Introduction to

Multiple-Point Simulation (MPS)

Well data
Training image
is a conceptual
geological
model
MPS model

Geological
interpretation
(well logs and
analogs)
Pattern
reproduction
conditional
to
available
data

Petroleum Geostatistics, Multiple-Point Simulation. 1

Institute of Petroleum Engineering,
University of Tehran, 2010.
 Variogram-based Modeling
(SIS)
• A complete and robust library of variogram-
based geostatistical programs has been
developed
• But variogram-based techniques do not easily
allow modeling properly reservoir architecture
with complex correlation patterns (e.g. channel
reservoirs)

Well data

Variogram-based
model
γ(h)
+ +
+ +
Variogram +
Modeling +
Variogram
h reproduction

Petroleum Geostatistics, Multiple-Point Simulation. 2

Institute of Petroleum Engineering,
University of Tehran, 2010.
 Alternative to Variogram-Based
Geostatistics: Object-Based Models

• Used when reservoir facies have characteristic

geometrical shapes
• But conditioning is very limited in object-based
models:
– difficult to integrate large number of wells
– poor integration of 3D seismic
• Modeling facies interaction can be extremely
tedious

Well data
Object-based
s model
w

Parameterization of
geological bodies
Model
built
object by
object

Petroleum Geostatistics, Multiple-Point Simulation. 3

Institute of Petroleum Engineering,
University of Tehran, 2010.
 MPS Approach

• MPS geostatistical approach that combines:

– Ability to reproduce “shapes” of object-based
algorithms
– Speed, flexibility and easy data conditioning of
variogram-based algorithms

Well data
Training image
is a conceptual
geological
model
MPS model

Geological
interpretation
(well logs and
analogs)
Pattern
reproduction
conditional
to
available
data

Petroleum Geostatistics, Multiple-Point Simulation. 4

Institute of Petroleum Engineering,
University of Tehran, 2010.
 Training Image to Simulation

From Multi-Point User’s Guide

Petroleum Geostatistics, Multiple-Point Simulation. 5
Institute of Petroleum Engineering,
University of Tehran, 2010.
 Training Image to Simulation

From Multi-Point User’s Guide

Petroleum Geostatistics, Multiple-Point Simulation. 6
Institute of Petroleum Engineering,
University of Tehran, 2010.
 MPS Advantages

• Realistic correlation patterns that

would be extremely difficult to
produce using conventional SIS
approach
• Conceptual (geological) input easy
to provide under the form of a
training image

MPS model
example
Training Image

MPS model
conditioned to 45 data

Petroleum Geostatistics, Multiple-Point Simulation. 7

Institute of Petroleum Engineering,
University of Tehran, 2010.
 MPS Advantages

• Facies interaction (correlation)

modeling
• Fast algorithm: no trial-and-error
process, no convergence issue
• Simulation of 1 million cell grid a few
minutes

Training Image MPS Model

Petroleum Geostatistics, Multiple-Point Simulation. 8

Institute of Petroleum Engineering,
University of Tehran, 2010.
 MPS Advantages

• Wide Range of Input Data:

– Well data
– Geobody data
– Facies proportion soft constraints
– Local deposition azimuth variations
– Local object size variations

Potential
Seismic
geobody
amplitude
to be used
data
as hard
data
Cluster
Analysis

Cluster for
geobody
picking

Petroleum Geostatistics, Multiple-Point Simulation. 9

Institute of Petroleum Engineering,
University of Tehran, 2010.
 Multiple-point Geostatistics

• MPS simulation program: pixel-based

approach building on sequential
simulation paradigm
Training image
Simulation grid
Look for patterns
matching
the conditioning
u2 data
u4 u?
u3
u1

P(sand) = 3/4
P(no-sand) = 1/4
Draw simulated
Updated simulation value

• Stochastic (multiple realizations)

• Easy to condition (pixel-based)
• No more variogram
• General (not specific to channels)
Go to next grid node...
(random walk)

Petroleum Geostatistics, Multiple-Point Simulation. 10

Institute of Petroleum Engineering,
University of Tehran, 2010.
 MPS Workflow
Training image

Generation of
a conceptual
deposition
model
Core data
Analogs MPS model
Geological Depositional trends
expertise
NTG curve/map, map
of facies deposition
directions

Well log data

Seismic derived
sand probability
map
Calibration
of seismic
to well data

Seismic data
Interpreted sand
bodies
Then each facies
can be populated
Seismic
geobody
with Porosity and
Interpretation Permeability using
SGS

Petroleum Geostatistics, Multiple-Point Simulation. 11

Institute of Petroleum Engineering,
University of Tehran, 2010.
 Training Image

• The training image reflects the facies

correlation pattern that we expect to
find in the reservoir, and thus, that
should be reproduced in the
simulation results.

