Upload
others
View
5
Download
4
Embed Size (px)
Citation preview
Geophysical Modelling for CO2 Storage, Monitoring and
Appraisal 2nd-3rd November 2015, London
www.ukccsrc.ac.uk
Agenda.................................................................................................................................................................................................... 2Delegate list.................................................................................................................................................................................................... 3Jonny Rutqvist - Modelling Fault Reactivation, Induced Seismicity....................................................................................................................................................................................................... 4Andrew Cavanagh - The sensitivity of CO2 storage simulations to....................................................................................................................................................................................................... 29Anna Stork - Passive seismic monitoring of CO2 storage sites.................................................................................................................................................................................................... 39Giorgos Papageorgiou - Advances in rock physics modelling....................................................................................................................................................................................................... 65Doug Angus - Assessing uncertainty of time-lapse seismic response....................................................................................................................................................................................................... 95Adriana Paluszny - Numerical modelling of fracture growth and caprock integrity....................................................................................................................................................................................................... 119
AGENDA 2nd November 2015 19:00 Onwards Networking dinner at VERITAS, 43-47 Great George St, City Centre, Leeds LS1 3BB 3rd November 2015 08:50 - 09:10 Registration with coffee 09:10 - 09:15 Welcome and Introduction 09:15 - 10:30 Session 1 - Chaired by Tom Lynch and Claire Birnie, University of Leeds
Modelling Fault Reactivation, Induced Seismicity, and Leakage during Underground CO2 Injection - Jonny Rutqvist (Lawrence Berkeley National Laboratory)
20 years and 20 Mt: Statoil storage experience - Andrew Cavanagh (Statoil Research)
10:30 - 11:00 Tea and Coffee break 11:00 - 12:30 Session 2 – Chaired by Claire Birnie, University of Leeds
Passive seismic monitoring for CO2 storage sites - Anna Stork (University of Bristol)
Advances in rock physics modelling and improved estimation of CO2 saturation - Giorgos Papageorgiou (University of Edinburgh/DISECCS)
Assessing uncertainty of time-lapse seismic response due to geomechanical deformation – Doug Angus (University of Leeds)
12:30 - 13:30 Lunch 13:30 - 15:00 Session 3 - Chaired by Tom Lynch, University of Leeds
Coupled flow, geomechanical and geophysical modelling: software tools and research gaps - Quentin Fisher (University of Leeds)
Monitoring, mapping and modelling thin layers of injected CO2 - Jim White (BGS)
Numerical modelling of fracture growth and caprock integrity during CO2 injection Adriana Paluszny Imperial College London/EPSRC CONTAIN project
15:00 - 17:00 Poster Session (ECRs) and Networking
DELEGATE LIST
First name Last name Institution/Organisation Juan Alcalde University of Edinburgh Mohammed Dahiru Aminu Cranfield University Doug Angus University of Leeds Domenico Bau University of Sheffield Hannah Bentham University of Leeds Claire Birnie University of Leeds Emilie Brady UKCCSRC Andrew Cavanagh Statoil ASA Laurence Cowton University of Cambridge Rami Eid University of Edinburgh Ismael Falcon-Suarez National Oceanography Centre, Southampton Quentin Fisher University of Leeds Mark Kelman Consulting Piroska Lorinczi School of Earth and Environment, University of Leeds Tom Lynch University of Leeds Rizgar Maolod University of Leeds John Midgley Energy Geoscience Andy Nowacki University of Leeds Opeyemi Oyewole Adriana Paluszny Imperial College London Giorgos Papageorgiou University of Edinburgh Sam Parsons University of Leeds Kazeem Rabiu Loughborough University Montserrat Recasens Heriot-Watt University Lisa Roach University of Leeds Jonny Rutqvist Lawrennce Berkeley National Laboratory Yong Sheng University of Leeds Anna Stork University of Bristol James White British Geological Survey
UK Carbon Capture and Storage Research Centre (UKCCSRC) The UKCCSRC brings together over 1000 members including over 200 of the UK’s world-class CCS academics to provide a national focal point for CCS research and development. The Centre is a virtual network where academics, industry, regulators and others in the sector collaborate to analyse problems devise and carry out world-leading research and share delivery, thus maximising impact. A key priority is supporting the UK economy by driving an integrated research programme and building research capacity that is focused on maximising the contribution of CCS to a low-carbon energy system for the UK. The UKCCSRC is supported by the Engineering and Physical Sciences Research Council (EPSRC) www.epsrc.ac.uk as part of the Research Councils UK Energy Programme, with additional funding from the Department of Energy and Climate Change (DECC) www.decc.gov.uk for the UKCCSRC PACT Facilities www.pact.ac.uk
www.ukccsrc.ac.uk
EARTH SCIENCES DIVISION • LAWRENCE BERKELEY NATIONAL LABORATORY
Modeling Fault Reactivation, Induced Seismicity, and Leakage during Underground
CO2 Injection
Jonny Rutqvist Antonio Rinaldi, Frederic Cappa
Lawrence Berkeley National Laboratory Berkeley, California
UK Carbon Capture and Storage Research Centre (UKCCSRC) specialist meeting , November 3, 2015
EARTH SCIENCES DIVISION • LAWRENCE BERKELEY NATIONAL LABORATORY
Geomechanics of CO2 Storage in Deep Sedimentary Formations
[Rutqvist (2012) Int J Geotechnical and Geological Engineering]
EARTH SCIENCES DIVISION • LAWRENCE BERKELEY NATIONAL LABORATORY
Geomechanics of CO2 Storage in Deep Sedimentary Formations
[Rutqvist (2012) Int J Geotechnical and Geological Engineering]
EARTH SCIENCES DIVISION • LAWRENCE BERKELEY NATIONAL LABORATORY
Geomechanics of CO2 Storage in Deep Sedimentary Formations
[Rutqvist (2012) Int J Geotechnical and Geological Engineering]
EARTH SCIENCES DIVISION • LAWRENCE BERKELEY NATIONAL LABORATORY
Potential Fault Reactivation and Notable Seismic Events
• An important issue from safety, storage security, and public acceptance perspectives.
• Release of stored energy triggered by the injection.
• Not just limited to seismically active areas, but could also occur within the seismically quiet intraplate crust (Zoback and Gorelick., 2012).
• Undetected minor faults relevant
[Rutqvist (2012) Int J Geotechnical and Geological Engineering]
EARTH SCIENCES DIVISION • LAWRENCE BERKELEY NATIONAL LABORATORY 6
Outline of Presentation
• Introduction • Modeling approach
• CO2 injection and fault activation
- Potential magnitudes? - Potential leakage?
