SPE 84078

Practical Approach in Modeling Naturally Fractured Reservoir: A Field Case Study

Asnul Bahar, SPE, and Harun Ates, SPE, Kelkar and Associates, Inc., Maged H. Al-Deeb, SPE, Salem E. Salem, SPE,
and Hussein Badaam, ADCO, and Mohan Kelkar, SPE, The University of Tulsa
Copyright 2003, Society of Petroleum Engineers Inc.
This paper was prepared for presentation at the SPE Annual Technical Conference and
Exhibition held in Denver, Colorado, U.S.A., 5 8 October 2003.
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, TX 75083-3836, U.S.A., fax 01-972-952-9435.

Abstract
This paper presents a practical approach in modeling a
naturally fractured reservoir. The approach was used for a
field study of a giant carbonate reservoir in the Middle East.
The method is shown to be practical and comprehensive and
yet has produced good results. It consists of a fully integrated
effort from geological, geophysical and engineering
disciplines. The overall goal of the study is to develop a
representative reservoir model to form the basis for reservoir
management and long-term development planning.
The approach consists of the following procedures:
Generation of multiple realizations of matrix property
using geostatistical techniques. The standard
cosimulation procedure was implemented to ensure
the consistency among reservoir properties, namely
rock type, porosity and permeability.
Generation of multiple realizations of 3D fracture
property by reconciling seismic, well logs and
dynamic data. These were obtained from curvature
analysis and seismic facies map validated by borehole
image and dynamic data. The fracture network was
described in the reservoir as lineaments (fracture
swarms) showing two major fracture trends.
Calibration of the model permeability with well testderived
permeability
considering
fracture
distribution. A newly developed technique was
implemented to ensure that the fine scale model (i.e.,
geological model) honors well test as well as
production data before it was subjected to the flow
simulation. The technique also generates permeability
anisotropy to account for fracture orientations.
Ranking of multiple realizations using streamline
simulation to select three representative realizations
(low, medium and high models).

Upscaling of reservoir properties, including vertical

upscaling level optimization using streamline
simulation.
History matching and future performance prediction
of the three selected realizations as a single media
model. The use of single media model was based on
the observation of relatively high matrix permeability
in the major producing zone. However, for
comparison purposes, a dual media model was also
developed.
Uncertainty analysis of the future dynamic
performance using a probabilistic approach.
The procedure described above has been implemented
successfully in a field study. The use of a calibration process
in the geological model reduces the number of parameters that
need to be adjusted during history matching. Consequently,
history matching may concentrate on the uncertainty in
parameters that have not been specifically accounted for in the
geological modeling stage, such as relative permeability and
aquifer size/strength.
Introduction
A reservoir model is a tool that can be used for better reservoir
management and long term development planning. A
representative reservoir model can only be achieved by
properly integrating various sources of data in a consistent
manner. This requires use of techniques that can quantify all
possible uncertainties of the future performance.
This paper shows a practical approach in modeling of
naturally fractured reservoirs; from geology to flow
simulation. The approach uses different techniques to integrate
various sources of data. The goal is to obtain a comprehensive
reservoir model, i.e., a model that honors the available data
and quantifies uncertainties in the future performance.
The strategy used to achieve the above-mentioned
objective is by generating alternate reservoir descriptions,
based on stochastic models. This way, various uncertainties of
the data can be accommodated and modeled. The type of
uncertainty being considered is selected based on hierarchical
system. Various types of uncertainties are first ordered from
the most influential to the least influential in terms of their
effect on the future performance. Each uncertainty type is
represented by discrete (two to five) realizations. The
combination of all realizations from each type makes up the
total number of multiple realizations to be used.

