• No results found

A compressible viscous three-phase model for porous media flow based on the theory of mixtures

N/A
N/A
Protected

Academic year: 2022

Share "A compressible viscous three-phase model for porous media flow based on the theory of mixtures"

Copied!
19
0
0

Laster.... (Se fulltekst nå)

Fulltekst

(1)

ContentslistsavailableatScienceDirect

Advances in Water Resources

journalhomepage:www.elsevier.com/locate/advwatres

A compressible viscous three-phase model for porous media flow based on the theory of mixtures

Yangyang Qiao

, Steinar Evje

University of Stavanger, Stavanger 4036, Norway

a r t i c le i n f o

Keywords:

Multiphase flow in porous media Three-phase flow

Viscous coupling Mixture theory Compressible model Water alternating gas (WAG) Waterflooding

a b s t r a ct

Inthispaperwefocusonageneralmodeltodescribecompressibleandimmisciblethree-phaseflowinporous media.TheunderlyingideaistoreplaceDarcy’slawbymoregeneralmomentumbalanceequations.Inparticular, wewanttoaccountforviscouscouplingeffectsbyintroducingfluid-fluidinteractionterms.In[Qiao,etal.

(2018)AdvWaterResour112:170–188]awater-oilmodelbasedonthetheoryofmixtureswasexplored.It wasdemonstratedhowtheinclusionofviscouscouplingeffectscouldallowthemodeltobettercaptureflow regimeswhichinvolveacombinationofco-currentandcounter-currentflow.Inthisworkweextendthemodel indifferentaspects:(i)accountforthreephases(water,oil,gas)insteadoftwo;(ii)dealwithboththecompressible andincompressiblecase;(iii)includeviscoustermsthatrepresentfrictionalforceswithinthefluid(Brinkman type).Amainobjectiveofthisworkistoexplorethisthree-phasemodel,whichappearstobemorerealisticthan standardformulation,inthecontextofpetroleumrelatedapplications.Wefirstprovidedevelopmentofstable numericalschemesinaone-dimensionalsettingwhichcanbeusedtoexplorethegeneralizedwater-oil-gasmodel, bothforthecompressibleandincompressiblecase.Then,severalnumericalexampleswithwaterfloodingina gasreservoirandwateralternatinggas(WAG)experimentsinanoilreservoirareinvestigated.Differencesand similaritiesbetweenthecompressibleandincompressiblemodelarehighlighted,andthefluid-fluidinteraction effectisillustratedbycomparisonofresultsfromthegeneralizedmodelandaconventionalmodelformulation.

1. Introduction Generally

Theprocessesofmultiphase flowin porousmediaoccurinmany subsurfacesystemsandhavefoundmanyapplicationsofpracticalin- terest,suchashydrology,petroleum engineering,geothermalenergy developmentandcarbonstorage(Bakhshianetal.,2019;Bakhshianand Hosseini,2019;Wu,2016).Theimmisciblethree-phaseflowisalways encounteredinwaterfloodingforoilreservoirswithgascap,inimmis- cibleCO2storageindepletedoilandgasreservoirs,andsteamfloods andwater-alternating-gas(WAG)processes(BentsenandTrivedi,2012;

Juanes,2008).Darcy’slawwasoriginallydevelopedforsingle-phase flow(Darcy,1856).Conventionalmodelingofmultiphaseflowisnor- mallybasedonDarcy’sextendedlaw(Rose,2000)byincorporationof relativepermeabilities(Muskatetal.,1937).However,recentexperi- mentalobservationsindicatethattheflowmode(co-currentorcounter- current)canhaveastrongimpactontheflowingphasemobilities.That istosay,therelativepermeabilitiesarenotonlyfunctionofsaturation butarealsorelatedtotheeffectofhowthefluidsflowrelativelytoeach other(BentsenandManai,1992;BourbiauxandKalaydjian,1990).

Correspondingauthor.

E-mailaddress:steinar.evje@uis.no(Y.Qiao).

Viscouscoupling

Viscouscoupling(i.e.,fluid-fluidinteraction)wasfirstlymentioned byYuster(1951)byusingtheoreticalanalysistoderivethatrelativeper- meabilityisafunctionofbothsaturationandviscosityratio.Inaddition, capillarynumberwasalsoproposedtobeafactoraffectingrelativeper- meabilities(Ehrlich, 1993;AvraamandPayatakes,1995).Ingeneral, momentumtransferduetodifferencesininterstitialvelocitiesinduces accelerationofthesloweranddecelerationofthefastermovingfluid whenthefluidsaremovingco-currently.Decelerationofbothfluidve- locitieswilloccuriftheyaremovingcounter-currently(Ayodele,2006;

BentsenandManai,1993;DullienandDong,1996;Lietal.,2004).

Inordertoextendthesingle-phaseDarcy’slawtomultiphaseflow, delaCruzandSpanos(1983)derivedtheoreticallyDarcy’sempirical extendedlawbyapplyingthemethodofvolumeaveragingtoStokes equation.InKalaydjian(1987,1990),Kalaydjiandevelopedflowequa- tionsusingtheconceptsofirreversiblethermodynamics(Katchalskyand Curran, 1975) from amacroscopic understandingof two-phase flow in porous media.Inaddition,some researcherstried togaininsight into howtwo immisciblephases flow through a porousmedium by using simpleanalogous modelssuch as tubular flow (Yuster, 1951;

Bacrietal.,1990).InLangaasandPapatzacos(2001)LangaasandPa-

https://doi.org/10.1016/j.advwatres.2020.103599

Received15May2019;Receivedinrevisedform21April2020;Accepted22April2020 Availableonline28April2020

0309-1708/© 2020TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense.(http://creativecommons.org/licenses/by/4.0/)

(2)

patzacosusedtheLatticeBoltzmann(LB)approachtoinvestigateef- fectsof viscouscouplingandconcludedthatcounter-current relative permeabilitiescausedpartlybyviscouscouplingarealwayslessthan thecorrespondingco-currentcurvesunderdifferentlevelsofcapillary forces.Usingthesamemethod,Lietal.(2005)showedthattheirmodel wasabletocapturemainexperimentaleffectscausedbyviscouscou- pling.Theyalsomentionedthattheinterfacialareabetweenthefluids isakeyvariableforrelativepermeabilityfunctionsfortwoimmisci- blefluidsflowinporousmedia.Ageneralizedmodelwasdevelopedin Qiaoetal.(2018)fortwo-phaseflowwithviscouscouplingeffect.Nu- mericalinvestigationsshowedabetteragreementwiththeexperimental tests(BourbiauxandKalaydjian,1990)comparedtotheconventional modeling.TheauthorsinBentsenandTrivedi(2012)constructedmod- ifiedtransportequationsforbothco-currentandcounter-currentthree- phaseflowthroughaverticalincompressiblemodelbasedonpartition concepts.Theirequationsareusedtoestimatetheamountofmodeler- rorbecauseofafailuretoaccountfortheeffectofinterfacialcoupling whichhastwotypes:viscouscouplingandcapillarycoupling.Moreover, SherafatiandJessen(2017)investigatedtheeffectofmobilitychanges duetoflow reversalsfrom co-currenttocounter-currentflow onthe displacementofWAGinjectionprocesses.

