You are on page 1of 10

The Wizard Lake Vertical Miscible Flood

Solvent Bank Redesign Concept


M.Y. KWAN, J.P. BATYCKY
Imperial Oil Resources

J.S. TANG
JT Petroleum Consulting Co.

Background

Abstract
The Wizard Lake D-3A Pool has been undergoing a tertiary
hydrocarbon miscible flood (HCMF) since 1983 after completing
primary production (1951 1967) and a secondary HCMF (1969
1983). The operation is comprised of three solvent injectors,
four push gas injectors, and 42 production wells. Fresh solvent
injection (2,500 m3/day) was initiated in 1991 to make up for significant solvent coning and changed the flood design from a vertical flood controlled by longitudinal diffusion to a horizontal
flood dominated by transverse diffusion. Recycling in this case
also allows the old propane-plus bank to be replaced by a more
economical ethane-plus blend. Under solvent recycling, the mixing zone would be stabilized (balanced injection and withdrawal)
with its thickness increasing from zero at the injector to the maximum at the producer. The design problem is thus reduced to
finding the solvent residence time at each producer and the corresponding minimum bank thickness needed to maintain miscibility. The 1D diffusion equation coupled with miscibility data
was used to calculate the required bank thickness and a full field
2D streamline model was used to predict the solvent residence
times in this study. New solvent bank thickness design curves
were generated by merging the residence times and the minimum
miscible bank thickness data. Operating strategies based on the
new design curves can potentially reduce the 1994 solvent bank
size of 5.5 million reservoir cubic metres (rm3) to 3.0 million rm3.
The concept applied in Wizard Lake can be extended to other
miscible floods that involve solvent recycling.

The Wizard Lake D-3A Pool is a dolomitized carbonate reef


with a current estimated original-oil-in-place of 62.4 million m3.
Reservoir pressure was decreased from 15.65 to 12.95 MPa during
primary production. Prior to the secondary miscible flood (1968),
water was injected to repressurize the reservoir to 14.80 MPa. At
this time, the water-oil contact rose to 1,201 m sub-sea (mSS) from
its initial position of 1,229.6 mSS. A first-contact miscible (FCM)
propane-plus solvent and a 95% methane push gas were injected to
sustain a vertical miscible flood from 1969 to 1983. The solventoil contact reached the 1,201 mSS depth by 1983 and a new
approval was granted to extend the flood to capture the tertiary oil
trapped between the 1,201 and 1,229.6 mSS levels. This stage is
known as the tertiary extension phase and the solvent was changed
to an ethane-plus blend. Figure 1 shows a well map of the Wizard
Lake D-3A Pool. Table 1 chronicles the history of the solvent bank
design for the Wizard Lake flood.
Extensive solvent coning commenced in 1991 (~ 3:1 solvent to
oil ratio). The produced solvent was purified and reinjected in
order to maintain solvent bank integrity. Updated geological data
indicated that the targeted reservoir had a lower porosity than was
estimated originally. Due to solvent re-injection, the reservoir flow
became largely horizontal with solvent mixing controlled by transverse diffusion. These events prompted the questioning of the original design assumptions and led to a redesign of the flood in 1994.
The 1983 solvent bank design of 6.13 million rm3 is considered
excessive because recycling sharpens the HC interfaces at the
injectors and limits solvent dispersion to that corresponding to the

TABLE 1: Wizard Lake solvent bank design history.

Date

Reservoir
Pore Volume
Estimate
(106 m3)

Solvent Bank Design


_________________________
Thickness at
corresponding depth
(m)
(mSS)

1969

90

max.
end

10.2
8.7

1,180
1,205.4

5.0

4.50

Secondary Miscible

1983

90

max.
end

11.2
4.8

1,182
1,228.4

7.5

6.13

Tertiary Extension
Phase

1991

83.5

actual 10.4
end
4.8

1,211.9
1,228.4

1994

83.5

end

1,229.6

3.0

Bank
Volume
(% PV)

Bank
Volume
(106 rm3)*

Stage of
Operation &
Assumptions

Revised Geology
3.00

Redesign Assuming
Trans. Dispersion Dominant
and Solvent Recycling

*rm3 = reservoir m3
PEER REVIEWED PAPER (REVIEW AND PUBLICATION PROCESS CAN BE FOUND ON OUR WEB SITE)
36

Journal of Canadian Petroleum Technology

the time of writing, the new design documented here has been
implemented into the operations of the Wizard Lake flood.

Diffusion Modelling
The 1D transverse diffusion equation(1) is given by
D ij

2 C i C i
=
t ...................................................................................(1)
z 2

where Ci is the concentration of fluid i, z is the coordinate normal


to flow, t is time, and Dij is the dispersion coefficient between fluids i and j and its value depends on the normal flow velocity. The
flow geometry in the solvent reinjection stage of the Wizard Lake
miscible flood is illustrated in Figure 2. The flow is predominantly horizontal and dispersion is normal to the flow in the vertical
direction. This mode of mass transfer is most suitably described as
transverse dispersion. For initially sharp gas-solvent and solventoil interfaces, Equation (1) can be solved to give the various HC
concentration profiles:
zz

sg
C g = 0.5erfc
2 D sgT t

FIGURE 1: Wizard Lake well location map.

