AI across Reservoir Modelling Workflows - a Hands-on Introduction

Instructor: Prof. Vasily Demyanov
Language: English 
Level: Intermediate
Duration and formats:
Online:  2 half days of 3-4 hours CPD Points: 8
Classroom: 2 days CPD Points: 10

Course Description

The course offers a hands-on practical introduction to Python AI reservoir modelling workflows for those familiar with reservoir data and eager to master the advantages of contemporary Python capabilities in gaining additional value from reservoir data with machine learning. We will take you through a number of common reservoir characterisation, modelling and monitoring tasks using open-access reservoir data.

Participants will consolidate the basic conceptual understanding of machine learning principles from an illustrative overview in the reservoir modelling context. Participants will also gain an experiential learning knowledge of AI applications to reservoir modelling workflows across geoscience and engineering tasks. The take-home learning will include examples of AI applications to basic reservoir data: wireline, seismic, and production (PTA).

The course can be adjusted for delivery options that vary in how many tasks of the reservoir modelling workflow are covered:

  • 1 day: a brief introduction to ML concepts with 1-2 hands-on Python notebook examples 

  • 2 days: a more concise introduction to ML with 2-3 hands-on Python notebook practical tasks and illustrative examples;

  • 3 days: a comprehensive introduction to ML with 3-4 hands-on Python notebook practical tasks and illustrative examples of a variety of ML algorithms.

  • Up to 5 days: a comprehensive introduction to ML with an illustrative overview of various AI applications across the reservoir modelling workflows - from geology to engineering, with 4-5 hands-on practical Python notebook tasks and paper exercises.

The course is delivered by Prof V. Demyanov and Ms F. Rabie.

Course Outline

The course will focus on several specific AI applications to reservoir data that demonstrate how AI can provide added value to conventional modelling workflows:

Participants will gain practice with pre-composed Python Jupyter-notebooks supplied for visualisation, modelling and analysis of reservoir data from a real field. The course includes a concise introduction of relevant machine learning principles and basic algorithms accompanied by illustrative examples of AI applications to a range of reservoir modelling datasets with different types of reservoir data. The learnings are reinforced through in-class hands-on practice with Python notebooks.

Course topics include:

  • Introduction to learning from data key concepts, good practice and pitfalls.

  • Over-learning and model complexity.

  • Unsupervised and supervised learning principles and algorithmics.

  • Deep learning and generative AI for reservoir applications.

  • Reservoir data visualisation and manipulation with Python.

  • Facies identification and classification from wireline data.

  • Seismic segmentation and object detection with unsupervised learning for seismic interpretation.

  • Pattern recognition for well dynamics performance in Pressure Transient Analysis data (optional).

  • How Machine Learning can account for domain knowledge in learning from data.

  • Physics-based learning (optional)

Participants’ Profile

This course is for anyone who is interested in getting familiar with running machine learning applications and data analytics for subsurface data. Domain knowledge of subsurface data, such as petrophysics, facies, seismic, reservoir engineering and production would be beneficial.

Prerequisites

The course does not imply any prior knowledge of machine learning and will include a basic introduction of algorithms. No prior Python coding experience is required – all template codes will be provided. Basic understanding of scripting language (like Python) would come handy. The participants will be expected to use their own laptops for the course exercises with Python Jupiter notebooks following the instructions provided.

Basic understanding of reservoir data: wireline, seismic, well pressure transients would be essential.

About the Instructor

Prof. Vasily Demyanov, leads GeoDataScience and the UQ group at the Institute of Geoenergy Engineering, Heriot-Watt University. Prof Demyanov has 30 years of academic experience in geostatistics and machine learning in geoscience applications. He has been running the EAGE EET course, Challenges and Solutions in Stochastic Reservoir Modelling.