2006 11 08 spe dl upscaling

Upload: iperm

Post on 05-Apr-2018

229 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    1/95

    SPE DISTINGUISHED LECTURER SERIESis funded principally

    through a grant of the

    SPE FOUNDATION

    The Society gratefully acknowledgesthose companies that support the program

    by allowing their professionalsto participate as Lecturers.

    And special thanks to The American Institute of Mining, Metallurgical,and Petroleum Engineers (AIME) for their contribution to the program.

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    2/95

    2 of 51

    Acknowledgements

    SPE International for the opportunity to participate in the 2006-07 DistinguishedLecturer Program

    BP America, Inc. for permission, and the Professional Recognition Program which has

    provided the time and resources to prepare and present this material

    Colleagues whose work is represented

    Mr. Escalante, the Shekou Section, and other local SPE chapters worldwide for their

    efforts in hosting these presentations

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    3/95

    Upgridding and Upscaling:

    Current Trends and Future Directions

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    4/95

    4 of 51

    Outline

    Introduction: Change of Scale & Upscaling

    Case Study: Magnus LKCF

    Validation and Analysis: What Went Wrong?

    Improved Upscaling: Understanding Permeability

    Boundary Conditions and Permeability Upscaling

    Transmissibility: Yes, Permeability: No

    Maintain the Well Injectivity & Productivity

    Magnus LKCF & Andrew Reservoir Case Studies

    Summary: What To Avoid & What Works Well?

    Future Trends:

    A Priori Error Analyses & Designer Grids

    Summary: Best Practice in Upscaling

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    5/95

    5 of 51

    Introduction: What is Upscaling?

    What is Upscaling?

    Assign effective properties to coarse scale cells from properties on

    fine scale grid

    Capture flow features of fine scale model

    Why Upscale?

    Reduce CPU time for uncertainty analysis and risk assessment

    Make fine-scale simulation practical

    geological models: ~10 -100 million cells

    Resolution?

    Image from Mike

    ChristieDW GOM

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    6/95

    6 of 51

    Why Upscale?: CPU Time Reduction

    Waterflood Field Example

    0.00

    0.10

    0.20

    0.30

    0.40

    0.50

    0.60

    0.70

    0.80

    0.90

    1.00

    0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00

    Uniform Layering Coarsen

    Optimum Layering Coarsen

    MCoarsen

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    7/957 of 51

    Upgridding and Upscaling: Context

    Structure from well picks &/or

    seismic horizons

    Properties from well logs &/or seismic

    attributes &/or field performance data

    Geologic description from facies, analogues and field data

    Performance prediction in the

    absence of dynamic data

    Starting point for a history matchwhen dynamic data is available

    , upscalingwill

    preserve the most important flow

    characteristics of a geologic model

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    8/958 of 51

    Why Upscale?: Length & Area

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    9/959 of 51

    Kanaalkop: Tanqua Karoo basin, South Africa

    Deepwater channel w/splay at top of photo

    ~15ft windmill

    ~10ft exposure

    ~250ft, which is about the size of a single cellin the areal direction of many simulation grids

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    10/9510 of 51

    10ft thick exposure of channel

    With 5 Components of a Bouma sequence

    ~10ft

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    11/9511 of 51

    Why Upscale?: Thickness

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    12/9512 of 51

    Reservoir Zones, Well Logs & Outcrop

    No Vertical Exaggeration

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    13/95

    13 of 51

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    14/95

    14 of 51

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    15/95

    15 of 51

    15 meters Geologist at Outcrop

    30 geologic model layers

    1-5 simulation model layers

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    16/95

    16 of 51

    Outline

    Introduction: Change of Scale & Upscaling

    Case Study: Magnus LKCF

    Validation and Analysis: What Went Wrong?

    Improved Upscaling: Understanding Permeability

    Boundary Conditions and Permeability Upscaling

    Transmissibility: Yes, Permeability: No

    Maintain the Well Injectivity & Productivity

    Magnus LKCF & Andrew Reservoir Case Studies

    Summary: What To Avoid & What Works Well?

    Future Trends:

    A Priori Error Analyses & Designer Grids

    Summary: Best Practice in Upscaling

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    17/95

    17 of 51

    LKCFLimit

    OWC

    MSM

    Limit

    Do We Have an Economic LKCF Waterflood Development?

    1 2 A - 9M5 :C4M1 6 :A4

    1 km

    M1 2 :A5

    -2700

    LKCF

    -2800

    UKCF

    -2900

    -3000

    MSM C-G

    -3100

    MSM A

    -3200

    B Shale

    -3300

    -3400

    Heat her / Brent

    -3500

    -3600

    -3700

    1 2 A - 9M5 :C4M1 6 :A4

    1 km

    M1 2 :A5

    -2700

    LKCF

    -2800

    UKCF

    -2900

    -3000

    MSM C-G

    -3100

    MSM A

    -3200

    B Shale

    -3300

    -3400

    Heat her / Brent

    -3500

    -3600

    -3700

    1 2 A - 9M5 :C4M1 6 :A4

    1 km

    M1 2 :A5

    -2700

    LKCF

    -2800

    UKCF

    -2900

    -3000

    MSM C-G

    -3100

    MSM A

    -3200

    B Shale

    -3300

    -3400

    Heat her / Brent

    -3500

    -3600

    -3700

    Yellow = Channel

    Red = Margins

    Blue = Non-pay

    64x64x450 = 1,843,200 cells 50mx50mx0.5m

    resolution

    M LKCF

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    18/95

    18 of 51

    Magnus LKCF

    Waterflood Development Study

    C ll P bili U li

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    19/95

    19 of 51

    A laboratory coreflood

    In three dimensions, we have three numerical corefloods

    Coreflood follows the coarse cell shapes

    No flow side boundary conditions are the most common

    (others are possible)

