Rock Physics Driven Machine Learning for Reservoir Characterization with GeoAI
OCTOBER 24 - 25, 2024.
North America (United States and Canada)
Learning Objectives:
GeoAI is a HampsonRussell product, encompassing a novel seismic reservoir characterization technology for projects with limited well control. Use Rock Physics theory and statistical simulations to model various geological scenarios. Train convolutional neural networks (CNNs) using simulated synthetic data and transfer learning to apply to real seismic gathers, estimating multiple rock and elastic reservoir properties in a simplified machine learning approach.
This course covers all aspects of GeoAI, starting with an introduction to the practical use of deep neural networks (DNNs) to predict elastic and rock properties, as well as an overview of relevant machine learning theory. In this supervised learning workflow, the relationships mapping the pre-stack seismic to the properties of interest are learned from the data itself. Key to deriving robust operators are big data sets of training data. GeoAI allows the generation and incorporation of synthetic data in the machine learning workflow and utilization of the synthetic data in the CNNs to get more reliable volume estimates.
Content:
- Supervised learning, multi-linear regression, and deep neural networks.
- Introduction to deep neural networks (DNNs) and the need for big datasets.
- Creating big data by generating synthetic wells and synthetic seismic gathers.
- Examples showing how the inclusion of synthetic data improves the property estimates.
- Lithofacies Classification, Rock Physics Modeling, Statistics, Variogram Modeling, and Well Simulations.
- Seismic data preparation, including correlation, generation of angle gathers, and scaling.
- CNN training and transfer learning.
- CNN application for both regression and classification studies.
- Quality control of the results. Benefits of GeoAI methodology and practical applications.
Duration: 2 days (all day)
Software used: HampsonRussell
Course Format: Instructor-led, workflow-based, virtual training.
Prerequisites: None
Who should attend: Geophysicists, geologists, engineers, and technical staff who would like to understand the theory behind essential Seismic Reservoir Characterization Workflow.
Price*: $2,000 USD per attendee
*Virtual classroom capacity is a minimum of 4 and a maximum of 10 attendees for this course; group discount may apply