Diese Präsentation wurde erfolgreich gemeldet.
Wir verwenden Ihre LinkedIn Profilangaben und Informationen zu Ihren Aktivitäten, um Anzeigen zu personalisieren und Ihnen relevantere Inhalte anzuzeigen. Sie können Ihre Anzeigeneinstellungen jederzeit ändern.
11/11/2015
1
david.lumley@uwa.edu.auLumley et al., 2014
Nonlinear Uncertainty Analysis:
4D Seismic reservoir monitoring
Pr...
11/11/2015
2
david.lumley@uwa.edu.auLumley et al., 2014
Outline
• Define “4D”
• Examples of 4D seismic
• Define “uncertain...
11/11/2015
3
david.lumley@uwa.edu.auLumley et al., 2014
Geophysics definition of “4D”
2D = f(x1,x2): eg. maps, cross-secti...
11/11/2015
4
david.lumley@uwa.edu.auLumley et al., 2014
Geophysics definition of “4D”
4D = f(x,y,z,t): hyper-cubes (x,y,z,...
11/11/2015
5
david.lumley@uwa.edu.auLumley et al., 2014
Rock properties can change over time
with fluids, stress, temperat...
11/11/2015
6
david.lumley@uwa.edu.auLumley et al., 2014
Seismic Image – zoom on reservoirs
OWC
Lumley
david.lumley@uwa.edu...
11/11/2015
7
david.lumley@uwa.edu.auLumley et al., 2014
TIME 2 amplitude map
extracted along top of reservoir structure
Lu...
11/11/2015
8
david.lumley@uwa.edu.auLumley et al., 2014
4D Monitoring of Steam Injectors
Sigit et al., 1999
steam costs > ...
11/11/2015
9
david.lumley@uwa.edu.auLumley et al., 2014
Injection Pressure Anomaly
Before After
Lumley et al., 2003Lumley ...
11/11/2015
10
david.lumley@uwa.edu.auLumley et al., 2014
Outline
• Define “4D”
• Examples of 4D seismic
• Define “uncertai...
11/11/2015
11
david.lumley@uwa.edu.auLumley et al., 2014
errors vs. uncertainty
various data results
A B
workflow
david.lu...
11/11/2015
12
david.lumley@uwa.edu.auLumley et al., 2014
errors vs. uncertainty
geophysics data image of the earth
A B
Ima...
11/11/2015
13
david.lumley@uwa.edu.auLumley et al., 2014
errors vs. uncertainty
geophysics data earth model
Inversion… F-1...
11/11/2015
14
david.lumley@uwa.edu.auLumley et al., 2014
errors vs. uncertainty
model simulated data
A + errors B + ???
Fo...
11/11/2015
15
david.lumley@uwa.edu.auLumley et al., 2014
Errors Uncertainty
Domain 1 Domain 2
A Definition of “uncertainty...
11/11/2015
16
david.lumley@uwa.edu.auLumley et al., 2014
Model space
+ errors
Data space
+ uncertainty
m +  d + 
F + 
F...
11/11/2015
17
david.lumley@uwa.edu.auLumley et al., 2014
* You may have the right model, but it may not fit the data!
* Yo...
11/11/2015
18
david.lumley@uwa.edu.auLumley et al., 2014
Example: low resolution data
Gravity data
david.lumley@uwa.edu.au...
11/11/2015
19
david.lumley@uwa.edu.auLumley et al., 2014
Model “null” space Data space
m + imi’ d + 
F(m’)≈ 0
* There ar...
11/11/2015
20
david.lumley@uwa.edu.auLumley et al., 2014
Inversion …imaging, estimation
Model space Data space
F-1
m d
dav...
11/11/2015
21
david.lumley@uwa.edu.auLumley et al., 2014
Inversion …imaging, estimation
Model space
+ uncertainty
+ non-un...
11/11/2015
22
david.lumley@uwa.edu.auLumley et al., 2014
“Mickey Mouse model”
Gravity data
david.lumley@uwa.edu.auLumley e...
