Professional Documents
Culture Documents
in Miscible Displacements
Olaoluwa O. Adepoju, and Larry W. Lake, SPE, The University of Texas at Austin, and
Russell T. Johns, SPE, The Pennsylvania State University
–30
0.90
–20
0.40
Dispersivity (ft)
1.5 0.0001 to 0.001
0.001 to 0.01
1.0 0.01 to 0.1
αL
αT
0.1 to 0.3
0.5 0.3 to 0.6
0.6 to 1
0.0
0 50 100 150
Time (days)
3
2 Analytical Solution
Simulation
2.5
Dimensionless Concentration
1.5
Dispersivity (ft)
2
αL
1
αT 1.5
0.5 1
0.5
0
10 20 30 40 50
0
Longitudinal Distance (ft) 0 50 100 150 200 250 300 350 400
Time (Days)
Fig. 4—Average of estimated local (gridblock) longitudinal dis-
persivity (aL) and transverse dispersivity (aT) across vertical Fig. 5—Analytical sample fit of local concentration history.
cross section. The matches validate the model. Concentration profile at y 5 4.75 ft and x 5 24 ft.
3.E–04 1.2
Concentration
Dispersivity (ft)
Vdp = 0.8
2.E–04 0.8 Vdp = 0.6
Vdp = 0.4
1.E–04
0.4
0.E+00
0 40 80 120 160 200 240 280 320 360 400
0
0 20 40 60 80
Time (Days)
Longitudinal Distance (ft)
1
Fig. 7—Average longitudinal dispersivity for uncorrelated het-
erogeneous medium. Dispersivity increases with distance trav-
eled and appears to approach an asymptotic value.
Concentration
Dispersivity (ft)
50 0.045
Vdp = 0.8
0.04 0.8 Vdp = 0.6
45
35
0.025
0
30 0 20 40 60 80
0.02
Longitudinal Distance (ft)
25
0.015
Fig. 10—Average longitudinal dispersivity for a medium with
20 0.01 permeability autocorrelation length of 25% of total in the x-
direction and 10% in the y-direction. Longitudinal dispersivity
15 0.005 increases with distance traveled and with levels of heterogene-
0 10 20 30 40 50 60 70
Length (ft) ity. Convective spreading at high level of heterogeneity can
dominate true mixing.
Normalized Concentration of 2-D CDE Simulation @ Day
19
LX = 0.0; LY = 0.0; VDP = 0.6 streak. A high-permeability streak results in convective spreading
55 0.03 of the miscible solute and reduced miscibility. This will be
noticed in field application with early solute breakthrough and
50
early production of high concentrations of the solute.
0.025
45
55
Effect of Input Dispersivity on Dispersion. Most conventional
8
compositional simulators allow for input dispersivity, which is
50 sometimes used as a fitting parameter. We investigated the effect
7
of input longitudinal and transverse dispersivity in an uncorrelated
45 medium. An uncorrelated medium was used to minimize the
6
effect of correlated regions on the estimated dispersivity.
40
Height (ft)
5
Results show that the effect of input longitudinal dispersivity
35 is to add to the original total dispersivity in the medium. This is
4
shown in Fig. 20, where the average estimated longitudinal dis-
30 persivity increased from 0.56 to 0.91 ft with an input of 0.5 ft.
3
This increase is caused essentially by the input dispersivity. The
25 effect of the input longitudinal dispersivity is to cause increased
20 2 mixing in the longitudinal direction. There is only a slight change
in the transverse dispersivity with input longitudinal dispersivity.
15 1 The effect of input transverse dispersivity is amplified spread-
0 10 20 30 40 50 60 70
ing in the transverse direction that is far greater than the input
Length (ft)
value. The average transverse dispersivity increased from 0.58 to
Fig. 9—Concentration distribution at 19 days superimposed on
2.6 ft with an input transverse dispersivity of 0.5 ft (Fig. 21). The
model streamlines (green lines). As heterogeneity increases, impact of input transverse dispersivity on the average longitudinal
streamlines become sinuous and transverse dispersion dispersivity is also not significant. As the solute particles spread
increases. to a new transverse position, the input local transverse dispersivity
causes the solute to spread farther transversely (Fig. 22). The esti-
is quite different. This difference is because the level of transverse mated vertical dispersivity is greater than the longitudinal disper-
dispersion is dependent on the structure of the flow paths, which sivity, indicating that transverse spreading dominates.
