Geological models, heterogeneities and scale relations
Transcription
Geological models, heterogeneities and scale relations
Geological models, heterogeneities and scale relations Peter Frykman Geologist & Geo-engineer GEUS - Geological Survey of Denmark and Greenland This is what you will hear • • • • • Motivation Why do geological modelling? How to do it? What do we learn? What is special for CO2 storage? Tax deduction for solar panels On June 1, 2005, Governor Arnold Schwarzenegger signed Executive Order # S-3-05, greenhouse gas targets: By 2010, Reduce to 2000 Emission Levels By 2020, Reduce to 1990 Emission Levels By 2050, Reduce to 80 percent Below 1990 Levels BUT, hydrogen has different colours! His hydrogen converted Hummer Sept 2009 • “We're investing billions to capture carbon pollution so that we can clean up our coal plants “ Why modelling? • It is the only way to ensure the public, regulators and politicians about safety – Predictions for the future – Estimate capacity – Quantify probability for events and risks Important? modelling How? • Model properties – Size – Geometry – Scale • Flow and reaction processes -> other presentations this week How to build a model • Building blocks (cells in grid) assigned a property – e.g. porosity or permeability Decisions on model characteristics • • • • How large gridblocks Size of model Computing capability What is the question for the modelling Model building • Saline aquifers for storage often have very few data at initial stage • We need methods to populate the whole model with properties – Deterministic (hand-drawing) – Interpolation – Geostatistical modelling (random+rules) • Object models • Voxel models Terminology • A geological model can be a qualitative characterisation – “Fluvial setting developing into estuarine and shallow marine” • But it can also be a numerical model of the facies geometry and their properties, which can be used for flow calculations and other calculations Geological model – Fluvial environment with sand deposits Numerical model – Fluvial facies geometry ready for flow simulation How to use geology in modelling Example: • Regional model = Flood plain with river deposits • Core description: sand, crossbedded, and nonmarine mudstone • Environment = Fluvial • Geometry = Channel belts with river deposits of sandy pointbar systems and channel fills • Translate that geometry into a numerical model and add some properties Input Scandinavian highlands Regional conceptual model Floodplain with incised valleys With fluvial deposits Berlin Marine influx Input for channel belt simulation Channel belts, not the individual river channel, used as objects in model 100 - 1600 m width Thickness = 1 - 8 m Channel belt param. Well data and N/G sand 3500 BM_a BM_b fluvsim_orig Robinson & McCabe 1997 12.1 1.9 Channel Belt Width 3000 2500 2000 1500 1000 500 0 0.0 2.0 4.0 6.0 8.0 Pointbar thickness m 10.0 12.0 Pointbar deposits http://frg.leeds.ac.uk/ http://frg.leeds.ac.uk/ Heterogeneities! Sand Floodplain clay and silt Crevasse splay Pointbar top 5m Sand Sand Triassic fluvial channel margin Floodplain Sand Point-bar Amalgamation Sand Shallow seismics of reservoir sands V. Kolla, H.W. Posamentier and L.J. Wood, (2007), Deep-water and fluvial sinuous channels characteristics, similarities and dissimilarities, and modes of formation, Marine and Petroleum Geology Deep-water example of meandering channel Simple solution: Low-sinuosity channel geometry ~ Channel belts Anisotropic patchy permeability with SGSIM ~ Pointbars mapview 5 km 3D model of fluvial system 1 layer Porosity Permeability Model upscaling • We like to model details to reflect the small-scale variations so that it looks realistic to the geologist • BUT – too many gridcells for efficient computations • Upscaling! From 48 to 6 blocks • Upscaling by averaging is common • Looks simple, but is usually more complicated – E.g. directional permeability etc Permeability model upscaled into flow model grid, having non-uniform grid Kx in top layer of model 2 3 4 Claudia Schiffer 2 There are models everywhere 3 4 Carla Bruni-Sarkozy Models look great at a distance, but they are never perfect • It is too easy to make models! Not symmetric Excess pore space Wrinkles Challenge for the modeller • How to best represent the most influential geological features in a model, • A model that can be used for flow simulation and prediction of CO2 injection & movement, • and can be used to evaluate site performance (capacity, injectivity, risk studies) • • • • Reservoir Caprock Faults (Wells) Site Model components + for Earth Model: • Overburden What else? • Site model looks OK “at a distance”, but what have we missed? 10 m – 100 scale Site model Region scale Meter scale 120x190 km region model Small scale processes Filling efficiency and Capillary trapping Small-scale effects due to sub-meter scale heterogeneities Flow simulation: 1) Fill with CO2 2) Let aquifer move in from below 3) Notice the above-endpoint residual saturations in the sand compartments Well known from oil/gas that finescale structure down to mm and cm scale has en effect on the displacement of fluids. This could also be the case for CO2 /water system, but not fully studied yet Reservoir rock at very small scale: This is what we rely on – pore space Large-scale interaction • The pressure front travels much faster and further than the actual flow • With a site model of 12x16 km we must be aware that the overpressure extends far into the region around the site • The extent and magnitude are influenced by lateral connectivity (facies geometry, faults), vertical communication (seals and caprock), and compressibility of fluids and rock Region scale 120x190 km region model, pressure propagation during injection When simulating industrial scale rates, the pressure effect reaches outside the site model Back to the site model • Given that we can account for both the smallscale and the large-scale interactions, we can now asses how our site behaves • The interplay between geological structure and the site behaviour Sleipner Utsira Fm. GR log North Denmark region with interfingering of facies: Nearshore sandy sediments with continuous marine mudstones as intraformational seals S E F Danish storage formation: Reservoir section at 1700-2100 m depth Sedimentary facies Shoreface Marine mud Sleipner GR log Shoreface Estuary Fluvial 350 m section Sleipner 125 m thick Homogeneous sand with scattered thin clay layers Typical Danish site 250 m thick sand layers and intraformational seals of mudstone CO2 distributed in layers CO2 goes to top layer Geology matters! • Geology has consequences: different geological models - different CO2 distribution • Other presentations this week will deal with flow, pressure, geochemical or geomechanical effects. • So, next 3 days – keep an eye on the geology input and model assumptions What is new for CO2 site modelling compared to oil/gas • • • • • • • Few data to start with in saline aquifers Longer time perspective for predictions Pressure rise from injection Mobility and gravity override Trapping mechanisms Close symbiosis with monitoring activities (Geochemical reactions and geomechanics) Future challenges • Improve the translation of geology into numerical models • Co-modelling involving different scales • Management and description of uncertainty • Adjustments of model to a range of different monitoring data Conclusions •Confidence: we have modelling tools that work well •Clarity: we need to state assumptions for input and uncertainty for output •Site specific modelling: is the only way to assess performance and assure safety of the operation