Vous êtes sur la page 1sur 58

HAGI

Bandung

23 to 24 June 2014

Velocity Model Building

Presenter : Jacques BONNAFE


1

Outline
Introduction Principles
Types of velocity models
Velocity model building methodology and tools

Anisotropy
Example
Conclusions

Depth Imaging

Preprocessing required (decon, demult, )

Acquisition geometry is determinant (mon/multi/wide


azimuth, maximum offset)
Prestack surface
gathers

Depth Migration
Algorithm

PSDM image

Velocity Model
Major concern of a PSDM project

Integrated Depth Imaging Toolkit

Interpretation

Velocity Model
Building

High Performance
Computing
4

Depth
Migration
Algorithm

Residual Moveout
Computation

Outline
Introduction Principles
Types of velocity models

Different types of models


How to make the initial model

Velocity model building methodology and tools


Anisotropy
Example
Conclusions

LAYERED MODEL

GRID MODEL
Smooth Models
With constraints

Hybrid parameterization
No velocity contrasts between layers
V0(x,y) and k(x,y) within each layer
Horizon not necessary geological

Outline
Introduction Principles
Types of velocity models
Velocity model building methodology and tools
reflection tomography, principle, gamma, updates
Other techniques: refraction tomography, scan, ....
Well ties

Anisotropy

Example
Conclusions

Principle of tomographic updating


Tomo Solver
for Smooth
Update To
Interval Vels

Velocity
Model

Iterate

PreStk
Depth Mig.

Are
CIPS
Flat?
YES
PreStk.
Depth Mig.
Image

10

NO

Auto Pick
Residual
Moveout
& Dip in z

3D Ray-trace
Linear
Tomography
Equations

RMO definition
x,y

depth

offset

CRP Gather

11

Too fast
Too slow

depth

OK
RMO section

Gamma definition
The ratio of the migration velocity to the
true (geological) velocity

12

Comparison of RMO before and after iteration on maps


Iteration #3 VTI anisotropic

0-1 km

Iteration #2 isotropic

+4

-4

+4

-4

+4

-4

+4

1-2 km

-4

13

Comparison
of RMO
before
RMO Statistic
Initial
VTIand after iteration on histograms
Iteration 3

Iteration 4
Interval 0 1
Km

Interval 1 2
Km

Interval 2 3
Km

Interval 3 4
Km

Interval 4 5
Km

Interval 5 6
Km

Interval 6 7
Km

Interval 7 8
Km

PSDM Stack Overlaid with Gamma: inline

PSDM Stack overlaid with Gamma : depth = 500m

GAMMA ON HORIZON

17

Example of cdp gathers evolution


Iteration 1

18

Iteration 2

Iteration 3

Iteration 4

EXAMPLE OF STACK AND VELOCITY MODEL


Iteration 1

19

Iteration 5

EXAMPLE OF STACK AND VELOCITY MODEL

Iteration 2

Iteration 3

2 km

20

2 km

One iteration illustrated:


Initial RMO residuals

21

One iteration illustrated:


Inversion + New PSDM + RMO residuals

22

One iteration illustrated:


Initial PSDM

23

One iteration illustrated:


New PSDM

24

Other tools
Refraction tomography
Velocity scan
Information from wells of geologists
FWI

25

Refraction tomography
FB in Shot Domain
First Break picking:
Automated picks in good seismic area
First Break picks

Manual picks in degraded seismic area


Full offset used

Application:
For shallow anomalies
Down to depth approx 25% of maximum offset

FB in various Offset Planes

FB in Offset Domain
Manual picking done

Shallow
Gas
200 262.5 m

1600 1762.5 m
Late arrival

2000 2162.5 m

2400 2562.5 m

Low

High

First Break Time


26

Low

First Break Time

High

Iterative refraction tomography updates


Smoothed PSTM velocity model as initial model
Several iterations of calculations run with
decreasing grid size
QCs performed:
Comparison with previous model with greater grid size
Visual observation by overlying the model with depthstretched PSTM section
Target line migration using final shallow model

