Hierarchical automated machine learning (AutoML) for advanced unconventional reservoir characterization.

Recent advances in machine learning (ML) have transformed the landscape of energy exploration, including hydrocarbon, CO2 storage, and hydrogen. However, building competent ML models for reservoir characterization necessitates specific in-depth knowledge in order to fine-tune the models and achieve the best predictions, limiting the accessibility of machine learning in geosciences. To mitigate this issue, we implemented the recently emerged automated machine learning (AutoML) approach to perform an algorithm search for conducting an unconventional reservoir characterization with a more optimized and accessible workflow than traditional ML approaches. In this study, over 1000 wells from Alberta's Athabasca Oil Sands were analyzed to predict various key reservoir properties such as lithofacies, porosity, volume of shale, and bitumen mass percentage. Our proposed workflow consists of two stages of AutoML predictions, including (1) the first stage focuses on predicting the volume of shale and porosity by using conventional well log data, and (2) the second stage combines the predicted outputs with well log data to predict the lithofacies and bitumen percentage. The findings show that out of the ten different models tested for predicting the porosity (78% in accuracy), the volume of shale (80.5%), bitumen percentage (67.3%), and lithofacies classification (98%), distributed random forest, and gradient boosting machine emerged as the best models. When compared to the manually fine-tuned conventional machine learning algorithms, the AutoML-based algorithms provide a notable improvement on reservoir property predictions, with higher weighted average f1-scores of up to 15-20% in the classification problem and 5-10% in the adjusted-R2 score for the regression problems in the blind test dataset, and it is achieved only after ~ 400 s of training and testing processes. In addition, from the feature ranking extraction technique, there is a good agreement with domain experts regarding the most significant input parameters in each prediction. Therefore, it is evidence that the AutoML workflow has proven powerful in performing advanced petrophysical analysis and reservoir characterization with minimal time and human intervention, allowing more accessibility to domain experts while maintaining the model's explainability. Integration of AutoML and subject matter experts could advance artificial intelligence technology implementation in optimizing data-driven energy geosciences.

On the generation of realistic synthetic petrographic datasets using a style-based GAN.

Deep learning architectures have transformed data analytics in geosciences, complementing traditional approaches to geological problems. Although deep learning applications in geosciences show encouraging signs, their potential remains untapped due to limited data availability and the required in-depth knowledge to provide a high-quality labeled dataset. We approached these issues by developing a novel style-based deep generative adversarial network (GAN) model, PetroGAN, to create the first realistic synthetic petrographic datasets across different rock types. PetroGAN adopts the architecture of StyleGAN2 with adaptive discriminator augmentation (ADA) to allow robust replication of statistical and esthetical characteristics and improve the internal variance of petrographic data. In this study, the training dataset consists of > 10,000 thin section images both under plane- and cross-polarized lights. Here, using our proposed novel approach, the model reached a state-of-the-art Fréchet Inception Distance (FID) score of 12.49 for petrographic images. We further observed that the FID values vary with lithology type and image resolution. The generated images were validated through a survey where the participants have various backgrounds and level of expertise in geosciences. The survey established that even a subject matter expert observed the generated images were indistinguishable from real images. This study highlights that GANs are a powerful method for generating realistic synthetic data in geosciences. Moreover, they are a future tool for image self-labeling, reducing the effort in producing big, high-quality labeled geoscience datasets. Furthermore, our study shows that PetroGAN can be applied to other geoscience datasets, opening new research horizons in the application of deep learning to various fields in geosciences, particularly with the presence of limited datasets.

Coevolution of diagenetic fronts and fluid-fracture pathways.

Diagenetic boundaries are paleo-reaction fronts, which have the potential to archive the termination of metasomatic processes in sedimentary rocks. They have not been extensively studied, perhaps because they appear simple in outcrop. Recent work has demonstrated the significance of paleo-reaction fronts to decipher multiphase recrystallization processes and provide high porosity zones. This paper provides a detailed documentation of reaction front evolution in a tectonically active salt basin and reveals a high level of complexity, associated with multiple fluid flow and tectonic events. Here, consistent patterns of increasing dolomite stoichiometry and ordering, along with a change from seawater-derived, fabric-retentive dolomite to fracture-controlled, fabric-destructive hydrothermal dolomite are observed vertically across the stratabound dolomite bodies. These patterns, coupled with a decrease in porosity, increase in ∆47 temperature and δ18Owater values indicate multiphase recrystallization and stabilization by warm, Mg-rich fluids. The stratabound dolomite bodies apparently terminated at a fracture-bound contact, but the presence of dolomite fragments within the fracture corridor suggests that fracturing post-dated the first dolomitization event. The termination of dolomite formation is therefore interpreted to be associated with a decrease in the capacity of the magnesium-rich fluids to dolomitize the rock, as indicated by the presence of non-stoichiometric and poorly ordered dolomite at the reaction fronts. The fracture corridors are interpreted to exploit dolostone-limestone boundaries, forming prior to a later, higher temperature, hydrothermal dolomitization event, which coincided with the formation and growth of the anticline. Karstification subsequently exploited these fracture corridors, widening fractures and leading to localized collapse and brecciation. The results demonstrate that an apparently simple reaction front can have a complex history, governed by the inheritance of prior diagenetic events. These events modified rock properties in such a way that fluid flow was repeatedly focused along the original dolomite-limestone boundary, overprinting much of its original signature. These findings have implications to the prediction of structurally controlled diagenetic processes and the exploration of naturally fractured carbonate reservoirs for energy exploration globally.

Evidence for the early Toarcian Carbon Isotope Excursion (T-CIE) from the shallow marine siliciclastic red beds of Arabia.

The Toarcian Oceanic Anoxic Event (T-OAE) and its corresponding Carbon Isotope Excursion (CIE) have been reported widely across the Tethyan region and globally. In Arabia, and based on ammonite dating, the time window of the T-OAE coincided with the deposition of the reddish siliciclastic unit of the Marrat Formation. However, no evidence of the T-OAE/CIE was ever reported from Arabia because these red beds were previously interpreted as continental deposits. Recently, these red beds have been recognized as shallow marine deposits which opened an opportunity to assess the occurrence and expression of T-OAE-CIE in Arabia. In this study, a multiproxy geochemical characterization was performed on the Toarcian Marrat Formation to infer the chemistry of the paleowater column and identify intervals of possible T-OAE/CIE in Arabia. While the low concentrations of redox-sensitive elements (Mo, U, V, Cr) may indicate a shallow oxic marine settings, the coupled negative δ13Corganic excursion and apparent increase in the chemical weathering suggests that the deposition of Marrat red beds coincided with the development of T-CIE and possibly time-equivalent to the T-OAE globally. The origin of reddening is interpreted to have occurred during the middle Marrat deposition due to the stabilization of unstable hydrous iron oxides to hematite under oxic marine conditions. The proposed model further indicates the possible development of source rocks in the deep, anoxic environment counterpart where the T-OAE may be expressed. Since our study documents the first record of the T-CIE and discuss the origin of shallow marine siliciclastic red beds in the Arabian Plate, this will have significant implications for the overall understanding of the T-CIE globally and for hydrocarbon exploration through realizations of potential new source rocks associated with the OAEs in the Toarcian and other time intervals.