11/11/2015
23
david.lumley@uwa.edu.auLumley et al., 2014
“Closing the loop… history matching”
Observed
Data; t++
Inversion...
11/11/2015
24
david.lumley@uwa.edu.auLumley et al., 2014
Sources of 4D error and uncertainty
• 4D Seismic data errors
• No...
11/11/2015
25
david.lumley@uwa.edu.auLumley et al., 2014
4D NR Noise
T1 T2 T2-T1
david.lumley@uwa.edu.auLumley et al., 201...
11/11/2015
26
david.lumley@uwa.edu.auLumley et al., 2014
after Landro
4D NRMS vs. position error
david.lumley@uwa.edu.auLu...
11/11/2015
27
david.lumley@uwa.edu.auLumley et al., 2014
Line A: before the platform Line B: after the platform
with Petro...
11/11/2015
28
david.lumley@uwa.edu.auLumley et al., 2014
Lumley et al., SEG, 1998
Baseline Monitor Difference
“Physics-bas...
11/11/2015
29
david.lumley@uwa.edu.auLumley et al., 2014
Local image difference Global image difference
CO2 injectors?
Lum...
11/11/2015
30
david.lumley@uwa.edu.auLumley et al., 2014
Reservoir container
david.lumley@uwa.edu.auLumley et al., 2014
Re...
11/11/2015
31
david.lumley@uwa.edu.auLumley et al., 2014
Up/down scaling
david.lumley@uwa.edu.auLumley et al., 2014
porosi...
11/11/2015
32
david.lumley@uwa.edu.auLumley et al., 2014
Rock physics param pdfs
david.lumley@uwa.edu.auLumley et al., 201...
11/11/2015
33
david.lumley@uwa.edu.auLumley et al., 2014
4D FD elastic modeling
david.lumley@uwa.edu.auLumley et al., 2014...
11/11/2015
34
david.lumley@uwa.edu.auLumley et al., 2014
real PSTM
1 layer CO2
StatoilLumley et al., 2008
synth PSDM
david...
11/11/2015
35
david.lumley@uwa.edu.auLumley et al., 2014
Multi-objective optimisation
find a reservoir model m such that:
...
11/11/2015
36
david.lumley@uwa.edu.auLumley et al., 2014
a) Reference Litho-facies model
b) Before MO model updating c) Af...
11/11/2015
37
david.lumley@uwa.edu.auLumley et al., 2014
How to quantify 4D errors + uncertainty?
>> Nonlinear stochastic ...
11/11/2015
38
david.lumley@uwa.edu.auLumley et al., 2014
4D seismic image of
water saturation (Sw)
a
b
david.lumley@uwa.e...
11/11/2015
39
david.lumley@uwa.edu.auLumley et al., 2014
{Sw} for 1000 realizations
a
b
david.lumley@uwa.edu.auLumley e...
11/11/2015
40
david.lumley@uwa.edu.auLumley et al., 2014
Probability that Sw > 0.5
a
b
david.lumley@uwa.edu.auLumley et a...
11/11/2015
41
david.lumley@uwa.edu.auLumley et al., 2014
Nächste SlideShare
Wird geladen in …5
×

David Lumley - 4D uncertainty - Nov 11, 2015

334 Aufrufe

Veröffentlicht am

Prof. David Lumley from the Centre of Energy Geoscience at the Uni. Of Western Australia presents his work on “Nonlinear Uncertainty Analysis: 4D Seismic reservoir monitoring”.

Veröffentlicht in: Wissenschaft
  • Als Erste(r) kommentieren

  • Gehören Sie zu den Ersten, denen das gefällt!