maximum residence time. A pseudo steady state mixing zone


forms between the injector and producers, and the maximum solvent residence time becomes the controlling parameter. This presented an opportunity to redesign the bank size with a simplified
analytical approach. The 1D diffusion model was combined with
phase behaviour data and a 2D streamline residence time model to
generate a series of solvent bank thickness design curves as a function of well distance and residence time (hence, operating rate).
This paper examines the steps taken in coming up with the
redesign. First, the appropriate diffusion model equations needed
to calculate bank thickness are described. Second, log and core
data leading to the correlation of field scale diffusion coefficients
are examined. Third, the rationale of combining diffusion model
and miscibility data to arrive at minimum bank thickness at firstand multiple-contact miscible conditions is discussed. Fourth, the
2D full field streamline residence time simulation is described.
Finally, the merging of the pieces to form the redesign and the
approaches to implement the design in the field are discussed. At

................................................................(2)

zz

zz

sg
so
C s = 0.5erfc

0.5erfc
2 D soT t
2 D sgT t
Co = 1 Cs Cg

........................(3)

..................................................................................(4)

where erfc is the complementary error function; z is the vertical


coordinate (flow is perpendicular to z); subscripts g, s, sg, so, and
T denote push gas, solvent, solvent-gas and solvent-oil contacts,
and transverse respectively. The above equations will be used to
determine the solvent bank thickness, which is the separation
between the oil-solvent and gas-solvent contacts. The locations of
the contacts are defined as the 50% concentration crossover points.

Dispersion Measurements
Wizard Lake field diffusion coefficients were determined from
three data sources, log data, core data, and published values.
Solvent-oil and push gas-solvent contact locations are monitored
regularly using gradiomanometer density logs. The measured density profiles can be converted to concentrations using

FIGURE 2: Well
fluid profile during
solvent reinjection.
June 2001, Volume 40, No. 6

37

TABLE 2: Density log dispersion coefficients summary.


Well
No.

Date
Logged

Dso,L
(cm2/s)

Correlation
Coefficient

Dsg,L
(cm2/s)

Correlation
Coefficient

15-9
13-15
9-21
2/2-27
15-9
13-15
1-9
2/2-27

02 Feb/90
13 Feb/90
22 Feb/90
02 Mar/90
04 Feb/91
09 Feb/91
22 Feb/91
04 Mar/91

2.82 10-5
1.32 10-5
1.49 10-5
1.10 10-5
1.66 10-5
4.23 10-6
-

0.944
0.973
0.985
0.989
0.997
0.989

9.76 10-5
2.83 10-4
1.24 10-4
2.54 10-4

0.997
0.941
0.982
0.997

Average
1.47 10-5

Co =

m s
o s

and
C s = 1 C o ...........................................................................................(5)
for the solvent-oil mixing zone. Similarly for the solvent-push gas
zone, the concentration equations are
Cs =

m g
s g

and
Cg = 1 Cs

...........................................................................................(6)

In Equations (5) and (6), is density and the subscript m


denotes mixture. Other symbols are previously defined. Ideal mixing with no volume shrinkage and no gas penetration into the oil
zone are assumed. These assumptions introduce an error of a few
metres in locating the HC interfaces. Figure 3 shows typical density log data converted to concentrations and curve-fitted to
Equations (2) and (3).
The measured profiles are occasionally non-symmetrical relative to the interface; this may be an artifact caused by logging tool
movement or rock heterogeneity. Equations (2) and (3) can be
rearranged into a linear form y =x/ D and the dispersion coefficient can be correlated using linear regression. The match between
data and model can sometimes be improved by shifting contact

FIGURE 3: Solvent, gas, and oil saturation profiles for Well 13-15
(February 13, 1990).
38

Average
1.90 10-4

location. There is an overlap of 4 m between the gas and oil data


in Figure 3, and this will produce some error in locating the contact because the theory assumes that only two components exist in
that mixing zone. Table 2 summarizes the statistics used in correlating the solvent-oil and solvent-push gas diffusion coefficients,
and the analytical expressions fit the concentration data quite well.
According to Perkins and Johnston(2), convective dispersion
dominates at high Peclet No. (Pe2 > 100) and molecular diffusion
dominates at low Peclet No. (Pe2 <1). The definition of Pe2 is
Udp/Do where U is interstitial velocity, dp is grain diameter of sand
pack, and Do is a molecular diffusion coefficient. Based on a horizontal velocity of 200 m/yr., Kh of 1,070 md, of 0.1, Do of 2
10-4 cm2/s, and an equivalent dp = K h / of 3.27 10-4 cm, a Pe2
value of 0.001 is obtained for the Wizard Lake operation. This data
indicates that molecular diffusion is dominant in the Wizard Lake
HCMF.
The Coats-Smith model(3) describes mass transfer involving
live and dead pores (e.g., carbonates) and it was used to correlate dispersion parameters from vertical miscible core flood data
to use in the original Wizard Lake solvent bank design. Yellig and
Baker(4) suggest that the Coats-Smith model can be reduced to the
simple diffusion equation with an equivalent dispersion coefficient
of
De
(1 f )u 2
= 1+
D
DK
..............................................................................(7)
when system length approaches field scale and D, f, and K are the
fitting parameters and u is the interstitial velocity. In the redesign,
the above method was used to transform the Coats-Smith parameters to a single dispersion coefficient so that the simple diffusion
model can be used. Furthermore, the combined effect of D, f, and

FIGURE 4: Miscible core flood dispersion coefficient scaling.