The reservoir being studied is characterized by naturally

fractured system located within a relatively high matrix
permeability, i.e., average permeability in excess of 100 mD.
Dynamic performance indicates that either fracture or matrix,
or both, may act as the dominant factor contributing to the
production at each well. Therefore, one of the most important
challenges for this study is how to properly integrate the
fracture and matrix systems based on the available
information.
Two fracture maps were generated during the course of the
study (Figure 1). The first map was generated in 1999 and the
second map was generated, as an updated version, in 2002.
The difference between these two maps was significant. This
is due to several factors such as improvement on the quality of
the seismic interpretation, the increase on the number of
available well data and improvements on the fracture mapping
software.
The implementation and results shown in this paper were
based on these two fracture maps, which are referred to as
Case-1 (based-on the 1999 Fracture Study) and Case-2 (basedon the 2002 Fracture Study). The results of Case-1 have been
previously published in three technical papers.1,2,3 It is
presented here for convenience in order to maintain the
integrity and continuity of the overall technique with respect
to the new results (Case-2).
The 1999 Fracture Study produced a single realization of
fracture network only. There was no quantitative information
(fracture properties such as fracture width, porosity,
permeability, etc.) available. On the other hand, the 2002
Fracture Study produced three stochastic realizations as well
as fracture properties, namely fracture porosity, and the
anisotropy representation of fracture permeability (kx, ky, and
kz) at the grid block level. The three stochastic realizations of
the 2002 Fracture Maps were named the 500m, 750m, and
1000m realizations, respectively. These names were derived
from the average fracture length used to get each realization.
It is worth mentioning that all tasks described in this paper,
except for the conventional flow simulation and fracture
modeling, were done on Personal Computers (PCs). This
includes the generation of multi-million cells of matrix
models, integration of matrix and fracture, ranking and
dynamic upscaling using streamline simulation.
Geological Background
The structure of the resevoir is characterized by a gentle,
simple elongated anticline plunging both in the NE and SW
directions. It covers an area of approx. 160 km2. Oil is mapped
over an area of approx. 130 km2. It consists of a carbonate
sequence of Maestrichtian age deposited on an actively
growing palaeohigh in a shallow marine environment, subtidal
to intertidal (bank like) and supratidal conditions. It is
bounded by two unconformity surfaces being overlain by the
basal shale of Palaeocene age and underlain by Upper
Cretaceous formations.
Traditionally, the reservoir was divided into three main
reservoir units U1, U2 and U3 on the basis of petrophysical
variations and characteristics. Subsequently, it was subdivided
into 11 flow units based on sequence stratigraphy.
The thickness of the reservoir increases from the crest to
the NW and SE flanks. The reservoir is highly heterogeneous,

SPE 84078

as both the diagenetic and tectonic activities have drastically

affected the formation and have had important impact on
reservoir characteristic both vertically and laterally.
Eight different rock types have been determined in the
whole reservoir section. They are characterized by specific
petrophysical properties. The determination was based on the
electrofacies, which was based on logs using a combination of
cluster and discriminant analyses techniques taking into
consideration the Stratigraphic Limits of the three reservoir
units, calibrated and supervised with core data (for lithology,
, K, and SCAL Data (for Pc, Kr). It has been observed that
rock type number 7 (sucrosic dolomite rock) was characterized
as the super-K rock that makes the matrix system of this
reservoir as important as the fracture system in controlling the
production.
Approach, Implementation and Results
The overall workflow of the study is shown in Figure 2. It
mainly consists of eight main stages, as follows:
1. Structural Framework Modeling
2. Hierarchical System of Multiple Realizations
3. Matrix Modeling
4. Fracture Modeling
5. Integration of Matrix and Fracture Systems
6. Ranking of Multiple Realizations
7. Upscaling
8. Flow Simulation (History Matching, Prediction and
Uncertainty Quantification)
The detail of each stage will be the subject of the following
sub-sections.
Structural Framework Modeling. The first stage in the
workflow is the building of framework model. The input data
required for building this model comes from all disciplines
(geophysics, geology, petrophysics and engineering). Several
aspects that need to be properly considered during this
modeling include faults, eroded surface, isochronal/
stratigraphy intervals, well markers/tops, fine-layering design,
and scale-up of well logs to match the grid block size.
For this study, the structural framework model was
developed using 4 seismic surfaces. These surfaces represent
the boundaries between the three major reservoir units (U1,
U2, and U3). Faults are considered neglible in this field since
they coincide with the fracture network and have relatively
small throw, i.e., about 5-30 ft.
The three major reservoir units were divided further into
11 flow units based on the stratigraphy interpretation.
Subsequently, the fine layering was done within each flow unit
by considering the resolution of well log of 2 ft. Final
dimension of the framework was 334 x 134 x 93 = 4,162,308
grid blocks, where each grid block has dimension of
approximately 100 m x 100 m in the areal direction.
Hierarchical System of Multiple Realizations. Multiple
realizations that are commonly generated using geostatistics
technique were mainly due to differences in the random seed
number used in the geostatistical simulation. This may be
considered as the representation of uncertainty because of the
simulation process. However, uncertainty exists in many other
aspects of the modeling process that comes from various data

