You are on page 1of 10

SPE 102165

Characterisation and Modelling of a Fractured Reservoir Using a Novel DFN Approach


F. Gouth and A. Toublanc, Total, and M. Mresah, CPTL

Copyright 2006, Society of Petroleum Engineers


of the flow mechanisms involved during production is
This paper was prepared for presentation at the 2006 Abu Dhabi International Petroleum underlined.
Exhibition and Conference held in Abu Dhabi, U.A.E., 5–8 November 2006.

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
Introduction and scope
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to The full workflow of a fracture study from characterization 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 modeling on a field operated by TOTAL in Africa is described
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
in this paper. Particular attention is given to the integration of
for commercial purposes without the written consent of the Society of Petroleum Engineers is static and dynamic data and the way it allowed us to achieve a
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 satisfactory characterization of the fracture network. A new
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.
DFN tool developed by TOTAL has been used for the
modeling of this field. For confidentiality purposes, the field
Abstract will be designated as reservoir ‘R’ in the paper.
Characterising and modelling of naturally fractured reservoirs The field is an oil and gas accumulation forming a WE
(NFR) with fracturing at different scales is usually a anticline truncated to the East by a fault (Fig 1). It is located in
challenging task as the specific response from each scale is a complex folded and faulted transfer zone, active during the
difficult to isolate. reservoir deposition, of Eocene age. The morpho-structural
We focus on a carbonate reservoir in North Africa, in pattern of the transfer zone is constrained by N150 normal
production for two years. There is evidence of fracturing at faults with possible salt diapir activity paralleling the rifts and
different scales (from diffuse fractures to conductive faults) bounding 3 sedimentation domains. They cut folds with N80
with significant implications for oil production. axes partially interpreted as reactivation of previous normal
We first present the way fractures have been fully faults. The western domain of the transfer zone is constrained
characterised using an extensive integration of static (FMI) by tectonic activity during the Eocene reservoirs deposition
and dynamic (Well test, Pressure Build Up, Mud losses) data. with:
The advanced use of well test signatures in understanding the • High subsidence under open lagoonal and marine
main flow mechanisms occurring within the reservoir is influence characterized by thick Upper ‘R’ Nummulitic
emphasized. limestone facies limited to paleo-highs generated by the
Then, we detail the way the dynamic model was build growing anticlines.
using a novel discrete fracture network (DFN) approach • Thin, low energy, bioclastic limestone and dolomite in
developed internally. The method populates the static Lower ‘R’.
parameters at the full field scale using a geo-statistical process The field was appraised by three wells. W01, drilled
guided by a geological driver. Using an innovative technique vertically on the crest, found gas in the Upper ‘R’ and oil in
(the Elementary Patch), a fast calculation of fracture the Lower ‘R’. W02, drilled vertically on the flank, was water
permeability in each grid cell and automatic calibration on the bearing. W03, drilled slanted and up dip of W02, crossed the
well test data can be made. We also discuss how the best WOC and triggered the decision to develop the field since it
choice for the simulation model (single porosity vs double proved an overall oil thickness of 220m and exhibited
porosity) depends not only on the fracture-to-matrix characteristics even better than those encountered in W01. The
permeability ratio but also on the magnitude of the shape total reservoir thickness is around 250m and most of the oil
factor and on the production mechanism. The final dual lies in the Upper ‘R’. The dip varies from 4° westward to 30°
porosity–dual permeability model integrates the intermediate on the northern and southern flanks. Rock data show better
scale fractures implicitly and the conductive faults explicitly properties in the Upper (φ=17%, k=30mD) than in the Lower
whereas the smallest scale is discarded as it has a negligible (φ=12%, k=5mD).
impact. This field has been on stream since 2003. It is currently
In summary, we present a complete and innovative producing at a plateau of 40000 bopd through 5 horizontal and
workflow process for a geoscience study on NFR and 1 vertical wells. Additional wells and gas re-injection are
demonstrate its application in a real case study. The need for planned. In this context, a good characterization of the
full data integration to achieve a comprehensive understanding fractures was deemed critical.
2 SPE 102165