Journal of Canadian Petroleum Technology

Scaled
Core
Data

Theory(4)
(Wilke &
Chang)

Density
Logs

Final
Selected
Values

Dsg,T
(cm2/s)

2 to 5 10-5

1.7 10-4

2.0 10-4

5 10-5

Dso,T
(cm2/s)

1 10-5

2.9 10-5

1.5 10-5

1 10-5

K is length dependent(4) and must be scaled for field application.


The scaling is accomplished in Figure 4 by plotting core-derived
De against De/uL. The solvent-oil data represent the displacement of kerosene and decane by heptane to simulate mixing in the
solvent-oil zone. The solvent-push gas data represent the displacement of C4 by C1 and N2 to simulate mixing in the gas-solvent zone. At a Peclet number of 10-3 corresponding to the Wizard
Lake miscible flood in Figure 4, the dispersion coefficients for the
solvent-oil and solvent-push gas mixing zones are extrapolated to
1 10-5 and 2 to 5 10-5 cm2/s, respectively. The validity of the
extrapolation is reaffirmed when the solvent-oil diffusion coefficient came out to be the same as that interpreted from density-logs
and from corrected bulk fluid diffusion coefficients (which will be
addressed next). At small De/uL values, De is independent of the
length scale (i.e., dead pore and end effects become insignificant).
Perkins and Johnston(2) reported a factor of 0.7 between porous
media and bulk fluid diffusion coefficients. The relevant bulk fluid
coefficients from Wilke and Chang(5) have been multiplied by 0.7
and included in Table 3 to be compared with the diffusion coefficients correlated from density log and core data. The solvent-oil
dispersion coefficients from all three sources are about the same.
The scaled core flood solvent-gas dispersion coefficients are considerably less than both Wilke and Changs and the density-log
data correlated coefficients. Some possible reasons for the discrepancy are as follows. The correction factor of 0.7 derived from
sand packs may not be appropriate for use with carbonate cores.
The density logging tool could induce additional mixing in the
well bore fluid during measurement. The unconventional method
of introducing methane gas as a lift gas through the casing may
have increased the level of mixing within the well bore and hence
becomes less representative of the level of mixing in the reservoir.
The core-derived solvent-gas dispersion coefficient is more reliable and is chosen over the other two data sources. Scaled core
flood solvent-oil and solvent-push gas dispersion coefficients of 1
10-5 and 5 10-5 cm2/s are respectively selected for the redesign
because they were obtained under controlled conditions.

New Solvent Bank Design Basis


Figure 2 illustrates the flood process in Wizard Lake as a horizontal flow dominated by transverse dispersion (the bold arrows)
occurring normal to the HC interfaces. Both field and laboratory
data indicate that molecular diffusion is the dominant mode of
mass transfer. Due to flow convergence, there is a small flow
region around the well bores that is controlled by convective dispersion. Using Perkins and Johnstons(2) criterion of Pe2 > 10
revealed that this region has a radius of ~ 0.25 m for the Wizard
Lake reservoir condition. This small region will be ignored in our
analysis.
The shaded boundaries in Figure 2 represent the mixing zones
in a solvent recycling scheme. Each mixing zone has no thickness
at the injector and has the maximum thickness at the producer.
For balanced injection and withdrawal the process reaches a
pseudo-steady state. If we move vertically at the same velocity as
the interfaces, then the mixing zone HC concentration profiles
will remain constant with time. Mixing zone thickness varies only
with the distance from the injector. As a result, solvent bank
design is simplified to maintaining miscibility at the producers. In
June 2001, Volume 40, No. 6

crit
ical
poin
t

TABLE 3: Wizard Lake dispersion coefficients


summary.

FIGURE 5: Pseudo-ternary diagram of the Wizard Lake fluids.

a diffusion-controlled process, the solvent required to separate


push gas from the oil depends only on the contacting time
between the HC layers. The critical solvent saturation to ensure
miscibility is dictated by phase behaviour. The design process is
reduced to using the analytical diffusion model to impose the
critical solvent saturation at the producers based on maximum
solvent residence time. Modelling residence time using a
streamline model will be dealt with later.

