in salah high resolution heterogeneous simulations of co2

In SalahHigh Resolution HeterogeneousSimulations of CO2 Storage

Andrew Cavanagh, HalliburtonPhilip Ringrose, Statoil ASA

GHGT10, Amsterdam

CSLF Global achievement award for In Salah in 2011

'One of the most important industrial scale CCS initiatives... helping to counter the view that the technology is unproven.'Carbon Storage Leadership Forum, Beijing 2011

'People oppose what they do not understand.'Al-Ghazali, 1058-1111

CSLF Global achievement award for In Salah in 2011

JIP Phase 1 – Eight Important Lessons

6. Injection and pressure management needto be linked to geo-mechanical modelling...

- requires high-resolution data.- requires advanced geo-mechanical models.- fluid dynamics, geochemistry, temperature.

CSLF Global achievement award for In Salah in 2011

JIP Phase 1 – Eight Important Lessons

JIP Phase 2 – Five aims

1. advanced coupled geomechanical modelling.2. dual porosity-permeability fracture models.3. surface deformation coupled to hydro-geomechanics.4. three-component deformation modelling.5. microseismic and dedicated caprock data wells.

Mona Lisa: 120,000 pixels

1

In SalahHigh Resolution Heterogeneous Simulations...

Mona Lisa: 120,000 pixels Mona Lisa: 120 pixels

1

In SalahHigh Resolution Heterogeneous Simulations...

(Iding & Ringrose, 2008) Mona Lisa: 120 pixels

1

In SalahHigh Resolution Heterogeneous Simulations...

In SalahHigh Resolution Heterogeneous Simulations...

● Context, megatonne CO2 Storage Sites

● Overview of In Salah

● Modelling Approach- Flow domains- Capillary numbers

● Field Scale Model

● Simulations- Pilot- Faulted- Faulted and Fractured

● Comparison with Observations Mona Lisa, Leonardo da Vinci, c. 1519

1

Weyburn-Midale: 50x40 km

Sleipner: 300x400 km

In Salah: 20x18 kmInjection depth Sleipner: 1000 mSleipner: 1000 mWeyburn: 1400 mWeyburn: 1400 mIn Salah: 1900 mIn Salah: 1900 m

Gorgon: Gorgon: 2300 m 2300 m

3 Mt3 Mt

22 Mt22 Mt

13 Mt13 Mt

IEA GHG

Megatonne Storage Sites

2

Tops •• Cretaceous Superieur, siltstones and limestones 0 m

•• Pan Saharan Aquifer, sandstones, mudstones, carbonates 170 m

•• Regional Hercynian U/C, top of Visean, overlain by thin anhydrite 940 m

•• C20-7... fractured mudstones with strong gamma ray response (Visean, Carb) 1590 m •• C20-3... fractured mudstones, overpressured (Visean, Carb) 1790 m •• C20-2... fractured mudstone with open fracture sets, unstable (Visean, Carb) 1820 m •• C20-1... fractured mudstone thin dolomites and siltstones (Visean, Carb) 1860 m •• C10-3... fractured siltstones and sandstones (Tournasian, Carb) 1885 m •• C10-2... fractured sandstone, under-pressured at 17.5 MPa (Tournasian, Carb) 1900 m

•• D70... dolomitic sandstone with thin silt and mudstones (Fammenian, Dev) 1920 m

Ages U Cretaceous 100 - 65 Ma Sakmarian 288 - 283 Ma Visean 342 - 327 Ma Tournasian 354 - 342 Ma Fammenian 364 - 354 Ma Northern Gondwana

0.5

1.0

1.5

km

In Salah Stratigraphy

••

3

Sandstone: massive, rippled, fracture-influenced, matrix-dominated Facies: tidal deltaic, deposited in a broad palaeovalleyPorosity: ~15% (13- 20)Permeability: ~10 md (0.1 - 300)

FaultsFault throw less than 20 metersNo faults cut the reservoir communicationE-W faults easiest to map, N-S faults, NE-SW faults Fault mapsProlific and sparse fault models A) Prolific

B) SparseFracturesσ1-σ2 plane orientation NW-SE caprock: 1-3/meter, aperture: 0.1-2 mm, length: 25-50 mConforms to recent stress field reservoir: 2-3/meter, aperture: 0.1-1 mm, length: 6-25 m

Wells, KB-501, -502, -503Injection rate: 0.2 Mt/yr/well... 30 Mmscf/day... 8 litres/well/secondHorizontal: 1.3 - 1.5 kmAzimuth: perpendicular to stress fieldObservations: logged, cored and monitored

Storage Site Observations3D seismic baseline, downhole geophysics, 4D seismic monitoring,InSAR: Interferometry Satellite Airborne Radar, geochemical tracers...