Fracture characterisation • Small extension, open, conductive fractures usually


In Naturally Fractured Reservoirs, characterization of the confined in tight beds, in clusters of 10-20ft. These
fractures is a difficult task that requires a rigorous workflow discontinuous conductive fractures represent 60 to 70% of
which can be summarized by the following steps. the overall natural fracture population.
• First, fractures sets with different orientation and scale • Large (decimeter scale in cores) conductive, open, highly
have to be identified. Scales typically range from diffuse dipping fractures in echelon, scattered in clusters of
fractures at one extreme to fracture corridors and discontinuous fractures or alone in more porous facies.
conductive faults at the other. These form 20-30% of the natural fracture population.
• Then for each fracture set, static parameters i.e. length Induced fractures are frequent in cores and images on the
and spacing distributions have to be defined. Well vertical wells. Breakout and induced fracture orientation is
imagery (FMI) is the most effective tool for assessing related to present day maximum horizontal stress, N160-N170,
spacing distributions but lengths can not be measured in agreement with the preferential wellbore elongation
directly in the reservoir. An outcrop analog is often the detected by SCHLUMBERGER CLIBAN module run in wells
best way to get an idea of this parameter. W04, W05 and W06. Natural fractures parallel to this
• Finally, the conductivity of each fracture set must be orientation are more likely to be widened by the drilling
estimated. This is usually the most difficult task. MDT is process.
the only tool enabling a selective fracture measurement Natural fracture orientations found by cluster analysis on
but it is often ineffective as radial flow within the fracture the full data set for reservoir ‘R’ are:
is concealed by the storage effect. Well test data are the • N70: 31% of all fractures, 77% of which are cemented.
most effective tool to assess the overall dynamic response • N150 to N180: 52% of all fractures, mostly open.
of a fractured reservoir but it is difficult to separate the • N110: 17% of all fractures, mainly in W08 but some in
individual conductivity of each fracture set. W03 & W07.
In terms of fracture density, cemented fractures are either
Static data. For reservoir ‘R’, all the available static data clustered (10 fractures/ft) in micro-faulted zones or scattered
were reviewed to build a conceptual fracture model. These elsewhere with intervals of 300 to 500ft (up to 1300ft at W08)
included outcrops, well imagery and conventional logs. without fractures.
Scales on outcrop. Very good outcrops of this Lower Open fractures are mainly clustered (10-15 fr/ft up to
Eocene Carbonate Platform have been found. They show 3 30fr/ft). Clusters are 10-30 ft thick, usually made of
main scales of deformation: discontinuous fractures, but occasionally of 100 to 200ft.
Bed Controlled Fractures (BCF): particularly well Continuous fractures (10 to 50%) are scattered in the density
developed in the multi-layered Lower ‘R’ formation, the BCF picks or individually in porous facies. Intervals without
are second-order heterogeneities in the thick (200m) and fractures range from 50 to 300ft.
hyper-compact Upper ‘R’. Fracture spacing has a reverse exponential distribution
Highly Persistent Fractures (HPF): these fractures have a with minimum spacing around 0.5ft, most of the values under
large horizontal extension (up to one kilometre) and cut the 5ft and maximum spacing around 200ft but up to 500ft.
platform over its whole thickness (> 200m). On the aerial In terms of lithology, best fracture density picks are
photographs, the mean spacing between HPF appears to range usually related to tight stringers but some thick tight levels (at
from 10 to 50m (2 or 3 sets of Highly Persistent Fractures bottom Upper ‘R’/ top Lower ‘R’) are not or only partially
generally co-exist.) These HPF are believed to exist in fractured. Density picks of diffuse, discontinuous conductive
reservoir ‘R’ and if open, they could act as gas/water conduits fractures are confined to tight beds while more continuous
during production. fractures are scattered in previous density picks or more
Faults: only a few faults have been observed on the individually in more porous facies.
outcrop. All these faults are normal faults. Vertical throw is Upper ‘R’ (more porous) is less fractured overall (0.15
generally small, always less than 20 meters (i.e. probably sub- fr/m) than Lower ‘R’ which has thick tight levels (0.25 fr/m).
seismic in the reservoir). 10 to 50 cm tectonic breccias always Static conceptual fracture model. From the above, it is
are associated to the fault planes. As these breccias are poorly possible to divide the deformation features into 3 classes (from
cemented, they could be also efficient gas/water carriers the largest to the smallest):
during oil production. • Faults (seismic and sub-seismic scale).
No “fracture corridors” (i.e. dense clusters of large • Highly persistent fractures (HPF): defined from the
extension fractures) were observed on the outcrop. Fracture frequency logs as isolated fractures with no other
clusters have been described in some reservoir R wells (W04 fractures within 10 ft on either side. Although they seem
for instance), but these are believed to be in fact Bed to cross the entire platform from outcrop analog, they
Controlled Fracture sets. could be confined to the sub-units (Upper or Lower ‘R’)
Well imagery. Image log data was available for 8 wells, 3 considering the rock mechanical differences based on
of them calibrated on cored intervals. Core and image log porosity logs.
studies reported 3 main types of fractures: • Bedding confined fractures (BCF): characterised on
• Highly persistent, thick, resistive cemented fractures (20- image logs by fracture clusters located in thin interbedded
40%), either scattered or in small clusters, and often porous and tight rocks. This mechanical context is very
related to micro-faults that are also cemented. favourable for the development of highly fractured beds.
SPE 102165 3

