1996.Peters.A Look at Dispersion in Porous Media Through Computed Tomography Imaging

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Journal of Petroleum Science and Engineering 15 (1996) 23-31

A look at dispersion in porous media through computed


tomography imaging
Ekwere J. Peters *, Ridha Gharbi, Nadeem Afzal
Department of Petroleum and Geosystems Engineering, The University of Texas at Austin, Austin, TX 78712, USA

Received 20 August 1994; accepted 15 May 1995

Abstract

This paper presents a method for measuring the longitudinal dispersion coefficient and the adsorption properties of
porous media by X-ray computed tomography (CT) imaging. The in situ solvent concentration profiles of tracer tests were
imaged and fitted to a convection-dispersion mathematical model. In addition to providing the average dispersion coefficient
traditionally measured with breakthrough curves, our method allows the dispersion phenomenon to be visualized and the
contribution of heterogeneity to the dispersion coefficient to be estimated. The method is demonstrated by measuring the
dispersion coefficients and adsorption properties of a sandpack and a Berea sandstone. Results show that heterogeneity
increases the dispersion coefficient above that of a homogeneous porous medium. Adsorption was found to be higher in the
Berea sandstone, a natural porous medium, than in the clean sandpack. CT imaging of the tracer tests has shed more light on
the dispersion and adsorption phenomena in porous media.

1. Introduction dispersion and the attendant problem of measuring


the dispersion coefficient for the porous medium.
When a miscible fluid displaces another in a Traditionally, the dispersion coefficient is usually
porous medium, the displacing fluid tends to spread measured in the laboratory by monitoring the efflu-
and mix with the displaced fluid as the displacement ent solvent concentration for a tracer test and then
progresses. This mixing and spreading is usually calculating the longitudinal dispersion coefficient
referred to as hydrodynamic dispersion. Dispersion is from the breakthrough curve using a convection-dis-
important in the transport of contaminant plumes in persion model (Blackwell et al., 1959; Bringham et
groundwater and in the efficiency of miscible dis- al., 1961; Blackwell, 1962). The average dispersion
placements for improved oil recovery from petroleum coefficient so obtained includes the contribution of
reservoirs. gross heterogeneity to the dispersion coefficient.
One of the fundamental issues in modeling misci- In this paper, we present an improved method for
ble displacements is the quantitative description of measuring the longitudinal dispersion coefficient of a
porous medium by imaging of a tracer test in the
medium. Imaging the tracer test allows the contribu-
* Corresponding author. tion of heterogeneity to the dispersion phenomenon

0920-4105/96/$15.00 Published by Elsevier Science B.V.


SD1 0920-4105(95)00054-2
24 E.J. Peters et al./ Journal of Petroleum Scw~r and Engineering 15 (1996) 23-31

to be evaluated. It also allows the adsorption proper- medium; L is the length of the porous medium; and
ties of the medium to be estimated. We demonstrate U is the constant solvent injection flux. Therefore,
the method by determining the longitudinal disper- the interstitial velocity is given by:
sion coefficients and the retardation factors for a
14 kAP U
sandpack and a Berea sandstone. _=_
(5)
;=;w 4

Eq. 5 can then be substituted into Eq. 3 to describe


2. Theory the longitudinal dispersion of the solvent in the
porous medium. To solve Eq. 3 for the case of
In order to focus attention on the dispersion phe- continuous injection of the solvent, we apply the
nomenon, we consider a tracer test consisting of a following initial and boundary conditions:
stable, first-contact miscible displacement of two
C( X,0) = 0, X> 0 (6)
incompressible fluids having equal viscosities and
densities in a homogeneous porous medium. For C(O,t) =c,, t20 (7)
such a displacement, viscous and gravity instabilities
C(x,t) = 0, t2 0 (8)
are suppressed and only dispersion will manifest
itself. The mathematical model for this displacement The analytical solution to Eq. 3 for the initial and
in one dimension consists of the continuity equation, boundary conditions given by Eqs. 6-8 is (Aronof-
Darcy’s law and the convection-dispersion equation: sky and Heller, 1957; Ogata and Banks, 1961):
au ut

