Limited Offer
Implementation and Interpretation of Machine and Deep Learning to Applied Subsurface Geological Problems
Prediction Models Exploiting Well-Log Information
- 1st Edition - January 1, 2025
- Author: David Wood
- Language: English
- Paperback ISBN:9 7 8 - 0 - 4 4 3 - 2 6 5 1 0 - 5
- eBook ISBN:9 7 8 - 0 - 4 4 3 - 2 6 5 1 1 - 2
Implementation and Interpretation of Machine and Deep Learning to Applied Subsurface Geological Problems: Prediction Models Exploiting Well-Log Information explores the implem… Read more
Purchase options
Institutional subscription on ScienceDirect
Request a sales quoteImplementation and Interpretation of Machine and Deep Learning to Applied Subsurface Geological Problems: Prediction Models Exploiting Well-Log Information explores the implementation of machine and deep learning models to a range of subsurface geological prediction problems commonly encountered in applied resource evaluation and reservoir characterization tasks. It provides readers with insight into how the performance of ML/DL models can be optimized, and sparse datasets of input variables enhanced and/or rescaled, to improve their prediction performances. The author covers a variety of topics in detail, such as regression models to estimate total organic carbon from well-log data, predicting brittleness indexes in tight formation sequences, trapping mechanisms in potential sub-surface carbon storage reservoirs, and several more. Each chapter includes its own introduction, summary, and nomenclature sections together with one or more case studies focused on prediction model implementation related to its topic. The first part of each topic chapter describes the geological issues related to the topic, including an up-to-date literature review. The remainder focuses on prediction modeling of that topic including suitable machine learning and/or deep learning approaches and configurations. Case studies form the latter part of each chapter. Readers in this field will find an invaluable resource to assist them in applying machine and deep learning to their work in sub-surface geoscience.
- Addresses common applied geological problems focused on machine and deep learning implementation with case studies
- Considers regression, classification, and clustering machine learning methods and how to optimize and assess their performance considering suitable error and accuracy metric
- Contrast the pros and cons of multiple machine and deep learning methods
- Includes techniques to improve the identification of geological carbon capture and storage reservoirs, a key part of many energy transition strategies
Geoscience and subsurface engineering are the main disciplines. Petroleum and energy operating companies, such as exploration and development geologists (for modeling, data mining, and interpretation), petrophysicists and well-log analysts (for dataset interpretation), and machine and deep learning analysts (for geological and model implementation insight). Industry service providers, such as energy service-company modelers and data analysts (for product optimization and development). Graduate students, post-doctoral researchers, and other academic researchers
Introduction
Types and objectives of subsurface geological prediction models
Statistical characterization of datasets
Dataset pre-processing and outlier considerations
Model configurations and control variable optimization
Prediction assessment methods and metrics
Chapter 1 Regression models to estimate total organic carbon (TOC) from well-log data
Geological description of TOC in organic-rich sediments
TOC distributions in shale reservoir gas and oil resource assessment
Feature selection of well-log input variables
Multi-K-fold cross-validation analysis and its purpose
Prediction model control/hyperparameter optimization
Regression prediction error analysis
Case study(s)
Chapter 2 Predicting brittleness indexes in tight formation sequences
Mineralogical versus mechanical brittleness indexes
Formation brittleness and shale reservoir productivity potential
Overcoming the challenge of sparse well-log data
Exploiting well-log attributes to improve predictions
Transparent data matching ML algorithm for outlier detection
Case study(s)
Chapter 3 Classifying lithofacies in clastic, carbonate, and mixed reservoir sequences
Challenges of macro-scale lithofacies predictions from well-log data
Appropriate well-log log data suites for lithofacies prediction
Optimizer-machine learning combinations for feature selection
Classification accuracy metrics and annotated confusion matrices
Case study(s)
Chapter 4 Permeability and water saturation distributions in complex reservoirs
Permeability and fluid saturation distributions are required for quantitative reservoir analysis
Complexities of heterogeneous carbonate reservoirs
Exploiting lithofacies information to improve permeability and fluid saturation predictions
Comparison of Pearson and Spearman correlation coefficients to characterize datasets
Benefits of tree-ensemble ML prediction models
Case study(s)
Chapter 5 Trapping mechanisms in potential sub-surface carbon storage reservoirs
Characteristic of sub-surface carbon storage reservoirs
Potential for carbon capture with enhanced oil /gas recovery
Determination and prediction of reservoir CO2 trapping indexes
Regression error characteristics and area-over-curve analysis
Case study(s)
Chapter 6 The accurate picking of formation tops in field development wells
Resource volume implications of determining field-wide reservoir boundaries
Problems associated with transitional reservoir formation boundaries
Overcoming the limited number of well-log curves in development wells
Partial-dependence plots between input and target variables
Case study(s)
Chapter 7 Assessing formation loss of circulation risks with mud-log datasets
Identification of subsurface formations prone to loss of circulation
Classifying the severity of loss of circulation
Exploiting mud-log drilling data variables to predict loss of circulation
Noise reduction to reduce the impact of outliers
Pros and cons of deep learning versus classical learning models
Case study(s)
Chapter 8 Delineating fracture densities and apertures using well-log image data
The influential role of natural fractures in resource recovery in carbonate reservoirs
Fracture characteristics and determination using formation image well-logs
Fracture modeling integrating image logs seismic and production data