Three orientations (2 major and 1 minor) are interpreted flow, which is usually associated with the so-called
from the image logs: ‘hydraulic fracture model’, corresponds in the present
• N150: regional “background” fracturing that is always case to the flow in the vicinity of the conductive faults.
present. These fractures are sub-parallel to present day Matching the derivative enables the lateral extension of
maximum horizontal stress and almost always open. They the fault (linear) and its conductivity (bi-linear) to be
have a fairly homogeneous density throughout the field assessed. The typical values for reseroir R are
with an average spacing of 300ft. respectively about 3000ft and 100000mD.ft.
• N70: these fractures are parallel to the axis of the Some wells that show massive losses while drilling do not
structure. They are interpreted as fold related fractures as display linear flow on the derivative. The likely explanation is
their distribution shows a fairly strong variability with that the flow mechanism is a competition between two types
structural location. The spacing ranges from 20 to 150ft. of flow: radial around the drain and linear around the fault, as
They are 90% cemented at BCF scale and 60% at HPF illustrated in Fig 4. The signature will show the flow that
scale. dominates, i.e. linear if the fault extension is large compared
• N110: This set is a minor one and seems to develop to the drain length or radial in the opposite case.
locally (SW). Although its origin is still unclear, it could PLT. PLTs were run on all the wells shortly after first oil.
be related to some major faults. They are mostly open. The oil production profiles along the well bore show a clear
These 3 orientations are present in both BCF and HPF sets. relationship between productivity and vicinity of a fractured
zone, i.e. the productivity is not controlled by the matrix but
Dynamic data. Dynamic data have been carefully looked at by the fracture parameters. One example is illustrated in Fig 5.
to check the consistency of the previously defined static model Well W05 shows a strong productivity decline for the intervals
and to characterize the dynamic behaviour of the different at the toe. The porosity log on the corresponding interval is
fracture sets. The dynamic data consist of well test (DST) and uniform. The HPF observed on well imagery at 11400 ft MD
PBU data, production data, production logging tool (PLT) is controlling the productivity of the nearby intervals: the
data, core data measurements and mud loss data. They provide further from the HPF, the lower the productivity.
insights into the relative conductivity of each fracture scale. Mud losses. Mud losses are an important source of data for
DST and PBU. DSTs were performed on two of the ranking the conductivity of the different fracture sets
vertical exploration-delineation wells, in both reservoir units encountered during the drilling of the wells. Major losses have
(Upper and Lower). No dual porosity effect can be seen on the been monitored while drilling across subseismic faults.
tests and the interpreted kh values suggest slightly larger Significant losses have also been reported at some, but not at
permeability than those measured on cores (within a ratio of all HPFs (Fig 6). No measurable losses were associated with
about 2). At the time the field was developed there was thus BCF. This information is of key importance for understanding
little information on the possible dynamic impact of the the hierarchy of conductivity among the different fractures
fractures. scales.
The field has been developed with 500 m long horizontal Production data. The production data are also key for
drains mainly targeting the Upper R unit of better understanding the major contribution of the fractures during
productivity. Two wells target both the Upper and Lower production. The major observations are:
units. The wells have been completed with 7’’ liners and up to • Productivities of the wells put on stream were better than
10 intervals were perforated along each drain. The initial anticipated.
completion strategy carefully avoided opening the wells in • When wells were produced beyond their critical rate, gas
fractured zones identified on well imagery. The minimum breakthrough was observed after a few hours only,
stand off between the completion and the fractures was about suggesting a very limited pore volume was swept by the
30ft. This strategy was aimed of course at avoiding direct gas before reaching the wells. This limited pore volume
communication with the fracture network and early gas break can only be associated with fractures and is hence a good
though from the gas cap. Since the start of the production, an indicator of continuity within the fracture network. It is
intensive monitoring program has been set up on all the worth emphasizing that this gas coning effect was
producers. The well completions are equipped with permanent reversible. As the well was choked and oil production
down-hole gauges and regular PBU have been recorded. A decreased, the GOR stabilised at the dissolved gas ratio
screening of the available derivative signatures on the early level.
data (not yet affected by gas break through effects) shows two
major types of behaviour: Dynamic impact of the different fractures scales. The
• Some wells show two radial stabilizations (Fig 2): the static synthesis has identified three scales of fractures that co-
early stabilization is interpreted as the vertical radial flow exist in the reservoir:
and is dependant on kv.kh whereas the later one is • BCF seems to have almost no dynamic impact. No mud
interpreted as the horizontal flow and yields in kh.h where losses reported in tight layers with BCF. Permeability
h is the investigated reservoir thickness. might be slightly enhanced only.
• Some wells show linear flows (preceeded sometimes by • HPF: Their dynamic impact seems to be significant
bi-linear) (Fig 3). This type of signature is associated with according to the mud losses.
wells cross-cutting or lying very close to conductive faults • Conductive Faults: They have the largest conductivity but
identified by heavy mud-losses while drilling. The linear they are isolated features.
4 SPE 102165

