Copyright 2001, Society of Petroleum Engineers Inc.

This paper was prepared for presentation at the 2001 SPE Middle East Oil Show held inBahrain, 17–20 March 2001.

This paper was selected for presentation by an SPE Program Committee following review ofinformation contained in an abstract submitted by the author(s). Contents of the paper, aspresented, have not been reviewed by the Society of Petroleum Engineers and are subject tocorrection by the author(s). The material, as presented, does not necessarily reflect anyposition of the Society of Petroleum Engineers, its officers, or members. Papers presented atSPE meetings are subject to publication review by Editorial Committees of the Society ofPetroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paperfor commercial purposes without the written consent of the Society of Petroleum Engineers isprohibited. Permission to reproduce in print is restricted to an abstract of not more than 300words; illustrations may not be copied. The abstract must contain conspicuousacknowledgment 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.

AbstractThe paper describes the upscaling and reservoir simulation ofa giant Middle East oilfield, whose geological modeling isdescribed in a companion paper (1). The main objective of thestudy was the simulation of the irregular water advanceobserved in some parts of the field, as a consequence ofperipheral water injection.

Three scales of heterogeneity were identified in thecharacterization phase, namely the matrix, the stratiformSuper-K intervals and the fractures. To accommodate thedifferent hydraulic properties of each heterogeneity system, adual-media approach (dual porosity and dual permeability)was used.

The assignment of the effective properties to thesimulation grids (matrix and fracture grids) was performedindependently for the three heterogeneity systems. Inparticular, the geostatistical facies model was upscaled usingalgebraic methods, while the stratiform Super-K layers andfractures properties were explicitly reproduced at thesimulation gridblock scale, through an original upscalingprocedure.

The history match was achieved in a short time, by a smallvariation of the fractal dimension of the fracture distributionand without resorting to any local modification.

Simulation results showed that the fracture system was thecontrolling factor in terms of water advance and breakthrough,while the impact of the stratiform Super-K layers proved to beof second order.

In a later stage, the model was utilized to run productionforecasts under different exploitation scenarios.

Conclusions of this study indicate that for such porous andfractured reservoirs with stratiform Super-K occurrences, adetailed characterization of all the heterogeneity systems,coupled with a dual-media formulation, are necessaryrequisites for accurate reservoir simulation and effectivereservoir management.

IntroductionHeterogeneities are always present, to some degree, in naturalpetroleum reservoirs (2). Their impact can be very importantin the overall dynamic behavior of the reservoirs, especiallywhen secondary recovery project are active, e.g., water or gasinjection.

In the Middle East area, many oil reservoirs are currentlyexperiencing unexpected production performance, especiallyearly water breakthrough, which started usually soon after theimplementation of waterflooding projects. In most studies,such unexpected behavior is generically related to thepresence of reservoir heterogeneity, in the form of some highpermeability conduits which link the injector and the producerwells. Note that, while such simplified understanding can besufficient for a history matching exercise, a much betterdescription of the reservoir heterogeneity is required, in termsof type and distribution, when the simulation model is used inforecasting mode.

This project concentrated on the geological description,upscaling and numerical simulation of a giant Middle Eastcarbonate reservoir, which experienced early waterbreakthrough in some parts of the field. Since it was felt thatreservoir heterogeneity was the driving factor behind thisunexpected behavior, most of the effort was devoted to thedescription and simulation of such heterogeneity.

The geological characterization of the reservoir isdescribed in a companion paper (1). Three main heterogeneitysystems were identified, namely:

1. The matrix. This is constituted by porous and permeablelimestones and dolomites. A rock-type classificationsystem was established by means of a multivariatestatistical algorithm. Vertical and horizontal proportioncurves were generated within a sequence stratigraphyframework, which showed a strong non-stationary

SPE 68184

Integrated Study of a Fractured Middle East Reservoir with Stratiform Super-KIntervals – Part 2: Upscaling and Dual Media SimulationL.Cosentino and Y.Coury (Beicip-Franlab), J.M.Daniel, E.Manceau, C.Ravenne and P.Van Lingen (IFP), J.Cole andM.Sengul (Saudi Aramco)

2 L.COSENTINO et al. SPE 68184

behavior. Eventually, a 3D facies geostatistical model wasgenerated. Petrophysical properties were assigned to thefine scale grid through an original inversion algorithmbased on flowmeter data (3).

