Lake and Hirasaki (1981) (SPE-8436-PA)
Lake and Hirasaki (1981) (SPE-8436-PA)
Lake and Hirasaki (1981) (SPE-8436-PA)
Abstract
During 1953-54, Taylor showed that if a certain planation for the large field-scale dispersion observed
criterion is met the combined effects of the transverse in tracer test studies. Moreover, it appears that the
profile of longitudinal velocity and transverse dif- grouping procedure could indicate a method for
fusion on a solvent slowly flowing through a tube obtaining layered reservoir models from core 9ata.
will manifest themselves as a longitudinal diffusion
phenomenon. A similar phenomenon exists in Introduction
stratified porous media where the transverse profile In laboratory displacements, longitudinal (parallel to
of longitudinal velocity and transverse dispersion can the bulk fluid velocity) dispersion is well charac-
produce an effective longitudinal dispersion, called terized as consisting of additive contributions of
Taylor's dispersion in this paper. Since this effective diffusion and convection:
longitudinal dispersion is larger than the
corresponding homogeneous longitudinal dispersion, Kp=Do+O'f V , ..... (1)
the quantitative description of this phenomenon Fe/>
would be important to dispersion-sensitive EOR where K f is the longitudinal dispersion coefficient,
processes, such as surfactant or miscible flooding. D is the molecular diffusion coefficient, F is the
Taylor's dispersion will occur in two-layer porous fo~mation electrical resistivity factor, v is the in-
media if a suitably defined dimensionless number is terstitial longitudinal velocity, and O'f is the
much greater than unity. When this condition holds, longitudinal dispersity. At a velocity above about 0.1
the effluent history of a constant-mobility equal- ftiD (0.35 /Lm/s) the convection term dominates Eq.
density miscible displacement is that of the same 1, so for practical displacements - about 1 ft/D (3.5
displacement in a homogeneous medium with in- /Lm/s) - K f depends only on the term O'fv. O'f, in turn,
creased dispersion. The resulting effective is a function of average particle size and local
longitudinal dispersion may be derived analytically heterogeneity, and averages 0.05 to 0.2 in. (0.l3 to
and verified numerically as a function of several 0.51 cm) for homogeneous laboratory displacements.
media properties. The most important of these are Similarly, transverse (perpendicular to the bulk
system thickness and permeability contrast. fluid velocity) dispersion is
In multilayer media, when two adjacent layers
have a large transverse dispersion number they Kt = Do + O't V, ............ (2)
behave as a single layer with suitably averaged Fe/>
properties. This observation suggests an algorithm where K t is the longitudinal dispersion coefficient
whereby Taylor's dispersion may be extended to and O't is the transverse dispersivity. Measurements
multilayer systems. The algorithm, or grouping of O't are much less common than O'f but they in-
procedure, gives effluent histories that are in dicate 2 that O't = O' f /30.
agreement with numerical solutions to the continuity In the scaled differential material-balance
equation and allow properties of the resulting ef- equations, both K f and K t become part of Peclet
fective dispersion to be investigated. From the results numbers, vL/Kf and (vH/Kt)H/L, which appear as
of this work, Taylor's dispersion can offer an ex- inverses in the equations. In laboratory
0197-7520/81/0008-8436$00.25
displacements, O'f and O't are of the order of fractions
Copyright 1981 Society of Petroleum Engineers of AIME of centimeters and L is the order of centimeters. In
AUGUST 1981 459
( LONGITUDINAL) dispersion. The results of the work then are expanded
L to include multilayered reservoirs.
E HIGH VELOCITY k,
l~
Basic Assumptions and Definitions
The derivations and results of this report conform to
the following assumptions.
~ LOW VELOCITY k2
jl 1. The displacing and in-situ fluids are miscible,
incompressible, and have equal mobilities and
densities.
