Professional Documents
Culture Documents
Condensate Recovery
A.S. Cullick, SPE, and H.S. Lu, Mobil E&P Technical Center; L.G. Jones, SPE, Consultant; and
M.F. Cohen, SPE, and J.P. Watson, SPE, Mobil E&P Technical Center
Summary. This paper shows that the water-alternating-gas (WAG) process may improve sweep efficiency and gas-condensate recov-
ery compared with continuous cycling in highly stratified reservoirs. The study used extensive numerical simulation to investigate the
sensitivity of the process to several variables, including reservoir layering, permeability, relative permeability, capillary pressure, and
trapped gas. The process mechanics were confirmed by laboratory displacements in layered core.
Introduction
Many WAG process applications have been proposed and applied The 2D cross-sectional model has a horizontal, 69-ft-thick, lO-md
to improve sweep efficiency of injected gas in miscible and immis- moderate-permeability stratum at the top; an 8-ft-thick, 100-md
cible floods in oil reservoirs. Use of WAG to improve sweep effi- high-permeability stratum in the middle; and a 23-ft-thick, I-md
ciency in a gas-cycling, pressure-maintenance process in a gas- low-permeability stratum at the bottom. Fig. 1 shows the "base-
condensate reservoir has not been reported. case" model [110 cells (22 x 5) and a 1,870-ft well-to-well distance]
Gas injected to maintain pressure in gas-condensate reservoirs representing an inverted five-spot well pattern (with constant y di-
can lead to early gas breakthrough, low sweep efficiency, disap- mension) of "" 160 acres.
pointing condensate yield, and high compression costs because of Continuous gas injection was 82,500 sefiD and reached a cumula-
gas channeling in high-permeability strata. In this computer simu- tive total of "" 1.22 HCPV after 23 years. Injection was balanced
lation study, we show that the WAG process improves gas sweep by production. WAG consisted of injecting water at a reservoir
and ultimate recovery. Water increases recovery by acting as a di- volume equal to the gas (240 days of 76-BID water slug injection)
verting agent by preferentially entering high-permeability channels alternately with the gas, beginning the first water cycle after initial
gas breakthrough. Each water slug was ",,0.035% HCPV, and the
and diverting injected dry gas to lower-permeability channels, by
gas/water ratio was 0.92. Fifteen WAG cycles were performed dur-
sweeping gas condensate out of the low-permeability strata through
ing the 23 years.
imbibition and because of the water's favorable mobility, and by
preferentially sweeping the lower part of the reservoir that is un-
Fluid and Rock Properties. Fluids representative of three typical
swept by gas. gas-cycling projects were chosen. Table 1 compares the fluid char-
In conventional practice, water is not injected into a gas-conden- acteristics with those in representative gas-cycling projects. The
sate reservoir because of the possibilities of losing reserves to gas condensate and dry gas were three-component synthetic fluids
trapped gas condensate, killing wells with water invasion, and reduc- of ethane, propane, and butane. They were assumed to be first-
ing injectivity. In contrast to a waterflood, in a WAG process, water contact miscible because the pressure was maintained above the gas-
follows and traps dry gas, not gas condensate; water production condensate dewpoint pressure; the hydrocarbons therefore remained
can be avoided by designing the process so that only small water single phase throughout the simulation (except for the blowdown
slugs and a small total water volume are used. Gas injectivity es- sensitivity case). In addition to relative permeability, fluid viscosity
sentially is restored after each water slug injection. and density ratios determine fluid flow characteristics (i.e., rela-
We present detailed results of a fully compositional reservoir tive mobility and gravity segregation).
simulation of a synthetic layered system and discuss the effects of For the base case, initial saturations were assumed to be 75%
reservoir and process parameters on WAG performance. We ad- gas condensate and 25 % water; trapped gas saturation by water was
dress concerns about potential adverse effects from water injection 28 %. The saturation endpoints and 0.1 water relative permeability
and present laboratory displacement data that demonstrate the proc- at trapped gas saturation used were reported by Chierici et al. 2
ess in a two-layer core with different permeabilities. to be representative of an outcrop limestone. Fig. 2 shows the rela-
tive permeability curve. Porosity was 12 %, and kv/kH was
Simulation ' assumed to be 0.5.
