,

Effects of Dispersion and Dead-End Pore Volume

in Miscible Flooding
L. E, BAKER
MEMBER SPE-AIME

ABSTRACT
The design of the solvent slug size /or a miscible
flood process
can be improved with data on holdup
(or capacitance
effects)
and dispersion
of the
solvent slug in the reservoir.
A modified version of
the Coats-Smith
dispersion-capacitance
mode I and
an improved
solution
method for the model were
used to study dispersion
avd capacitance
effects
in cores.
The velocity
dependence
of the model
parameters
is sbotvn. A correlation
is developed
for
estimating
effective
dispersion
coefficients
for
field application.
The method
described
provides
a means
for
characterizing
the properties
of dispersive
mixing
and micro heterogeneity
of reservoir cores and aids
in the design of the volume o~ solvent for miscible
floods.
INTRODUCTION
The amount of solvent that must be injected
is a
critical
factor in the success
of a miscible
flood.
Because
of the cost of miscible
solvents
such as
carbon dioxide or rich gas, slug processes
generally
are used. If the solvent
slug used is larger than
necessary,
the solvent
cost
will be increased
without compensatory
increases
in oil recovery.
If
too small a slug is used, some of the oil that could
have been moved will be left behind. The slug size
required
is affected
by many variables,
including
reservoir
geometry,
interwell
spacing,
gravity
effects,
mobility
ratios,
etc. Slug degradation
is
caused
by mixing (by dispersion)
of solvent
with
oil at the leading edge of the solvent bank and with
chase fluid (for example,
dry gas) at the trailing
edge. Trapping
of oil and solvent in microscopic
heterogeneities
(regions of dead-end pore volume or
relatively
stagnant
flow) also contributes
to the
mixing-zone
growth.
This
trapping,
known
as
capacitance,
may be caused
by rock heterogeneities 1 or by shielding
of oil and solvent by water
films. 293
Original manuscript received in Society of Petroleum Engineers
office July 23, 1975. Paper accepted for publication Nov. 21,
197S. Revised manuscript received Jan, 24, 1977. Paper (SPE
5632) was first presented at the SPf+A2ME SOth Annual Fall
Technical Conference and Exhibition, held in DaI1as, Sept. 28Oct. 1, 1975. (@ Copyright 1977 American Institute of Mining,
Metallurgical, and Petroleum Engineers, Inc.
2%1spaper will be included In the 1977 ?%msactfone
volume.
JUNE. 1977

AMOCO PF?ODUCT/ON CO.

TULSA-

This paper is concerned

with predicting
solvent
slug requ~rements
in an idealized
linear system
where gravity,
mobility
ratio,
and areal
sweep
effects
are unimportant,
but where longitudinal
dispersion
(mixing at the leading and trailing edges
of the bank) and capacitance
effects are significant.
An example might be a miscible
displacement
In
the pinnacle reef formations of Alberta. A prediction
of the effects
of dispersion
and capacitance
was
needed for the design of a miscible
flood of this
type. The oil-column
height was about 350 ft, and
the flood advance
rate was to be downward
at
0.0384 ft/D. The oil/solvent
viscosity
ratio of 10
was unfavorable;
however, it was expected that the
unfavorable
mobility
effects
would
be largely
compensated
for by the stability
effects of gravity
at the low flow rate.
Publish ed data4-6 relating
to similar reservoirs
volume that could cause
indicated
that stagnant
trapping and degradation
of the solvent slug mjght
be as much as 10 percent of the reservoir
volume,
calculations
were
Based Oi] these data, preliminary
made using the Coats-Smithl
dispersion-capacitance
model to predict
the mixing-zone
profiles.
The
results indicated
that this level of stagnant volume
might cause the solvent requirement
to be increased
by 30 to 90 percent over the amount predicted
by a
simple
dispersion
model
without
capacitance
effects
if the peak solvent
concentration
in the
enriched
gas bank did not drop below 99 percent
throughout the life of the flood.
Coats and Smithl indicated
that tests in short
cores
would
show
extended
mixing
zones
if
capacitance
effects
were present,
but that if the
magnitude
of the traq,$fer group MD = ML/u was
large (as it woulck be i; a field situation,
where L
may be very large),
the influence
of capacitance
wou Id be minimized.
The prediction
of a 30- to
90-percent
increase
iii solvent requirements
for the
case described
above prompted a review of methods
for measuring
capacitance
effects and a search for
a more convenient
method for predicting
the severity
effects
in field application.
An
of capacitance
improved method for modeling data from short core
and experimental
work was
tests
was developed,
performed to investigate
the factors influencing
the
capacitance-model
parameters.

