Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
April 7, 2016
Multipoint Geostatistics - Outline
Fundamental question
– How to quantify geological variability?
– How to bridge the gap between “geological concepts, ideas and
interpretations” and actual 3D numerical modeling?
Variogram:
– a mathematical equation (not optimal way to capture geology)
Training Image:
– a single (or multiple) 3D volume (image) explicitly describing the
geological concept
A. Geological Continuity - Training Image
1 2
configurations
3
3. Data events organized in a search tree
Simulation grid
Background Mud ?
? Unknown Location
Data Configuration
B. Multipoint Geostatisitcs - SNESIM
Data Configuration
Training Image
?
?
A = Channel Sand
?
Replicates = # of times the
data configuration occurs
in the TI ?
?
Pr(A | B) = ???
Black/Total
A) 1/4 (Strebelle, 2000)
B) 3/4
C) 1/2
B. Multipoint Geostatisitcs - SNESIM
1. Randomly Visit a Block
2. Calculate Probability
– P(A|B)
A = “event occurs” (e.g. sand occurs at a location)
B = conditioning data (well data & geologic information)
3. Randomly Draw a Value
P(A|B) = 0.75
Mud Sand
B. Multipoint Geostatisitcs - SNESIM
1. Randomly Visit a Block
2. Calculate Probability
– P(A|B)
A = “event occurs” (e.g. sand occurs at a location)
B = conditioning data (well data & geologic information)
3. Randomly Draw a Value
4. Put Value in Block and
Retain for Subsequent
Data Configurations ?
5. Go to next block and
repeat steps 2-5
Data Configuration
B. Multipoint Geostatisitcs - SNESIM
Data Configuration
Training Image
?
A = Channel Sand
PA | B
1
5
Replicates = # of times the
data configuration occurs
in the TI
B. Multipoint Geostatisitcs - SNESIM
1. Randomly Visit a Block
2. Calculate Probability
– P(A|B)
A = “event occurs” (e.g. sand occurs at a location)
B = conditioning data (well data & geologic information)
3. Randomly Draw a Value
4. Put Value in Block and
Retain for Subsequent
Data Configurations
5. Go to next block and
repeat steps 2-5
Reservoir Model
P(A|B) = ____
(A=black square)
2
Random #1 = 0.72
Node 1 = ____ 1
Random #2 = 0.56
Node 2 = ____
Class Example
Training Image
Data Configuration
Reservoir Model
P(A|B) = ____
(A=black square)
2
Random #1 = 0.72
Node 1 = ____ 1
Random #2 = 0.56
Node 2 = ____
Implementation Note - Store probabilities
Training image
Data template
(data search neighborhood)
Search tree
14 11
Construction requires scanning
training image one single time
5 7 5 3
3 1 2 5 3 0 1 1
Minimizes memory demand
Level 1 (1 CD) 1 1 1 1
u 5 7
u u 5 3
u
.
. 1 1 1 1 1 1 1 1
. u2 3 1 u2 u2 2 5 u2 u2 3 0 u2 u2 1 1 u2
.
1 0 1 1 0 3 1 0 1 0 2 0 1 1 1 1
. u 2 u2
. 3 3
Level 4 (4 CD) 1 1 0 2 1 0 1 0 2 0 0 1 1
4u2
3
C. Large Scale Continuity
full
empty
coarse fine
template template
Reproducing geology at all scales
Freeze coarse grid nodes and use them
Training image as conditioning data to simulate finer
grid nodes
Finer simulation grid
channel
shale
Stationarity
Dealing with prior proportions
u? u3
u4
Corrected conditional pdf Conditional pdf inferred u1
Data template
23 14 9 10 11
22 8 1 2 12
21 7 0 3 13
20 6 5 4 14
19 18 17 16 15