a’4P w

._ m I

IISodotY of PetroIeumEr@nwrsI

SPE 29118

Petroleum Reservoir Simulation in a Virtual EnvironmentJ.S. Jacobsen, E.W. Bethel, Akhil Datta-Gupta,* and P.J. Holland, Lawrence Berkeley Laboratory

●SPE Member

TM paper was preperedfor presentationat the 13fh SPE Sym~ium on ReservoirSimulationheld in San Antonio,TX, U.S.A., 12-15 February1S95,

This paper wee selectedfur presentationby an SPE ProgramCommitteefollowingreviewof informationoonteinedin an abetrecfsubmittedby the eufhor(a).Contantaof the paper,= mnfd ~W ~ *n m~awed by the *lety of petrokum EngineeJreand am aubjecfto oormofionby the author(a).The material,ee pnwmed, dam not rwceewily refkct~Y P@~ of the S@@ ~ petroleumEngineers,Ifaoffbem,or members.Pepempreeentedet SPE meetirva em subiaofto publicationreviewby EditorialCommitteesof the SocietyofPefmkum Engineem.Pefmiaaiontocopyiareefrktadtoan ebatr@fofIW4momthenSt)Owords.Illuefratk%wmaynotbecopied,TheabatmcfahoufdcontainoonepicuouaeoknMe@mmfof whereandby whomthe paper is presented.Write Libmrian,SPE, P.O. Sox SSSSSS,Rkherdson,TX 7SOSMS2S, U.S.A. Telex, 1S2245SPEUT.

ABSTRACT

In this paper, we describe an approach to combining areservoir simulation with 3-D visualization andvirtual reality technology. Our prototypeVR/visualization system minimizes human-machineinterface barriers and provides enhanced control

●L-ever 4.:-, ,1-+:ARLIIC allllulatwil, therebjj maximizing scientific

judgment and use of intuition. We illustrate thepractical advantage of using the VR/visualizationprototype system in reservoir engineering byvisualizing the results of a waterflood in an oil field~~th a ?~r~~.~imnneinnai, s~~t~ai!~ cnrr~lated,,, ,”,,”,”,, --- . -.-.-

heterogeneous permeability field.

INTRODUCTION

With the advent of multi-processor computers andthe increased use of stochastic methods tocharacterize heterogeneities, 3-D visualization hasbecome an integral part of modern reservoirsimulation. For the most part, however,visualization of simulation results is a postprocessing step that occurs only after the simulationhas run to completion. Given the size and complexityof problems in reservoir engineering, completion ofa simulation may take severs! hours or even days,

Often times simulation results are discarded becauseof errors in input files or because the given inputparameters did not give the desired simulationresults. In these cases both the reservoir engineer’stime and computer time are wasted.

In this work, we describe an approach to combining areservoir simulator with visualization software toolsand virtual reality (VR) technology. Our goal is tomake simulators easier to use and simulation results_-_! -easier to uride~sitiind. The benefits d 3- Dvisualization are fairly well known, however, thevisualization environment that we use provides arich assortment of visualization techniques forexamining different aspects of a simulation. Inaddition, VR output technology provides more flexibleand convincing ways to view simulation results.

The key advantage of the prototype VR/visualizationsystem is that the researcher or reservoir engineersees the output from the simulation as soon as it hasbeen calculated. This allows early detection of errorsin input parameters, poorly formulated reservoirscenarios, and other cases in which unexpectedproblems arise during the simulation that result inthe input needing to be revised and the simulationrerun. The VR input device facilitates specificationof new three-dimensional input parameters, such as

.- ——- ________References and illustrations at end of paper.

2 P=ROLEUM RESERVOIR SIMULATION IN A VIRTUAL ENVIRONMENT SPE 29118

well locations without having to resort to countinggrid blocks. Incorporating the simulator into thevisualization environment means input parameterscan be changed and the simulation restarted withouthaving to exit the visualization environment. Thusour VR/visualization system increases the reservoirengineers interaction with the simulator anddecreases the amount of human and computer timelost due to errors in or pooriy formulated input.

The following sections describe the basic concepts ofvirtual reality and visualization incorporated in thiswork, the reservoir simulator used, and how theresults of the simulation have been visualized. Weconclude with an application of the VR/visualizationsystem to waterflooding in an oil reservoir.