219

DISCUSSION
CAPACITANCE

MODEL

Dispersion
during flow in porous media has been
studied
from both experimental
and theoretical
viewpoints.
Perkins
and Johnston7
and Greenkorn
and Kessler8
have reviewed much of the literature
on this
topic.
Of particular
interest
are the
the
effects
of
papersl~2~6~9- 13 dealing
with
heterogeneity
and capacitance
on the development
of mixing zones in porous media.
Capacitance
or stagnant
pore-volume
effects are
most apparent
when a concentration
step-change
miscible-displacement
test is performed. This is a
test in which a continuous
solvent bank is injected,
the in-situ fluid
beginning at time t = O, to displace
from a core. The simple dispersion
model characterized by Eq. 1,

K*IU

ac

ax2

JC

ac=ac
-ax%

.(1)

aX2

f%

+ (1-f)

g+

. . . . . . . . . . . . . . . . . (3a)
and

when solved
using
boundary
conditions,b
effluent concentration:
C = 0.5 erfc

The models proposed by Turnerg and Gottschlic~10

assumed a velocity-independent
transfer coefficient,
while the Coats-Smith
model does not impose this
limitation.
Brigham6 showed that Eq. 1 could describe either
the flowing (effluent)
or the in-situ concentration,
depending
on the boundary conditions
chosen for
solution of the equation.
Appendix A shows that a
similar choice of boundary conditions
determines
whether
the in-situ
or the effluent
concentration
profile is obtained from the Coats-Smith
capacitance
model.
The method described
in Appendix
A is
much simpler than the method proposed by Brigham6
to determine
the effluent
concentration
for the
capacitance
model, and results
in the following
equations:

the appropriate
initial
and
yields
a solution
for the

ac+
~=

~~-f)

to be solved

()

+ 0.5 e K) erfc

w
m

M(C-C+),

with the initial

. . . ..(3b)

and boundary

conditions

.(2)

()

c+

The concentration
profile predicted
by Eq, 2 is
nearly
symmetrical
around the PV = 1 point, as
shown in Fig. 1. Also shown in Fig. 1 IS a typical
effluent concentration
profile exhibiting capacitance
effects.
The asymmetric
profile
is attributed
to
fluid in regions
of stagnant
or
holdup of in-situ
very SIOW flow, with subsequent
bleeding
out of
the in-situ fluid as the mixing zone passes.
Several
models
incorporating
dispersion
and
have
been
proposed.
Each
model
capacitance
and convection
in a flowing
includes
dispersion
region of the core and fluid transfer between flowing
and stagnant
regions.
The model described
by
Coats
and Smith 1 was chosen
for the present
analysis.
Experimental
studies performed by Coats
and Smith 1 and by Goddard 12 indicated
the transfer
coefficient,
M, of the model was velocity-dependent,

(xJo) = o
(X,o) = o

c (Ott)

t)=o

c(oo,

@x>o.

o..

..

(4a)

@x>o

. . . . ..(4b)

@t>()

. . . . ..(4c)

Ant

. . . .

..(4cI)

These
equations
are derived
from the equations
obtained
by Brigham by means of a simple transformation of variables.
Eqs, 3a and 3b may be solved in the s (complex
frequency) domain by taking the Laplace transforms
of the equations
and combining
the resulting
ordinaty
differential
equations.
The solution
with
initial and boundary conditions
of Eqs, 4a through
4d is

1.00
,0

o.w

,/

IL
/

0.60

. . . . . . . . . . . . . . . . . . . . .(5)

OISPIRSION

--WITH

CAPACITANCE

Eq. 5 may be inverted to the time domain

use of the Cauchy Integral Theorem 14:

0.40

by the

If

0, xl

=+
c (x)t)

o. m

0.0

0.2

0.4

0.6

HYDROCARBON

FIG.

EXAMPLE

0.8

1.0

1.4

1.6

1.8

PORE VOLUMES OF FLUIO INJECTIO

EFFLUENT

PROFILES.

2m

i.?

~jw

~(c),

stds

. . . . . .(6)

-P

CONCENTRATION

2.0

the numerical
integration
involved
However,
quite time-consuming,
especially
when it must
done many times during the iterative
process
fitting the ~odel to the experimental
data.

is
be
of

SOCIETY OF PETROLEUM ENGINEERS JOURNAL

3- WAY VALVE

The form of Eq. 5 suggests

that, rather than
transforming
the model solution to the time domain,
the data could be transformed
into the frequency
domain, An algebraic
curve-fitting
procedure
then
could be used, reducing
the computing
time and
inaccuracies
introduced by the numerical integration
of Eq. 6. A method for transforming
time-domain
data to the frequency
domain and for fitting the
model is outlined in Appendix B. This method was
An additional
found
to be quite
satisfactory.
advantage
of this procedure
is that the frequency
response of the flow system can be obtained directly
from step,
impulse,
sinusoidal,
or many other
concentration
signals.
EXPERIMENTAL