2. The stratiform Super-K intervals. These are thin, highpermeability layers with conductivity in excess of 500bbl/day/feet. Such intervals are defined through theanalysis of the available flowmeters and they have beenincluded into the geostatistical model using aplurigaussian algorithm. A strong relationship betweenstratigraphic position and stratiform Super K intervalswas demonstrated. The estension, the shape and thecontinuity of these bodies are largely unknown and arekey factors of the characterization phases.

3. The fractures. The presence of fractures has beeninferred through the analysis of data from seismicinterpretation, curvature analysis and wellbore logs.Fractures do cluster in swarms that can be represented asheavily fractured lineaments (called fracture in thefollowing). All the available data concerning thesefracture swarms were integrated in a stochastic fault andfracture model and alternative images were generatedthrough a fractal approach.

In the next sections, we describe the impact of thesereservoir heterogeneities in the field behavior and theimplementation of the available geological characterization inthe simulation model.

Problem StatementThe field under study is a giant carbonate reservoir, which hasbeen on stream for almost 30 years. The reservoir has beenproducing under fluid expansion and weak aquifer drive until1995, when a waterflooding project started. Fig. 1 shows amap of the reservoir structure. Note that the field is a part of amuch larger oil accumulation which extends to the South andto the North, therefore these limits do not correspond to realreservoir boundaries. The inner window shows the arearetained for numerical simulation (pilot area).

The main problem of the field under study is early waterbreakthrough, especially in some wells located in the Westernflank. Such behavior has been observed in recent years, afterthe start of the water injection project, while no significantwater production had been measured before that date.

The situation is depicted in Fig. 2, where all the dry andwet producers of the studied zone are indicated, together withthe approximate position of the actual water front in the twoflanks. As it can be appreciated, the water has moved fasterand in an irregular way in the West flank, while in the Eastflank the water front appears in general smoother and moreregular.

In particular, some of the wells of the West flankexperienced early water breakthrough only few months afterproduction start-up, despite being several kilometers distantfrom the original contact. This corresponds to an approximatewater velocity of 20-40 meters/day. On the other hand, some

wells located much closer to the original water contact are stilldry.

The above observations suggest that this reservoir behavesas a Dual Media reservoir. Actually, 2 flow systems can beidentified, with completely different hydraulic properties, onemedium with high storage capacity and relatively lowconductivity and one medium with low storage capacity andhigh conductivity.

The former medium corresponds to the matrix, i.e. theporous and permeable limestone and dolomite facies. Here, inthe absence of significant reservoir heterogeneity, thesweeping process is rather stable and will eventually lead togood conformance factors.

The latter medium includes those reservoir heterogeneitiesthat are deemed responsible for the high water velocityobserved in some areas of the field. This medium thereforeincludes the faults and fractures described in the structuralstudy and the stratiform Super-K layers.

For such complex reservoirs, the dual-media approach(dual porosity – dual permeability) represents a convenientsimulation approach, since it explicitly accounts for thedifferent hydraulic properties of each heterogeneity system. Italso provides the necessary flexibility for testing the impact ofeach reservoir feature independently. Furthermore, the modelused in this study allows for the explicit modeling of thedifferent displacement mechanisms acting in a Dual Mediareservoir (expansion, capillarity, gravity, viscosity), thusproviding additional flexibility to the simulation study (4). Forthese reasons, since the beginning of the project, the DualMedia approach was considered the reference framework forsimulation purposes.

Alternative simulation strategies were also investigated,based on a single-medium approach and original pseudoisationprocedures (5). In some cases, these options provide a viablealternative to the more general Dual Media formulation.

