28th October 2021
Topic: Leveraging Supervised Machine Learning for Rapid Lithology Prediction
Speaker: Rebecca Head, Halliburton
Human beings are creating vast quantities of new data every year, and the oil and gas industry is no exception. These new data are adding to the vast legacy datasets already in existence. Machine learning provides an opportunity to interpret these data rapidly, and gain insights and value from them.
A supervised learning approach for rapid lithology assessment is presented, which applies the detailed expertise of petrophysicists to numerous wells in minutes. Prior and posterior probability calculations provide confidence information enabling the user to identify well sections that could benefit from further human interpretation.
This approach was applied to 113 wells from the Alaskan North Slope, yielding >70% lithological match to the human petrophysical interpretation. This is a good starting point, demonstrating the utility of this approach in unlocking the value of large datasets.
Rebecca provides technical geoscience support to potential and existing Halliburton customers. She specialises in the Halliburton Landmark suite of geoscience software and content, working with customers to help them better understand the subsurface. She has eight years of industry experience, having worked as a geoscientist in the research and development of Halliburton’s geoscience content portfolio, Neftex® Predictions. From 2018 to 2021, she was also managing editor of the Subsurface Insights magazine. Rebecca holds an MSci in Natural Science from the University of Cambridge.