BENZENE

D
PUMP

II

SAMPLE
COLLECTIONl!!!?
FIG,

Step-change
tests
of miscible
displacement
in
cores were made with benzene (p = 0.604 cp, p =
0.8734 gin/cc) and meta-xylene
(p = 0.584 cp, p =
0,8601 gin/cc).
The flow system is shown in Fig.
2.
A constant-rate
positive-displacement
pump
manifolded
to two fluid reservoirs
controlled
flow
rates at 1.45 x 10-5 to 9.88 x 10-3 cm/sec (0.041
to 28 ft/D). A three-way
valve at the core inlet
determined
which fluid was injected
into the core.
The pump maintained
equal pressure
on both fluid
reservoirs
to minimize pressure
surges when flow
was switched
from one fluid to the other, Volumes
of the core end-caps
and inlet and outlet tubing
were less than 1 percent of the core volume.
A Berea sandstone
core (chosen for its homogelimestone
core
(which
neity)
and
a vugular
preliminary
tests
indicated
was heterogeneous)
were used. The cores were sealed with epoxy resin
and reinforced
with Fiberglas
tape. The sandstone
core (Core 1) was 22,86 cm long and 5.o8 cm in
with a permeability
of 175 md and a
diameter,
core (Core
porosity of 21.4 percent. The limestone
2) was 9.4 cm long and 7.60 cm in diameter,
and
had a permeability
of 5,4 md and a porosity of 11.9
percent.
No brine saturation
was present
in the
cores.
The cores
were evacuated
before
being
saturated
with liquid.
The procedure
for each test was to saturate
the
core with one of the miscible
fluids and establish
a stable flow rate. The injection
valve was then
switched to the other fluid, and injection
continued
at the same rate as samples
of the effluent fluid
were taken
for analysis.
Concentration
profiles
were monitored by gas chromatographic
analysis
of
1 RUN

CORE

WATER

PROCEDURE

TABLE

!YLENE

SYSTEM.

FLOW

the effluent
samples.
The precision
of the sample
analysis
were ~ 1 percent.
For the initial
runs,
samples
were collected
manually.
For later tests,
an in-line gas chromatography with an automatic
sample valve was used. In
these later tests,
a backpressure
of JO psig was
maintained
at the sample-valve
outlet. Repeat tests
showed no effect of the change in backpressure
or
sampling procedure.
The run variables
are listed
in Table
1, and
example
effluent
profiles
are plotted
in Figs.
3
an; 4.
DISCUSSION

OF RESULTS

The capacitance
model was fitted to the data of
each run as described
in Appendix B, Fig. 3 shows
the experimental
data and the best-fit
calculated
points for Run 2. The data of Runs 1 through 3 (in
the 23-cm Berea core) were fitted very well by the
simple dispersion
model (to which the capacitance
model reduces when / = 1.0). This is illustrated
by
the straight-line
fit of the data of Run 2 on the
probability
plot 1s of Fig. 4.
Runs 4 through
10 were made v.ith a vugular
limestone
core at flow velocities
of 0.04 to 19.9
An
ft/D (1.45 x 10-5 to 7.02 x 10-3 cm/see).
example
concentration
profile and best-fit
calculated curve (for Run IO) are shown in Fig. 3. The
concentration
profile is also shown on the probability
plot of Fig. 4. The curvature of the data as shown
that some type of holdup
on this plot indicates
(capacitance)
occurred.
Fig.
5 shows
the dispersion
coefficients
for
Runs
1 through
10. The dispersion
coefficients
VARIABLES

Model Parameters
Run tire
.
11

21
31
4

52

62
72
82
9
10
J!.NE, 1977

2
2

Flow Rate

Displacing
Fluid

Displaced
Fluid

(CIII/S@C X 105)

(ft/D)

Mets-xylene
Benzene
Benzene
Mets-xylene
Benzene
Mets-xylene
Eerrzene
Benzene
Mets-xylene
Met&xylene

Benzene
Mets-xyl ene
Mets-xylene
Benzene
Mets-xylene
Benzene
Mete-xylene
Met&xylene
Benzene
Benzene

104
490
988
1.45
14.7
30.6
31.9
109
123
702

2,95
13.9
28.0
0,047

0,416
0.668
0.803
3.08
3.49
19,9

(m

Cn&c

19.3
157
301
1,36

28.6
85.3
46.4
225
174
1,590

105)

(s@c-lX 106)

1.000
1.000
1,000

1,80
5,46
13,8
20,0
25.5
53.6
222

0.363

0.670
0.404
0.562
0.691
0.579
0.666

221

increase
with velocity;
the line shown connecting
the points for the limestone
core has a slope of 1.0
and that for the sandstone
core has a slope of 1.25.
This indicates
that the correlations
suggested
for
homogeneous
cores (Eq. 7) is also valid for cores
exhibiting
capacitance
effects:
~ , ~l.

oto 2,0 . . . . . . . . . .(7)

Fig. 5 also shows the transfer

coefficients,
M,
computed for Runs 4 through 10. The data indicate
a dependence
of M on velocity
to the 0.84 power,
Coats and Smith* also noted a possib!e
velocity
variation
of M, as did Goddard. 12 If fluid transfer
diffusion
alone,
M
was dependent
on transverse
should
be independent
of velocity.
The velocity
that
the
transfer
dependence
of M confirms
mechanism
involves
mixing other than diffusion,
The flowing fractions
(/) for Runs 4 through 10
are shown in Fig. 6. A large amount of scatter in
the data is evident;
/ values range from 0.383 to
0.691, with no apparent velocity correlation.
If the
two lowest values are disregarded,
the other data
points are fit fairly well by an average value of / =
0.64. Some variation
of / with velocity
might be
expected;
however, the data of Fig. 6 do not justify
any conclusion
except
that the flowing fraction
appears to be constant
in these runs for velocities
greater than 10-4 cm/sec (O. 28 ft/D).