2. Near wellbore transverse pressure gradients,
FLOW
such as those that might be caused by partial
Fig. 1 - Definition of quantities.
penetration or a limited perforation interval, are
negligible; this assumption is necessary for the theory
presented below since Taylor's dispersion is an
asymptotic phenomenon for large well spacings.
field displacements, L is of the order of several tens Assumptions 1 and 2 eliminate all cross flow except
of meters; hence, many engineers, assuming that that caused by transverse dispersion. In particular,
dispersivities are roughly equal in the laboratory and the absence of transverse pressure gradients means
field, have concluded that dispersion is of little that vertical permeability will not appear in the
importance in field displacements. theory and that the transverse fluid velocity will be
Field measurements of C<:f' on the other hand, tend zero. The latter factor means that off-diagonal terms
to give values somewhat greater than the laboratory in the dispersion tensor will be zero, regardless of the
values. These measurements are few, being com- orientation of its major axes, since these are all
plicated by nonlinear flow, reservoir heterogeneities, proportional to transverse velocity. 10
and operational problems; however, recent work has 3. The injected fluid's concentration is zero
yielded field-measured values from 0.002 to 0.11 ft initially and the system is subjected to a unit step
(6.1 X 10- 4 to 13.4 X W- 2 m) in single-well tests3,4 function at t = O.
to 8 ..1 ft (2.5 m) in two-well tests. 5 Such large values 4. Throughout this paper we take the intralayer
C<:t = c<: f I30. The dispersivities are assumed
are of concern to those designing mixing-sensitive
displacements such as surfactant flooding, miscible homogeneous (independent of position) and
flooding, or the displacement of condensate fluids by anisotropic (since c<:e;C. C<:t).
nitrogen-enriched natural gas. 6 5. Longitudinal and transverse dispersion coef-
Large C<:f values in field displacements are at- ficients are proportional to the first power of the
tributable to the larger-scale heterogeneities present interstitial velocity as given by Eqs. 1 and 2.
in field-scale porous media. By assuming the porous 6. Longitudinal transport of injected fluid by
medium to be made up of heterogeneities randomly dispersion is much slower than longitudinal transport
distributed statistically, other workers 7 have shown by bulk flow.
that field and laboratory-scale dispersion coefficients 7. When dealing with two layers only, we assume
can differ by factors of 10. Given the stratified that the porous medium consists of two
nature of some reservoirs, however, conclusions homogeneous layers of contrasting permeability (k),
based on heterogeneities randomly distributed are porosity (cI, and thickness (h). From Fig. 1 the
less satisfactory than those that account for reservoir relevant quantities are II
layering. It is one of the purposes of this paper to use
stratification to explain large field-measured values Fk = kllk2' permeability contrast,
of c<:/'. F </> cI> II cl>2' porosity contrast,
Fh hl/(h l +h 2), layer-thickness fraction,
Approach F </>h cl>1 hi 1cl>2h2' porosity-thickness contrast,
Our approach is to explore the very close analogy and
between transverse dispersion in a two-layer medium Fkh k I'h l Ik2h2' permeability-thickness con-
and molecular diffusion in a capillary tube. The trast,
latter problem was presented in a classical set of
papers by Taylor 8 ,9 dealing with the effect that where the Subscripts 1 and 2 refer to high and low
molecular diffusion transverse to the bulk fluid velocity layers, respectively, and Fk/F</> is always
velocity has on the average concentration of solvent greater than one. Note that because the fluids have
slowly flowing through a tube. Basically, Taylor equal densities it does not matter which layer is on
showed that if a certain criterion is met, transverse top. Only three of these quantities are independent.
diffusion and longitudinal bulk flow will manifest We realize that very few actual displacements
themselves as a longitudinal diffusion phenomenon, satisfy these assumptions, particularly Assumption 1.
called Taylor's diffusion. It would seem that a This paper's objective, however, is to describe a
similar effect also might exist in two-layer porous phenomenon that will be present to some extent in all
media. Therefore, it is the purpose of this paper to displacements; hence, we invoke these simplifying
present the criterion under which Taylor's dispersion assumptions so that we may isolate this effect from
will apply and to describe the resulting effective those caused by unequal densities, mobilities, etc.
(Fig.2C).
Fig. 2 - Schematic illustration of Taylor's dispersion in
Limits. Fig. 2 shows the limits toward which two layer porous media.
transverse dispersion will drive a miscible
displacement in two-layer media. When transverse
dispersion is small, the medium behaves as a layered dimension driven by dispersion, and n t and nf are
heterogeneous medium; when it is large, the medium fluxes in the longitudinal and transverse directions,
behaves as a single-layer medium with increased respectively:
dispersion.
