SPE 25233 Reservoir Simulation Grids: Clppmtunitiw and Problems
SPE 25233 Reservoir Simulation Grids: Clppmtunitiw and Problems
SPE 25233 Reservoir Simulation Grids: Clppmtunitiw and Problems
w
9
Se@Myof
Petreleunr
Eflglneers
.
SPE 25233
Reservoir Simulation Grids: clppmtunitiw;
and Problems
3, 1993.
This paper was solectod for prosontcd!on by on SPE Pro~am Commliteo following rovlew of btformidlon contained br an abstract 8ubn?lHod by tho ou!hor(s), Comonto of lho paper,
as presemod, have not berm reviewed by the Society of Polrolm!m Englnoors and are subject to correction by tho author(s), The material, ae PIQ80nled, doos not nocessmlly reflect
any position of Iho Socloty 01 Poboleum Enghwors, its ofllcow, or members. PaPQrOFr090nted et SpE m~etln9s are sublect to Publlcatlon ro~iow by ~ditorl~l Committees of the SOclOtY
of Petroleum Engineers. Porm&Aon to copy Is rs$irlctsd to an abstract of not mom thsn 300 words. Illustra![on$ may not be copied, The abstract should contah! conspicuous rmknowl~dg,
monf of where and by whom the psper [s presented, Write Publlcritions Manogor, SPE, P.O. Box 833330, Rlchardsorr, TX 750?3-3830, U.S.A. Telex, 1S3245 SPEUT.
ABSTRACT
I!NTRC)DUCTION
A reservoirsimulator predicts reservoir performance
by solving flow equations on a discrete grid that is
chosen by the simulation engineer to represent the
reservoir.The grid is normally selected with one or
moreof the followingconsiderationsin mind:
I .OCitl ({ridRCfillCIUCIlt
Iiybiid
(;rifJ
RIMIRVOIRSWULATKONGRIDS: OPPORTUNITIES
ANDPROBLEMS
* Curvilinear(Stream.Tube)Gtid
@ Voronoi or PEBI Grid (GeneraHzationof PointDistributedGrid)
e CornerPointGeomehy
o DynamictMd
e AutomaticGrid Generation
CNher techniques available in
simulatorsanx
.
some research
ElasticGrid AdjustmentMethods
ControlVohmmFinite ElementMethods
Free LagrangianMethods
WE 25233
FLOW ECNJATK)P4S
The conservationof mass for componentc (for BlackOil models c = oil, gas and water at standard
condhions) combined with Darcys law yields the
followingset of flowcqmtions:
,,,.
(1)
Ckthl!d
,*
1
t,
P{J
[1
M%%,,.!
::: ..-.
dpp
..-..-..
..s----
.....
. i,]
,,.,
(2)
8
SPE 25233
KHALIDAZIZ
SWwAmor+1
GRIDS
Giobdy Orthogonal Grids: Most commonly used
grids are constructedby aligning the gridblock along
orthogonal coordinate directions, and then distorting
the grid, where nemmary, to fit major rescrvcrir
features (e.g. fllp). Examples of these Idnds of grids
are the sttildard Cartesian block-centeredand pointdktributed grids shown in l%g.2.
Y4
?-
. . . . .
I
,<,J:-
-T
...L..
I
-
.....
TR:
--.-...:TJ
---
.--.:
--
-fi.l-
--
-Q
.c.
a)
A-.
11.IJ
I
~~J-.
Rlock.wmmxlSrirl
!$:: :;J;:
..-.$....a
..
--
-L--
, I+lJ
.- -.:
.&
-j4_-
,+*J
-l----
..-
IJ+l
. -f--
.-_-,_
-.
--
b) l)obit4iNrilMllcd
grid
L...+
RJLWRVOIR
SM.JLATIONGIWS: OIWMWUNITIESANDPROBLEMS
SPE
25233
(1)hexagonal
Fig.3:
HybNK.I
Grid (Pedrosaand Aziz[8])
(a)Cwcdan
(f)
,,,.,,...
..
,...
....
l:,
,,
(4) cylindrical
(d)IlcK#gwl*l
liyww.iwbn
(g)
.....
...
,...,
,/
....
..<
(5) eraser
:.
....-.
..
----.
........,
.