APPLICATION

The number of data points that can be used in

matching
the capacitance
model to experimental
data may be limited by the repetitive
and timeconsuming
numerical
integration
of the inverse
Laplace
transform
(Eq. 6). The frequency-domain
curve-fitting
procedure outlined in Appendix B does
not require the inversion
of the Laplace
transform;
thus, for a given amount of computing
time, the
frequency-domain
fit allows the use of many more
data points. The result is a more precise fit of the
model to the data.
Application
of the capacitance
model to prediction
of solvent concentration
profiles in field application
also can be simplified.
Stalkup2 and Brigham G have
shown that for reservoir-scale
systems,
the concentration curves predicted
by the capacitance
model
(Eqs. 3a and 3b) approach the symmetrical
profile
predicted
by the simple dispersion
model (Eq. 1).
The apparent
or effective
dispersion
coefficient
(Ke), however, is greater than the true dispersion
coefficient
(K). To confirm this, mixing-zone lengths
(the distance between the IO- and 90-percent solvent
concentration
points for a step-change
test) were
calculated
(using
the
capacitance
model)
for
several
values of K and M at 100 and 1,000 ft of
For each set of parameters,
the
flood advance.

,0-2

l.al ~

-r-@--w+-!

..-

1 DISPERSION
2 OISPIRSION

&

m 0)S3
X 03SS

CORE 2 MASS

TRANSFER

m
@

x
*

~o-i
&
10+

a
1~

10+

~
JJ

o.IxJ
0.0

1.

0.5

.L

..-.

1.5

1.0

..1.

20

25

FOR[ vOLUMIS

FIG.

3 EFFLUENT

Zm

l_

3.0

4.0

3.5

lNJftl[O,

4.5

CONCENTRATION

10-3

,0-2

VELOCITY, WSEC

FIG. 5 VELOCITY
DEPENDENCE
OF THE DIS.
PERSION
COEFFICIENT
AND MASS -TRANSFER
COEFFICIENT.

:r-

10

A RLN

PROFILES.

mRW2

1.54

5.0

]~+
..

Lal [

I-1 o,~ .

+.30

0.50 -

~a.aemmm /
*
. *
*

~emeeao

-Lm -

flu

*S

0.al

I
~y

-1,54-.
-zoo -250 ,
0.010

,
0.100

,
O.m

0.6UI

awl

o. Wo

EFFUKNTCONCLNTR41IW

FIG,
222

4 EFFLUENT

CONCENTRATION

FIG,
PROFILES.

6VELOCITY

DEPENDENCE
FRACTION.

10-2

OF THE

FLOWING

SOCIETY OF PETROLEUM ENGINEERS JOURNAL

mixing-zone
length increased
as the square root of
the
distance
that
was
traveled.
The
simple
dispersion
model also predicts growth of the mixing
zone with the square root of distance
traveled;
thus, it appeared that it should be possible
to use
the simple dispersion
model (with an effective
dispersion
coefficient)
to predict the mixing-zone
growth for field applications.
To determine
the effects
of M and / on the
effective dispersion coefficient,
mixing-zone lengths
(10- to 90-percent
solvent concentration)
were calculated for a range of values of the dimensionless
parameters
NBO, MD, and /. The effective dispersion
coefficient,
K@, was defined by 15

LMZ 3.62S

MDINBO

LMZ2
()LMZr

for r3C+/dt
dispersion-

$-ug=%

(11)

This is analogous
to the simple dispersion
model
(Eq. 1), with an effective
dispersion
coefficient,
Ke, defined by
&
&

o-;(9)

as shown in Fig. 7. The curves

shown in this
figure are for values of NBO greater than 1,000, and
will overestimate
Ke for lower values of NBO.
It follows from Eq. 8 that the ratio Kc/K is also
the square of the ratio of the mixing-zone
lengths
with and without capacitance
effects:
Ke
T

expression
3a, the

W........(8)

For constant
values
of K, M, /, and u, the
coefficient
increased
asympeffective
dispersion
totically
to a limiting value, Klim, as NBO increased
(increasing
L), The limiting values,
expressed
as
K1im/K, were correlated
as a function of f and the
dimensionless
group
~=

#~+
,,+g.~
[1ax;..........

respect to t and the resulting

substituted
into
Eq.
is
capacitance
model becomes

. . . . . . . . . ..

