2023 PESA WEBINAR SERIES: Machine Learning and Intelligent Augmentation Applied to Legacy Subsurface Data

Recent emphasis on low carbon energy projects in Australia  means large volumes of unstructured subsurface data become valuable to operators. . The need to  evaluate reservoirs for energy or carbon storage or light gas extraction means a requirement to de-risk geographic search and  make machine-readable text available for advanced data science algorithms. Companies with access to data gain a competitive advantage in licensing activities. Exisiting data management workflows may be too costly and resource intensive to support these emerging requirements.

We highlight how a  workflow using Machine Learning (ML) and Intelligence Augmentation (IA) can improve the efficiency of a typical data management workflow. Our example involves manually identifying relevant well logs in images from collection of tens of thousands of file types, a typical task for an operator using open-file data in Australia. The current process may involve downloading from multiple state and federal repositories and manually entering indexing metadata into internal data repositories.

We have developed internal ML/AI data extraction workflows that that streamline automatically separating well log image files, extracting mandatory indexing meta-data, and presenting model-trained choices to end-users for adjustment and verification. The full solution involves image classification, enhanced Optical Character Recognition(OCR), Natural Language Processing (NLP) for keyword labels and values, and a User Interface (UI) that delivers and documents increased data quality and confidence metrics.