AAGGP Bs As, ArgentinaNovember 12, 2004
Recent Advances and Current Challenges in Geoscience Technology
Fred Aminzadeh,
Outline
KEY PRACTICAL PROBLEMS
MAJOR ADVANCES IN GEOPHYSICAL TECHNOLOGIS
FUTURE TRENDS
A FEW EXAMPLES
Major Challenges•• 1- Accurate positioning & detection of sub-salt /salt-flank plays• 2- Characterization of thinly laminated sand / shale sequence• 3- Deep exploration and accurate depth imaging
• 4- Distinguishing commercial gas from non-commercial gas• 5- Fault detection and their types (e.g. sealing vs. non-sealing)• 6- Environmental issues, geohazard and remediation
• 7- Exploration in difficult areas (gas clouds, mud volcano, basalt)• 8- Fluid and permeability prediction, detection and monitoring• 9- Fractures: type, orientation, frequency and connectivity• 10-Prediction of over-pressured reservoir zones.
New Technology Trends
1- Broader use of statistics / soft computing
2- Depth imaging and modeling
3- Dynamic reservoir characterization
4- Linking seismic patterns to rock properties
5- New acquisition methods / Instrumented oil fields
6- Integrating Seismic and Basin Modeling Information
7- Processing in the compressed domain/ data mining
8- Seismic while drilling and real time imaging
9- True integration data and knowledge
10-Immersive Visualization and Interpretation
Crossbedding – two scales
Well (Logs)
Scale, Uncertainty Resolution Environment
Rock (core)
Field (Seismic Attributes)
SURE CHALLENGE
TRADITIONAL & UNCONVENTIONAL STATISTICS
Wave Equations with Random VariablesMultiple Realizations of Seismic SectionsFuzzy Logic, Neural Networks, Genetic Algorithms, DAIChaos Theory, Complexity Theory
“As complexity increases, precise statements lose meaning and meaningful statements lose precision.”
L.A. Zadeh
Balanced training set
Impact of balanced data set
Unbalanced training set
Unbalanced training set Balanced training set
Impact of Balancing on Prediction
Seismic character classes for different lithologies
1
2
3
4
0
10
20
30
40
%
A B C D E F G
Class
Seismic patterns Divided into 4 classes
A-Evaporates, B-Silt/Shales, C- Shore line, D- Wet Dunes, E- Dry Dunes, F- Fanglomerates, G- Volcanics
Classification Confidence
Inherent fuzziness in geology / seismic data
From Aminzadeh and Wilkinson (2004)
Fuzzy Boundaries in Rock Type Classification
From J. Gerard, 2001
THE CONCEPT OF PSEUDO-Xn*
Pseudo-number
n
(non-precisiable granule)
Pseudo-function
X• if f is a function from reals to reals, f* is a function from reals to pseudo-numbers
(non-precisiable)
f*
Y
f
O
O
O
O
O
O
111
1
Energy
Input Attributes:Energy, Frequency, Cube Similarity Continuity, Dip Var., Azimuth Var.,
Absorption, Curvature, ..
Interpreter’sKnowledge
HAIAttribute
ANN
HAI: Combining Human Intelligence and AI
Can we model seismic patterns by DNA
From Aminzadeh (2004)
Depth Imaging & ModelingNext-generation earth modeling will incorporate quantitative representations of geological processes and stratigraphic / structural variability. Uncertainty will be quantified and built into the models The shared earth model will become the centerpiece for end-to-end technology integration from seismic imaging through reservoir and well performance simulation. As cycle time and costs for reservoir modeling and prediction fall, integration and feedback with real-time operational data become practical.
“So far as the law of mathematics refer to reality they are not certain; and so far as they are certain, they don’t refer to reality” Albert Einstein, 1951
CONSTRUCT MODEL HORIZON AND FAULTS
3D Reservoir Modeling
3-D Structural Model Building
SEG/EAGE Salt and OverthrustModeling Project 3 TB data, $25Million1994-1996
Courtesy of SEG
Highly Faulted Salt Area
Curtesy of Seislink
Seismic Imaging
Seismic imaging technology will continue to improve interpretation reliability in complex imaging environments, suchas sub-salt and overthrust playsInteractive 3D depth imaging and velocity modeling will become a standard procedure.Imaging while drilling will help fine tune subsurface images through new image updating techniques. Multi-component and time lapse measurements will grow in use for the reservoir management business, driving down unit costs, expanding application, and stimulating advances in imaging and interpretation technologies.
3D TIME/DEPTH MIGRATION TO CONSTRAIN SALT DOME
Courtesy of Unocal
DYNAMIC RESERVOIR CHARACTERIZATION
Isolating changes in reservoirs from acquisition foot prints
Sensitivity of rock properties to fluid/temperature
Permanent sensors
Incorporating time lapse seismic, log and production data
Efficient reservoir updating and visualization methods
Time: What a wonderful dimension, alas for now, we can only go in one direction: Carl Sagan, 1996
Temperature
0
0
1000
2000
3000
4000
5000
6000
7000
8000
9000 10000
11000
Am
plitu
de
‘96 ‘97
Tem
pera
ture
Incr
ease
( F
)
0
50
1100
1200
250
300
150
‘96 ‘97
Time-lapse welllog, seismic &production data
Steam Thickness Saturation
Dynamic ReservoirCharacterization
&Production OptimizationGeology & Core
Statistical Methods(Regression, clusters, cross plot, ANN, …)
Reservoir PropertiesData
Physical Methods(Rock physics, bright spot, S wave, …)
Hybrid Method
seismic, log, …)Uncertainty
Reservoir PropertiesData
Dynamic Reservoir Model Building
seismic, log, time laps data, production data
Uncertainty
RecursiveUpdating &
Visualization
FROM ATTRIBUTES TO ROCK PROPERTIES
Statistical Approach
Physics-Based Approach
Indirect Relationships (Formation Character)
Hybrid Methods
We can recognize a pattern far more easily than we can explain how we recognize it. We don’t even have reliable numerical confidence measures of patterns we think we can recognize. Bart Kosko, 1997
Reservoir PropertiesData
Physical Methods(Impedance contrast, bright spot, absorption, Bio-Gassmann, shear wave splitting,, etc.)