(10)

for a given
dispersion
7 shows
that,
Fig.
coefficient
and velocity,
the mixing-zone
length is
less for a larger value of / (less stagnant
volume)
or for a larger value of the transfer coefficient.
The
range
of KMln2
values
shown
includes
those
measured
in the present work and those mentioned
in earlier papers. 1~2*4-G~12
The correlation
of Fig. 7 can be derived from
Eqs. ja and 3b. If Eq. 3b is differentiated
with

.l+~2

,..

&_
3X2

-(2

(12)

If we assume
that d2C/dx2
and d2&/dx
2 are
equal (at a large rravel distance,
L), the values of
Kc/K given by Eq. 12 agree within a few percent
with the values shown in Fig. 8 for Khf/u2 values
greater than 0.001. For smaller values of KM/u2,
the approximation
d% /dx2 = &C+/dx2
is apparently
invalid;
as M approaches
zero (no fluid
transfer between flowing and stagnant
regions),
the
effective
dispersion
coefficient
should
again
approach the true dispersion
coefficient.
Deansll
used a different
approach
to derive a
relation similar to Eq. 12, but limited in application
to small values of the group KM/u2. His correlation
predicts
thar Kc/K will approach
zero for large
values
of KM/u2 or for values
of / approaching
unity.
This correlation
shows that the solvent concentration is a function of the dimensionless
groups
NBO, KM/u2, and f, and that it is the group KM1u2
rdther than MD that determines
whether capacitance
effects will be negligible
in a reservoir application.
Thus, capacitance
effects
do not disappear
with
~~creasing system length, as implied by Coats and
~mith I and Stalkup. 2

t\\\x

-w

1?
0.001

a01

0.10

1.0

Y
FIG. 7 DEPENDENCE
PERSION
COEFFICIENT
COEFFICIENT
AND
JWE, 1977

OF THE EFFECTIVE
DISON THE MASS-TRANSFER
FLOWING FRACTION.

10-111

, , ,1
10-3

104
Ke,

FIG.

8 ERROR
EFFECTIVE

-i
1 , , I,

10-2

CM21SEC

IN MODEL FIT AS A FUNCTION

DISPERSION
COEFFICIENT.

OF

223

For the field example cited in the Introduction,

the dispersion
group NBO was about 12,250. The
transfer group MD was estimated
to range from 46
(Case
1) to 132 (Case 2). Concentration
profiles
predicted
by the
simple
dispersion
and
the
capacitance
models for these values,
and with a
of 0.9 (10 percent
{stagnant
flowing
fraction
volume), are shown in Fig. 9. The 10- to 90-percent
mixing zone is 11.2 ft long for the simple dispersion
case, increasing
to 15.45 ft (Case 1) or to 21.7 ft
(Case 2) with the inclusion
of capacitance
effects.
The same prediction
can be obtained
from Fig. 7;
at KM/u2 = 1,07 x 10-2 (Case 1), the Kc/K ratio is
1.96, corresponding
to a mixing-zone
length ratio of
w=
1.4. At KM/u2 = 3.74 x 10-3 (CaSe 2), the
Kc/K ratio is 3.61, corresponding
to a 90-percent
.
increase
in mixing-zone
length.
EFFECT

OF PARAMETER VALUES ON MODEL FIT

The correlation
of Eq. 12 suggests
that it may not
be possible
to specify unique K, M, and / values to
fit the data of any given run. If one of the values
is specified
arbitrarily,
the others can be chosen
to obtain any desired
value of Ke (within certain
limits). To investigate
this possibility,
the error in
fitting
the model to the experimental
data was
calculated
for a range of values of the parameters
(K and M values ranging from 0.1 to 10.0 times the
best-fit
values
found by the fitting
procedure
described
in Appendix B, and values of / ranging
f 0.2 from the best-fit
values).
The results
are
shown in Fig. 8 for a representative
case as a plot
of E (sum of squared
errors)
vs K@. Nea: the
there
best-fit region (Ke = 1.2 x 103 sq cm/see),
are several combinations
of K, M, and / that give
relatively
good fits. However, there are mar,y more
combinations
giving the same valuee of Ke that do
not fit the data well, Examination
of the error in fit
as a function of each of the variables
K, M, and /
shows that there is a relatively
narrow region for
each parameter within which values can be chosen
to produce acceptable
fits.

was developed
and used to model unit-viscosity
ratio, laboratory miscible-displacement
tests over a
wide
rc.nge of velocities.
The method
allows
frequency
response
data
to be obtained
with
nonsinusoidal
input signals.
The dispersion
coefficients,
K, and fluid transfer
coefficients,
M, for the dispersion-capacitance
model were found to be velocity-dependent,
while
the flowing fractions,
/, were relatively
independent
of velocity.
A correlation
was developed
for estimating
the
effective dispersion
coefficient
for field application
from the values
of the dispersion
coefficient,
transfer
coefficient,
and flowing fraction
derived
from short core tests. It was shown that / and the
dimensionless
group KM/u2 determine
the value of
the effective
dispersion
coefficient.
Application
of the capacitance
model to an
example
reservoir
shows
that
the exaggerated
tailing noted in laboratory
tests using short cores
is not as pronounced
in longer systems.
However,
mixing-zone
lengths and solvent require~n~s
may
be increased
significantly
by capacitance
effects
in reservoirs.
NOMENCLATURE

c.