    Cell Permeability Upscaling:

    Laboratory and Reservoir Model

    Q

    A

    K P

    L

    Darcys Law:

    1 2

    3 4

    S li i h U l d LKCF M d l

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    20/95

    20 of 51

    Time of Flight & Pressures after conventional

    2x2x6 upscaling:

    Loss of 95% of effective permeability

    Loss of internal reservoir heterogeneity

    Coarse Scale Time of Flight

    Coarse

    Pressure

    Streamlines in the Upscaled LKCF Model

    How Well Did 2x2x6 Upscaling Work?

    3D Streamlines, Time of Flight & Pressures

    calculated in the fine scale geologic model

    2xInjectors & 2xProducers at a typical

    waterflood well spacing

    Fence diagram traced within the 3D geologic

    model

    Pressure constrained wells used to validate

    permeability

    Fine Scale Time of Flight

    C ll P bili U li

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    21/95

    21 of 51

    600

    200

    0

    0

    500

    600

    0

    300

    300

    0

    0

    300

    0 0 600 0

    Cell Permeability Upscaling

    What Went Wrong?

    KX Permeability0 100 200 300 400 500 600

    0 0 0 0 600 600 600 600 600 0 0 0

    0 0 0 0 600 600 600 600 600 0 0 0

    0 0 0 0 600 600 600 600 600 0 0 0

    0 0 0 0 600 600 600 600 600 0 0 0

    0 0 0 0 0 600 600 600 600 600 0 0

    0 0 0 0 0 0 600 600 600 600 600 0

    300 300 300 0 0 0 600 600 600 600 600 0

    300 300 300 0 0 0 0 600 600 600 600 600

    300 300 300 300 300 0 0 600 600 600 600 600

    300 300 300 300 300 300 0 0 600 600 600 600

    300 300 300 300 300 300 0 0 600 600 600 600

    300 300 300 300 300 300 0 0 600 600 600 600

    Sealed Side coreflood boundary conditions systematically expand barriers and reduce thecontinuity of pay

    Example 12x12=>4x4 (3x3 Upscaling):

    Continuous channel replaced by marginal sands

    Highly productive well replaced by poor producer

    C ll P bilit U li

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    22/95

    22 of 51

    Cell Permeability Upscaling

    Streamline Flow Visualization

    Each cell in isolation

    No cross-flow

    Equilibrium at cell faces

    Preserves & expands barriers

    12x12 => 3x3

    4x4 Upscaling

    Example

    KX Cell Permeability KY Cell Permeability

    C ll P bilit U li

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    23/95

    23 of 51

    Cell Permeability Upscaling

    Errors & More Subtle Errors

    Sealed Side Boundary Conditions do not adequately represent fluid flow in the finescale model

    Reservoir quality is not preserved

    This is the most significant error

    However, there are more subtle errors

    Needless loss of spatial resolution

    Transmissibility Upscaling

    Well Productivity (or Injectivity) is not preserved

    Well PI Upscaling

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    24/95

    24 of 51

    Outline

    Introduction: Change of Scale & Upscaling

    Case Study: Magnus LKCF

    Validation and Analysis: What Went Wrong?

    Improved Upscaling: Understanding Permeability

    Boundary Conditions and Permeability Upscaling

    Transmissibility: Yes, Permeability: No

    Maintain the Well Injectivity & Productivity

    Magnus LKCF & Andrew Reservoir Case Studies

    Summary: What To Avoid & What Works Well?

    Future Trends:

    A Priori Error Analyses & Designer Grids

    Summary: Best Practice in Upscaling

    B d C diti d

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    25/95

    25 of 51

    Boundary Conditions and

    Upscaled Permeability - 1/2

    Upscale a simple sand/ shale reservoir

    Sealed side BCs

    expand barriers

    Open linear pressureBCs allow barriers to

    leak

    Pizza box (Wide

    BCs) allow global flow

    tortuosity

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    26/95

    Tr n mi ibilit U ling 1/3

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    27/95

    27 of 51

    Transmissibility Upscaling 1/3

    Preserves Spatial Resolution

    Transmissibility can be calculated by direct upscaling instead of from the harmonic averageof cell permeabilities

    Link Permeability is upscaled from cell center to cell center and has double the lateral

    resolution compared to cell permeability upscaling

    Harmonic average of a zero cell permeability is always zero

    1

    12121

    2

    ii

    iiii

    DXKXDXKX

    DXKXDXKXATX

    1

    21

    2121

    2

    ii

    i

    iiDXDX

    KXATX

    121 )()( iii MinusKXKXPlusKX

    i 1i

    21i

    Transmissibility Upscaling 2/3

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    28/95

    28 of 51

    KX

    Sealed Cell

    KXY

    Wide No

    Shift

    Transmissibility Upscaling 2/3

    KX Streamline Flow Comparisons

    KX

    Wide

    Shifted

    KX

    Sealed

    Shifted

    Transmissibility Upscaling 3/3

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    29/95

    29 of 51

    Transmissibility Upscaling 3/3

    Captures fine scale juxtaposition

    0 MD 0 MD 38 MD 0 MD 50 MD50 MD

    Well Productivity Upscaling

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    30/95

    30 of 51

    Well Productivity Upscaling

    Used to Preserve Reservoir Quality

    Simulator well productivity calculated from sealed sidecoreflood permeability?