SPE 84078

sources, such as uncertainty of seismic structural

interpretation, faults location/conductivity, geological
continuity, well log measurement/interpretation, etc. Each of
this uncertainty may impact reservoir performance at different
influence levels.
The design of the multiple realizations should cover the
entire range of possible uncertainties that are considered
important for a particular field study. These uncertainties may
be listed from the highest to the least influences on the
reservoir performance. Several discrete realizations (from two
to five) for each uncertainty types may be assigned to the
overall model. The multiple combinations of all variables
make up the overall multiple realizations scheme of the model.
The number of variations for each uncertainty type should be
assigned so that the total number of realizations is still
manageable, e.g., less than 100 realizations.
Four types of uncertainties were considered for this study.
They are structural uncertainty, geological uncertainty, petrophysical property uncertainty and simulation technique
uncertainty.
Three cases of structural uncertainty were modeled to
represent the low, most-likely and high structures. The low
and high structures were generated by considering the amount
of uncertainty observed during the development of the field,
i.e., the mismatch between seismic prediction and the actual
depth found during drilling of new wells.
Two cases of geological uncertainty were considered
implemented as two different sets of variogram models. The
first set of variogram model was obtained purely based on the
well data. The second set of variogram model was obtained
based on the combination of well data and the soft geological
knowledge of the depositional model of the reservoir. More
details on this topic can be found in Ref. 4.
Two cases of petrophysical properties were modeled. This
is to capture the uncertainties of the measurement and
interpretation of the well logs. The first property was obtained
directly from the well-log measurement/interpretation,
whereas the second property was obtained by re-sampling the
original well log within the error-band of the
measurement/interpretation.
Four cases of simulation paths were considered to account
for simulation uncertainty. The combination of all of the above
four uncertainty types produces (3 x 2 x 2 x 4) 48 realizations.
Matrix Modeling. The petrophysical properties of the matrix
system are commonly generated by integrating well-log, cores,
and in some cases seismic derived-properties using
geostatistical methodology. Furthermore, prediction at the
uncored wells for rock type and permeability should be done
to improve data coverage, which in turn will improve the
statistics of the result.
Prediction at the uncored wells may be done using various
techniques, such as cross-plot, neural-network, electro-facies
followed by non-linear regression, or combination of linear or
non-linear regression with geostatistics. The use of
geostatistics in permeability prediction provides an additional
uncertainty group that may be added into the hierarchical
system of the multiple realizations scheme.5
An important feature that needs to be satisfied for each
realization is that the petrophysical properties have to be

consistent with the underlying geological description of the

field. That is, the generated properties, namely porosity and
permeability, should be in agreement with the underlying
facies/rock type description.6
Another important aspect of generating realizations with
reference to the geology is how to properly model the
continuity of the rock and property consistent with the
geological interpretation, namely the depositional environment
and diagenesis. A representative reservoir model should have
good agreement between the conceptual geology, as it is
viewed by geologist, and the simulation result. This goal can
be achieved by properly modeling the spatial relationships of
the rock type as well as the properties.4
In addition to constraining to the geological information,
matrix model should also be constrained to all available
data/information, such as seismic-derived porosity. The main
difficulty in integrating this information is how to reconcile
scale differences between well log and seismic measurement.
A technique described by Doyen et al. was implemented in
this study to solve this problem.7 It was found that seismic
porosity constraint has significantly improved the property
distribution at the poor-well coverage areas (i.e. the flanks of
the reservoir).3
Figure 3 (a), (b) and (c) show the examples of longitudinal
cross-section of rock type, porosity, and permeability
respectively. The consistency among these properties is shown
clearly in this figure. For example, the areas of high porosity
and permeability coincide with the good rock types (RRT 4
and 7), whereas the areas of poor porosity and permeability
coincide with the bad rock types (RRT 1 and 2).
Fracture Modeling. The presence of fractures has always
been an issue for the assestment of field productivity and
reserves. Techniques to model fracture distribution are
available in the industry.8,9 These are based on seismic data
together with well-cores, borehole image log and welldynamic performance to map fracture locations in the
reservoir using the curvature analysis principle.
In practice, curvature corresponds to the second derivative
of topography for a given orientation. For fracture mapping
purposes, curvature at a certain point is assigned to be positive
or negative. Based on the observation at the well location
(core and borehole image log), the positive curvature value
may indicate the possible fracture locations whereas negative
curvature may indicate non-fracture locations. Furthermore,
fracture locations determination may be enhanced by the
seismic facies analysis that uses other seismic attributes such
as Dips, Edge, Curvature and Trace Dissimilarity. The details
of such techniques are beyond the scope of this paper and can
be found elsewhere.8,9
The availability of information about the location of
fractures across the entire reservoir indeed gives valuable
information for the overall modeling process. It may serve as a
guide for better modeling of permeability.1 However, this
information alone may not be enough to quantitatively
describe the fracture system. For this reason, the hydraulic
characterization of the fracture system, based on the well test
and production data is required in order to get fracture
properties that can be used to derive fracture porosity and
fracture permeability.