We initially simulate water flooding in aheterogeneous oil reservoir. Visualization of the 3-Doutput of the simulator shows a region of persistentlyhigh oil saturation. Using the VR input device, we areable to reposition one of the production wells and thusto quickly and easily evaluate an alternative drillingstrategy.

VIRTUAL REALITY AND VISUALIZATION

In spite of a constant media bombardment that wouldhave us believe otherwisel, the essence of virtualreality is not “goggles and gloves.” Implementationsof VR encompass a wide gamut of input and outputdevices. Rather than the fundamental ideas that drivethe technology, the devices themselves have capturedthe attention of the media.

The basic premise of VR, visualization, and how thetwo fit together, can best be understood by focusingon the basic concept of the architecture of a typical“graphics pipeline.” In the graphics pipeline, a set ofgeometric objects, such as points, lines, triangles, issubmitted to a “renderer.” The renderer thenprocesses each of these objects, producing an image.The image is presented to the user on a display device.The user is able to interact with the renderer inorder to specify display parameters such as viewer...posmon and orientation. in contrast, visualization isthe process whereby the object database isconstructed using algorithms in which abstract datais represented geometrically.

The goal of VR is to allow the user to specify, in anintuitive and easy-to-use manner, three or higher

dimensional parameters to the visualization process.This idea has been in the graphics community since atleast the early 1960’s2 and more recently hasbecome popular as display device technology hasrapidly advanced. The result, however, is that mostpeople think of VR as being synonymous with a head-mounted display (HMD) and a data glove. Clearly,many different input devices may be employed by theuser to specify parameters to the renderer, to avisualization algorithm, or to a numerical simulator.These include keyboard, mouse, button panel, and dialbox. None of these devices lend themselves to“natural” specification of three or higherdimensional input and thus are cumbersome to usefor that purpose. Many low-cost VR input devicesthat may be used for this purpose are available.There is a noticeable absence, however, of widespreaduse of these devices in the field of scientificcomputing.

The manner in which graphical images are presentedto the user is the subject of VR output devices.Clinical studies3 have shown that currently availableHMD devices result in impaired vision and Iimbicsystem dysfunction even after short periods of use.Further, the best HMD devices available (militarygrade), cost upwards of $100,000 and still don’tdeliver the spatial resolution required by the mostordinary of visualization tasks. These advanceddisplay devices leave much to be desired both interms of practical use and affordability. Otherergonomic issues of concern pertain to integration ofthese devices into a person’s workspace. “Suitingup,” that is, donning goggles and gloves, takes timeand interferes with the work environment in whichtelephones must be answered, fire alarms go off, etc.

The premise of VR, the ability to interact with acomputing process in an a minimally-unobstructedfashion, is consistent with advances in recent yearsof more conventional Graphical User Interfaces andhas the potential to change scientific computing in thesame way that the “point and click” user interfacebrought the power of personal computers to thepublic.

COMBINING VIRTUAL REALITY ANDVISUALIZATION

Our prototype system consists of a few discretecomponents: the visualization environment, thenumerical simulator, and the VR input device. These

234

.

SPE 29118 J.S. JACOBSEN, E.W. B=HEL, A. DAITA GUPTA, P. J. HOLIAND 3

components have been integrated into a system thatmay be used to study the effects of well placement in asimulation of a water or chemical flood in aheterogeneous oil reservoir.

The visualization environment consists of a collectionof separately compiled and linked software modules.The organization of these modules into a visualizationprogram is facilitated using visual programming. Asimple point-and-click interface is used to arrangethe modules in a workspace (Fig. 1) as well as toroute the flow of data between the modules. Ascheduler computes the correct module executionorder from a directed graph in which the nodes aremodules and the edges are data flow paths.

Most important to our work is the fact that thissystem is extensible, that is, users can add their ownsoftware modules to this environment. It is thisflexibility that has made it possible for us toincorporate both the simulator and the use of the VRinputloutput device into the visual environmem.

-—. . —---in

addition, the visualization environment that we areusing supports the execution of modules on remotecomputers. The simulation may be run on a computeserver, such as a massively parallel computer, withthe visualization and user interaction beingperformed on a graphics workstation. (Other suchmodular visualization environments also are

available4’5.) The simulator was ported to run inthis environment for the purposes of interactivecontrol and to provide the flexibility thisenvironment offers in terms of visualization tools.