Miscibility Limits
A pseudo-ternary diagram representation of the Wizard Lake
HC system at reservoir condition is given in Figure 5. The twophase envelope was generated through a series of flash calculations. Solvent, oil, and push gas were mixed in predetermined
ratios and then flashed to obtain equilibrium vapour and liquid
compositions. The current reservoir oil, solvent, and push gas compositions are represented respectively by points A, C, and D in
Figure 5. Lines A-C and C-D represent, respectively, pure oil/solvent and solvent/gas mixtures. Lines A-B and B-D represent the
first-contact miscible (FCM) limit for the tertiary solvent-oil-push
gas mixture. A minimum solvent concentration of 94% (95 mole
%) is required in the solvent-oil mixture to maintain the FCM condition. In Figure 5, this requirement is represented by the ratio of
length AB to AC. The multiple-contact miscible (MCM) limit is
represented by the intercept between the critical tie-line and line
A-B. A minimum solvent concentration of 73.7 mole % is needed
to establish MCM. A limit of 1 to 99% saturation has been chosen
to define the thickness of a mixing zone (the shaded lines in
Figure 2).
Minimum solvent bank thickness at either FCM or MCM condition can be correlated to solvent residence time. The optimal
condition is illustrated in Figure 6. When a critical solvent concentration of 94% (see above) is maintained at the 1% push gas
front, the oil zone is sufficiently isolated from gas invasion to
ensure FCM condition. The required bank thickness is the vertical
separation between the two HC interfaces. Solvent-oil and solventgas diffusion coefficients of 1 10-5 and 5 10-5 cm2/s were used,
respectively, with Equations (2) to (4) to generate the miscibility
limits in Figure 8 in terms of required solvent thickness for a given
residence time. The calculation procedure is illustrated in
Appendix I. The area below the FCM line in Figure 8 yields a bank
size thicker than the FCM limit. The area between the FCM and
MCM curves represents MCM designs. The area above the MCM
curve represents partially miscible and immiscible designs. When
39

FIGURE 6: HC concentrations at optimal bank thickness.

the flood crosses the MCM limit into the immiscible region, there
is a gradual increase in the residual oil saturation. A gas flood
residual saturation will not occur until the oil meets a 100% push
gas front.

Areal Streamline Modelling


Various computer models were considered before settling on the
use of the streamline model. Using the Todd-Longstaff(6) approach
in a computer model, we were not able to simulate the thin transverse mixing zones associated with molecular diffusion.
Concentration oscillation and convergence problems were also
encountered when using small time steps. Further testing revealed
that a full field simulation of the miscible flood using a numerical
simulator was simply out of the question at the time. Since the
redesign problem has been reduced to maintaining a critical solvent concentration for a given residence time, the problem is now
amenable to the use of a streamline model to simulate the duration
of solvent flow in the reservoir. Once the streamline model results
are verified using computer modelling (tracer option) and found to
agree with selected field residence times, full field simulation of
solvent residence time can then be undertaken.
Based on potential theory, a 2D streamline model was developed to track solvent residence time between wells in the Wizard
Lake miscible flood. The model is described in Appendix II. The
result is a key input for reworking the floods solvent bank requirement. A full 3D field simulation was too costly to run. The model
uses sources and sinks to simulate injection and production wells,

FIGURE 7: Wizard Lake solvent residence time contours.

respectively, and their relative strengths are proportioned according to actual well rates. Injection and production data for the
month of April 1994 were used to assign source and sink potentials
(well locations in Figure 1). Areal geometry is faithfully duplicated, but the vertical solvent profile is not modelled. The streamline
model predicted radial flow out of the three injectors even though
the wells were drilled on square patterns, and there was considerable variation in well rates.
Areal contours of equal solvent residence time have been generated for the entire field and are plotted in Figure 7. The results
have been used to generate the new solvent thickness redesign
curves of Figure 8. These curves are straight lines correlating residence times to the bank thickness required to sustain FCM and
MCM conditions at fixed radial distances from the central injector.
Each straight line traces the shift in residence time as the solvent
bank thickness changes for a fixed location. When a residence time
line crosses the FCM limit, the process becomes MCM. If the line
crosses the MCM limit, then the process will eventually become
immiscible. Solvent residence time increases directly with increasing zone thickness and distance from the injector provided injection rate is kept constant. The residence times in Figure 8 have
been adjusted for the effects of sweep efficiency, gravity, and the
vertical dimension, as will be described in the next section.

Streamline Model Calibration

FIGURE 8: Wizard Lake solvent bank redesign curves.


40

The vertical variation in the solvent path (Figure 2) might cause


a large discrepancy between the actual solvent residence times and
the streamline model predicted values. The effect of phase density
would be insignificant if miscibility is maintained between the
HCs. To verify the accuracy of the streamline model residence
time, black oil numerical simulations were conducted. One set of
radial flow simulations (push gas and solvent) with a central injector and two producers (radii of 490 and 980 m from the injector)
produced solvent arrival times 14% higher than the corresponding
streamline model predicted values. The effect of gravity and the
vertical dimension accounts for the extra travel time. A second set
of simulations added a third component (oil) to model the same
radial geometry. The 2-component results are equivalent to the 3component simulation results. Varying the height of the injection
perforations above the production interval had no effect in the simulations. The third set of simulations used a 3D rectangular grid to
simulate a segment of the Wizard Lake reservoir. The flow boundary was defined by the outer streamlines from Injector 10-9 to
Journal of Canadian Petroleum Technology

TABLE 4: Application of the streamline model to modify bank size (central injector with two typical offset
producers).

Production
Rate
(m3/d)

Solvent
Bank
Thickness
(m)

Equivalent
Well*
Distance
(km)

Base
Residence
Time
(yr.)