    Does not describe radial flow and

    logarithmic pressure drop near a well

    Instead, use three (hypothetical)X, Y, and Z wells for each coarse cell

    w

    ZZ

    rr

    HKYKXWI

    0ln

    2

    w

    Y

    Y rr

    HKZKXWI

    0ln

    2

    w

    XX

    rr

    HKZKYWI

    0ln

    2

    Improved Upscaling:

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    31/95

    31 of 51

    Lack of pay continuity resolved through Well Index Upscaling

    Preserves injectivity and productivity of horizontal and vertical wells

    But, expands channels and removes barriers

    Contrast and barriers reintroduced through Transmissibility Upscaling

    Repeat in all three directions for 2x2x2=8-fold factor of improved flow resolution compared to cell permeabilities

    Improved Upscaling:

    Well Index + Transmissibility

    KX Permeability0 100 200 300 400 500 600

    0 0 0 0 600 600 600 600 600 0 0 0

    0 0 0 0 600 600 600 600 600 0 0 0

    0 0 0 0 600 600 600 600 600 0 0 0

    0 0 0 0 600 600 600 600 600 0 0 0

    0 0 0 0 0 600 600 600 600 600 0 0

    0 0 0 0 0 0 600 600 600 600 600 0

    300 300 300 0 0 0 600 600 600 600 600 0

    300 300 300 0 0 0 0 600 600 600 600 600

    300 300 300 300 300 0 0 600 600 600 600 600

    300 300 300 300 300 300 0 0 600 600 600 600

    300 300 300 300 300 300 0 0 600 600 600 600

    300 300 300 300 300 300 0 0 600 600 600 600

    600

    467

    200

    200

    533

    600

    0

    300

    300

    200

    67

    300

    0 400 600 0

    600

    467

    200

    200

    533

    600

    0

    300

    300

    200

    67

    300

    0 400 600 0

    Coreflood Cell Permeability OR

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    32/95

    32 of 51

    Coreflood Cell Permeability OR

    Well Index + Transmissibility Upscaling

    Coreflood Cell Permeability Upscaling

    Well Index + Transmissibility Upscaling KX Permeability0 100 200 300 400 500 600

    600

    200

    0

    0

    500

    600

    0

    300

    300

    0

    0

    300

    0 0 600 0

    0 0 0 0 600 600 600 600 600 0 0 0

    0 0 0 0 600 600 600 600 600 0 0 0

    0 0 0 0 600 600 600 600 600 0 0 0

    0 0 0 0 600 600 600 600 600 0 0 0

    0 0 0 0 0 600 600 600 600 600 0 0

    0 0 0 0 0 0 600 600 600 600 600 0

    300 300 300 0 0 0 600 600 600 600 600 0

    300 300 300 0 0 0 0 600 600 600 600 600

    300 300 300 300 300 0 0 600 600 600 600 600

    300 300 300 300 300 300 0 0 600 600 600 600

    300 300 300 300 300 300 0 0 600 600 600 600

    300 300 300 300 300 300 0 0 600 600 600 600

    600

    467

    200

    200

    533

    600

    0

    300

    300

    200

    67

    300

    0 400 600 0

    0 0 0 0 600 600 600 600 600 0 0 0

    0 0 0 0 600 600 600 600 600 0 0 0

    0 0 0 0 600 600 600 600 600 0 0 0

    0 0 0 0 600 600 600 600 600 0 0 0

    0 0 0 0 0 600 600 600 600 600 0 0

    0 0 0 0 0 0 600 600 600 600 600 0

    300 300 300 0 0 0 600 600 600 600 600 0

    300 300 300 0 0 0 0 600 600 600 600 600

    300 300 300 300 300 0 0 600 600 600 600 600

    300 300 300 300 300 300 0 0 600 600 600 600

    300 300 300 300 300 300 0 0 600 600 600 600

    300 300 300 300 300 300 0 0 600 600 600 600

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    33/95

    33 of 51

    Outline

    Introduction: Change of Scale & Upscaling

    Case Study: Magnus LKCF

    Validation and Analysis: What Went Wrong?

    Improved Upscaling: Understanding Permeability

    Boundary Conditions and Permeability Upscaling

    Transmissibility: Yes, Permeability: No

    Maintain the Well Injectivity & Productivity

    Magnus LKCF & Andrew Reservoir Case Studies

    Summary: What To Avoid & What Works Well?

