youli quan & jerry m. harris stanford university

26
Stochastic Seismic Inversion using Waveform and Traveltime Data and Its Application to Time-lapse Monitoring Youli Quan & Jerry M. Harris Stanford University November 12, 2008

Upload: aliza

Post on 11-Feb-2016

70 views

Category:

Documents


0 download

DESCRIPTION

Stochastic Seismic Inversion using Waveform and Traveltime Data and Its Application to Time-lapse Monitoring. Youli Quan & Jerry M. Harris Stanford University. November 12, 2008. Outline . Introduction Motivations Kalman filter Seismic inversion with EnKF - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Youli Quan & Jerry M. Harris Stanford University

Stochastic Seismic Inversion using Waveform and Traveltime Data and

Its Application toTime-lapse Monitoring

Youli Quan & Jerry M. HarrisStanford University

November 12, 2008

Page 2: Youli Quan & Jerry M. Harris Stanford University

• Introduction • Motivations• Kalman filter

• Seismic inversion with EnKF

• An example of CO2 storage monitoring

• Conclusions

Outline

Page 3: Youli Quan & Jerry M. Harris Stanford University

• Seismic inversion recovers subsurface elastic properties (e.g., acoustic impedance and velocity) from seismic data.

• Inversion using both waveform and traveltime improves the estimation of velocity.

• Estimation of absolute velocity helps quantitative interpretation of seismic data.

Introduction

Page 4: Youli Quan & Jerry M. Harris Stanford University

Amplitude Traces Impedance Traces

-0.2 0 0.2

0

100

200

300

400

500

600

700

8000 0.5 1

0

100

200

300

400

500

600

700

800

SeismicInversion

• Images of seismic inversion may be more meaningful for interpretation.

Page 5: Youli Quan & Jerry M. Harris Stanford University

• Deterministic inversion methods normally need less computation.

• Stochastic inversion uses more computing power but can integrate other information (e.g., sonic logs.)

• Ensemble Kalman Filter (EnKF) is a stochastic method used in this study for seismic inversion.

Page 6: Youli Quan & Jerry M. Harris Stanford University

• Seismic monitoring

- To integrate time-lapse seismic data - Dynamic imaging

• Reservoir characterization

- Integration of sonic logs and seismic data

Motivations to Use EnKF

Page 7: Youli Quan & Jerry M. Harris Stanford University

• Dynamic imaging

• Integration of sonic logs and seismic data

Kalman Filter

)( )()1()()1( tttt Gmdkmm

)(ˆ )()()( sonicseismicsonic Gmdkmm

1)( RGCGCGk TTKalman gain

Page 8: Youli Quan & Jerry M. Harris Stanford University

Seismic Inversion with EnKF

],...,[ 1 NddD

],...,[ 1 NmmM

• Define observation function

• Create model & data ensembles using their probability distributions

)(md g

d – poststack seismic data in this case

Page 9: Youli Quan & Jerry M. Harris Stanford University

)(md g

m dPoststack

Full waveform Convolution

Define observation function

Page 10: Youli Quan & Jerry M. Harris Stanford University

Create model ensemble

ii εmm 0

0 1000 2000 3000 4000 5000 60000

0.2

0.4

0.6

0.8

1

1.2

1.4x 10

-3

Velocity (m/s)

f(m)

m0

],...,[ 1 NmmM

Page 11: Youli Quan & Jerry M. Harris Stanford University

Create data ensemble

ii γdd

],...,[ 1 NddD

-1000 -500 0 500 10000

1

2

3

4x 10

-3

Aplitudef(m

)

m0d

Page 12: Youli Quan & Jerry M. Harris Stanford University

• Estimate model parameters with EnKF

)]([ )()1()()1( tttt g MDKMM

K can be simply calculated from the ensemble covariance

It can handle large model and non-linear inverseThis is a Monte Carlo approach

)1/()]()][([ NEE TMMMMC

)1/()]()][([ NEE TDDDDR

)(Mg

and observation function

Page 13: Youli Quan & Jerry M. Harris Stanford University

• CO2 sequestration provides a possible solution for reducing the green gas emission to the atmosphere.

• For safety and operational reasons, we need to monitor the containment of the CO2 storage in the subsurface.

An Example of CO2 Storage Monitoring

CO2 Sequestration

Page 14: Youli Quan & Jerry M. Harris Stanford University

• Find model parameters from unmineable coalbeds in Powder River Basin

• Build a stationary geology model

• Run flow simulation with GEM

• Convert flow simulation results to time-lapse seismic velocity models

Creation of Time-lapse Models

Page 15: Youli Quan & Jerry M. Harris Stanford University

Distance(m)

Dep

th(m

)

0 200 400 600 800

200

400

600

800

50

100

150

200

Distance(m)D

epth

(m)

0 200 400 600 800

200

400

600

800

Distance(m)

Dep

th(m

)

0 200 400 600 800

200

400

600

800

Four time-lapse P-wave velocity modes created based on CO2 flow simulation in the coalbeds. A: time=0; B: time=3 months; C: time=1 year; D: time=3 years.