• Deep fracture/fault responses at In Salah
• Concluding remarks
EARTH SCIENCES DIVISION • LAWRENCE BERKELEY NATIONAL LABORATORY 7
Modeling Fault Reactivation and Seismicity Typical Fault Discretization
Typical Model Domain • Anisotropic plasticity model allowing shear (Coulomb)
failure along the fault plane
• Shear-induce fault permeability change
• Strain-softening plasticity to represent slip-weakening fault behavior (sudden slip)
• Seismic moment and moment magnitude calculated from Kanamori et al (e.g. M0 = µAd)
Strain-softening Fault FLAC3D Geomechanical Simulator
TOUGHMultiphase Flow
Simulator
FLAC3D Geomechanical Simulator
TOUGHMultiphase Flow
Simulator
TOUGH-FLAC Simulator
EARTH SCIENCES DIVISION • LAWRENCE BERKELEY NATIONAL LABORATORY
Simulated CO2 Injection and Fault Activation
• Reactivation at about 7.5 MPa overpressure
• 4 cm fault slip over 0.4 seconds, peak slip 0.6 m/s
• 290 m fault rupture corresponding to Mw = 2.53
(Cappa and Rutqvist, GJI, 2012)
Reservoir 7.5 MPa
Overpressure
EARTH SCIENCES DIVISION • LAWRENCE BERKELEY NATIONAL LABORATORY
no top soil 50 m top soil
100 m top soil
Top soil
• PGV 30 mm/s at 6-12 Hz
• PGV for one jolt at a lower frequency
9
Ground Surface Motion at Top of the Fault
• PGA 0.6g at 30-40 Hz
• High frequency acceleration damped for soil
(Rutqvist et al., IJGGC, 2014)
EARTH SCIENCES DIVISION • LAWRENCE BERKELEY NATIONAL LABORATORY
Building Damage and Human Perception US Bureau of Mines (USBM) ground vibration criteria for building damage and human-perception limits for vibration
Rutqvist et al 2014 (Int. J Greenhouse Gas Control)
USBM RI 8507
OSM REGULATIONS
HUMAN PERCEPTION OF STEADY STATE VIBRATION
PERCEPTIBLE
UNPLEASANT
INTOLERABLE
DRYWALL (19.1 mm/s)
PLASTER (12.4 mm/s)
PEA
K P
AR
TIC
LE V
ELO
CIT
Y (m
m/s
)
FREQUENCY (Hz)
PEA
K P
AR
TIC
LE V
ELO
CIT
Y (ip
s)
(50.8 mm/s)
1 10 100
100
10
1
1.0
0.1
0.01
10.0
0.1
USBM RI 8507
OSM REGULATIONS
HUMAN PERCEPTION OF STEADY STATE VIBRATION
PERCEPTIBLE
UNPLEASANT
INTOLERABLE
DRYWALL (19.1 mm/s)
PLASTER (12.4 mm/s)
PEA
K P
AR
TIC
LE V
ELO
CIT
Y (m
m/s
)
FREQUENCY (Hz)
PEA
K P
AR
TIC
LE V
ELO
CIT
Y (ip
s)
(50.8 mm/s)
1 10 100
100
10
1
1.0
0.1
0.01
10.0
0.1
USBM RI 8507
OSM REGULATIONS
HUMAN PERCEPTION OF STEADY STATE VIBRATION
PERCEPTIBLE
UNPLEASANT
INTOLERABLE
DRYWALL (19.1 mm/s)
PLASTER (12.4 mm/s)
PEA
K P
AR
TIC
LE V
ELO
CIT
Y (m
m/s
)
FREQUENCY (Hz)
PEA
K P
AR
TIC
LE V
ELO
CIT
Y (ip
s)
(50.8 mm/s)
1 10 100
100
10
1
1.0
0.1
0.01
10.0
0.1• In this example vibrations could
cause cosmetic building damage and clearly felt by humans
Simulated ground motion
frequency spectrum
PGV 30 mm/s at 6-12 Hz
EARTH SCIENCES DIVISION • LAWRENCE BERKELEY NATIONAL LABORATORY
TOUGH-FLAC modeling of events triggered by injection (Cappa and Rutqvist, 2011)
Rupture Size of a Notable (Felt) Seismic
• Largest magnitude when fault exposed to the highest shear stress (horizontal/vertical stress ratio = 0.6)
• A notable (felt) event, e.g. magnitude 4, requires a km-sized fault rupture • What about 2D model simplification?
σH
σV
Stress ratio = σH /σV = 0.6, 0.7 or 0.8
EARTH SCIENCES DIVISION • LAWRENCE BERKELEY NATIONAL LABORATORY 12
3D Modeling of Fault Reactivation and Seismicity
• Flow in the third dimension (not confined within 2D plane strain model) • Longer time for pressure buildup before reactivation ⇒ slightly larger magnitude
EARTH SCIENCES DIVISION • LAWRENCE BERKELEY NATIONAL LABORATORY 13
• Rupture area elongated along strike of fault • Large fault area pressurized at rupture ⇒ felt events, e.g. M = 3 - 4
3D Modeling of Fault Reactivation and Seismicity Contours on Fault Plane
EARTH SCIENCES DIVISION • LAWRENCE BERKELEY NATIONAL LABORATORY
Shear slip and stress drop associated with a seismic event:
Zoback and Gorelick (2012)
Rutqvist and Stephansson (2003)
Shear-induced permeability
Potential Shear-Induced Permeability and Leakage
EARTH SCIENCES DIVISION • LAWRENCE BERKELEY NATIONAL LABORATORY
Seismicity and Leakage Seismicity CO2 leakage
In this simulation example we simulated a seismic event that might be felt but with no upward CO2 leakage
Rinaldi et al 2014 (nt. J Greenhouse Gas Control)
EARTH SCIENCES DIVISION • LAWRENCE BERKELEY NATIONAL LABORATORY
Under higher stress normal to fracture the permeability decreases with shear (at stress level higher than the uniaxial compressive rock-strength)
Example of Shear-Permeability Tests on Shale Fractures Gutierrez et al. (2000)
EARTH SCIENCES DIVISION • LAWRENCE BERKELEY NATIONAL LABORATORY
Cretaceous Sandstones and mudstones (900 m thick)
Four gas producing wells
Three CO2injection wells
Gas zone
Carboniferous mudstones (950 m thick)
Carboniferous reservoir (20 m thick)
Water zone
CO2 is reinjected into the reservoir at Krechba for long term sequestration
Cretaceous Sandstones and mudstones (900 m thick)
Four gas producing wells
Three CO2injection wells
Gas zone
Carboniferous mudstones (950 m thick)
Carboniferous reservoir (20 m thick)
Water zone
CO2 is reinjected into the reservoir at Krechba for long term sequestration
• The CO2 injected at a depth of about 1,8 to 1,9 km into a 20 m thick formation of relatively low permeability.
• Nearly one million tonnes CO2 per year injected from 2004 to 2011 at 3 horizontal injection wells
• Bottom hole pressure limited to below the fracturing gradient ⇒ maximum pressure increase of about 100 bar (160% of hydrostatic)
• 950 m thick caprock with multiple low permeability formations
5 km
Gas-water contact at a depth of 1.8 km
KB503
KB502
KB501
Horizontal CO2 injection wells
KrechbaGas Field
5 km
Gas-water contact at a depth of 1.8 km
KB503
KB502
KB501
Horizontal CO2 injection wells
KrechbaGas Field
The In Salah CO2 Storage Project, Algeria
Plane view of Krechba Gas Field
Krechbagas field
In Salah Gas Project
Algeria
Spain
Mali
Libya
Niger
Marocco
Krechbagas field
In Salah Gas Project
Algeria
Spain
Mali
Libya
Niger
Marocco
σ1
Stress σ3 Fractures
In Salah Gas JV (BP, Staoil, Sonatrach)
EARTH SCIENCES DIVISION • LAWRENCE BERKELEY NATIONAL LABORATORY
In Salah Ground Surface uplift 2004-2007 from Satellite (InSAR)
Rutqvist et al (2010)
EARTH SCIENCES DIVISION • LAWRENCE BERKELEY NATIONAL LABORATORY
In Salah Ground Surface uplift 2004-2007 from Satellite (InSAR)
Rutqvist et al (2010)
Double-lobe uplift
EARTH SCIENCES DIVISION • LAWRENCE BERKELEY NATIONAL LABORATORY
In Salah Deep Fault or Fracture Zone Responses
1
Vasco et al. (GRL, 2010) interpreted observed double-lobe (uplift) response to be caused by a tensile opening feature at the injection zone.
u(x,t )
Tensile opening
u(x,t )
Tensile opening
u(x,t )
Tensile opening
Seismic contour in caprock 150 m above injection zone and surface uplift after 3 years .
(Rutqvist, 2012)
EARTH SCIENCES DIVISION • LAWRENCE BERKELEY NATIONAL LABORATORY 21
1
Rinaldi, Rutqvist (2013) TOUGH-FLAC modeling with simultaneous matching of transient uplift and injection data
Modeling indicates that the fracture zone extends a few hundred meters up from the reservoir (contained within the 900 m thick caprock)
TOUGH-FLAC
Data
In Salah Deep Fault or Fracture Zone Responses
EARTH SCIENCES DIVISION • LAWRENCE BERKELEY NATIONAL LABORATORY
However, CO2 injection at In Salah has not resulted in any felt seismic events or substantial strike-slip shear movements (Max magnitude 1.7, (Stork et al. 2014))
In Salah Deep Fault or Fracture Zone Responses
Injection pressure sufficiently high to induce deep fracture zone opening
Minor faults indicated from 3D seismic (Ringrose et al., 2011)
Theoretically close to critically stressed for shear reactivation (Morris et al., 2011)
σ1
Stress
σ3
Ringrose et al. (2011)
EARTH SCIENCES DIVISION • LAWRENCE BERKELEY NATIONAL LABORATORY
Concluding Remarks
Major Fault
Stress and strain changes beyond area of pressure change
CO2 plumePressure change far
beyond CO2 plume
Injection well
Minor faultsMajor Fault
Stress and strain changes beyond area of pressure change
CO2 plumePressure change far
beyond CO2 plume
Injection well
Minor faults
- We used numerical modelling to induce reactivation of steeply dipping faults at a high injection pressure in an unfavourable stress regime.