For this study, fractures were observed from cores, bore

hole images, mud-losses, production log and pressure transient
analysis. The fractures were found to be grouped as clusters
that can be interpreted as fracture swarms or as sub-seismic
faults, oriented in two main directions N40E and N70E
(Figure 1). There was no diffuse fracturing determined from
the available data. Based on the geological observations, the
two fracture trends are likely to be open.
The hydraulic characterization of the fracture, that includes
productivity and water breakthrough analysis, has been
performed for the 2002 Fracture Study. It was concluded that
the production in the upper part of the reservoir (Unit U1) was
mainly enhanced by fracture. On the other hand, for the
middle reservoir unit (Unit U2), the production was enhanced
by both fracture and super-K rock (sucrosic dolomite). The
sucrosic dolomite rock was represented by rock type number 7
in the model. The lower reservoir unit (Unit U3) was a nonproductive unit, except for the upper section which shows
similar characteristic with the lowest section of the middle
reservoir unit U2.
Additionally, fracture properties have also been calculated
based on the PLT and well test data. The procedure of this
calculation is beyond the scope of this paper. The calculated
fracture porosity was in the range between 0.000015% to
0.3797 % whereas fracture permeability was in the range
between 215 to 5700 mD, 70 to 5700 mD, and 0.5 to 12000
mD, for x, y, and z directions, respectively. The permeability
anisotropy ratio ranges between 0.05 to 6.4, 0.0001 to 16.9,
and 0.001 to 61.4, for (ky/kx), (kz/kx), and (ky/kz), respectively.
Integration of Matrix and Fracture Systems. One of the
main challenges that need to be addressed in this study is how
to integrate matrix and fracture systems using the available
information. For this reservoir, it was observed that the ratio of
the flow potential, kh, between well test permeability and the
core-derived permeability is very significant (up to 1100) for
the wells located in a fractured area and insignificant (close to
1) for the wells located at non-fractured areas. This ratio is
referred as Enhancement Factor, EF, and it is given in the
following equation:
kh
(Eq. 1)
EF = welltest
khmod el
Case-1. Two cases were addressed in this study to
reconcile the differences in the flow potential. The first case
was based on the 1999 Fracture Study where the only
information available was the fracture location. The approach
used in this study was to find a correlation between areal
fracture density (AFD) and log(EF) required to match the well
test permeability. Figure 4 shows the typical correlation found
for this reservoir. The intercept of the correlation line (which
corresponds to AFD equal to 0.0) is the background
enhancement, which reflects the difference between well test
and core measurement in the absence of macro fractures. Areal
fracture density was calculated based on the number of
fracture lines that penetrate each grid block per the crosssectional area of a particular grid block. The detail of the
approach and results for this case has been previously
published and can be found in Ref. 1.

SPE 84078

Case-2. The second case was based on the 2002 Fracture

Study where fractrure properties were available at every grid
block. The integration procedure used for this case is shown in
Figure 5. This workflow consists of two main steps, namely
Pre-Calibration and Calibration. In brief, the Pre-Calibration
consists of some analysis that will be used as the guide in
performing permeability calibration. The Calibration process
is a looping system that links the static geological model with
the dynamic flow simulation model. The purpose of the
looping system is to ensure that changes made to permeability
are done within the geological understanding of the reservoir.
Thus, consistency between the two models (static and
dynamic) can be maintained. The looping system is terminated
once satisfactory production capacity match, as evaluated from
flow simulation, is achieved.
Pre-Calibration. The pre-calibration step consists of three
tasks, namely Dominant Factor Determination, Upscaling and
Flow Simulation. The objective of the Dominant Factor
Determination is to determine whether a well is predominantly
controlled by matrix or fracture system. This can be done by
performing one of the following two analyses.
(1) Enhancement Factor (EF) Analysis. The EF analysis
was done by calculating the EF (see Eq. 1) for two different
systems. The first one is the system with matrix only, i.e.,
assuming there is no fracture in the reservoir. The second
system is the one that has both matrix and fracture.
Figure 6 shows the results of the EF Analysis. For the first
system (matrix only), it can be observed that significant
variation exists with the EF. On the other hand, for the second
system (matrix + fracture), the EF distribution has changed.
The changes vary from well to well. They are significant for
some wells and are insignificant for other wells. The changes
of EF can be crosschecked by evaluating the location of each
well. The wells that show significant change of EF is
classified as Fracture Well, whereas the wells that are not so
sensitive are classified as the Matrix Well.
Figure 7 shows the dominant factor at all wells for the
three stochastic fracture realizations. It can be seen that most
of the realizations predicted the same dominant factor except
at 6 wells. This shows the presence of uncertainty with respect
to the determination of fracture location.
(2) Sensitivity Analysis. The objective of the sensitivity
analysis is to determine the sensitivity of the EF with respect
to the changes on matrix or fracture permeabilities. This
analysis provides the confirmation of the results obtained from
the previous analysis with regard to the dominant factor. The
analysis was conducted by multiplying the original matrix or
fracture permeability (globally) with several constants. The
wells that are sensitive to the changes in fracture permeability
are classified as the fracture wells and likewise the wells that
are sensitive to the changes in matrix permeability are
classified as the matrix well. The results of the dominant
factor determined from the EF Analysis (Figure 7) has been
confirmed from this analysis.
Upscaling and Flow Simulation. Upscaling serves as the
link between the static and dynamic models due to differences
on the model size. An appropriate upscaling technique such as
flux-based (numerical simulation) method is required to
ensure proper upscaling of permeability.