These observations suggest the following as a n fj = Vj C - K fj ( ~~) . . . . . . . . . . . . . . . . . . . (6)
transverse dispersion index.
and
- (QD) C=O.5 - (QDi)C=O.5
I TD = , ......... (3)
1.0- (QDi)C=O.5 ntj = -Ktj(~~)' ...................... (7)
where (QDi) C=O.5 is the cumulative injection of a
nondispersing displacement when the effluent From Eqs. 1 and 2, K f and K t are functions of v
concentration is 0.5, and (QD) C=O.5 is the which is, in turn, dependent on the system's
cumulative injection of a dispersing displacement at heterogeneity:
the same effluent concentration. (QDi) C=O.5 is ki cf>_
VJ. = --"-
rPj . k v ........................ . (8)
given by -0.
(QDi)C=0.5=
Fkh
----'-=---
+1 . F<f>
. (4a) since the displacement is at unit mobility ratio. By
(F<f>h + 1) Fk Assumption 6, the second term in Eq. 6 is negligible
For Fkh > 1, and by and these equations may be written in normalized
form
F kh + 1
(QDi) C=O.5 = F . . . . . . . . . . . . . . . . (4b) nfj = VjC . . . . . . . . . . . . . . . . . . . . . . . . . . . . (9)
<f>h+ I
for Fkh < 1. The quantities given in Eqs. 4a and 4b and
are the cumulative injection at breakthrough of the
high- and low-velocity layers, respectively. ntj =- K;; (~~), . . . . . . . . . . . . . . . . . . . (10)
The quantity (QD) C=O.5 is intended to be a
measure of the breakthrough cumulative injection of whereYD =y/H.
a displacement in which transverse dispersion is Putting these into Eq. 5 gives
present. When the system behaves as a single 1ayer,
[(QD)C=O.5]=I and ITD=1. When transverse N TD =( ~)( - ~~)(aC~YD) . ....... (11)
dispersion is small, ITD=O since (QD)C=O.5=
(QDik=o.5 The magnitude of the last factor in Eq. 11 is unclear;
but when the terms multiplying this factor are large,
Transverse Dispersion Number. Fig. 2 anticipated
N TD also generally will be large. In the scaled form
that transverse dispersion may be characterized by a
of Eq. 11, the derivative ac/aYD is of order C;
suitably defined number, N TD' Following Taylor 8
therefore, we omit the last term from the definition
tf L nt of N TD and compensate for the omission by selecting
N TD = - = - - , .................. (5)
tt H nf the remaining terms to give a minimum value of
N TD :
where t f is the time required for fluid to cross the L min (K tj ) _ L Kt2
medium longitudinally driven by bulk flow, t t is the N TD = H2 max ( Vj )
- H2 -VI . . . . . . . . (12)
time required for fluid to cross the transverse
AUGUST 1981 461
The reasoning for this is that if K(2 IVI will give a
large N TD , any other combination (K IV2' for
h
example) will give an even larger N TD . This
definition is analogous to that derived by Taylor
except for a numerical factor that is included below.
The transverse dispersion number, therefore, using
Eq. 2, is
N TD = (HL )2 [Do
LVIFcf>2 + L
Ci( v2 ]
~ ....... (13)
~ LOLOLOLOLO LOLO LOLO LOLO
Frequently, it is a good approximation to ignore
~~~~~~~~~~~~~~~~
.,000000000000000
,,-000000000000000
molecular diffusion; using Eq. 8 to eliminate the
<: C! C! C! C! C! C! C! C! C! C! C! C! C! C! C! velocity ratio in Eq. 13 yields
~OOOOOOOOOOOOOOO
2 2'1 QD (N pe ) - J 08
DISPERSION
(N",~=O.OOO475
----
where
Ke 1vL,
04
porosity-thickness weighted average,
and EXPliCIT FINITE DIFFERENCE
SOlUTI()\j
K t2t
D.