- We simulated events of magnitudes < 4 that would not result in any structural damage, but could likely be felt and cause concern in the local community. - At In Salah, injection pressure was relatively high indicating minor faults being critically stressed for reactivation, but no felt seismic event has been reported.
EARTH SCIENCES DIVISION • LAWRENCE BERKELEY NATIONAL LABORATORY
Concluding Remarks - We used numerical modelling to induce reactivation of steeply dipping faults at a high injection pressure in an unfavourable stress regime.
- We simulated events of magnitudes < 4 that would not result in any structural damage, but could likely be felt and cause concern in the local community. - At In Salah, injection pressure was relatively high indicating minor faults being critically stressed for reactivation, but no felt seismic event has been reported.
- At future large-scale CO2 operations (much larger than In Salah), it is the large-scale and long-term pressure buildup, associated crustal straining, and potential undetected (minor) faults that might be of greatest concern.
Major Fault
Stress and strain changes beyond area of pressure change
CO2 plumePressure change far
beyond CO2 plume
Injection well
Minor faultsMajor Fault
Stress and strain changes beyond area of pressure change
CO2 plumePressure change far
beyond CO2 plume
Injection well
Minor faults
EARTH SCIENCES DIVISION • LAWRENCE BERKELEY NATIONAL LABORATORY
Thank you! Jonny Rutqvist ([email protected])
Rutqvist J. The geomechanics of CO2 storage in deep sedimentary formations. International Journal of Geotechnical and Geological Engineering, 30, 525–551 (2012).
Rutqvist J., Cappa F., Rinaldi A.P., and Godano M. Modeling of induced seismicity and ground vibrations associated with geologic CO2 storage, and assessing their effects on surface structures and human perception. International Journal of Greenhouse Gas Control 24, 64–77 (2014).
Rinaldi A.P. and Rutqvist J. Modeling of deep fracture zone opening and transient ground surface uplift at KB-502 CO2 injection well, In Salah, Algeria. International Journal of Greenhouse Gas Control 12, 155–167 (2013).
Cappa F. and Rutqvist J. Seismic rupture and ground accelerations induced by CO2 injection in the shallow crust. Geophysical Journal International, 190, 1784–1789 (2012).
Cappa F. and Rutqvist J. Impact of CO2 geological sequestration on the nucleation of earthquakes. Geophysical Research Letters, 38, L17313, (2011).
Major Fault
Stress and strain changes beyond area of pressure change
CO2 plumePressure change far
beyond CO2 plume
Injection well
Minor faultsMajor Fault
Stress and strain changes beyond area of pressure change
CO2 plumePressure change far
beyond CO2 plume
Injection well
Minor faults
The sensitivity of CO2 storage simulations to pressure artifacts: indications from the Sleipner Benchmark model
Geophysical Modelling for CO2 Storage, Monitoring and Appraisal, University of Leeds, November, 2015
Andrew Cavanagh
Principal Researcher
Statoil RDI
[email protected] +47 9027 9715
Workflow...
2
Decide the model purpose
Establish conceptual geological models
Build rock models
Build property models
Assign flow properties and functions
Upscale flow properties and functions
Make forecasts
Assess and handle uncertainties
Re-iterate as necessary: 1. Maintain subsurface
database; 2. Preserve model build
decision track; 3. Discard or archive the
model results; 4. Address the next
question…
Compare simulations to observations
2.0
0.0
To
tal m
ass, C
O2 (
Mt)
Simulated time: 100 yrs
0.25
0.00
Dis
so
lve
d C
O2 (
Fra
ctio
n)
VE x10
The Sleipner plume
• Seismic monitoring has allowed for significant
improvements in understanding CO2 flow dynamics
• An improved basis for predicting the future plume
distribution and estimation of dissolved CO2
High-resolution model
(Layer 9 circa 2008)
Good match to observed distribution (red line)
Permedia BOS
3 Classsification: Draft 2014-04-22 (Cavanagh, Energy Procedia 2013)
Sleipner Benchmark (IEAGHG)
0.25
0.00
Dissolution estimate
4
Dis
so
lve
d C
O2 (
Fra
cti
on
)
2.0
0.0
To
tal m
ass
, C
O2 (
Mt)
Simulated time: 100 yrs
(2010)
10%
20%
VE x10
Permedia CO2 BOS
(Cavanagh, EP, 2013)
Implementation after
Hassanzadeh et al.
(IJGGC, 2008)
Plume calibration
• Darcy flow approach:
− Viscous forces, reservoir simulation
− Vertical equilibrium assumption (VE)
− Poor match, strong pressure artifact
• Percolating flow approach:
− Capillary forces, basin modeling
− Gravity assumption for migration (MGN)
− Equally poor match, but is buoyancy closer?
• We then allow the pressure to dissipate in the VE reservoir simulation,
and the plume redistributes to its buoyant equilibrium position. A much
better match to the footprint of the seismic observation is achieved.
Flow modeling based on seismic
5 Classsification: Draft 2014-04-22
Reservoir simulation 2-phase black oil model (CO2 BOS)
6 Classsification: Draft 2014-04-22
• Calibrating for 2008 seismic footprint
based on pressure equilibrium
• Simulation time in years:
• Pressure field at the end of injection:
~ 460 to 710 kPa (65-100 psi) overpressure
~ 250 kPa (36 psi) drop over 3 km
10 15 20 30 40 50 60 70 80 90 100 10
710
460
X
Conclusion
7
1999 2001 2002 2004 2006 2008
Dynamic equilibrium
The simulation results clearly indicate that the plume beneath
the caprock is gravity-dominated, and close to equilibrium at
every observation point (Cavanagh, Energy Procedia, 2013)
Reservoir simulations for CO2 storage may be susceptible to
significant pressure artifacts that distort the model outcome.
Implication
8
Pressure
Spatial distribution
Timing
Without calibration and correction, reservoir simulations are
highly likely to be misleading with respect to CO2 storage.
9
• Area of Interest: 3x6 km
• Cell resolution: 50x50x0.5 m
• Geocellular mesh: 550,000 cells
Sleipner Benchmark II
Cap Rock
Sand Wedge
Thick Shale
Utsira Sand
Thin Barrier
Base Utsira
(Cavanagh & Haszeldine, IJGGC, 2014)
10
NORWAY
A big thank you to... Philip Ringrose (Statoil) Varunendra Singh Hilde Hansen Bamshad Nazarien Martin Iding Neil Wildgust (IEAGHG… PTRC… GCCSI!!!) Chris Leskiw (Permedia) Jason Wudkevich
The sensitivity of CO2 storage simulations
to pressure artifacts: Indications from the
Sleipner Benchmark model
Andrew Cavanagh
Principal Researcher
Tel: +47 2097 2715 www.statoil.com
Passive Seismic Monitoring of CO2 Storage Sites
Anna L. Stork1, James P. Verdon1, J.-‐Michael Kendall1, Claire Allmark2, Andrew Cur@s2 and Don J. White3
UKCCSRC Geophysical modelling for CO2 storage, monitoring and appraisal mee@ng Leeds
3 November 2015
2 November 2015
1. University of Bristol 2. University of Edinburgh 3. Geological Survey Canada
Passive seismic monitoring
2
2 November 2015
geomechanical deformation experienced at these large “mega-tonne” storage sites, as these will inform us of the potential geo-mechanical issues that will be experienced as commercial-scale,megatonne injection sites are developed in the coming decades.
Geomechanical Response to CO2 InjectionThe effective stress, σ′ij, acting on porous rocks is defined byTerzaghi (13) as follows:
σ′ij ¼ σij − βW δijP; [1]
where σij is the stress applied by regional tectonic stresses andthe overburden weight, βW is the Biot–Willis coefficient, δij is theKroenecker δ, and P is the pore pressure. Therefore, any in-crease in pore pressure induced by injection will reduce the ef-fective stress, which will in turn lead to inflation of the reservoir.The magnitude of this inflation will be controlled by the magni-tude of the pore pressure increase, and the geometry and mate-rial properties of the reservoir (14).As well as directly changing the effective stress acting on
reservoir rocks via Eq. 1, inflation of the reservoir will lead tochanges in the applied stress both in and around the reservoir.Small amounts of deformation are common in many settings, andwill not pose a risk to storage security. However, if deformationbecomes more substantial, it can affect storage operations ina number of ways, illustrated schematically in Fig. 1. The prin-cipal risks posed by geomechanical deformation to secure storageare summarized below.