{in-situ

c.

flowing

C*

h-=

A rapid and efficient

method for matching
a
dispersion-capacitance
model to experimental
data

dispersion

Klim

= limiting
uL/K)

I
350

360

9 PREDICTED

MIXING

ZONE

370

AT 350 FTi

B)

sq cm/sec
coefficient

of Ke (at

(defined

by

values

of

large

cm

transfer

seel
group,

ML/u
sol-

= mixing-zone
length
without
capacitance
effects, cm (reference
length)
integral number of cycles
Appendix B)

NBO & dimensionless

Bodenstein
%=
s=

DI STANCE, FE~

available
fraction

= mixing-zone
length ( 10- to 90-percent
vent concentration),
cm

n=

FIG.

Appendix

= Laplace transform of C
fluid transfer coefficient,

o.al -

224

value

core length,

LMZ,

I
Ml

stag-

in model fit

coefficient,

K= = effective dispersion
Eq. 8), sq cm/sec

LMZ

~
~

in

volume readily
is the stagnant

ratio (see

MD = dimensionless

a mm

errors

M=

am -

fluid

fluid

9= imaginary part of Laplace transform

/= pore volumes of fluid injected
j= W-

f?(c)
%

of injected

by C+ = C* - (K/u) (dC*/dx)

sum of squared

FA . ampIitude

L=

Lm

fluid

flttid at core inlet

concentration

fraction
of pore
for flow; (1 -/)

/=

of injected

of injected

concentration
of in jetted
nant region

c+ . defined

K=
SUMMARY

concentration

Cj = concentration

dispersion
number

real part of Laplace

Laplace

transform

of phase
group

lag (see
(uL/K);

transform
variable,

SOCIETY OF PETROLEUM ENGINEERS JOURNAL

t = time, sec
tn = time at which the output
returns to its initial value

tl = time at which the output concentration

reaches steady-state
value

tq = time at which the input concentration

reaches a steady-state
value

Ci

u = velocity,
distance

a =

proportionality

~ = phase

relative

p = density,

constant,

defined

Appendix

1964) 73-84

Trans., AIME,

Vol.

231.

Shelton,
J. L, and Schneider,
F. N.: The Effects
of
Water Injection
on Miscible
Flooding
Methods
Using
Hydrocarbons
and Carbon Dioxide, ~ Sot, Pet. Eng.
J. (June 1975) 217-226.

4.

AERCB
Texaco,

Application 4404, Supplementary Information,

Inc., Wizard Lake field (Oct. 1969, 197 1).

5.

AERCB
Application
Gas Co., 2ama-Virgo

6.

Brigham,

W. E.:

5065, Hudsons
Bay
field (Msrch 1970).

{tMixing
Equationa
Etz& ].

SO=. Pet.

Vol.

Oil

and

1974)

9.

Turner, G. A,: The Flow Structure in Packed Beds,

Cbetn. Ens Sci. (1958) 156-165.
Disperaiofl
in a Packed
Gottschlich,
C. F.: /lAxial
Jour.

(1963)

88-92.

Deana, H. A,: A Mathematical

Model for Dispersion
in the Direction
of Flow in Porous Media, Sot. Pet,
Eng. ]. (March 1963) 49-52; Tratis., AIME, Vol, 228,

12.

Goddard,
R. R.: Fluid
Diapersion
and Distribution
in Porous
Media
Using
the Frequency
Response
Method With a Radioactive
Tracer, Sot. Pet. Eng.

(June 1966) 143-152;

Trans.,

AIME,

Vol.

237.

Determination
of Diapersion
and
Goaa,
M. J.:
Diffusion
of Miscible
Liquids
in Porous
Media by a
Frequency
Response
Method, paper
SPE
3525
presented
at the SPE-AIME
46th Annual
Fall Meeting, New Orleans,
Oct. 3-6, 1971.
Complex Variables and ApplicaBook Co., Inc., New York (1960)

14.

Churchill,
R. V.:
tions, McGraw-Hill
118.

1 s.

Brigham,
W. IL, Reed,
P. W., and
~~ExPeriments on Mixtig During Miscible
Media,
Sot. Pet. Ens
8; Trans., AIME, Vol. 222.
in Porous

Clements,
Testing
Proc,

JUNS, 1S77

T.

R.:

AIChE

Jour. (1967)

W. C., Jr., and SchnelIe,

for Dynamic

Analysis,

.,

equation

. . ..(A-l)

at

b%

which describes
the in-situ concentration,
C, of a
fluid flowing through a porous medium, has a form
to that of the equation
identical

32C

_aJ

, ac

ax

TiT

.(A-2)
....

This equation describes

the flowing concentration,
C, which is equivalent
to the effluent or cup-mixing
concentration.
The two concentrations
are related
by

Dew,

J.

N.:

Displacement

J. (March

1961)

S ln~

& t%g.