Upscaling proceduresUpscaling aims at defining the effective reservoir properties atthe simulation grid scale. A sound upscaling procedure shouldreproduce, at the coarser scale of the simulation grid, the flowperformances expected if the corresponding fine scale modelwould be used. In a Dual Media context, the problem isdecoupled, since the simulation grids are actually two: thematrix and the fracture grids. This decoupling can be seen as afurther advantage of the Dual Media formulation, sincedistinct and specific procedure can be established for the twosystems, thus reducing the risk of excessive homogeneisation.

In the context of this study, the 3 heterogeneity systemsidentified in the geological phase, has been assigned to thesimulation model as a function of the expected flowperformances. Therefore, the matrix properties have beenupscaled to the matrix grid, while fractures and stratiformSuper-K properties have been upscaled to the fracture grid(even though in the geological modeling stratiform Super-Kintervals were modeled within the geostatistical matrix grid).Note that in the remaining of the text, whenever a reference is

SPE 68184 INTEGRATED STUDY OF A FRACTURED MIDDLE EAST RESERVOIR WITHSTRATIFORM SUPER-K INTERVALS–PART 2: UPSCALING AND DUAL MEDIA SIMULATION 3

made to the fracture grid, this actually includes the fracturesand the stratiform Super-K intervals.

In this section, the upscaling procedures utilized for the 3heterogeneity systems will be described. The model has beenbuilt using cells with dimension of 250*250 meters in the Xand Y directions, corresponding to an increase of a factor 2with respect to the geological model grid, which was 125*125meters. In the vertical direction, 13 layers were considered.This layering was designed in order to have sufficient detailfor the most important reservoir units, while keeping areasonable number of total cells. The number of cells in the X,Y and Z directions is 88*40*13. The total number of cells istherefore 45,760 for each of the simulation grids (matrix andfracture). Fig 3 shows two images of the simulation grids ofthe matrix and the fracture media (bottom layer). Note that thefracture grid can be imagined as a 3D, interconnectedframework of vertical drains (faults and fractures) andhorizontal drains (stratiform Super-K intervals).

MatrixThe upscaling of porosity to the simulation matrix grid hasbeen performed using a simple weighted average operator.The stratiform Super-K intervals of the geological model wereremoved, while a Net/Gross was computed as the ratiobetween matrix thickness (excluding the stratiform Super-Kthickness) and total thickness.

Permeability has been assigned to the matrix simulationgrid on the basis of the fine-scale permeability valuesavailable in the existing geostatistical model. As in the case ofporosity, stratiform Super-K intervals were removed from thecomputation. The upscaling procedure applied to compute theeffective permeability values for the simulation grid is basedon the method of Cardwell and Parsons (5). This algebraicmethod was deemed adequate, considering the relatively lowdegree of dispersion of the individual facies averagepermeability.

Stratiform Super-KStratiform Super-K were upscaled to the fracture grid using anoriginal procedure. In the XY plane, the presence and/or theabsence at the simulation grid scale was assessed on the basisof the number of cells existing at fine scale. Whenever 2 ormore cells were defined as stratiform Super-K in the fine scalegeological model, this was extended to the whole gridblock,no matter their relative disposition. This empirical rule isshown in Fig. 4.

In the vertical direction, for each simulation layer, thelocation of the stratiform super-K intervals, as well as theirthicknesses (or cumulative thickness, when more than 1interval is present) are computed, and then directly integratedin the calculation of fracture transmissivities and verticalblock sizes. In all cases, the vertical position of eachindividual stratiform super-K. In all cases, the vertical positionof each individual stratiform Super-K was maintained in thesimulation gridblock, to account for lateral communicationbetween distinct Super-K bodies.

From a petrophysical point of view, stratiform Super-Kintervals have been characterized with a constant porosityvalue equal to 35% and one single average value ofpermeability of 2 Darcies. This is the average permeabilityvalue for such bodies, as it was found through the analysis ofthe available flowmeter data. These values were assigned tothe simulation grid blocks where a stratiform Super-K intervalwas found.