Reservoir Inflation and Alteration of Flow Properties. Pore pressureincrease and inflation can influence the flow properties ofa storage reservoir. Laboratory experiments show that perme-ability is sensitive to pressure (15). Furthermore, pore pressureincreases may open existing fracture networks in the reservoir, orcreate new ones, along which CO2 can flow more rapidly. Per-meability increases within the reservoir will not pose a directleakage risk. Nevertheless, if permeability is increased duringinjection, this will reduce the accuracy of fluid flow simulationsused to predict the resulting CO2 distribution. The result may bethat CO2 reaches spill-points or breaks through at other wellsfaster than anticipated, reducing the amount of CO2 that can bestored. For example, Bissell et al. (16) have shown that injectivityat In Salah is pressure dependent, implying that CO2 flow iscontrolled at least in part by the opening and closing of fracturesin the reservoir.
Fracturing of Sealing Caprocks. Deformation in a reservoir isgenerally transferred into the surrounding rocks. This can leadto the creation or reactivation of fracture networks around andabove a reservoir. Fractures running through an otherwiseimpermeable caprock could compromise the storage integrity,providing permeable pathways for CO2 to escape from thereservoir. This is probably the greatest risk to storage securityposed by geomechanical deformation. Leakage of gas throughfractured caprock has been observed above hydrocarbon res-ervoirs (17, 18) and at natural gas storage sites (19).
Triggering of Seismicity. Beginning with the earthquakes triggeredby waste fluid injection at the Rocky Mountain Arsenal (20), ithas been recognized that subsurface fluid injection is capable oftriggering felt (of sufficient magnitude to be felt by nearbypopulations, so typically ML > 2) seismic events on preexistingtectonic faults (21). Recently, examples of tectonic activity trig-gered by disposal of waste water from hydraulic fracturing havebeen noted. Of course, it should be kept in mind that, of thou-sands of fluid injection wells, only a handful have experiencedsuch seismic events (22). Even if felt seismicity is induced duringCO2 injection, it is unlikely that events would be of sufficient
magnitude to damage property or endanger life. Nevertheless,regular triggering of felt seismic events would represent a sig-nificant “own-goal” from a public relations and political per-spective, and local opposition has already proved to be a significantobstacle to planned CCS projects (23). More significantly,triggering of larger seismic events will indicate that the failurecondition on small faults has been met due to anthropogenicpressure changes, with implications for caprock integrity issues asdiscussed above.
Wellbore Failure and Casing Damage. Geomechanical deformationin producing reservoirs has been observed to cause failure ofwellbore casing (24). It is conceivable that either bedding-parallelslip in layers above the reservoir, or expansion of the reservoiragainst the overburden, could cause shearing of the wellbore. Aswell as the associated costs, damaged well casing in the over-burden presents a significant leakage risk. Although the authorsare not presently aware of any incidence of geomechanically in-duced wellbore failure during CO2 injection, the risk to storageintegrity posed by mechanical effects in the wellbore is an issuethat must be considered at future storage sites.
Monitoring Geomechanical DeformationFig. 1 also illustrates the variety of methods that can be used tomonitor geomechanical deformation in the field. Although theimportance of geomechanical deformation in oil production isbecoming increasingly appreciated, monitoring it in the fieldremains something of a niche activity. Nevertheless, a number of
SATELLITE GEODESY
SEISMIC MONITORING
MICROSEISMICMONITORING
BOREHOLETILTMETERS
BEDDING PARALLEL SLIP
SURFACE UPLIFT
WELLBOREFAILURE
FAULTREACTIVATION
SEALFRACTURING
INFLATION OFRESERVOIR
Fig. 1. Schematic illustration showing how geomechanical deformation caninfluence CO2 storage sites (red text), and potential monitoring options(blue text). Adapted from Herwanger and Horne (34).
2 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1302156110 Verdon et al.
Verdon et al., 2013
• Generally small magnitude, M<0,
• Iden@fy poten@al leakage pathways,
• Near real-‐@me analysis to provide early-‐warning,
• Understand geomechanical response & verify models,
• Aid seismic hazard assessment.
Major CCS projects 1. Weyburn, Saskatchewan
• >30 Mt since 2000 • Constraining geomechanical models
with microseismic observa@ons
2. In Salah, Algeria • ~4 Mt 2004-‐2011 • Changes in fracture characteris@cs during injec@on
3. Aquistore – Boundary Dam, SK • Began injec@on April 2015 • Using ambient noise to determine seismic veloci@es
3
2 November 2015
Image PTRC
4
Weyburn passive seismic monitoring 2003 – 2011
200m
250m
Injection: 2000 – present
1430m
Weyburn microseismic event locations
Injection well
5
Magnitudes -3<Mw<-1
Producing wells
Verdon et al., PNAS, 2013
Geophone array
During injection After shut-in
6
2 November 2015
% change in fracture poten
@al
Reservoir
Overburden
Injec@on well
Producing well
Modelling stress changes
To match observed seismicity pa_ern an updated geomechanical model in required with a so`er reservoir.
Verdon et al., EPSL, 2011
Weyburn - Summary • Microseismic observa@ons can provide important constraint on geomechanical models.
• Model with a so`er reservoir than expected from core samples • Increases fracture poten@al in overburden above producing wells;
• Decreases fracture poten@al in overburden above injec@on wells.
• Seismicity caused by stress transfer, not fluid migra@on.
7
2 November 2015
8
2 November 2015
In Salah passive seismic monitoring 2009 – 2011
Injection 2004 – 2011
Stork et al., IJGGC, 2015
9
2 November 2015
In Salah, Algeria
Stork et al., IJGGC, 2015
10
2 November 2015
Fracture zone detected at injection depth (2km)
Rutqvist, 2012
In Salah – Geophysical observations Ground movement detected by satellites (InSAR)
Rucci et al., 2013
2 November 2015
11
In Salah – Passive seismic observations
>9000 events Mw=1.7
Stork et al., 2015
12
2 November 2015
In Salah – Event locations
A.L. Stork et al. / International Journal of Greenhouse Gas Control 32 (2015) 159–171 165
Fig. 9. Event depths and horizontal distances from the observation well for dif-ferent tsp times, estimated using E3D. The colours represent the inclination of theP-arrival measured from the synthetic waveforms. The caprock and reservoir layersare shaded as in Fig. 2 and the approximate injection interval is between the twothicker black lines at ∼1.9 km deep. (For interpretation of the references to color inthis figure legend, the reader is referred to the web version of this article.)
Fig. 10. Estimated depth and horizontal distance of events from observation well.Locations are projected onto a SE-NW plane. The colours indicate the time of theevent in number of days since the earliest plotted event. Locations are estimatedfor events with io < 15◦ , linearity ≥0.95 and signal-to-noise ratio >3.0. The caprockand reservoir layers are shaded as in Fig. 2 and the approximate injection interval isbetween the two thicker black lines at ∼1.9 km deep. The green triangle indicates thelocation of the geophone used in the analysis. (For interpretation of the referencesto color in this figure legend, the reader is referred to the web version of this article.)
a larger area to become seismically active. We do not observe anysystematic shortening of tsp times over time this suggests that thereis no systematic migration of seismicity through the cap rock. Thisis reassuring for the containment of CO2. We do observe a smallnumber of events (11) with shorter tsp times (<0.5 s) (Fig. 6). Thesedo not satisfy our criteria to estimate locations but their significanceis discussed below.
0 0.5 1 1.5 2 2.5 300.10.20.30.40.50.60.70.80.9
1
Horizontal Distance (km)
tsp tim
e (s
)
0 0.5 1 1.5 2 2.5 3−2.5
−2
−1.5
−1
−0.5
0
Horizontal Distance (km)
Dept
h (k
m)
Fig. 11. Raytracing results for P- (red) and S-waves (green) (lower panel) and tsp
times as a function of distance (upper panel), estimated using the provided isotropic1-D velocity model and a source at 2.4 km deep (star). (For interpretation of thereferences to color in this figure legend, the reader is referred to the web version ofthis article.)
To provide additional evidence for the approximate locationsobtained through finite-difference modelling we conduct a ray-tracing exercise. The results from ray-tracing through the isotropic1-D velocity model using the method of Kendall and Thomson(1989) show that events with hypocentres at 2.4 km depth and1.2 km horizontal distance from the array (Fig. 11).