.(A-3)

O, X=

C(X, O)=
t)=o,

C(o,t)

O . . . . . .

t=o

= I +:

. . . . . .
~

, t 2?0

..(A-~a)
..(i%dq

(A-4C)

The flowing
concentration
(which is the one
is
usually
measured
in laboratory
core tests)
obtained when the conditions
C(X, O)=O,

C(w,

t)=

C(O, t)=l,

x20,

(l,

t~(j

t=

. . . .

.( A-5a)

, . . . . .( A-5b)

O , . . . . .( A-5c)

1-

K. B., Jr.: rPulse

Design and Deu, (1963) 94-102.

ac

Whether the in-situ or the flowing concentration

is obtained upon solution of Eq. I or Eq. 2 depends
on the initial and boundary conditions
chosen.
The
in-situ concentration
is obtained when the following
conditions
are used:

C(m,

11.

16.

ac=ac ..

._

c=c-~~....

Greenkom,
R A, and Kessler,
D. P.: Dispersion
in Heterogeneous
Non-Uniform
Porous
Media, k!
& f?ng. Cbem. (Sept. 1969) 14-32.

~.

Systems,

Harris,

of Mathem$tk&l

out that the dispersion

91-99;

257.

8.

13.

and

Evaluation

in Short Labora-

( Feb.

Perkins,
T. K., Jr,, and Johnston,
O. C.: A Review
of Diffusion
and Dispersion
in Porous
Media, Sot.
Pet. Eng. ], (March 1963) 70-84; Trans., AIME, Vol.
228.

Bed, AICbE

pointed

*C

K%

3.

10.

for Dynamic

Brighamb

radians/see

of Oil by Solvent
at
Stalkup,
F. I.: Displacement
High Water Saturation, jr SO=, Pet. Errg. J. (Dec. 1970)
337-348.

7.

W. C,,

Domain

DETERMINING CONCENTRATION
PROFILE
OBTAINED

2.

AIME,

Clements,

gin/cc

Eng. J. (March

Trans.,

of Edinburgh

. and Experiences
in
Eng. Prog,, Monograph

APPENDIX

B)

Coats, K. f-f. and Smith, B. D.: tDead-End Pore

Volume and Dispersion in Porous Media,$ Sot. Pet.

Cores, !)

R.,

by K = au

REFERENCES

tory

J.

Frequency

Modela
374-378.

cp

u = frequency,

1.