In its original implementation, the simulator was abatch-oriented code. The user would supply inputdecks, submit the job, and look at the results offlineat a later time, Detection of errors in the simulationinput did not occur until after the simulation had runto completion, using up valuable compute cycles, anddelaying interpretation of results for days or evenweeks. In the current environment, the user ispresented with immediate visual feedback depictingthe input and the results of the simulation, thusshortening the time between parameter specificationand studying the effects of those parameters on thesimulation. In addition, the placement of the wells inthe simulation is under control of the VR inputdevice. Use VR input devices to specify visualizationparameters is not new6. The use of VR input tocontrol a simulation is not only a natural extension of?~~~ nre~~nl IC“.. wn rk b~t g~~a?!y inc~~~~es the

effectiveness of ~h; “&e of the scientist’s intuition by

minimizing the barriers between scientist andsimulation, allowing the scientist to focus more onthe science and less on technology.

DESCRIPTION OF THE NUMERICALSIMULATOR

The numerical simulator incorporated into theVR/visualization system is an adaptation of UTCHEM,a multicomponent, multiphase chemical floodcompositional simulator developed at the Universityof Texas and widely used in the oil industry’) 8, 9.Although the primary focus of the application in thispaper is waterflooding in heterogeneous permeablemedia, the simulator is capable of modeling a varietyof processes including micellar-polymer flooding,caustic flooding, cosurfactant-enhanced alkalineflooding, and profile control processes involving in-situ reactions with polymers and crosslinkingcations. The simulator allows the use of multipletracers in any of the above processes anti forestimating residual oil saturation for eitherinterwell or single well backflow tracer tests.

Material conservation equations are solved for up tonineteen components that can form up to threeseparate phases: aqueous, oleic, and microemulsion.For chemical flooding, the phase environment isdetermined by an effective salinity parameter thattakes into account the presence of divalent cations andcosurfactants. Major physical phenomena modeled bythe simulator are: dispersion, dilution effects,adsorption, interracial tension, relativepermeability, capillary trapping, capillarypressure, compositional phase density and viscosity,cation exchange, alcohol partitioning, and variouspolymer properties such as non-Newtonian rheology,permeability reduction and inaccessible porevolumes. Dissolution/precipitation reactions aremodeled assuming that the aqueous and solid(mineral) phases are in thermodynamic equilibrium.

The basic solution strategy used in the simulator isanalogous to the IMPES (implicit pressure, explicitsaturation) method. A pressure equation, derivedbased on the overall mass-continuity and Darcy’slaw, first is solved implicitly to obtain phasepressures and phase velocities. The conservationequations then are solved explicitly for totalconcentrations of components. Finally, phasesaturations and concentrations are obtained usingflash calculations based on the pseudoquaternary

235

.

P~ROLEUM RESERVOIR SIMULATION IN A VIRTUAL ENVIRONMENT SPE 2911

representation of the phase behavior. A third-ordertotal variational diminishing (TVD) scheme is used tominimize numerical dispersion during solution of theconservation equations.

VISUALIZATION OF SIMULATION RESULTS

In the prototype VR/visualization system, the usersupplies the names of the input files for thenumerical simulator via a graphics user interface.Once the input files have been specified, the displayshows a wireline grid of the numerical grid used bythe simulator, the location of the production andinjections wells, and an isosurface of the magnitudeof the permeability distribution (Fig. 2).Conventional post syrnbois are used to distinguish---

between production and injection wells. At thispoint, prior to running the simulator, the user maychange the value of the permeability isosurfacedisplayed, the frequency with which the output of thesimulator is displayed, and the locations of the wells.Dial boxes are used to change the first twoparameters. To change the location of a well, the userselects the well to move using the mouse and movesthe well to a new location using the VR input device.T1.. ~~awd;~m+~~ nf ?~e ~~pneitinnnd ~~!! ~~~I IIG rm%v euulull IaLGQ ul ““n. mw, ,“”

automatically written to a file accessible to thesimulator.

When the simulator begins to run, the value of the oilsaturation is indicated by the color of small cubespositioned at the center of each grid point. Each cubeindicates not the point value of the oil saturation atthe grid point, but rather the oil saturation (obtainedby interpolation) of the region occupied by the cube.The range of values of oil saturation is indicated by acolor legend in the lower left comer.