Complexmultiphaseflowinporousmediaanduseofthetheoryofmixtures

Motivated by petroleum related applications variousattempts to solvethethree-phaseporousmediaflowmodelhavebeenreporteddur- ingthepastdecade(FallsandSchulte,1992;GuzmánandFayers,1997a;

1997b;JuanesandPatzek,2004).Aninterestinginvestigationwascar- riedoutinLieandJuanes(2005)whereafront-trackingalgorithmwas proposedforconstructing veryaccuratesolutionstoone-dimensional problems(forexampleWAGtesttherein).Thiswasexploredinthecon- textof streamline simulationwhich decouples thethree-dimensional problemintoasetofone-dimensionalproblemsalongstreamlines.This workislimitedtothree-phaseimmiscible,incompressibleflowandalso gravityandcapillaritywereignored.Differentnumericalmethodshave beenimplementedtosimulatethree-phaseflowinporousmedia.Afinite volumemethodwasusedinLeeetal.(2008)forsolvingcompressible, immiscibleflowwithgravityinheterogeneousformationsbyusingthe blackoilformulation.Ahybrid-upwindingschemeforphasefluxwas proposed inLee andEfendiev (2016)fora finitedifferenceapproxi- mationtosolvethethreephasetransportequationsinthepresenceof viscousandbuoyancyforces.Afiniteelementmethodwasappliedto simulatefluidinjectionandimbibitionprocessesinadeformableporous media(Gajoetal.,2017).Moreover,(DongandRivière,2016)applied asemi-implicitmethodwithdiscontinuousGalerkin(DG)discretization tosolvetheincompressiblethree-phaseflowintwodimensions.Addi- tionalphysicaleffectsarealsodiscussedandexploredforthree-phase porousmediaflow,suchashysteresiseffectsofrelativepermeabilities (Ranaee etal., 2019) andellipticregions(JuanesandPatzek, 2004;

Juanes,2008;LeeandEfendiev,2016).InJuanes(2008)Juanespre- sentedanonequilibriummodelofincompressiblethree-phaseflowin porousmedia.Thenonequilibriumeffectsbyintroducingapairofef- fectivewaterandgassaturationsintotheformulationshavetheability tosmearsaturationfrontsfromnumericalsimulations.

Thetheory ofmixturesoffersageneralframeworkfordeveloping modelsforcomplexmultiphaseflowsystems(Rajagopal,2007).More lately,biomedicalapplicationshavebeenadriverforthedevelopment ofvariousmodelsrelyingonthisapproach.Forexample,thestudyhow cancercellsareabletobreakloosefromaprimarytumorinvolvesasolid matrix(theso-calledextracellularmatrix),differenttypeofcells(can- cercells,stromalcells,immunecells),andinterstitialfluid(Evje,2017;

EvjeandWaldeland,2019).Arecentexample ofthis isdescribed in WaldelandandEvje(2018b);Urdaletal.(2019)where,respectively,a cell-fluidtwo-phasemodelandacell-fibroblast-fluidthree-phasemodel aredevelopedtoshedlightontheexperimentallyobservedtumorcell behaviorreported in Shieh etal. (2011). Themodel thatis derived

relies on replacingDarcy’slaw bymore generalmomentumbalance equationswhichincorporateboththecell-matrixresistanceforceand the cell-fibroblastinteraction. Thelatter is understood asa ”viscous coupling” effectcausedbyamechanicalcouplingthatcan occurbe- tweentumorcellsandfibroblastsandhasbeenreportedinexperimental studies(Labernadie,2017).Anotherexamplehowgeneralizedmomen- tumequationscanbeusedtocapturenon-standardmultiphasebehav- iorinthecontextofaggressive tumorcellsis exploredinWaldeland andEvje(2018a).InPolachecketal.(2011)twocompetingmigration mechanismswereobserved,oneintheupstreamdirectionandanother inthedownstreamdirection.Theuseofgeneralizedmomentumequa- tionsallowedustoaccountforboththisfluid-stressgeneratedupstream migrationandachemotactivemigrationinthedirectionofincreasing concentrationofchemicalconcentrations(WaldelandandEvje,2018a).

Theaimofthiswork

Theobjective ofthis paperistoinvestigate amixturetheory ap- proach tosimulatethreeimmisciblefluidsflowingina1Dreservoir.

Weshallconsiderboththecasewithcompressibleandincompressible fluids.Themodelwhichisintroducedisquitegeneralsinceitcanau- tomaticallycaptureflowthatinvolvesacombinationofco-currentand counter-currentflow.Thecurrentworkrepresentsextensionofprevious workintwoways:

Extendtheincompressibletwo-phasemodelthatwasexploredin Qiaoetal.(2018);Andersenetal.(2019)toincludethreephases.

Extend the compressible two-phase model studied in Qiaoetal.(2019a)toincludethreephases.

Inaddition,themodelswestudyinthecurrentworkaremoregen- eralthanthosestudiedinQiaoetal.(2018);Andersenetal.(2019)since we considerStokeslike momentumequationswhich involveviscous termsthataccountforinternalfrictionduetoviscosity.Inparticular, appropriatenumericalschemesareintroducedtoinvestigatecompress- ibleandincompressiblethree-phaseflowscenariosthataremotivated byinjection-productionflowscenarios.

Mainobservationsfrom ournumericalexperiments withtwoand three-phaseflowscenarioswheretheflowdynamicsaregeneratedby injectionof waterorgasin thecenterofthedomainandproduction offluidsattheleftandrightboundaryare:(i)Thesimulationcasesin- volvecompetitionbetweenpressuredrivenco-currentflowandcounter- currentgravitydrivenflow;(ii)Boththeincompressibleandcompress- iblediscreteversionofthemodelappeartohavegoodstabilityprop- erties.Thenumericalexperimentsindicatethatthenumericalschemes canbeusefulasatooltodeepentheinsightintotherelationbetweenthe incompressibleandcompressibleversionofthemodel.Themodeland itsdiscreteapproximatecounterpartsappeartobeagoodstartingpoint forextendingtomorecomplexflowsystems,asmentionedabove,that involvecompetitionbetweendifferentdistinct,non-standardtransport mechanisms.

The rest of this paper is organized as follows. In Section 2 we briefly describe the mixture flux approach in a three-phase setting.

In Section 3we summarize the generalizedthree-phase porous me- dia model, botha compressibleandanincompressible versionofit.

Section 4is devotedtonumerical simulationstodemonstratethree- phase dynamics andverifybasic featuresof the numericalschemes.

Thedetailsofthecompressibleandincompressibleschemearegiven inAppendixA–AppendixD.

2. Mixturetheoryframework

2.1. ConventionalmodelbasedonDarcy’slaw

We firstly describe the traditional formulation of incompressible multiphaseflowmodelwithoutsourceterms.Transportequationsfor

(3)

incompressibleandimmisciblephasesoil(o),water(𝑤)andgas(g)in porousmediaarenormallygivenby:

𝜕𝑡(𝜙𝑠𝑖)+∇⋅𝐔𝑖=𝑄𝑖, (2.1)

𝐔𝑖=𝜙𝑠𝑖𝐮𝑖, (𝑖=𝑤,𝑜,𝑔), (2.2)

where𝜙isporosity,si isphasesaturation,Qiisthesourceterm,and UianduiaretheDarcyvelocityandinterstitialvelocityofeachphase 𝑖=𝑜,𝑤,𝑔,respectively.Forsimplicitytheirreducible(immobile)phase saturation(sir)isnotconsideredintheequationsbyassumingitisequal to0.Hence,thenormalizedphasesaturation(= 𝑠𝑖𝑠𝑖𝑟

1−𝑠𝑤𝑟𝑠𝑜𝑟𝑠𝑔𝑟)equalsthe phasesaturationvaluesi. Thetraditionalmacroscopicformulationof Darcy’slawthatrelatesthevolumetricfluxofaphasetothepressure gradientofthatphaseisgivenby:

𝐔𝑖=−𝐾𝑘𝑟𝑖

𝜇𝑖 (∇𝑝𝑖𝜌𝑖𝐠), (𝑖=𝑤,𝑜,𝑔), (2.3) whereKistheabsolutepermeabilityofporousmedia,piisphasepres- sure,gistheaccelerationofgravityandkri,𝜌iand𝜇iarephaserelative permeability,densityandviscosity,respectively.

