Third Comparative Solution Project - Gas Cycling of Retrograde Condensate Reservoirs
Third Comparative Solution Project - Gas Cycling of Retrograde Condensate Reservoirs
Third Comparative Solution Project - Gas Cycling of Retrograde Condensate Reservoirs
Summary. Nine companies participated in this artificial modeling study of gas cycling in a rich retrograde-
gas-condensate reservoir. Surface oil rate predictions differ in the early years of cycling but agree better late
in cycling. The amount of condensate precipitated near the production well and its rate of evaporation varied
widely among participants. The explanation appears to be in K-value techniques used. Precomputed tables for
K values produced rapid and thorough removal of condensate during later years of cycling. Equation-of-state
(EOS) methods produced a stabilized condensate saturation sufficient to flow liquid during the greater part of
cycling, and the condensate never completely revaporized. We do not know which prediction is more nearly
correct because our PVT data did not cover the range of compositions that exists in this area of the reservoir
model.
Introduction
SPE conducted two earlier solution projects, 1,2 both de- because not all companies had facilities for simulating gas
signed to measure the state-of-the-art simulation capability plant processing with gas recycling in their composition-
for challenging and timely modeling problems. The first al simulators.
project involved a three-layer black-oil simulation with Nine companies responded to the invitation for partici-
gas injection into the top layer. 1 Both constant and vari- pation. Table 1 is a list of the participants in this project.
able bubblepoint pressure assumptions were used. Model Participant responses were well prepared and required a
predictions were in fair agreement. No simulator perform- minimum of discussion. We invited all the companies to
ance data (run times, timestep size, etc.) were given. use as many components as necessary for the accurate
Seven companies participated in the project. The second match of the PVT data and for the simulation of gas cy-
project was a study of water and gas coning with a radial cling. Companies were asked to give components actually
grid and 15 layers.2 Authors of the project felt that un- used in the reservoir model, how these components were
usual well rate variations and a high assumed solution characterized, and the match to the PVT data obtained
GOR contributed to the difficulty of the problem. Some with the components.
significant discrepancies in oil rate and pressure were ob- We first outline the problem specifications, including
tained. Eleven companies joined in the project. sufficient data for others who may wish to try the prob-
For the third comparative solution project, the Com- lem. the pertinent PVT data are given. We show each
mittee for the Numerical Simulation Symposium sought participant's components, the properties of these compo-
a compositional modeling problem. Numerical compari- nents, and the basic PVT match obtained. In many cases,
sons of the PVT data match were considered important. EOS methods were used exclusively, but in others, a com-
Speed of the simulators was not to be of major interest. bination of methods was applied. The results of the reser-
The problem we designed is the outcome of this fairly voir simulation are given and comparisons are shown
general request. Some features of interest in current pro- between companies for both cycling-strategy cases. Fi-
duction practice of pressure maintenance by gas injection nally, some facts regarding simulator performance are
are included. The results confirm the well-known trade- given, although this information was voluntary.
off between the timing of gas sales and the amount of con-
densate recovered. Several features of interest in a more Problem Statement
complete examination of production from gas-condensate The two major parts to a compositional model study are
reservoirs are ignored. These include the effects of near- the PVT data and the reservoir grid. For the PVT data,
well liquid saturation buildup on well productivity and of participants were supplied with a companion set of fluid
water encroachment and water production on hydrocar- analysis reports. The specification of the reservoir model
bon productivity. We did not address the role of numeri- is given in Tables 2 and 3 and the grid is shown in Fig.
cal dispersion. In addition, the surface process is 1. Note that the grid is 9x9x4 and symmetrical, indicat-
simplified and not representative of economical liquid ing that it would be possible to simulate half the indicated
recovery in typical offshore operations. We simplifiedthe grid. Most participants chose to model the full grid. Note
surface process to attract a larger number of participants also that the layers are homogeneous and of constant
porosity, but that permeability and thickness vary among
Copyright 1987 Society of Petroleum Engineers layers.
