Streamlining petrophysical workflows with machine learning
Lucy MacGregor1, Juan Berrizbeitia1, Nick Brown2, Anna Roubickova2 & Marc Sabate2
1 – Rock Solid Images 2 – EPCC, University of Edinburgh
The oil and gas industry is not short of data, in the form of wells, seismic and other geophysical information. However, the industry is notoriously poor at utilizing this information. The complexity of workflows required to take raw information that is available in public or proprietary data stores and turn this into decision ready information on sub-surface geology and properties means that such workflows are time consuming, so that often only a fraction of available information is used. Making better use of information, using modern data analytics techniques, and presenting this information in a way that is immediately useful to geologists and decision makers has the potential to dramatically reduce time to decision and the quality of the decision that is made.
In this presentation we concentrate on one aspect of the general problem described above: using machine learning approaches to streamline petrophysical workflows. In order for a well log to be used in a quantitative interpretation workflow, a full petrophysical interpretation is required, followed by careful rock physics analysis. Erroneous data must be identified and corrected, and missing information estimated on the basis of the data available. This can be a time-consuming process.
Machine learning approaches have the potential to dramatically streamline such workflows. However, to do this requires a rich and diverse training dataset of wells that have been consistently processed for geophysical analysis. This project makes use of RSI’s global rock physics database as the training dataset. The work discussed in this paper has focused on the estimation of clay volume, determination of mineral volumes and determination of porosity and water saturation. A variety of machine learning techniques and algorithms have been tested to find the one most suited to this application. Initial analysis is regionally focused, but we plan to investigate whether the approaches and models developed can be generalized across regions, basins and geological settings.