2.2. Ageneralizedmultiphaseflowmodelbasedonmixturetheory

Forourinvestigations,themassbalanceequationswithsourceterms inthecaseofcompressiblewater-oil-gastransportcanbegivenby:

(𝜙𝑛𝑤)𝑡+∇⋅(𝜙𝑛𝑤𝑢𝑤)=−𝑛𝑤𝑄𝑝+𝜌𝑤𝑄𝐼𝑤, 𝑛𝑤=𝑠𝑤𝜌𝑤 (𝜙𝑛𝑜)𝑡+∇⋅(𝜙𝑛𝑜𝑢𝑜)=−𝑛𝑜𝑄𝑝, 𝑛𝑜=𝑠𝑜𝜌𝑜

(𝜙𝑛𝑔)𝑡+∇⋅(𝜙𝑛𝑔𝑢𝑔)=−𝑛𝑔𝑄𝑝+𝜌𝑔𝑄𝐼𝑔, 𝑛𝑔=𝑠𝑔𝜌𝑔 (2.4) whereui, (𝑖=𝑤,𝑜,𝑔) representstheinterstitialvelocityof phasei in theporousmedia.Inaddition,Qpistheproductionrateand𝑄𝐼𝑤,QIg representtheinjectionrateofwaterandgas,respectively.

Thestartingpointfordevelopingour modelthatcanaccount for moredetailedphysicalmechanismsforwater-oil-gasporousmediaflow thanconventionalmodeling,isthetheoryofmixtures.Thisisatheory basedonbalancelawsandconservationprinciples,whichiswellknown incontinuummechanics(Bowen,1976;RajagopalandTao,1995;Byrne andPreziosi,2003; AmbrosiandPreziosi,2002; PreziosiandFarina, 2002),andhasbeenwidelyappliedtothebiologicaltumor-growthsys- temswhichcanbecharacterizedasamixtureofinteractingcontinua.

Neglecting inertialeffects (accelerationeffects),as isusual when dealingwithcreepingflowinporousmaterials,themechanicalstress balanceisgivenbyAmbrosiandPreziosi(2002):

0=∇⋅(𝑠𝑖𝜎𝑖)+𝑚𝑖+𝐺𝑖, (𝑖=𝑤,𝑜,𝑔), (2.5) where𝜎ireferstotheCauchystresstensor,mirepresentstheinteraction forcesexertedontheconstituentsbyotherconstituentsofthemixture, and𝐺𝑖=𝑠𝑖𝜌𝑖𝑔istheexternalbodyforceduetogravity.Thestandard expressionforthestressterms𝜎i,isgivenby

𝜎𝑖=−𝑝𝑖𝛿+𝜏𝑖, (𝑖=𝑤,𝑜,𝑔), (2.6)

where𝛿istheunitarytensorand 𝜏𝑖=2𝜇𝑖𝑒𝑖, 𝑒𝑖=1

2(∇𝑢𝑖+∇𝑢𝑇𝑖), (𝑖=𝑤,𝑜,𝑔). (2.7) Theviscouspart𝜏i reflectsthatthewater,oilandgasbehaveasavis- cousfluid.Accordingtogeneralprinciplesofthetheoryofmixtures,the interactionforcesmibetweentheconstituentsappearingin(2.5)maybe describedasinPreziosiandFarina(2002);AmbrosiandPreziosi(2002); ByrneandPreziosi(2003):

𝑚𝑜=𝑝𝑜𝑠𝑜+𝐹𝑤𝑜𝐹𝑜𝑔+𝑀𝑜𝑚, 𝑚𝑤=𝑝𝑤𝑠𝑤𝐹𝑤𝑜𝐹𝑤𝑔+𝑀𝑤𝑚,

𝑚𝑔=𝑝𝑔𝑠𝑔+𝐹𝑤𝑔+𝐹𝑜𝑔+𝑀𝑔𝑚, (2.8)

whereFij(𝑖,𝑗=𝑜,𝑤,𝑔),denotestheforce(drag)thattheiphaseexerts on thejphase.Thejphaseexerts anequalandoppositeforce −𝐹𝑖𝑗. Similarly,Mom,𝑀𝑤𝑚andMgmrepresentinteractionforces(dragforces) betweenfluidandporewalls(solidmatrix),respectively,foroil,water andgas.Thetermpisi isrelatedtointerfacialforceexertedbyother phaseson phasei, arisingfrommathematical derivationofaveraged equations(DrewandSegel,1971).Toclosethesystemwemustspecify thedragforce term𝐹𝑤𝑜,𝐹𝑤𝑔,andFog andthestresses 𝜎i (𝑖=𝑜,𝑤,𝑔) andinteractionforcetermsMimbetweenfluid(𝑖=𝑤,𝑜,𝑔)andmatrix.

Dragforce representstheinteractionbetweenonephaseandanother phaseandismodelledasRajagopal(2007);PreziosiandFarina(2002); AmbrosiandPreziosi(2002):

𝐹𝑤𝑜 =̂𝑘𝑤𝑜(𝑢𝑤𝑢𝑜), 𝐹𝑤𝑔 =̂𝑘𝑤𝑔(𝑢𝑤𝑢𝑔),

𝐹𝑜𝑔=̂𝑘𝑜𝑔(𝑢𝑜𝑢𝑔), (2.9)

wherê𝑘𝑖𝑗(𝑖,𝑗=𝑜,𝑤,𝑔),remainstobedetermined.Typically,̂𝑘𝑖𝑗𝑠𝑖𝑠𝑗 toreflectthatthisforcetermwillvanishwhenoneofthephasesvan- ishes.Similarly,theinteractionforcebetweenfluidandporewall(ma- trix,whichisstagnant)isnaturallyexpressedthenas(Rajagopaland Tao,1995; Rajagopal,2007;PreziosiandFarina,2002;Ambrosi and Preziosi,2002):

𝑀𝑖𝑚=−̂𝑘𝑖𝑢𝑖, (𝑖=𝑜,𝑤,𝑔). (2.10)

Thecoefficients ̂𝑘𝑖𝑗 and̂𝑘𝑖(dimensionPa· s/m2),thatcharacterizethe magnitudeofinteractionterms,canbechosensuchthatthemodelre- coverstheclassicalporousmediamodelbasedonDarcy’slaw.Atthe sametimetheapproachusedherewillopenfordevelopmentofreser- voirmodelswheremoredetailedphysicscanbetakenintoaccount.