120

120

5.6

0.87
0.56

5.3
1.7

1.00

5.3
1.7

B. Transition Start
90

60

120

5.6

0.87
0.56

5.3
1.7

1.33
(= 120/90)

7.1
2.3

C. Transition End
90

60

120

3.0

0.92
0.56

3.2
1.2

1.33
(= 120/90)

4.3
1.6

120

120

5.6

0.87
0.56

5.3
1.7

1.00

5.3
1.7

D. Transition Start
150

120

180

5.6

0.87
0.56

5.3
1.7

0.80
(= 120/150)

4.2
1.4

E. Transition End
150

120

180

3.0

0.87
0.56

3.2
1.2

0.80
(= 120/150)

2.6
1.0

Step

Average
Rate
(m3/d)

Injection
Rate
(m3/d)

Correction
Factor

Corrected
Residence
Time
(yr.)

Option I (Reduced Injection)


A. Base
120

Option II (Increased Production)


A. Base
120

Steady-State Operation
F.

120

120

120

3.0

0.87
0.56

3.2
1.2

1.00

3.2
1.2

G.

90

90

90

3.0

0.87
0.56

3.2
1.2

1.33
(= 120/90)

4.3
1.6

H.

60

60

60

3.0

0.87
0.56

3.2
1.2

2.00
(= 120/60)

6.4
2.4

* Streamline model distances were decreased by 14% to correct for flow effects and increased by 9% to correct for sweep efficiency.

Producers 16-9 and 1-16 (Figure 1) generated using the streamline


model. The 3D simulation model residence times are 12 to 14%
higher than the corresponding streamline model values.
The streamline model residence times need to be increased by
14% to match the numerical simulation data. Solvent-oil contact
measurement indicated, however, that the overall sweep efficiency
in Wizard Lake is 91%. Therefore solvent residence times must
also be reduced by 9% to correct for sweep efficiency. The net correction to the streamline model residence times works out to be an
increase of 3.7%.
Field data indicated that cumulative solvent produced at the
start of 1993 was equal to the total solvent injected up to 1983.
This data yields an average solvent residence time of ten years for
Wizard Lake. By averaging 272 streamline residence times from a
full-field streamline simulation, an average solvent residence time
of 9.5 years was obtained. The simulation was based on the actual
average bank thickness from flood start to 1987 (8.2 m). Increasing
this average residence time by 3.7% (correcting for vertical flow
and sweep efficiency) yielded 9.9 years. Both average field and
streamline residence times are nearly identical. It is therefore concluded that the streamline model after calibration can adequately
quantify solvent residence times in the Wizard Lake reservoir.

Implementation Approach
A correction factor of + 3.7% has been applied to all of the residence time lines in Figure 8, as was justified previously. The
redesign is based on a typical reservoir segment encompassing
June 2001, Volume 40, No. 6

injector 10-9 and producers 1-16 and 16-9 (see Figure 1). Adding
Township 48 and Range 27W4M to the well name completes their
LSD convention. The producing wells represent two rings of wells
with radii of 0.56 and 0.87 km. At the outer wells, it would require
a minimum solvent bank of 3.0 m to operate at the FCM condition.
The redesign is based on the solvent bank reaching the widest part
of the reservoir at flood end. The 1994 injected solvent volume of
5.5 million rm3 would have a thickness of 5.6 m if it reaches the
1,229.6 mSS level. This bank is much thicker than the 3.0 m
required at the outer wells. The 3.0 m bank target is translated into
a bank volume of 3.0 million rm3. Total allocated solvent injection
and withdrawal for the above three-well group is 120 rm3/day. The
typical rate for each producer is 60 rm3/day (2.4% of the total yearly production of 1% HCPV). Two options were studied (see Table
4). Option I reduces injection by 50% to 60 rm3/day while maintaining production in each well at 60 rm3/day. Option II increases
total production to 180 rm3/day while maintaining injection at 120
rm3/day. Boths options require 5.5 years to produce the excess 2.5
million rm3 of solvent, since the over-production is 30
rm3/day/producer. Selection of either option will be influenced by
facility constraints.
The two strategies are illustrated in Figure 8 at the 0.87 km well
location. The starting point is A and it represents the 1994 projected bank thickness at the 1,229.6 mSS level. Calculated residence
times for the 0.56 and 0.87 km producer locations are given in
Table 4. For Option I, reducing the average flow rate from 120 to
90 rm3/d instantaneously increases solvent residence times, and
this is shown as a step change from point A to B. Over-production
of solvent causes the solvent bank to thin and the residence time
41

decreases along path BC. At point C, the 3.0 m bank target is


reached on the MCM limit curve and balanced operation is
restored. Option II also starts at point A and jumps to D following
the overall rate increase. Thinning of the bank is represented by
line D-E in Figure 8. Points C, F, and E will not line up at the 3.0
m line if they were to be located right at the MCM and FCM limits, respectively. Point E is stopped short of the FCM limit, as it is
easier to restore balanced operation at the 3.0 m thick bank without overshooting the target at F. The higher rate in Option II allows
the excess solvent to be produced faster, but the lower rate in
Option I reduces solvent cycling requirements. After thinning the
solvent bank, steady operation can be restored by rebalancing solvent injection and withdrawal. If the original rate of 120 rm3/day
is restored, then operation will resume at point F (FCM). If the
rates are lowered to 75% or 50% of the initial target, then operation will shift, respectively, to point G (MCM) or point H
(immiscible).
Solvent bank thickness can be further reduced to 2 m if injectorproducer distances are kept within the 0.6 km range (Figure 8).
However, at the late stage of the flood, injector reconfiguration is
not an economical option. This condition would enable a further
reduction of solvent volume of 0.9 million rm3 down to 2.1 million
rm3. Thus, there is a further opportunity to reduce solvent requirements and optimize operations by redistributing solvent injection.