To estimate errors in our reported locations we tested the effectof the velocity model on the travel-times and, for example, welocate Cluster 2 up to 450 m shallower if the velocity model is 10%slower overall, if the near surface layer is 20% slower or if the modelis anisotropic (see Stork et al., 2015 for a detailed description). Thiswould place the events in this cluster between 1.65 km – 2.25 kmdeep and therefore extending up to 150 m unto the lower caprock.As an estimate of the error in horizontal distances from the arraywe take the maximum horizontal distance between grid points inFig. 9, this is 174 m when tsp = 0.60s near 0◦ incidence. Event loca-tions obtained using the two methods, finite difference modellingand ray-tracing, agree within the estimated errors.
Overall, the results for the estimated location of Cluster 2 showthat the seismicity occurred at depths over a range of ∼600 m at orbelow the injection interval and at azimuths from the monitoringwell consistent with the activation of a pre-existing wide fracturezone at the injection depth and extending into the lower caprock(as reported by Iding and Ringrose (2010) and Rutqvist (2012)) withevents occurring on similarly oriented fractures within the zone. Aninaccurate velocity model significantly affects seismic event loca-tions and if the velocity model is 10% slower this would imply thatthe events extend into the lowermost 150 m of the caprock, consis-tent with the previous fracture zone interpretation. An anisotropicfractured medium may also affect interpretation of the data.
We note that a few events occur outside the two main clustersand example seismograms are shown in Fig. 5. We find 11 eventswith 0.31 s <tsp< 0.5 s (Fig. 6 and example seismograms in Fig. 5c).According to our model locations in Fig. 9 these events are between1.1 km and 1.8 km deep but, as with all locations reported here,there are significant uncertainties in these locations. The eventsoccur over the whole monitoring period and there is no correlation
Stork et al., 2015
• Constant depth events • No evidence of migra@on to surface.
Injection interval
Geophone
13
27 May 2015 Fracture characterisation from shear-wave splitting
Concern: Is injection creating new fractures, allowing CO2 migration?
Del
ay ti
me
betw
een
sp
lit w
aves
Stork et al., IJGGC, 2015
Dominant fracture orientation in direction of σH
Delay time increases after high injection
Returns to original value
In Salah - Summary
• Proof-‐of-‐concept • Limited array but provides useful results
• No evidence of shallower seismicity with @me • No evidence of shallow migra@on of CO2
• CO2 injec@on opens fractures that close when pressure decreases • Limits poten@al of CO2 leakage
14
2 November 2015
15
2 November 2015
Aquistore passive seismic monitoring 2012 – 2015
2.5km x 2.5km array 50 – 64 1C/3C geophones 6m/20m deep 3 broadband stations
Injection: Since April 2015
mileskm
12
InjObs
1km
16
2 November 2015
BOUNDARY DAM - AQUISTORE CO2 INJECTION PROJECT
World’s first commercial power plant CCS project
Image PTRC
17
2 November 2015
BOUNDARY DAM - AQUISTORE CO2 INJECTION PROJECT
2982 Ben Rostron et al. / Energy Procedia 63 ( 2014 ) 2977 – 2984
Figure 4. Schematic of the 01/5-6-2-8W2 injection well. Approximate locations of major geological units are indicated (left) drawn to vertical
depth scale.
One week after completion of the injection well, the drill rig was moved 150m northeast (Figure 1) and over the period October 1st to November 9th, 2012 the observation well (41/5-6-2-8W2) was drilled 3400m deep through the entire Phanerozoic section (schematic on Figure 5). A similar suite of geophysical well logs was collected from the observation well as was the injection well.
After their interpretation, geological, hydraulic, and petrophysical data collected during the drilling and well
evaluation were incorporated into a revised geological model of the Aquistore site.
2.3. Post drilling activities - downhole
Subsequent geological information was obtained between, and around, the newly-drilled wells via two different downhole seismic surveys conducted as part of baseline surveys to start the CO2 Measurement, Monitoring, and Verification (MMV) program at the site. The first (February, 2013), was a crosswell seismic survey between the two wells over the interval 3100 to 3400m that provided detailed (metre-scale) tomography of the geology between the wells. The second survey (Fall, 2013) was a 3D vertical seismic profile (VSP) that utilized both a conventional 60-level, three-component geophone over the interval 2550-3400m and the well-installed optical fibre system. The 3D VSP provided subsurface information between the resolution the detailed scale from the crosswell survey, and the standard surface 3D seismic survey conducted previously [6].
Rostron et al., 2014
18
2 November 2015
mileskm
23
1km
Ambient seismic noise interferometry
Ambient seismic noise interferometry • Use noise recorded at receivers to produce velocity map • Cross-‐correlate noise at pairs of receivers (Bensen et al., 2007)
• Create virtual source at one receiver
19
2 November 2015
Ambient Noise Tomography
• Cross-‐correlate noise at pairs of receivers
20
2 November 2015
49.10
256.95256.90
0.24 0.33 Velocity km/s
Preliminary Tomography Results
21
2 November 2015
Depth sensitivity for periods 0.6 – 1.0s Rayleigh wave group velocity 0.7s period
Dep
th
Fast Marching Surface Wave Tomography (Rawlinson et al., 2008)
Preliminary Tomography Results
22
2 November 2015
Dep
th
49.10
256.95256.90
0.24 0.33 Velocity km/s
Rayleigh wave group velocity 0.7s period
Aquistore – Summary
• Excellent baseline • Background seismicity • Near-‐surface characterisa@on • Allows @me-‐lapse studies
• On-‐going array detec@on and loca@on studies • Broadband vs near-‐surface geophones vs downhole geophones vs fibre op@c
• Similar geology to Weyburn – similar response?
23
2 November 2015
• Large CCS sites exhibit differing microseismic responses. • Weyburn 100s seismic events up to MW = -‐1.0 • In Salah 1000s seismic events up to Mw~ 1.7 • Aquistore?
• Baseline data is crucial to • Highlight any ac@ve structures; • Evaluate effect of injec@on. • In Salah – ac@ve fracture zone iden@fied if earlier installa@on.
• Use passive seismic monitoring to • Calibrate geomechanical models; • Determine fracture characteris@cs; • Observe changes in seismicity, velocity, fracture characteris@cs.
• Conduct careful array design appropriate for purpose. 24
2 November 2015 Conclusions
25
We thank the In Salah JIP, BP, Statoil and Sonatrach, for providing the microseismic data recorded at the In Salah site & for their permission to present this work. We thank the PTRC for providing permission to work with and present the Weyburn and Aquistore microseismic data.
The author would like to acknowledge the financial support of the UK CCS Research Centre (www.ukccsrc.ac.uk) in carrying out this work. The UKCCSRC is funded by the
EPSRC as part of the RCUK Energy Programme.
26
BRISTOL UNIVERSITY MICROSEISMICITY PROJECTS BUMPS
We thank the sponsors of the Bristol University Microseismicity Projects (BUMPS) consortium for supporting this research.
Advances in rock physics modellingand improved estimation of CO2
saturation
Giorgos Papageorgiou
University of Edinburgh
UKSCCSRC Geophysical Modelling for CO2 Storage,Monitoring and Appraisal Specialist Meeting
Leeds, 2015
Introduction
Partial FluidSaturation
Applicability
Conclusions
The squirt flow mechanism
• Seismic waves create pressuregradients
• Depending on time/lengthscale, different types of flow(hence dispersion) occur
• Model “local” flow usingidealised pore geometries:
Squirt-flow Mechanism 429
For a lateral pressure gradient to develop, we assume thatpore pressure on the sides of a small homogeneous repre-sentative volume of the rock does not change in time. Thiscondition is strictly valid in apparently fully saturated rockswith only small amounts of high compressibility gas in thepores-a situation quite typical in natural reservoirs. Thesquirt-flow pattern becomes more complex in saturatedrocks without residual gas: pore fluid is squeezed from thincracks into surrounding large pores or adjacent cracks ofdifferent orientation (Mavko and Nur, 1975). Pressure on thesides of a representative volume does change in time. Theamplitude of pressure variation in large pores is muchsmaller than that in thin cracks. Therefore, in this case theBISQ model will give realistic quantitative estimates tovelocity dispersion and attenuation.
A natural choice of the representative volume for the caseunder consideration is a cylinder with its axis parallel to thedirection of wave propagation. The radius of this cylinder isthe characteristic squirt-flow length (Figure la). The physi-cal meaning of the characteristic squirt-flow length is theaverage length that produces the squirt-flow effect identicalto the cumulative effect of squirt flow in pores of variousshapes and sizes. This parameter is intimately related to thepore space geometry of a given rock. We assume that it is afundamental rock property that does not depend on fre-quency and fluid characteristics, and thus can be determinedexperimentally. This concept is similar to the permeabilityconcept where permeability cannot be measured directly,but can be found by matching the Darcy formula’s predic-tions with fluid Row rate and pressure gradient measure-ments.