I:
-= 0
ax (‘1 X-G
C( x,t) = ; erfc
k aP 2\1’(
K,/R, >t
u=--- (2)
IJ- ax
ac u ac K, a*c
_-= / Ut

11
0 (3) .I--
-G+---
$4 ax R,- ax2
lz- 1erfc ! ($4
1:x

In the above equations, u is the superficial velocity


+exp 21°C
K./R,) t
L
(9)
(Darcy veloc’tI y ) and R, is a retardation factor that
accounts for the adsorption of the tracer by the
where erfc is the complementary error function, an
porous medium. If there is no adsorption of the
integral that is tabulated in mathematical handbooks.
tracer by the porous medium, the retardation factor is
An approximate analytical solution normally used to
unity whereas if there is adsorption, the retardation
determine the dispersion coefficient from break-
factor is greater than unity. It can be seen from Eq. 3
through data is:
that the effect of the retardation factor is to reduce u
and the K, for the displacement. Thus, the speed of

I_
the solvent concentration is retarded by adsorption.
The retardation factor is the ratio of the injection C( XJ) = ; erfc (10)
interstitial velocity and the retarded velocity of the
solvent concentration profile.
For a constant rate injection, Eqs. 1 and 2 lead to
In dimensionless form, Eq. 10 becomes:
the following solution for the superficial velocity:

k AP
u=----_u C(x,,t,) = ~[erfc{~~[XD~~~‘) )‘i]
(4)
P- L

where a P is the pressure drop across the porous (11)


E.J. Peters et al./ Journal of Petroleum Science and Engineering 15 (1996) 23-31 25

where the dimensionless variables are defined as the formation electrical resistivity factor; $ is the
follows: porosity; 01 is a mixing coefficient that characterizes
x the heterogeneity of the porous medium; d, is the
X D=- average particle size of the porous medium; u is the
L
interstitial velocity (U/+); and A is an exponent
between 1 and 2 depending on the lithology of the
(13)
porous medium. Eq. 18 relates a dimensionless lon-
gitudinal dispersion coefficient to a P&let number
(14) based on the mean grain diameter of the porous
medium and the binary diffusion coefficient of the
Eq. 14 defines a P&let number based on the total bulk fluids without a porous medium. The first term
length of the porous medium which is the ratio of on the right-hand side of Eq. 18 is the molecular
convective to dispersive transports. Eq. 11 suggests a diffusion term, whereas the second term is the me-
self-similarity transformation variable for first-con- chanical or convective dispersion term. At low P&let
tact miscible displacement of the form: numbers, the molecular diffusion term dominates the
dispersion behavior whereas at high P&let numbers,
the mechanical or convective term dominates the
dispersion behavior.
(1%
Blackwell et al. (19.59) and Perkins and Johnston