• Training images may be combined to

produce more complicated patterns

Petroleum Geostatistics, Multiple-Point Simulation. 12

Institute of Petroleum Engineering,
University of Tehran, 2010.
 Training Image

• The training image may be generated

from a variety of geometric elements.
• Some simple elements

Ellipse Ellipse
Half-ellipse
Constant thickness

Quarter-ellipse Upper half-ellipse

Upper sigmoid
Rectangle
Lower half-ellipse
Lens
Delta Lower sigmoid

Double sigmoid

Petroleum Geostatistics, Multiple-Point Simulation. 13

Institute of Petroleum Engineering,
University of Tehran, 2010.
 Training Image

• Parameters for geometric elements

(e.g. thickness, length, etc.) may be
given probabilistic distributions

Width Orientation
Length

Thickness

Uniform distribution Triangular distribution Gaussian distribution

Mean=(Min+Max)/2

Sigma

Min Max Min Mode Max Mean

Petroleum Geostatistics, Multiple-Point Simulation. 14

Institute of Petroleum Engineering,
University of Tehran, 2010.
 Training Image

• Sinuosity “components” may also be

specified allow realistic channel
geometries to be included in the
training image

No sinuosity C-sinuosity S-sinuosity User-specified

sinuosity

Length
Amplitude scale
Amplitude
Amplitude

Petroleum Geostatistics, Multiple-Point Simulation. 15

Institute of Petroleum Engineering,
University of Tehran, 2010.
 Training Image

• Facies “interactions” may also be

defined as the training image is built

Facies 2 Facies 2 is Facies 2

erodes eroded contained
facies 1 by facies 1 within facies 1

Petroleum Geostatistics, Multiple-Point Simulation. 16

Institute of Petroleum Engineering,
University of Tehran, 2010.
 Training Image

• Facies vertical constraints may also be

defined as the training image is built

Facies 2 Facies 2
No vertical
above below
constraint
facies 1 facies 1

Petroleum Geostatistics, Multiple-Point Simulation. 17

Institute of Petroleum Engineering,
University of Tehran, 2010.
 Training Image

• May be as simple or complicated

as appropriate. Tools exist (and
others are being developed) to
build complicated training
images efficiently. Many Fluvial deltaic
complicated training images deposition
have been generated already environment
and can be shared. “tool”

Conceptual model Training Image

Lagoon
(4) Red
Bed (4)
Abando
ned
lobe

Shorefa
Fan Lobes
ce (0,1) Outer Fan
(3)
(2)

Petroleum Geostatistics, Multiple-Point Simulation. 18

Institute of Petroleum Engineering,
University of Tehran, 2010.
 Training Image

• Additional constraints may be added to

control the distribution and orientation
of the training image

Training Image MPS model

3
0
6
0

9
0

1.5 .5

Petroleum Geostatistics, Multiple-Point Simulation. 19

Institute of Petroleum Engineering,
University of Tehran, 2010.
 MPS Simulation

• Constrained to hard data during

simulation
MPS model MPS model
Conditioned to 9 Conditioned to
Training Image
data 45 data

SIS model
Conditioned to 45 data

Petroleum Geostatistics, Multiple-Point Simulation. 20

Institute of Petroleum Engineering,
University of Tehran, 2010.
 MPS Simulation

• Soft data may be used as additional

constraints

Facies
proportion MPS model
Map

20% Sand

50% Sand

20% Sand

Petroleum Geostatistics, Multiple-Point Simulation. 21

Institute of Petroleum Engineering,
University of Tehran, 2010.
 MPS Simulation

• Construction of training images

– Geological information
– Simulation and analysis of geological
processes (e.g. event-based simulation)
Event-based channelized sheet complex model

Streamlines and Channel and Levee, Sheets, and Frontal Splay Cross Sections

Trend in Element Geometries

(increasing size of CH and LV)

Longitudinal Hierarchical Trends, and Architectural Elements

Petroleum Geostatistics, Multiple-Point Simulation. 22

Institute of Petroleum Engineering,
University of Tehran, 2010.