Journal of Petroleum Technology, August 1987 981
TABLE 1-COMPANIES PARTICIPATING IN THIRD TABLE 2-RESERVOIR GRID AND SATURATION
SPE COMPARATIVE SOLUTION PROJECT INPUT DATA
E~~~~~I~~
1.00 1.000 0.0
7360 II
7400 II
733011
7450 It Capillary pressure for gas/oil is assumed to be zero.
- - - DATUM = 7500 II (subsurface)
'Gas synthetically prepared in 1he laboratory, liquid is random condensate sample; gas and liquid not in equilibrium at 2,000 psia.
Deviation factor, Z
Equilibrium gas 0.803 0.798 0.802 0.830 0.877 0.924
Two-phase flow 0.803 0.774 0.748 0.730 0.703 0.642
Liquid Volume-
Pressure Relative (% saturated
(psig) Volume- volume)
6,000 2.2435
5,500 2.3454
5,000 2.4704
4,880 2.5043 0.0 (dewpoint)
4,800 2.5288 Trace
4,600 2.5946 0.1
4,400 2.6709 0.3
5000
4,000 2.8478 0.7
3,500 3.1570 1.4
3,000 3.5976 3.6
Fig. 2-Relative total volume in constant-composition ex- ~Re[ativevolumes and liquid volume percents are based
pansion at 200F (see Table 6). on the original hydrocarbon PV at 3,428 psig and 200F.
culations do not perform material balance on all 16 com- hood of 2,500 psi [17.2 MPa] with peak liquid volume
ponents. Table 16 indicates the component groups selected varying between about 18 and 22% of the initial (dew-
by each participant. point) gas volume. Actually, the reservoir models predict
Tables 17 through 25 give summary data for each com- liquid volumes higher than this value in the vicinity of
pany's representation of component properties and the ba- the production well because of convection of heavy end
sic PVT match obtained with this set of components. More products into this low-pressure area and subsequent depo-
detailed matches of PVT data are included in Figs. 2 sition. The increased heptanes-plus content leads to com-
through 6. positions and flash behavior not available in the laboratory
Fig. 2 shows pressure/volume data in constant- data provided. Results given later show disagreement in
composition expansion of the reservoir gas at 200F the predicted liquid buildup in this area, which we attrib-
[93C]. While there are some minor discrepancies at the ute to the absence of flash data for such compositions.
lowest pressures shown, there is rather good agreement Liquid yield by multistage surface separation of equi-
in the pressure range in which most of the gas cycling librium gas produced during constant-volume depletion
takes place, between 2,500 and 3,400 psi [17,2 and 23.4 is given in Fig. 4. Separator conditions for the problem
MPa]. differ slightly from the separator conditions in the labo-
Fig. 3 shows liquid dropout in constant-volume deple- ratory reports distributed to the participants in two areas:
tion. The greatest discrepancies occur in the neighbor- (1) the primary separator pressure is switched from 815
Core Elf
Component Arco Chevron Laboratories CMG Aquitaine Intercomp Marathon McCord-Lewis Petek
-- --- -- --
~+ t *t
CO 2 X X X X
~ * t *
X
N2 X X X
Cl X X X X
C2 X X X X X X
C3 X X X X X
C4
C5
Cs t t X
X
X
X
X
X
XX
XX
X
X
X
X
C7
Cs
Cg
C lO
C ll
Hl X
I X
1 t
X *
X
IX X
X
X
X
X
X
X
X
X
X
X X
H2 X X X X X
H3 X X X
H4 X X
H5 X X
Total number of
components 9 5 5 10 6 8 12 16 13
Total C s+ components H 2 H H 2 H 6 7 H
Total C 7 + components 3 H 2 4 H 5 H H 5
!ci
>'" p molecular weight is about 17. Differences in the swell-
ing characteristics obtained with these gases would be ex-
15
pected.
2. The reservoir pressure falls continuously with time
in both cases of interest. Thus volumetric behavior at pres-
500 1000 1500 2000 2500
sures above the initial reservoir pressure is unimportant
SCF/BBL: OF DEW POINT GAS in the context of the model.
Fig. 5-Relative total volume in swelling of reservoir gas
with lean hydrocarbon gas at 200F (see Table 11).