is quite different for each realization, as shown in Fig. 14. We also considered flow where the model has input longitudi-
For autocorrelated media, care must be taken in planning mis- nal and transverse dispersivity of 0.5 ft. The effect of transverse
cible enhanced-oil-recovery injection to avoid a high-permeability dispersivity dominates the model. Therefore, care must be taken
5.5
50
1.5 Vdp = 0.8
5
Vdp = 0.6 45
1 Vdp = 0.4 4.5
40
4
0.5
Height (ft)
35 3.5
0 3
0 20 40 60 80 30
60 4.5
55 4
Effect of Autocorrelation Lengths on Dispersion. We investi-
gated the effect of increasing autocorrelation in the longitudinal 50
3.5
direction for dimensionless correlation lengths of 0.25, 0.5, and 5
45
at the same dimensionless correlation in the y-direction of 0.1. 3
The higher value of 5 indicates a layered model. We also consid- 40
Height (ft)
60
Normalized Concentration of 2-D CDE Simulation @ Day 3.5
239 55
–4
LX = 0.25; LY = 0.1; VDP = 0.8 x 10
50 3
65 14
60 45
2.5
Height (ft)
12 40
55
2
35
50
10
30 1.5
45
Height (ft)
25
40 8 1
20
35
15 0.5
6 0 10 20 30 40 50 60 70
30
Length (ft)
25
4 Fig. 12—Concentration distribution at 99 days superimposed
20 on model streamlines (green lines) for correlated medium. As
heterogeneity increases, streamlines become sinuous and coa-
15 2 lesce into high-flow-rate regions. High-flow-rate regions can
0 10 20 30 40 50 60 70 enhance convective spreading.
Length (ft)
Fig. 13—Concentration contour for autocorrelated medium quickly with little mixing but then transitions to a new layer
showing trapped solute plume in slow-flow-rate region. This because of the correlation in the vertical direction. It is thus the
trapped solute plume usually results in tailing and asymmetry crossflow from one layer to another that is responsible, for the
in the effluent concentration. most part, for the mixing in a layered medium. The average
11 9
50 45
10
8
45 40
9
7
40 35
8
Height (ft)
Height (ft)
6
35 7 30
5
6
30 25
4
5
25 20
3
4
20 15 2
3
15 2 10 1
0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70
Length (ft) Length (ft)
Normalized Concentration of 2-D CDE Simulation @ Day Normalized Concentration of 2-D CDE Simulation @ Day
88 88
–3
LX = 0.25; LY = 0.1; VDP = 0.8; R4 –3 LX = 0.25; LY = 0.1; VDP = 0.8; R5 x 10
x 10
70 4 60 5.5
55 5
60 3.5
50 4.5
50 3 45 4
40 3.5
Height (ft)
Height (ft)
40 2.5
35 3
30 2 30 2.5
25 2
20 1.5
20 1.5
10 1
15 1
0 0.5 10 0.5
0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70
Length (ft) Length (ft)
Fig. 14—Streamlines of four different permeability realizations with the same autocorrelation and permeability variation. The flow
paths are obviously different and result in different transverse dispersivities.
transverse dispersivity for the case of a layered medium, which direction equilibrate more efficiently (Fig. 25). The transverse dis-
shows some scatter, is oscillating at approximately a constant av- persivity tracks the autocorrelation in the longitudinal direction
erage value (Fig. 24). (Fig. 26).
When the dimensionless correlation is 0.5, the longitudinal
dispersivity increases at a steep slope initially, and the slope
changes as it enters another region of correlation. The transverse Effect of Anisotropic Dispersivity on Upscaling. Conventional
dispersivity of each region is also different, as shown in Fig. 24. upscaling focuses on petrophysical upscaling without considera-
Increasing autocorrelation in the transverse direction reduces tion of mixing, especially for miscible displacement. Garmeh
longitudinal spreading because concentrations in the transverse
4
2
3
R2
Dispersivity (ft)
1.5
R1
Dispersivity (ft)
R1
R2 2 R3
1 R3 R4
R4 1 R5
0.5 R5
0
0 0 10 20 30 40 50 60 70
0 10 20 30 40 50 60 70 Longitudinal Distance (ft)
Longitudinal Distance (ft)
Fig. 16—Average transverse dispersivity for different realiza-
Fig. 15—Average longitudinal dispersivity for different realiza- tions of the same autocorrelation (25% in x-direction and 10%
tions of the same autocorrelation (25% in x-direction and 10% in in y-direction) and VDP 5 0.8. Transverse dispersivity depends
y-direction) and VDP 5 0.8. Longitudinal dispersivity increases on the nature of the flow path, which is different for each
in all realizations. realization.
