Book Description
Development of both conventional and unconventional energy resources remains a significant component in the US and world energy portfolio. From well log data at the reservoir level to nanoscale characterization, each step informs our understanding of the petrophysical behavior of the system. Due to the multiscale heterogeneity in shale reservoirs especially, characterization at various scales is necessary to capture the full physical behavior. Electron and x-ray microscopy techniques are commonly used to image and characterize the microstructure and composition of porous media. Including information about the microstructure and pore connectivity of the inorganic and organic matrix improves estimations of hydrocarbon production and storage capacity in shale reservoirs. In order to capture the effect of multiscale features on bulk properties, a combination of experimental characterization techniques must be employed. The main objective of this research is to develop a multiscale method to describe porosity, connectivity, and chemistry in complex, microporous systems. We first examine a selection of experimental techniques used to characterize complex microporous media. We combine direct measurements of the pore space with image analysis methods to create a multiscale understanding of a Texas cream carbonate and a Vaca Muerta shale sample. We show how these methods are used to obtain a more accurate representation of the pore network for systems with microporosity or features that are difficult to segment. We propose a series of methods that are used together to improve our understanding across multiple length scales. Next, we apply this workflow to study how kerogen evolution and migration during an artificial maturation process is tied to changes in the shale microstructure. We observe an increase in porosity across the shale surface, coupled with a decrease in organic matter-rich regions. This result, in addition to SEM imaging of microcrack development along intraparticle and organic-rich areas, provides important insight into the possible pathways for kerogen to escape the sample. We then investigate how imaging parameters, specifically resolution, affect the petrophysical properties that are calculated directly from the imaged pore networks of a sandstone sample. We present a workflow to downsample a high-resolution image dataset, segment the pore network, and calculate the single-phase permeability using a direct numerical method. We show directly how changes in imaging resolution affects uncertainty in the porosity and permeability of a sample. Finally, given the imaging intensive work performed above, we investigate the potential of using recently developed deep learning based methods for generating realistic pore volumes. These pore volumes are segmented and used for further petrophysical analysis without requiring additional imaging or sampling of the actual reservoir. We develop a generative flow network and apply the model to create 2D and 3D representations of the sandstone pore network with morphological and petrophysical properties that mimic those of real rock. We illustrate the advantages of using such a model for its rapid generation capability and scalability during training.
