGwf + iCwf o + qof f Sof c f + cof ( ) poft = f
Sof
t
high-permeability, heterogeneous flow paths with small ............................................................................................ (21)
storage capacity. We accommodate for the variability in
fracture direction and directional permeability by using the We can write equations identical to 12 through 21 for the flow
perpendicular bisector grid shown in Fig. 6. This formulation in the matrix, with exception that we must replace the
is practical and leads to efficient computations. In the subscript f with m and the negative sign in front of the
following, we only present the water-oil flow equations, which transfer functions, , will become positive.
degenerate to single-phase flow when necessary. For the special case where the intra-matrix flow is
negligible compared to the fracture flow and the fracture-to-
Water-Oil Flow Case: The pressure equation for the matrix flow, we will have the following simple relationships
water-oil flow in the fracture network has the following form: for the fracture-to-matrix flow:
ikf t f pof ( wf w + of o ) D wf pcf
.................... (12) pom S
p
( w + o ) + q t = ( ct )f o w + qwm m S wm ( c m + cwm ) = m wm .................... (22)
t t t

where w + o = 0 .......................................................................... (23)

v t f = -kf t f pof ( wf w + of o ) D wf pcf ............. (13) Five equations, the pressure equation, Eq. 12, the total velocity
equation, Eq. 13, the water saturation in the fracture equation,
The water phase flow equation in the fracture network is:
Eq. 16, the fracture-matrix transfer function, Eq. 20, and the
water saturation in the matrix, Eq. 22, are the main equations
(
- v wf w + q wf f S wf c f + cwf ) poft = f
S wf
t
........... (14) for building the flow simulator. Thus, we will write down the
finite-difference form of these equations for further
discussion:
where
Computational Algorithm
v wf = -kf wf ( pof w D pcf ) ...................................... (15) The computation of the flow models presented above requires
a finite-difference formulation and an appropriate
Eq. 14 can be re-written in terms of the total velocity vtf as computational sequence.
follows:
Finite-Difference Formulation
The finite difference formulation for pressure and saturation
-if wf v t f iGwf iCwf w + q wf
equations are given below:
.................................. (16)
(
f S wf c f + cwf ) poft = f
S wf
t Pressure Equation: We choose the following finite-
difference equation as the analog of Eq. 12:
For incompressible flow, Eq. 16 reduces to the following
equation:
Ttnf pof (Twf w + Tof o ) D Twf n
n+1 n n
pcf

S wf VR ....................... (24)
-v tf if wf iGwf iCwf w + q wf = f ................... (17) + qwf = ( ct )f t pof
t t ( 24 )

where When Eq. 24 is written for every grid node, the result is a set
of linear equations for the oil pressure in the fracture, pofn +1 , at
Gwf = f wf of ( w o ) kf D ................................................. (18)
each node. These equations can be efficiently solved using
several available solvers.
Cwf = f wf of kf pcf .............................................................. (19)
Saturation Equation: The water saturation equation, Eq.
w = km
( wf om ) ( pcm pcf ) + w ( hwf hwm ) o ( hof hom )
16, can be written in the following finite-difference analog.
+ om )
n +1
( wf This equation is explicit in water saturation, Swf , for each
............................................................................................ (20) node; thus, it can be easily solved for water saturation.

The oil-phase flow equation in the fractures has the following

form:
( n
f wf )
v tx f yz x f wn pox

n
( w o ) x D x f wn pox
n
(
x pcn f + .... )
VR wn + qwf
VR n
t
( ) VR
f S wf c f + cwf t pof = f t Swf
t
............................................................................................ (25)
6 SPE 93053

2. Reorganize the fine grid into a coarse grid. For our

where, q wf = f w q t example problem we used 10x10x3. Thus, every
The water saturation in the matrix can be calculated by an coarse grid contained 25 fine-grid nodes.
explicit finite-difference analog of Eq. 22 as follows: 3. Use a flow-based permeability averaging technique
to calculate the components of the equivalent
(upscaled) permeability tensor, kx and k y , for each
VR wn + qwm
VR
t
n
m S wm ( )
c m + cwm t pom =

VR
t
m t Smw . (26)
coarse grid (Figs. 6, 8 and 9). The coarse grid may
internally consist of rectangular or PBI fine grid
Please note that in Eq. 24, 25, and 26, the time difference elements. The flow rate profile computed in the
operators t pof , t Swf , and t Som are part of solution vector upscaling process should be stored for future use as
the weight factors for velocity calculation when
and are defined as: t pof = pofn +1 - pofn , t Swf = Swf
n +1
- S wf
n
, interpolating from the coarse-grid pressure solution
n +1
and t Som = Som - Som
n
. to the fine-grid saturation calculations.
4. Solve the pressure equation, which is based on total
mobility on the coarse grid, and subsequently
calculate the total velocity components at each face
of the coarse grid.
5. Solve the pressure equation, which is based on total
mobility for the fine grids within each coarse grid,
using the velocity or flow rates allocated at the coarse
grid boundaries. The pressure at the center of the
coarse grid nodes will be assigned to the central fine-
grid node as an additional constraint. Once the fine
grid nodes are calculated, the total velocity or flow
rates at the fine-grid faces are calculated.
6. The x, y, and z-components of the total velocity ( vtx ,
vty , vtz ) calculated in the last step are used in the
saturation equation to calculate the fine-grid water
saturations.