A

B

In Salah Reservoir

4

A

B

In Salah Models for Geomechanics and Flow

2008Geomechanical uplift with ABACUS (Vasco et al., 2008)

Overburden migration model with PERMEDIA (Ringrose et al., 2008)

Discrete fracture network model with FRACA (Ringrose et al., 2008)

Reservoir simulation with ECLIPSE E300 (Iding et al., 2008)

Reservoir simulation with STARS (Bissell et al., 2008)

2011Coupled reservoir simulation with NUFT-SYNEF (Morris et al., 2011)

Reservoir model STARS with geomechanical inversion analysis (Bissell et al., 2011)

Reservoir simulation with PERMEDIA (Cavanagh & Ringrose, 2011)

Coupled surface-to-basement with TOUGH-FLAC (Rutqvist, 2011)

5

In Salah Observations by InSAR

max change: 2 cm in 2 yearsmean change: 0.5 cm/year

Pinnacle data for surface changes from winter 2003 to spring 2008 6

A

B

Modelling ApproachFlow Domains and Dimensionless Numbers

Flow Domain Flux rate; occurrence Dimensionless numberTurbulent flow Very high; near-well, unusual Re > 10--- --- ---Darcy flow High-to-moderate; near-well Re < 10, Ca > 0.0001--- --- ---Capillary flow Low; reservoir and basin-scale Ca < 0.0001, Kn < 1--- --- ---Knudsen flow Extremely high; unconventional Kn > 1

Key: Re, Reynolds No.; Ca, Capillary No.; Kn, Knudsen No.

7

Modelling ApproachCapillary Number Calculations

Ca = μq/γ [/]

μ, viscosity [Pa.s] [N/m2].[s]q, flux [m/s]γ, interfacial tension [N/m]

Capillary flow dominates when Ca < 1/10,000

(Henri Bouassé, Capillarity and Wetting Phenomena, 1924)(Pierre-Gilles de Gennes, Brochard-Wyart & Quere, ibid., 2003)

● In Salah injection, Ca ~E-06 (8 litres/well/second)

● In Salah migration, Ca ~E-07 (1.3 km/2 years)

8

12 million cells, 17.5x25 km, 20x20x2 m

Field Scale Model, Capillary Flow

8

12 million cells, 17.5x25 km, 20x20x2 m

Field Scale Model, Capillary Flow

8

Stochastic analysis of probable migration paths (N=120)

Simulations Migration beneath the Caprock

9

Stochastic analysis for KB-501 (N=55)

Southern AreaSimulations

10

Stochastic analysis for KB-502 & KB-503 (N=55)

Northern AreaSimulations

10

Prolific fault scenario (BP seismic mapping team)

Migration with FaultsSimulations

11

5-10-15-20 meter column height (frequency analysis, N=60)

Migration with FaultsSimulations

11

5-10-15-20 meter column height (frequency analysis, N=60)

Migration with FaultsSimulations

11

5-10-15-20 meter column height (frequency analysis, N=60)

Migration with FaultsSimulations

11

5-10-15-20 meter column height (frequency analysis, N=60)

Migration with FaultsSimulations

11

5-10-15-20 meter column height (frequency analysis, N=60)

Migration with FaultsSimulations

11

Comparison of field-scale migration with and without faults

SimulationsSimulations

11

Map input Curvature Analysis Edge Detection

Fracture Model

12

Northern area contour map Log-normal fracture distribution Contour map after fabric added

Fracture Model

12

Migration with Fractures

Reservoir with fabric added

Simulations...

13

Migration with Fractures

Before After

Simulations...