David Lumley - 4D uncertainty - Nov 11, 2015

  1. 1. 11/11/2015 1 david.lumley@uwa.edu.auLumley et al., 2014 Nonlinear Uncertainty Analysis: 4D Seismic reservoir monitoring Prof David Lumley + various colleagues and students over the years... UWA School of Physics; School of Earth & Environment david.lumley@uwa.edu.auLumley et al., 2014 Outline • Define “4D” • Examples of 4D seismic • Define “uncertainty” • Nonlinear uncertainty analysis + examples
  2. 2. 11/11/2015 2 david.lumley@uwa.edu.auLumley et al., 2014 Outline • Define “4D” • Examples of 4D seismic • Define “uncertainty” • Nonlinear uncertainty analysis + examples david.lumley@uwa.edu.auLumley et al., 2014 Geophysics definition of “4D” 1D = f(x1): eg. well logs f(z) f = porosity, clay content, age facies, velocity, density…
  3. 3. 11/11/2015 3 david.lumley@uwa.edu.auLumley et al., 2014 Geophysics definition of “4D” 2D = f(x1,x2): eg. maps, cross-sections (x,z) f = porosity, clay content, age structural depth, facies, reflectivity… Lumley et al. david.lumley@uwa.edu.auLumley et al., 2014 Geophysics definition of “4D” 3D = f(x,y,z): volumes (x,y,z) f = porosity, clay content, age velocity, reflectivity… Niri & Lumley, 2013
  4. 4. 11/11/2015 4 david.lumley@uwa.edu.auLumley et al., 2014 Geophysics definition of “4D” 4D = f(x,y,z,t): hyper-cubes (x,y,z,t) f = porosity, clay content, age velocity, reflectivity… Lumley, 1995 david.lumley@uwa.edu.auLumley et al., 2014 Outline • Define “4D” • Examples of 4D seismic • Define “uncertainty” • Nonlinear uncertainty analysis + examples
  5. 5. 11/11/2015 5 david.lumley@uwa.edu.auLumley et al., 2014 Rock properties can change over time with fluids, stress, temperature etc… Duffaut et al.2011 david.lumley@uwa.edu.auLumley et al., 2014 Seismic Image – 2D cross-section Lumley
  6. 6. 11/11/2015 6 david.lumley@uwa.edu.auLumley et al., 2014 Seismic Image – zoom on reservoirs OWC Lumley david.lumley@uwa.edu.auLumley et al., 2014 TIME 1 amplitude map extracted along top of reservoir structure Lumley et al., 2003
  7. 7. 11/11/2015 7 david.lumley@uwa.edu.auLumley et al., 2014 TIME 2 amplitude map extracted along top of reservoir structure Lumley et al., 2003 david.lumley@uwa.edu.auLumley et al., 2014 4D amplitude difference map extracted along top of reservoir structure Lumley et al., 2003
  8. 8. 11/11/2015 8 david.lumley@uwa.edu.auLumley et al., 2014 4D Monitoring of Steam Injectors Sigit et al., 1999 steam costs > $2 MM / day david.lumley@uwa.edu.auLumley et al., 2014 Before… After! courtesy of Statoil Monitoring Injection
  9. 9. 11/11/2015 9 david.lumley@uwa.edu.auLumley et al., 2014 Injection Pressure Anomaly Before After Lumley et al., 2003Lumley et al. david.lumley@uwa.edu.auLumley et al., 2014 Permanent Array 4D example Map of amplitude changes Map of compaction 2003 2004 2005 200820072006 LoFS Survey Timeline 1 2 3 4 5 6 7 8 9 1 0 Map of amplitude changes Map of compaction 2003 2004 2005 200820072006 LoFS Survey Timeline 1 2 3 4 5 6 7 8 9 1 0 Map of amplitude changes Map of compaction 2003 2004 2005 200820072006 LoFS Survey Timeline 1 2 3 4 5 6 7 8 9 1 0 Map of amplitude changes Map of compaction 2003 2004 2005 200820072006 LoFS Survey Timeline 1 2 3 4 5 6 7 8 9 1 0 Map of amplitude changes Map of compaction 2003 2004 2005 200820072006 LoFS Survey Timeline 1 2 3 4 5 6 7 8 9 1 0 Map of amplitude changes Map of compaction 2003 2004 2005 200820072006 LoFS Survey Timeline 1 2 3 4 5 6 7 8 9 1 0 Map of amplitude changes Map of compaction 2003 2004 2005 200820072006 LoFS Survey Timeline 1 2 3 4 5 6 7 8 9 1 0 Map of amplitude changes Map of compaction 2003 2004 2005 200820072006 LoFS Survey Timeline 1 2 3 4 5 6 7 8 9 1 0 Map of amplitude changes Map of compaction 2003 2004 2005 200820072006 LoFS Survey Timeline 1 2 3 4 5 6 7 8 9 10