Reservoir Model Performance
Table 26 gives the initial surface fluids in place with mul-
tistage separation. Stock-tank oil rates for constant gas
sales rate and for deferred early gas sales are shown in
Figs. 7 and 8. The corresponding cumulative liquid pro-
0
0 3 4 5 6 7 S 9 10 11 12 13 14 15
YEARS OF PRODUCTION
duction for these cases is given in Figs. 9 and 10. All year- Fig. 7-Reservoir model stock-tank oil rate, Case 1.
ly production data were connected with straight-line
segments in Figs. 7 through 10. Most models were already
below the dewpoint pressure at 1 year of production, and
surface liquid rate had already dropped below initial rate. STOCK TANK OIL RATE - CASE 2
Pc Acentric
Tc Molecular Specific Tb
Gravity (OF)
Component ~ ~ Factor Weight
C1 673.1 343.3 0.0130 16.04 0.3250 201.0
C2 708.3 549.8 0.0986 30.07 0.4800 332.2
C3 617.4 665.8 0.1524 44.09 0.5077 416.0
iC 4 529.1 734.6 0.1848 58.12 0.5631 470.6
nC 4 550.7 765.4 0.2010 58.12 0.5844 490.8
iC s 483.5 828.7 0.2223 72.15 0.6248 541.8
nC s 489.5 845.6 0.2539 72.15 0.6312 556.6
C6 457.1 910.1 0.2806 84.00 0.6781 607.0
C7 432.2 969.6 0.3220 101.20 0.7100 657.0
C8 419.7 1,011.7 0.3495 114.60 0.7340 692.0
C9 391.6 1,064.8 0.3912 128.80 0.7570 740.0
C lO 365.0 1,118.9 0.4354 142.30 0.7800 790.0
C 11 359.2 1,154.7 0.4568 155.00 0.8010 820.0
C 12 + 348.0 1,219.7 0.4946 210.00 0.8380 875.0
CO 2 1,071.3 547.6 0.2250 44.01 0.4200 350.7
N2 492.3 227.2 0.0400 28.02 0.4800 139.6
Interaction Coefficients (with methane)
~ ~ ~ ~ ~ ~ ~ ~ ~
0.02813 0.03260 0.03596 0.03918 0.04240 0.04534 0.12860 0.10000 0.1000
PVT Methods
McCord-Lewis used Watson 16 characterization factors for each fraction with correlations
of Whitson and Haaland. 11,17,18 Boiling points, with some minor changes, came from Katz
and Firoozabadi. 19 Specific gravities and methane binary interaction coefficients of the
heavy ends were estimated from the Watson K factor. Lee_Kesler 2o ,21 correlations were
used for critical pressure and temperature and molecular weight.
Initialization Results
Initial wet gas in place, Bscf 26.52
Initial separator gas in place, Bscf 23.18
Initial stock-tank oil in place, MMSTB 3.56
Basic PVT Matc;h
Dewpoint pressure, psia 3,443
Dewpoint Z factor 0.803
Simulator Description
The McCord-Lewis simulator is based on a partial density model that includes condensa-
tion from the reservoir gas phase to the reservoir liquid phase. The basic assumption
is that each reservoir phase can be viewed as a binary mixture of its surface products,
termed partial densities, when forming any reservoir phase. The detailed discussion of
the model is not possible here, but it is important to note that it doesn't require K values
per se.
2100 2100
1800 1800
(D (D
t;; 1500 t:; 1500
:::;; A = ARea :::;; A = ARGO
1200 CH = CHEVRON 1200 CH = CHEVRON
eM = CMG eM = CMG
CL : CORE LAB CL = CORE LAB
900 900 E = ELF
E =ELF
I ~ INTERCOMP I = INTERCOMP
600 MA = MARATHON 600 MA s MARATHON
Ml = McCORD-LEWIS ML = McCORD-LEWIS
P = PETEK P = PETEK
3 4 5 6 7 8 9 10 11 12 13 14 1S 4 5 6 7 8 9 10 11 12 13 14 15
YEARS OF PRODUCTION YEARS OF PRODUCTION
Fig. 9-Cumulative reservoir model stock-tank oil pro- Fig. 10-Cumulative reservoir model stock-tank oil pro-
duced, Case 1. duced, Case 2.