The water flood is shown as an isosurface of waterconcentration. The isosurface is translucent,allowing the user to look “through it,” A wirelinemesh is superimposed on the isosurface in order toenhance its three-dimensionality. Fig. 3 shows aclose-up view of the region near the injection well atan early time. This image is obtained by using thespaceball to navigate close to the injection well andprovides an example of how the user can get a close-up look at early time data.

While the simulation is running the user has controlover many aspects of the display, thus enabling theuser to interact with the output of the simulator andto investigate different aspects of the output,

In addition to being able to rotate and/or translate timage, the user may change

“ the size of the cubes that indicate the value of toil saturation and/or select a subset of the cubei.e., display every other cube, every fifth cubetc.,

● the value of the isosurface of the watconcentration shown,

● the value of the isosurface of the permeabilidistribution,

● the value of the isosurface of the oil saturation,● the location of 2-D slices through the

saturation,s (not shown in Figs 2 and 3).

ThA . .-A. -I-A -au i ma tha \/R inmA tiauifis to n~@~I I Ic U==l al=u 11lay us= W[u . I , II p.”. ““. s”” .

through the reservoir. This is equivalent to beiable to zoom in, rotate, and translate. Each of theseparate steps in the visualization system (withothe added VR interface) would require multiple clicon different mouse buttons, together with pressinthe shift key (for zooming in). In contrast, tspaceball provides an intuitive way to move throuthe reservoir without having to remember “whibutton does what.” The effect of clicking severdifferent mouse buttons in sequence canaccomplished with a single smooth movement of tVR input device.

APPLICATION OF VR/VISUALIZATION SYSTEM

We illustrate here the application of thVR/visualization system to convert batch operationthe reservoir simulator to an interactive, visuenvironment for investigating reservoircharacteristics and processes. The example casewaterflooding in a five-spot pattern in the presencof spatially-correlated, permeability heterogeneitiegenerated using the Turning Bands Method wspecified variance and correlation lengths. ThVR/visualization system was used to visualize tperformance of waterflooding and to dynamicalalter the position of one of the production wells prto restarting the simulation.

Generation of the PermeabilityDistribution

The three-dimensional, heterogeneous permeabilidistribution used in this study was generated by tturning bands method (TBM)l 0“ The TBM is a meth

236

.

SPE 29118 J.S. JACOBSEN, E.W. BETHEL, A. DAITA GUPTA, P. J. HOLIAND

for simulating spatially correlated random fields,either two or three dimensional. In this work, weuse an implementation of the TBM developed byA. Tompsonl 1, which uses an exponential-decaycovariance funciton. Because the method convertswhat would be a calculation in two or threedimensions to a series of one-dimensional problems,it is computationally efficient12.

THE TBM generates random fields from a normaldistribution with zero mean and a specified variance.lm tha nraeant annlifidinn thn ranrlnm fkkf rrmwiratd,,, ,,, e p“””, ,. “p~,s”-..”. ., ., ,“ .-. .“.-. . . ----- ~-. .. . . . . .

is converted to a Iognormal field with a geometricmean of 0.5 and a log-variance of 0.35. Theresulting permeabilities are scaled to give values inunits of millidarcies. The correlation lengths for thehorizontal permeabilities are 40% of the size of thereservoir in the horizontal directions and 20% of thesize of the reservoir in the vertical direction. Thecorrelation lengths for the vertical permeabilities-.- onol- -4 bh~ mi~m -f +ha rac. am.., niw ~fi ~~t~ t~,~aI= cv /0 VI LII= =ILS7 WI LII= J=QUI VUIi

horizontal and vertical directions. In Fig. 4,permeability magnitudes in the range of 600 to 700md are shown in white.

Simulation Input Parameters

The five-spot pattern used for simulation studies isshown in Fig. 2. it has a centrai injection weiisurrounded by four production wells. The injectionrate is set at 10’% pore volumes/year (33.2ft3/day). Pattern balance is maintained by settingthe production rate for each well at one-quarter of

the total injection rate (8.3 ft3/day). Each well isperforated along its entire length. The distancebetween the injection well and each production wellis 52.5 ft. The reservoir is 105 ft by 105 ft and 55ft in depth. The reservoir has been divided into 21grid blocks in each horizontal direction and 11 gridblocks in the vertical direction for a constant gridspacing of 5 ft. The reservoir has a constant porosityof 0.2, an initial oil saturation of 70Y0, and an initialwater saturation of 300A.