and 18. In all cases, when NTD in Table 1 is greater SIMULATION RESULTS
!TABLE I)
02
than five the agreement is good. Moreover, the ef-
fective dispersion is usually much larger than the
homogeneous longitudinal dispersion coefficient. 16
2
16 1 '0' '0' 0'
3 II 13 15
Fig. 5 shows the sensitivity of Ole to three of its F IF
constituent parameters: F h, FkIF</>, and H. Clearly, o 02 04 06 08
k "
~~~~~
~
~ ~ ~!:!~
LAYER k 3,5,1,4,1,2,5,2,1,4
(HOMOGENEOJS POROSITY)
rI
08
ANALYTIC,
EFFECTIVE DISPERSION--
EXPLICIT FINITE DIFFERENCE
(RUN 14,TABlE I)
10 LAYERS, I FT THICK
(7~~) 1750 1483 18.54 0141 188 1750 1483 1&54 0741 8.9 06 + NUMERICAL DISPERSION
( 3m)
(N~)~~T=00027
1180 263
04
0.636 7.95 D318 597
07 08 09 10 II I2 13
PORE VOLUMES INJECTED
Fig. 6 - Schematic of grouping procedure. Fig. 7 - Effluent history from 10layer porous media.
from 0.5 to 11.7 ft (0.15 to 3.6 m). For the first (one '0 EQUAL POROSITY
r
01 10 01 10 OOr~.:n..".~,..;.; 10
'
often-speculated dependence of Ci on reservoir
heterogeneities.
Field-Scale Mixing. Using the grouping procedure,
it is possible to deduce the effect of system length on
44ft
reservoir mixing. A measure of reservoir mixing is (134m)
L
the maximum relative mixing zone length, M/ L,
defined as the fraction of total system length between
C=O.1 and C=0.9 at 1 PV cumulative injection:
L'IOOfT(305m) L,SOOtl(152Sm) l'IOOOft (305m)
MIL = 3.625--./ Ci e / L. . ................. (27) (AVElll/L'026) (AVE tJ/L,OI7)
(HOMCGENEOUS lJ/l=026) (HOM:X>ENEOOS M/L=O II)
lAVE llt!L=OI9)
(HOMOGENEOUS M/L:OOB) (HOMCX3ENEOUS Mtt.:004)
o- MIXING ZONE LENGTH IN EXCESS OF I-OAOGEt'OOS SYSTEM
2](a~~)'
16. Lantz, R.B.: "Qualitative Evaluation of Numerical Diffusion
(Truncation Error)," Soc. Pet. Eng. J. (Sept. 1971) 315-320. .( v; -I)[YD2 -2YD + I-Fh
17. Todd, M.R. and Chase, C.A.: "A Numerical Simulator for
Predicting Chemical Flood Performance," paper SPE 7687 I-Fh 5,YD5,I
presented at the SPE Symposium on Reservoir Simulation,
Denver, Jan. 31-Feb. 2,1979.
18. Lake, L.W., Johnston, J.R., and Stegemeier, G.L.: and
"Simulation and Performance Prediction of a Large-Scale
Surfactant/Polymer Project," paper SPE 7471 presented at
the SPE 53rd Annual Technical Conference and Exhibition, c=C[ + ~ [ vL ](~)2 (V~ -I)YD2
Houston, Oct. 1-3, 1978. YD=O 2 K L v
t2
APPENDIX
.( a:: ), 05,YD < I-Fh . ............. (A-4)
Derivation of Effective Longitudinal aXD
Dispersion Coefficient Now C must be continuous at Y D = 1 - F h' Using this
Our approach is first to derive a transverse con- condition, the first part of Eq. A-4 becomes
centration profile that we then use to compute the
total longitudinal convective flux. This flux, when
substituted in the one-dimensional material balance, c=C[ _ + ~(~)(~)2(V~ -1)
results in a diffusion equation whose coefficient YD-O 2 K L v
t2
determines the effective dispersion.
Eq. 15, with longitudinal dispersion neglected and '(1-Fh )2( a:: )+ ~2 ( KtlvL )( ~
aXD L
)2
with x transformed to the moving dimensionless
x
coordinate D' where
A x vf .( v; -I)~D2 -2YD + I-Fh 2)
XD=L-L=XD-QD ............... (A-I)
ac )f( v2
_ WHvi!> ( H
1Jc- - 6 - L )2 ( aXD -;] -1
)2 (I-Fh )3
( 2)[
. -=-
- VL] +3 (V2
Kt2
--=- -1 )(Vl
v
--=--1 )
v
These substituted into Eq. A-7 give Ke=v(L)(Nh)-1 +1([, ............. (A-I4)