seismic, log, core, geology Degree of Match
λ/ λ/ λ24
Att
ribu
te m
agni
tude
Amp.Freq.Att3
Target zone thickness
λ/8
Model-Based Reservoir Characterization
Geologic Framework
Data Segmentation: Real wells =>Pseudo wells => Synthetic
Geologic Framework
Fluid Factor Attributes
Pow
er
Frequency
•Avg. Freq. Squared (AFS)•Max. Spect. Amp. (MSA)•Freq. Slope Fall (FSF)•Absorption Qual. Factor (AQF)•MSA*Dom. Freq. (MDA)
Shift of dominant frequency
NEW ACQUISITION METHODS
Ocean Bottom Surveys4 Component Data
Instrumented Oil Fields (4D Seismic)ROV’s and Robots for Placing PhonesContinuous PhonesVery Large Offset DataCross-Well Seismic+Cross Well EM
Time-lapse Inversion
φSw
Reflectivity
91
97
Acoustic imp.
Elastic imp.
InversionInput
Reservoir simulator
Gassmann
Shuey’sequation
91Timeequivalentlogs
AVO-modelled time-eq. logs
97 91
97
91
97
Water saturation volumes
0
1
Saturation 91
Saturation 97
Difference 97 -91
0.3
0
NONLINEAR SIGNAL PROCESSING
AnisotropyChaotic RegimesFractalsWavelets
Our seismic signal processing methods are all based on the Convolution models. What if, mother earth refuses to
convolve?Sven Treitel, 1995
A slice of fault cube
Faults classified bysupervised MLP neuralnetwork using 22 attributes
Fault Blocks Responsible for Hydrocarbon Migration Derived from Non-Linear Signal Processing
1 km
Non-Linear Generalized Median Filter
Before After
DATA COMPRESSION / DATA MINING
Data ExplosionRemote/ Web-Based ProcessingProcessing in the Compressed DomainZeroing on the Information-Rich Part of the DataIT trends
Where is the beef? Burger King Commercial, 1983
IT TRENDS
Seismic processing and interpretation technology will continue to ride the exponential growth curve in computational capability and connectivity. Virtually all work will be done via the Web, with interactive, on-demand access to applications, compute engines, and data. The Web will become the key enabler, reducing costs, providing worldwide access to technology, and supporting adaptation to local exploration and reservoir management requirements. More sophisticated human-machine-data interfaces will be developed beyond the capabilities of today’s visualization centers. Advanced measurement systems will create a new flood of data about reservoirs, wells, and operating facilities
Neural Network Energy Attribute
Dip Variance Similarity (Coherence type)
Attributes Along Fault Plane
Saturation > 3%
Chimney Cube Data
Integrating Chimney Analysis & Basin Modeling
Basin Model (Petromod) Accumulation of Gas Present Day
Source Rock ExpulsionSource Rock Expulsion
Fluid migration detection:
provides information on where hydrocarbons are being generated
Local fluid migration activity in source rocks
Assumed to be related to active hydrocarbon expulsion
Outline of ‘kitchen’ areaOutline of ‘kitchen’ area
Fault Fault Outline of modelled maturation area
Chimney activity
Other ‘kitchen’?
Analyse observations in Basin model and update if required
SEISMIC WHILE DRILLING
MWD and beyond Imaging and Image updating while drilling Oscillatory Drilling (Roto-Rooter)Use of Passive seismic dataUse of production noise as seismic source
3D Model Updating while Drilling
STRUCTURAL + THICKNESS
+ PETROPHYSICAL
INTEGRATION OF DATA, KNOWLEDGE AND DISCIPLINES
True Integration: Disciplines not results Data Fusion Methods Knowledge IntegrationGeo-Engineer: The Wave of the Future
Let’s Integrate
Least Wrong (Not Least Square)
Solution?
s
GeologyPetro-Physics
GeophysicsRock Physics
Geostatistics
Integration of Disciplines vs Results
Geology Geophysics Petrophysics
Geochemistry ReservoirEngineering
INTEGRATED RESERVOIR PROPERTIES AND STRUCTURAL MODELING
VISUALIZATION AND EMMERSIVE TECHNOLOGY
Virtual Reality: where are we? Next Generation VR in the oil industryVisualization as an data integration toolReduce cost, improve efficiency
A picture is worth a thousand words, Chinese expression
3D seismic and log attribute upscaling and girding
Courtesy of T-surf
E
AVO ( I*G)
Seal in a key block
Chimneys
Courtesy of Pemex
AVO on selected horizons with chimneys and wells
Sea floor
Deep Reservoir Sands
Salt
Chimneys
Shallow Reservoir Sands
The Chimney CubeThe Chimney Cube
Conclusions
Many new opportuntiesEstablishing the value of new technology applications
Reducing costDoing things fasterSolving more complicated problems
Risk reductionTrue integrationData Mining
Top Related