Company Year Layer 1 Layer 2 Layer 3 Layer 4 Company Year Layer 1 Layer 2 Layer 3 Layer 4
Arco 1 3.06 3.20 3.12 2.78 Arco 1 0.97 0.86 0.73 0.51
5 3.57 7.80 10.22 7.32 5 1.40 3.62 5.30 3.36
10 2.12 6.51 10.78 5.61 10 1.31 5.18 9.93 4.94
15 1.87 5.36 8.55 4.71 15 1.20 4.43 7.98 4.20
Chevron 1 1.76 1.89 1.78 1.57 Chevron 1 0.55 0.46 0.39 0.34
5 3.32 6.82 8.96 6.94 5 1.04 2.55 3.63 2.71
10 2.10 6.87 11.89 6.79 10 1.26 5.37 10.99 6.01
15 2.18 6.78 11.01 6.58 15 1.33 5.52 10.28 5.77
Core Laboratories 1 1.1 1.1 0.8 0.5 Core Laboratories 1 0 0 0 0
5 3.1 7.6 9.6 7.0 5 0.9 3.1 5.0 2.9
10 1.9 7.0 10.8 5.9 10 0.9 5.2 10.0 4.9
15 1.5 5.6 8.5 4.6 15 0.8 4.2 7.8 3.9
CMG 1 2.27 2.48 2.37 2.03 CMG 1 0.0 0.0 0.0 0.0
5 2.73 7.95 11.08 8.37 5 0.69 2.44 4.20 2.95
10 1.26 5.56 11.27 6.43 10 0.51 3.56 9.70 5.71
15 1.10 4.55 8.44 5.23 15 0.45 3.13 7.44 4.60
Elf Aquitaine 1 3.34 3.60 3.59 3.06 Elf Aquitaine 1 0.76 0.76 0.63 0.36
5 3.24 7.82 10.30 7.72 5 1.23 3.91 6.70 4.20
10 1.56 6.55 10.84 5.83 10 0.75 4.66 9.89 5.02
15 1.31 5.28 8.65 4.74 15 0.66 3.92 7.95 4.24
Intercomp 1 0.78 0.81 0.75 0.62 Intercomp 1 0.10 0.08 0.03 0.01
5 2.74 6.58 8.97 6.57 5 0.66 1.83 3.03 1.94
10 1.52 5.55 9.89 5.42 10 0.76 3.77 8.76 4.55
15 - - - - 15 - - - -
Marathon 1 4.28 4.60 4.60 4.04 Marathon 1 1.89 1.98 1.94 1.57
5 4.76 10.70 13.10 10.40 5 2.40 6.08 9.43 6.82
10 2.55 8.16 11.94 7.01 10 1.71 6.40 11.45 5.73
15 1.91 6.24 9.22 5.25 15 1.35 4.94 8.79 4.35
McCord-Lewis 1 1.10 1.07 1.01 0.81 McCord-Lewis 1 0.0 0.0 0.0 0.0
5 4.68 9.10 11.90 8.72 5 1.33 3.05 5.15 3.48
10 2.87 6.30 11.42 5.98 10 1.10 3.43 9.30 4.62
15 2.80 5.72 9.90 5.28 15 1.00 3.38 8.40 4.20
Petek 1 3.34 3.55 3.40 2.87 Petek 1 0.08 0.01 0.02 0.0
5 3.58 8.26 11.15 8.72 5 1.16 3.32 5.76 4.07
10 1.73 6.18 11.01 6.29 10 0.89 4.39 9.91 5.54
15 1.45 5.04 8.75 5.24 15 0.76 3.76 8.01 4.61
where between the two. Discrepancies in pressure are af- common trends of relatively uniform saturations for the
fected by both wet-gas rate (determined by yield, Fig. 4, first year. For other times, Layer 3 (the tight layer) shows
and surface-liquid density) and assumed gas Z factors in high. saturation because little injected gas sweeps this
early years of cycling. layer. Conversely, Layer 1 (a high-permeability layer)
A partial compensation for this sensitivity to injection shows almost no liquid. Layers 2 and 4 are intermediate
and production-gas Z factors is numerical dispersion, in sweep efficiency.
which tends to smear out the initial molecular weight and Condensate saturation in Node (7,7,4) is shown in Figs.