0.08 1.2
Dispersivity (ft)
32.5 kv/kh = 0.0
0.07
kv/kh = 0.01
0.8
32 0.06
kv/kh = 0.1
Height (ft)
0.03
31 0
0.02 0 20 40 60 80
30.5 0.01 Longitudinal Distance (ft)
0 10 20 30 40 50 60 70
Length (ft)
Fig. 18—Average transverse dispersivity at different levels of
Normalized Concentration of 2-D CDE Simulation @ Day crossflow. Transverse dispersivity increases with increasing
99
LX = 0.25; LY = 0.01; VDP = 0.6; KVKH = 0.01
crossflow.
33.5 0.055
1.2
0.8
Dispersivity (ft)
0.8
Dispersivity (ft)
kv/kh = 0.0
0.4 kv/kh = 0.01
0.4
kv/kh = 0.1
kv/kh = 0.2 Input Long. Disp. = 0.0 ft
Input Long. Disp. = 0.5 ft.
0
0 0 10 20 30 40 50 60 70
0 20 40 60 80 Longitudinal Distance (ft)
Longitudinal Distance (ft)
Fig. 20—Comparison of average longitudinal dispersivity with
Fig. 19—Average longitudinal dispersivity at different levels of an input longitudinal dispersivity. The effect of input longitudi-
crossflow. Longitudinal dispersivity increases with increasing nal dispersivity is to increase the mixing in the longitudinal
crossflow. direction.
3 2
Dispersivity (ft) 50
1.8
45
1.6
2 Input Trans. Disp. = 0.0 ft
40
1.4
Input Trans. Disp. = 0.5 ft.
Height (ft)
35 1.2
1
1
30
0.8
25
0 0.6
0 10 20 30 40 50 60 70 20
Longitudinal Distance (ft) 0.4
15 0.2
0 10 20 30 40 50 60 70
Fig. 21—Comparison of average transverse dispersivity with an Length (ft)
input of transverse dispersivity of 0.5 ft. The effect of input
transverse dispersivity is to increase vertical spreading Normalized Concentration of 2-D CDE Simulation @ Day
significantly. 59
–3
LX = 0.0; LY = 0.0; VDP = 0.6 x 10
55 10
9
3. Dispersion increases with increasing heterogeneity. Heteroge- 50
Height (ft)
same correlation lengths, mean permeability, and Dykstra- 35
Parson coefficient may have different flow structures. This dif- 5
2 2
Lx = 0.25
Lx = 0.5
1.5 1.5
Dispersivity (ft)
Dispersivity (ft)
Lx = 5
1 1
Lx = 0.25
0.5 Lx = 0.5
Lx = 5 0.5
0
0 50 100 150 200 250 300 0
Longitudinal Distance (ft) 0 50 100 150 200 250 300
Longitudinal Distance (ft)
Fig. 23—Average longitudinal dispersivity for 10% autocorrela-
tion in the y-direction, but within a different dimensionless Fig. 24—Average transverse dispersivity for 10% autocorrela-
autocorrelation (Lx) in the x-direction. The slope of the disper- tion in the y-direction, but with a different dimensionless auto-
sivity changes at locations roughly equal to the autocorrelation correlation (Lx) in the x-direction. Different trends of transverse
length. mixing are noticed in different autocorrelated regions.
Dispersivity (ft)
Ly = 0.5
Dispersivity (ft)
1.5
1.5
1
Ly = 0.05
1 Ly = 0.1
0.5
Ly = 0.5
0.5 0
0 50 100 150 200 250 300
Longitudinal Distance (ft)
0
0 50 100 150 200 250 300
Fig. 26—Average transverse dispersivity for 25% autocorrela-
Longitudinal Distance (ft) tion in the x-direction, but with a different dimensionless auto-
correlation (Ly) in the y-direction. Transverse dispersivity
Fig. 25—Average longitudinal dispersivity for 25% autocorrela- tracks the autocorrelated regions in the longitudinal direction.
tion in the x-direction, but with a different dimensionless auto-
correlation (Ly) in the y-direction. Increasing autocorrelation in
the transverse direction suppresses growth of longitudinal dis- 6.00E–03
DIMENSIONLESS CONCENTRATION
persion. Changing slope of the longitudinal dispersivity corre-
sponds to different correlated regions.
512X64
4.00E–03
1.60E–03 32X8
DIMENSIONLESS CONCENTRATION
1.20E–03
512X64 2.00E–03
64X16
8.00E–04
0.00E+00
0 200 400 600 800 1000
4.00E–04 TIME (DAYS)