vRf (1963) have presented correlations for dimensionless


longitudinal dispersion coefficient versus P&let
Defining a mixing zone length as the distance be- number obtained experimentally by various authors.
tween C = 0.1 and C = 0.9, it can be shown from These correlations show that at P&let numbers less
Eq. 11 that the growth of the mixing zone is given in than 0.02, molecular diffusion dominates the disper-
dimensionless form by (Lake, 1989): sion behavior and the dimensionless dispersion coef-
ficient becomes constant and equal to (l/F+). For
t,
Ax, = 3.625 - unconsolidated media, (l/F+) was found to be about
NPe R, 0.67. Bringham et al. (1961) point out that (l/F+)
or in dimensional form by should vary from 0.15 to 0.7, depending on the
lithology of the medium. At Peclet numbers between
0.02 and 6, both molecular diffusion and mechanical
Ax=3625 K,t (17)
Rf or convective dispersion contribute to the dispersion
J
coefficient and Eq. 18 applies in full. At P&let
Thus, by measuring the length of the mixing zone as numbers above 6, the dispersion coefficient is domi-
a function of time, Eq. 17 can be used to calculate nated by mechanical or convective dispersion and
the longitudinal dispersion coefficient for the porous the effect of molecular diffusion can be neglected.
medium. The length of the mixing zone can easily be Thus, at sufficiently high P&let numbers, Eq. 18
measured by imaging the experiment. reduces to:
Theoretical and experimental studies have shown
that the longitudinal dispersion coefficient consists
of a molecular diffusion term and a mechanical or (19)
convective dispersion term that can be expressed in
an equation of the form (Bringham et al., 1961; Bringham et al. (1961) have presented experimental
Perkins and Johnston, 1963; Bear, 1972): data in which CL varied from 0.69 for a 0.044-mm
glass beadpack to 53 for a Berea sandstone. They
(18) also found an average value of 1.2 for A for glass
beadpacks. Blackwell et al. (1959) report a value of
where Do is the molecular diffusion coefficient; F is 1.17 for X for unconsolidated sandpacks.
26 E.J. Peters et al./Journal ofPetroleum Science and Engineering 15 (19961 23-31

In the groundwater literature (Bear, 1972; 3. Experiments


Domenico and Schwartz, 1990) it is customary to
approximate Eq. 19 as: Traditionally, the longitudinal dispersion coeffi-
cient is normally determined by measuring the sol-
K, = (YLC (20)
vent concentration at the outlet end of the porous
where (Y,_ is termed the longitudinal dispersivity of medium (breakthrough curve) for a tracer test and
the porous medium. Longitudinal dispersivity is re- then applying Eq. I1 at the outlet end (xu = 1) to
garded as a primary petrophysical property of the calculate K, or more correctly, KJR, if retarda-
porous medium that quantifies the mechanical or tion is not explicitly accounted for. This method
convective dispersion in the medium. Laboratory and gives an average K, that includes the effects of
field measurements show longitudinal dispersivity to heterogeneity and adsorption on the dispersion coef-
be highly scale-dependent, ranging from < 1 cm at ficient. We present herein a method of determining
laboratory scale to > 10 km at field scale (Arya et K, that allows the contribution of heterogeneity to
al., 1988). the dispersion coefficient to be evaluated. This is
Assuming a linear sorption isotherm, the retarda- accomplished by imaging the tracer test experiment
tion factor is related to the distribution coefficient, in time and space. We use Eq. 11 to determine the
K,, as (Domenico and Schwartz, 1990): average K,~ (that includes the effect of heterogene-
ity) and R, by history-matching the average concen-
(21) tration profiles and use Eq. 17 to determine the
component of K,, without heterogeneity by measur-
where pg is the grain density of the porous medium; ing the growth of the mixing zone length with time
and I:, is the retarded velocity of the solvent concen- from the image data.
tration profile (at C = 0.5). The distribution coeffi- To demonstrate our method, two tracer tests were
cient can be calculated from Eq. 21. performed and imaged by CT. The first test was in

Table I
Experimental conditions for tracer tests
Experiment I Experiment 2

Porous medium:

Type unconsolidated sandpack Berea sandstone


Length (cm) 54.2 60.2
Diameter (cm) 4.8 s.1
Absolute permeability (darcies) 6.4 0.160
Average porosity from CT (c/o) 29.7 17.3

Displacing fluid distilled water + 13% NaCl distilled water + 10% Nat
Density of displacing fluid (g/cm’) 1.089 1.078
Viscosity of displacing fluid (mPa s) 1.262 1.029
Displaced fluid distilled water + 10% BaCl, distilled water + I .4% NaCl + IO%KCI
Density of displaced fluid (g/cm’) 1.089 I .078
Viscosity of displaced fluid (mPa s) 1.127 I.028

Viscosity Ratio 0.9 1.0


Darcy velocity (cm/s) 3.037 x lo-’ 2.742 X 10-s
Interstitial velocity (cm/s) 1.023 x lo-’ 1.714 x IO-’
Breakthrough recovery (%o) 95.0 84.4
E.J. Peters et al./ Journal of Petroleum Science and Engineering 15 (1996) 23-31 27