    Future Trends:

    A Priori Error Analyses & Designer Grids

    Summary: Best Practice in Upscaling

    LKCF Upscaling Validation

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    34/95

    34 of 51

    LKCF Upscaling Validation

    Well Index + Transmissibility

    3D Streamlines & Time of Flight

    Comparison of:

    Fine Scale Model

    Coreflood Cell Perm Upscaling

    WI + Transmissibility Upscaling

    Fine Scale Time of Flight

    Coarse Scale Time of Flight

    Coarse

    Pressure

    Coarse Scale Time of Flight

    Coarse

    Pressure

    Transmissibility Multipliers:

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    35/95

    35 of 51

    Transmissibility Multipliers:

    Double the Spatial Resolution

    A transmissibility multiplier can represent a barrier without using a cell

    In contrast, zero vertical permeability prevents flow both up AND down and

    impacts flow in three layers

    Andrew Reservoir:

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    36/95

    36 of 51

    Andrew Reservoir:

    Validation & Impact of Thin Barriers

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    37/95

    37 of 51

    Outline

    Introduction: Change of Scale & Upscaling

    Case Study: Magnus LKCF

    Validation and Analysis: What Went Wrong?

    Improved Upscaling: Understanding Permeability

    Boundary Conditions and Permeability Upscaling

    Transmissibility: Yes, Permeability: No

    Maintain the Well Injectivity & Productivity

    Magnus LKCF & Andrew Reservoir Case Studies

    Summary: What to Avoid & What Works Well?

    Future Trends:

    A Priori Error Analyses & Designer Grids

    Summary: Best Practice in Upscaling

    Summary:

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    38/95

    38 of 51

    Summary:

    What to Avoid

    Flow based coreflood upscaling for cell permeabilities

    Sealed side boundary conditions will not preserve flow tortuosity & will under-

    estimate reservoir quality

    Open linear pressure boundary conditions will not preserve reservoir barriers

    A single upscaling calculation cannot be used to preserve:

    Reservoir quality

    Reservoir barriers

    Tortuosity of reservoir fluid flow around barriers

    Unfortunately, using coreflood permeability upscaling is the most common practice in

    the industry

    Summary:

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    39/95

    39 of 51

    Summary:

    What Works Well

    Preserve connectivity and flowwithinthe reservoir using flow based transmissibilityupscaling

    Select boundary conditions to either preserve flow tortuosityorflow barriers

    Preserve reservoir quality and flowbetweenreservoir and wells using algebraic well

    index upscaling

    This combination of techniques has worked well within BP & similarly elsewhere in the

    industry

    Streamline calculations provide detailed validation based on pressures, sweep, and

    time of flight

    Validation after upscaling is always necessary

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    40/95

    40 of 51

    Outline

    Introduction: Change of Scale & Upscaling

    Case Study: Magnus LKCF

    Validation and Analysis: What Went Wrong?

    Improved Upscaling: Understanding Permeability

    Boundary Conditions and Permeability Upscaling

    Transmissibility Yes, Permeability No

    Maintain the Well Injectivity & Productivity

    Magnus LKCF & Andrew Reservoir Case Studies

    Summary: What To Avoid & What Works Best?

    Future Trends:

    A Priori Error Analyses & Designer Grids

    Summary: Best Practice in Upscaling

    Future Trends:

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    41/95

    41 of 51

    Assumption Source of Error(Missing Physics)

    Pressure equilibrium withinthe coarse cell

    Disconnected pay within the coarse cellwill not be in equilibrium

    Fluid velocity is parallel to

    the pressure drop

    Flow may depend upon the transverse

    pressure drop on the coarse grid Single velocity within a

    coarse cell Distribution of multiphase frontal

    velocities replaced by a single value

    Future Trends:

    A Priori Error Analyses & Designer Grids

    Wouldnt it be nice to know if an upscaling calculation would be a goodapproximation beforeyou performed the upscaling calculation?

    Sources of Upscaling Error

    Error from Layer Coarsening:

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    42/95

    42 of 51

    Error in the velocity distributionis introduced while upscaling

    Different fluid velocities are replaced by a single value

    F(S) Kx/Phi is the frontal speed in each layer This is the property whose heterogeneity we will analyze

    Analysis applies to the net sands

    Vertical equilibrium within each coarse cell

    Error from Layer Coarsening:

    Flood Front Progression

    Fast

    Slow

    Medium

    Fast

    Slow

    Medium

    XW KSF *

    Designer Grids within the Flow Simulator

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    43/95

    43 of 51

    g

    Static Boundary Conditions

    Design simulation layering from 3D geologic model to minimize variation in local multiphase

    frontal velocities

    249, 80%336, 86%

    0

    10

    20

    30

    40

    50

    60

    70

    80

    90

    100

    0 200 400 600 800 1000 1200 1400 1600

    Model Layers

    %-Heterogeneity

    0

    2

    4

    6

    8

    10

    12

    14

    16

    18

    20

    RMSRe

    gression-Error

    Li & Beckner % Heterogeneity

    % Heterogeneity ; B-Variation

    % Heterogeneity: Uniform Coarsen

    Diagonal Guide

    Solution Total RMS Regression

    Solution Weighted RMS Regression

    Total RMS Regression

    Weighted RMS Regression

    Number of Coarse Layers

    %-Heterogeneity

    RMSRegressionErrorOptimal Layering

    Uniform Coarsening: Not Efficient

    Designer Grids within the Flow Simulator

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    44/95

    44 of 51

    g

    Upscale During Initialization (Static)

    General trend shows that uniform coarsening does not perform well

    Optimal (293 layers) is the best layering scheme

    Flexible 3D grid (MCOARSE) provides even better results

    Tight Gas Layer Coarsening

    Fine Scale Model 22x23x1715 (Geological Scenario 5)

    0

    5

    10

    15

    20

    25

    30

    0 200 400 600 800 1,000 1,200 1,400 1,600 1,800

    Model Layers

    Cum.