----I-In
.. ---
(c)CyliMd
(3) irregular
(2) Cartesian
(c) (Xlnuhwar
liybdd.hexa#3nd
SPE 25233
AZIZ
KI~ALI~
(PaJagiandAziz[11])
Even greateraccuracywith the samekind of flexibility
is possible with finite element and Control Volume
Finite J31ement(CVFE) methods(ForsytJl[15]; Fung
et id. [16]; Kocberber and Collins [17]), but the
computational cost resulting from the additional
complexity of the flow equations is probably not
justified for generalfiehJscaleapplications,
Corner Point Geometry, h is possibleto accuratdy
complex rcsmvoir gconmtriesby specifying
the corners of each grir.ibbek this is known w
Comer Point Ckometry (Eclipse 100 91A Maawd).
While !Jle calculations arc more involved than in
standard Cartcsinn grids, till of the geometric
qwmtiticsin Eq. 1can be cahxdatcd.Ilc rwdproblem
wi!h this type of grid is that now flow across a block
flwcdependson more than two pressureon d.hcr side
of that face. This complicatesthe flow term iil J%. 1.
Ihc reason for this is that when the grid is skewed,
connectionsbelwecnblocks are no longer orthogonal
to the block faCcSt[Jllkxs fill COJIlpOllCIltS Of the
potonthd gradientat tlw bhck farxare consideredthis
kind of non-orthogonalgrid can lead to seriouserrors
in the calculationof interbloekflow,Almostthe swne
rcprcscnt
(Sharpaud Anderson[22])
11
RWEMKMIZSIMULATIONGIUOS: OPPORTUNITIJZS
ANDPROIILtiMS
Since this algorithm may produce a skewed (nonorthogomd) grid when the orthogonality condition
cannotbe satisfied,it has someof the sameproblem,as
the comer poirrt geometry with respect to the
depenrlcnceof flowon all cotnponcntsof the gradient
of the Potential. The main advantage of Sharp and
Andersonsmethod is that it tries to make the grid as
orthogonalas possibIe.
Grids
with
Grmeration
Homogeneous Mocks. M the transmissibility
(Eq,
2) an cffcctivc
value
25233
Autematic
calculation
SW
of permeability
PRACTICAL LEE
Il]eneed to satisfyia field scale applicationsa waricty
---2.=-T-T7----1of coustr,aintsthat arc often conflicting nuakes the
process of grid generation both difficult and time
consuming, fhis is particuhirly true of geologically
complex field like the C.Mlfaksin the Norwegian
sector of the North Sea (Pcttcrson [30]), In ordc!rt.o
rcprcscnt tho complex systcm of faults in this ticld
severalmonthsarc requiredto dwclop an appropriiltc
grid, For such reservoirs corner point ,gcomctryand
Voroi~oigrids providethe requiredflexibility,but the
simulationengineerwill llOrMillly select the onc that is
[30], tOOh$
easier to USC, As pointed out by Pcttcrson
that
can
help
the
reservoir
cnginccr
in
quickly
12
.,.
SPE 25233
~WLIDAZIZ
NOMENCLAT!.NiE
A
d
At
o
fc
i
[32].
CO?KXJJCNIW3REMARKS
lle simulation engineer has the opportunity to use
many khlds of flexible grids, However, from a
practical standpoint the most important problem is:
Whatgrid shouldone selectfor a specificprubhmi?As
mentionedearlier in this paper,increasein the number
of grkiblocks does not automatically translate into
increased accuracy, Here are some comments and
guidelirres:
i, Local grid refinement (Cartesian or hybrid)
improvesthe predictionof WOR and GOR when
sharp saturation gradients exist,ne,arWCIIS as
ia coning problms. The refined region shouldbe
large enoughto include the extentof the reservoir
with sharp gradientsin saturation,
2, Unless there are compeilirrgreasons,grid should
h orthogonal at least locally - and as
uniform as possible, Large blocks next to small
blocks should Lwavoided,Irregular Voronoi grid
(generalkzatioi~
of point-distributedgrid)is usually
more reliablethan block-ccntcredgrid.
