Sei sulla pagina 1di 138

A

Velocity model building


using Petrel software

Universiti Teknologi Petronas


Centre Of Excellence In Subsurface Seismic Imaging
& Hydrocarbon Prediction (CSI)

Amir Abbas Babasafari


November 2019 1
Outline

• Velocity modeling the principles and pitfalls

• Well and seismic velocity data

• Incorporating velocity data to build a reliable model in Petrel software

• Time to Depth conversion (Map and reservoir property)

• Residual error correction and well marker adjustment

• Structural uncertainty

In this presentation some figures adapted from Dr. Badley, Dr. Robertson, Dr. Abdollahi far and Dr. Nosrat
and courtesy of Schlumberger, CGG, Jason and dGB. 2
Seismic Structural Interpretation

• Data gathering, loading and QC


• Well top correlation
• Data conditioning
 Seismic data conditioning
 Well data conditioning
• Well to seismic tie and horizon identification
• Time structural interpretation
 Seismic attribute generation
 Horizon picking
 Fault interpretation
• Velocity model building
• Depth conversion and mapping

3
Seismic dataset:
• Isotropic/Anisotropic Time migrated seismic data
• Isotropic/Anisotropic Depth migrated seismic data 4
Depth Conversion

 Geometric distortions due to velocity changes (pitfalls) will be removed

 To predict drilling depth to the target horizon

 More accurate Reserve Calculations and Uncertainty Quantification

 For basin modelling purpose

5
Pitfalls and issues in seismic data interpretation affecting seismic data quality and S/N ratio
Inherent : steep dip
fault zone
reflectivity
Acquisition : acquisition footprint
surface condition
navigation
receiver problem
shot problem
missed shots
recording problem
crooked line
feathering in marine
Processing : time mismatches
mute
polarity differences
vertical anomalies
static problem
filtering
Others : migration & sideswipe
display
tuning
velocity effects
multiples and bottom simulating reflectors
6
llimits of software packages
Common Velocity Pitfalls:

• Anomalous high/low velocity zone (lithology)

• Lateral lithofacies changes

• Fault zones

• Gas effect

7
Seismic data acquisition

½ * Two-Way Time * Velocity = Depth

8
Velocity effects

Velocity pull up

Velocity push down

Variations in velocity produce apparent structures which may not exist.

9
Velocity effects and depth migration

Time migrated section

Depth migrated section

Depth migration accounts for lateral variations


in velocity and can minimise the appearance
of spurious structures
10
Drastic lithology changes

Lateral lithofacies changes

11
Fault shadows

A subtle form of velocity effect can


produce not just spurious folds but
also apparent faults

12
Velocity Distortion

Increasing velocity downdip


- the interval appears to thin

13
DEPTH Planar faults appear
Listric

Uniform thickness
beds appear to thin
with depth

TIME
Distortion of Structure
on Time Sections

14
Time and Depth Sections

Salt Layer – 4600m/s

15
Time section

Note that the water depth increases from 100m on the


right to 2.2km on the left

Depth section Prospect

The prospect is now imaged as a structural closure. The rapid lateral variations in
water depth and overburden are responsible for the distortion of the time section.

Depth Conversion
16
Velocity push down
due to gas cloud

17
Input data

1. Well data (markers and velocity)


2. Seismic velocity (Stacking or Migration)
3. Time (TWT) surfaces

Well velocity data include check-shot and VSP

18
19
20
21
22
1. Well velocity data

Time-Depth Curve

3000
2500
(millseconds)
Two way Time
2000
1500
1000
500
0
0 5000 10000 15000 20000
Depth (ft)

23
In addition, VSP data provides corridor stack which can be compared with a synthetic seismogram and seismic data
at a well location.

24
2. Seismic velocity data

Stacking velocity

25
Velocity Definition
Interval velocity Vi Stacking velocity

Root-mean-square (RMS) velocity


Average velocity

RMS velocity Dix conversion Interval and Average velocity

V1, 𝝙t1

V2, 𝝙t2 Horizontal isotropic


layering
V3, 𝝙t3
V4, 𝝙t4
V5, 𝝙t5
26
Stacking velocities are
typically a few percent higher
than well velocities

Well velocity
Stacking Velocity

Well and Seismic Velocities


27
Fundamentals of Geostatistics
1. PDF

28
Probability distribution histogram
29
Skewness
Kurtosis

30
31
32
2.Variogram
Variogram (γ)

“Sill”

“Range” Distance (h)