SPE 84078

During the pre-calibration analysis, full-field flow

simulation was executed for the entire production history. The
purpose of flow simulation at this stage is for the evaluation of
production capacity match only. It is not intended for full
history matching purposes, such as matching water/gas
production. Production capacity match is evaluated from the
capability of wells in producing oil at the specified rate and at
the observed flowing bottom hole pressure, for a given
permeability distribution. The evaluation produces three
possible conditions, namely under-capacity, matched-capacity,
or over-capacity. Depending on the condition at each well, it
either needs further enhancement (under-capacity), does not
need further enhancement (matched-capacity), or needs
reduction in permeability (over-capacity).
Figure 8 shows the flow simulation results for the well
with EF 1. It can be seen that this well is capable to
produce the oil at the specified flowing bottom hole pressure.
Wells with this condition is classified as matched capacity.
On the other hand, for wells that has EF >> 1 the flow
simulation results clearly show that those wells were not able
to produce the oil at the specified condition (Figure 9). The
wells with this condition were classified as under capacity.
Lastly, for wells with EF << 1, they are classified as over
capacity. Thus, good link exists between the concept of EF
and production capacity. This provides a good way in
predicting the amount of adjustment required during the
calibration process to convert all wells into the matchedcapacity category.
Figure 10 shows the result of the production capacity
match for all wells. The results show that most of the fractured
wells have been identified to match the production capacity.
This may indicate that the 2002 Fracture Map has successfully
predicted the location of major fractures that can be captured
by seismic data. On the other hand, the production capacity at
most of the matrix wells were classified as under capacity.
This may be due the presence of micro fracture (observed as
chicken-wire fracture), which cannot be captured by seismic
data.
Calibration. The calibration step also contains three tasks,
namely Permeability Adjustment, Upscaling and Flow
Simulation. The objective of the permeability adjustment is to
adjust matrix and fracture permeabilities separately to match
the well test and production capacity. The adjustment can be
done in two steps. The first step is for matrix adjustment and
the second step is for fracture adjustment. Matrix adjustment
can be done globally using kriging of the EF and then applied
to the original matrix distribution. The control points for the
kriging should be from the matrix wells or from wells that are
both influenced by matrix and fracture. Fracture enhancement
can be done in several ways, either globally or locally.
The results of the calibration process show significant
improvement on the production capacity match. All of the
under capacity wells have been transformed into the
matched capacity category. However, some of the over
capacity wells remain the same. For this case, it appears that
the introduction of fractures has added too much permeability
around those wells. There are several possible reasons as to
why this occurs, such as the possibility of closed fractures
around those wells and the possibility of incorrect fracture
locations. Additional production data were required to solve

this problem and it was decided to solve this issue during the
history matching process.
Ranking of Multiple Realizations. One of the important
consequences of generating multiple realizations is the
difficulty in evaluating them (i.e., dynamic evaluation). This is
true since the geological model is commonly built using multimillion cells for which rigorous flow simulation cannot be
conducted and history matching cannot be performed for all
the realizations considered.
This problem can be overcome by ranking all realizations
using streamline simulation that can run multimillion cell
models in a quite efficient, however approximate manner. The
streamline simulation has been tested successfully in the
literature for such tasks.10 Three realizations may be then
selected to represent the low, medium and high cases for
detailed history matching process.
Case-1. The goal of the ranking for the Case-1 was to rank
the 48 realizations to select 3 representative realizations. The
details of this ranking process can be found in Ref. 2.
Case-2. For the Case-3, there are only 9 realizations to be
ranked. These 9 realizations were the result of the combination
between the 3 matrix realizations (low, medium and high), i.e.,
the three realizations selected from Case-1, and the 3 fracture
realizations from the 2002 Fracture Study. Figure 11 shows
the results of the ranking, plotted as normalized sweep
efficiency versus STOIIP. One realization is selected for each
group. For low case, the selected realization is the one that
gives the lowest sweep. For medium case, the selected one is
the one with middle sweep, and for the high case, the selected
one is the one with highest sweep.
Upscaling. Upscaling from a fine scale geological model into
a coarser scale simulation model is a step that cannot be
avoided due to the computational cost and limitation on CPU
resources. A reduction from multimillion cells into several
tens or hundreds of thousands grid blocks is a common
practice in the industry. The challenge that exists in this step
can be viewed as two-fold;
1- Optimizing the vertical upscaling level
2- Upscaling technique to be used for petrophysical
properties
Both of these aspects have a unique goal, which is to
maintain the heterogeneity of the geological model as much as
possible while optimizing the number of grid blocks. Most of
the current reservoir model softwares available in the industry
have several options to conduct upscaling of the petrophysical
properties properly. Therefore, it would not be discussed here.
However, the first aspect of the upscaling is not usually
properly considered. The optimization of the vertical upscaling
level can be achieved by evaluating the dynamic performance
of the model. This can be done again using the streamline
simulation technique that is capable in calculating layer sweep
efficiencies quickly.2
Another important aspect that also needs to be considered
is how to choose the grid blocks that need to be combined. A
practical way in performing this task is to evaluate the layer
sweep efficiency and combine only two or more subsequent