Z-factor contrast between injection and production gases. 12 and 13. Most of the models achieve a fairly stable satu-
This and the merging of the participants' depletion match- ration of slightly more than irreducible oil saturation
es (Fig. 4) at pressures far below dewpoint pressure ex- (24 %). This indicates a condition of reservoir condensate
plain the near-parallel oil production rates in the advanced flow in this area. Before this time, liquid dropout in the
stages of cycling and blowdown. Another factor compen- low-pressure region strips liquid from the gas stream. This
sating for discrepancies in Z factors and separator fac- continues until a small liquid flow begins and the surface
tors is the reservoir response to falling pressure. Any yield stabilizes. The stabilized yield value depends on the
model with a high rate of decline in pressure produces mixing of injection gas with reservoir gas around the
a rapid loss in surface liquid yield. This reduces reser- producer and the contribution of the depleted area behind
voir voidage and tends to lessen subsequent pressure the producer to production.
decline. Later, during cycling, the condensate around the
Actual recovery efficiencies achieved by the models are producer is partly revaporized, and reservoir oil ceases
atypical of field values in view of the homogeneous na- to flow. Liquid yield is partly sustained as some heavy-
ture of the model grid. The results for liquid recovery end fractions continue to vaporize and are produced. What
are 55 to 74% of the initial oil (condensate) in place. The is perhaps surprising is the widely different predictions
incremental production achieved by gas-sales deferral is for oil saturation at advanced depletion levels in the
shown in Fig. 11. These exhibit a considerable range- models, ranging from 0 to more than 22%. We believe
from 3 to 8 % of the initial condensate in place. that this can be explained by the K values used.
Layer average oil saturations for selected times are We made two supplementary runs with COMPIII 23 for
given in Tables 27 and 28. Not surprisingly, these show Case 2 to demonstrate the importance of the K-value tech-
21
~ 18
~
~
o
~ 15 (f) 15
o
(f)
12 12
9
6
2 3 4 5 6 7 8 9 10 11 12 13 14 15 345678
YEARS OF PRODUCTION YEARS OF PRODUCTION
Fig. 12-Reservoir model condensate saturation in Block Fig. 14-Reservoir model condensate saturation in Block
(7,7,4), Case 1. (7,7,4) with precomputed K-value tables, Case 2.
9 9
_____c.!-___ _
3 4 5 6 7 8 9 10 11 12 13 14 15 2 3 4 5 6 7 8 9 10 11 12 13 14 15
YEARS OF PRODUCTION YEARS OF PRODUCTION
Fig. 13-Reservoir model condensate saturation in Block Fig. 15-Reservoir model condensate saturation in Block
(7,7,4), Case 2. (7,7,4) with in-line fugacity-based K values, Case 2.
Average
Timestep
Numerical Number of Size Average CPU Material Balance
Solution Computer Timesteps (days) per Timestep Error at 15 years
Company Method Used 1 2 2 1 2 1 2
3
Arco IMPES IBM 4341 325 16.8 121 2.0 x 10-
Chevron IMPES VAX-11/780 375 383 14.6 14.3 116 103 4.6x10- 3 8.0x10- 4
4
Core Laboratories IMPES CDC 6600 251 244 21.7 22.3 6.7 6.3 7.5x10- 6.6x10- 4
5
CMG IMPES Honeywell DPS 68 200 194 27.4 28.2 185.7 163 6.0 x 10- 3.1x10- 5
Elf Aquitaine IMPES IBM 3081 185 199 29.4 27.4 2.2 1x 10- 4
Intercomp IMPES Harris 800 128 114 39.9 44.8 66.4 67.1 5.6x10- 6
Marathon IMPES Burroughs B7900 365 347 14.7 15.4 8.0 7.5 5.5x 10- 4 5.7x10- 4
4
McCord-Lewis IMPES VAX-11/780 91 91 60.0 60.0 13.9 13.9 7.5 x 10-
Petek IMPES ND-560 5;19 509 10.5 10.7 168 192 1.1x10- 3 2.0 x 10- 3