It is important to emphasize that the dynamic impact of a on test data starts to be significant. The denser the
given fracture scale at the reservoir stage will be dependent on network, the higher the equivalent test permeability.
conductivity but also on connectivity. As a result, three This kind of relationship has been already published for small
quantitative parameters will be of the utmost importance to scale (diffuse) fractures in the North Sea (1).
characterize the possible impact of the fracture sets: One may want to split the matrix and frature dynamic
• The conductivity of the fractures. contribution on the above plot. The basic idea is to consider a
• The spacing and length of the fractures, which control the roughly uniform matrix contribution of about 15000mD.ft
connectivity of the set. throughout the field (which is suggested by well data) and to
A good example is provided by the comparison of the subtract this value from the test kh. This new relationship (Fig
relative dynamic impacts of faults and HPF in reservoir R. The 7) has been used later on for the dynamic calibration of the
faults will have a very high local impact but since their fracture conductivities in our DFN model.
spacing is large, they do not form a connected network and To bolster the above analysis, we have tried to simulate the
their overall impact at field scale will be moderate. HPFs behaviour illustrated by the cross-plot using a DFN. The
which are much denser and which form a connected fracture procedure was the following:
network should have a more significant impact. • For a given HPF N70 FF, create several (10) geostatistical
The comparison between the core data and the realisations of the N70-N150 HPF system and calculate
permeability derived from the tests has not revealed a large the equivalent permeability tensor on a 150m by 150m
discrepancy between matrix and fracture permeability and thus domain (size of a grid block). The fracture lengths of both
is not conclusive in proving the dynamic impact of the sets were taken as 300m..
fractures. To better estimate the impact of the HPF, we have • Check the maximum and minimum eigenvalues of the
checked the possible correlation between parameters permeability tensors (kmax and kmin) and use flags to
controlling the fracture distribution and dynamic data. On the characterize them: “Non” if kmax=0, “Partially” if kmax>0
graph below, we have plotted the test kh on the y-axis vs the and kmin=0 and “Fully” for kmin>0.
HPF N70 fracture frequency (FF) on the x-axis. HPF N150 FF • Re-do the process for different N70 FF values (from 0 to
is uniform throughout the reservoir and its analysis is thus of 0.15 by 0.01 increments).
limited interest. The results are plotted on the graph below:
KH vs N70 frac density Connectivity function of HPF N70 FF
50000
10
45000 9
40000 8
Nb of realisations

35000 7
6
KH test (mD.ft)

30000
5
25000
4
20000 3
15000 2
10000
1
0
5000
0,04

0,05

0,1

0,15
0

0,01

0,02

0,03

0,06

0,07

0,08

0,09

0,11

0,12

0,13

0,14
0
0 0,02 0,04 0,06 0,08 0,1 0,12 0,14 0,16 FF N70
N70 frac density (frac/m)
Fully Partially Non