    Ga

    sProd.

    (BCF)

    Regular-CoarsenNextVar-OneStepNextVar-SequentialOptimal-12LOptimalLi-Map-12LLi-Ave-MaxLLi-Ave-12LMCOARSE

    Fine Scale

    Li and Beckner

    UniformMCOARSE

    Optimal

    Layer Coarsening:

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    45/95

    45 of 51

    y g

    Waterflood Example

    Waterflood Field Example:

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    46/95

    46 of 510%

    10%

    20%

    30%

    40%

    50%

    60%

    70%

    80%

    90%

    00%

    0 2000 4000 6000 8000 10000 12000

    FineScale

    Coarsen_54

    Coarsen_22

    Coarsen_22U

    Coarsen_31

    Coarsen_19

    Coarsen_07

    p

    Oil Recovery and Watercut

    Optimal Simulation Model has 22 layers

    7 layers and 22 uniform layers are each too coarse

    0%

    5%

    10%

    15%

    20%

    25%

    30%

    0 2000 4000 6000 8000 10000 1200

    FineScale

    Coarsen_54

    Coarsen_22

    Coarsen_31

    Coarsen_19

    Coarsen_07

    0%

    5%

    10%

    15%

    20%

    25%

    30%

    0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00

    FineScale

    Coarsen_54

    Coarsen_22Coarsen_31

    Coarsen_19

    Coarsen_07

    1760

    1780

    1800

    1820

    1840

    1860

    1880

    1900

    1920

    1940

    1960

    1980

    0 2000 4000 6000 8000 10000 12000

    FineScale

    Coarsen_22

    Coarsen_22U

    Designer Grids within the Flow Simulator

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    47/95

    47 of 51

    g

    Static Boundary Conditions

    Design 3D simulation grid to prevent different sands from merging

    Designer Grids within the Flow Simulator

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    48/95

    48 of 51

    g

    Upscale During Initialization (Static)

    General trend shows that uniform coarsening does not perform well

    Optimal (293 layers) is the best layering scheme

    Flexible 3D grid (MCOARSE) provides even better results

    Tight Gas Layer Coarsening

    Fine Scale Model 22x23x1715 (Geological Scenario 5)

    0

    5

    10

    15

    20

    25

    30

    0 200 400 600 800 1,000 1,200 1,400 1,600 1,800

    Model Layers

    Cum.

    Ga

    sProd.

    (BCF)

    Regular-CoarsenNextVar-OneStepNextVar-SequentialOptimal-12LOptimalLi-Map-12LLi-Ave-MaxLLi-Ave-12LMCOARSE

    Fine Scale

    Li and Beckner

    UniformMCOARSE

    Optimal

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    49/95

    49 of 51

    Outline

    Introduction: Change of Scale & Upscaling

    Case Study: Magnus LKCF

    Validation and Analysis: What Went Wrong?

    Improved Upscaling: Understanding Permeability

    Boundary Conditions and Permeability Upscaling

    Transmissibility Yes, Permeability No

    Maintain the Well Injectivity & Productivity

    Magnus LKCF & Andrew Reservoir Case Studies

    Summary: What To Avoid & What Works Best?

    Future Trends:

    A Priori Error Analyses & Designer Grids

    Summary: Best Practice in Upscaling

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    50/95

    50 of 51

    Summary: Best Practice in Upscaling

    Check transport properties in initial geologic model

    By Facies: NTG, Porosity, Horizontal Permeability, Kv/Kh ratio

    When upscaling permeability

    Preserve reservoir quality

    Preserve reservoir barriers

    Preserve flow around reservoir barriers

    Streamline-based flow validation after upscaling

    Iteration: Is there a need to change resolution?

    Future trends: A Priori Error analysis & Designer Grids

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    51/95

    51 of 51

    Summary: A Personal Literature Review

    Individuals whose work and questions have shaped my understanding of

    permeability & upscaling

    Chris Farmer

    Kirk Hird

    Lars Holden

    Peter King

    Dave MacDonald

    Colin McGill

    John Barker

    Karam Burns

    Dominic Camilleri

    Tianhong Chen

    Mike Christie

    Lou Durlofsky

    Don Peaceman

    Jens Rolfsnes

    Kefei Wang

    Chris White

    John K Williams

    Mike Zerzan

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    52/95

    52 of 51

    Backup

    Upscaling within the Flow Simulator

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    53/95

    53 of 51

    p g

    Dynamic Boundary Conditions

    Utilize actual well positions, flow rates and an iterative global solution on the coarse simulation grid to provide

    local pressure boundary conditions for the upscaling calculation, including the transverse pressure drop

    100x100x50 => 20x20x10 upscaling for a variogram-based fine scale model

    Material provided by Lou Durlofsky (Stanford) & Yuguang Chen (Chevron)

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    54/95

    54 of 51

    Future Trends:

    Calculate your errors beforeupscaling

    Designer simulation grids that minimize these errors

    Best coarse layering

    Best unstructured 3D grids

    Upscaling in the Simulator (Static)

    Transmissibility is calculated from the fine model by upscaling

    Done at model initialization

    Upscaling in the Simulator (Dynamic) Utilize well locations and well rates on the coarse grid to define the fine scale boundary

    conditions

    Iterative calculation per time step

    A Priori Error:

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    55/95

    55 of 51

    Lack of Pressure Equilibrium

    Assumption Source of Error(Missing Physics)

    Pressure equilibrium withinthe coarse cell Disconnected pay within the coarse cellwill not be in equilibrium

    Fluid velocity is parallel tothe pressure drop

    Flow may depend upon the transversepressure drop on the coarse grid

    Single velocity within acoarse cell

    Distribution of multiphase frontalvelocities replaced by a single value

    Designer Grids within the Flow Simulator

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    56/95

    56 of 51

    Static Boundary Conditions

    Design simulation layering from 3D geologic model to minimize variation in local multiphase

    frontal velocities

    249, 80%336, 86%

    0

    10

    20

    30

    40

    50

    60

    70

    80

    90

    100

    0 200 400 600 800 1000 1200 1400 1600

    Model Layers

    %-H

    eterogeneity

    0

    2

    4

    6

    8

    10

    12

    14

    16

    18

    20

    RMSRe

    gression-Error

    Li & Beckner % Heterogeneity

    % Heterogeneity ; B-Variation

    % Heterogeneity: Uniform Coarsen

    Diagonal Guide

    Solution Total RMS Regression

    Solution Weighted RMS Regression

    Total RMS Regression

    Weighted RMS Regression

    Number of Coarse Layers

    %-Heterogeneity

    RMSRegressionErrorOptimal Layering

    Uniform Coarsening: Not Efficient

    Designer Grids within the Flow Simulator

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    57/95

    57 of 51

    Upscale During Initialization (Static)

    General trend shows that uniform coarsening does not perform well

    Optimal (293 layers) is the best layering scheme

    Flexible 3D grid (MCOARSE) provides even better results

    Tight Gas Layer Coarsening

    Fine Scale Model 22x23x1715 (Geological Scenario 5)

    0

    5

    10

    15

    20

    25

    30

    0 200 400 600 800 1,000 1,200 1,400 1,600 1,800

    Model Layers

    Cum.GasProd.

    (BCF)

    Regular-CoarsenNextVar-OneStepNextVar-SequentialOptimal-12LOptimalLi-Map-12LLi-Ave-MaxLLi-Ave-12LMCOARSE

    Fine Scale

    Li and Beckner

    UniformMCOARSE

    Optimal

    Future Trends:

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    58/95

    58 of 51

    Upscale in the Simulator (Static)

    3x3x3 coarsen used to reduce run-time

    Resolution re-introduced to preserveFault block boundariesResolution near wellsFluid contactsHeterogeneity via statistical measures

    More accurate flow simulation thanwith uniform coarsening

    Workflow Implications

    Single Shared Earth Modelused for both static and

    dynamic calculations

    Negligible time spent building coarse grid

    Extremely flexible grid design

    Simulation speed improvement

    comparable to model rebuild

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    59/95

    59 of 51

    Complex Flow in a Vertical Cross-Section

    In many ways, the unconfined boundary conditions are more typical of flow in the full three

    dimensional model. For example, look at the flow patterns calculated by

    as part of their work on Compositional Upscaling.

    The detailed velocity field shows significant local variation, and only rarely aligns with the coarse grid

    block boundaries.

    Transmissibility Upscaling

    http://../My%20Webs/3DKNOW/upscaling/spe37986/spe37986.htmlhttp://../My%20Webs/3DKNOW/upscaling/spe37986/spe37986.htmlhttp://../My%20Webs/3DKNOW/upscaling/spe37986/spe37986.htmlhttp://../My%20Webs/3DKNOW/upscaling/spe37986/spe37986.html
  • 7/31/2019 2006 11 08 SPE DL Upscaling

    60/95

    60 of 51

    KX

    Sealed Cell

    KXY

    Open Wide

    No Shift

    KX Streamline Flow Comparisons

    KX

    Open Wide

    Shifted

    KX

    Sealed

    Shifted

    KY Upscaling Comparisons

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    61/95

    61 of 51

    Streamline Flow Visualization

    KY

    Open

    Wide

    Shifted

    KY

    Sealed

    Shifted

    KY

    Sealed

    Cell

    KYX

    Open

    Wide No

    Shift

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    62/95

    Layer Coarsening:

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    63/95

    63 of 51

    Waterflood Example

    Waterflood Field Example:

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    64/95

    64 of 510%

    10%

    20%

    30%

    40%

    50%

    60%

    70%

    80%

    90%

    00%

    0 2000 4000 6000 8000 10000 12000

    FineScale

    Coarsen_54

    Coarsen_22

    Coarsen_22U

    Coarsen_31

    Coarsen_19

    Coarsen_07

    Oil Recovery and Watercut

    Optimal Simulation Model has 22 layers

    7 layers and 22 uniform layers are each too coarse

    0%

    5%

    10%

    15%

    20%

    25%

    30%

    0 2000 4000 6000 8000 10000 1200

    FineScale

    Coarsen_54

    Coarsen_22

    Coarsen_31

    Coarsen_19

    Coarsen_07

    0%

    5%

    10%

    15%

    20%

    25%

    30%

    0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00

    FineScale

    Coarsen_54

    Coarsen_22

    Coarsen_31

    Coarsen_19

    Coarsen_07

    1760

    1780

    1800

    1820

    1840

    1860

    1880

    1900

    1920

    1940

    1960

    1980

    0 2000 4000 6000 8000 10000 12000

    FineScale

    Coarsen_22

    Coarsen_22U

    Tight Gas Example: Cum. Recoveryl

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    65/95

    65 of 51

    Coarsening Results

    Tight Gas Layer CoarseningFine Scale Model 22x23x1715 (Geological Scenario 5)