The purpose of the simulations is to use synthetic, prototype models
Base-Case Simulator Results. Fig. 3 shows the recovery curves
to study process mechanisms in general, not for a particular reser-
at 1 HCPV injection; WAG recovery is "" 78 % of original hydrocar-
voir. Our results from one cross-sectional model compare pressure- bon in place (OHIP) after 19.7 years compared with ",,61 % for
maintenance operations by continuous gas injection with those from gas injection after 19. 1 years (a 28 % increase. Initial gas break-
the WAG process. The model has three strata and approximately through is at "" 3.3 years (i.e., at 0.17 HCPV).
represents a 160-acre inverted five-spot pattern. We use a fully com- Relative permeability, viscosity (mobility), and gravity mecha-
positional simulator that incorporates a Peng-Robinson equation of nisms contribute to this improvement. Injected water preferentially
state (EOS) for fluid properties. The simulator was described enters the l00-md layer at the injection wellbore, reducing gas rela-
previously. 1 tive permeability in that stratum. Injected dry gas therefore is divert-
ed to the top lO-md stratum, where it sweeps that stratum. Because
Model Configuration. The model, called the' 'three-permeability of a favorable mobility ratio, injected water sweeps the l00-md
layer model," is a "layer-cake" model with three different per- stratum efficiently and displaces the dry gas from the gas-swept
meability layers (Fig. 1). The model is a prototype developed from regions; gas use efficiency is increased ~ignificantly with WAG.
reservoir kh core and well-log data that were averaged into three Water also crossflows into the bottom I-md layers because of gravi-
permeability strata (high, moderate, and low permeability) by taking ty. At 1 HCPV total injection, WAG required 40% less gas than
a geometric average within the permeability ranges. The high- continuous injection, but recovered 28 % more original condensate.
permeability layer is "" 8 % of the total thickness and has average Fig. 4 compares the fraction of gas condensate remaining at 1
permeability ratios of "" 100: 1 and"" 10: 1 compared with the low- HCPV injection for both cases and illustrates the superior sweep
and moderate-permeability strata, respectively. achieved with WAG. Almost 90% of the top layer is swept by dry
gas in WAG compared with"" 80% with continuous gas injection.
Copyright t 993 SOCiety of Petroleum Engineers In both cases, sweep is less efficient in the second layer because
Laboratory
10 MD 34.5 and Simulator Reservoir
Fluid Properties Model A B C
Viscosity
10 MD 34.5 P,H,olp, Rf 6.0 3.8 5.7 5.5
P, H,olp, /g 13.5 10.0 16.3 16.5
Density
11.5 p H,olp Rf 2.1 2.2 2.1 2.0
I MD 11.5 P H,olp /g 3.4 3.3 5.0 4.3
'""1--------1870 feet---------i.!
pares the average reservoir pressures for simulator runs with and
Fig. 1-Three-permeability layer model. without capillary pressure.
Because use of individual-stratum properties did not significantly
affect the results, we used the base-case properties for the other
of gas override. Gas channeling through the high-permeability stra-
sensitivity cases. (No runs were made with relative permeability
tum is predominant in the gas injection case but is reduced in WAG.
hysteresis assumed. A trapped-gas hysteresis potentially would low-
Although dry gas almost completely bypasses the I-md stratum in
er WAG recovery slightly, but the key result trends would not
both cases, water crossflows downward into the bottom I-md stra-
change.)
tum from the 100-md stratum. A stable water displacement occurs
Fine-Grid Model. To test the effect of the number of vertical-
in the lower stratum; consequently, recovery from the I-md stra-
direction grids on gas override prediction, we conducted simula-
tum is improved significantly.