3. Asummaryofthegeneralthree-fluidmodelforporousmedia flow

3.1. Thecompressiblecase

Weareinterestedinstudyinga1-Dmodelforthreecompressibleim- misciblefluidsmovinginaporousmedia.Aftercombining(2.4)-(2.10) themodeltakesthefollowingform:

(𝜙𝑛𝑤)𝑡+(𝜙𝑛𝑤𝑢𝑤)𝑥=−𝑛𝑤𝑄𝑝+𝜌𝑤𝑄𝐼𝑤, 𝑛𝑤=𝑠𝑤𝜌𝑤, (𝜙𝑛𝑜)𝑡+(𝜙𝑛𝑜𝑢𝑜)𝑥=−𝑛𝑜𝑄𝑝, 𝑛𝑜=𝑠𝑜𝜌𝑜, (𝜙𝑛𝑔)𝑡+(𝜙𝑛𝑔𝑢𝑔)𝑥=−𝑛𝑔𝑄𝑝+𝜌𝑔𝑄𝐼𝑔, 𝑛𝑔=𝑠𝑔𝜌𝑔

𝑠𝑤(𝑃𝑤)𝑥=−̂𝑘𝑤𝑢𝑤̂𝑘𝑤𝑜(𝑢𝑤𝑢𝑜)−̂𝑘𝑤𝑔(𝑢𝑤𝑢𝑔) +𝑛𝑤𝑔+𝜀𝑤(𝑛𝑤𝑢𝑤𝑥)𝑥,

𝑠𝑜(𝑃𝑜)𝑥=−̂𝑘𝑜𝑢𝑜̂𝑘𝑤𝑜(𝑢𝑜𝑢𝑤)−̂𝑘𝑜𝑔(𝑢𝑜𝑢𝑔) +𝑛𝑜𝑔+𝜀𝑜(𝑛𝑜𝑢𝑜𝑥)𝑥,

𝑠𝑔(𝑃𝑔)𝑥=−̂𝑘𝑔𝑢𝑔̂𝑘𝑤𝑔(𝑢𝑔𝑢𝑤)−̂𝑘𝑜𝑔(𝑢𝑔𝑢𝑜) +𝑛𝑔𝑔+𝜀𝑔(𝑛𝑔𝑢𝑔𝑥)𝑥,

Δ𝑃𝑜𝑤(𝑠𝑤)=𝑃𝑜𝑃𝑤, Δ𝑃𝑔𝑜(𝑠𝑔)=𝑃𝑔𝑃𝑜 (3.11) withcapillarypressureΔ𝑃𝑜𝑤definedasthepressuredifferencebetween theoilandwaterandcapillarypressureΔPgo definedasthepressure differencebetweenthegasandoil.Wemaychoosetousethefollowing expressionsforcapillaryforce

Δ𝑃𝑜𝑤=𝑃𝑜𝑃𝑤𝑃𝑜𝑤(𝑠𝑤)=−𝑃𝑐1ln(𝛿1+𝑠𝑤 𝑎1

) and 𝛿1,𝑎1>0, Δ𝑃𝑔𝑜=𝑃𝑔𝑃𝑜𝑃𝑔𝑜(𝑠𝑔)=𝑃𝑐2𝑠𝑎𝑔2 and 𝑎2>0 (3.12) withnon-negativeconstants 𝑃𝑐𝑖 representinginterfacialtension.This allowsustomimiccapillarypressurefunctionsthatprevisouslyhave beenproposedforthree-phasereservoirflow(ChenandEwing,1997;

(4)

OddandDavid,2010).Inaddition,wehavethefundamentalrelation thatthethreephasesfilltheporespace

𝑠𝑜+𝑠𝑤+𝑠𝑔=1. (3.13)

Theabovemodelmustbecombinedwithappropriateclosurerelations for𝜌𝑖=𝜌𝑖(𝑃𝑖).Werepresentthethreephasesbylinearpressure-density relationsoftheform

𝜌𝑤̃𝜌𝑤0=𝑃𝑤

𝐶𝑤, 𝜌𝑜̃𝜌𝑜0= 𝑃𝑜

𝐶𝑜, 𝜌𝑔= 𝑃𝑔

𝐶𝑔 (3.14)

where𝐶𝑤,CoandCgrepresenttheinverseofthecompressibilityofwa- ter,oilandgas,respectively.

WerefertoAppendixBforasemi-discreteapproximationof(3.11)as wellasafullydiscretescheme.

Remark3.1.Wemayalsostudyahigherdimensionalcase(e.g.,2D) wherethe modelconsists of three massbalanceequations for three phases(water,oilandgas)andsixmomentumequations(eachphase hastwodirectionssuchasxandy).Theschemehasbeentestedin2D fortwophasesandshowssimilarpropertiesasin1D.

3.2. Theincompressiblecase

3.2.1. Viscousflow

Wemaylet𝐶𝑤,𝐶𝑜,𝐶𝑔 gotoinfinityin(3.14).Thenweobtainthe incompressibleversionofthemodel(3.11).WerefertoAppendixCfor asemi-discreteaswellasafullydiscreteschemeforthisincompressible case.

3.2.2. Inviscidflow

Moreover,inordertorelatethisincompressibleversiontotheclas- sicalDarcy-basedformulationweignoretheviscositytermsinthemo- mentumequations by setting 𝜀𝑖=0(𝑖=𝑤,𝑜,𝑔)in (3.11)4,5,6. Solving momentumequationswithrespecttointerstitialphasevelocitiesui,the Darcyvelocitiesoffluidphaseareexpressedasfollowsbasedon(2.2): 𝑈𝑤 =𝜙𝑠𝑤𝑢𝑤=−𝜆𝑤𝑤(𝑃𝑤𝑥𝜌𝑤𝑔)−𝜆𝑤𝑜(𝑃𝑜𝑥𝜌𝑜𝑔)−𝜆𝑤𝑔(𝑃𝑔𝑥𝜌𝑔𝑔),

𝑈𝑜 =𝜙𝑠𝑜𝑢𝑜=−𝜆𝑤𝑜(𝑃𝑤𝑥𝜌𝑤𝑔)−𝜆𝑜𝑜(𝑃𝑜𝑥𝜌𝑜𝑔)−𝜆𝑜𝑔(𝑃𝑔𝑥𝜌𝑔𝑔), 𝑈𝑔 =𝜙𝑠𝑔𝑢𝑔=−𝜆𝑤𝑔(𝑃𝑤𝑥𝜌𝑤𝑔)−𝜆𝑜𝑔(𝑃𝑜𝑥𝜌𝑜𝑔)−𝜆𝑔𝑔(𝑃𝑔𝑥𝜌𝑔𝑔), (3.15) andthefollowingrelationsaredefined:

𝜆𝑤𝑤 = 𝜙𝑠2𝑤

𝑅 (𝑅𝑜𝑅𝑔̂𝑘2𝑜𝑔), 𝜆𝑤𝑜=𝜆𝑜𝑤=𝜙𝑠𝑤𝑠𝑜

𝑅 (̂𝑘𝑤𝑜𝑅𝑔+̂𝑘𝑜𝑔̂𝑘𝑤𝑔), 𝜆𝑜𝑜 = 𝜙𝑠2𝑜

𝑅 (𝑅𝑤𝑅𝑔̂𝑘2𝑤𝑔), 𝜆𝑤𝑔=𝜆𝑔𝑤= 𝜙𝑠𝑤𝑠𝑔

𝑅 (̂𝑘𝑤𝑔𝑅𝑜+̂𝑘𝑜𝑔̂𝑘𝑤𝑜), 𝜆𝑔𝑔 = 𝜙𝑠2𝑔

𝑅 (𝑅𝑤𝑅𝑜̂𝑘2𝑤𝑜), 𝜆𝑜𝑔=𝜆𝑔𝑜=𝜙𝑠𝑜𝑠𝑔

𝑅 (̂𝑘𝑜𝑔𝑅𝑤+̂𝑘𝑤𝑔̂𝑘𝑤𝑜), (3.16) where

𝑅𝑤= ̂𝑘𝑤+̂𝑘𝑤𝑔+̂𝑘𝑤𝑜, 𝑅𝑜= ̂𝑘𝑜+̂𝑘𝑤𝑜+̂𝑘𝑜𝑔, 𝑅𝑔 = ̂𝑘𝑔+̂𝑘𝑤𝑔+̂𝑘𝑜𝑔,

𝑅= ̂𝑘𝑤̂𝑘𝑜̂𝑘𝑔+(̂𝑘𝑤+̂𝑘𝑜+̂𝑘𝑔)(̂𝑘𝑤𝑔̂𝑘𝑤𝑜+̂𝑘𝑜𝑔̂𝑘𝑤𝑜+̂𝑘𝑤𝑔̂𝑘𝑜𝑔)