    0

    5

    10

    15

    20

    25

    30

    0 200 400 600 800 1,000 1,200 1,400 1,600 1,800

    Model Layers

    Cum.GasProd.(BCF

    )

    Regular-CoarsenNextVar-OneStepNextVar-SequentialOptimal-12LOptimalLi-Map-12LLi-Ave-MaxL

    Li-Ave-12LMCOARSE

    Fine Scale

    Li and Beckner

    UniformMCOARSE

    Optimal

    k

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    66/95

    66 of 51

    Backup

    b l l

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    67/95

    67 of 51

    Permeability Upscaling

    Volume average of Darcys Law defines the effective permeability tensor for eachcoarse cell

    Flow calculation region can be >> than averaging region

    Results depend upon the choice of boundary conditions

    Coarse grid superimposed on fine grid and fine

    cell properties

    Darcys Law:

    Volume Averageof Darcys Law:

    pku

    pku *1

    1 2

    3 4

    Boundary Conditions andU l d P b l /

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    68/95

    68 of 51

    Upscaled Permeability - 2/3

    Vertical permeability, with a 4x4 coarse

    grid overlay

    Kz varies from 0 to 150 mD

    Open boundaries over-estimate flow

    capacity

    Calculations Courtesy of VoluMetrix FasTracker

    What Works Well?T b l U l

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    69/95

    69 of 51

    q

    1

    2

    Transmissibility Upscaling

    Upscale from coarse cell center to coarse cell center

    Replaces harmonic average of permeability with link permeability

    Captures fine scale juxtaposition of properties within the reservoir

    21pp

    q

    p

    qT

    fEffective

    q

    1

    2

    Transmissibility UpscalingP S i l R l i

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    70/95

    70 of 51

    Preserves Spatial Resolution

    Transmissibility can be calculated by direct upscaling instead of from the harmonic average

    of cell permeabilities

    Link Permeability is upscaled from cell center to cell center and has double the lateral

    resolution compared to cell permeability upscaling

    Harmonic average of a zero cell permeability is always zero

    1

    12121

    2

    ii

    iiii

    DXKXDXKX

    DXKXDXKXATX

    1

    21

    2121

    2

    ii

    i

    iiDXDX

    KXATX

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    71/95

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    72/95

    Transmissibility UpscalingC t fi l j t iti

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    73/95

    73 of 51

    Captures fine scale juxtaposition

    0 MD 0 MD 38 MD 0 MD 50 MD50 MD

    Permeability Upscaling does not preserve fine scaleti it

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    74/95

    74 of 51

    connectivity

    1x5 Upscaling Example

    Arithmetic average for horizontal permeability

    Harmonic average for vertical permeability

    Horizontal flow over-represented: too much sweep

    Arithmetic average of the transmissibility is preferred

    Vertical permeability reduced by lower perms

    Harmonic average preserves local barriers

    KZ KZ

    100 0.01

    100 0.011 1

    0.01 100

    0.01 1

    Harmonic Average

    0.024873 0.024751

    KX KX TX

    100 0.01 0.019998

    100 0.01 0.0199981 1 1

    0.01 100 0.019998

    0.01 1 0.019802

    Average: 0.215959

    Arithmetic Average

    40.204 20.404 27.06977

    Cell Permeability UpscalingM S btl E

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    75/95

    75 of 51

    More Subtle Errors

    Simulator well productivity calculated from

    sealed side coreflood permeability

    Does not describe radial flow and

    logarithmic pressure drop near a well

    Transmissibility could have been calculated by direct

    upscaling instead of from the harmonic average of cell

    permeabilities

    Link Permeability doubles the lateral resolution of

    the calculation

    Harmonic average of a zero cell permeability is always

    zero

    w

    ZZ

    rr

    HKYKXWI

    0ln

    2

    11

    2121

    2

    ii

    ii

    ii DXKXDXKX

    DXKXDXKXATX

    1

    21

    2121

    2

    ii

    i

    iiDXDX

    KXATX

    w

    YY

    rr

    HKZKXWI

    0ln

    2

    wX

    Xrr

    HKZKYWI0ln

    2

    What Works Well?W ll I d U li

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    76/95

    76 of 51

    Well Index Upscaling

    Three hypothetical directional wells (X, Y & Z) for each coarse cell

    Algebraic upscaling preserves reservoir quality & continuity of pay

    Use well index upscaling to define cell permeability in the simulator

    Ensures that fluids correctly enter and leave the reservoir

    ijk

    ijk

    ijk

    ijkijk

    Effective

    DZDYDXNTG

    DYDXDZNTGKYKX

    KYKX

    B k

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    77/95

    77 of 51

    Backup

    Upscaling Overview:In Review

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    78/95

    78 of 51

    In Review

    Understand, Validate and/or

    Challenge the Reservoir Model

    Gridding

    Grid Alignment

    Static Properties

    Upscaling: Quality Control

    Multiphase Flow & Pseudoization

    Iteration & Learning

    Future Trends:Upgridding and Upscaling

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    79/95

    79 of 51

    Upgridding and Upscaling

    Design of the simulation grid at run-time

    Fine scale model initialized in the simulator

    Resolution chosen as required by calculation

    Error estimates used to design grid

    Regular grid

    Layer grouping

    Unstructured grid

    Designed composite corner point grids in

    3D

    How to Combine Well Index & Transmissibility Upscaling

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    80/95

    80 of 51

    How to Combine Well Index & Transmissibility Upscaling

    Well Index upscaling defines cell permeability

    Algebraic average (close to arithmetic average)

