Multiscale, image-based interpretation of well logs acquired in a complex, deepwater carbonate reservoir

Carbonate formations hold a large percentage of the world's hydrocarbon reserves. Their petrophysical evaluation via well logs and core data faces significant technical challenges because of the coexistence of multiscale pore features affecting fluid-transport phenomena. The combined effects of diagenesis, solid dissolution, and recrystallization in carbonate rocks give rise to pores ranging from centimeter-size vug openings to submicron microporosity. In turn, the wide variability of pore sizes, pore shapes, pore textures, and pore connectivity gives rise to multiple length-dependent flow regimes. % New data sources have allowed many porous-media processes to be observed or numerically simulated in detail for the first time; they have also helped to understand flow mechanisms that have a direct impact on hydrocarbon recovery. Spatial imaging techniques such as computed tomography (CT) are now applied routinely to acquire 3D images of both laboratory samples and whole core. Those images reveal fine features of rock structure, pore topology, and mineral spatial distribution, in addition to enabling the numerical simulation of several physical phenomena taking place inside the pore space. Methods used to analyze rock properties based on 2D and 3D digital images are collectively known as digital rock petrophysics, and are being used at an accelerated pace to quantify the storage and production potential of spatially complex rocks.

This dissertation introduces new quantitative methods for the analysis of whole-core CT images to improve the interpretation of well logs acquired in carbonate formations. % The first method focuses on the estimation of density and atomic number from dual-energy CT core scans. A new Monte Carlo-based inversion algorithm for estimating such properties is developed to account for uncertainties in X-ray attenuation coefficients in addition to delivering uncertainty estimates of inversion products. Estimation of electron density and effective atomic number from CT core scans enables direct deterministic or statistical correlations with salient rock properties for improved petrophysical evaluation. Verification tests of the inversion method performed with CT-generated density and PEF logs yield very good agreement with borehole measurements of density and photoelectric factor.

Next, a new workflow is introduced for image segmentation and interpretation. Reliable classification of image voxels in components representing grains, pores, and sub-resolution features remains challenging in images with multiscale features such as those of carbonate whole cores. The new workflow reduces statistical bias introduced by interpreter subjectivity, and allows automation for the analysis for a large number of samples.

Segmentation of vug space in CT images also enables close inspection and reliable interpretation of well logs in vuggy reservoir regions. Connected vugs are expected to exhibit high and dominant fluid production capacity, whereby the ability to properly identify such reservoir zones via well logs is very important. Ultrasonic borehole images have been extensively used to assess rock texture and multiple geometrical and sedimentary features. Comparison of ultrasonic borehole images to CT data confirms specific well-log responses across vuggy depth segments. New feature-enhancing methods are introduced for the interpretation of ultrasonic borehole images. However, no strong correlation was found when attempting to quantify vuggy porosity from various image attributes. % The segmentation of CT images across vuggy space is also explored for estimating vug flow properties. A statistical description of segmented vuggy space is suggested to estimate permeability given the relatively low image resolution of the available CT data. Results confirm the hypothesis that connected vugs dominate fluid flow, whereby the assessment of vug geometrical properties provides sufficient information for estimating permeability.

Finally, a new method is introduced for the interpretation of nuclear magnetic resonance (NMR) logs. Adverse borehole conditions such as mud-filtrate invasion and large washouts in vuggy zones are usually neglected in conventional interpretation procedures of NMR logs. To circumvent the latter problem, we describe the measured distribution of transverse relaxation times as the superposition of a finite set of log normal components where each component accounts for specific relaxation rates for drilling mud and original formation fluids. Estimated permeabilities in vuggy zones from NMR logs with the new method are more accurate than those rendered by conventional techniques based on cutoff values or logarithmic averages. The method also explicitly quantifies vuggy porosity, which is found to be in good agreement with values obtained from segmented CT data.

The combined use of the above interpretation methods confirms the value of digital rock techniques to improve the interpretation of well logs acquired in complex carbonate formations, specifically in the calculation of permeability across vuggy depth segments. Results can be used to improve the interpretation of well logs acquired in wells devoid of core data and/or high-resolution borehole images.