Short Courses

Geological Applications of Compositional Data Analysis: A Practical Introduction

  • organized by: Mark Engle (Eastern Energy Resources Science Center, U.S. Geological Survey, El Paso, TX) and Madalyn Blondes (Eastern Energy Resources Science Center, U.S. Geological Survey, Reston, VA)
  • Venue:
  • Date: August 11, 2019
  • Time: 9am-12am, 1pm-4pm
  • Min. number of participants: 10
  • Max. number of participants: 50
  • costs per participant: US$ 250

Many types of data used in the earth sciences are compositional, meaning they are composed of relative variables or parts. Common examples include soil texture or grain size distribution, elemental or oxide concentrations, relative abundance of minerals or coal macerals, etc. It has been shown that, for mathematical reasons, analysis of compositional data using classical tools, such as correlation or even simple plotting, can provide results that are misleading or incorrect. Methods based on log- ratio transformations, so-called Compositional Data Analysis (CoDA), have been developed in an attempt to prevent some of these known problems in the interpretation and analysis of compositional data. In this short course, we will introduce attendees to the basic concepts of CoDA including the 3 basic log-ratio transformations, run through real-world geological examples using the CoDaPack software package. he latter portion of the course will cover some more advanced topics and current areas of research and rovide citations for future information

schedule:
  1. Introduction to compositional data
  2. Basic concepts of compositional data analysis (CoDA)
  3. Univariate and bivariate analysis
  4. Multivariate exploratory data analysis
  5. Creation and uses of the isometric log-ratio (ilr)
  6. Advanced topics
For more information click here.

Introduction to Digital Rock Physics

  • organized by: Dr. James McClure (Virginia Tech)
  • Venue:
  • Date: August 10, 2019
  • Time:
  • Min. number of participants: 10
  • Max. number of participants: 50
  • costs per participant: US$ 250

This workshop will provide an introduction to computational approaches for digital rock physics based on capabilities within the Digital Rocks Portal and the LBPM software package. Participants will be introduced to the capabilities of the Digital Rocks Portal and learn how to manage, publish and visualize digital rock data sets using the web-based framework. Existing data sets published within Digital Rocks Portal will be used to demonstrate simulation workflows used to measure permeability and relative permeability based on 3D rock geometries.  Traditional workflows for segmentation of micro-CT data sets will be presented alongside emerging approaches to train neural networks to perform volumetric segmentation.

  • Overview of Digital Rocks Portal capabilities and data management best practices
  • Visualization of data with ParaView
  • Permeability and Relative permeability simulation using LBPM
  • Segmentation of micro-CT data
  • Python workflows to train deep learning models using digital rock data

Machine Learning for Geoscience Modelling: Introduction and Advanced Topics with Case Studies

  • organized by: Prof. Mikhail Kanevski (University of Lausanne, Switzerland) and Prof. Vasily Demyanov (Heriot-Watt University, UK)
  • Venue:
  • Date: August 10, 2019
  • Time:
  • Min. number of participants: 10
  • Max. number of participants: 50
  • costs per participant: US$ 250

Machine Learning (ML) algorithms have gained a great popularity in geoscience studies and applications for analysis, modelling and visualization of complex high dimensional data to improve understanding of natural systems’ behavior and their interaction with human activity. The short course presents both an introduction to the basics of ML and insights into advanced ML topics and their applications in geoscience modelling. The presentations are well illustrated by simulated and real data case studies across various geoscience fields. The course focuses on the important issues of predictive learning: high dimensional and multivariate data visualization, predictability, data and models complexity, feature selection, uncertainty characterization, which are considered within a holistic framework of data driven modelling. We will demonstrate how domain geoscience knowledge adds value and improves efficiency of machine learning application. No prior knowledge of machine learning is required. Details of the course:

  • Introduction to machine learning for geo- and environmental data
  • High dimensional multivariate data visualization.
  • Basic machine learning algorithms for solving clustering, classification, and regression problems
  • Advanced ML topics and models: active learning, semi-supervised learning, ensemble learning, ML and uncertainties
  • Case studies: environmental pollution (air, water, soil), natural hazards, subsurface reservoir characterization, renewable energy
  • General discussion

Shale Analytics

  • organized by: Shahab D. Mohaghegh (Intelligent Solutions, Inc. & West Virginia University)
  • Venue:
  • Date: August 11, 2019
  • Time:
  • Min. number of participants: 10
  • Max. number of participants: 50
  • costs per participant: US$ 250

Data-driven analytics is becoming an important competitive differentiation in the upstream oil and gas industry. When it comes to production from unconventionals, specifically, shale, companies are realizing that they are in possession of large amount of facts and information in the form of the data they have been collecting in the past several years. It has been proven beyond any reasonable doubts that when it comes to analysis and modeling of production from shale, our traditional techniques (Numerical Simulation, RTA, and Decline Curve Analysis) leave much to be desired. Relevant and domain-based implementation of Artificial Intelligence and Machine Learning in analysis and modeling of the collected data via field measurements can provide much needed insight that would overcome the biases, preconceived notions, and overwhelming assumptions that have dominated our traditional techniques.

Data-driven analytics is the set of tools and techniques that provides the means for extraction of patterns and trends in data and construction of predictive models that can assist in decision-making and optimization. Shale Analytics is the domain (reservoir completion and production engineering) based application of the state of the art Artificial Intelligence and Machine Learning for production and recovery optimization from shale wells. Shale Analytics integrates all relevant data (well, reservoir, completion, frac job, placement and stacking, and operation) with production history in order to model the complex physics of hydrocarbon production from shale. As the number of wells in an asset increases, so does the accuracy and reliability of the analytics.

Attendees will become familiar with the fundamentals of data-driven analytics, Artificial Intelligence and Machine learning including the most popular techniques used to apply them such as artificial neural networks, evolutionary computing, and fuzzy set theory.

This course will demonstrate through actual case studies (and real field data from thousands of shale wells) how to impact well placement, completion, and operational decision-making based on field measurements rather than human biases and preconceived notions.

Topics:
  • Basics of artificial intelligence (AI) and machine learning
  • Descriptive analytics
  • Predictive analytics
  • Prescriptive analytics
  • Introduction to AI-based dynamic modeling
For more information click here.

Subsurface Fracture Characterization

  • organized by: Paul Lapointe
  • Venue:
  • Date: August 11, 2019
  • Time:
  • Min. number of participants: 10
  • Max. number of participants: 50
  • costs per participant: US$ 250

Sponsors

EMS Energy Insitute, Penn State University

John and Willie Leone Familty Department of Energy and Mineral Engineering

Department of Metereological Sciences at Penn State

Penn State Institute of Energy and the Environment