The plot clearly illustrates a correlation which indicates It can be seen that a steep change occurs in the proportion of
that the N70 HPF is the key fracture set that controls the “Fully” flags above an FF of 0.05. which can be considered as
dynamics of the field. The plot shows two different trends: the percolation threshold. Beyond this value, the proportion of
• For N70 FF below 0.07frac/m, the plot is flat with a tensors with two non-zero components is large enough to
plateau of about 15000mD.ft. This kh value is roughly ensure continuity within the fracture network.
equal to the kh associated with the matrix according to
core data. Fracture modeling
• For N70 FF above 0.07frac/m, the plot shows a linear The fracture characterization described above has helped us to
increase of kh with fracture frequency. understand the key features that control the dynamic
These observations can be analysed the following way: behaviour of the field. We have emphasized the important role
• The 0.07frac/m N70 FF is interpreted as the percolation played by the HPF in increasing locally the equivalent
threshold of the HPF network i.e. the value beyond which permeability and in adding heterogeneity to the sweep
the N70 and N150 HPFs are dense enough to build up a efficiency.
connected fracture network. We emphasize also the need for using a dual porosity
• Below the threshold value, the HPFs are isolated in the model for modelling Reservoir ‘R’ even though the equivalent
matrix and the equivalent test permeability is permeability of the fractures is only slightly larger than the
representative of the matrix contribution only. permeability of the matrix. As soon as two phase flows are
• Beyond the threshold value, the HPFs form a connected involved (gas coning from gas cap and later gas injection), it is
fracture network and the dynamic impact of the fractures of the utmost importance to be able to reproduce with the
simulator, the differential sweep efficiency between the
SPE 102165 5

fractures and the matrix. A dual porosity-dual permeability A specific conductivity calibration module is also available
model has therefore been built. An internal DFN tool named in GoFraK. Condutivities of the different fractures sets are
GoFraK has been used to calculate the fracture permeability seldom measured data. Practically, the only available dynamic
grids. In the next sections, we first detail the main features of information comes from test data. GoFraK’s calibration
the tool, then we explain its use for the modeling of reservoir module calculates the equivalent test permeability within a
‘R’. user defined investigation area. The value is compared to the
actual data and GoFraK iterates on the conductivity of the
GoFraK. GoFraK is a new DFN developed by TOTAL & different fracture sets till a good match is achieved. The
FSS International. The full Workflow, from static to dynamic simulation of the test data assumes steady-state conditions and
modelling has now been available for more than a year. is not suited to reproduce the transient flow observed during a
Previous publication (2,3 &4) shows the use of the tool’s static PBU or DST. This calibration module is associated with a
module on a field case study. The current paper describes the sensitivity module that allows checking the influence of the
application of the full GoFraK workflow for modelling of main input parameters on the equivalent fracture permeability.
reservoir R. It requires a manual match but an automatic inversion module
Workflow. The main steps are detailed here after: will be available soon.
• Horizons of the main mechanical units and the grid into
GoCad. Reservoir modelling. As mentioned earlier, the HPF has
• Import the well trajectory as well as the imagery log data the major dynamic impact in the reservoir and it was thus
(strike directions and dip) for each fracture set. decided to model this specific scale with the DFN. The HPF
• Process the data (automatic correction of the well spacing is small compared to the grid block size, so they can
deviation) and define histograms distributions for the be represented implicitly in the simulation grid using GoFraK.
different fractures sets. Conductive faults also have a strong dynamic impact, but
• Use the principal component analysis module to define due to their size and spacing were modeled explicitly in the
the best suited driver using seismic data, curvature maps. fracture grid. The modeling was limited to the faults showing
GoFraK delivers a full field gridded map with the significant mud losses.
principal component and bi-variate histograms for each BCF have not been represented in the dynamic model
fracture set. because their dynamic impact is not significant.
• Geostatistical conditional simulation of FF to deliver full Static grid building. As HPFs are thought to have a large
field grids of FF for each fracture set. vertical continuity, the reservoir has been divided in two
• Definition of conductivities and lengths per fracture set. mechanical units: one for the Upper and one for the Lower R.
Uniform values throughout the field are usually assumed Well imagery has been analysed and statistics calculated
but an option allows for the conductivity to be correlated separately for each mechanical unit. Three sets of HPF were
with spacing via a power-law model. previously defined with orientations N70, N150 and N110.
The latter is not well developed and mainly located in the
• Processing for each grid block of the equivalent fracture
permeability tensor and shape factors. south-western part of the reservoir. It is assumed to have little
impact and was discarded from the modeling. The Fracture
Main strengths of GoFraK. The main strength of GoFraK
Frequency distributions for the other sets show:
compared to its competitors lies in the use of the Elementary
Patch technique which enables to focus on the smallest • A quite uniform distribution throughout the reservoir for
periodic element in the fracture network i.e. the patch is build the N150 set which is considered as background regional
for each grid block and the equivalent properties are processed fracturing. These fractures are open, oriented in the main
on the patch only as illustrated below: stress direction and hence assumed to have the maximum
conductivity.
• A distribution strongly related to the folding for the axial
N70 set. The maximum curvature turned out to be the best
fracture frequency driver. Fig 8 provides the curvature
map showing the largest values are in the area of rapid dip
variation between the crest and the flanks. Fig 9 shows
the correlation between the fracture frequency and the
curvature established on actual data.
The GoFraK geostatistical module was used to populate
the grid in fracture frequency distributions and several
realizations were generated. A ranking of the realisations was
performed based on the frequency predicted by GoFraK in the
This method is effective in:
vicinity of the wells where no imagery data was available but
• Saving simulation time in the processing of the grid.
whose kh was used to estimate the N70 fractures frequency.
• Assessing the exact permeability tensor i.e. without any The N70 fracture frequency grid selected for the dynamic
assumptions on the boundary conditions since the modeling is given in Fig 10.
periodicity of the patch ensures a periodicity in the Dynamic grid building. In GoFraK, each fracture set is
boundary conditions. assumed to have constant length and conductivity parameters
6 SPE 102165