The BISQ model does not require an individual poregeometry: pore fluid dynamics are linked to permeability andthe characteristic squirt-flow length. Therefore, we model
the squirt-flow mechanism by using its macrorcopic ratherthan microscopic description.
In this paper, we analyze and simplify our earlier solution(Dvorkin and Nur, 1993) to show that for frequencies smallerthan Biot’s characteristic frequency the viscoelastic proper-ties of rocks can be expressed through a single dimensionlessparameter that is a combination of angular frequency, thecharacteristic squirt-flow length, and hydraulic diffusivity.We theoretically explore the relative importance of the Biotand the squirt-flow components of fluid flow on a high-porosity sandstone sample (the Biot dispersion and attenu-ation typically increase with increasing porosity). The exam-ple shows that the squirt-flow component dominates even inhigh-porosity rocks.
We explore the influence of permeability on attenuationand show that the BISQ model can explain experimentallyobserved relations between these two parameters.
Finally, we modify the formulas for velocity and attenua-tion for partially saturated rocks. To do so, we assume thatthe saturated part of the representative cylindrical volume isalso a cylinder of a smaller radius (Figure lb), whichdecreases with decreasing saturation. We find good agree-ment between experimental attenuation data and our theo-retical predictions.
THE BISQ MODEL-VELOCITY AND ATTENUATION
BISQ and Biot formulas
The BISQ model gives the following expressions for thefast P-wave velocity Vp and attenuation coefficient a(Dvorkin and Nur, 1993):
FIG. 1. The mechanical image of a representative cylinder used in the BISQ model: (a) FluidRow in the cylinder-the Biot and the squirt components; a P-wave propagates parallel tothe cylinder’s axis. (b) Partial saturation-the radius of the fluid-filled cylinder decreaseswith decreasing saturation.
Introduction
Partial FluidSaturation
Applicability
Conclusions
The squirt flow mechanism
• Seismic waves create pressuregradients
• Depending on time/lengthscale, different types of flow(hence dispersion) occur
• Model “local” flow usingidealised pore geometries:donut+disk
Introduction
Partial FluidSaturation
Applicability
Conclusions
The squirt flow mechanism
• Seismic waves create pressuregradients
• Depending on time/lengthscale, different types of flow(hence dispersion) occur
• Model “local” flow usingidealised pore geometries:coins+spheres
Introduction
Partial FluidSaturation
Applicability
Conclusions
A minimal model
Minimally, to model the squirt flow effect replace the rock bya collection of coin-shaped cracks and sphere-shaped pores�
Introduction
Partial FluidSaturation
Applicability
Conclusions
Extending to two fluids
How do we model partial saturation?
Introduction
Partial FluidSaturation
Applicability
Conclusions
Assume two fluids in each pore
Solve Darcy’s law in the frequency domain:
∂tm1 =
ρ1k1ζ
η1(P�1 − P1 ), m1 = S1ρ
1 φ
∂tm2 =
ρ2k2ζ
η2(P�2 − P2 ), m2 = (1− S1)ρ2 φ
.
and use the result in Eshelby’s expansion (obtain complexvalued bulk modulus):
Keff(ω) = Kd +
φ0
(Km
σc+ 1)
P(ω)
σ(ω)+ φ�0
(3Km
4µ+ 1)
P�(ω)
σ(ω).
Appeal of this method is that Keff(0) = KGassmann
There issome ambiguity as to which to use here!
Introduction
Partial FluidSaturation
Applicability
Conclusions
Assume two fluids in each pore
Solve Darcy’s law in the frequency domain:
∂tm1 =
ρ1k1ζ
η1(P�1 − P1 ), m1 = S1ρ
1 φ
∂tm2 =
ρ2k2ζ
η2(P�2 − P2 ), m2 = (1− S1)ρ2 φ
.
and use the result in Eshelby’s expansion (obtain complexvalued bulk modulus):
Keff(ω) = Kd +
φ0
(Km
σc+ 1)
P(ω)
σ(ω)+ φ�0
(3Km
4µ+ 1)
P�(ω)
σ(ω).
There is some ambiguity as to which pressure to use here!
Introduction
Partial FluidSaturation
Applicability
Conclusions
Assume two fluids in each pore
Solve Darcy’s law in the frequency domain:
∂tm1 =
ρ1k1ζ
η1(P�1 − P1 ), m1 = S1ρ
1 φ
∂tm2 =
ρ2k2ζ
η2(P�2 − P2 ), m2 = (1− S1)ρ2 φ
.
and use the result in Eshelby’s expansion (obtain complexvalued bulk modulus):
Keff(ω) = Kd +
φ0
(Km
σc+ 1)
P(ω)
σ(ω)+ φ�0
(3Km
4µ+ 1)
P�(ω)
σ(ω).
There is some ambiguity as to which pressure to use here!
Introduction
Partial FluidSaturation
Applicability
Conclusions
Assume two fluids in each pore
Solve Darcy’s law in the frequency domain:
∂tm1 =
ρ1k1ζ
η1(P�1 − P1 ), m1 = S1ρ
1 φ
∂tm2 =
ρ2k2ζ
η2(P�2 − P2 ), m2 = (1− S1)ρ2 φ
.
and use the result in Eshelby’s expansion (obtain complexvalued bulk modulus):
Keff(ω) = Kd +
φ0
(Km
σc+ 1)
P(ω)
σ(ω)+ φ�0
(3Km
4µ+ 1)
P�(ω)
σ(ω).
There is some ambiguity as to which saturation to use here!
Introduction
Partial FluidSaturation
Applicability
Conclusions
Inclusion-Dependent Saturation
The “observable” saturation can differ from the saturation inthe cracks/pores. This leads to a way of modellingimbibition/drainage phenomena.1
��� ��� ��� ��� ��� ������
���
���
���
���
���
������� ����������
���������������
imbibition
drainage
1G Papageorgiou and M Chapman. “Multifluid squirt flow andhysteresis effects on the bulk modulus–water saturation relationship”.In: Geophysical Journal International 203.2 (2015), pp. 814–817.
Introduction
Partial FluidSaturation
Applicability
Conclusions
Inclusion-Dependent Saturation
The “observable” saturation can differ from the saturation inthe cracks/pores. This leads to a way of modellingimbibition/drainage phenomena.1
����������
��������
��� ��� ��� ��� ��� ������ × ����
��� × ����
��� × ����
��� × ����
��� × ����
��� × ����
��� × ����
����� ����������
�����������(��)
��� ω = �
1G Papageorgiou and M Chapman. “Multifluid squirt flow andhysteresis effects on the bulk modulus–water saturation relationship”.In: Geophysical Journal International 203.2 (2015), pp. 814–817.
Introduction
Partial FluidSaturation
Applicability
Conclusions
Inclusion-Dependent Saturation
The “observable” saturation can differ from the saturation inthe cracks/pores. This leads to a way of modellingimbibition/drainage phenomena.1
����������
��������
��� ��� ��� ��� ��� �������
����
����
����
����
����
����� ����������
�����������
��� ω = �
1G Papageorgiou and M Chapman. “Multifluid squirt flow andhysteresis effects on the bulk modulus–water saturation relationship”.In: Geophysical Journal International 203.2 (2015), pp. 814–817.
Introduction
Partial FluidSaturation
Applicability
Conclusions
Inclusion-Dependent Saturation
The “observable” saturation can differ from the saturation inthe cracks/pores. This leads to a way of modellingimbibition/drainage phenomena.1
��������
���� ����
-� -� � � �����
����
����
����
����
����
��� ���������
�����������
� = ����
1G Papageorgiou and M Chapman. “Multifluid squirt flow andhysteresis effects on the bulk modulus–water saturation relationship”.In: Geophysical Journal International 203.2 (2015), pp. 814–817.