tion runs using nine vertical grids instead of five by subdividing
Fig. 5 shows that the water front is still 340 ft away from the
the top lO-md strata vertically into six equal-thickness grids. Re-
producer at 1 HCPV injection and that no water has been produced.
sults from the model were almost identical to those from the base-
Its slow movement is the result of effective sweep, low water rela- case model.
tive permeability, and high viscosity. 3D Model. A 3D model showed recovery improvement with
WAG over continuous gas injection similar to that shown by the
Sensitivity Cases. In addition to the base-case model, we compared 2D model, a 23% increase in recovery at 1 HCPV injected com-
WAG with continuous gas for several model variations: an pared with that of continuous gas injection. The 3D model had the
individual-stratum-property model, a fine-grid model, a coarsening- same layering as the 2D model, 3 cells in the y dimension and 330
upward sequence, a coarsening-downward sequence, a limited total cells. The model simulated one-eighth of an inverted 160-acre
vertical-crossflow case, a 3D case, and a model with four layers five-spot pattern with 1,870 ft between injector and producer. The
with different permeabilities (the "four-permeability layer model"). injection and production conditions were the same as those used
Table 2 summarizes the recovery results. in the 2D study.
Individual-Stratum-Property Model. Each stratum was assigned Both gas injection and WAG achieved good areal sweep because
a relative permeability and a capillary pressure appropriate for its of favorable mobility displacements. The dry-gas mobility is about
permeability. Table 3 lists the individual-stratum residual satura- twice that of gas condensate. The areal sweep of a miscible dis-
tions. The relative permeability curves for each stratum were com- placement on a five-spot pattern with mobility ratio of two should
puted by normalizing the curves in Fig. 2 with the individual-stratum reach 75 % to 85 %,3 while that for water, which has a mobility
residual saturations. Fig. 6 shows the capillary pressure curve for ratio of "'" O. 1, should approach 100 % .4
the lO-md stratum. The same capillary pressure adjusted for the Coarsening-Upward Sequence. Simulation of a coarsening-
residual saturations was used for the other two strata. upward sequence (i.e., with the high-permeability stratum at the
Fig. 7 compares the simulation recovery curves for the continuous top of the model and permeability decreasing with depth) shows
gas and WAG ca~es. The results are essentially the same as those that gas cycling recovers only 38 % OHIP at 1 HCPV injected be-
for the cases where the residual saturations and relative permeabil- cause of severe gas override. Gas does not crossflow effectively
ities were the same for all strata. Fig. 8 compares the water pro- into the lower strata. Recovery improves by 42 % to 54 % OHIP
duction curves with and without capillary pressure. As expected, with WAG because of reduced gas channeling and increased recov-
water breaks through somewhat earlier when there is no capillary ery from the lO-md stratum. However, any incremental recovery
pressure to retain the water. In both cases, however, the water pro- from that stratum is from water displacement because little dry gas
duction remains low until the very end of the flood. Fig. 9 com- is diverted into the bottom I-md stratum. Therefore, water sweeps
1.0 .-------------....,.,,~---=:=,.,
0.9
0.8
0.8
0.7
0.6 0.6
RECOVERY 0.5
RELATIVE FRACTION OHIP
PERMEABILITY 0.4
0.4
0.3
0.2
0.2
0 0.1
0 0.2 0.4 0.6 O.S
0.0'~--L---L----'----'.-----L--~
WATER SATURATION 0.0 4.0 8.0 12.0 16.0 20.0 24.0
TIME (YEARS)
Fig. 2-Base-case relative permeability curve. Fig. 3-Three-permeability layer model recoveries.
