Slide 1

Acknowledgements

Quote

Overview

Part I

Limitations of traditional geostatistics

Stochastic sequential simulation

Practice of sequential simulation

Multiple-point Geostatistics

Extended Normal Equations

Single Normal Equation

The training image module

The SNESIM algorithm

Probabilities from a Search Tree

Example

Where do we get a 3D TI ?

Modular training image

Properties of training image

Part II

Simple question, difficult problem…

Combining sources of information

Conditional independence

Correcting conditional independence

Permanence of ratios hypothesis

Advantages of using ratios

Simple problem…

Example reservoir

P(A|C), A = single-point !

Concept of MODULAR training image

Local rotation angle from seismic

Results

Constrain to local “channel features”

Part III

Production data does not inform geological heterogeneity

Approach

Methodology: two facies

Define a Markov chain

Transition matrix

Parameter r_D

Determine r_D

r_D determines a “perturbation”

Complete algorithm

Examples

Single model

r_D values,
single 1D optimization

Different geology

More wells

Hierarchical matching

Example

Results

Results

More realistic

Conclusions

More on conditional independence