Automated Rock Classification Using High-Resolution CT-Scan Images and Core Photos in a Siliciclastic Formation Used for CO2 Storage

Abstract

Abstract Carbon capture and storage (CCS) is an attractive alternative to reduce the concentration of greenhouse gases in the atmosphere with the objective of preventing further increases in global temperature. Accurate estimation of Petrophysical properties and detection of rock types are critical for the assessment of key aspects of CCS projects in geological formations such as storage capacity, injectivity, trapping mechanisms, and containment. The objectives of this paper are (a) to use whole-core computed tomography (CT) scan images and core photos, conventional well logs, nuclear magnetic resonance (NMR) logs, and core-measured properties for automated rock classification, (b) to develop class-based rock physics models for enhanced petrophysical properties estimation, and (c) to provide a method to expedite the detection of quantitative image-based rock classes. First, we conducted conventional formation evaluation for initial assessment of petrophysical properties. Then, we employed image analysis techniques to remove non-core material visual elements from the available image data (i.e., CT-scan images and core photos). Afterwards, we extracted rock-fabric related features from the available image data. We characterized the pore structure of the evaluated interval using NMR logs. We integrated conventional well logs and routine core analysis (RCA) data with image-based features and NMR pore structure parameters to automatically detected rock classes by means of a physics-based cost function. Finally, we updated the estimated petrophysical properties employing class-based rock physics models and compared the obtained result against conventional formation evaluation estimates. We applied the proposed workflow to the pilot well drilled in a saline water aquifer formation that will be used for CO2 injection and storage in the Northern Lights CCS project. The extracted image-based rock fabric features were in agreement with the visual aspect of the evaluated depth intervals. The detected rock classes captured the fluid-flow behavior using a permeability-based cost function, the variation in petrophysical and compositional properties trough well logs, and quantitative rock fabric of the evaluated depth interval through the core image data. Finally, the use of class-based rock physics models improved permeability estimates decreasing the mean relative error by 27% compared to formation-based permeability estimates from a conventional method (formation-based porosity-permeability correlations). One of the key contributions of the proposed workflow is that it integrates conventional well logs, core-measured properties, NMR logs, and high-resolution image data. As a result, the obtained integrated rock classes capture key petrophysical and geological parameters of the evaluated depth intervals that are typically not included in rock classification efforts. The obtained integrated rock classes can potentially improve the development of accurate geological models, which are employed in simulation efforts as a screening tool for selection of geological formations for CO2 storage as well as for storage capacity, selection of CO2 injection intervals, and containment forecasting.