Conclusions
Based on the foregoing analyses, it is concluded that:
1. Due to the predominantly horizontal cycling of solvent in
Wizard Lake operations, the mixing of miscible fluids (solvent, oil, and push gas) is best modelled as a transverse dispersion process, rather than the more commonly used longitudinal dispersion representation.
2. Based on field fluid density log correlation, scaled core flood
dispersion data and bulk fluid diffusion data, solvent-oil and
solvent-gas dispersion rates are very slow and are dominated
by molecular diffusion. As a result, the push gas-solvent-oil
dispersion process can be modelled as a purely time-dependent process.
3. A calibrated one-dimensional diffusion equation coupled
with a tuned equation-of-state representation of the phase
behaviour of Wizard Lake reservoir fluids has been applied
to uniquely specify the required thickness of the solvent bank
that would maintain either an FCM or MCM condition as a
function of solvent residence time.
4. A two-dimensional areal streamline model, calibrated with
numerical simulation and reservoir data, has been developed
to generate solvent residence times at constant solvent bank
thickness. A process has been outlined (c.f. Figure 8) for
applying the streamline model to develop strategies to reduce
solvent bank size.
5. Within current operations, assuming FCM and MCM conditions are maintained at distances of up to 0.9 km from current
injectors, the 1994 solvent bank can be reduced from 5.5 million rm3 to 3.0 million rm3.

Ao, B
C
C*
d
D
erf-1
erfc
f
FCM
HC
HCMF
HCPV
h
K
Kh
Kr
Kx, Ky
LSD
MCM
Pe
Pe2
Q
r
t
u
Vr ,w
v
x
xo, yo
y
z

= integration constants in Equation (II-3)


= concentration, integration function in
Equation (II-15)
= stagnant pore volume C in Coats-Smith model
= sand particle diameter (m)
= dispersion or diffusion coefficient (cm2/s, m2/s)
= inverse error function
= complementary error function
= mobile fluid pore fraction (Coats-Smith Model)
= first-contact miscible
= hydrocarbons
= hydrocarbon miscible flood
= hydrocarbon pore volume (m3)
= solvent zone thickness (m)
= material transfer constant in Coats-Smith model (1/s)
= horizontal permeability (m2)
= polar co-ordinate integration function in
Equation (II-5)
= permeability in x, y co-ordinate
= legal sub-division
= multiple-contact miscible
= Peclet number (Do/Udp)
= Peclet number (Udp/Do)
= well flow rate (m3/day, m3/s)
= radius (m)
= time co-ordinate (s, day)
= interstitial velocity, velocity in x direction (m/s)
= radial velocity (m/s)
= velocity in y direction (m/s)
= horizontal dimension (m)
= source and sink location in streamline model (m)
= areal dimension normal to x (m)
= vertical dimension (m)

Greek Symbols

=
=
=
=
=

ratio of Kx / Ky
angular segment around injectors
fluid dynamic viscosity (Pas)
polar coordinate
fluid density (kg/m3)

Subscripts
a
c
e
g
i,j
m
o
s
T
w

=
=
=
=
=
=
=
=
=
=

1% push gas concentration location


solvent-oil 50% concentration line
equivalent D [Yellig and Baker(4)]
push gas
fluids i, j
oil-solvent mixture
oil
solvent
transverse direction
well

Appendix IZone Thickness Calculation


Acknowledgements
We thank Imperial Oil Resources for permission to publish this
paper and the Battle River Operations for sponsoring this project.
We thank Murray Sykes and Drew Irwin for their constructive
comments and reviews. Operating and PVT data from Denise Furst
and Dr. Jimmy Yu, respectively, are greatly appreciated. We
acknowledge Dr. Bruce S. Careys input on streamline modelling.

NOMENCLATURE
English Symbols
a
42

= length for normalizing streamline model (m)

We considered the hypothetical case of placing a one time solvent slug at the top of the Wizard Lake D-3A Pool and displacing
this bank downward with push gas to the final depth of 1,229.6 m.
Figure 6 illustrates the bank geometry at this position. Based on a
vertical frontal rate of 2 m per year, it would take 27 years to complete the flood. The required final bank thickness can be obtained
by applying Equation (2) at the 1% gas front (z = za) to give
z

g
0.01 = 0.5erfc
2 D sg t

or
Journal of Canadian Petroleum Technology

z g = z a z gs = 2 Dsg t erf 1 (0.98)

.........................................(I-1)

where t is the solvent residence time. Applying the critical solvent concentration at the 1% push gas front za (Cs = 0.99 0.94 =
0.93) using Equation (3) yields
z z
so
0.93 = 0.5erfc a
0.01
2 D so t

v=

(y y o )
2B
(x x o )2 (y y o )2
+
Kx
Ky

......................................................(II-4)

Transforming Equation (II-4) into polar coordinates yields


Vr =

2B
K r ( )
r

where

or
zso z a = 2 Dso t erf 1 (0.88)

.....................................................(I-2)

The optimal solvent bank h = zso - zgs is obtained by adding


Equations (I-1) and (I-2) so that
1

h = 2 Dsg t erf (0.98) + 2 Dso t erf (0.88)

K r ( ) =

cos sin 2
+
Kx
Ky
2

...............................................................(II-5)

2
2
Integrating Vr = u + v around the well radius rw yields the
total flow Q in or out of the well such that

...................(I-3)

Using Dso = 1 10-5 cm2/s, Dsg = 5 10-5 cm2/s and t = 27


years in Equation (I-3) gives 8.82 m. This method was used to generate the FCM and MCM limit curves in Figure 8.