Faults and FracturesThe faults and fractures system was generated through astochastic approach (7), based on the integration of differentkind of data. Since no properties were generated at fine scale,no explicit upscaling procedure has been applied to thefractures, and all the main fracture properties were defineddirectly at the simulation grid scale. This was performedthrough an original procedure, that automatically provide theinput files for simulation purposes.

First, the geometry of the fracture network generated bythe stochastic model is superimposed onto the simulation grid.The length and the location of the fractures in each cell arecomputed. Each fracture is approximated by a discretizedfracture as illustrated in Fig. 5. Then the effective fractureporosity, Φi,j, is computed as:

Φi,j = (di,j * efrac) / (∆Xi,j * ∆Yi,j)

Where:di,j : effective length of the fracture in cell (i,j)efrac: fracture thickness∆Xi,j : cell dimension in X direction∆Yi,j : cell dimension in Y direction

Note that if there are several fractures in a given cell, therelevant properties are added.

As far as fracture permeability is concerned, the procedureallows for the direct calculation of the transmissibility terms inthe fracture grid. When a fracture crosses two adjacent cells(i,j) and (i+1,j), the transmissibility term is given by:

TRANX(i,j)->(i+1,j) = (Kfrac * efrac * ∆Zi,j * 2) / (di,j + di+1,j)

Where:Kfrac : fracture permeability∆Zi,j : cell dimension in Z direction

Note that this formulation is normalized for the actuallength of the fracture, which is always less or equal to thediscretised length (see Fig. 5).

Fracture permeability has been chosen on the basis of theaverage value of hydraulic conductivity, i.e. the product ofaperture and fracture permeability:

CF = KF * e

After some tests, a constant conductivity value of 40Darcy*meter has been applied to all the fractures. This valuehas been derived on the basis of the available flowmeter data.

4 L.COSENTINO et al. SPE 68184

Another important parameter in Dual Media simulation isthe matrix block size, since this largely controls the exchangesbetween matrix and fractures. The methodology adopted tocompute the matrix block size is based on an ideal imbibitionprocess and it is described in detail in Ref. 8.

Phenomenological ModelBefore starting the pilot area simulation, a phenomenologicalmodel was built and run with the main objective of studyingthe relative impact of matrix, stratiform Super-K and fractureson the dynamic behaviour of the reservoir, particularly withrespect to the timing of the water breakthrough. Two versionsof this model were built: a fine scale (reference) SingleMedium model and a coarse grid, Dual Media model. Thiswas done with the aim of testing the validity of the procedurefor fracture properties determination described in the previoussection.

The phenomenological model was based on a twokilometers wide half-transversal section of the pilot model,located in the West flank of the field. The geometrical andpetrophysical characteristics were derived from the pilotmodel, in order to guarantee the consistency of the results.

The main results of the phenomenological model areshown in Fig. 6. The following points can be appreciated:

• In the matrix case, water breakthrough (BT) occurs 10years after the injection start.

• The existence of a typical stratiform Super-K (2 Darcy)has little impact on BT but only on water cut (Fw)evolution.

• The presence of a stratiform Super-K accelerates the BTonly when permeability in higher than 2 Darcy.

• In the presence of a fracture, water breakthrough occursmuch earlier than in the previous cases. In all the runanalyzed, water BT happens in the first year after theinjection start.

• Different fracture conductivities have an impact on Fwevolution, while the BT time is not significantly different(not shown in the figure).

• The Dual Media formulation is in remarkable agreementwith the Single Medium (fine scale) reference case. Thisconfirmed the validity of the upscaling proceduresdescribed in the previous section.