Experiments
J. 0.:
18. Hougen,
*~ Cbem.
Process
Dynamics,
Series (1960) Vol. 4, 33-34.
~~~e

Society

to core inlet

angle (see

p = viscosity,

Royal

vol.

19. Hays,

cm/sec

x =

L. N. G.: Proc.,
49, 38-47.

17. Filon,
( 1928)

concentration

Chem.

are used. The differences

in the two solutions
are
minor except
when the dimensionless
dispersion
20 or
group NBO . ux/K is small (for example,
22s

less). However, this is precisely

the range of NBO
that is most likely
to be encountered
In tests
performed in reservoir core samples.
The same problem appears in the application
of
the Coats-Smith 1 three-parameter
capacitance
model
to short core data. The model may be written as

A more satisfactory
approach is the transformation
of the experimental
data to the frequency
domain
and fitting of the model in the frequency
domain.
This can be accomplished
easily using the following
method, as described
by Clements. 16
The Laplace
transform
of a function
C(x, t) is
defined by

. . . . . . . . . . . . . . . . . (A-6a)

a co
~~

(1 -f)

= M(C

or, with a transformation

-C).

, . . .( A-6b)

of variables,

where s = jm. The frequency

arbitrarily,
within certain limits.
be written as
co

as
X(C)

m .f ac + (l-f)

a*c
-uax~
ax2

=M(C -C+}.

(l-f) ~
Again,

-#

C and C are related

C,.CX3C
u

. . . . (A-7b)

. . . . . . (A-8)

ax

and C+ and C * are related

= {

(A-is)

by

,..

(A-9)

FREQUENCY-DOMAIN
EXPERIMENTAL
The solution
of the
time domain
is given
integral:

c (x, t)=

z%

B
FITTING
DATA

OF

capacitance
model in the
by a complex
numerical

~jmX(C)
-jce

et ds

by

J3c,)

. . . . . . . . . . . . . . . . . .(B-2)
The computing time and round-off errors involved
in evaluating
this integral
are undesirable,
especially
for an iterative
curve-fitting
process
that
requires the integration
to be performed many times.
226

dt

- j

sk

(tit)]

dt
(B-5)

function,
jd(x,

~)

. . . . .

(B-6)

with real and imaginary parts 9 and 9, respectively.

The numerical
integration
of Eq. !3-5 can be
accomplished
quickly and accurately
using Filons
method. 17
For an input signal that does not return to its
initial value but does reach and hold a steady-state
value within a finite time (such as a step-change),
the method of Walsh and Wiesner18 m~y be used to
transform
the data. This method is equiva!.ent
to
transforming
the derivatives
of the input and output
signaIs,
and is performed as follows:

. . . . .(B-l)

,w,+/~j
X((v
.. .

is a complex

ac) = t?(x,td)+

( m fx,
C

X(c)

where .$?(C) is given

C(L t) [COS (Wt)

. . . . . . . . . . . . . . . . . . .
The result

Substitution
of Eqs. A-8 and A-9 into Eqs. A-6a
and A-6b leads directly to Eqs. A-7a and A-7b.
The proper boundary conditions
for solving Eqs.
A-7a and A-7b to obtain C are Eqs. A-5a through
A-~c.

sin (Wt)]

Consider
the response
of a flow system
to a
concentration
input pulse that begins at time t = 0,
rises to a finite value, and decreases
again to zero.
The output concentration
also will rise from zero
to a finite value, then will drop again to zero at a
finite time tn. Thus, the integral of Eq. B-4 need
be evaluated
only for the finite period for which
C(X, t) + o:
Z(C)

C+=c+$.......

C(x, t) [COS (@t) -j

. . . . . . . . . . . . . . . .(B-4)

by

APPENDIX

u may be chosen
The trmtsforrn may

f=

t)e

Ci (O, t)

Stdt

e t dt

o
ymacg
p e -5t dt
.>,
t
~c. (O,t)e -Stdt

(B-7)

The equivalence
of these ratios is based on the
fact that C(Y., O) = Ci(O, O) = O. Since the input and
output functions reach steady-state
values at finite
then
times t2 and tl, respectively,
SOCIETY OF PETROLEUM ENGINEtiRS JOURNAL

ac (x t)

~=ot?t=t
at

l......

(B-8~)

and
2Ci (O, t)

at
Thus,

O@tt2

. . . . . ..(B-8b)

Eq. B-7 can be integrated

by parts

x(c)
:

C (x, tl)e

E=

Ci (O,t 2) e

-Stz +5
J

X(c)
Zza,c

I
s

C (xjt) e tit

.&u2!x(Ci)

~XPtl

. . . . . . . . . . . . . . . . . . (B-IO)

0 t2

~(Ci)

to give

tl
-Stl + ~

initial velocity
estimate,
Ul, is made by dividing
the concentration
vs time integral
into the core
length.
3. Estimates
of Kl, /1, and Ml are made.
4. Eq. B-2 is evaluated
for each value of s = j~
and the sum of squared errors

Cl (O, t) e

-St ~

. . . . . . . . . . . . . . . . . .(B-9)
Both these integrals are finite, and can be evaluated
from the original experimental
data without further
manipulation.
It should be noted that the input signal Ci may
be a step, pulse,
or any other wave form with
frequency
content
to be
Laplacesufficient
transformed.
It is not limited to a step change as
used in this paper.
Fitting of the model to the data then proceeds
by
the following steps:
1. The experimental
data are transformed
to the
complex (frequency)
domain using Eq. B-9 for a
series of values O ~ 0< timax (~max is chosen as
the value of ~ at which the absolute
value of the
transfer function Y (C)/l? (C~) is equal to 0.3).16
Values
of o are chosen
by trial
and error,
beginning at u = O, such that for each succeeding
value
of w the absolute
value of the transfer
function !? (C)/!? (Ci) is abour 93 to 95 percent of
the transfer-function
magnitude
for the preceding
value of ~. Other techniques
used to choose the
frequencies
(within the range specified)
resulted
in
different
frequency
sets,
but the same best-fit
values of K, f, and M.
2. A material balance is made by integrating
the
dimensionless
effluent
concentration
profile with
respect
to time. (In most cases,
the integral
was
found to be within
2 percent
of its theoretical
value: core pore voIume divided by injection
rate.)
The model fit, in some cases, may be very sensitive
to errors in the assumed
velocity;
therefore,
an

is calculated.
5. A Taylor-series
expansion
of Eq. B-2 around
the parameter
estimates
of Ki, /i, Mi, and ui is
is made to determine
the dmection
of steepest
descent of the sum of squared errors.
6, New parameter
estimates
of Ki+l} /i+ 1, Mi+ 1,
and ui+ ~ are made, and Steps 4 through 6 are
repeated
until a minimum sum of squared errors is
reached.
In all cases,
the final (best-fit)
veIocity differed
from the estimated
(material-balance)
velocity only
by a few percent.
Hays et al. 19 have shown that the method of
least-squares
fitting in the frequency
domain is
equivalent
to least-squares
fitting
in the time
domain. In this study, few problems with convergence
resulted
when reasonable
initial
estimates
were
used,
and the substitution
of different
initial
estimates
led to essentially
the same final parameter
values.
The transformed
data are equivalent
to the data
obtained
from frequency-response
testing
using a
sinusoidal
input signal for exam le, the methods
used by Goss 13 and by Goddard. 1$ The amplitude
ratio of the concentration
signal at any frequency
u is given by

and the phase

$ =
where

lag is given

tan-l

by

(d?/~)+Zn~

n is the integral

number

of cycles

.( B-12)

of lag.
*x~*