Reservoir Characterisation and Modelling for CO 2 Storage

Transcription

1 Reservoir Characterisation and Modelling for CO 2 Storage Tess Dance IEA CCS Summer School Perth, December 2015 ENERGY

2 Why build subsurface models? To simulate fluid flow To estimate capacity Predict reservoir response To ascertain uncertainty To know when and where to monitor To impress stakeholders: communication tool

3 Seek understanding! The purpose of computing is insight, not numbers Richard Hamming. The hardest part of modelling is to develop an intuition for the physical processes.

4 Models should be fit for purpose 1. To address scientific questions in a generic context e.g.: the effect of barriers on vertical migration of CO 2 the effect of a hydrodynamic gradient on CO 2 migration Theory of convection mixing 4 Tess Dance Image Reference: Cardoso& Andres, 2014, Nature Communications Volume: 5

5 Generic modelling example: Convection of dissolved CO 2 k h =100 md, k v =50 md 27 years 90 years 5 Slide courtesy of Jonathan Ennis-King, CSIRO

6 Models should be fit for purpose 2. To make technical predictions in a site-specific context to support decisions e.g.: What is the breakthrough time of CO 2 in an EOR project for deep injection? What is the effect of wellspacing on the maximum injectivity? What is the predicted seismic response? 6 Tess Dance Image source:

7 Site Specific Modelling Example: Breakthrough at monitoring well Reference: Hosseini et al., Static and dynamic reservoir modeling for geological CO 2 sequestration at Cranfield, Mississippi, U.S.A. Int. J. Greenh. Gas Control, 7 Tess Dance

8 Models can be simple... e.g. Analytical models q r c,max h(r,t) h to the mobility contrast between brine and CO 2 (N r c, max ( t ) k rc c w k qt rw h Nordbotten et al. (2005)

9 Example A basin-scale region investigated. Bottom hole pressure the limiting factor on injectivty and capacity. Faults sealing or not sealing the uncertainty. Multiple scenarios investigated Source: K. Michael, CSIRO, using MonteCarbon modelling 9 Tess Dance

10 or more complex e.g. 3D Numerical Models Solves a large set of linear equations at a number of given time-steps for a large number of cells Computationally demanding Can be coupled Example from the Gorgon Project: Planned to inject and store 3.4 MT PA Site located beneath an A Class nature reserve Pressure management employed Long-term modelling and monitoring required Reference: Flett, M. A., et al. (2008, January 1). Gorgon Project: Subsurface Evaluation Of Carbon Dioxide Disposal Under Barrow Island. Society of Petroleum Engineers. doi: / ms 10 Tess Dance

11 Governing equations & computer codes Starts with initial conditions: temp, pressure, salinity, saturations Mass and energy conservation Transport law (e.g. Darcy s law) Equation of state for fluids The heart of this method is the solution of a big set of linear equations What (physical) processes would you like to simulate? + heat transport Flow + transport + mineral reactions + fault reactions + dissolution + temperature Code comparison Tested against analytical models Tested against previous field studies

12 Typical issues Code limitations Non-uniqueness Inappropriate boundary Over-simplification Not properly calibrated No relative permeability and capillary pressure data Zhou et al. (2008)

13 The Foundation: A Geological Model Aims to: Capture effects of structure, stratigraphy, sedimentary architecture, petrophysical properties Reservoirs and seals Lateral and vertical heterogeneity Faults & fractures 13 Tess Dance

14 What it is we are trying to represent? Pore geometry Sedimentary structures Rock texture 7mm

15 Reservoir heterogeneity: bedding 2 m Top Images: Bottom image: 15 Tess Dance

16 Parasequences: Intraformational seals (baffles) increase length of CO 2 migration pathways & potential for Sgr and dissolution CO 2 injection well Lakes Entrance Formation Image from: C. Gibson-Poole

17 Flow Unit Architecture From: 17 Tess Dance

18 Conceptual geological model Simple 1D models K v / K h << 1 18 Tess Dance Chris Green, CSIRO

19 Data Point Set Some Alternatives Simple 3D models Pure Object Models Network Models 19

20 The 3D static geological model 3D representation(s) of the subsurface Each cell contains values for geographic position, depth, volume, rock type, poro/perm, and other static properties. Size and complexity may be a limiting factor. Grid resolution a key decision: detail vs. computational limits.