  10. 10. 11/11/2015 10 david.lumley@uwa.edu.auLumley et al., 2014 Outline • Define “4D” • Examples of 4D seismic • Define “uncertainty” • Nonlinear uncertainty analysis + examples david.lumley@uwa.edu.auLumley et al., 2014 errors vs. uncertainty input output A B transform
  11. 11. 11/11/2015 11 david.lumley@uwa.edu.auLumley et al., 2014 errors vs. uncertainty various data results A B workflow david.lumley@uwa.edu.auLumley et al., 2014 errors vs. uncertainty earth model simulated data A B Forward modeling… F
  12. 12. 11/11/2015 12 david.lumley@uwa.edu.auLumley et al., 2014 errors vs. uncertainty geophysics data image of the earth A B Imaging… F* david.lumley@uwa.edu.auLumley et al., 2014 errors vs. uncertainty geophysics data earth model A B Inversion… F-1
  13. 13. 11/11/2015 13 david.lumley@uwa.edu.auLumley et al., 2014 errors vs. uncertainty geophysics data earth model Inversion… F-1 A + errors B + ??? david.lumley@uwa.edu.auLumley et al., 2014 errors vs. uncertainty geophysics data earth model Inversion… F-1 A + errors B + uncertainty!
  14. 14. 11/11/2015 14 david.lumley@uwa.edu.auLumley et al., 2014 errors vs. uncertainty model simulated data A + errors B + ??? Forward modeling… F david.lumley@uwa.edu.auLumley et al., 2014 errors vs. uncertainty model simulated data A + errors B + uncertainty! Forward modeling… F
  15. 15. 11/11/2015 15 david.lumley@uwa.edu.auLumley et al., 2014 Errors Uncertainty Domain 1 Domain 2 A Definition of “uncertainty” david.lumley@uwa.edu.auLumley et al., 2014 Forward Modeling Model space Data space m d F
  16. 16. 11/11/2015 16 david.lumley@uwa.edu.auLumley et al., 2014 Model space + errors Data space + uncertainty m +  d +  F +  Forward Modeling david.lumley@uwa.edu.auLumley et al., 2014 Model space + errors Data space + uncertainty m +  dmod +  F +  dobs +  Forward Modeling
  17. 17. 11/11/2015 17 david.lumley@uwa.edu.auLumley et al., 2014 * You may have the right model, but it may not fit the data! * You may have the wrong model, but it may fit the data! Model space + errors Data space + uncertainty m +  dmod +  F +  dobs +  david.lumley@uwa.edu.auLumley et al., 2014 Non-uniqueness, null space… Model “null” space Data space m + m’ d +  F(m’)≈ 0
  18. 18. 11/11/2015 18 david.lumley@uwa.edu.auLumley et al., 2014 Example: low resolution data Gravity data david.lumley@uwa.edu.auLumley et al., 2014 “Mickey Mouse model” Gravity data
  19. 19. 11/11/2015 19 david.lumley@uwa.edu.auLumley et al., 2014 Model “null” space Data space m + imi’ d +  F(m’)≈ 0 * There are infinitely many models that fit the data! david.lumley@uwa.edu.auLumley et al., 2014 Uncertainty ≠ Non-uniqueness
  20. 20. 11/11/2015 20 david.lumley@uwa.edu.auLumley et al., 2014 Inversion …imaging, estimation Model space Data space F-1 m d david.lumley@uwa.edu.auLumley et al., 2014 Inversion …imaging, estimation Model space + uncertainty Data space + errors F-1 +  m +  d + 