500 m

400 x 400 m
400 m

200 x 200 m

Initial velocity model

400 m

Forward modeler module for refracted wave


First Break picks
4 km

400 m

4 km

Refraction tomography equation builder and


solver

4 km
100 x 100 m

27

Main elements of refraction tomography:

50 x 50 m
400 m

4 km

Final Shallow Model


Update is effective < 400m (~1/6 cable length)

model for scanning

Carbonate
flooding

-28-

Carbonate V=2750 m/s

-29-

Carbonate V=2900 m/s

-30-

Carbonate V=3100 m/s

-31-

Map of reefs

-32-

VELOCITIES CONTROL AT WELL


Initial velocity
Iteration 1 velocity
Iteration 2 velocity
Iteration 3 velocity
Iteration 4 final velocity

TVD TWT

Horizon A

Horizon B

1400

33

5000 m/s

Basic Depth Imaging Workflow

Velocity model may be iteratively built using a layer stripping approach


Simplifies velocity update process
May rather build downwards

PSDM migration performed at each iteration


Moveout extracted from the Gathers for velocity update

Workflow and tools adapted to study


Objectives : Exploration/Reservoir imaging
Geological environment
Data : Marine/land acquisition, mono/multi/wide azimuth

34

Importance of seismic interpretation in the VMB


Interpretations have to be very consistent with the seismic information
Used for ray-tracing
Used for velocity contrasts positioning, e.g. salt bodies, carbonate build-ups

The accuracy of the inversion result depends on these interpretations


The RMO extracted (along interpretations) have to represent reliable
seismic information:
If multiples are picked by RMO, velocity model will be updated wrongly
Use interpretation to discriminate
In area of poor signal to noise ratio, RMO analysis can be erroneous

If maximum offset is not long enough, RMO can be inaccurate


In very complex areas or when initial model is very wrong, the moveout can be far from
hyperbolic; sometimes gathers lack coherence
Direct interpretive input into the model

35

Outline
Introduction Principles
Types of velocity models
Velocity model building methodology and tools

Anisotropy
Example
Conclusions

36

Anisotropy

37

Velocity Issue: isotropic, anisotropic STI or VTI?


Fast
Fast

Salt

Salt

Salt

Slow

Slow

True Model (STI)

Isotropic Migration

Velocity is increasing with angle


to structural dip normal

Velocity does not vary


with angle of propagation

39

VTI Migration
Velocity is increasing
with angle of propagation

Thomsens VTI parameters estimation


Delta estimation (relates vertical and imaging vels.)

Epsilon estimation (relates vertical and horizontal vels.)

Comparison of depth markers on seismic and on wells

Measurements based on hockey sticks on gathers

Requires accurate calibrations on wells

Then Epsilon averaged to obtain single epsilon functionvalue per


layer
Sometimes set to a constant percentage of delta (150% or 200%)

Iter.2 isotropic

Iter.2: delta applied

Iter.3 VTI anisotropic

Averaged Epsilon function (z)


Averaged Delta function d (z)

d,

z
40

Anisotropic model Building


In the Isotropic case:
Vnmo = Vvertical by layer
Vnmo = Vrms multi-layer
Isotropic model/migration, anisotropic earth:
Vertical velocity over estimated
Mismatch between horizon and well marker (typically, horizons deeper)
Mismatch between sonic and seismic velocities
In the anisotropic case : Vnmo Vvertical or Vrms
Match horizon and well marker
Match between sonic and seismic velocities
Calibrated PSDM image

Very delicate and time consuming work

41

3D Depth Migration

Vertical stretch is not sufficient !!