layers that have similar sweep numbers. It is also important to

remember that grids from different flow unit should not be
combined. For this study, 30 layers have been observed to be
the optimum number of layers.2 The final dimension of the
upscaled model was 147 x 41 x 30 = 180,810 grid blocks.
Flow Simulation. Flow simulation through history matching
and future prediction is ultimately the main objective of any
reservoir study. It is logical to expect that all efforts that have
been spent in building the model to show their influence
during the flow simulation process. That is, the history
matching processs should not be done with any arbitrary
changes to match production history. Therefore, the history
matching should concentrate on the parameters that have not
been accounted for during the static modeling process, such as
the relative permeability, aquifer strength, faults
conductivities, etc.
The choice of flow simulation technique used for the
fractured reservoir has become the subject of many debates.
Both single medium and dual medium have been used to
simulate this type of reservoir.11,12 The single media model has
been chosen in this study. However, for comparison purposes,
a dual media model was also developed.
Case-1. Comprehensive history matching and future
prediction have been successfully completed for the three
selected realizations using ECLIPSE-100 Black Oil Simulator
within reasonable amount of time.
Field-wise, the history match results have shown excellent
results. On the well-level, the water cut match at 52 out of the
55 well strings were obtained without the need to modify
permeability around the wells during history history matching
process. The three wells where permeability needs to be
further adjusted to obtain the water cut match were horizontal
wells. These wells were not used as the control points in
obtaining the correlation line of enhancement process. See
Ref. 3 for the details.
Case-2. Two flow simulation models, single and dual
media, have been developed and executed. The dual media has
been used for comparison purposes only. Various
simplifications of single media model do not seem to affect the
capability of the model in predicting reservoir performance. In
terms of running simulation time, the single media model was
completed in about 3 hours whereas the dual media model
took about 3 days. Therefore, from practical point of view, the
single media model is very attractive while producing good
results.
The history matching result of Case-2 (single media) has
shown significant improvement on the ability of the model in
predicting water cut performance. Figure 12 shows the
comparison of water cut performance at one of the fracture
wells between (a) Case-1 and (b) Case-2. The matches shown
in this example are similar, i.e., both cases are able to predict
water cut performance properly. However, the match in Case1 was obtained after applying multiplying the y-direction
permeability, ky, around the well with a factor of 60. This well
is one of the three wells that needed local adjustment during
the history match of Case-1. On the other hand, the match on
Case-2 was obtained immediately, i.e., after the calibration

SPE 84078

process. Comparison for other wells also show similar

encouraging results.
The model has also been used to design the location of
new wells. Some of these wells were completely drilled as
new infill wells but some others were rehorizontalizationworkover of old horizontal wells that show high water cut.
The results of the workover wells were very good. It
successfully eliminated water production. Figure 13 shows the
production performance of a rehorizontalization work-over
well. It can be seen that water cut has been steadily increasing
prior to the workover and was immediately drop to zero upon
the completion of the work-over. The results from the newly
infill wells are not yet available at the time this paper was
written because they have not been put into production yet.
Uncertainty Quantification. The last, but certainly not the
least, challenge in a study that uses a stochastic approach is
how to quantify the uncertainty in future performance as given
by the results of the different realizations. The answer to this
question may be the most sought by Senior Management.
Therefore, a proper evaluation is required in quantifying this
uncertainty. A technique described by Misra, et al.13 may be
used for this purpose. In this technique, each realization is
given a weighting factor that can be used to calculate the
expected value of the production. The weighting factor may be
calculated through a surrogate function such as the sweep
efficiency that has been used in ranking the multiple
realizations.
For this study, it was found that the weighting factors to be
applied for the three realizations are p1 = 0.27, p2 = 0.51 and p3
= 0.22.3 Therefore, the expected value of the future production
can be calculated using the results of the three simulations
weighted with these probability factors.
Summary
The study presented in this paper can be summarized as
follows:
1. Practical, but comprehensive, reservoir modeling
procedure using the concept of an integrated study has
been presented with a case study for a naturally fractured
carbonate reservoir.
2. The approach used to achieve the objective was to generate
alternate descriptions based on the stochastic techniques by
integrating various data source from different disciplines.
3. The hierarchical system was designed and implemented in
designing the scenario for the multiple realizations in order
to capture all possible uncertainties.
4. Matrix and Fracture properties are modeled separately and
then integrated using newly developed techniques. Two
cases were presented based on the available information.
The integration process was done within the geological
understanding of the reservoir to ensure consistency
between the static and dynamic model.
5. Streamline simulation technique has been used for two
purposes, namely ranking and upscaling.
6. Flow simulation of single media model has been used to
simulate the naturally fractured reservoir. The results show
the capability of the single media model in simulating the
naturally fractured reservoir.

SPE 84078

7. The integration process has reduced the effort in history

matching by concentrating on working with the parameters
that have not been included in the static description of the
model.
8. The model has been used to design the location of new
wells. Results from the rehorizontalization-workover
program show a good indication of the representativeness
of the model in predicting the future performance.

13. Mishra, S., Choudary M. K., and Datta-Gupta, A., A Novel

Approach for Reservoir Forecasting Under Uncertainty, SPE
62926, paper presented at SPE Annual Tech. Conference and
Exhibition, Dallas, TX Oct. 1-4, 2000.

