Evgeny Tolstukhin*, Reidar Midtun, Pål Navestad, ConocoPhillips Norway

Slides:



Advertisements
Similar presentations
Title Petrophysical Analysis of Fluid Substitution in Gas Bearing Reservoirs to Define Velocity Profiles – Application of Gassmann and Krief Models Digital.
Advertisements

Analysis of High-Throughput Screening Data C371 Fall 2004.
3-D Seismic Waveform Analysis for Reservoir Characterization
U D A Neural Network Approach for Diagnosis in a Continuous Pulp Digester Pascal Dufour, Sharad Bhartiya, Prasad S. Dhurjati, Francis J. Doyle III Department.
Software Quality Ranking: Bringing Order to Software Modules in Testing Fei Xing Michael R. Lyu Ping Guo.
Report on Intrusion Detection and Data Fusion By Ganesh Godavari.
Reclamation Mid-Term Operational Modeling Seasonal to Year-Two Colorado River Streamflow Prediction Workshop CBRFC March 21-22, 2011 Katrina Grantz, PhD.
Chapter 1: Introduction
Business Process Performance Prediction on a Tracked Simulation Model Andrei Solomon, Marin Litoiu– York University.
Introduction to machine learning and data mining 1 iCSC2014, Juan López González, University of Oviedo Introduction to machine learning Juan López González.
Report on Intrusion Detection and Data Fusion By Ganesh Godavari.
Feature selection LING 572 Fei Xia Week 4: 1/29/08 1.
Aemen Lodhi (Georgia Tech) Amogh Dhamdhere (CAIDA)
Today Ensemble Methods. Recap of the course. Classifier Fusion
Potential Benefits of Multiple-Doppler Radar Data to Quantitative Precipitation Forecasting: Assimilation of Simulated Data Using WRF-3DVAR System Soichiro.
How to create property volumes
Bed Contained Tectonic Fold-Related Fractures Flank of Teton Anticline Sawtooth Mnts. W. Montana Miss. Madison Ls.
VALIDATION AND IMPROVEMENT OF THE GOES-R RAINFALL RATE ALGORITHM Background Robert J. Kuligowski, Center for Satellite Applications and Research, NOAA/NESDIS,
Introduction to The Lifting Scheme. Two approaches to make a wavelet transform: –Scaling function and wavelets (dilation equation and wavelet equation)
Classification Ensemble Methods 1
Dario Grana and Tapan Mukerji Sequential approach to Bayesian linear inverse problems in reservoir modeling using Gaussian mixture models SCRF Annual Meeting,
Chapter 12: Correlation and Linear Regression 1.
Hybrid Bayesian Linearized Acoustic Inversion Methodology PhD in Petroleum Engineering Fernando Bordignon Introduction Seismic inversion.
Chapter 11 – Neural Nets © Galit Shmueli and Peter Bruce 2010 Data Mining for Business Intelligence Shmueli, Patel & Bruce.
CMPS 142/242 Review Section Fall 2011 Adapted from Lecture Slides.
SEISMIC ATTRIBUTES FOR RESERVOIR CHARACTERIZATION
68th EAGE Conference and Exhibition, Vienna 1 Impact of Time Lapse Processing on 4D Simultaneous Inversion The Marlim Field Case Study C. Reiser * 1, E.
Howard Community College
Software Defects Cmpe 550 Fall 2005
Data Mining, Machine Learning, Data Analysis, etc. scikit-learn
Machine Learning with Spark MLlib
Hasan Nourdeen Martin Blunt 10 Jan 2017
Chapter 11 Simple Linear Regression and Correlation.
Building Adaptive Basis Function with Continuous Self-Organizing Map
Michigan Technological University
Fast Kernel-Density-Based Classification and Clustering Using P-Trees
Amit Suman and Tapan Mukerji
An ETP Studentship with University of Strathclyde and TNO (NL)
A strategy for managing uncertainty
Discrimination between pressure and fluid saturation using direct non-linear inversion method: an application to time-lapse seismic data Haiyan Zhang,
INVESTIGATION OF SQUEEZE CAST AA7075-B4C COMPOSITES
Automatic Picking of First Arrivals
Advanced Analytics Using Enterprise Miner
SA3202 Statistical Methods for Social Sciences
Agenda Motivation. Components. Deep Learning Approach.
CMPT 733, SPRING 2016 Jiannan Wang
Céline Scheidt, Jef Caers and Philippe Renard
Assessing uncertainties on production forecasting based on production Profile reconstruction from a few Dynamic simulations Gaétan Bardy – PhD Student.
Upscaling of 4D Seismic Data
10701 / Machine Learning Today: - Cross validation,
Implementing Six Sigma Quality
Analytics: Its More than Just Modeling
SPE DISTINGUISHED LECTURER SERIES
Machine Learning Interpretability
Data Mining, Machine Learning, Data Analysis, etc. scikit-learn
Data Mining, Machine Learning, Data Analysis, etc. scikit-learn
Part I Review Highlights, Chap 1, 2
Analysis for Predicting the Selling Price of Apartments Pratik Nikte
Microarray Data Set The microarray data set we are dealing with is represented as a 2d numerical array.
Upscaling Petrophysical Properties to the Seismic Scale
Multivariate Methods Berlin Chen, 2005 References:
Topological Signatures For Fast Mobility Analysis
Poverty Maps for Sri Lanka
Improving Drilling Performance using innovative ways to analyze the downhole information Hani Ibrahim.
by Thea Hincks, Willy Aspinall, Roger Cooke, and Thomas Gernon
Srinivas Neginhal Anantharaman Kalyanaraman CprE 585: Survey Project
Advisor: Dr.vahidipour Zahra salimian Shaghayegh jalali Dec 2017
Outlines Introduction & Objectives Methodology & Workflow
Creating Trends for Reservoir Modelling Using ANN
L. Glimcher, R. Jin, G. Agrawal Presented by: Leo Glimcher
Presentation transcript:

Data-driven estimates of reservoir properties from 3D/4D seismic A brown field study Evgeny Tolstukhin*, Reidar Midtun, Pål Navestad, ConocoPhillips Norway Tetyana Kholodna, CapGemini Norway Evgeny.Tolstukhin@conocophillips.com October 13, 2019

Project: Properties from Seismic Motivation and Business Case Idea / Innovation: Use Machine Learning to predict reservoir properties directly from 3D seismic and well logs data Motivation: Increase value of G&G data through data-driven approach Build a Machine Learning and AI platform for Subsurface Domain Project scope: Duration: 3 months Prove the concept Evaluate business impact Business impact: Drill less water-wet wells and side-tracks Better understanding of reservoir properties and mechanisms October 13, 2019

Digitization of Subsurface: seismic and wells data Survey 2010 Survey 2016 A09 A18 A27 Elastic Impedance Poro well log Swe well log RFT pressure Voxels Table October 13, 2019

Illustration of resolution and sampling effects Sample bias Swe 100 ft average - Observed Well log vs. Seismic Pressure Swat > 0.5 - Observed - Polynom - Linear - Swe Fly by Pluto with the New Horizons probe | New Scientist October 13, 2019

Data available: poro, swe, pressure, A09, A18,A27 and ratios EI A09 EI A09 Poro Swe Swe EI A09 Poro EI A09/A27 October 13, 2019

Causation vs Correlation: poro, swe, pressure, A09, A18,A27 and ratios Properties, available only from well logs EI A09 Seismic, available in 3D volume Poro EI A09/A27 October 13, 2019

Causation vs Correlation Swe EI A09 Color is Pressure Poro EI A09/A27 October 13, 2019

Causation vs Correlation Swe EI A09 Color is Pressure Low Swe High Poro Poro EI A09/A27 October 13, 2019

Causation vs Correlation Swe EI A09 High Swe Low Poro Color is Pressure Poro EI A09/A27 October 13, 2019

Causation vs Correlation: NEXT LEVEL, division into groups or clusters Pore intervals Color is Formation Scope: Try alternative clustering methods: Density Dbscan K-means Normal Mixture Hierarchical Dimesionality reduction: Multi-Dimensional Scaling Principal Component Analysis Try alternative ML methods: Multi-Adaptive Regression Neural Networks Decision Trees Support Vector Machine Random Forest etc. Pressure Intervals A09 Swat 0-1 October 13, 2019

Concept illustration: clustering and prediction SeismicType 1 RockType 1 Cluster 1 Category H Cluster 1 Category H RockType 1 SeismicType 1 RockType 2 Only seismic Use ML model from wells A09, A18, A27 Cluster 2 Category L RockType 2 SeismicType 2 SeismicType 2 Category H Cluster 2 Cluster 2 Category L Category L Cluster 1 RockType 1 Poro, Press, Swe RockType 2 October 13, 2019

Software architecture SAS JMP / R-scripts Select Transform Filter Interpolate CARET Predicted Factor: Category Well Master table K-means Clustering 3 Seismic 3 props per formation Random Forest Predict Cluster With Factor: Water and Factor: Category Predict Swe, Poro, Press within Cluster using Random Forest Clusters Software used in the project: • SAS Enterprise Guide 7.1 • JMP 14.2.0 • JMP.R version 14.0 Distributions Within Clusters Factor: Category High, Medium, Low Scoring in 3D October 13, 2019

Agenda Introduction Methodology review Results Validation Summary October 13, 2019

Results of new Clustering at well level Clusters Swe Poro Pressure Seismic A09 October 13, 2019

Results of new Clustering: zoom Clusters Swe Poro ed ed ed ed Pressure Seismic A09 October 13, 2019

BLIND TEST Prediction stage Validation at well level: future wells drilled in 2017-2018 October 13, 2019

Effect of faults and fault shadows: Well 4 example (prediction validation) Well 4 water saturation Faulted zone Well 4 water saturation Faulted zone October 13, 2019

Summary of prediction at well level: wells drilled in 2017-2018 Property Abs. misfit Quality of prediction, percent of correct voxels Poro, unit < 0.03 82% Swe, unit < 0.1 84% Pressure, psi < 1000 psi 81% What data do we have: ML models trained on wells drilled in 2010-2016 Reservoir properties scored in 3D using 2016 seismic What data do we compare to: «Blind test» or validation wells drilled in 2017-2018 Observed properties from well logs (lumped into «voxels») October 13, 2019

3D Comparison with simulation model: prediction of 3D Water Saturation, formation average Polygons are manual interpetations of water fronts based on well and production data October 13, 2019

Agenda Introduction Methodology review Results Validation Summary October 13, 2019

Summary The methodology allowed to predict 3D volumes of porosity, water saturation and pressure: Predictions show good results at wells and in 3D This data-driven model can be further utilized for: Further quantitative analysis Reservoir characterization Multi-disciplinary communication Key learnings from the project: Agile project management Collaboration between data and geoscientists «Test fast, fail fast, adjust fast» Quick feed-in of more data from Subsurface Data Lake: New wells, new seismic, new simulation models, other observations Consider addition of tracers, pressure, temperature and other data Strength of the methodology: Quick to run and update Overall 10 min from training at well level to scoring in 3D October 13, 2019

Acknowledgements October 13, 2019