CHOA Upcoming Special Events

1
CHOA Upcoming Special Events

• Technical Luncheon
• Wednesday, November 15, 2006
• Topic AEUB Update on 2 new Directives
• Performance Presentations, Audit, and
  Surveillance of In Situ Schemes
• Compliance Assurance - Enforcement
• Speakers Andrew MacPherson Cam Taylor, AEUB
• registration
  forthcoming ..

www.choa.ab.ca
Canadian Heavy Oil Association
2
CHOA Upcoming Special Events
Annual Fall Conference Infinite Possibilities
Thursday, November 9, 2006 Telus Convention
Centre (registration details are
available on web site)

www.choa.ab.ca
Canadian Heavy Oil Association
3
CHOA Corporate Sponsors
4
Strategies for Effective Geomodeling for Thermal
Recovery in Oilsands
David Garner Chevron Canada Resources
5
Strategies for Effective Geomodeling for Thermal
Recovery in Oilsands
David Garner Chevron Canada Resources
6
Contents

• Introduction
• The McMurray Geomodel Workflow
• The 1D Facies Model Core, Logs, Electrofacies
• Facies Proportions and Trends
• Stratigraphy Structure
• Variograms and Simulation
• Gridding, Data, and Scaling
• Conclusions and Recommendations
• Discussion

7
Goals and Objectives of Geomodeling

• The classical statement from the mid 1990s
• Render the geologic interpretation and conceptual
  model into a digital format suitable for input to
  reservoir simulation software and for
  post-processing.
• Quantitatively simulate reservoir properties.
• Provide an image of the reservoir that has the
  same variability as the natural phenomenon.
• Today Lets Add
• As is fit for understanding the physical process

8
RESERVOIRS ARE HETEROGENEOUS!!

• As new data are introduced, conceptual models
  take on more complexity.

9
Concept on Stratigraphic Framework
http//www.searchanddiscovery.net/documents/2005/g
arner/index.htm

• Sequence Layering

Channels
10
Contents

• Introduction
• The McMurray Geomodel Workflow
• The 1D Facies Model Core, Logs, Electrofacies
• Facies Proportions and Trends
• Stratigraphy Structure
• Variograms and Simulation
• Gridding, Data, and Scaling
• Conclusions and Recommendations
• Discussion

11
Example Geomodeling Workflow
Athabasca Oilsands developments are Engineering
Driven with complexities from thermal physics and
heterogeneous geology.
Post-Processing Analysis Performance Prediction
Risk
Pre-processing EDA GG Data Interpretation
Geomodel
Structure Modeling
3D Facies Simulations in Stratigraphy
Re-scaling to Pad/pair Simulation Grids
Attributes
Property Modeling
Facies Assignments in Wells
Thermal Simulation
Petrophysics Core
Hrzntl Well Placement
Checking Results Feedback loops
12
Model Parameters (Commercial Area Models)

• Grid
• Lithostrat-units
• Stratigraphic Grid Reference
• Facies
• Porosity
• SW
• Permeability (SPE 103083)

• 25 x 25 x 0.5 m
• 2 4 Layers
• Parallel to Top (typical)
• 5 (SS, Breccia, 2xIHS, Mud)
• Pixel simulation w/3D Trends
• SGS in Stratigraphy
• By facies w/3D Trend (LVM)
• SGS from OWC reference
• By facies w/3D Trend (LVM)
• Horizontal Perm
• P-Field Cloud by facies, OR
• co-SGS with Porosity by facies
• Regression Poro-Perm by facies
• Vertical Perm
• SGS Kv/Kh

Data availability dependent
13
Model Parameters (Local Pad/Pair Models)

• Grid
• Lithostrat-units
• Stratigraphic Grid Reference
• Facies
• Porosity
• SW
• Perm

• 1 x 5 x 0.5 m (or 1x10x0.5)
• 1 (or match Area Layers)
• Parallel to Top
• Repeat steps from area model using the area model
  as the conditioning data, i.e. pseudo wells.
• If necessary simplify geostatistical steps to
  global means, by facies

Upscale for Simulator
14
Geomodeling Workflows for SAGD

• Even if you're on the right track, you'll get
  run over if you just sit there.
• Will Rogers

15
Contents

• Introduction
• The McMurray Geomodel Workflow
• The 1D Facies Model Core, Logs, Electrofacies
• Facies Proportions and Trends
• Stratigraphy Structure
• Variograms and Simulation
• Gridding, Data, and Scaling
• Conclusions and Recommendations
• Discussion

