Oil and Gas
Oil and Gas | Data Analytics
Business Impact: In order to increase production and drive down the cost per barrel of oil equivalent (BOE), this course will enable participants to build reliable earth models of unconventional reservoirs using analytics for data insight and geostatistics for assessing uncertainty.
This class addresses the application and integration of data analytics to subsurface geomodeling for unconventional resources, including oil, gas, and geothermal. Deterministic and stochastic methods used to create static models and uncertainty assessment will be reviewed to establish a common knowledge baseline. This is followed by skill development in data analysis methods such as multivariate statistics and machine learning. Topics include kriging, conditional simulation, principal components, cluster analysis, regression, recursive portioning, neural networks, and other practical methods. The class focus is to go beyond traditional static modeling through the integration of geostatistics and data science to produce reliable models for reservoir and completion engineers.
Participants will learn how to create mappable quality indices to optimize successful well placement and formation stimulation strategies. This course pulls together geoscience, engineering, and data science to build synergistic teams, optimize successful drilling programs, reduce uncertainty, and drive down cost per barrel of oil equivalent.
Duration and Training Method
This is a classroom or virtual classroom course comprising a mixture of lectures, discussion, case studies, and practical exercises.
Participants will learn to:
- How to apply data analytics and machine learning for data QC.
- Assess critical relationships between petrophysical, geochemical, geothermal, and mechanical data using multivariate analysis
- Evaluate workflows for geocellular modeling of unconventional Reservoirs.
- Tell the difference between interpolation, simulation, imputation, and prediction.
- How to properly integrate data science and geostatistical static modeling.
- Develop spatial models using quality indices from integrated geomechanical, geochemical, geothermal, and petrophysical properties.
- Assess the uncertainty of geomodels of unconventional reservoirs.
- Post process static geomodels al reservoirs for dynamic flow simulation.
- Provide valuable model information to well completion and stimulation engineers.
- Understand the value of C.I.P. languages (e.g. R and Python) in geomodeling.
This course will address questions about using geocellular models for unconventional reservoirs such as:
- The difference between Conventional and Unconventional Reservoirs
- Are common modeling tools such as kriging and conditional simulation appropriate?
- Can Machine Learning be used to create geomodels?
- I’m not a developer, how can I use R and Python to help me build models?
- Is it necessary to model “shale facies” or “geothermal facies” and if so, how do we define them?
- What if my data are sparse and there is no seismic?
- How do I deal with missing values when I am building a geomodel?
- Geocellular grids or meshes; what dimensions, layering types, and granularity are required?
- How to integrate natural fractures - key considerations
- What is the role of micro-seismic data and how can it be used in a geocellular model?
- How can geocellular models be used to improve geosteering and real-time drilling?
- What are the limitations to geocellular grids and stratigraphic layering styles?
- How do geocellular grids of unconventional reservoirs respond in dynamic simulators?
- How can Completion Engineers benefit from geomodels?
The specific topics to be addressed are:
- Role of 3D geocellular models for unconventional resources
- Considerations for constructing geocellular grids or meshes for unconventional resources
- Exploratory data analytics for unconventional resources
- Types of variables
- Univariate and bivariate statistics for data quality assessment
- Review of principles of spatial modeling
- Review of Kriging and Conditional Simulation
- Review of geomodeling steps and the importance of frameworks and stratigraphy
- Workflows for shale plays and how they differ from conventional plays
- Understanding multivariate analysis techniques and machine learning
- Applications in R and Python
- Post processing - Building probability maps to identify “Sweet Spots” and improve well placement
Topic 3 - Modeling Workshop
This workshop uses simple exercises on a variety of multivariate techniques in order for participants to better understand the underlying principles. Some of the examples are from unconventional shale reservoirs and some are not. Those that are not, are classic data sets used in texts, classrooms, seminars, and online short courses and are not necessarily geologic in nature, but clearly illustrate the methods.
Practical computer-based exercises demonstrating Multivariate Data Analytics that form the basis of the workshop include the following:
- A brief introduction to R and R Studio
- Exploratory Data Analytics
- Spatial Modeling, Kriging and Conditional Simulation
- Analytics for data reduction and data redundancy
- Principal Component Analysis and Factor Analysis
- Classification Systems and Regression Systems
- Discriminant Analysis
The actual number of exercises and methods will vary depending on available time.
Who Should Attend and Prerequisites
This course has been designed for mid to senior level geoscientists (specifically geomodelers), and data scientists who are working with geomodelers. Familiarity with geostatistics and practical geocellular modeling is assumed. Managers and others who have previous experience in building geomodels and wish to develop a better understanding of how to apply geomodeling and data science techniques to unconventional reservoirs could also benefit from this course.
Dr. Yarus has 45 years’ experience in quantitative statistical analysis of geoscience data and computational earth modeling. Noted expert in the applied field of subsurface reservoir characterization and spatial statistics, Dr. Yarus is well known for his various publications and lectures on applied geostatistics which are given worldwide. In 2016, he received the distinguished George Matheron Lecture Award from the International Association for Mathematical Geosciences. In 1994 he published the first book on principles, methods and case studies in geostatistics for the American Association of Petroleum Geologists entitled, “Stochastic Modeling and Geostatistics.” This was followed by 2 additional edited volumes on graphic information systems and geostatistical applications in 2000, and 2006, respectively.
In 2007, Dr. Yarus was invited to contribute the first-ever dedicated chapter on “Geologically Based Geostatistics” to the Society of Petroleum Engineering’s prestigious “Petroleum Engineering Handbook.” He has worked for both Major Oil and Service Companies including Amoco, Marathon, Beicip-Franlab, and ROXAR, where he cultivated experience covering production, exploration, consulting, and software development. His experiences also included initiating and managing his own consulting company, Quantitative Geosciences (QGSI) which provided consulting and research services across many disciplines including Petroleum, Environmental, Water Resources, and Climatology. QGSI was acquired Halliburton in 2006 where Dr. Yarus was asked to manage a team of professional geoscientists, engineers, and software developers to build the successful, “DecisionSpace Earth Modeling” software. In 2013, Dr. Yarus became a Senior Technology Fellow at Halliburton where he reported to the executive staff.
Throughout his career, he has maintained strong university relationships working as an adjunct professor at the University of Colorado, Colorado School of Mines, and the University of Houston. His university responsibilities included lecturing at the graduate level and advising graduate students on their theses. Dr. Yarus’ university connections include University of Texas, Bureau of Economic Geology, Texas A&M, Stanford University, University of Utah, Tulsa University, University of Alberta, Norwegian Computing Center, University of Neuchâtel, Imperial College, and the Paris School of Mines (Mines ParisTech). He retired from Halliburton at the end of 2018 and is currently a full research professor at Case Western Reserve University engaged in research and teaching geospatio-temporal modeling of near surface, surface and subsurface phenomena, data science, and informatics systems.
Affiliations and Accreditation
PhD University of South Carolina
N058: Reservoir Characterization and Geostatistical Modeling in Field Development
N345: Geomodeling for Unconventional Reservoirs
N557: Geostatistics and Data Science; a Practical Introduction to Quantitative Spatial Modeling