21 Pillar Gridding Structural model Stratigraphic zones and horizons Grid orientation Producer Injector

22 Discretisation & parameterisation Each grid block only has one value for porosity, permeability, saturation, composition etc. This has two important consequences: We cannot resolve anything in the results below the size of a grid block, i.e. may need to refine grid in areas of interest. Geological data measured on different scales e.g. core data, has to be upscaled or averaged in an intelligent way.

23 Stochastic modelling Truncated Gaussian Simulation Sequential Indicator Simulation Indicator Kriging

24 Object Facies modelling No matter what prediction technique we apply to a variable we are unlikely to achieve an acceptable result unless we take geological effects into account. (Houlding, 1994)

25 Input data types Hard Data: Direct measurement from the sub surface: Cores (metres), cuttings (a few mms), Plugs (10s cms) fluid samples

26 Data types Soft data Indirect interpretation of the rock and fluid properties from geophysical & petrophysical measurements: Well logs 3D seismic, MT, gravity, Electromagnetics, Remote sensing Eumeralla Timboon Paaratte Pember Formation Massacre Dilwyn Sandstone Formation Mudstone Waarre Shale - SB1 Frm A Eumeralla Pebble Skull Mepunga Timboon Paaratte Pember Belfast Dilwyn Creek Formation Massacre PointSandstone Formation mudstone Mudstone Shale - SB1 NAYLOR SOUTH 1 [MD] MD GR litho_facies [U] (From property) MD 2020 Flaxmans - Waarre D Flaxmans - Waarre D Eumeralla Pebble Skull Timboon Mepunga Paaratte Pember Belfast Dilwyn Creek Formation Massacre PointSandstone Formation mudstone Mudstone Shale - SB1 CRC-1 [MD] GR Vsh 2.88 litho_facies 1.40 GR VCL 0.81 litho_facies litho_facies [U] (From property) NAYLOR 1 [MD] MD 2004 Flaxmans - Waarre D GR Vsh 1.26 litho_facies GR litho_facies [U] (From property) Flaxmans - Waarre D Flaxmans - Waarre D Top Potential Reservoir Top Potential Reservoir Top Potential Reservoir Waarre - Waarre Cb Waarre - Waarre Cb Top Potential Reservoir Top Potential Reservoir Waarre - Waarre Cb Waarre - Waarre Cb Waarre - Waarre Cb 2100 Top IVF Top IVF Top IVF 2080 Waarre B - SB 2 Waarre B - SB Top IVF Top IVF Attribute 2120 Waarre B - SB 2 Waarre A Waarre B - SB Waarre A Waarre B - SB 2 Waarre A

27 Shot Receiver Layer 1 Seismic - Trough + Peak Layer 2 Layer 1 Layer 2 Impedance Increase Layer 3 Layer 4 Layer 2 Layer 3 Impedance Decrease Impedance = Velocity * Density Layer 3 Layer 4 Impedance Increase Miall, A.D., Principles of Sedimentary Basin Analysis

28 Exploration wells are on the highs, injection wells in the lows Data Point Exploration Well Location Proposed CO 2 Injection Well Location Proposed CO 2 Injection Well Location, The reservoir (and seal) characteristics seen in the exploration well may not be the same as those that will be encountered in the injection well(s) or by the migrating CO 2

29 Outcrop Analogues Top Image Source: Austin Chalk Getting Another Look Bottom Image Source: Outcrop analog for an oolitic carbonate ramp reservoir

30 Modelling workflow Small-scale (pilot projects) Vs Basin-scale Anderson & Woessner (1992)

31 Small is beautiful? Resist the one big model temptation. Multi-disciplinary workflows encourage big models you can t easily iterate. Early models should be small so you can run a lot of them, and investigate sensitivities. Mature models can be bigger, but still allow for a suite of models.

32 Closing remarks Models must be fit for purpose Models are useful tools BUT be aware of the shortcomings The static model is a living repository for integrating all relevant information: feed the beast! Get to know your uncertainty Models provide insight not answers. 32 Tess Dance

33 Thank you Acknowledgments: Jonathan Ennis-King Karsten Michael Tara La Force IEA CCS SUMMER SCHOOL Tess 33 Dance