These results demonstrate that early water breakthrough isvirtually always related to fractures. This is what typicallyhappens in the West flank of the field, where the fracturing isintense, and the fractures often link up the injectors andproducers. Note that these results were anticipated through thesimple superposition of the interpreted fracture network andthe production performances (Fig. 7). As it can be appreciated,the wet wells happens to lie always on a fracture swarm, whilethe dry wells are located in unfractured areas.

On the other hand, stratiform Super-K may contribute toearly BT only when the Super-K permeability is high and

when their extension is significant, connecting for exampleinjector/producer pairs.

Dual Media Simulation ModelThe simulation model covers an area of approximately 220squared kilometers, corresponding to the inner window of Fig.1. This zone includes 39 producer and 18 injector wells.

The model was initialized with a bubble point map, toreflect the trends of saturation pressure and viscosity observedin the field. A relatively simple model of saturation functionswas used, with one curve and variable end points for thematrix medium and one single curve (cross-shaped) for thefracture medium. The model was run through the completehistorical period (28 years).

The history match of the saturations was achieved in ashort time frame. Actually, the initial geological model, interms of matrix, fracture and stratiform Super-K, provided avery reliable framework for simulation purposes. As it will beshown, only minor modifications were necessary to obtain asatisfactory match.

The sensitivity runs that were performed showed therelative impact of the various reservoir parameters in the finalresults. These are shown in Tab. 1. As it can be appreciated,three main parameters crop out as the most important reservoirfeatures, i.e. structural lineaments, hydraulic conductivity andmatrix capillary pressure.

Structural lineaments. These were defined in thegeological model by combining the faults coming fromseismic interpretation and the lineaments picked during thecurvature analysis. Stochastic fractures are generated aroundthese lineaments using a fractal technique. A better match ofthe field performance was obtained through a slight reduction(1.4 1.3) of the fractal dimension. This corresponds to anincrease of stochastic fracture clustering (less spread) aroundthe main lineaments.

Hydraulic conductivity. This is the product of fractureaperture and permeability. Several combination of the twoparameters were tested, within the conductivity range of 10 to40 D*meter. Eventually, a value of 40 D*meter was retained.

Matrix capillary pressure. This value is important in thatit defines the amount of water that is prone to be imbibed inthe matrix as a consequence of capillary exchanges. A highvalue corresponds to water wet rocks, where significantimbibition takes place. Lower values, corresponding to oil wetconditions, would prevent such imbibition and would allowthe water to run faster into the fracture system

Results of the simulation runs confirmed that the fracturesystem is by far the controlling factor in the waterdisplacement process. This is especially true in the West flank,where fracture swarms are more developed. Stratiform Super-K intervals, on the other hand, play a minor role and arerelated to early water breakthrough only in those cases wherethe Super-K directly connects injector/producer pairs, as wasalready shown in the phenomenological model. This is mostlikely to occur in the East flank, where fracturing is lessintense and several stratiform Super-K intervals have beenobserved.

SPE 68184 INTEGRATED STUDY OF A FRACTURED MIDDLE EAST RESERVOIR WITHSTRATIFORM SUPER-K INTERVALS–PART 2: UPSCALING AND DUAL MEDIA SIMULATION 5

Fig. 8 and Fig. 9 show the results of the simulation for twotypical wells of the West flank, in terms of water cut. The firstwell (Fig. 8) experienced early water breakthrough fewmonths after the injection start, despite being 4 Km far fromthe injectors line and high in the structure. On the other hand,the second well is located downstructure from the previousone, much closer to the injector wells (Fig. 9). As it can beappreciated, the model restitutes very well the behavior ofboth wells.

Total field results are shown in Fig. 10. The agreementbetween the measured data and the simulated profile isremarkable, both in terms of water breakthrough and water cuttrend. The total quantity of water is also well matched. Theseresults prove that the reservoir behavior has been correctlyreproduced in the simulator and give confidence in the resultsof the prediction phase.

