Seismic Inversion: Theory and usage in Petrel
Contents:
•Introductions
•Amplitude variation with offset: basic example
Schlumberger Public
•Seismic inversion: post-stack and pre-stack
•Petrel genetic inversion
Amplitude Variation with Offset (AVO)
•AVO – response is a function of material properties.
•Seismic data considerations – processing and stacking.
Schlumberger Public
Receivers/Geophones
ω
φ φ≠ω
ω φ
AVO Analysis: Methodologies
•AVO classification
Class 2 AVO
Schlumberger Public
•Gradient and intercept
analysis
Class 3 AVO
•Invert seismic for rock Class 1 AVO
properties
Seismic Inversion
•Aim: use seismic data to infer
rock and fluid properties –
beyond well control
Schlumberger Public
•Calibrate with well log data
• Sonic and density
Deterministic Inversion
•Well log calibration using synthetics
Schlumberger Public
Forward
•Fine tune time-depth; Model
•Estimation of the part of
the seismic, i.e. a wavelet
or ‘signal energy’; x =
•The logs and seismic are
‘fixed’.
Inversion
Post Stack vs Pre Stack
Post Stack:
– Can not resolve any angle dependent properties
– Better signal to noise
Schlumberger Public
– Acoustic Impedance
Pre Stack (AVO inversion):
– Can resolve angle dependent properties
– Extract shear information
– Poisson’s Ratio, Lambda, Mu, Density
Acoustic Impedance Inversion
α1ρ1=AI1
Schlumberger Public
α2,ρ2=AI2
AI 2 − AI1
R=
AI 2 + AI1 R (t ) • W (t ) = S (t )
α3,ρ3=AI3
Normal incidence reflectivity
Convolutional model
Schlumberger Public
Petrel Genetic Inversion
Genetic Inversion: an innovative combination
of Neural Nets and Genetic Algorithm
• A new approach to derive an Acoustic Impedance volume by solving
the non linear relation between data;
• Genetic Inversion required input is limited to the seismic amplitudes
Schlumberger Public
cube and well logs.
• Strengths:
• Runs fast and the process is straight forward;
• The Genetic algorithm gives a good chance to achieve the global
minimum error;
• It can be extended to any kind of petrophysical parameter which is
link to the seismic amplitude or other derived seismic attribute (e.g.
Velocity, Density, Porosity, etc).
Artificial Neural Networks
•ANN are using the behavior of the human neuron;
•The processing units are wired together in a complex communication network;
•X = input, W = weights, Threshold, Y = output.
Schlumberger Public
Genetic Algorithm
• Uses Chromosomes’ behavior;
• Genes from parents form are in some way a
whole new chromosome by using mutation
and crossovers;
Schlumberger Public
• The fitness of an organism is measured by
success of the organism in its life
Selection, Cross-over, Mutation
Theory of the Neural Network - Genetic Inversion
Schlumberger Public
Parameters in Genetic Inversion Process
• Genetic Inversion is part of
the Volume attribute library
• Default settings have been
set-up to fit the best, most of
Schlumberger Public
the datasets
• The parameter tab is divided
into 3 categories:
• Learning input
• Settings
• Advanced options
Genetic Inversion in Petrel 2009
Schlumberger Public
Schematic approach of the link between seismic traces and property logs,
through Neural Nets
Time to Depth Relation
The Time Depth
relation for the wells
Schlumberger Public
should be accurate
Data gathering and pre-processing
Schlumberger Public
Genetic Inversion: Range of application
Schlumberger Public
Acoustic Impedance cube generation
Summary for Genetic Inversion
• Robust, easy to use, and produces high quality results;
• The Genetic Inversion can be extended to various
parameters linked to the seismic, for reliable prediction of
reservoir properties;
Schlumberger Public
• It can be use as a first pass before running the Deterministic
or Stochastic Seismic Inversion;
• Results can be used like trends or co-kriging properties for
simulation.