Introduction
Partial FluidSaturation
Applicability
Conclusions
Pressure discontinuity
Use capillary pressure equation ∆C = q∆Pw constrainedwithin −1 < q < 0. Assume, the balancing pressure inEshelby’s formula can jump from that of the non-wetting tothat of the wetting fluid in a discontinuous way. Think of thelow frequency limit (Gassmann limit) of this model. 2
2presented in SEG 2015 and under revision in GJI
Introduction
Partial FluidSaturation
Applicability
Conclusions
A wet Gassmann model
Different effective pressure choices correspond to differentmodels:
P(1) 'Pw
P(2) 'Pnw
... and different effective fluid moduli:
1
K(1)f (q)
' SwKw
+Snw(1− q)
Knw=
1KGW
− q1− SwKnw
1
K(2)f (q)
'Sw(1 + q)
Kw+
SnwKnw
=1KGW
+ qSwKw
That depend on this parameter q
Introduction
Partial FluidSaturation
Applicability
Conclusions
A wet Gassmann model
Different effective pressure choices correspond to differentmodels:
P(1) 'Pw
P(2) 'Pnw
... and different effective fluid moduli:
1
K(1)f (q)
' SwKw
+Snw(1− q)
Knw=
1KGW
− q1− SwKnw
1
K(2)f (q)
'Sw(1 + q)
Kw+
SnwKnw
=1KGW
+ qSwKw
That depend on this parameter q
Introduction
Partial FluidSaturation
Applicability
Conclusions
A wet Gassmann model
Think of these as a non-wetted and wetted extremes and jointhem somewhere in between. Depending on where thistransition happens and how fast, different models areobtained (keep q as a scaling parameter):
�/��
��� ��� ��� ��� ��� ���
-�/�
�
��
��� ��� ��� ��� ��� ���
��
�
��
�
Introduction
Partial FluidSaturation
Applicability
Conclusions
A wet Gassmann model
As a result, a jump appears in the bulk modulus VSsaturation relationship:
Wet Gassmann
Gassmann
�� ± δ�
���� (��)
���� (�)
��
Here φ = 30%,Km = 4Kd = 8Kw = 800Knw similar togas/water in sandstone. Still not clear if parameter q affectsthe frequency dependence of the theory and how.
Introduction
Partial FluidSaturation
Applicability
Conclusions
Ultrasonic experiments
Do these models have any reason to exist? Observed “jump”3
in Keff normally attributed to frequency effects but could beexplained using the static wet Gassmann described here.
��� ��� ��� ��� ��� ����������
�������
�������
�������
�������
�������
��
���
3Kelvin Amalokwu et al. “Water saturation effects on P-waveanisotropy in synthetic sandstone with aligned fractures”. In:Geophysical Journal International 202.2 (2015), pp. 1088–1095.
Introduction
Partial FluidSaturation
Applicability
Conclusions
A speculative explanation
How much saturation is needed to transition from thenon-wetted to wetted regime ↔ pore raggednessHow smooth the transition ↔ pore size distribution
But not quantified! Hope is this is the path to petrophysicalparameters in this context.
Introduction
Partial FluidSaturation
Applicability
Conclusions
Saner approach: pressure averaging
Scale capillary pressure equation a little differently:
Pnw = αKnw
KwPw, 1 ≤ α ≤ Kw
Knw.
Assume pressure averaging in the inclusions balances stressP = SwPw + (1− Sw)Pnw.
Introduction
Partial FluidSaturation
Applicability
Conclusions
Pressure averaging - Low Frequency
At low frequency approximation the effective fluid modulusdepends on α:
K̃f =SwKw + α(1− Sw)Knw
Sw + α(1− Sw), 1 ≤ α ≤ Kw
Knw
which looks like Brie’s empirical model.4
α = 1
α = 2α = 3
α = 5α = 10
α =w
g
��� ��� ��� ��� ��� ���
�
�
��
�
4Work under review for GP special issue in rock physics
Introduction
Partial FluidSaturation
Applicability
Conclusions
Pressure averaging - Frequency dependence
The characteristic frequency depends on α as well so thismodel attenuates differently depending on the value of α
0.036
0.072
0.108
0.144
0.180
Introduction
Partial FluidSaturation
Applicability
Conclusions
Pressure averaging - Frequency dependence
The characteristic frequency depends on α as well so thismodel attenuates differently depending on the value of α
0.036
0.072
0.108
0.144
0.180
Introduction
Partial FluidSaturation
Applicability
Conclusions
Pressure averaging - Frequency dependence
The characteristic frequency depends on α as well so thismodel attenuates differently depending on the value of α
0.036
0.072
0.108
0.144
0.180
Introduction
Partial FluidSaturation
Applicability
Conclusions
Using these models
If this interpretation is correct, this parameter is of crucialimportance. Even a slight departure from harmonic law,improves gas estimation using rock-physics based inversions:• f-AVO5
• trace inversion6
• ...?We are currently using these ideas to determine if CO2saturation in the Sleipner field can be estimated moreaccurately.
5Xiaogyang Wu et al, 20046Current work by Zhaoyu Jin in Edinburgh
Introduction
Partial FluidSaturation
Applicability
Conclusions
Estimating the parameter
Parameter q is given as a function of capillary pressure7:
q =1− Sw(1− Sw)C ′(Sw)/Kw
1− Sw(1− Sw)C ′(Sw)/Knw
Is it a fiddle factor, is it realistic, can it be tuned with C (S)experimental results?See whether it is measurable from rock physics experiments8
7Juan E. Santos, Jaime M Corbero, and Jim Douglas Jr. “Static anddynamic behavior of a porous solid saturated by a two-phase fluid”. In:J. Acoust. Soc. Am. 87.4 (1990), pp. 1426–1438. DOI:10.1121/1.1908239.
8Data from K. Amalokuw, I. Falcon-Suarez at SOC
Introduction
Partial FluidSaturation
Applicability
Conclusions
To Conclude
• The effect of capillary pressure in rock physics may besignificant
• Choice of different saturation in pores/crack with fixedoverall saturation, leads to modelling ofimbibition/drainage
• Choice of pressure jump leads to modulus discontinuity• Choice of averaged pressure leads to Brie’s law at lowfrequency and appealing frequency dependent model
• No need to resort to patches• Feedback welcome!
Introduction
Partial FluidSaturation
Applicability
Conclusions
Thanks!
Thank you!
Acknowledgments:
• Mark Chapman
• EPSRC DiSECCS grant
Assessing uncertainty of time-lapse
seismic response due to geomechanical
deformation
Doug Angus,
School of Earth & Environment, University of Leeds, [email protected]
Acknowledgements:
Claire Birnie, Yanxiao He, Tom Lynch & Dave Price
Outline
• Context
• Multi-physics solution
• Time-lapse seismics/geomechanics
• Overburden imaging
• Valhall example
• Geomechanics and uncertainty
• Way forward
2
Context
Sleipner
• Negligible pressure effect
• High porosity/high permeability
Snohvit/In Salah
• Pressure effect
• Snohvit – compartmentalisation
• In Salah - ? Top layer 2010
thic
kne
ss (
m)
stru
ctu
ral t
hic
knes
s (
met
res)
tem
po
ral t
hic
knes
s
(mse
c)
3
4
e.g., microseismicity – static velocity model
Event location SI Event location MTMI Waveforms Velocity model
e.g., microseismicity – non-static velocity model
Static velocity model
True dynamic velocity model
Event location MTMI
5
• Change in pore pressure (DP)
• Leads to change in horizontal
total stress
• Difficult to predict stress evolution
D ¢sV = DsV -aDP
D ¢s H = Ds H -aDP
Herwanger (2007)
Dynamic view
(syn- and post-production)
6
Multi-physics: porous deformable media
• Integration (hydro-mechanics):
• Coupled fluid-flow/geomechanics
– Fully-coupled
– One-way coupling (flow to geomechanics)
– Two-way coupling
Geophysical
attributes
D Pp
D k, ø, c
Stress
changes
Petroleum
Production
Reservoir
property
changes
D Si
4D Seismics
Microseismics
7
Multi-physics: porous deformable media
Stress dependent velocities
Rock physics transforms
8
Reservoir model with high fault transmissibility
Rock physics
Shapiro 2003 Tod 2002
9
Seismic monitoring
Marine seismic
Land seismic
Acquisition (instrument) geometry
10
Time-lapse or 4D seismic • Change in:
• Saturation
• Pressure/stress
• Mechanical properties
P-wave velocity change for true earth model
Baseline - Monitor 1 Baseline - Monitor 2
Baseline model
11
Time-lapse or 4D seismic • What is measured:
• Time differences
• Amplitude differences
• “Devil is in the detail”
12
Estimated P-wave reflection amplitude (strength)
changes using full-offset seismic data
Estimated P-wave reflection amplitude (strength)
changes using near-offset seismic data
Time-lapse or 4D seismic • Extract time-lapse amplitude changes
13
P-wave velocity change for
true earth model
Estimated P-wave velocity