  21. 21. 11/11/2015 21 david.lumley@uwa.edu.auLumley et al., 2014 Inversion …imaging, estimation Model space + uncertainty + non-uniqueness Data space + errors F-1 +  m + + m’ d +  david.lumley@uwa.edu.auLumley et al., 2014 Inversion …imaging, estimation Model space + uncertainty + non-uniqueness * Regularization * * Model-shaping * Data space + errors F-1 +  m + + imi’ d + 
  22. 22. 11/11/2015 22 david.lumley@uwa.edu.auLumley et al., 2014 “Mickey Mouse model” Gravity data david.lumley@uwa.edu.auLumley et al., 2014 Outline • Define “4D” • Examples of 4D seismic • Define “uncertainty” • Nonlinear uncertainty analysis + examples
  23. 23. 11/11/2015 23 david.lumley@uwa.edu.auLumley et al., 2014 “Closing the loop… history matching” Observed Data; t++ Inversion Estimated model; t++ Simulation Predicted data david.lumley@uwa.edu.auLumley et al., 2014 4D amplitude difference map extracted along top of reservoir structure Lumley et al., 2003
  24. 24. 11/11/2015 24 david.lumley@uwa.edu.auLumley et al., 2014 Sources of 4D error and uncertainty • 4D Seismic data errors • Non-repeatable noise • Source-receiver positioning errors • Changes in the water column / near-surface / overburden • Changes in acquisition geometry • Changes in source-receiver characteristics • Non 4D-compliant processing flow • Etcetera… david.lumley@uwa.edu.auLumley et al., 2014 3D Noise data = “signal” + noise
  25. 25. 11/11/2015 25 david.lumley@uwa.edu.auLumley et al., 2014 4D NR Noise T1 T2 T2-T1 david.lumley@uwa.edu.auLumley et al., 2014 4D Repeatability Saul & Lumley, 2013
  26. 26. 11/11/2015 26 david.lumley@uwa.edu.auLumley et al., 2014 after Landro 4D NRMS vs. position error david.lumley@uwa.edu.auLumley et al., 2014 image image difference Image difference after 40-60 cm tidal corrections Eiken et al., EAGE 1999 4D tidal corrections
  27. 27. 11/11/2015 27 david.lumley@uwa.edu.auLumley et al., 2014 Line A: before the platform Line B: after the platform with Petrobras david.lumley@uwa.edu.auLumley et al., 2014 “Statistical” image processing Baseline Monitor Difference Lumley et al., SEG, 1998
  28. 28. 11/11/2015 28 david.lumley@uwa.edu.auLumley et al., 2014 Lumley et al., SEG, 1998 Baseline Monitor Difference “Physics-based” image processing david.lumley@uwa.edu.auLumley et al., 2014 1999 Local Diff Global Diff ? 4D Local vs. Global optimization Lumley et al. 2003
  29. 29. 11/11/2015 29 david.lumley@uwa.edu.auLumley et al., 2014 Local image difference Global image difference CO2 injectors? Lumley et al. 2003 4D Local vs. Global Optimization david.lumley@uwa.edu.auLumley et al., 2014 Sources of 4D error and uncertainty • Model definition • Model parameterization (acoustic, elastic, aniso, attenuation…) • Physical property relationships (velocity-pressure…) • Model discretization/sampling (fine, coarse, up/down-scale…) • Model relationships (geology, seismic, fluid flow…) • Etcetera…
  30. 30. 11/11/2015 30 david.lumley@uwa.edu.auLumley et al., 2014 Reservoir container david.lumley@uwa.edu.auLumley et al., 2014 Reservoir properties
  31. 31. 11/11/2015 31 david.lumley@uwa.edu.auLumley et al., 2014 Up/down scaling david.lumley@uwa.edu.auLumley et al., 2014 porosity clay fraction