l f
lLl
HIGH
PERM
LAYER
~~~MH~~~~nJ
c::::J 0 - 10% II!Il!IIi 51 - 90 %
the I-md stratum in addition to diverting gas from the 100-md lay- WAG as an Alternative to Early Blowdown
er, improving overall vertical sweep and recovery. WAG also may be an alternative to blowdown in older cycling
Coarsening-Downward Sequence. With the 100-md stratum at projects. To compare blowdown with WAG, we cycled gas for 15
the bottom of the model and permeability increasing with depth, years, or ""0.60 HCPV injected, using the base-case reservoir
at 1 HCPV, WAG improves vertical sweep and increases recovery model, followed by either 25 years of blowdown or 15 years of
25% compared with gas injection (61 % recovery for continuous WAG plus 10 years of blowdown.
gas and 78 % for WAG). A noticeable difference between this case To account for liquid-condensate-dropout effect, we used an EOS
and the base case is that the water front moves faster in the bottom model description for a reservoir fluid with"" 6, 170-psi dewpoint
100-md stratum. At 1 HCPV, the water front has moved to only pressure at 175°P that yields"" 175 bbl/MMscf at the surface. The
200 ft from the producer, compared with 340 ft away for the base fluid has a maximum liquid dropout of "" 40 vol % at 5,981 psia.
case. The top I-md stratum is not invaded efficiently by gas or water The Stone II model 5 was used to account for three-phase relative
in this case. permeability; critical condensate saturation was 0.30.
Crossflow (Vertical Permeability). If vertical crossflow is restrict- Fig. 10 shows that WAG improved condensate-liquid recovery
ed between reservoir strata, recovery improves a dramatic 54 % with by about 25 % over blowdown alone. In addition, ultimate gas recov-
WAG. With restricted crossflow, gas injection recovers only 46% ery was 10% higher at the end of the blowdown after WAG (gas
OHIP at 1 HCPV injection. Little recovery from the bottom I-md relative permeability hysteresis effects reported by Hawes et al. 6
stratum is obtained; however, with WAG, water flows preferen- were not included).
tially into the 100-md stratum and little goes into the I-md stra-
tum. Consequently, water breaks through after only 0.2 HCPV Laboratory Corelloods
water has been injected, with only 41 % OHIP recovered. Laboratory floods in a layered core demonstrated that WAG per-
To prevent water production, we stopped water injection after formance was analogous to that in the simulations. We construct-
five cycles (Le., at ""0.175 HCPV water injected). Gas injection ed a core of two 12-in.-long hemicylindrical outcrop limestone slabs:
was continued for 23 years. At 1 HCPV injection for 19.4 years, a lO-md air permeability, 26 % porosity Texas cream limestone and
almost 71 % OHIP is recovered, a 54% improvement compared with a 0.3-md air permeability, 11.3% porosity Indiana limestone. The
continuous gas injection. Although water production was not avoided
completely, only 167 bbl of water was produced (approximately TABLE 2-SIMULATION CASE SENSITIVITIES
0.04% of HCPV), indicating that water movement slowed signifi-
cantly when water injection stopped. Continuous Incremental
The injection profile improved significantly during the WAG Gas WAG WAG
Recovery' Recovery' Recovery
process. The amount of gas channeling through the l00-md stratum
Case (%OHIP) (% OHIP) (%)
was reduced from 65 % of that injected in the gas injection case
Base case 61 78 28
to 27% at the end of the WAG operation and to <2% at 1 HCPV
Coarsening upward 38 54 42
injected. Coarsening downward 62 78 25
Limited crossflow,
kv1kH =0.1 46 71 54
120 krw = 1 at Srg 61 81 33
3D, one-eighth of five-
100
spot pattern 61 75 23
80 Four-permeability
layer model 45 72 60
60
CAPILLARY
• At 1 HCPV injected.
PRESSURE 40
psi
20
0:-
0.8 80 t------I~~ 1-
J:
0
z 1346 psi.
0 0.6 CONDENSATE
§ RECOVERY
1156 plio
~ (0,. OF ORIGINAL) 40
!!:.