33
Variogram
0 5 10 15 20
0
V ariable h=1 h=2 h=3
1
2
9.4
12.7
xi  xi 1 2x  xi  2x  x 2
2
i
i i 3
3 8.6
4 9.5 14.0

5 5 10.3
12.0
6 10.8
7 7.7
10.0
8 6.9

バリオグラム(γ)
9 9.7
8.0
10 10 11.3
11 12.7
6.0
12 10.5
13 12.3
4.0
14 9.6
15 15 14.6 2.0
16 15.4
17 14.5 0.0
1 N
18
19
15.3
16.4
 1   xi  xi h 2 0 5 10
距離(h)
15 20

20 20 9.9
2 N i 1
21 8.2

1 N
25  h    xi  xi h 2
2 N i 1 34
Variogram
Variogram(γ) Experimental Variogram

“Sill”
Variogram Modeling
• Spherical Function
• Exponential Function
“Nugget” • Gaussian Function

Distance(h)
“Range”

Vertical Variogram (Min Range)


Horizontal Variogram (Max/Med Range)

35
Variogram(γ) Covariance(C)

“Sill”
C h    2   h 

“Nugget”

Distance(h)
“Range” Distance(h)

36
3. Interpolation algorithm
Kriging

37
Kriging

• Well data / Primary variable


+ Seismic data/ Secondary variable

CoKriging

38
39
Seismic data contribution

Why is this important? Well 1 Well 2 Well 3

In Field Development:

Example Field Study

• Water breakthrough problems


in all 3 wells

After Weber et al.,


• Decision made to inject water 1995
in well 2 to stimulate
production in well 3

Grainstone distribution

40
Why is this important?
Well 1 Well 2 Well 3
Wrong decision because:

• Original correlation based on


lithostratigraphy

• New correlation based on


chronostratigraphy using
seismic data

After Weber et al.,


1995

Grainstone distribution

41
Objective: Incorporating well and seismic data for a reliable velocity model

a) Well data b) Seismic data

c) Incorporating well and seismic data

N
2000 m

42
Structural Uncertainty

NW

1
2
3

43
Some QC steps for horizon interpretation
before velocity modeling
 Seismic data conditioning
• Using DSMF volume to enhance auto tracking quality and time horizon interpretation
• Using variance and ant track cubes to illustrate faults trend

 Tying loops
• Various inline, crossline and arbitrary lines passing through all wells to cover the entire field

 Auto tracking / Manual Picking


• 2D Auto tracking/ Manual Picking

 Using paint brush by setting parameters for 3D tracking


 Displaying next & previous horizons as a guidance
 Flattening horizons to find reflector’s continuity
 Quality Controlling in the cross line directions to follow reflectors
 Using seismic surface attribute such as extract amplitude value
 Isochrone map generation to control thickness variations
 TDR creation for interval velocity checking at well locations
44
Some QC steps for fault interpretation
before velocity modeling

1. Extracting Steered cube for Dip and Azimuth calculation based on seismic events.
2. Generating Variance, chaos and curvature attribute volumes to illustrate fault trends and orientations.
3. Providing Ant track cube and confining dip and azimuth to evaluate minor faults and fractures on the
basis of seismic data resolution.
4. Generating surface attribute maps of Variance and Ant track.
5. Fault interpretation on seismic sections using co-volume cubes which were generated.
Interval 10 inline by 10 inline or 5 by 5 (depends on tectonic setting) and quality checked on Variance
attribute maps.
6. Building fault sticks and fault planes in time domain.

45
Well (red color point) and seismic (green color point) velocity data in Petrel

46
Seismic stacking velocity grid: 200 * 200 or 100 * 100 meters
Interval Average Seismic
velocity at velocity at Stacking
well location well location velocity
47
Data preparation in Petrel

1. Sonic log (DT) correction with check-shot


2. Well to seismic tie using corrected sonic log
3. Applying the obtained TDR (Time Depth Relation) on well

More appropriate match between markers and predicted depth map is achieved at well locations after
conducting the sequences above.

48
1

49
1

50
2

51
2

52
3

53
Velocity Modeling in Petrel

54
55
56
57
58
59
60
61
62
63
64
Velocity Modeling in Petrel

1.Function approach

2.K approach

3.Layer Cake approach

4.Average velocity approach


(segy or property format)

5.F_Anisotropy Approach
65
1. Function approach (simple)

1
2

66
3

67
or

68
TDR for more than 1 well
Deficiency: Fitting only 1 function that can represents the velocity variation
of all wells is not possible.