- Experiment
an unconsolidated sandpack whereas the second test
‘-- Calculated
was in a consolidated Berea sandstone. In the tracer
tests, brine containing an X-ray contrast agent was
used to displace or was displaced by another brine of
the same viscosity and density. The experiments p 0.8
were designed to approximate a one-dimensional d
displacement in accordance with the theoretical 5 0.6
derivations given above. Table 1 shows the pertinent
H
experimental parameters. The procedure for CT
8
0.4
imaging of a coreflood has been presented by Peters
s
and Hardham (1990).
E
2; 0.2
w

4. Results and discussion


0.0
0.0 0.2 0.4 0.6 0.8 1.0
Fig. 1 shows the solvent concentration images for
DIMENSIONLESS DISTANCE
the tracer test in the sandpack at 0.2, 0.5 and 0.9
Fig. 2. Experimental versus computed solvent concentration pro-
pore volumes injected. The images show a vertical
files for tracer test in a sandpack. K, = 100X 10m5 cm*/s,
R, = 1.00.

slice through the center of the sandpack. The growth


of the mixing zone with distance or injection time is
apparent. The distortion in the mixing zone is caused
by inhomogeneities in the sandpack. Such distortions
or heterogeneities serve to increase the average dis-
persion coefficient measured by the traditional break-
through curve method.
Fig. 2 compares the experimental and calculated
solvent concentration profiles based on Eq. 11 using
an average longitudinal dispersion coefficient of 100
X 10m5 cm’/s and a retardation factor of unity. The
agreement between the experimental and calculated
profiles is good at early times but poor at late times.
At late times, the calculated profiles traveled farther
than the experimental profiles. However, the experi-
mental and calculated profiles are essentially parallel
at late times, indicating that the average longitudinal
dispersion coefficient is correct but the retardation
factor of unity is incorrect. Fig. 3 compares the
experimental and calculated profiles with the same
average dispersion coefficient but with a retardation
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.6 0.9 1.0
saturation factor of 1.04. The agreement between the experi-
ment and Eq. 11 is good at all times. It should be
C noted that the dispersion coefficient estimated from
Fig. 1. Solvent concentration images for a tracer test in a sandpack
the average solvent concentration profiles contains
at: (A) 0.2 pore volume injected; (B) 0.5 pore volume injected; the effect of heterogeneity in the sandpack and is
and (C) 0.9 pore volume injected. equivalent to the dispersion coefficient that would be
28 E.J. Peters et al./ Journal of Petroleum Science and Engineering 15 (19961 23-31

Experiment
*“-----. Calculated

0.0 0.2 0.4 06 06 1.0

DIMENSIONLESS DISTANCE

Fig. 3. Experimental versus computed solvent concentration pro-


files for tracer test in a sandpack. K, = 100X 10-’ cm’/s,
R, = 1.04.

obtained from the breakthrough curve. The average


0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.P 1.0
dispersivity for the sandpack was calculated to be 6AtUrAtiOn
0.098 cm using Eq. 20.
C
Fig. 4 shows the growth of the mixing zone
Fig. 5. Solvent concentration images for a tracer test in a Berea
length with time for the sandpack experiment. The
sandstone at: (A) 0.2 pore volume injected; (B) 0.5 pore volume
average mixing zone length at each time was mea- injected: and (Cl 0.8 pore volume injected.
sured from the three-dimensional CT images of the
tracer test. Thus, the effect of the distortion of the
mixing zone caused by heterogeneity in the sandpack was excluded from the mixing zone length. It can be
seen that the mixing zone grows linearly with the
square root of time as predicted by Eq. 16 or Eq. 17.
y = 0.72204 +0.10156x R”Z = 0.993 From the slope of the straight line of Fig. 4, K,/R,
was calculated to be 78.5 X lo-’ cm’/s. Thus, the
dispersion coefficient without the effect of hetero-
geneity in the packing is 82 X 10 -’ cm2/s. There-
fore, heterogeneity accounts for about 18% of the
total dispersivity of the sandpack.
Fig. 5 shows the solvent concentration images for
the tracer test in the Berea sandstone at 0.2, 0.5 and
0.8 pore volumes injected. As in the sandpack, the
growth of the mixing zone with distance or injection
time is apparent. The distortion in the mixing zone is
caused by heterogeneity in the sandstone. The lower
half of the sandstone was more permeable than the
lTIME (s’q upper half.
Fig. 4. Growth of mixing zone length with time for a tracer test in Fig. 6 compares the experimental and calculated
a sandpack. solvent concentration profiles based on Eq. 11 using
E.J. Peters et al./ Journal of Petroleum Science and Engineering 15 (1996) 23-31 29