The numerical simulator requires many other inputparameters describing chemical and physicalprocesses, etc. For the purpose of the applicationdescribed here, only parameters describing spatialreservoir properties (e. g., porosity andpermeability) and well locations andinjection/production rates are given. A quadraticfunctional form for relative permeabilities has beenused for both water and oil.

Description of Simulation Results

Fig.3 shows a close-up view of the region near theinjection well at an early time (26 days). The effectof the heterogeneity in the permeability distributionis evident as shown by the asymmetry in theisosurface of water. Fig. 5 shows a later (49 days)view of the reservoir from above it, looking downinto. Because the cubes that indicate the value of theoil saturation do not touch, the user is able to lookthrough the reservoir and see the cone of watersurrounding the injection well. Once again, the effectof the heterogeneous permeability distribution isevident from the asymmetry of the water flood.

A user sitting at the color workstation on which theVR/visualization system is running would be able torotate the display in order to get a better view of theoutput results and to look for areas in which lowpermeability zones have resulted in higher oilconcentrations than in surrounding areas. Fig. 6shows such a view at a simulation time of 75 days.The dark isosurface spreading across the reservoir isan isosurface of relatively high (0.69) oilsaturation. Between the rightmost (foreground) welland the central injection well, there is a pocket of oilextending over half of the depth of the reservoir.This pocket is evident at earlier times and persistsfe~ lnnmar timae (nnt ehnwn)

Iul l~wm Lmlllwu ,, B“. “m ,“.. , 8,.

In order to demonstrate the value of theVR/visualization system for identifying candidate.--: --- f-. :-<:114 ..:ll:--- &k-regions Ior Irr III I urllllrrg,

-:- ..1-4:-.. ..#-elIIG SIlllulullull w a=

gf)pp~~ , ~n~ using the VR input device, the rightmost

production well was moved closer to the oil pocket asshown in Fig. 7. In addition, a change in theperforation of the injection well was made so that thewell would not be perforated near the top or bottom,but only for the middle 50 ft.

After making these changes, the simulation wasrestarted from the visualization environment. Fig. 8(26 days) shows a nearly identical view of theregion near the injection well as shown in Fig. 3.The water isosurface near the injection well is nolonger pinched in the middle as in Fig. 3, but is moresymmetrical with respect to the injection well.Fia. 9 shows a view comparable to Fig. 6. The slight.=. _ _.. -.. – ..r–- –.–repositioning of the production well has had thedesired effect. The pocket of oil so prominent inFig. 6 has been produced.

237

6 PHROLEUM RESERVOIR SIMUIATIC)N IN A VIRTUAL ENVIRONMENT

CONCLUSION

The VR/visualization system described in this paperprovides the following advantages to the reservoirengineer:

● the ~pp~rt~nit~ to view simulation results asthey are generated,

● a rich assortment of visualization tools andtechniques for examining simulation results,

“ the use of a VR input device to specify 3-D inputparameters such as well locations and perforationzones,

● the use of VR/visualization to study fielddevelopment strategies (e. g., infill drilling,perforations, etc.)

Taken together, these advantages enhance thereservoir engineers intuition for formulatingdifferent strategies for solving reservoir problems.!-k or she is ab!e t~ start to see simulation resultssoon after commencing a run, while the thoughtprocess that led to the formulation of the inputscenario is still fresh in his/her mind. Similarly,kq~ using a varie%~ of ~i~ua!~~a~ien !e~hniq~~~ toexamine simulation results at a single time step, thereservoir engineer is able to get a better feel forwhat is happening in the reservoir as the simulationproceeds. Seeing the simulation results while thesimulator is running sparks the reservoir engineer’scuriosity about what would be the effect of changingone or another parameter, so that the thread ofreasoning is not lost between the start and finish ofthe simulation. Not only does the reservoirengineer’s attention stay focused on the strategy thathe/she has formulated to solve the problem at hand,but less time is lost because input errors or poorsolution strategies could be detected early in asimulation.

In the simple case used to demonstrate theapplicability of our VR/visualization system, theneed for repositioning one of the production wells wasindicated by the persistence of a region of high oilsaturation. Where to reposition the well also wasfairly clear. Moving the well 30 ft. gave the desiredeffect of removing the oil pocket and convincinglydemonstrates the practical advantage of using aVR/visualization system, such as ours, to understandsimulation results and to formulate new solutionstrategies.