Combining deep learning and cluster analysis to predict fractures
Committee machine ensemble machine learning models
Case study(s)
Chapter 9 Determining reservoir microfacies using photomicrograph and computed tomography image data
Reservoir microfacies characterization techniques
Clastic and carbonate reservoir microfacies and diagenetic features
Color adjustments to enhance features of interest
Feature segmentation with nearest neighbor machine learning algorithms
Clustering to delineate microscopic features and pore areas
Case study(s)
Chapter 10 Characterizing coal-bed methane reservoirs with well-log datasets
Compositional, geomechanical, and calorific value variability of coal formations
Beneficial characteristics of coal seam gas reservoirs
Predicting coal proximate analysis from well log data
Estimating coal gross calorific value from proximate data
Benefits of transparent prediction auditing to explain outliers
Case study(s)
Types and objectives of subsurface geological prediction models
Statistical characterization of datasets
Dataset pre-processing and outlier considerations
Model configurations and control variable optimization
Prediction assessment methods and metrics
Chapter 1 Regression models to estimate total organic carbon (TOC) from well-log data
Geological description of TOC in organic-rich sediments
TOC distributions in shale reservoir gas and oil resource assessment
Feature selection of well-log input variables
Multi-K-fold cross-validation analysis and its purpose
Prediction model control/hyperparameter optimization
Regression prediction error analysis
Case study(s)
Chapter 2 Predicting brittleness indexes in tight formation sequences
Mineralogical versus mechanical brittleness indexes
Formation brittleness and shale reservoir productivity potential
Overcoming the challenge of sparse well-log data
Exploiting well-log attributes to improve predictions
Transparent data matching ML algorithm for outlier detection
Case study(s)
Chapter 3 Classifying lithofacies in clastic, carbonate, and mixed reservoir sequences
Challenges of macro-scale lithofacies predictions from well-log data
Appropriate well-log log data suites for lithofacies prediction
Optimizer-machine learning combinations for feature selection
Classification accuracy metrics and annotated confusion matrices
Case study(s)
Chapter 4 Permeability and water saturation distributions in complex reservoirs
Permeability and fluid saturation distributions are required for quantitative reservoir analysis
Complexities of heterogeneous carbonate reservoirs
Exploiting lithofacies information to improve permeability and fluid saturation predictions
Comparison of Pearson and Spearman correlation coefficients to characterize datasets
Benefits of tree-ensemble ML prediction models
Case study(s)
Chapter 5 Trapping mechanisms in potential sub-surface carbon storage reservoirs
Characteristic of sub-surface carbon storage reservoirs
Potential for carbon capture with enhanced oil /gas recovery
Determination and prediction of reservoir CO2 trapping indexes
Regression error characteristics and area-over-curve analysis
Case study(s)
Chapter 6 The accurate picking of formation tops in field development wells
Resource volume implications of determining field-wide reservoir boundaries
Problems associated with transitional reservoir formation boundaries
Overcoming the limited number of well-log curves in development wells
Partial-dependence plots between input and target variables
Case study(s)
Chapter 7 Assessing formation loss of circulation risks with mud-log datasets
Identification of subsurface formations prone to loss of circulation
Classifying the severity of loss of circulation
Exploiting mud-log drilling data variables to predict loss of circulation
Noise reduction to reduce the impact of outliers
Pros and cons of deep learning versus classical learning models
Case study(s)
Chapter 8 Delineating fracture densities and apertures using well-log image data
The influential role of natural fractures in resource recovery in carbonate reservoirs
Fracture characteristics and determination using formation image well-logs
Fracture modeling integrating image logs seismic and production data
Combining deep learning and cluster analysis to predict fractures
Committee machine ensemble machine learning models
Case study(s)
Chapter 9 Determining reservoir microfacies using photomicrograph and computed tomography image data
Reservoir microfacies characterization techniques
Clastic and carbonate reservoir microfacies and diagenetic features
Color adjustments to enhance features of interest
Feature segmentation with nearest neighbor machine learning algorithms
Clustering to delineate microscopic features and pore areas
Case study(s)
Chapter 10 Characterizing coal-bed methane reservoirs with well-log datasets
Compositional, geomechanical, and calorific value variability of coal formations
Beneficial characteristics of coal seam gas reservoirs
Predicting coal proximate analysis from well log data
Estimating coal gross calorific value from proximate data
Benefits of transparent prediction auditing to explain outliers
Case study(s)
- No. of pages: 475
- Language: English
- Edition: 1
- Published: January 1, 2025
- Imprint: Elsevier
- Paperback ISBN: 9780443265105
- eBook ISBN: 9780443265112
DW
David Wood
David A. Wood has more than forty years of international gas, oil, and broader energy experience since gaining his Ph.D. in geosciences from Imperial College London in the 1970s. His expertise covers multiple fields including subsurface geoscience and engineering relating to oil and gas exploration and production, energy supply chain technologies, and efficiencies. For the past two decades, David has worked as an independent international consultant, researcher, training provider, and expert witness. He has published an extensive body of work on geoscience, engineering, energy, and machine learning topics. He currently consults and conducts research on a variety of technical and commercial aspects of energy and environmental issues through his consultancy, DWA Energy Limited. He has extensive editorial experience as a founding editor of Elsevier’s Journal of Natural Gas Science & Engineering in 2008/9 then serving as Editor-in-Chief from 2013 to 2016. He is currently Co-Editor-in-Chief of Advances in Geo-Energy Research.
Affiliations and expertise
Owner/Consultant, DWA Energy Limited, UK