Fig. 11 shows the results of the simulation in terms ofvertical distribution of oil and water rates for one key well.This is compared with the actual flowmeter results measuredat the same date. As it can be appreciated, the match isexcellent, which implies that the fluid withdrawal along thevertical direction (i.e., the vertical displacement) is alsocorrectly reproduced in the model. Note in this figure thepresence of a stratiform Super-K intervals (depth 5580 ft),which accounts for almost 70% of the total well production.This behaviour is well restituted in the simulation.

It should also be noted that these results were obtainedwithout resorting to any local modification of the inputparameters, neither in terms of static parameters (permeabilitydistribution), nor in terms of saturation functions. The onlytuning that was realized was related to global parameters (i.e.,fracture fractal dimension and matrix capillary pressure). Thisgives good confidence about the use of this model in theforecast phase.

Fig. 12 shows two images of the displacement process inthe fracture grid (water is in light grey). These images refer tothe present saturation conditions in different reservoir layers(left image the bottom layer, right image the top layer). Thecomparison of the 2 images highlights the segregation processtaking place in the fracture network. The analysis of severalimages of this type as a function of time (Fig. 13) revealedthat the water movement in the field is controlled by thecomplex intersecting network of stratiform (horizontal) Super-K and tectonic (vertical) fractures and faults, while thebackground limestone-dolomite matrix only plays a minorrole.

Finally, it should be remarked that simulation runs areperformed in a reasonable time frame (about 2 hours in a NECvector machine), despite the complexity of the reservoirdescription (total amount of cells), the duration of thehistorical production (28 years) and the Dual Mediaformulation.

Presently, this model is being used in a series ofproduction forecasts under different exploitation scenarios,both on a medium and a long term basis.

ConclusionsThe main results and conclusions of the present work can besummarized as follows:

• The analysis of the production performance of the fielddemonstrated that the reservoir behaves as a Dual Mediasystem, characterized by localized early breakthrough insome wells updip in the structure.

• Three heterogeneity systems have been identified andcharacterized in this field: matrix, stratiform Super-Kintervals and fractures.

• Original upscaling procedures have been developed tocompute the effective properties of each heterogeneitysystem. These allows for the preservation of the extremevalues of conductivity at the simulation grid scale.

• A phenomenological model has been set up with the mainobjective of studying the relative impact of the differentheterogeneity systems. Results of this study indicated thatwater breakthrough occurs much faster in the fracturesthan in the matrix and stratiform Super-K intervals.

• A Dual Media simulation has been performed in thecentral part of the field. This simulation allowed for aremarkable match of all the critical wells, as well as totalfield performance, without resorting to localmodifications. This model represents a sound basis forreliable production forecasts.

• The analysis of the results showed that water movementin the field is controlled by the complex network ofstratiform Super-K and tectonic fractures, while thematrix only plays a minor role.

• The available reservoir characterization, in terms ofmatrix, stratiform Super-K and fractures, provides anaccurate model of the reservoir. From this point of view,the model validates the characterization procedure utilizedin the previous phases of the study.

• The quality of the simulation results demonstrates that theDual Media approach represent a rigorous and flexibleapproach for such complex reservoirs. The additional timerequired to build the fracture grid is largely paid back bythe time saved during the history match (no well by welltuning and no local grid refinement required).

AcknowledgementsThe authors want to thank Marie Christine Cacas, SylvainSarda, Bernard Bourbiaux, Julien Seguin, Daniel Mas andJean Claude Sabathier for their essential contributions to thedevelopment of the ideas here presented. We also would liketo thank Beicip-Franlab, IFP and Saudi Aramco managementsfor the permission to publish this paper.

References1. Ravenne, C., Daniel J.M., Lecomte J.C., Camus D., Chautru J.M.,

Cosentino L., Coury Y., Cole J., Sengul M.: Integrated Study ofa Fractured Middle East Reservoir with Stratiform Super-KIntervals – Part 1 : Geological Model. SPE paper 68183,presented at the 2001 Middle East Oil Show, Bahrein.