+̂𝑘𝑔̂𝑘𝑤𝑜(̂𝑘𝑤+̂𝑘𝑜)+̂𝑘𝑤̂𝑘𝑜𝑔(̂𝑘𝑜+̂𝑘𝑔)+̂𝑘𝑜̂𝑘𝑤𝑔(̂𝑘𝑤+̂𝑘𝑔). (3.17) Usingcapillarypressurerelations(3.12)itfollowsthat(3.15)take thefollowingequivalentform:

𝑈𝑤 =−̂𝜆𝑤𝑃𝑤𝑥−(𝜆𝑤𝑜+𝜆𝑤𝑔)Δ𝑃𝑜𝑤𝑥𝜆𝑤𝑔Δ𝑃𝑔𝑜𝑥 +(𝜆𝑤𝑤𝜌𝑤+𝜆𝑤𝑜𝜌𝑜+𝜆𝑤𝑔𝜌𝑔)𝑔,

𝑈𝑜 =−̂𝜆𝑜𝑃𝑤𝑥−(𝜆𝑜𝑜+𝜆𝑜𝑔𝑃𝑜𝑤𝑥𝜆𝑜𝑔Δ𝑃𝑔𝑜𝑥+ (𝜆𝑤𝑜𝜌𝑤+𝜆𝑜𝑜𝜌𝑜+𝜆𝑜𝑔𝜌𝑔)𝑔, 𝑈𝑔 =−̂𝜆𝑔𝑃𝑤𝑥−(𝜆𝑔𝑔+𝜆𝑜𝑔𝑃𝑜𝑤𝑥𝜆𝑔𝑔Δ𝑃𝑔𝑜𝑥

+(𝜆𝑤𝑔𝜌𝑤+𝜆𝑜𝑔𝜌𝑜+𝜆𝑔𝑔𝜌𝑔)𝑔. (3.18)

Herewedefinethefollowingnotationforgeneralizedphasemobilities

̂𝜆𝑖:

̂𝜆𝑤=𝜆𝑤𝑤+𝜆𝑤𝑜+𝜆𝑤𝑔,

̂𝜆𝑜=𝜆𝑜𝑜+𝜆𝑤𝑜+𝜆𝑜𝑔,

̂𝜆𝑔=𝜆𝑔𝑔+𝜆𝑤𝑔+𝜆𝑜𝑔.

(3.19)

Bysumming𝑈𝑤,UoandUgin(3.18)andusingthenotationintroduced in(3.19),thetotalDarcyvelocitycanbeexpressedasfollows:

𝑈𝑇=−̂𝜆𝑇𝑃𝑤𝑥−(̂𝜆𝑜+̂𝜆𝑔𝑃𝑜𝑤𝑥̂𝜆𝑔Δ𝑃𝑔𝑜𝑥+(̂𝜆𝑤𝜌𝑤+̂𝜆𝑜𝜌𝑜+̂𝜆𝑔𝜌𝑔)𝑔 (3.20)

wherewehaveused

̂𝜆𝑇= ̂𝜆𝑤+̂𝜆𝑜+̂𝜆𝑔. (3.21)

Therefore,thewaterpressuregradientcanbederivedfrom(3.20): 𝑃𝑤𝑥=− 1

̂𝜆𝑇𝑈𝑇−(𝑓̂𝑜+𝑓̂𝑔𝑃𝑜𝑤𝑥𝑓̂𝑔Δ𝑃𝑔𝑜𝑥+(𝑓̂𝑤𝜌𝑤+𝑓̂𝑜𝜌𝑜+𝑓̂𝑔𝜌𝑔)𝑔 (3.22)

withgeneralizedfractionalflowfunction:

𝑓̂𝑖= ̂𝜆𝑖̂𝜆𝑇, (𝑖=𝑤,𝑜,𝑔). (3.23)

Inserting(3.22)into(3.18)weget:

𝑈𝑤=𝑓̂𝑤𝑈𝑇+(𝑊𝑜+𝑊𝑔𝑃𝑜𝑤𝑥+𝑊𝑔Δ𝑃𝑔𝑜𝑥−(𝑊𝑤𝜌𝑤+𝑊𝑜𝜌𝑜+𝑊𝑔𝜌𝑔)𝑔, 𝑈𝑜=𝑓̂𝑜𝑈𝑇+(𝑂𝑜+𝑂𝑔𝑃𝑜𝑤𝑥+𝑂𝑔Δ𝑃𝑔𝑜𝑥−(𝑂𝑤𝜌𝑤+𝑂𝑜𝜌𝑜+𝑂𝑔𝜌𝑔)𝑔, 𝑈𝑔 =𝑓̂𝑔𝑈𝑇+(𝐺𝑜+𝐺𝑔𝑃𝑜𝑤𝑥+𝐺𝑔Δ𝑃𝑔𝑜𝑥−(𝐺𝑤𝜌𝑤+𝐺𝑜𝜌𝑜+𝐺𝑔𝜌𝑔)𝑔, (3.24) where

𝑊𝑖= ̂𝜆𝑤𝑓̂𝑖𝜆𝑤𝑖, 𝑂𝑖= ̂𝜆𝑜𝑓̂𝑖𝜆𝑜𝑖,

𝐺𝑖= ̂𝜆𝑔𝑓̂𝑖𝜆𝑔𝑖, (𝑖=𝑤,𝑜,𝑔). (3.25)

Itshouldbenotedthat𝑊𝑖+𝑂𝑖+𝐺𝑖=0(𝑖=𝑤,𝑜,𝑔)inlightof(3.16), (3.21),and(3.23).

4. Numericalexamples

Wemainlyfocusonareservoirmodelwherethereareoneinjection well atthecenterandtwoproductionwellsdistributedattwosides.

Theinjectionrateisequaltothetotalproductionrateandtheratesof twoproductionwellsarealsosame(SeeFig.1).Inaddition,reservoir inclination𝜃isalsoaccountedforinthemodel.

Fig.1. Reservoirmodelwithinjectionandproduction.

(5)

Fig.2. Left:Capillarypressurebetweenwater andoil.Right:Capillarypressurebetweenoil andgas.Wereferto(3.12)fortheirexpressions andTable1fortheinputparameters.

Fig.3. Waterfractionalflowfunction𝑓𝑤(𝑠𝑤,𝑠𝑜)(definedin(4.27))witheffectsofmodelinclination𝜃andtotalflowdirectionofUT.(A):𝑠𝑖𝑛𝜃=0,𝑈𝑇=𝑄𝑝∕2;(B):

𝑠𝑖𝑛𝜃=1,𝑈𝑇=−𝑄𝑝∕2;(C):𝑠𝑖𝑛𝜃=1,𝑈𝑇 =𝑄𝑝∕2.