Fig. 8 Two possible schemes for coarse mesh, square grid
construction superposed on fine mesh, perpendicular bisector
grid.
Note that the time steps used in the coarse grid pressure
solution could be fairly large, while the time steps used in the
fine-grid saturation calculations are fractions of the pressure
solution time-steps. The fine-grid time step size is constrained
by the Courant-Friedrichs-Lewy (CFL) stability criterion.13
The coarse-grid pressure solution is only updated periodically.

Results and Discussions

Fig. 9 Coarse mesh / fine mesh superposed on fracture matrix
network.
Computation Sequence
The computation sequence is as follows:
1. Construct a fine grid (either rectangular or
perpendicular bisector, PBI) for the reservoir. For our
example problem we used a 50x50x3 grid.
To illustrate the viability of our formulations, we simulated a
two-well, dual-porosity, heterogeneous reservoir segment
(1500 ft x 1500 ft x 30 ft) consisting of an injector and a
producer. The fine grid was 50 x 50 x 3 and coarse grid 10 x
10 x 3. The results of several 3D simulations are shown in Fig.
10 and the oil recovery fractions are shown in Fig. 11. These
comparisons show that the dual mesh results are essentially
the same as the results from the conventional fine-grid
simulation.
In the dual mesh, the pressure equation is solved implicitly
but the pressure vector is much smaller than the fine-grid
simulation requiring proportionately less computing time. The
computing time for the velocity interpolation from coarse- to
fine-grid mesh may be reduced with vectorization and parallel
processing. Similarly, the saturation calculations are explicit
and amenable to vectorization and parallel processing. The
entire process should lead to much smaller total computing
time than that for fine-grid simulation.
SPE 93053
Fig. 10 Saturation distribution in the fractures calculated by four
different methods.
Fig. 11 Comparison of recovery versus time for dual porosity,
incompressible system for the four methods used.
The approach used for single-mesh simulation in this paper
is very similar to the streamline approach used in an earlier
paper13 but with less complexity. As in streamline simulation,
the computing advantage comes from the fact that the pressure
solution is updated at much larger time intervals than the time
step size used to calculate the saturations. Similar results have
been obtained in a companion paper.14
To simulate pressure transient tests, a logarithmic time-
stepping routine is used and the pressure calculations are
performed on the fine grid level at every time step. The results
of a pressure drawdown test conducted in a heterogeneous,
dual-porosity reservoir are presented in Fig. 12. It can be seen
that the pressure response of this heterogeneous reservoir is
very similar to a homogeneous reservoir. This is very
consistent with field observations and merits further study in
conjunction with other observations in the field, as we alluded
to in the early part of this paper.
7
Fig. 12 Log-Log pressure derivative plot for the drawdown test
in the example, heterogeneous, fractured reservoir.
Fig. 13 Semilog pressure plot for the drawdown test in the
example, heterogeneous, fractured reservoir.
Conclusions
We have presented several approaches for modeling fractured
reservoirs. These approaches capture the effect of small-scale
reservoir heterogeneity (via multi-level fracture distribution)
with simpler computational algorithms than conventional
methods. For instance, the pressure is solved implicitly and
saturations are solved explicitly in fractured reservoir in
contrast to the common fully implicit approach. Furthermore,
the method used for multi-phase flow is similar to the
streamline modeling method with the same high degree of
resolution, but with added advantage of not requiring
saturation transfer from the streamline mesh to the pressure
solution mesh. The approaches presented in this paper should
enhance our ability to incorporate the fine-grid level
characteristics of the reservoir into flow simulation.
Acknowledgement
This work has been performed under Marathon Center of
Excellence for Reservoir Studies (MCERS) and the Slope and
Basin Consortium at Colorado School of Mines. Marathon Oil
Company, REPSOL YPF, Saudi Aramco, Yukos, and Kerr-
McGee have provided the financial support for the graduate
students involved in this research and other related topics at
MCERS.
8 SPE 93053