Q = hrw Vr ()d =
0

2 Bh

K ( ) d
r

..........................................(II-6)

where h is zone thickness. Constant B is given by

Appendix IIAreal Streamline Model


[Muskat(7) and Hurst(8)]

B=

2 h K r ()d

Solvent velocities u, v in the x and y flow directions are

...........................................................................(II-7)

K P
u= x
x

For Kx = Ky = K, B reduces to Q/4hK. Substituting Equation


(II-7) into Equation (II-4) yields

and
K y P
v=
y ....................................................................................(II-1)
where Kx and Ky are x and y permeabilities, P is pressure and is
solvent viscosity. The continuity equation is
u v
2 P
2 P
Kx 2 + Ky 2 = 0
+
=0
x y
x
y
or
......................................(II-2)

u=

(x x o )
(x x o )2 (y y o )2
+
h K r ()d
Kx
Ky
Q

.....................................(II-8)

and
v=

(y y o )
(x x o )2 (y y o )2
+
h K r ()d
Kx
Ky
Q

.....................................(II-9)

For this problem, rate boundary conditions are applicable.


Integration is performed on a unit circle around each well (source
or sink) and the total flow in or out of each well is equated to the
allocated flow. The pressure field can be, but is not, calculated in
this approach since we are primarily concerned with solvent residence time and hence the inverse velocity field. As will be shown
later in this section, a given set of relative well rates will uniquely
define a unique set of streamlines for a given reservoir.
The solution to Equation (II-2) is

The velocity components at a point (x,y) are made up of the


algebraic sums of the contributions from each well. By letting
2

A = 1 / K r ( ) d
0

..........................................................................(II-10)

the velocity components for a multi-well system become


(x x o )2 (y y o )2
P = A o + B ln
+

K y
K x
........................................(II-3)

dx
= u( x, y) = A
dt

i =1

(x x o )
2B
(x x o )2 (y y o )2
+
Kx
Ky

and
June 2001, Volume 40, No. 6

o ,i )

Kx

where Ao and B are constants and (xo,yo) is the location of the well.
Substituting Equation (II-3) into (II-1) yields
u=

(Q i / h )( x x o,i )

(x x

( y y o ,i ) 2
Ky

.........................(II-11)

and
dy
= v( x, y) = A
dt

(Q i / h )( y y o,i )

(x x
i =1

o ,i )

Kx

( y y o ,i ) 2
Ky

.........................(II-12)
43

FIGURE 9: Calculating partial flow around an injector well.

For injectors, Qi/h > 0 (injection) and for producers, Qi/h < 0
(production). The subscript i denotes each well. Equations (II-11)
and (II-12) can be normalized by dividing the x,y coordinates by a
length scale a, and the individual well rates by the total injection
or production rate Q. The ratio of = Kx/Ky fixes the shape of the
streamlines. Equations (II-11) and (II-12) then become
x x
n
(q i / q )( o,i )
x y
q
a
a
u( , ) = C( )
y y o ,i 2
a a
a i =1 x x o,i 2
(
) + (
)
a
a
a
a
......................(II-13)

x y
q
v( , ) = C( )
a a
a i =1

y y
(q i / q )( o,i )
a
a
x x o ,i 2
y y o ,i 2
(
) + (
)
a
a
a
a
......................(II-14)

Numerical integration is required to solve Equation (II-16)


when Kx Ky. When Kx = Ky, then the normalized flow q12/q is
equal to (2 - 1)/2.
To generate the streamlines, a small circle is drawn around each
injector as shown in Figure 10. Equally spaced rays are drawn
from the centre of the circle. A streamline is constructed from each
ray by numerically integrating Equations (II-13) and (II-14) with
respect to time. First, a small constant displacement z in the x-y
plane is selected as the incremental length. A ratio of z/a (a is the
normalizing length) of 0.005 or less is recommended. The radius
of the generating circle ra should be around 0.001a. The starting
point for Ray No. 2 in Figure 10 is given by

(x o + ra cos 2, y o + ra sin 2) .
The incremental time required to reach (x,y) by displacing z
from the starting point is
t =

and
C=

FIGURE 10: Streamline generation around an injector.