16
The 1D Facies Model

• Combining logs and core description to train the
  logs
• Provides Statistics on Reservoir Heterogeneities
• eFacies Conditioning data for the Geological
  Model
• Need Rocktypes to unlock Thermal Simulator

17
Electro-facies Points

• Porosity is a better indicator of facies than
  vshale.
• When using Core Description, assign
  Electro-facies to wells using a non-parametric
  discriminant analysis.
• If there is a vertical trend, separate into
  layers for assignments.
• Try to honour the proportions of the input
  training samples during classification.
• Fluid effect on logs may cause erroneous
  classifications.
• Will show which logs to use for classification.

18
Databases and Choices

• Should we
• Assume the Core Facies description is Correct?
• Assume the Core depths are adjusted Correctly?
• When we have imprecision in data, probabilistic
  methods help
• Partial Core and logs Use Discriminant
  analysis
• Full Core Coverage and Logs Use Discriminant
  analysis, at least to check the description
  consistency and depth errors.
• What about core photo scanning? Vshale driven.

19
Discriminant Analysis

• Find an orientation along which the two classes
  have the greatest separation and the least
  inflation.

Variable 2 Gamma Ray
Discriminant function
Variable 1 Porosity
After Davis, 1986

• Assign unknown samples

20
Non-Supervised Discriminant Analysis

• Mode mapping or bump hunting

• Find a high density of points grouped together.
• These might be interpreted as petrofacies.
• Seed the points.
• Shown in color in the crossplots.
• These samples can be used as a training set for
  discriminant analysis.

Neut
Neut
Rhob
Rhob
GR
21
Non-Supervised Method

• Samples Assignment

• Assigned remaining samples according to a
  function of 3 variables

22
Non-Supervised Method

• In this example
• Electrofacies have a better visual relationship
  to the effective porosity curve than to the
  Vshale curve.
• Porosity was computed from density and GR.
• A fast electrofacies model can be a practical
  way to proceed.

23
Logs to use for Classification

• Neutron, Density, GR, Phie, Sonic
• GR is better than Vshale
• Phie is the most statistically descriptive
• Typically Phie f(Density, Vshale)
• If no Phie, GR is most descriptive.
• Sonic helps with degraded reservoir description,
  i.e. IHS, Breccias
• No Resistivity fluid effects are serious.

24
Electro-facies Conclusions

• Porosity is a better indicator of facies than
  vshale.
• When using Core Description, assign
  Electro-facies to wells using a non-parametric
  discriminant analysis.
• If there is a vertical trend, separate into
  layers for assignments.
• Try to honour the proportions of the input
  training samples during classification.
• Fluid effect on logs may cause erroneous
  classifications.

25
Contents

• Introduction
• The McMurray Geomodel Workflow
• The 1D Facies Model Core, Logs, Electrofacies
• Facies Proportions and Trends
• Stratigraphy Structure
• Variograms and Simulation
• Gridding, Data, and Scaling
• Conclusions and Recommendations
• Discussion

26
Well Discretization
What Vertical Cell Size? Preserve Proportions and
Heterogeneities
Upscaling(per facies)
27
Vertical Proportion Curve (VPC)
Influence of Sequence Stratigraphy
W5
W1
VPC
W4
W2
W3
Facies proportion ()
28
Proportion Curves

• Graphic representation of stratigraphic layering
• Fixes probability of lithofacies along layers
• Defines local facies proportions in Facies
  simulations (SIS, TG, PG, MPS, Obj)
• Defines cut-offs in truncated (Pluri-)Gaussian
  simulation
• Controls volume of each lithofacies in the model
  layers

29
Vertical Trends
30
Non-stationary Facies Modeling

• Trend Modeling Does net-to-gross change
  spatially?
• Stationary Geostatistics
• mean proportion curve represents all locations
• Facies proportions are uniform across study area
• Non-Stationary Geostatistics
• facies proportions change across study area
• proportion curves from subsets of wells are
  distinct
• net-to-gross across the study area varies