throughout the field. These parameters can not be measured We emphasize that the only way to model this reversible gas
directly. We use the following as a best guess: coning mechanism was to use a double pososity model. The
• An average length of 1000ft for both orientation sets. This principal task of the history match consisted in tuning the
data is consistent with observations on outcrops. parameters controlling the magnitude of the gravity drainage
Furthermore, it turned out to be consistent with the mechanism to match the GOR rises. Those parameters are:
percolation threshold observed on actual data (see • The matrix-block height. The base case value was taken
Fracture characterization above). as the good matrix layer thickness i.e. about 30 ft (Fig
• Initial conductivity derived from the equivalent fracture 13). Indeed, when the gas flows down the HPF, it will
permeability from well tests i.e. C=kf/FF. A ratio of 10 also diffuse horizontally within the tight layers where
has been considered between N70 and N150 sets to BCF enhance locally the permeability. The gas saturation
account for likely anisotropy effects related to current day increase in the tight layers will partially disconnect the
stress (parallel to N150), giving respectively 3000 and porous matrix blocks by creating a baffle zone with
30000mD.ft. lowered oil mobility. This creates a “dynamic block”
These input parameters have been processed by GoFraK to effect where the block height is equal to the thickness of
generate permeability tensors for each grid block of each the porous layers.
mechanical unit. • The shape factor, which impacts the exchange rate.
Dynamic grid calibration. The calibration module of It is important to point out that those parameters will have
GoFraK has been used to match the test permeability within the same effect on the GOR and as a result multiple matching
the user defined investigation area playing on conductivities of solutions are possible. Gas saturation measurements (TDT
the different sets. Some slight modifications of the initial best logs) in the gas swept area would help to constrain the match
guess values were required. For W07, it was not possible to by adding valuable information on the capillary trapping of oil
match the low permeability of the test with the fracture in the matrix i.e. the block height. Unfortunately, such
frequency associated with the static grid. It is thought that with measurements are not possible given the layout of the
the fracture length considered, the percolation threshold is horizontal development wells in this reservoir. Since the
slightly too optimistic in GoFraK and results in a connected uncertainty on the matrix-block height is limited compared to
network in the vicinity of W07 whereas the actual data suggest the uncertainty on the shape factor, it was decided during the
the opposite. However, since the only mismatch was observed history match to tune the latter parameter only.
for this well, it was decided to not reprocess GoFraK with A global multiplier for the whole field was applied to the
different length assumptions. Instead, local modifications to shape factor values calculated by GoFraK. A decrease by a
the fracture grid were made manually. factor of 50 was required, which means that the gravity
The principle directions of the permeability tensors were in drainage block size is 7 times larger than the block sizes
line with the orientation of the grid and hence the principal calculated by GoFraK, which are based on the geometry of the
components were exported directly to the reservoir simulator. fracture network, which in turn depends on fracture frequency.
The tensors were calculated for the two mechanical units i.e. Though this kind of discrepancy is common in fractured
Lower and Upper: there was assumed to be no vertical reservoirs, the analysis of the causes is seldom developed.
variation within a given unit. The resulting map for the Naturally fractured reservoirs are a particular case of very
permeability in the N70 direction is shown in Fig11. heterogeneous reservoirs. Fractures and matrix are tightly
History Match. The production history is short but the imbricated and their properties are very different. One way to
early time data are rich in information in terms of critical rate tackle the complex issue of reservoir modeling is to consider a
and GOR variations. First a match in pressure was performed dual porosity model which represents separately the fractures
by adjusting the aquifer activity and the transmissivity and the matrix. This means that properties are modeled
between Upper and Lower. Then, the GOR variations of the separately and pressure and saturation are handled separately.
wells were matched. Refined historical data (on a daily basis) Though this represents a huge improvement over a single
were required to capture the fast production changes. Since the porosity model, the heterogeneity commonly observed within
permeability field had already been calibrated on the well test the fracture network itself is only partially accounted for. In
data, the viscous forces are correctly simulated in the reservoir the present case, the assumption that all the HPFs within a
and so the critical coning rate should also be correct. given set share the same conductivities and lengths might not
However, the rise in GOR depends on the exchange between be realistic. This assumption is made because of the limited
the fracture and the matrix. Indeed, two mechanisms compete amount of available data. We have some evidence from mud
within the gas cone developing between the gas cap and the losses of low conductive HPF but the organization of
producers: conductivity within the reservoir is poorly understood. When
• The gas tends to flow from the GOC to the well through two phase flows are involved, the gas will likely move into a
the fracture network. The gas front velocity depends on few HPFs leading to an equivalent dynamic block size much
kf/φf. larger than the static one.
• The gas column within the fracture creates a gravity The final match was acceptable (typical result given for
drainage mechanism which tends to force the gas into the well W03 in Fig 12) and most of the major GOR changes have
matrix and expel the oil. This mechanism mitigates the been well captured. It is important to point out that GOR
differential sweep efficiency between the fractures and changes are related to oil rate fluctuations around the critical
the matrix. value. One or two wells have a constant GOR slightly higher
than the dissolved GOR. This may be due to gas channelling
SPE 102165 7