Nomenclature
ct = total compressibility, psi-1
Cw = capillary force flow velocity vector, ft/day
D = depth, ft
f w = water fractional flow, fraction
Gw = gravity force flow velocity vector, ft/day
h = gravity head, ft
k = absolute permeability, md
kr = relative permeability, md
kr = relative permeability endpoint, me
k = permeability tensor in fractures, md
f Conductivity of High Permeability Channels in Naturally
Lx = matrix block x-dimension, ft
Ly = matrix block y-dimension, ft
Lz = matrix block z-dimension, ft
no = oil relative perm exponent
nw = water relative perm exponent
p = phase pressure, psi
3 Gilman, J. R., Meng, H. Z., Kazemi, H. and Ozkan, E.:
q = surface flow rate, ft /day
3 3
q = specific flow rate, ft /day/ ft of reservoir
R = recovery factor, fraction
S = saturation, fraction
t = time, day
u = interstitial velocity vector, ft/day
v = Darcy velocity vector, ft/day
x = x-direction grid dimension, ft
y = y-direction grid dimension, ft
z = z-direction grid dimension, ft
Greek Symbols
= fluid gradient, psi/ft
= porosity, fraction
= density, lbm/ ft3
= viscosity, cp
= transfer function, day-1
= shape factor, ft-2
= mobility coefficient, cp 1
x , y , z = finite-difference spatial operator
t = finite-difference temporal operator
= ( x , y ,z )
= gradient operator
= divergence operator
References
1. Horne, R. and Raghavan, R.: Reservoir Description and
Dynamics, JPT (March 2004) 32-33.
2. Liu, Y., Harding, A., Abriel, W., Strebelle, S.: Multiple-point
simulation integrating wells, three-dimensional seismic data,
and geology, AAPG Bulletin (July 2004) v. 88, No. 7, 905-921.
3. Adibrata, B.W. H. and Hurley, N. F. : Flow-Unit Modeling Using
Neural Network, Logs and Cores in a Vuggy Dolomite
Reservoir, Dagger Draw Field, New Mexico, SPWLA 44th
Annual Logging Symposium, June 22-23, 2003.
4. Dershowitz, B., LaPointe, P., Eiben, T., and Wei, L.: Integration
of Discrete Feature Network Methods with Conventional
Simulator Approaches, SPEREE (April 2000) 165-170.
5. Kazemi, H. and Shinta, A. A.: Determining Orientation and
Fractured Reservoirs, Reservoir Characterization III, PennWell
Books, 1993.
6. Gogarty, W. B.: Enhance Oil Recovery Through the Use of
ChemicalsPart 2, SPE Distinguished Author Series (Dec.
1981-Dec. 1993) Vol. 1, 158-159.
7. Beliveau, D.: Pressure Transients Characterize Fractured Midale
Unit, JPT (Dec. 1989) 1354-1362.
8. Araujo, A., Lacentre, P. Zapata, T., Del Monte, A., Dzelalija, F.,
Dynamic Behavior of Discrete Fracture Network (DFN)
Models, SPE 91940, 2004 SPE International Conference,
Puebla, Mexico, Nov. 8-9, 2004.
9. Kazemi, H. and Gilman, J. R.: Multiphase Flow in Fractured
Petroleum Reservoirs, Flow and Contaminant Transport in
Fractured Rocks, Academic Press, San Diego, CA (1993) 267-
323.
10. Rodriguez, F., Arana-Ortiz, V., and Cinco-Ley, H.: Well Test
Characterization of Small- and Large-Scale Secondary Porosity
in Naturally Fractured Reservoirs, SPE 90287, 2004 SPE
Annual Technical Conference and Exhibition, Houston, TX,
Sept. 26-29, 2004.
11. Dreier, J., Ozkan, E., and Kazemi, H.: New Analytical Pressure-
Transient Models to Detect and Characterize Reservoirs With
Multiple Fracture Systems, SPE 92039, 2004 SPE International
Conference, Puebla, Mexico, Nov. 8-9, 2004.
12. Pascal, A. and Blunt, M. J.: Dual Mesh Method in Upscaling,
SPE 79681, 2003 SPE Reservoir Simulation Symposium,
Houston, TX, Feb. 3-5, 2003.
13.Moreno, J., Kazemi, H., and Gilman, J. R.: Streamline
Simulation of Countercurrent Water-Oil and Gas-Oil Flow in
Naturally Fractured Dual-Porosity Reservoirs, SPE 89880,
2004 SPE Annual Technical Conference and Exhibition,
Houston, TX, Sept. 26-29, 2004.
14. Atan, S., Almatrook, M., Kazemi, H., Ozkan, E. and Gardner, M.:
Dual-Mesh Simulation of Reservoir Heterogeneity in Single-
and Dual-Porosity Problems, SPE 93294, 2004 SPE Reservoir
Simulation Symposium, Woodlands, TX, Jan. 31-Feb. 2, 2005.

Subscripts
f = fracture
m = matrix
o = oil
w = water

Superscripts
n = old time level
n + 1 = new time level