1.

2.

3.

4.

5.

6.

7.

8.

9.

Negromonte,Pleonasm?”

N.: “Virtual Reality:Wired (Dec. 1993)

Bishop, G., and Fuchs, H.

SPE 2911

Oxymoron1(6).

(co-chairs)“Research Directions in” VirtuaEnvironments,” Report of an NSF InvitationWorkshop, University of North CaroiinaChapel Hill, March 1992. “ResearcDirections in Virtual Environments,” Reportan NSF Invitational Workshop, UniversityNorth Carolina, Chapel Hill, March 199Computer Graphics (Aug. 1992) 26(3).

Warm, J., Rushton, S., Monwilliams, MHawkes, R., Smyth, M.: “What’s Wrong wyour Head Mounted Display,” CyberEdgeJournal, (Sept./Ott. 1993).

Upson, C., Faulhaber, T., Kamins, D., LaidlaD., Schlegel, D., Vroom, J., Gurwitz, R., VDam, A.: “The Application VisualizationSystem: A Computational Environment fScientific Visualization,” /EEE ComputeGraphics and Applications (July 1989), 9(4)

Rasure, J., Williams, C., Argiro, D., and SaueT.: “A Visual Language and SoftwarDevelopment Environment for ImagProcess ing,” International Journal of ImaginSystems Technology (1990) 2, 183-199.

Bryson, S., and Levit, C.: “The VirtuWindtunnel: An Environment for thExploration of Three-Dimensional UnsteadFlows,” Proceedings of IEEE Visualizatio(1991), 17-24.

Pope, G. A. and Nelson, R. C.: “A ChemicFlooding Compositional Simulator,” Sot. PEng. J (1978),18(5), 339-354.

Datta-Gupta, A., Pope, G. A., Sephernoori,and Thrasher, R. L.: “A Symmetric PositiDefinite Formulation of a Three-dimensionaMicellar/Polymer Simulator,” SPE Res. Eng(1986), 1(6), 622-632.

Saad. N., Po~e. G. A. and Se~ehrnoori,“Simulation of Big Muddy SurfactantSPE Res. Engg. (1989), 4(1), 24-34.

Pilo

238

SPE 29118 J.S. JACOBSEN, E.W. BETHEL, A. DATTA GUPTA, P. J. HOLLAND

10.

11.

12.

Mantoglou, A. and Wilson, J. L.: “The TurningBands Method for Simulation of Random FieldsUsing Line Generation by a Spectral Method,”Water Resour. Res. (1982), 18(5), 1379-

1394.

Tompson, A. F. B.: “User’s Guide for TurningBands Codes,” Lawrence Livermore NationalLaboratory. Unpublished.

Tompson, A. F. B., Ababou, R., Gelhar, L. W.:“implementation of the Three-DimensionalTurning Bands Random Field Generator,” Water,-----Flesour. Hes. (l ww~, 25(1 0), 2227-2243.

ACKNOWLEDGMENTS

T1-- -... L--- . ..-..1A 1:1. -

I ne aulncrrs wrJuIa llKe to tha~k the Ci3ntEW fCIT

Petroleum and Geosystems Engineering at theUniversity of Texas at Austin for allowing the use ofthe UTCHEM simulator in this study, Jane Long forreviewing this paper and her encouragement topursue this work, and Alexandra Yurkovsky for thepreparation of this paper. This work was supportedby the Director of the Scientific Computing Staff,Office of Energy Research, of the US DOE underContract No. DE-AC03-76SFOO098.

239

240

FIGURE 2 Plaoement of injetilon well (center well) andsurrounding productionwells in numerical grid.Isosurface of permeability magnitude (670 md) isalso shown.

FIGURE 3 Close-up-view of the regionof the reservoirnearthe injectionwell at a simulationtime of 26 days.The value of the isoconcentrationof water is 0.6.

241

FIGURE 6 Isosurface of high oil saturation (value of 0.69)shown by dark horizontal surface and verticalfinger at a simulation time of 75 days.

244

FIGURE 7 Repositioning of production well, prior to restartingthe simulation, to counter vertical finger of high oil

The ~~rn~lation output is identical tosaturation. . ----that of Figure 6.

245

.,

FIGURE 9 Effect of repositioning production well on region ofhigh oil saturation. Simulation is shown at a timecomparable to that of Figure 7.

247