13

Northern Area early filling sequence

Comparison to Observations

14

Northern Area early filling sequence

Ringrose et al. (EAGE, 2008)

Comparison to Observations

14

Northern Area early filling sequence (N=20)

Simulations

15

Northern Area early filling sequence (N=20)

Simulations

15

Northern Area early filling sequence (N=20)

Simulations

15

Northern Area early filling sequence (N=20)

Simulations

15

Northern Area early filling sequence (N=20)

Simulations

15

Northern Area early filling sequence (N=20)

Simulations

15

Northern Area early filling sequence (N=20)

Simulations

15

Match to expected CO2 distribution

17Comparison to Observations

Ringrose et al. (EAGE, 2009)

16

Topography Faults Fractures Fractured and Faulted

Question: why model at low resolution?

17

Topography Faults Fractures Fractured and Faulted

Question: why model at low resolution...

...when heterogeneity matters?

17

Because we need geomechanics to explain uplift...

18

Because we need geomechanics to explain uplift and flow

18

Challenges: are coupled reservoir simulators the way forward?

Mesh design and resolution issuesSurface-to-reservoir problem requiresa large spatial domain, and sufficientgeological detail.

Even with scottish meshing, typical cell resolutions are in the range 100-1000 m.

(Morris et al., 2011)

17

Challenges: are coupled reservoir simulators the way forward?

Mesh design and resolution issuesSurface-to-reservoir problem requiresa large spatial domain, and sufficientgeological detail.

Even with scottish meshing, typical cell resolutions are in the range 100-1000 m.

(Rutqvist, 2011)

17

Challenges: are coupled reservoir simulators the way forward?

Mesh design and resolution issuesSurface-to-reservoir problem requiresa large spatial domain, and sufficientgeological detail.

Even with scottish meshing, typical cell resolutions are in the range 100-1000 m.

(Rutqvist, 2011)Pressure and flow issuesAt 50x50 m2 for the detailed sectors,the pressure field and saturationare distributed on both sides of KB-502.

Large cells cause numerical dispersionof pressure and flow.

(Iding & Ringrose, 2008)17

Challenges: are coupled reservoir simulators the way forward?

Mesh design and resolution issuesSurface-to-reservoir problem requiresa large spatial domain, and sufficientgeological detail.

Even with scottish meshing, typical cell resolutions are in the range 100-1000 m.

(Rutqvist, 2011)Pressure and flow issuesAt 50x50 m2 for the detailed sectors,the pressure field and saturationare distributed on both sides of KB-502.

Large cells cause numerical dispersionof pressure and flow.

(Durucan et al., 2011)17

Challenges: are coupled reservoir simulators the way forward?

Mesh design and resolution issuesSurface-to-reservoir problem requiresa large spatial domain, and sufficientgeological detail.

Even with scottish meshing, typical cell resolutions are in the range 100-1000 m.

(Rutqvist, 2011)Pressure and flow issuesAt 50x50 m2 for the detailed sectors,the pressure field and saturationare distributed on both sides of KB-502.

Large cells cause numerical dispersionof pressure and flow.

Complexity of couplingAre CO2 distribution and pressure coupled?Is there a strong geochemical component?

(Cavanagh & Ringrose, 2010)17

Conclusions: coupled geomechanics – difficult to crack

Simulation as an aid for engineering/operational decisions matures over a decade.The highest operational pressures and geomechanical risk occur within months-to-years.

Inversion modeling Behaviour in, properties out:Typically non-unique. Uncertainty remains high.

Forward modeling Properties in, behaviour out:Data models are sparse.Fracture permeability is dynamic.

Simulation outcomes Retrospective, not predictive:Bring understanding but no definitive forecast.

Reservoir conditions Exception rather than the rule?Pemeability is marginal at 10-20 mD.

Geochemistry: A wild card:A camel is stronger, an elephant wiser. Al-Ghazali, 1058-1111

18

Ira Ojala,Geophysicist,specialisingin reservoirgeomechanicsand skepticism.

Ira's Proof...

First, we are all aware that models require time and money:

And, we have already proven that time = money:

THEREFORE:

And, we all know that money is the root of all evil:

THEREFORE:

And thus,

19

Ira Ojala,Geophysicist

Acknowledgements

Neil Wildgust, IEAGHG Programme

Phil Ringrose