Interactionterms

Themodel(3.11)4,5,6shouldbearmedwithappropriatefunctional correlationsforfluid-rockresistanceforcê𝑘𝑤,̂𝑘𝑜,̂𝑘𝑔andfluid-fluiddrag forcê𝑘𝑤𝑜,̂𝑘𝑤𝑔,̂𝑘𝑜𝑔.Hereweusetheinteractiontermssuggestedinthe recentworks(Standnesetal.,2017;Qiaoetal.,2018;Andersenetal., 2019):

̂𝑘𝑤 ∶=𝐼𝑤𝑠𝛼𝑤𝜇𝑤

𝐾𝜙, ̂𝑘𝑜∶=𝐼𝑜𝑠𝛽𝑜𝜇𝑜

𝐾𝜙, ̂𝑘𝑔∶=𝐼𝑔𝑠𝛾𝑔𝜇𝑔 𝐾𝜙,

̂𝑘𝑤𝑜 ∶=𝐼𝑤𝑜𝑠𝑤𝑠𝑜𝜇𝑤𝜇𝑜

𝐾 𝜙, ̂𝑘𝑤𝑔∶=𝐼𝑤𝑔𝑠𝑤𝑠𝑔𝜇𝑤𝜇𝑔

𝐾 𝜙, ̂𝑘𝑜𝑔∶=𝐼𝑜𝑔𝑠𝑜𝑠𝑔𝜇𝑜𝜇𝑔 𝐾 𝜙.

(4.26)

Alltheinteractiontermŝ𝑘𝑖and̂𝑘𝑖𝑗havedimensionPa· s/m2.Theparam- eters𝛼,𝛽and𝛾aredimensionlessexponentswhereas𝐼𝑤,IoandIgare dimensionlessfrictioncoefficientscharacterizingthestrengthoffluid- solidinteraction.Finally,𝐼𝑤𝑜,𝐼𝑤𝑔andIogarecoefficientscharacterizing thestrengthofthefluid-fluiddragforcewithdimension(Pa· s)−1.

Inputdata

TheinputparametersusedinthesimulationsarelistedinTable1. We use 101 grid cells for a 100-meter reservoir layer. We refer to AppendixDforaconvergencetest.Themagnitudeoftheinteraction coefficients𝐼𝑤𝑜,𝐼𝑤𝑔,andIogarechosenasinQiaoetal.(2018)where weappliedageneralizedtwo-phasemodeltomatchtheexperimentally measuredrelativepermeabilitycurvesandobtainedvaluesfortheinput parameterssuchas𝐼𝑤𝑜whosemagnitudeisaroundseveralthousands.

Inorder toavoid toomanycomplicatingeffectsatthesametimein thesubsequentdiscussion,wehavesettheviscositytermstozero,i.e., 𝜀𝑤=𝜀𝑜=𝜀𝑔=0.

We use the similar capillary pressure relations as Qiaoetal.(2019b)forwaterandoilandLewisandPao(2002)foroil andgas(seeFig.2).Theexpressionofaneffectivewaterfractionalflow function𝑓𝑤(𝑠𝑤,𝑠𝑜)intheconventionalwater-oil-gasmodel(assuming nocapillarypressure,i.e.,Δ𝑃𝑜𝑤𝑃𝑔𝑜=0)is

𝑓𝑤(𝑠𝑤,𝑠𝑜)∶=def 𝑈𝑤 𝑈𝑇 =

̂𝜆𝑤

̂𝜆𝑇𝑈𝑇−(𝑊𝑤𝜌𝑤+𝑊𝑜𝜌𝑜+𝑊𝑔𝜌𝑔)𝑔sin𝜃

𝑈𝑇 (4.27)

(6)

Fig.4. Resultsofthehorizontalcompressiblethree-phasemodelduringa400-daywaterfloodingperiod.Thesourcetermeffectscanbeseenclearlyinallplots whereproductionwellsarelocatedat10mand90mandinjectionwellat50m.(A)Waterpressureplotshowsastrongpressuregradientregionattheearlystage (before130days).(B)Watervelocityprofile.Itcanbeseenthatwater—-frontreachestheproductionwellafteraround100days.(C)Normalizedwatersaturation showsthatthewaterfrontisfastwhereastheotherphases(oilandgas)areproducedslowly(takesalmost300days).(D)Oilpressureprofilegivesasimilarresult aswaterpressure.(E)Oilvelocitybehaviorissimilartowatervelocity.(F)Normalizedoilsaturationplotillustratesthatoilisdisplacedquiteslowly.(G)Thegas pressuregradientisverylowinthegas-displacedregionattheearlystageduetothehighmobilityofgas.(H)Thereisnogasadvancingfrontsincegasflowseasily.

(I)Gasisdisplacedfastlyandalotofgasisrecoveredbefore130days.

Table1

Referenceinputparametersinthesimulations.

Parameter Dimensional value Parameter Dimensional value

L 100 m 𝐼 𝑤 2.5

𝜙 0.25 I o 1.8

̃𝜌𝑤0 1 g/cm 3 I g 1.1

̃𝜌𝑜0 0.8 g/cm 3 𝐼 𝑤𝑜 3000 (Pa · s) −1

̃𝜌𝑔0 0.018 g/cm 3 𝐼 𝑤𝑔 3000 (Pa · s) −1

𝑠 𝑤𝑟 0 I og 3000 (Pa · s) −1

s or 0 𝛼 0.01

s gr 0 𝛽 0.01

𝜇𝑤 1 cP 𝛾 0.01

𝜇o 1.5 cP 𝑃 𝑐1 4 10 4Pa

𝜇g 0.015 cP a 1 2

K 1000 mD 𝛿1 0.08

𝑘 max𝑟𝑤 0.4 𝑃 𝑐2 10 5Pa

𝑘 max𝑟𝑜 0.5556 a 2 2

𝑘 max𝑟𝑔 0.9091 𝐶 𝑤 10 6m 2/s 2 𝑄 𝐼𝑤 0.125 m 3/day C o 5 10 5m 2/s 2 𝑄 𝐼𝑔 0.125 m 3/day C g 10 5m 2/s 2 𝑄 𝑝 0.0625 m 3/day 𝜀 𝑤 0.0 cP

N x 101 𝜀 o 0.0 cP

A 1 m 2 𝜀 g 0.0 cP

𝑃 𝑤𝐿 10 6Pa x I 50 m

△t 1570 s x P(1,2) 10 (1)&90 (2)m

wherewehaveused(3.24)and(3.25)where𝑈𝑇=∫0𝑥(𝑄𝐼𝑄𝑝)𝑑𝑥.Sim- ilarly,foandfgcanalsobeexpressedinthesamemanner.Inorderto illustratethephaseflowfraction𝑓𝑤(seeFig.3)werepresentUTbya

referencetotalvelocity𝑈𝑇∈ [−𝑄𝑝

2 ,+𝑄𝑝

2].WerefertoTable1forother inputdatathatareused.

Initialconditions

For the waterflooding case, we assume the reservoir initially is mostlyfilledwithgas(90%)andsomeoil(10%):

𝑠𝑔(𝑥,𝑡=0)=0.9, 𝑠𝑜(𝑥,𝑡=0)=0.1. (4.28) FortheWAGinjectioncase,thereservoirisassumedinitiallyfilledwith oil(90%)andsomeextrawater(10%):

𝑠𝑜(𝑥,𝑡=0)=0.9, 𝑠𝑤(𝑥,𝑡=0)=0.1. (4.29) Forthecompressiblecase,areferencepressure𝑃𝑤𝐿attheleftboundary ofthelayerisgivenatinitialstate,

𝑃𝑤𝐿(𝑥=0,𝑡=0)=106Pa. (4.30)

Boundaryconditions

Weassumeaclosedboundaryforbothcompressibleandincompress- iblemodels,whichmeansthat

𝑢𝑖(𝑥=0,𝑡)=0, 𝑢𝑖(𝑥=𝐿,𝑡)=0, 𝑖=𝑤,𝑜,𝑔. (4.31) Fortheincompressiblecase,wegiveareferencepressure𝑃𝑤𝐿attheleft boundaryofthelayer,

𝑃𝑤𝐿(𝑥=0,𝑡)=106Pa. (4.32)

(7)

Sourceterms

ForWAGexperiments,gasandwaterareinjectedatdifferenttime periodsduringthewholeoilrecoveryprocess.WeassumethatQI(x)and Qp(x)taketheform

𝑄𝐼𝑤,𝐼𝑔(𝑥)= 𝑄𝐼𝑤,𝐼𝑔 𝜎

{1, if|𝑥−𝑥𝐼|≤𝜎∕2;

0, otherwise. , 𝑄𝑝(𝑥)= 𝑄𝑝

𝜎

{1, if|𝑥−𝑥𝑝,𝑖|≤𝜎∕2;

0, otherwise. (4.33)

where(𝑖=1,2)and𝑄𝐼𝑤,𝐼𝑔=0.125m3∕dayand𝑄𝑝=0.0625m3∕day.The widthofthesmallregionassociatedwiththeinjectorandproduceris 𝜎.Inthenumericalscheme𝜎𝑥.