N40E

References
1. Charfeddine, M., et. al., Reconciling Well Test and Core
Derived Permeability using Fracture Network: A Field Case
Example, SPE 78499 presented at the 10th Abu Dhabi
International Petroleum and Exhibition Conference held in Abu
Dhabi, U.A.E., 13-16 October 2002.
2. Ates, H., et. al., Ranking and Upscaling of Geostatistical
Reservoir Models Using Streamline Simulation: A Field Case
Study, SPE 81497 presented at the 2003 SPE Middle East Oil
Show, held in Bahrain, 9-12 June 2003.
3. Al-Deeb., et. al., Fully Integrated 3D-Reservoir Characterization
and Flow Simulation Study: A Field Case Example, SPE 78510
presented at the 10th Abu Dhabi International Petroleum and
Exhibition Conference held in Abu Dhabi, U.A.E., 13-16 October
2002.
4. Bahar, A., Ates, H., Kelkar, M., and Al-Deeb, M., Methodology
to Incorporate Geological Knowledge in Variogram Modeling,
SPE 68704, presented at the SPE Asia Pacific Oil and Gas
Conference and Exhibition, held in Jakarta, Indonesia, 17-19
April 2001.
5. Silva, F., Ghani, A., Al-Mansoori, A., and Bahar, A., Rock Type
Constrained 3D Reservoir Characterization and Modeling, SPE
78504 presented at the 10th Abu Dhabi International Petroleum
and Exhibition Conference held in Abu Dhabi, U.A.E., 13-16
October 2002.
6. Bahar, A. and Kelkar, M. Journey From Well Logs/Cores to
Integrated Geological and Petrophysical Properties Simulation: A
Methodology and Application, SPE Reservoir Evaluation and
Engineering 3(5), October 2000.
7. Doyen, M., Psaila, D.E., Den Boer, L.D., and Jans, D.,
Reconciling Data at Seismic and Well Log Scales in 3D Earth
Modeling, SPE 38698, presented at the 1997 SPE Annual
Technical Conference and Exhibition held in San Antonio, Texas,
5 8 October 1997
8. Bourbiaux, et. al. An Integrated Workflow to Account for MultiScale Fractures in Reservoir Simulation Models: Implementation
and Benefits, SPE 78489, presented at the 10th Abu Dhabi
International Petroleum and Exhibition Conference held in Abu
Dhabi, U.A.E., 13-16 October 2002.
9. Bloch G., et. al., Seismic Facies Analysis for Fracture Detection:
A New and Powerful Technique, SPE 81526, presented at the
2003 SPE Middle East Oil Show, held in Bahrain, 9-12 June
2003.
10. Idrobo, E. A., Choudhary, M. K., Datta-Gupta, A., Swept
Volume Calculations and Ranking of Geostatistical Reservoir
Models Using Streamline Simulation, SPE 62557 presented at
the 2000 SPE/AAPG Western Regional Meeting held in Long
Beach, California, 19 23 June 2000
11. Van-Lingen, P., Sengul M., Daniel J-M., and Cosentino, L,
Single Medium Simulation of Reservoirs ith Conductive Faults
and Fractures, SPE 68165, presented at the 2001 SPE Middle
East Oil Show, held in Bahrain, 17-20 March 2001.
12. Gurpinar, O. M., and Kossack, C. A., Realistic Numerical
Models for Fractures Reservoirs, SPE 68268 / SPE Journal Vol.
5 (4), December 2000.

N70E

(a) 1999 Fracture Map

N40E

N70E

(b) 2002 Fracture Map

Figure 1
Fracture Maps Shown the Two Main Directions (N40E
and N70E); (a) 1999 Fracture Map, (b) 2002 FractureMap

Fracture
Modeling
Geological
Interpretation
Petrophysical
Interpretation
Seismic
Interpretation

Figure 2

Hierarchical
Realizations

Structural
Framework

Integration of
Matrix/Fracture

Predictions At
Uncored Wells

Matrix
Modeling

3 Selected
Real. Upscaling
Of Prop.

Ranking of
Realizations

Overall Workflow of the Reservoir Study

Selective
History
Matching

SPE 84078

Pre-Calibration
Dominant Factor
Upscaling
Flow Simulation

Calibration
Permeability Adjustment
Upscaling
Flow Simulation

(a)

NO

Production Capacity
Match
Accepted ?