1
2

cos

d
+ sin 2

...............................................................(II-15)

where qi = Qi/h. The normalized velocity equations show that the


streamline pattern depends only on relative flow distribution,
source and sink location, and permeability ratio. The combined
velocity and zone thickness effect is represented by q/a. The flow
angle tan-1(u/v) is independent of q/a because this parameter is
cancelled out in the velocity ratio. This model can therefore handle changing flow rates with time as well as varying zone thickness. The function A determines the flow between various streamlines. The geometry is illustrated in Figure 9.
The flow enclosed by the radial streamlines with angles 1 and
2 relative to the total flow is
q 12
=
q

K ( ) d / K ( ) d
r

2
2
where w = u + v is calculated using Equations (II-13) and
(II-14) based on the coordinates of the previous time step. The
coordinates of the streamline at the ith time step is, therefore, given
by

x i = x i 1 + u( x i 1 , y i 1 )t i ...........................................................(II-18)

y i = y i 1 + v( x i 1 , y i 1 )t i ...........................................................(II-19)
and
t i =

z
( u( x i 1 , y i 1 )) + ( v( x i 1 , y i 1 )) 2 ..................................(II-20)

The cumulative residence time along a streamline at a given


location is the sum of the incremental times up to that point, i.e.,

z
w ...........................................................................................(II-17)

.....................................................(II-16)

t(x i , y i ) =

t
k =1

where

K r () = 1 / cos 2 / K x + sin 2 / K y
44

.........................................................................(II-21)

Iteration is terminated when a streamline reaches a producer. A


small tolerance of + z/2 can be used to test for this condition. The
number of rays in the generating circles will dictate the number of
streamlines that will be generated. An infinite boundary is assumed
Journal of Canadian Petroleum Technology

in this approach. For the well configuration in Wizard Lake, typically only 4% of the streamlines generated by the model flowing
out from the injectors would miss all of the producers. A no flow
boundary condition would only alter the shape of the streamlines
slightly. Once the solvent residence time is mapped, then the dispersion equation can be used to redesign the solvent bank size for
the Wizard Lake depletion scheme.

REFERENCES
1. BEAR, J., Dynamics of Fluids in Porous Media; American Elsevier
Publishing Company, Inc., New York, 1972.
2. PERKINS, T.K. and JOHNSTON, O.C., A Review of Diffusion and
Dispersion in Porous Media; Society of Petroleum Engineers Journal,
pp.70-84, March 1963.
3. COATS, K.H. and SMITH, B.D., Dead-End Pore Volume and
Dispersion in Porous Media; Society of Petroleum Engineers Journal,
pp.73-84, March 1964.
4. YELLIG, W.F. and BAKER, L.E., Factors Affecting Miscible
Flooding Dispersion Coefficients; Journal of Canadian Petroleum
Technology, pp.69-75, October December 1981.
5. WILKE, C.R. and CHANG, P., Correlation of Diffusion Coefficients
in Dilute Solutions; American Institute of Chemical Engineers
Journal, Vol. 1, No. 2, pp.261-270, June 1955.
6. TODD, M.R. and LONGSTAFF, W.J., The Development, Testing,
and Application of a Numerical Simulator for Predicting Miscible
Flood Performance; Journal of Petroleum Technology, pp.874-882,
July 1972.
7. MUSKAT, M., Physical Principles of Oil Production; First Edition,
McGraw-Hill Book Company, Inc., New York, Chapter 5, 1949.
8. HURST, W., Reservoir Engineering and Conformal Mapping of Oil
and Gas Fields; The Petroleum Publishing Company, Tulsa, OK,
1979.

ProvenanceOriginal Petroleum Society manuscript, The


Wizard Lake Vertical Miscible Flood Solvent Bank Redesign
Concept (99-48), first presented at the 50th Annual Technical
Meeting, June 14 18, 1999, in Calgary, Alberta. Abstract submitted for review October 21, 1998; editorial comments sent to the
author(s),May 29, 2000; revised manuscript received June 20,
2000; paper approved for pre-press November 30, 2000; final
approval June 4, 2001.

June 2001, Volume 40, No. 6

Authors Biographies
Mori Kwan received B.A.Sc. and Ph.D.
degrees in mechanical engineering from the
University of Waterloo and is a member of
APEGGA. He is currently a senior research
specialist with Imperial Oil Resources in
Calgary and has worked on steam-based oil
recovery processes, heavy and conventional
oil core analysis, miscible floods, dispersion, hydrocarbon phase behaviour, and
non-thermal heavy oil recovery processes.
With his untimely passing in 1998, Dr. Jim
Batycky left behind a legacy of outstanding
technical contributions to the Petroleum
Society and the oil industry. Jim was a senior advisor at Imperial Oil Resources and he
also worked for the Petroleum Recovery
Institute in the early 1980s. Jims career
encompassed the areas of reservoir engineering, simulation, miscible flood, reservoir depletion, interfacial phenomena, and
core analysis.
Joseph Tang, the principal of JT Petroleum
Consulting Co., is a fellow of the China
Institute of Atomic Energy and an appointed expert scholar of the United Nations
International Atomic Energy Agency. He
worked for 20 years in oil and gas production research for a major Canadian oil company. Dr. Tang is a distinguished expert in
tracer applications. His other areas of interest include enhanced oil recovery, micellar
flood, multi-phase porous media flow, miscible displacement, flow stability, numerical simulation, non-thermal heavy oil recovery, formation damage, groundwater drift and
contamination, thermal hydrocracking of bitumen, and upgrading.
He holds a Ph.D. degree in chemistry and a M.Sc. degree in chemical engineering from the University of Alberta and is a member of
the Petroleum Society and APEGGA.

45

You might also like