Distance(m)

Dep

th(m

)

0 200 400 600 800

200

400

600

800

3000

4000

5000

Distance(m)

Dep

th(m

)

0 200 400 600 800

200

400

600

800

Distance(m)

Dep

th(m

)0 200 4000 600 800

200

400

600

800

Distance(m)

Dep

th(m

)

0 200 400 600 800

200

400

600

800

Page 16: Youli Quan & Jerry M. Harris Stanford University

A Simple Synthetic Test

“Observed” data calculated by convolution

Distance(m)

Dep

th(m

)

0 500 1000

200

400

600

800

Distance(m)D

epth

(m)

0 500 1000

200

400

600

800

Distance(m)

Dep

th(m

)

0 500 1000

200

400

600

800

Distance(m)

Dep

th(m

)

0 500 1000

200

400

600

800

Page 17: Youli Quan & Jerry M. Harris Stanford University

Inversion with Waveform Data

Inversion with Waveform and Travel Time Data

Use constant initial model

Page 18: Youli Quan & Jerry M. Harris Stanford University

A Full Waveform Synthetic Test

• Run FD for time-lapse Vp models derived from flow simulation.

• Process complete shot gathers and get depth and time images.

• Extract wavelet.

• Use convolution as the modeling in the inversion.

• Perform seismic inversion with EnKF.

• Compare the inverted Vp with given models.

Page 19: Youli Quan & Jerry M. Harris Stanford University

Tim

e (s

ec)

0

0.1

0.2

0.3

0.4

0.5

0.6

Samples of the shot gathers calculated using the finite difference

Page 20: Youli Quan & Jerry M. Harris Stanford University

Distance(m)

Dep

th(m

)

220 380 580 780

100

200

300

400

500

600

700

800

Tim

e (s

ec)

Distance (m)100 300 500 700 900

0

0.1

0.2

0.3

0.4

0.5

0.6

Depth image Time image

Reflector 1 2 3 4 5

Depth (m) 270 310 550 670 750

Time (sec) 0.1675 0.1918 0.3340 0.4173 0.4595

Traveltime picks used for the inversion

Page 21: Youli Quan & Jerry M. Harris Stanford University

Distance (m)200 400 600 800

2500

3000

3500

4000

4500

5000

v (m/s)Distance (m)

Dep

th (m

)

200 400 600 800

200

400

600

800

Dep

th (m

)

200

400

600

800

Time-lapse velocity models inverted using EnKF

time=0 time=3 months time= 1 year time=3 years

Page 22: Youli Quan & Jerry M. Harris Stanford University

Vp differences between time-lapse models and base modelD

epth

(m)

200

400

600

800D

epth

(m)

200

400

600

800

Distance(m)

Dep

th(m

)

200 300 400 500 600 700 800

200

400

600

800

Distance(m)200 300 400 500 600 700 800

-200

-100

0

100

dv(m/s)

time=3 months time= 1 year time=3 years

True

Inverted

Page 23: Youli Quan & Jerry M. Harris Stanford University

2000 3000 4000 5000

100

200

300

400

500

600

700

800

Velocity (m/s)

Dep

th (m

)

TrueInit.Inv.

A comparison between true model and inverted model

Solid black line: Ture model; Dash-dot blue line: inverted model;Dotted yellow line: Initial model; At distance x=500m

Page 24: Youli Quan & Jerry M. Harris Stanford University

-0.5 0 0.5

100

200

300

400

500

600

700

800

Amplitude

Dep

th (m

)

GivenConv.

A comparison between “observed” data and modeled data

Solid line: “Observed” seismic trace Dotted line: Modeled seismic trace from inverted model

Page 25: Youli Quan & Jerry M. Harris Stanford University

• The ensemble Kalman filter is a useful tool for stochastic seismic inversion, especially for dynamic inversion in seismic monitoring (field data tests will be done.)

• Integrating travetime data into the inversion makes the estimation of absolute velocity possible.

• Fast forward modeling and true amplitude processing are essential.

Conclusions

Page 26: Youli Quan & Jerry M. Harris Stanford University

• We would like to thank the sponsors (ExxonMobil, General Electric, Schlumberger, and Toyota) of Global Climate & Energy Project at Stanford University for their support to this study.

• Eduardo Santos, Yemi Arogunmati, and Tope Akinbehinje helped for the creation of time-lapse velocity models.

Acknowledgements