  32. 32. 11/11/2015 32 david.lumley@uwa.edu.auLumley et al., 2014 Rock physics param pdfs david.lumley@uwa.edu.auLumley et al., 2014 Sources of 4D error and uncertainty • Physics (modeling/inversion operators/code) • Linear versus nonlinear • Acoustic, Elastic, Anisotropy, Attenuation… • Convolution, Raytracing, Finite Difference… • Algorithm implementations • Etcetera…
  33. 33. 11/11/2015 33 david.lumley@uwa.edu.auLumley et al., 2014 4D FD elastic modeling david.lumley@uwa.edu.auLumley et al., 2014 1 layer CO2 Lumley et al., 2008 synth PSDM
  34. 34. 11/11/2015 34 david.lumley@uwa.edu.auLumley et al., 2014 real PSTM 1 layer CO2 StatoilLumley et al., 2008 synth PSDM david.lumley@uwa.edu.auLumley et al., 2014 Sources of 4D error and uncertainty • Optimization criteria (inversion/estimation) • Least squares (L2), Least absolute (L1), hybrid… • Optimal fit to data, amplitude, phase… • Multiple objectives, weighting schemes… • Inversion constraints (soft, hard, weighted…) • Optimization method (gradients, stochastic, MC, PSO, genetic…) • Resolution, Null space, Non-uniqueness… • Etcetera…
  35. 35. 11/11/2015 35 david.lumley@uwa.edu.auLumley et al., 2014 Multi-objective optimisation find a reservoir model m such that: min E = (seismic)p + (logs)q + (geology)r + … david.lumley@uwa.edu.auLumley et al., 2014 13.22 13.27 13.32 13.37 13.42 13.47 13.52 13.57 36 41 46 ObjectiveFunction2 Objective Function 1 Generation 1 13.22 13.24 13.26 13.28 13.3 13.32 13.34 13.36 35 36 37 38 39 40 ObjectiveFunction2 Objective Function 1 Generation 10 13.12 13.14 13.16 13.18 13.2 13.22 13.24 13.26 30 32 34 36 38 ObjectiveFunction2 Objective Function 1 Generation 20 13 13.02 13.04 13.06 13.08 13.1 13.12 26 27 28 29 30 ObjectiveFunction2 Objective Function 1 Generation 50 12.96 12.97 12.98 12.99 13 13.01 13.02 13.03 13.04 13.05 13.06 24 25 26 27 28 29 ObjectiveFunction2 Objective Function 1 Generation 70 12.95 12.96 12.97 12.98 12.99 13 13.01 13.02 13.03 13.04 13.05 24 25 26 27 28 ObjectiveFunction2 Objective Function 1 Generation 100 Multi-objective optimization – “Pareto front” 70 Niri & Lumley 2013
  36. 36. 11/11/2015 36 david.lumley@uwa.edu.auLumley et al., 2014 a) Reference Litho-facies model b) Before MO model updating c) After MO model updating  Average Mismatch error reduced from 36.5% to 14.6% Multi-objective optimization – “Pareto front” Niri & Lumley 2013 david.lumley@uwa.edu.auLumley et al., 2014 How to quantify 4D errors + uncertainty?
  37. 37. 11/11/2015 37 david.lumley@uwa.edu.auLumley et al., 2014 How to quantify 4D errors + uncertainty? >> Nonlinear stochastic error propagation david.lumley@uwa.edu.auLumley et al., 2014 d1 d2 p N=1000 realizations Sw Pp 4D inversion statistics N=1000
  38. 38. 11/11/2015 38 david.lumley@uwa.edu.auLumley et al., 2014 4D seismic image of water saturation (Sw) a b david.lumley@uwa.edu.auLumley et al., 2014 4D seismic inversion for water saturation (Sw) Sw a b
  39. 39. 11/11/2015 39 david.lumley@uwa.edu.auLumley et al., 2014 {Sw} for 1000 realizations a b david.lumley@uwa.edu.auLumley et al., 2014 signal to noise ratio “S/N” of Sw a b
  40. 40. 11/11/2015 40 david.lumley@uwa.edu.auLumley et al., 2014 Probability that Sw > 0.5 a b david.lumley@uwa.edu.auLumley et al., 2014 “Closing the loop… history matching” Observed Data; t++ Inversion Estimated model; t++ Simulation Predicted data
  41. 41. 11/11/2015 41 david.lumley@uwa.edu.auLumley et al., 2014

×