69
2. K approach

Vertical variation of velocity

70
71
72
73
74
75
76
77
78
79
80
81
82
83
3. Layer Cake approach

1. Seismic interval velocity extraction between main horizons


2. Outlier points elimination using Time vs. Int. velocity cross plot
3. Interpolation, smoothing and interval velocity map creation
4. Calibrating with well interval velocities using co-kriging collocated method
5. Depth conversion using velocity grid
6. Well top adjustment
7. Performing blind test and cross validation for depth conversion
8. Cross section QC
9. Thickness map QC

Note:
• Average velocity surface for the first horizon by incorporating well and seismic
• Interval velocity surface for the second horizon onward by incorporating well and seismic
84
85
ASCII format: Right click and open Spreadsheet
Interval velocity calculation using stacking velocity
1

86
87
88
89
90
Average velocity calculation of markers at well
3

91
4

6
92
7 93
Interval velocity calculation of markers at well
bold

94
3

Anomaly?

95
Velocity surface generation using only well data

96
97
Velocity surface generation using well and seismic data

98
Well interval velocity Seismic interval velocity

99
Incorporating well and seismic interval velocity (Velocity surface)
Make a velocity model using velocity surface

Residual errors

100
Well top adjustment (1)

101
Well top adjustment (2)

102
Depth map after well top correction

Depth map before well top correction

103
Depth conversion

Horizon
Fault
Seismic section
/2 …

104
/3

105
/4

Horizon
Fault
Seismic section
Model including reservoir property

106
Note: Once the reservoir property e.g. porosity and water saturation is
converted to depth domain, the correlation coefficient and error
between measured and predicted reservoir property at well locations
should be checked.
Slight change in correlation and error between time and depth domain is
acceptable, while in the case of observing significant change the velocity
model needs to be updated.

107
Making thickness map

Isochore
Isopach

108
4. Average velocity approach

109
110
111
112
113
114
115
116
117
118
5. F_Anisotropy Approach

119
120
121
122
123
124
Structural Uncertainty

125
• Make contact
• Volume calculation (base case)
• Std. Dev derived from depth error estimation
• Uncertainty and Optimization Process
• Uncertainty results

Managing drilling risk

126
Case study

127
Stacking velocity
128
Velocity model methods

Method1 Method2 Method3 Method4 Method5 Method6


(Average) (Average) (Average) (Interval) (Interval) (Interval)

Calibrated Co-kriging Trend Layer Cake Anisotropy Trend (inversion)

129
Calibrated method

1. A simple grid construction and layering


2. Scaling up well average velocity (TDR) at well locations
3. Interpolation and smoothing of average velocity derived from seismic
stacking velocity and average velocity map generation for each interval
separately
4. Calculation of a fraction from dividing well average velocity (TDR) by
average velocity derived from seismic stacking velocity maps at well
locations
5. Interpolation of fraction values using kriging method by determination of
major/minor direction and range for variography (interpolated fraction)
6. Multiplying the average velocity derived from seismic stacking velocity (3)
by interpolated fraction (5) to calibrate it at well locations (velocity model)
7. Depth conversion using velocity model
8. Well top adjustment
9. Performing blind test and cross validation for depth conversion
10. Cross section QC
11. Thickness map QC
130
Co-kriging method

1. A simple grid construction and layering


2. Scaling up well average velocity (TDR) at well locations
3. Interpolation and smoothing of average velocity derived from seismic
stacking velocity and average velocity map generation for each interval
separately
4. Velocity model building through geostatistical method combination of well
average velocity (2) as primary data and average velocity derived from
seismic stacking velocity (3) as secondary data (trend using co-kriging
algorithm). “Using Petrophysical modeling in Petrel”
5. Depth conversion using velocity model
6. Well top adjustment
7. Performing blind test and cross validation for depth conversion
8. Cross section QC
9. Thickness map QC

131
Trend method

1. A simple grid construction and layering


2. Scaling up well average velocity (TDR) at well locations
3. Interpolation and smoothing of average velocity derived from seismic
stacking velocity and average velocity map generation for each interval
separately
4. Velocity model building through geostatistical method combination of well
average velocity (2) as primary data and average velocity derived from
seismic stacking velocity (3) as secondary data (trend using calculation of a
fraction via subtraction of well average velocity (TDR) from seismic average
velocity at well locations, subsequently interpolation and adding to seismic
stacking velocity for calibration). “Using Petrophysical modeling in Petrel”
5. Depth conversion using velocity model
6. Well top adjustment
7. Performing blind test and cross validation for depth conversion
8. Cross section QC
9. Thickness map QC
132
133
Structural Uncertainty

134
Well Method1 Method2 Method3 Method4
Blind test 1 2.04 -4.07 2.1 -3.58
2 2.89 4.84 3.15 -0.4
3 -7.14 -17.78 -7.74 0.08
4 11.54 2.78 12.12 2.91

135
Checking mean and skewness in distribution histogram of residual depth errors to avoid
over/under estimation of bulk and reserve calculation

136
Distribution histogram of Dip map
137
Thanks for your attention

a.babasafari@yahoo.com

138

Potrebbero piacerti anche