6 L.COSENTINO et al. SPE 68184

2. Weber, K.J.: How heterogeneity Affects Oil Recovery. InReservoir Characterisation, Academic Press, 1986.

3. Mezghani, M., Van Lingen P., Cosentino L., Sengul M.:Conditioning Geostatistical Models to Flowmeter Data. SPEpaper 65122, presented at the Europec 2000, Paris.

4. Quandalle, P., Sabathier J.C.: Typical Features of a NewMultipurpose Reservoir Simulator. SPE paper 16007

5. Van Lingen, P., Daniel J.M., Cosentino L., Sengul M.: SingleMedium Simulation of Fractured Reservoirs. SPE paper 68165,presented at the 2001 Middle East Oil Show, Bahrein.

6. Cardwell, W.T., Parsons R.L.: Average Permeability ofHeterogeneous Oil Sands. Trans. AIME 1945.

7. Sabathier, J.C., Bourbiaux B.J., Cacas M.C., Sarda S.: A NewApproach of Fractured Reservoirs. SPE paper 39825.

8. Bourbiaux, B.J., Cacas M.C., Sarda S., Sabathier J.C. : A Fast andEfficient Methodology to Convert Fractured Reservoir Imagesinto Dual Porosity Models. SPE paper 38907.

SPE 68184 INTEGRATED STUDY OF A FRACTURED MIDDLE EAST RESERVOIR WITHSTRATIFORM SUPER-K INTERVALS–PART 2: UPSCALING AND DUAL MEDIA SIMULATION 7

Table 1 – Uncertainty attached to various simulation parameters and their impact on final results

Fig. 1 – Structural top map of the field. The inner window shows the area retained for numerical simulation

Uncertainty ImpactStatic parameters

Phi/K matrix * *Phi Super-K *** *K Super-K ** **Super-K distribution ** **Structural lineaments ** ***Stochastic Fractures *** **Hydraulic conductivity ** ***Dynamic parameters

PC matrix *** ***Kr matrix ** **Viscosity * *

8

15

16

17

18

19

20

21

22

23

24

29

64

6567 6869 70 71 72

78

7980

87

88

89

90

91

92

93

94

95

96

97

98

99

100

101102

103

105106

107 109

110

111

112

113116

117

118

120

121122

123

124

125

126

127

128

129

130

132

133

134

135

136

137

138

139

140

141 142

143

144146

147

148

149

150

152

9

42

55

85

300

301

302

303

304

305

306

307

308

309

310

311

312

313314

315

316

317

318

319

320

321

322

323

3246

329

330

331

8 L.COSENTINO et al. SPE 68184

Fig. 2 – Present water contact position. Circles indicate wet wells, squares dry wells. The grey bands towards the flanks show the original position of the oil-water contact.

Fig. 3 – Dual Media simulation grids: matrix grid (left) and fracture (fracture+stratiform Super-K) grid (right)

17

18

19

78

88

89

91

92

93

96

97

99

100

103

105106

107 109

110

111

81

120

124

125

126

127

129

132

134

135

36

137

139

140150

42

300

301

302304

308

309

11

312

313

316318

319

320

330

21

20

22

243

32

21

2 314

SPE 68184 INTEGRATED STUDY OF A FRACTURED MIDDLE EAST RESERVOIR WITHSTRATIFORM SUPER-K INTERVALS–PART 2: UPSCALING AND DUAL MEDIA SIMULATION 9

Fig. 4 – Upscaling of the stratiform Super-K facies in the XY plane

Fig. 5 – Procedure for determining the effective fracture porosity and transmissibility for the simulation (fracture) grid

1

2

3

1 2 3

d13

d23

For the fracture:

TRANX (13) (23) = Kfrac. efrac. ∆ Z .2/(d13+d23)

TRANY (23) (22) = Kfrac. efrac. ∆ Z .2/(d23+d33)

POROS (13) = d13 . efrac . / ( ∆ X . ∆ Y)

d22

Actual fissure

Discretized fissure

Discretization of a fracture

10 L.COSENTINO et al. SPE 68184

Fig. 6 – Results of the phenomenological model: water cut behaviour for the matrix system, the matrix plus one stratiformSuper-K interval (both in Equivalent Single Medium and Dual Media formulation) and the matrix plus one fracture.