4.1. Waterfloodinginagasreservoir

Wefirsttesttheproposedcompressiblethree-phasemodelappliedto agasreservoirdevelopment.Inthisexample,waterisinjectedat50m intoagasreservoirlayeroflength100mwithalittleproportionofoil (10%).Twocases,respectively,forthehorizontal(Fig.4)andvertical reservoir(Fig.5)areshownbelow.

Theresultsofthehorizontalcompressiblethree-phasemodelwith waterinjectionforatotalperiodof400daysareshowninFig.4where pressures (first column), velocities (middle column) and saturations (rightcolumn)aresymmetricwiththeinjectionwelllocatedatthecen- terofreservoirlayer.Thegasismostlyrecoveredduringthefirst130 days,see(I),whereasoilrecoverytakesplaceovermorethan300days, see(F),duetoitslowermobilitythangas.Itisalsoobservedthatat earlystagegaspressurealongthereservoirlayerhaslessgradientthan boththewater’sandtheoil’s(seefirstcolumninFig.4).Theinjected waterdisplacesbothoilandgasinthereservoirneartheinjectionwell regionwhereahighpressuregradientisnecessaryforbothwaterand oiltoflow,seepanel(A)and(B),becauseoftheirlowmobilities.After waterhasarrivedtheproductionwellsataround100days(seeC),wa- terandoilpressuresdropowingtothefactthatwaterthencanfindan easyflowpathtotheproductionwells.

InFig.5,weshowtheresults(phasepressures,velocitiesandsat- urations)ofacompressibleverticalthree-phasemodelwitha400-day waterfloodingdisplacement.Waterisinjectedtodisplaceoilandgasat bothsidesofthereservoirlayer.Itquicklyfillsthebottompart,then starts accumulating,see panel(C).Correspondingly,gas is displaced faster inthelowerpartthanin theupperpartbecausethereservoir layerisvertical.Gravitysegregationisseeninthelowerpartwheregas

Fig.5. Resultsoftheverticalcompressiblethree-phasemodelduringa400-daywaterfloodingperiod.Thesourcetermeffectsareidentifiedinthevelocityand saturationplotswhereproductionwellsarelocatedat10mand90mandinjectionwellat50m.(A)Waterpressureplotindicatesthatalotofwaterflowstoward thebottomandbythatgreatlyincreasesthepressureinthatregion.(B)Duetothestronggravityeffectwaterflowsfastertowardsthebottomoflayercompared thewaterdisplacementintheupperlayer.(C)Normalizedwatersaturationshowsthatwaterflowsfastlytothebottomwhereitisaccumulatedbeforeitbegins toefficientlydisplacetheupperpartofthelayer.(D)Oilpressurefollowsthesimilarbehavioraswater.(E)Waterdisplacestheoiltowardsbothsidesfromthe center.However,atearlytimesomeoilintheupperpartofthelayerwillmovedownwardlyduetogravity.Later,thewaterfrontwilldisplaceoilupwardly.(F) Theoiladvancingfrontbehavessimilarasthewaterfront.(G)Gaspressurebehavessimilartothewaterpressure.(H)Atanearlystagegasisdisplacedtowardsthe productionwellfromthecenter.Afterthewaterfronthasreachedthebottomproductionwellthewholebottompartofgas(50mto100m)startsmovingupwards.

(I)Gasisrecoveredslowlyintheupperpartwhereasgasrecoveryinthelowerpartconsistsoftwostages:initially,gasisdisplacedbywatertothebottomproduction well.Then,gasinthelowerzonestartsflowingupwardly.

(8)

Fig.6. Comparisonbetweenthecompressibleandincompressiblemodelwithverticalthree-phaseflow.(A,D)Phasevelocity𝑢𝑤andugforwaterandgas,respectively.

(B,E)Pressure𝑃𝑤andPgforwaterandgas,respectively.Thecompressiblemodelaccountsforthefactthatgasissignificantlycompressedandstoresenergywhich isremovedfromthesystemasgasisproduced.Thisgivesrisetolowerpressureprofilesforthecompressiblecaseascomparedtotheincompressiblecase.Thisgives risetoalowerpressurelevelforthecompressiblemodelascomparedtotheincompressible.(C,F)Saturation𝑠𝑤andsgforwaterandgas,respectively.

issqueezedupwardly,see(H)and(I).Incontrasttowhatisshownin Fig.4G,gaspressuredistributionshowsasimilarbehavioraswaterand oil(higheratbottomandlowerattop),seefirstcolumninFig.5.We refertothefiguretextformoredetails.

4.1.1. Comparisonofthecompressibleandincompressiblemodels WecontinuethediscussionofthecaseshowninFig.5.Inparticular, wewanttocomparethebehaviorofthecompressibleandincompress- iblemodel.Constantdensityvalues 𝜌𝑤=1000kg∕m3,𝜌𝑜=800kg∕m3 and𝜌𝑔=18kg/m3areusedintheincompressiblemodel.

Fig.6showsacomparison between thecompressibleandincom- pressiblemodelafter30 and120days.(A)shows thatatearlystage theinjectedwaterinthecompressiblemodelpreferstodisplacegasin thelowerpart(highpositivevalue)sincewaterleadstohigherpressure atthebottomsuchthatthegasiscompressedthere.Withcompressed gasproducedatthebottomandgasexpandingintheupperpart,gas willonly slowlymigratetowardstheupperpartresultingin compa- rablylowervelocity(negative)inthecompressiblemodel.Thevelocity differenceshownin(D)fitswellwiththesaturationdifferenceafter120 days.Attheearlytime(30days)thesaturationdifferencesarenotdis- tinct,see(C).However,afteralongtime(120days)thedifferencesare moresignificant,especially,inthewaterdisplacingpart,see(F).This isduetotheincreasingphasepressuredifferencebetweencompressible andincompressiblemodel,see(B)and(E).Theremovalofcompressed gasfromthegasreservoiras(almostincompressible)waterisinjected clearlygeneratesadditionalspaceforthewatertofillwhichgivesrise toalowerpressure.

4.2. Thecompressiblethree-phasemodelwithaWAGexperiment

InWAGprocesses,theinjectedwaterwillmigratetowardsthebot- tomoftheformationwhiletheinjectedgaswillflowupwardly.There- fore,counter-currentflowoccursintheverticaldirectionofthereser- voir duetothegravity segregationof water, oilandgas. Significant differencesintermsofsaturationdistributionandproducingGOR(gas- oil-ratio)havebeenreportedbetweenaconventionalmodelandmodels thatbettercanaccountforthemixofdifferentflowregimes(co-current andcounter-current).Forexample,inSherafatiandJessen(2017)an explicit representation of flow transitions between co-current and counter-currentflowwasusedtoimprovethedesignofWAGinjection processes.