YES

Finish

Figure 5
Workflow for Matrix and Fracture Integration used for
the Case where Fracture Property is Available

Matrix Well

Fracture Well
1000

(b)

100

EF

10

0.1

Well
Matrix

Matrix + Fracture

Figure 6
Variation of Enhancement Factor for Two Different
Models; System with Matrix Only and System with Matrix +
Fracture

(c)
Figure 3
Longitudinal Cross-Section of Matrix Properties: (a)
Rock Type, (b) Porosity, and (c) Permeability

(Legend indicates Well Number)

Figure 4
Typical Correlation between Fracture Density and
Enhancement Factor used to Match Core-Derived Permeability
with Well Test Permeability

35H-U2

34H-U2

32H-U2

27H-U2

26H-U2

25H-U2

6H-U2

21H-U2

11H-U2

25H-U1

24H-U1

22H-U1

19H-U1

18H-U1

4V-U3

12H-U1

20V-U2

19V-U2

18V-U2

15V-U2

14V-U2

13V-U2

6V-U2

12V-U2

4V-U2

11V-U2

1V-U2

20V-U1

19V-U1

18V-U1

15V-U1

13V-U1

12V-U1

11V-U1

1V-U1

0.01

SPE 84078

Index

String

500 m

750 m

1000 m

1V-U1

11V-U1

12V-U1

13V-U1

15V-U1

18V-U1

M
F
M
F
M
M
M
F
F
F
F
M
F
M
F
M
F
F
F
M
F
M
M
M
M
F
F
M
M
F
F
F
F
M
M
M
F
F
F

F
F
M
F
M
M
M
F
F
F
F
F
F
M
M
F
F
F
F
M
F
M
M
M
M
F
F
M
M
F
F
F
F
M
M
M
F
F
F

M
F
M
F
M
M
M
M
F
F
F
F
F
M
F
M
F
F
F
M
F
M
M
M
M
F
F
M
F
F
F
F
F
M
M
M
F
F
F

20V-U1

12H-U1

10

18H-U1

11

19H-U1

12

22H-U1

13

24H-U1

14

25H-U1

15

39H-U1

16

1V-U2

17

4V-U2

18

6V-U2

19

11V-U2

20

12V-U2

21

13V-U2

22

14V-U2

23

15V-U2

24

18V-U2

25

19V-U2

26

6H-U2

27

11H-U2

28

21H-U2

29

25H-U2

30

26H-U2

31

27H-U2

32

32H-U2

33

34H-U2

34

35H-U2

35

37H-U2

36

Figure 7

19V-U1

38H-U2

37

4V-U3

38

4V-COM

39

6V-U3

M = Matrix
F = Fracture

Dominant Factor for All Wells

Figure 9

Flow Simulation Result for Well with EF >>1.0

Index

Figure 8

Flow Simulation Result for Well with EF

1.0

2
4
8
9
10
11
13
15
17
18
19
21
26
27
30
31
32
33
37
38
39
1
3
5
6
7
12
14
16
20
22
23
24
25
28
29
34
35
36

String
11V-U1
13V-U1
20V-U1
12H-U1
18H-U1
19H-U1
24H-U1
39H-U1
4V-U2
6V-U2
11V-U2
13V-U2
6H-U2
11H-U2
26H-U2
27H-U2
32H-U2
34H-U2
4V-U3
4V-COM
6H-U3
1V-U1
12V-U1
15V-U1
18V-U1
19V-U1
22H-U1
25H-U1
1V-U2
12V-U2
14V-U2
15V-U2
18V-U2
19V-U2
21H-U2
25H-U2
35H-U2
37H-U2
38H-U2

Dominant
Factor
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
M
M
M
M
M
M
M
M
M
M
M
M
M
M
M
M
M
M

Production
Capacity
MATCH
UNDER
UNDER
OVER
OVER
MATCH
UNDER
MATCH
MATCH
MATCH
MATCH
UNDER
MATCH
MATCH
MATCH
MATCH
UNDER
MATCH
MATCH
UNDER
MATCH
UNDER
UNDER
UNDER
UNDER
UNDER
MATCH
UNDER
UNDER
UNDER
UNDER
UNDER
UNDER
UNDER
MATCH
UNDER
UNDER
UNDER
UNDER

M = Matrix
F = Fracture

Figure 10 Result of the Production Capacity Match during the

Pre-Calibration Analysis

10

SPE 84078

1.5

10000

1.4

500 m

W/C (%)

Horizontalization
Workover for ESP Downsizing.

8000

55
50
45

7000

1.1

750 m

Selected (1000m)

1000 m
500 m

0.9

1000m
500m
750m

0.8

40

6000

35

5000

30

4000

25

Water Cut (%)

1.2

Oil Rate (BOPD)

Normalized Sweep Efficiency

Oil Rate (BOPD)

9000

750 m
1000m

1.3

60
Flow with ESP

Selected (500m)

20

3000

15

0.7

Selected (750m)

2000

0.6

10

1000

0.5
0.5

0.6

0.7

0.8

0.9

1.1

1.2

1.3

1.4

1.5

Normalized STOIIP

Figure 11 Result of the Ranking of Realizations for Case-2

60*Ky

(a)

(b)
Figure 12 Comparison of the Flow Simulation Result At a
Fracture Well between (a) Final Match for Case-1 and (b) First Run
for Case-2

0
1995

5
1996

1997

1998

1999

2000

2001

2002

2003

Time - Years

Figure 13 Production Performance from a Re-HorizontalizationWorkover Well