ISOTROPIC (stretched to well)

ANISOTROPIC

Outline
Introduction Principles
Types of velocity models
Velocity model building methodology and tools
Anisotropy
Example: BEKAPAI VMB

Step 1. First Break picking


Step 2. Refraction tomography
Step 3. Initial PSDM velocity building
Step 4. Reflection tomography
Results

Conclusions
43

Step1. First Break picking and QC


FB in Shot Domain
First Break picking:
Automated picks in good seismic area
First Break picks

Manual picks in degraded seismic area


Full offset used

QC performed:
Visual observation of picked events in shot domain,
offset domain and various offset planes
Edit manually anomalous picks

FB in various Offset Planes

FB in Offset Domain
Manual picking done

Shallow
Gas
200 262.5 m

1600 1762.5 m
Late arrival

2000 2162.5 m

2400 2562.5 m

Low

Low

High

First Break Time


44 - IPA11-G-121 Combined Refraction Tomography and Reflection Tomography for PSDM Velocity Model Building

First Break Time

High

Step 2. Iterative refraction tomography updates


Smoothed PSTM velocity model as initial model
Four iterations run with decreasing grid size
(400m, 200m, 100m, 50m)
QCs performed:
Comparison with previous model with greater grid size
Visual observation by overlying the model with depthstretched PSTM section
Target line migration using final shallow model

500 m

400 x 400 m
400 m

200 x 200 m

Main elements of refraction tomography:


Initial velocity model

400 m

Forward modeler module for refracted wave


First Break picks
4 km
100 x 100 m
400 m

4 km

Refraction tomography equation builder and


solver

4 km
50 x 50 m
400 m

4 km

45 - IPA11-G-121 Combined Refraction Tomography and Reflection Tomography for PSDM Velocity Model Building

Final Shallow Model


Update is effective < 400m (~1/6 cable length)

Step 3. Initial PSDM velocity building

500 m

Final Shallow Model

Initial PSDM velocity model

2 km

Merge at
Z = 400 m

3 km

1500 m

1500 m

Smoothed PSTM velocity*

Model is ready to be updated through


iterative reflection tomography
3 km
* PSTM velocity field was first converted to interval velocity and
then smoothed spatially

46

Step 4. Iterative reflection tomography updates


(Woodward, et. al., 2008, A decade of tomography)

Four iterations carried-out:


Two isotropic model followed by two VTI
anisotropic updates
Kirchhoff PSDM to depth-migrate the
gathers
On 50 x 50m grid size
Best model was selected based on
comparison with previous model using:
Gamma map & histogram
Gamma/stack section overlay

Next slide

Main elements of reflection tomography:

Gathers display
Velocity model/stack overlay
Gamma = ratio of the migration
velocity to the true geological velocity

Velocity needs to
be slower
47

Velocity is
correct

Velocity needs to
be faster

Initial velocity model


Pre-Stack Depth Migration
Automated RMO and Dip picking
Reflection tomography equation builder and solver

Gamma section: initial model

RMO picking and QC

Gamma section: 1st update

Automated RMO pick (50 x 50m):

QC done through:
Visual observation of picked events on gathers

1000 m

Precondition gathers with de-multiple and offset


muting prior to picking

Visual observation of RMO attribute in section


and map views
Histogram plot of RMO attribute

Significant improvement of
Gamma after 1st iteration
6 km

Gamma QC: iteration 1


Gamma = 0.9 (10% too slow)

Vel. error map (1-2km): iter.2

6 km
Gamma = 1.1 (10% too fast)

Vel. Error map (1-2km): iter.3

500 m

+10%

-10%

Significant improvement of
Gamma after 3rd iteration (VTI)

0.8

Too slow

1.0
0

4 km

1.2
48 - IPA11-G-121 Combined Refraction Tomography and Reflection Tomography for PSDM Velocity Model Building

-4

+4

Too fast
-4

+4

Step 5. Introducing anisotropy


Thomsens VTI parameters estimation
Delta estimation (relates vertical and imaging vels.)

Epsilon estimation (relates vertical and horizontal vels.)