change using full-offset
seismic data
Estimated P-wave velocity
change using near-offset
seismic data
Time-lapse or 4D seismic • Extract time-lapse velocity changes
14
Time-lapse or 4D seismic • Tau-p pre-stack approach
15
-7.0
-6.0
-5.0
-4.0
-3.0
-2.0
-1.0
0.0
1.0
1982 1987 1992 1997 2002 2007 2012
time (year)
subsid
ence (
m) QP - Measured
QP - Numerical
North - Numerical
South - Numerical
Vertical displacements predicted vs measured • Subsidence evolution recorded at seafloor
• From PO-035105 document
– QP (524204,6237070)
– North (522181,6242020)
– South (527011,6231800)
Vertical division of the domain for parallel analysis
Horizons surfaces Isosurfaces of vertical displacement
16
AVOA modelling • Map reflection coefficients and
fast P-wave direction
– Using layer-matrix approach
– Elasticity between chosen layers
Base Miocene
Top Chalk 2130ms horizon
17
Microseismicity: • Valhall reservoir
• Geomechanics and
microseismicity
18
• Likely not all CCS sites will be like Sleipner
• Large scale projects:
– Not always ideal porosity/permeability
– Injection rates and volumes
– Fluid-rock (e.g., geochemical) alterations
• Must be capable of monitoring strains and overburden
effects
– Monitor seal integrity
– Evaluate CO2 containment
Recap
19
• Rock physics transforms – Models for multiphase fluids, patchy saturation, attenuation, anisotropy,
plasticity
– More data (e.g., shallower depths – overburden)
– Calibration with in-situ data (not only core samples)
– Scaling (static measurements to dynamic measurements) – significant knowledge gaps, but can apply site specific empirical relationships
• Hydro-mechanical models: – Calibration/history match fluid-flow and geomechanical simulation models
– Systematic approach to model building (i.e. geometry & meshing)
– Constitutive models from rock and petro-physics with up-scaling
– Uncertainty of model parameters and geophysical/seismic attributes
20
Challenges, uncertainties & way forward
21
Valhall reservoir, North Sea
• ELFEN-VIP one-way coupling
– Hexahedra mesh (6 million elements)
– Cap plasticity, non-associated flow rule, water
weakening SR3 adjusted to Valhall model (ISAMGEO
PO-035105)
– Soft coupled to 500,000 grid VIP model
– VIP pore pressure output monthly
• Pore pressure used by Elfen as applied load
• ELFEN solved in parallel (4 domains)
• 3 element groups are mapped (VIP-Elfen)
– Reservoir Tor
– Reservoir Hod
– Reservoir faults
Overburden
Reservoir &
Sideburden
Underburden
Domain partitions for
parallel analysis
VIP model
22
Elastic/Elasto-plastic properties
Elastic
Elasto-
plastic
Elastic
Elasto-
plastic
Elastic
23
Microseismicity: • Weyburn CCS pilot
• Geomechanics and microseismicity
Overburden after production
Overburden after CO2 injection
Overburden after shut-off
Ove
rbur
den
frac
ture
pot
entia
l
Mic
rose
ism
ic lo
catio
ns
24
Numerical modelling of fracture growth and caprock integrity during CO2 injection
UKCCS RC – “Geophysical Modelling for CO2 storage, monitoring and appraisal” University of Leeds, Leeds, UK November 3rd, 2015 Adriana Paluszny, Saeed Salimzadeh, Thibaut Defoort, Morteza Nejati, Robert W Zimmerman
CONTAIN – EPSRC
British Geological Society – Imperial College – Cardiff University Experimental – Numerical – Societal
… to undertake and disseminate research in the computational modelling
of poro-elastic behaviour of the caprock during reservoir depletion and its subsequent reinflation due to CO2 injection …
Specifically, the aim of IC is to:
• evaluate caprock failure as a function of long-term geomechanical deformation for a range of injection scenarios.
These will be validated using data generated by the British Geological
Society, will be used as a basis to inform the public about CCS
CSMP++
• C++ based numerical library for finite element & finite volume methods • Unstructured grids • Discrete fracture representation • THMC applications • Core developers (~2-3) • Numerical methods developers (~4-6) • Application programmers (~17) • Users (~100)
Two-phase flow Numerical methods Core development
Black-oil Parallelization
Computational mechanics
Reactive, compositional high temperature transport
IC Geomechanics Library and CSMP++
This project will utilise CSMP++ (Complex Systems Platform), an object-oriented finite-element based library that is specialised to simulate complex multi-physics processes.
It has already been validated to model transport, single-phase and
multiphase flow in three dimensions, and can operate on workstations as well as on high performance computing systems.
The geomechanics library developed at Imperial College, integrated with
CSMP++, is capable of simulating the growth and interaction of multiple discrete 3D meso-scale fractures.
Numerical modelling of fracture growth
Compatible with flow within the fractures
Growth Principles
Fracture pattern (2D)
[Paluszny & Matthai, IJSS, 2009]
Polygonal patterns
Polygonal fracture growth due to shrinkage of the matrix. Mean stress contours are plotted with the polygonal fracture pattern. Stress concentrates ahead of the fracture tips.
Limitation in 3D: SIF Computation mesh
Reduced Virtual Integration Technique (RVIT)
[Paluszny & Zimmerman, CMAME, 2011]
I-integral for stress intensity factor computations
[Paluszny & Zimmerman, CMAME, 2011; Nejati et al., IJSS, 2015]
This allowed to reduce computation time, increase accuracy and improve robustness of the growth engine
Volumetric Domain J-Integral now is I-Integral over virtual disk
Fracture Growth (3D)
[Paluszny & Zimmerman, CM, 2013]
Fracture set growth
[Paluszny & Zimmerman, Engineering Fracture Mechanics, 2013]
Fracture-driven fragmentation (3D)
shapes
Velocity-dependent fragmentation pattern [Paluszny & Zimmerman, Computational Mechanics, 2013]
Key CCS Improvements to the ICGT core 2015
(1) Accurate fluid pressure dependent stress intensity factor computations (2) Poroelastic coupled deformation (3) High-accuracy friction model (4) Initial validation using Goldeneye field data (Shell) With work contributed by AP+SS and PhD students: Morteza Nejati and Thibaut Defoort
CONTAIN: “During the first two years, the focus will be on the extension, integration, and validation of existing flow and propagation kernels.”
Multiple fracture growth – coarse mesh but accurate SIFs
- Peak: 320k nodes - Runtime: 10 hours
(minutes to run)
All simulations run on a Dell Precision Workstation (2013) with a maximum of 8 cores dedicated to one job.
Mechanical Variables Poisson’s ratio Density Young’s Modulus UCS Fault Friction
Related Publications
Nejati, M., Paluszny, A., Zimmerman, R.W. (2015) “A disk-shaped domain integral method for the computation of stress intensity factors using tetrahedral meshes”, Int. J. Solids Struct., 69-70, 230-251.
Nejati, M., Paluszny, A., Zimmerman, R.W. (2015) “On the use of quarter-point tetrahedral finite elements in linear elastic fracture mechanics”, Eng. Fract. Mech., 144, 194-221.
Tang, X.H., Paluszny, A., Zimmerman, R.W. (2014) “An impulse-based energy tracking method for collision resolution”, Comput. Meth. Appl. Mech. Eng., 278, 160-185.
Paluszny, A., Tang, X.H., Zimmerman, R.W. (2013) “Fracture and impulse based finite-discrete element modeling of fragmentation”, Comput. Mech., 52(5), 1071-1084.
Paluszny A, Zimmerman RW (2013) Numerical fracture growth modeling using smooth surface geometric deformation", Engineering Fracture Mechanics (available online).
Nejati M, Paluszny A, Zimmerman RW (2013) Theoretical and Numerical Modeling of Rock Hysteresis Based on Sliding of Microcrack" 47th U.S. Rock Mechanics / Geomechanics Symposium (ARMA), San Francisco, USA, 23-26 June.
Zimmerman RW, Paluszny A (2012) Some New Developments in Modelling the Failure, Fracture and Fragmentation of Rocks", 7th Asian Rock Mechanics Symposium, Invited Paper, Seoul, Korea, 15-19 October.
Paluszny A, Zimmerman RW (2011) Numerical simulation of multiple 3D fracture propagation using arbitrary meshes", Computer Methods in Applied Mechanics and Engineering, Vol:200, Pages:953-966.
Disclaimer
Some slides have been removed from the original presentation.