    Adjust transmissibility at cell faces according to flow-based upscaling calculations

    Retain two flow calculations as sensitivities

    Pizza Box boundary conditions will preserve tortuosity

    Sealed side barriers will preserve local barriers

    1

    1212121

    2222

    ii

    iiiii

    DXKXDXKXDXKXDXKXATMXTX

    Face Property Cell Properties

    Backup

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    81/95

    81 of 51

    Backup

    Magnus LKCFWaterflood Development Study

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    82/95

    82 of 51

    Waterflood Development Study

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    83/95

    83 of 51

    LKCF Upscaling Streamline Validation

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    84/95

    84 of 51

    LKCF Upscaling Streamline Validation

    -10.0%

    -5.0%

    0.0%

    5.0%

    10.0%

    15.0%

    20.0%

    25.0%

    30.0%

    2x2x6 2x2x4 2x2x2 1x1x6 1x1x4 1x1x2

    Upscale ( NX * NY * NZ)

    Error on injection rate

    Error on connected volume

    CPU time scale =1 x1.8 x2.4 x3.7x6.8 x15.6

    42285 active cells 152734 active cells

    Backup

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    85/95

    85 of 51

    Backup

    A h A

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    86/95

    86 of 51

    Arithmetic Average

    Think of all of the reservoir re-stacked and placed immediately adjacent to a well.

    All the rock feels the same pressure gradient

    K1

    KN

    P

    L

    PAKQ jj

    H A

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    87/95

    87 of 51

    AQ

    KLP

    j j

    j

    Harmonic Average

    Think of all of the reservoir sliced and stacked into one amazingly long core.

    All the flow must run through each piece of rock.

    K1 KNP

    U li E i Fl Pi

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    88/95

    88 of 51

    Upscaling Exercise: Flow Pictures

    Geometric Average: Permeability follows a log normal distribution. In others words, the logarithm of

    permeability follows a Gaussian distribution, and the average of the data provides an unbiased

    estimate of the mean.

    Important Exceptions:

    What if we lose all of our unconsolidated core samples?

    What if we never make permeability measurements of our muds?

    Log PermPerm

    Fre

    que

    ncy

    Mode,

    Median& Mean

    A i h i H i

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    89/95

    89 of 51

    Arithmetic-Harmonic

    Harmonic followed by Arithmetic: Turn off all cross-flow between layers. Now you

    have the sum of many core floods!

    P

    H i A i h i

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    90/95

    90 of 51

    Harmonic-Arithmetic

    Arithmetic followed by Harmonic: Think of perfect vertical pressure equilibrium. This

    generates mixing at each column of the model, and a single average core flood

    P1 PN

    Coarsen in 3D:Preserve Pay/Non-Pay in Each Column

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    91/95

    91 of 51

    Preserve Pay/Non Pay in Each Column

    Tight Gas Field ExampleLayer Coarsening Analysis

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    92/95

    92 of 51

    Layer Coarsening Analysis

    1715 Geologic Layers Coarsened to 1 Simulation Layer

    249, 80%336, 86%

    0

    10

    20

    30

    40

    50

    60

    70

    80

    90

    100

    0 200 400 600 800 1000 1200 1400 1600

    Model Layers

    %-Heterogeneit

    y

    0

    2

    4

    6

    8

    10

    12

    14

    16

    18

    20

    RMSRegression-Error

    Li & Beckner % Heterogeneity

    % Heterogeneity ; B-Variation

    % Heterogeneity: Uniform Coarsen

    Diagonal Guide

    Solution Total RMS Regression

    Solution Weighted RMS Regression

    Total RMS Regression

    Weighted RMS Regression

    Effective Vertical PermeabilityImpact of Boundary Conditions

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    93/95

    93 of 51

    pac o ou da y Co d o s

    Upscale shales on a

    sand background

    Sealed sides capture

    local flow barriers

    Linear pressure allows

    barriers to leak

    Pizza box allows

    global flow tortuosity

    Summary:What to Avoid

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    94/95

    94 of 51

    Flow based upscaling for cell permeabilities

    Sealed side boundary conditions used for flow based upscaling of permeability

    Using the same upscaled flow based permeability to calculate both well indices and

    intercell transmissibility

    Linear pressure (open) boundary conditions used for flow based upscaling of

    permeability

    Unfortunately, these steps describe the most common upscaling approaches in the

    industry

    Summary:What Works Well

  • 7/31/2019 2006 11 08 SPE DL Upscaling

    95/95

    Use different upscaling techniques to extract different flow characteristics from the

    fine scale geologic model

    Well Index Upscaling preserves continuity of pay and provides a measure of the reservoir

    quality

    Transmissibility Upscaling provides higher spatial resolution

    Different boundary conditions preserve either flow tortuosity or flow barriers

    This combination of techniques has worked well within BP & elsewhere in the industry

    Streamline calculations provide detailed validation based on pressures, sweep, and

    time of flight

    Validation after upscaling is always necessary