Sleipner Benchmark Dataset and Model Comparison IEAGHG Modeling and Risk Assessment Network Meeting 10-13 June 2013 Sarah Gasda, Center for Integrated Petroleum Research, Uni Research, Bergen, Norway
Sleipner CO 2 Injection Injection began in 1996 0.9 Mt/yr injection rate special topic Over 14 Mtons stored through 2012. C0 2 Sequestration Arts et al. 2008 Figure 2 Scheme of the CO 2 injection at Sleipner. www.co2crc.com.au Arts et al. 2008
Injection with Data Collection Accumulated mass (million tons) 11 10 9 8 7 6 5 4 3 2 1 CO 2 Injection rate ~0.9 Mt/yr Injected CO2 Cumulative volume stored >14 Mt Continuous monitoring programme seismic surveys 0 1996 1997 1998 1999 2001 2002 2003 2004 2005 2006 2007 2008 Date seafloor Seismic mapping surveys highres seismic CSEM survey gravity surveys CSEM survey Gravity surveys Seafloor mapping
Depth Thickne Utsira Formation Shallow marine sandstones ~200-m thick saline aquifer Overlain by Nordland shale sequence Average porosity 34% Permeability 1-5 Darcy 98% sand content Depth Intra-formational shales Upper sand wedge above 5-m shale Extension of the Utsira Sand 100 km Thickness Rob Arts, Andy Chadwick & Ola Eiken E 134891 3 e reservoir consists of two reservoir units: a main lower 'Utsira Sand' unit and an upper 'Sand Wedge' unit, separated by a 6.5 m le unit (Figure 1). A number of thin intra-formational shale layers (approximately 0.5 to 2 m thick) are present within the Utsira nd unit. These appear as spikes on the gamma-ray log, and affect the flow behavior of the CO 2 in the reservoir. Chadwick et al. 2004 Singh et al. 2010 100 km Total area of the Utsira Sand is approximately 26.000 km 2
Utsira Modeling and Monitoring Pre-injection simulation study Eclipse 100, 250 m gridblocks 20 yr injection plume < 3 km in diameter 30 bar wellhead pressure increase 18% dissolution Since then Arts et al. 2008 Baklid and Korbøl, 1996
Layer 9 Benchmark Cumulative Mass (Mt) 1.8 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 Singh et al. 2010
CO 2 Storage Benchmarks Pruess, et al. An Intercomparison Study of Simulation Models for Geologic Sequestration of CO 2, Energy 2004. Numerous problems covering a multitude of storage issues. Class et al. A benchmark-study on problems related to CO 2 storage in geologic formations, Comp. Geosci., 2009. Flow simple leaky well, and more realistic formations Nordbotten et al. Uncertainties in practical simulation of CO 2 storage. IJGCC, submitted. CO 2 flow and trapping in a simple brine aquifer
Layer 9 Simulations Conformance: History match observations and simulations observation flow simulations depth contours Core permeabilities 2-3 Darcy Well permeabilities 1-8 Darcy 3 Darcy 3 Darcy E-W 10 Darcy N-S 3/10 Darcy (higher temp) 1. Single feeder in model 2. Baseline depth map assumes laterally uniform seismic velocity in overburden. Introducing velocity changes within Chadwick known range of velocity et variation al. 2009 could modify depth map to allow observed lateral migration. TOUGH2 Migration simulator Eclipse 100/300 Others SPE 134891 Singh et al. 2010
Sleipner Benchmark Downloads 38 downloads as of early 2013 11 countries Europe, N. America, and Asia Number of downloads US Norway Canada Australia Germany UK France Russia Denmark Spain Korea
Layer 9 Benchmark Results Model Institution Processes Assumptions Sensitivities Comments MUFITS (Multiphase Filtration Transport Simulator) Moscow State University CO 2 dissolution in water viscous and gravity forces nonisothermal (conduction and heat of dissolution) no NaCl solution in water, no capillary pressure, no mineralization initial formation temperature (305-320 K) assumes CO 2 is heated upon injection VESA (Vertical Equilibrium w/subscale Analytical Model) Uni CIPR/UiB Vertical Equil Isothermal Sharp Interface Upscaled Convective Dissolution gravity segregated no capillary pressure porosity permeability topography CO 2 density MRST (Matlab Reservoir Simulator Toolbox) SINTEF Vertical Equil Black Oil Compositional depending on modeling choice temperature temporal available as opensource Matlab simulator (MRST- VE) download from SINTEF MPATH/ ECLIPSE Permedia/Statoil invasion percolation Eclipse 100/300 viscous flow is neglected Darcy-flow dominates pressure MPATH used to model full-field percolation More to come
Seismic Data Migration Simulator MUFITS Temperature=36.8 C, Mean CO 2 Density = 224-436 kg/m 3 VESA Porosity = 36%, CO 2 Density = 700 kg/m 3 5 0 Incomp seq exp 5 Fully implicit VE 0 Fully implicit comp5 Fully implicit VE 0 Fully implicit comp5 Fully implicit VE 0 Fully implicit comp Fully implicit VE 4 4 4 4 MRST 4000 4000 4000 4000
Black Oil with pressure decay 2006 calibration 2010 prediction 2008 calibration 2012 prediction Cavanagh, 2012
Sensitivity to CO 2 density and porosity (VESA) Top of Utsira Base Case Low Porosity Low CO 2 density Low porosity and 36%, 700 kg/m 3 27% 400 kg/m 3 CO 2 density
Uncertainty in Temperature (MUFITS)
Sensitivity to Temperature and Simulation Method (MRST) 5 0 4 4000 T = 289.85 rhow = 1200 T = 289.85 rhow = T = 317.5 rhow = 1200 T = 317.5 rhow = 700 600 Disolved Struct. residual Residual Free residual Struct. movable Free movable Leaked Volume (M m 3 ) 400 300 200 100 0 10 20 30 40 50 60 70 80 90 100 Years since simulation start
Sensitivity to top surface topography (VESA) (i) (ii) (iii) (iv) Minimize traps and spillpoits Minimize traps Minimize spillpoints Smoothed surface
2004 5 0 4 2008 2014 2040 2250 Fully implicit comp Fully implicit VE 5 0 4 Fully implicit comp Fully implicit VE 5 0 4 Fully implicit comp Fully implicit VE 5 0 4 Fully implicit comp Fully implicit VE Fully implicit comp Fully implicit VE 4000 4000 4000 4000 Volume (M m 3 ) Fully implicit comp Fully implicit VE Disolved Struct. residual Residual Free residual Struct. movable Free movable Leaked 30 years of injection 220 years of post-injection trapping increases, while a good portion of CO2 flows out of the domain 0 50 100 150 200 250 Years since simulation start
Summary and Conclusions See a good scientific focus from the participants Gravity-dominated flow well-suited for percolation and VE models Understanding uncertainty is important for calibrating the simulators Progress with the Sleipner reference model is good, despite relatively few results to date. Feedback to Statoil can help improve the utility of the dataset in the future.
Acknowledgements The Sleipner Benchmark was provided by Statoil and is managed by IEAGHG Geomechanics benchmark, please contact if interested.