Computed based on time-picked horizons at nine wells

Automated based on far offsets at gathers located in nine wells

Workflow consists of:

Individual Epsilon function at well was averaged to obtain single


epsilon function

1) Seismic to well calibration to obtain calibrated vertical


velocity field
2) Extraction of horizons TWT and the corresponding depth
TVDSS (Zo)

QC done by comparing gathers before and after application of


Delta and Epsilon

3) Convert the horizons TWT to depth (Ziso) using isotropic


velocity field

Iter.2 isotropic

Iter.2: delta applied

Iter.3 VTI anisotropic

4) For each horizon, compute the delta


Single delta function (depth-variant) was used in 3rd VTI iteration
Delta set inactive (0.0001) at last iteration; considering quite
important depth errors in deep interval after 3rd VTI iteration

Averaged Epsilon function (z)

z
49

VTI anisotropy model: improvement of depthing & focusing


1st iteration model

2nd iteration model

3rd iteration VTI model

Improvement of gather flatness


= more focused stack

600 m

Initial model

Multiple

Isotropic

50

Isotropic

Anisotropic

Final PSDM velocity model

Velocities in time domain (TWT)

1000 m

Initial velocity
Iteration 1 velocity
Iteration 2 velocity
Iteration 3 velocity
Iteration 4 final velocity

600 ms

Measured interval velocity

Initial model for reflection tomography

2 km

Low velocity anomaly


captured by refraction
tomography

Velocity reversal well captured


thanks to reflection tomography

Final model

5000 m/s

Gamma reduction

1000 m

1400 m/s

Converging RMO (projected)

2 km

Iteration #
51

Improvement of image at shallow interval


Enable the delineation of shallow gas anomaly

500 m

2002 Post-STM: IL20600


converted in depth

Shallow slice
2002 dataset

1.5 km
2010 PSDM: IL20600

2.5 km
Shallow slice
2010 PSDM dataset

500 m

Low velocity anomaly


imaged properly

52

1.5 km

2.5 km

Improvements of image at target interval

COHERENCY
@ BETA

2002 Post-STM: IL20500

500 ms

2.5 km

2 km
Note: AGC was applied on the dataset
53

Improvements of image at target interval

COHERENCY
@ BETA

2010 PSDM: IL20500

600 m

2.5 km

More focused fault image


Higher S/N ratio
More preserved amplitude
Clearer shallow gas image

Push-down and statics not


fully solved

2 km

54

Operational added-values brought by PSDM dataset


More accurate post-mortem evaluation of recently drilled wells
SW

NE

Well G

WEST

CENTRAL

EAST

SW

NE

Well G

WEST

CENTRAL

EAST

Well G drilled before PSDM (targeting


Central Panel)
Well G results (sedimentology, fluid,
etc) are more coherent with the West
Panel consistent with the new
structural scheme on PSDM
Structural interpretation on PSDM
performed independently from well G
results

Old
interpretatio
n overlaid on
PSDM

750 m

BETA old map

750 m

BETA new map

Well G

1 km
55

Well G

1 km

Two other wells drilled after PSDM


showed acceptable consistency in
term of structural scheme and
expected HC column

Conclusions of Bekapai PSDM example

Reliable full-field velocity model was obtained by combining refraction


tomography at shallow depth and reflection tomography at deeper depth
Refraction tomography proved to be a solution to build a reliable shallow
velocity model that would not be achieved by using reflection tomography
Efficiency of refraction tomography relies on the quality of first break picks;
QC and editing are therefore fundamental, but time consuming
PSDM final product has improved drastically structural image and
understanding of the field

56

Outline
Introduction Principles
Types of velocity models
Velocity model building methodology and tools
Anisotropy
Example:
Conclusions

57

Conclusions on VMB

Great variety of tools


Strong involvements of interpreters
The survey should be big enough to allow proper tomography
Time consuming process
A good model is the key to the success of the PSDM project

58

Vous aimerez peut-être aussi