- Experiment - Experiment
*-- Calculated -- Calculated

0.0
0.0 02 0.4 0.6 0.6 1 .o 0.0 0.2 0.4 0.6 06

DIMENSIONLESS DISTANCE DIMENSIONLESS DISTANCE

Fig. 6. Experimental versus computed solvent concentration pro- Fig. 7. Experimental versus computed solvent concentration pro-
files for tracer test in a Berea sandstone. K, = 600 X 10-s cm2/s, files for tracer test in a Berea sandstone. K, = 600X 10-s cm2/s,
R, = 1.OO. R, = 1.11.

an average dispersion coefficient of 600 X lop5 length with time for the Berea sandstone experiment.
cm*/s and a retardation factor of unity. It is seen It can be seen that the mixing zone grows linearly
that the calculated profiles travel farther than the with the square root of time as predicted by Eq. 16
experimental profiles at all times, the separation of or Eq. 17. From the slope of the straight line of Fig.
the two profiles increasing with time. The results 8, K,/R, was calculated to be 388 X lo-” cm*/s.
indicate a satisfactory average dispersion coefficient Thus, the dispersion coefficient without the effect of
but an incorrect retardation factor. Fig. 7 compares heterogeneity in the porous medium is 431 X 10m5
the experimental and calculated profiles with the cm*/s. Therefore, heterogeneity accounts for about
same average dispersion coefficient but with a retar- 28% of the total dispersivity of the sandstone.
dation factor of 1.11. The agreement between the Table 2 summarizes the results for the sandpack
experiment and Eq. 11 is excellent at all times. The and the Berea sandstone. As may be expected, the
average dispersivity for the Berea sandstone was Berea sandstone which is a natural porous medium
found to be 0.379 cm. has a higher dispersion coefficient (dispersivity) and
Fig. 8 shows the growth of the mixing zone a higher retardation factor than the clean sandpack.

Table 2
Summary of results
Experiment 1 Experiment 2
Porous medium unconsolidated sandpack Berea sandstone
Longitudinal dispersion coefficient with heterogeneity (cm2/s) 100 x 10-s 600 x lo-’
Longitudinal dispersivity with heterogeneity (cm) 0.098 0.379
Longitudinal dispersion coefficient without heterogeneity (cm*/s) 82 x 1om5 431 x 10-s
Longitudinal dispersivity without heterogeneity (cm) 0.080 0.272
Distribution coefficient (cm3/g) 0.0057 0.0087
Retardation factor 1.04 1.11
P&let number 554 159
30 E.J. Peters et al. /Journal of Petroleum Science and Engineering 15 (19%) 23-31

ber in the sandpack experiment ( NPe = 554) than in


the sandstone experiment (N, = 159).