31
Trends in Facies Proportions (local)
N
32
Making Trends (by Facies)

• Combine 2D XY and 1D vertical

3D
Z

Y
0
1
Prop
X
Combine different sources of 3D Trends
Wells
Seismic
Updated

33
Contents

• Introduction
• The McMurray Geomodel Workflow
• The 1D Facies Model Core, Logs, Electrofacies
• Facies Proportions and Trends
• Stratigraphy Structure
• Variograms and Simulation
• Gridding, Data, and Scaling
• Conclusions and Recommendations
• Discussion

34
Influence of Sequence Stratigraphy
Top Reference Onlapping Sequence
Proportional Unit
Bottom Reference Erosion on top
35
Structure and Grids

• Defines the correlation lines.
• Geostatistical simulations are run in the
  stratigraphic grid

Stratigraphic
Structural
36
Contents

• Introduction
• The McMurray Geomodel Workflow
• The 1D Facies Model Core, Logs, Electrofacies
• Facies Proportions and Trends
• Stratigraphy Structure
• Variograms and Simulation
• Gridding, Data, and Scaling
• Conclusions and Recommendations
• Discussion

37
Variograms

• Defines orientation and relative geometry of
  facies heterogeneities
• Related to continuity of facies, with exceptions
• Vertical variogram
• McMurray facies and petrophysics often need
  nested variogram models

Figure from McLennan and Deutsch, 2005, Guide to
SAGD (Steam Assisted Gravity Drainage) Reservoir
Characterization Using GeostatisticsCentre for
Computational Geostatistics (CCG) Guidebook
Series Vol. 3
38

• Good judgment comes from experience,
• and a lot of that comes from bad judgment.
• Will Rogers

39
Simulation Methods Heterogeneity and Data Density

• Pixel Methods
• Object-based
• Process-based
• MPS

40
Contents

• Introduction
• The McMurray Geomodel Workflow
• The 1D Facies Model Core, Logs, Electrofacies
• Facies Proportions and Trends
• Stratigraphy Structure
• Variograms and Simulation
• Gridding, Data, and Scaling
• Conclusions and Recommendations
• Discussion

41
Rescaling

• (Geocellular Model-gt Reservoir Model)

42
Upscaling Geometry

• Coarse grid cells intersect the fine grid
• Complex structures may yield incorrect local tops
  positions

Lost Volume
Cell included in a lower layer
Mixed Stratigraphy
Cells of the geocellular grid may or may not have
an equivalent cell location in the reservoir grid
What if there are FAULTS?
43
Topological Upscaling

• Honouring stratigraphy and properties

Each cell of the geocellular grid is given a
topologically equivalent cell location in the
reservoir grid
Avoids mixing of layer properties!
44
Downscaling

• Re-simulate stochastically

Reservoir Grid
Downscaling
Geological Grid
45
Discretized Well Bore for Start-up and Grids
Sugar Cube Grids cut across stratigraphy and
average properties inappropriately. Try to use
small cell sizes along the well bores.
Try to preserve heterogeneity.
46
Permeability

• Three main permeability modeling challenges
• Limited core permeability measurements
• Core porosity and permeability measurement bias
• Core plug, geological, and flow simulation scale
  differences

A good rock types model in our Pad well pair
models allows us to implement geomechanical
effects in the simulator.
47
Permeability Upscaling

• Permeability is a non-additive parameter
• Power law averages and CP method are often used
  as a starting estimates of the upscaled values
• SPE 103083, McLennan et al. 2006
• Core measurements are biased and may not
  represent the true distribution at the small
  scale.
• One workflow is to model Permeability through
  statistically related variables at a small scale.
  
• The method is probabilistic using geostatistical
  rules.

48
Contents

• Introduction
• The McMurray Geomodel Workflow
• The 1D Facies Model Core, Logs, Electrofacies
• Facies Proportions and Trends
• Stratigraphy Structure
• Variograms and Simulation
• Gridding, Data, and Scaling
• Conclusions and Recommendations
• Discussion

49
Conclusions and Recommendations (1)

• Quantitatively simulate reservoir properties as
  is fit for understanding the physical SAGD
  process.
• Our understanding of reservoir heterogeneities
  changes over the life of the reservoir
  development.
• Use the best facies, efacies, lithotypes, and
  rocktypes model possible to feed the thermal
  simulator.

50
Conclusions and Recommendations (2)

• Honour facies proportions at all stages.
• Use trend models if data supports them.
• Scale models and all the parameters for
  simulation.
• Grid size matters to the simulator. Keep it
  small.
• A beer to quench the chat.