3. So f,ardynamicgrids have provca to be of limited
value in tkld applications, The ability to
efficientlyadd or removeblockswith the opening
or shuttingof wellsis howevera usefulfeature,
4, [JnsuwcturcdVoronoigrids requirespecialmatrix
solution techniques. Domain dr~ompmitionis a
useful technique for taking advantage of tbc
structure of the grid, A krrgc problem caa be
rcduwxlinm several smailcr proihns. This :ilsu
provides a natural appmacli to doing pilri!ll~i
compatalionsand localtirncstepping,
5, Iixpcrt systems arc needed for interactivc grid
gcncrittiotito takeadwuitagcof availnblcflriddilig
possibilities ilnd to rcduco the time involved in
buildingf.yidsfor geologicallycomplexrcwrvoirs.
6. Anidysis of results requires powerful dUW(Iimensional I%W ViSUilliZi{tiOll 0[1 Con)f)lCX flrids,
SoflWiU(>
for thC Visualizidiollt)t !IOWSOVCl
irfcgnlargrids is accdcdo
13
k
kr
Mc
N,l
Np
P
~c
r
~b
(ac
ACKNOWLEDGMENTS
Reservoirsirmilationresearchat StanfordUniversityis
su@porled by N1 international consortium of
organizations, dwougb the Stiinford University
Rrxetvoir Simulation Industrial Affiliates Program
(SUpR1-B).
FWEWNENCES
1. Paiirgi,C. L. and Aziz,K, I1andlin.gof Wells in
Simulators,Fourth InternationalForm on
ReservoirSim{ilaticrn,
Salzburg,Austria,August
31- September4, 1992,
2. Scttari,A, and Aziz, K, USCof Irrcgu]arGrid iii
ReservoirSimulation,[SPIiJ,Vol, 12,No, 2, 103114(1972).
3, Nw.11, 11. C. iin(l Azix,K, 1JWof I[~~~uli~h
ReservoirSinnilalion,S1}l;22886,Wh Al(:&l;,
l)i{llM,1X,(_)CtOl)CF 6-9, 199 I ,
4, QuirINldlc,P, itildIWsct, P. lllc IJSCof lWxiblc
firiddingfor hilprovcctMxcrvoirPcrformancc,
S1ll;12239,7th ReservoirSimulationSymposium,
$i]~)lkaacisco,CA, November15-18,1983,
.,$ 1kincniann, Z 11,, Gcrkcn,(i, iit]dIhltlcIN~oo,
(;.,[ JsingJAA grid Rcfiacmontin MultiplcApplicationReservoirSinluli\to14,
SPi112255,7i11
ReservoirSilllUlilli(}ll
Symposiuin,Siln1liUICiSCO,
(:A, November15-18,1983.
6, IItiilll!ll)ill) ll,Z, 1{. ad illalld, (~, W, (lridding
1cchni(iucs
in ReservoirSiln\lliltiOll,i>rocccdints
RXXU?RVOIR
!MMULATKON
Gwos: 43PPOMTJNITIKS
ANDPROBLEMS
May 1992.
11. Palagi,C. L. and Aziz,K, Useof VoronoiGrid
in ReservoirSimulation,SPB22889,66th
ATC&E,Dallas,TX, October6-9, 1991.
12, Palagi,C. L, and Aziz, K. TheModelingof
Horizontaland VerticalWells withVoronoi
Grid, SPE 24072,WesternRegionalMeeting,
Bakersfield,CA, March30-April 1, 1992,
13. Palagi,C, L. and Aziz,K. ADualllrnestcpping
Techniquefor ModelingTracerF1ow,Submitted
to SPE for publication,1991.
14. Consonui,P., Pa]agi,C. L,, Thiclc,M., DallOlio,
D., (hdi, A., and Pilotli, M. FlexibleGridding
Techniquesfor ConingStudiesin Horizontaland
VerticalWells, SUPRI-BReporton Joint
Stanford-AgipProjec4June 1992.
Forsyth,P.A. .4ControlVolumeFinite Element
MethodforLocal MeshRefinement,SPII 18415,
10thRcsrxvoir$imulationSymposium,Ikmston,
TX, Fobrunry6-8, 1989.
Fung, L, S,, IIicbert,A, D, and N.ghicm,1.,X,
ReservoirSimulationWitha Control-Vcdunw,
Finite-lllcmentMethod,SIW21224, 1Ith
ReservoirSimulationSymposium,An:dmim, CA,
February
SPE 25233
17-20, 1991.
14
.