Upload
imogen-miles
View
216
Download
0
Embed Size (px)
Citation preview
Classification: Internal Status: Draft
Simulation Challenges with WAG Injection
Presentation at FORCE WAG SeminarStavanger, 18 Mar 2009
Vilgeir DalenSenior Advisor, StatoilHydro
3
General
• The most important results of WAG simulations:
– (incremental) oil production
– gas retention (because of the impact on gas sales and/or gas import)
– location of remaining oil
• Simulation challenges will depend on:
– miscible vs immiscible
– degree of gravity domination
– sand/permeability distribution (e.g. massive vs layered vs labyrinthic)
• Typically, simulation would be done at (at least) two scales:
– high-resolution simulation of segment(s) (2D or 3D)
– low-resolution simulation of entire field (”full-field simulation”)
4
Simulated gain by WAG injection
• 7-comp EOS; close to MMP• 2D – 100x80 grid (10mx0.5m)• Homogeneous; except with some barriers handled as transmissibility multipliers (MULTZ)• Including hysteresis on krg
Difference between remaining oil (at std bet) in the grid cells
Difference between remaining oil (at std bet) in the grid cells
WAG vs WI Effect of increasing the gas rate of the WAG
(white means loss)
5
Simulated gain by WAG injection
• Same; except with the barriers handled as tight shales with some holes
Difference between remaining oil (at std bet) in the grid cells
Difference between remaining oil (at std bet) in the grid cells
WAG vs WI Effect of increasing the gas rate of the WAG
(white means loss)
6
WAG simulation challenges are related to both…
• Reservoir description (ref previous example)
– Size of attics
– Vertical communication (kv/kh, shales)
– Contrasts in horizontal permeability
– Impact of faults on attics, roofs and vertical communication
• Mechanistic parameters
– Relative permeability (3-phase, hysteresis, dep. on surface tension)
– Capillary pressure
– PVT (compositional)
– Diffusion/dispersion
• Resolution
– incl. the assumption of instantaneous equilibrium
7
PVT (EOS model)
• 7-8 components is usually a good compromise
• Important to match MMP and miscibility mechanism (usually C/V)
• Multi-contact experiments are considered useful
8
What about core to lithofacies scale?
• Laminations could have an impact on trapping of gas and apparent relperm in general.
• Fairly little is known for gas-oil and 3-phase flow – more for water-oil.
SCAL dataCore plug data
Pore scale results
3D Permeability model
Outcrop dataCore data
2 m
3D Lithofacies model
Ripple Planar Trough
Lithofacies scale results
9
Establishing high-res model(s)
• What part of the reservoir to pick? How many models?
• 2D or 3D? How large model(s)? Grid?
• How accurate ”boundary conditions” (in a broad sense)?
• Grid-refining a full-field model may retain history matching features
• Geomodel has more hetero-geneities but are they sufficient?
• WAG simulation may require a renewed look into thief zones, shale distribution and baffles for vertical flow from core and log data
Geodata
Geomodel
Simulation model
History-matched simulation model
Upscaling
HM
?
High-res simulation model
?
?
?
10
On grids for high-res sector models
• “Optimal” grid resolution for compositional simulation models:
– Lateral resolution ~10 meters; vertical resolution < 1 meter
– For gravity-dominated cases, the sensitivity to the vertical resolution is stronger than the lateral resolution
• Prudhoe Bay Gravity Drainage Miscible Injection Pilot (Waldren, SPE-101455)
– Lateral resolution in pilot area ~10 meters (implemented by LGR)
• WAG Pilot Hassi Berkine South Field (Lo et al., SPE-84076)
– Lateral resolution ~10 meters, vertical resolution 0.375 meter
– Vertical resolution above 1 meter did not match gas breakthrough and saturation profiles
• Ongoing Snorre work: ~1 meter vertically; ~50m horizontally
11
Common black-oil alternatives (low-res)
• BO with swelling (variable Rs); possibly with a limited DRSDT which
– reduces the recovery effect of the gas
– capture in a sense limited mixing of the gas and oil in large grid cells
– do not capture the full ”cycle” of miscible flooding
– hysteresis for gas required to capture gas retention
• BO with swelling and vaporization;
– Presently both DRSDT and DRVDT can be specified in ECLIPSE
– Caution required not to vaporize too much oil too fast
– What we see in compositional simulation is that vaporization potential has like an exponential decline
• Should tuning on high-res be through DRSDT (and DRVDT) and/or relperm?
DRSDT = dRs/dt. A value of e.g. 0.1 Sm3/Sm3/day means that it will take at least 1 year for Rs to increase from 100 to 136.5 Sm3/Sm3 in a grid cell.
12
Cell 11-20
Cell 61-70
Cell 40
Cell 100
0 2 4 6 8 10 12 140
0.5
1.0
1.5
2.0
2.5
RO
SA
T,
RG
SA
T a
nd
R_
BO
0
50
100
150
200
250
300
350
400
450
500
R_
RS
0
0.0002
0.0004
0.0006
0.0008
0.0010R
_R
V
ONE100_WAG_E100 ROSAT RGSAT R_BO R_RS R_RV
ROSAT 2 RGSAT 2 R_BO 2 R_RS 2 R_RV 2
0 2 4 6 8 10 12 140
0.5
1.0
1.5
2.0
2.5
RO
SA
T,
RG
SA
T a
nd
R_
BO
0
50
100
150
200
250
300
350
400
450
500
R_
RS
0
0.0002
0.0004
0.0006
0.0008
0.0010
R_
RV
ONE100_WAG ROSAT RGSAT R_BO R_RS R_RV
ROSAT 2 RGSAT 2 R_BO 2 R_RS 2 R_RV 2
Black-oil run
Compositional run
Bo
Bo
Rs
Rs
Sg
So
So
Sg
Rv
Cell 40 vs time (years)
13
More ”black-oil” alternatives
• Todd-Longstaff (and other miscible options)
– Simple and easy (with the mixing parameter omega in some sense corresponding to DRSDT)
– Originally formulated to capture viscous fingering, but can be looked upon as a simplified representation of a more general ”mixing zone” between virgin oil in front and remaining, stripped oil behind a miscible front/zone.
– Interpolation between miscible and immiscible conditions can be done.
– Could be an alternative for screening studies, but generally not flexible enough for other cases.
• The GI-option; with Rs and Rv correlated with the amount of gas that has flowed through a grid cell.
– Obsolete; not recommended
14
Streamline simulation
• Mechanistic high-res simulation in combination with stream-line simulation is an alternative.
• Results will be highly dependent on the type-curves generated from 2D or 3D segments
• May be difficult to distinguish between acceleration and increased recovery effects
• Probably best suited for a fixed, regular well pattern
15
Concluding remarks
• Simulation of WAG injection is a real challenge
• A combination of segment (high-res) and full-field simulation is regarded as the best approach
– even if today’s computing power can permit millions of grid cells
– thin layers below shales are especially important if the high-res step is skipped
• For typical NCS WAG injection schemes, geometry and heterogeneities are the largest challenges
• Injectivity and ”smart well” issues may represent additional challenges