5. Concluding remarks

In this paper, we have addressed the problem of


describing the mixing or dispersion that occurs in
first-contact miscible displacements. A method based
on CT imaging has been presented to simultaneously
determine the longitudinal dispersion coefficient and
-20 30 40 50 60
the retardation factor for a porous medium. The
L

method allows the contribution of heterogeneity to


miE (s’q the dispersion coefficient to be estimated. The method
Fig. 8. Growth of mixing zone length with time for a tracer test in was demonstrated by determining the dispersion co-
a Berea sandstone. efficients (dispersivities) and retardation factors for
an unconsolidated sandpack and a consolidated Berea
sandstone. CT imaging of the tracer tests has shed
The effect of heterogeneities is to increase the aver- more light on the dispersion and adsorption phenom-
age dispersion coefficient over that which would be ena in porous media.
obtained in a homogeneous medium.
Fig. 9 shows the solvent concentration data at all
times for the two experiments plotted against the 6. Nomenclature
self-similarity variable. As expected from Eq. 11, the
data transform into unique dimensionless response c= solvent concentration
functions characteristic of the two miscible displace- c; = inlet solvent concentration
ments. The curve for the sandpack is steeper than for K, = longitudinal dispersion coefficient
the sandstone, a reflection of the higher P&let num- d, = average particle size
Do = molecular diffusion coefficient
F= formation electrical resistivity factor
K, = distribution coefficient
NPC= P&let number
Rf = retardation factor
t= time
t, = dimensionless time
Ll= Darcy velocity
u= constant solvent injection flux.
I’ = interstitial velocity
I?, = retarded velocity of solvent concentration
profile
x= longitudinal coordinate
xn = dimensionless longitudinal coordinate

Greek symbols:
X D - (tD/Rf)

diz (Y= mixing coefficient


Fig. 9. Similarity transformation of solvent concentration profiles (YL= longitudinal dispersivity
for tracer tests in a sandpack and a Berea sandstone. AP = pressure drop
E.J. Peters et al. / Journal of Petroleum Science and Engineering 15 (1996) 23-31 31

grain density Dispersion and reservoir heterogeneity. Sot. Pet. Eng. Reser-
Pp =
voir Eng. (Feb.), pp. 139-148.
+= porosity
Blackwell, R.J., 1962. Laboratory studies of microscopic disper-
lJ= viscosity
sion phenomena. Sot. Pet. Eng. J. (Mar.), 2: l-8.
A= characteristic exponent Blackwell, R.J., Rayne, J.R. and Terry, V.M., 1959. Factors
5= self-similarity variable Influencing the efficiency of miscible displacement. Trans.
Am. Inst. Min. Metall. Eng., 216: l-8.
Bringham, W.E., Reed, P.W. and Dew, J.N., 1961. Experiments
on mixing during miscible displacement in porous media. Sot.
Acknowledgements
Pet. Eng. J. (Mar.), 1: l-8.
Domenico, P.A. and Schwartz, F.W., 1990. Physical and Chemi-
This paper is based in part upon work supported cal Hydrogeology. Wiley, New York, NY.
by the Texas Advanced Research Program under Bear, J., 1972. Dynamics of Fluids in Porous Media. Elsevier,
Grant No. 259 and the U.S. Department of Energy New York, NY.
under Contract No. DE-AC22-90BC14650. The au- Lake, L.W., 1989. Enhanced Oil Recovery. Prentice Hall, Engle-
wood Cliffs, NJ.
thors are grateful for this support.
Ogata, A. and Banks, R.B., 1961. A solution of the differential
equation of longitudinal dispersion in porous media. U.S.
Geol. Surv. Prof. Pap. 411-A.
References Perkins, T.K. and Johnston O.C., 1963. A review of diffusion and
dispersion in porous media. Sot. Pet. Eng. J. (Mar.), 21:
Aronofsky, J.S. and Heller, J.P., 1957. A diffusion model to 70-80.
explain mixing of flowing miscible fluids in porous media. Peters, E.J. and Hardham, W.D., 1990. Visualization of fluid
Trans. Am. Inst. Min. Metall. Eng., 210: 345-349. displacements in porous media using computed tomography
Arya, A., Hewett, T.A., Larson, R.G. and Lake, L.W., 1988. imaging. J. Pet. Sci. Eng., 4: 155-168.

You might also like