?iw: 0.4
BLOWDOWN
e;
0
20
5008 psis
w
a: 0~_66_7_3~ps_i.~____~__~_______ L_ _~_ _~
0.2 o 10 20 30 40
TIME (YEARS)
k(md)
HIGH k LAYER
10
LOW k LAYER
0.3
.-r-'-
POROSITY 26% 11.3 e/. 1.5"
TRAPPED GAS 29.S-/.
._._._._J._
GAS
SEPARATOR CHROM.
o~------~------~--------~------~~~--~
,oo...:::-JJ
o 5 10 15 20 25
TIME (YEARS)
2200 1.0
0.9
2000 -
0.8
WAG
~ 0.7
~ 1800 -
w 0.8
!!i RECOVERY,
~ 1600 -
FRACTIOr. 0.5
OHIP
"- 0.4
1400 - 0.3
0.2
1200
0 10 15 20 25 0.1
TIME (YEARS)
slabs were placed in a rubber sleeve to form a 2-in. -diameter core. Fig. 12-WAG vs. continuous-gas-injection recovery from
The lO-md layer contains 29 % of the PV. Fig. 11 is a schematic layered core, 10-md layer on bottom.
of the core end cross section. The interface between the layers was
filled with epoxy only at the core inlet. Fig. 11 also is a schematic in the Texas cream achieved essentially the same 90+ % recovery
of the coreflood apparatus. at 1.2 HCPV and the same breakthrough at 0.65 HCPV for linear
Fluids used were 5% brine, propane to represent gas conden- velocities of 0.05, 2, and 10 ftlD. Endpoint saturations of25% for
sate, and a 95% ethane/5% butane mix to represent dry gas. Fluid connate water and 28 % for trapped gas and '" 0.1 endpoint water
density and viscosity ratios at 1,500 psia and 150°F were approxi- relative permeability were measured for the Texas cream. The
mately the same as those for the simulation fluid (Table 1); gas trapped-gas saturation measured for the Indiana limestone was
displacement was first-contact miscible. The corefloods were con- 29.5%.
ducted at a nominal linear velocity of '" 2 ftlD, although actual fluid Initial fluid saturations in the layered core were established by
velocity depended on the degree of channeling or bypassing. We filling the core with brine, then flushing out the brine with up to
tested the injection rate effect and found that gasfloods conducted 25 PV of mineral oil. Water saturations after the oil flush were
0 .•
90
0.8
80
0.7
70
mol·;'
0.8
ORIGINAL 60
RECOVERY,
FLUID IN FRACTION 0.5
EFFLUENT 50 QHIP
0.4
40
0.3
30
0.2
20 . . . . . WAG
,. . . . WATER
~""""~t---t I---t~~
INJECTION
10
0 0.2 0.4 0.6 0.8 1.2
0.2 0.4 0.6 0.8 1.0 1.2
FRACTION TOTAL HCPV PRODUCED
FRACTION TOTAL HCPV PRODUCED
Fig. 13-Effluent-concentration history from layered core- Fig. 14-WAG vs. continuous-gas-injection recovery from
flood, 10-md layer on bottom. layered core, 10-md layer on top.
"" 27 % of total PV; additional water was not produced from the
100
core until WAG injection. The mineral oil was flushed with
Soltrol™ solvent followed by propane. We conducted four ex- 90
perimental floods: a continuous gasflood with the lO-md layer on 80
the top, a continuous gasflood with the 10 md-Iayer on the bottom,
70
a WAG flood with the lO-md layer on the top, and a WAG flood Dlole/.
with the lO-md layer on the bottom. The WAG floods used injec- ORIGINAL 60
tion cycles of a 0.05 HCPV gas slug alternating with a 0.05 HCPV FLUID IN 50
EFFLUENT
water slug. Fluids were produced at a constant I,500-psi back- 40
pressure.
30