effects the casing. The gas saturations after one year of The major outcomes of this study will help to optimize the
production are griven in Figs 14 & 15 for both the fractures way this newly developed reservoir will be produced. This
and the matrix. They show significant differences in term of includes reservoir management activities and the future gas
sweep efficiency. The average gas saturation in the matrix injection project.
behind the front remains high in the model but this might be
different if a block height other than 30 ft was used. For the Acknowledgments
forecast simulations, a range of 15 to 60 ft has been retained. The authors thank NOC, CPTL, Wintershall and Total for the
For each value in this range, the shape factors have been permission to publish this paper.
readjusted to keep a consistent match of the historical GOR.
Nomenclature
Conclusions f,m Fracture, Matrix
This study has shown how the integration of all dynamic and BCF Bed Confined Fractures
static data is essential for the characterization of a complex C Conductivity of a fracture
fractured reservoir. Characterization of fractured reservoirs FF Fracture Frequency
requires a large amount of data as the level of heterogeneity is HPF Highly Persistent Fractures
very high. Unfortunately, those data are seldom available at k Permeability
the time the field development decision is taken. It is thus φ Porosity
important to phase the development and perform an intensive
monitoring program during the early life of the field. References
Specifically, outcrop data was essential in understanding [1] A. Toublanc & al. Ekofisk Field: fracture permeability evaluation
the overall organisation of fracturing with the reservoir and in and implementation in the flow model. Published in Petroleum
interpreting the well imagery data. The latter provided Geoscience, Vol 11 2005, pp 321-330.
information on fracture frequency and orientation for each set [2] B. Gauthier, M. Garcia. Static and dynamic characterization of
fracture pattern in the Arab reservoirs of the ABK Field,
of fractures within the reservoir itself. Mud loss and PLT data offshore Abu Dhabi: from well data to full field modeling. Paper
showed which fracture sets were conductive locally, but well ADIPEC 10-22, October 2002.
test and PBU data were needed to assess the connectivity of [3] Gauthier & Al (2000), ‘Integrated reservoir characterization: a
the fracture sets. Finally, matching GOR variations with rate case study in North African Field’ SPE 65118.
during the early stages of production provided information on [4] Gauthier B, Garcia M, Daniel JM (2002) ‘Integrated reservoir
the effective matrix block size. characterization: a case study in a North African Field’ SPE
Analysis of linear flow regimes observed on PBU 79105.
derivatives enabled fault conductivities (up to 100000 mD.ft)
and extensions (up to 3000ft) to be inferred. However, the
faults do not form a connected network, so have only a local
impact on production.
A correlation between well test derived kh values and
fracture frequency for the N70 HPF set revealed that it is the
HPF that have the most significant impact on production. The
existence of a frequency threshold for connectivity of the
fracture network was inferred from the test data, consistent
with results from the DFN model.
A correlation between HPF fracture frequency and
curvature was demonstrated and used in GoFraK to generate
fracture permeability for input to a dual porosity-dual
permeability simulation model. The elementaty patch
technique and the automatic calibration module in GoFraK
have considerably improved the traditional workflow and have
been very time effective. BCF have not been represented as
this scale did not show a significant dynamic impact but the
matrix block height of the model accounts for a gas
disconnection effect related to BCF. Conductive faults have
been represented explicitely within the fracture grid itself.
Model history matching was relatively easy thanks to the
calibration against test permeability performed in GoFraK.
The only tuning parameters used for matching the GOR rise
are those controlling the gas gravity drainage mechanism i.e.
the block height and the shape factor. A dynamic shape factor,
significantly different from the static shape factor, has been
obtained. The difference is likely related to the heterogeneity
in the dynamic properties of the HPF fracture set itself
throughout the reservoir.
8 SPE 102165