Inthispart,weconductawateralternatinggas(WAG)injectionina 1Dreservoir(250mD)layerwhichinitiallycontains90%oiland10%

water.Thewaterandgasinjectionwellislocatedat50mandtwopro- ductionwellsaresetat10mand90m.Gasisinjectedforthefirst10 daysfollowedbythewaterinjectionthenext10days.Fluidscanbepro- ducedinbothproductionwells.ThewholeWAGexperimentcontinues withaninjectioncirculationofwaterandgas(eachfor10days).

Fig.7showstheresultforaWAGinjectionprocessproducedbythe compressiblethree-phaseverticalmodelwheregravitysegregationhas asignificanteffect.Fromthesimulationweseethatpressureincreases withtime(firstcolumninFig.7).Moreover,pressurevaluesatthelower partofthelayerarelargerthanattheupperpart.Duetothedensity difference,waterdisplacesoilfasterinthebottompart,see(B)and(C).

Inaddition,gasflowsquicklytowardstheupperpartofthereservoir

(9)

Fig.7.Resultsoftheverticalcompressiblethree-phasemodelfora400-dayWAGinjectionprocess.Thesourcetermeffectsarevisibleinthevelocityandsaturation plotswhereproductionwellsarelocatedat10mand90mandinjectionwellat50m.(A)Ahighpressureregioninthelayercenterduetothewaterorgasinjection andgravityeffect.(B)Wateradvancingfrontimpliesthatwaterflowsfastertowardsthebottomoflayercomparedthewaterdisplacementintheupperlayerdue togravitysegregation.(C)Waterpreferstoflowtowardsthebottomoflayerwheretheedgeregion(90m-100m)isalsosweptbywater.(D)Oilpressurefollows similarbehavioraswaterpressure.(E)Theupperpartofoilisrecoveredfasterthanthelowerpart.(F)Duetothelargedensitydifferencebetweenoilandgas,the upperpartoilisrecoveredveryquickly,evenfortheedgeregion(0m-10m).(G)Gaspressure.(H)Gasadvancingfrontisfastintheupperpartoflayerbecauseof thestronggravitysegregation.(I)Gasreachesthebottomproductionwellwhereasalotofgasisaccumulatedinthetopregion.

layer,seethesaturationplots.Intheupperpartoilisrecoveredfaster thaninthelowerpartbecauseofthelargerdensitydifferencebetween gasandoilthantheonebetweenwaterandoil,seethesecondcolumnin Fig.7.Wealsoobservethatgasreachesthebottomproductionwellbut doesnotmovefurther.Thiscanbeexplainedbythefactthatgravity segregationeffectovercomesthecapillarity.However,alotofgasis accumulatedintheupperedgeregion(0m-10m)duetothebuoyancy force,see(I).

4.3. Comparisonofcompressibleandincompressiblethree-phasemodels withWAGexperiments

Inthispart,wecomputesolutionsfromincompressiblethree-phase modelswithsameWAGinjectionprocessandcomparetherelevantre- sultswiththosefromthecompressiblethree-phasemodel.Constantden- sityvalues𝜌𝑤=1000kg∕m3,𝜌𝑜=800kg∕m3and𝜌𝑔=18kg/m3areused intheincompressiblemodel

Fig.8showsacomparison between thecompressibleandincom- pressiblemodeloftheverticalthree-phasereservoirwithaWAGpro- cess.Similartowhatwasobservedin Fig.6,differencesareseenfor

phasevelocity,pressureandsaturation.Withincreasingtime,thisdif- ferencewillbeenhanced,especiallyforthepressure.Thisismainlydue tothegascompressibility.See(B)and(E)andthefiguretextformore explanation.Becauseofthedensitydifferencewaterpreferstoflowto- wardsthebottomofthelayerwhereas gasmovesfastertowards the upperpartoflayer,see(C)and(F).

4.3.1. Effectoffluid-fluidinteractions

Here wewanttoillustratetheimpactfrom fluid-fluidinteraction termsonthecompressiblemodelwithaWAGprocess.Twosituations arecompared below:onewith𝐼𝑤𝑜=𝐼𝑤𝑔=𝐼𝑜𝑔=0(𝑃𝑎𝑠)−1 andone with𝐼𝑤𝑜=𝐼𝑤𝑔=𝐼𝑜𝑔=5000(𝑃𝑎𝑠)−1.

Fig.9comparestheresultsforthehorizontalmodelforaWAGpro- cesswithandwithoutfluid-fluidinteractioneffectat60and120days.

In(B)and(E),weobservethatduetothefluid-fluidinteraction,pres- sureiselevatedcomparedwiththecasewithnofluid-fluidinteraction.

Thewatervelocity(A)andsaturationprofiles(C)showthatwaterto alessextentdisplacesoilandinsteadflowsthroughtheoriginalwa- terchannelswhenfluid-fluidinteractionisincluded.Thedifferencein thewatersaturationprofilesbetween(C)and(F)isenhancedwithtime

(10)

Fig.8. Comparisonbetweenthecompressibleandincompressiblemodeloftheverticalthree-phasereservoirwithaWAGprocess.Resultsareshownafter60and 120days.(A)Gravitysegregationresultsinafastadvancingfrontofgasintheupperpartoflayerandafastadvancingfrontofwaterinthelowerpartoflayer.

(B)Phasepressureinthecompressiblemodelishighersincethecompressedgaswantstoexpandwhenitmovestoaregionwithlowerpressurebutcannotexpand duetotheconstrainedspaceforgas.(C)Gaspreferstomovetowardstheupperpartoflayerandwaterpreferstoflowtowardsthelowerpart.(D)At120days,gas reachestheupperproductionwellandwaterarrivesatthebottomwell.(E)Phasepressureinthecompressiblemodelincreaseswithtimecomparedwith(B).(F) Thedifferencebetweenthetwomodelsisenhancedwithtime.

Fig.9. ComparisonforthehorizontalcompressiblemodelforaWAGprocesswithandwithoutfluid-fluidinteractioneffects.Thesituationafter60and120days areplotted.(A)Phasevelocityat60days.(B)Phasepressureat60days.(C)Normalizedsaturationat60days.(D)Phasevelocityat120days.(E)Phasepressureat 120days.(F)Normalizedsaturationat120days.

Referanser

RELATERTE DOKUMENTER

The system can be implemented as follows: A web-service client runs on the user device, collecting sensor data from the device and input data from the user. The client compiles

alumina/epoxy material system data is applied in the plot. The calculated stiffness is compared to and found to agree with the model results using the general three-phase

The dense gas atmospheric dispersion model SLAB predicts a higher initial chlorine concentration using the instantaneous or short duration pool option, compared to evaporation from

Based on the above-mentioned tensions, a recommendation for further research is to examine whether young people who have participated in the TP influence their parents and peers in

Overall, the SAB considered 60 chemicals that included: (a) 14 declared as RCAs since entry into force of the Convention; (b) chemicals identied as potential RCAs from a list of

The SPH technique and the corpuscular technique are superior to the Eulerian technique and the Lagrangian technique (with erosion) when it is applied to materials that have fluid

Azzam’s own involvement in the Afghan cause illustrates the role of the in- ternational Muslim Brotherhood and the Muslim World League in the early mobilization. Azzam was a West

The ideas launched by the Beveridge Commission in 1942 set the pace for major reforms in post-war Britain, and inspired Norwegian welfare programmes as well, with gradual