Fig. 7 – Interpreted fracture pattern for the West flank and production results. Note that the wet wells(circles)lie in fractured areas, while dry wells are located in unfractured areas. The first row of wells to the left represents the injectors.

0

0.2

0.4

0.6

0.8

1

1981 1986 1991 1996 2001 2006 2011

Time

Matrix only (No SK)

SK perm=2D (SM)

SK perm=2D (DM)

Fracture conductivity20Dm

1

19

20

78

92

105

106

107

118

120

140

304

319

320

321

322

132

19

106

140

92

118

SPE 68184 INTEGRATED STUDY OF A FRACTURED MIDDLE EAST RESERVOIR WITHSTRATIFORM SUPER-K INTERVALS–PART 2: UPSCALING AND DUAL MEDIA SIMULATION 11

Fig. 8 – History match of a key well of the West flank which experienced early water breakthrough: water cut evolution.Above:position of the well (larger circle) with respect to the fault pattern and the injectors (small circles)

Fig. 9 – History match of a key well of the West flank which did not experienced early water breakthrough: water cut evolution.This well is located approximately 2 kilometers downdip with respect to the previous well

1975

1980

1985

1990

1995

1998

0

1

0.4

0.2

0.6

0.8

Wat

er c

ut (

adim

)

simulated

measured

1975

1980

1985

1990

1995

1998

0

1

0.4

0.2

0.6

0.8

Wat

er c

ut (

adim

)

simulated

measured

1975

1980

1985

1990

1995

1998

0

1

0.4

0.2

0.6

0.8

Wat

er c

ut (

adim

)

simulated

measured

1975

1980

1985

1990

1995

1998

0

1

0.4

0.2

0.6

0.8

Wat

er c

ut (

adim

)

simulated

measured

12 L.COSENTINO et al. SPE 68184

Fig. 10 – History match: total field

Fig. 11 – Match of the flowmeter results for one key well (total liquid rate and water cut)

050

100

5500

5600

5700

5800

5900

water (field)

total (field)

water (sim)

total (sim)

1975

1980

1985

1990

1995

1998

0

1

0.4

0.2

0.6

0.8

Wat

er c

ut (

adim

)

0

10

0.4

0.2

0.6

8

Standard cum

.water prod. (1e6bbl)

Measured water cut

Simulated water cut

Measured total water

Simulated water

1975

1980

1985

1990

1995

1998

0

1

0.4

0.2

0.6

0.8

Wat

er c

ut (

adim

)

0

10

0.4

0.2

0.6

8

Standard cum

.water prod. (1e6bbl)

Measured water cut

Simulated water cut

Measured total water

Simulated water

1975

1980

1985

1990

1995

1998

0

5000

0

100

150

Measured oil rate

Simulated oil rate

Cumulative oil

200000

150000

100000

50

200

250

300

Standard cum

.oil prod. (1e6bbl)S

tand

ard

surf

ace

oilr

ate

(bbl

/d)

1975

1980

1985

1990

1995

1998

0

5000

0

100

150

Measured oil rate

Simulated oil rate

Cumulative oil

200000

150000

100000

50

200

250

300

Standard cum

.oil prod. (1e6bbl)S

tand

ard

surf

ace

oilr

ate

(bbl

/d)

SPE 68184 INTEGRATED STUDY OF A FRACTURED MIDDLE EAST RESERVOIR WITHSTRATIFORM SUPER-K INTERVALS–PART 2: UPSCALING AND DUAL MEDIA SIMULATION 13

Fig. 12 – Present water position in the fracture grid: bottom layer (left) and top layer (right)

Fig. 13 – Cross sections showing water advance and segregation in the fracture grid (light grey).Note how fracture planes are connected through Stratiform Super-K intervals (horizontal links)