Fault extension Xf

GOC
OWC Radial flow Linear flow
B07

B3 9
9 5 B1a
B06
BD1 platform
Horiz Wellbore
5

B04

B2 B05
Drain Length L
20
B08 18
Pl an e s
(Stri k e) We i g hte d H i s to g ra m

View from Top to Bottom Fault


Axi a l N = 29

Fig 1: Reservoir R TOP structural map Fig 4: Flow models in the vicinity of wells and faults

W05: intervals PI

PI (bopd/psi)
30,0

25,0

20,0
Upper R
10 -2

Lower R
15,0
DP & DERIVATIVE (PSI/STB/D)

10,0

5,0
10 -3

PD =1/2

0,0
1 2 3 4 5 6 7 8 9 10 11 12 13 14

Isolated fractures
10 -4

« Isolated No Fractures
10 -5

10 -3 10 -2 10 -1 10 0 10 1
Delta-T (hr)

poor image quality

Fig 2: PBU derivative signature without conductive fault

ENDWBS
-1
10

Fig 5: PLT per interval along W05 – Imagery logs


DP&DERIVATIVE(PSI/STB/D)

Fractures Flow in - Flow out

200 Clustered fractures (BCF) in bedding Isolated N150 fracture


-2
10

150 No mudlosses Mudlosses


100
50
0
-3
10

-50
-100
-150
-4

-200
10

14050

14100

14150

10 - .5 10 .5 10 1.5
Delta-T (hr)

Fig 3: PBU Derivative signature with conductive fault Fig 6: Mud losses log
SPE 102165 9

KH vs N70 frac density

35000

30000

25000
F
KH frac (mD.ft)

20000 *

15000

10000

5000

L o w H ig h
0
0 0,02 0,04 0,06 0,08 0,1 0,12 0,14 0,16
N70 frac density (frac/m)

Fig 7: KH fracture vs N70 HPF FF Fig 10: Fracture frequency grif for N70

High Low

Fig 8: Maximum curvature maps Fig 11: Axial fracture permeability map

Correlation coeff =0.73


Actual data

Simulation

Fig 9: N70 HPF Fracture Frequency vs maximum curvature Fig 12: GOR match of well W03
10 SPE 102165

L
Hectometric

Block height P
BCF

HPF

Fig 13: Matrix block height

Fig 14: Gas saturation in the matrix fater 1 year of production

Fig 15: Gas saturation in the fracture after 1 year of production