A deep-learning-based surrogate model for data assimilation in dynamic subsurface flow problems (1908.05823v1)

Abstract: A deep-learning-based surrogate model is developed and applied for predicting dynamic subsurface flow in channelized geological models. The surrogate model is based on deep convolutional and recurrent neural network architectures, specifically a residual U-Net and a convolutional long short term memory recurrent network. Training samples entail global pressure and saturation maps, at a series of time steps, generated by simulating oil-water flow in many (1500 in our case) realizations of a 2D channelized system. After training, the `recurrent R-U-Net' surrogate model is shown to be capable of accurately predicting dynamic pressure and saturation maps and well rates (e.g., time-varying oil and water rates at production wells) for new geological realizations. Assessments demonstrating high surrogate-model accuracy are presented for an individual geological realization and for an ensemble of 500 test geomodels. The surrogate model is then used for the challenging problem of data assimilation (history matching) in a channelized system. For this study, posterior reservoir models are generated using the randomized maximum likelihood method, with the permeability field represented using the recently developed CNN-PCA parameterization. The flow responses required during the data assimilation procedure are provided by the recurrent R-U-Net. The overall approach is shown to lead to substantial reduction in prediction uncertainty. High-fidelity numerical simulation results for the posterior geomodels (generated by the surrogate-based data assimilation procedure) are shown to be in essential agreement with the recurrent R-U-Net predictions. The accuracy and dramatic speedup provided by the surrogate model suggest that it may eventually enable the application of more formal posterior sampling methods in realistic problems.

Analysis of the Empty Paper Structure

This document appears to be a preprint manuscript formatted for submission to the Nuclear Physics B journal, utilizing the elsarticle class likely provided by Elsevier. Beyond the basic structure, such as sections for title, author, abstract, keywords, and bibliography, the document remains unpopulated and lacks substantive content to analyze. The presence of these structural components suggests an intention to present research findings in a conventional academic format, yet they are devoid of scientific content, methods, results, or discussions.

Insights and Possible Interpretation

1. Formatting and Structure: The document exhibits standard formatting expectations for an academic paper submitted to a reputable journal. This includes designated spaces for essential components such as the abstract, keywords, main text, and references. Such a format ensures that the eventual contents will be accessible, standardized, and prepared for peer review.
2. Potential for Content Development: The absence of content invites speculation on possible topics within the field of nuclear physics, such as theoretical advancements, experimental investigations, or computational simulations. The utilization of a thorough journal structure implies an adherence to scientific rigor, hypothetically guiding future contributions in ensuring clarity and reproducibility in research within the discipline.

Implications for Future Work

Given the current empty state of the document, any potential future work would need to encompass the following components:

• Abstract and Introduction: A concise summary of the research objectives, methods, and chief findings is crucial. It would set the stage for the rest of the paper, offering a snapshot of the research conducted.
• Methodologies and Experiments: Detailed explication of the experimental or computational methodologies employed provides necessary transparency for reproducibility, an integral aspect of scientific inquiry.
• Results and Discussions: Clear presentation of the results, supported by data or theoretical analysis, offers a foundation for critical discussion on the implications and relevance of findings within the broader scientific community. The discussion should ideally connect results back to the theoretical framework or to practical applications in nuclear physics.
• Conclusion and Future Directions: Summarizing the key contributions of the work and suggesting avenues for further research enhances the relevance and impact of the paper.
• Robust Referencing: Citing relevant literature places the paper within the ongoing scientific discourse and acknowledges prior work that underpins the research.

Speculation on Future Developments

The paper's potential completion could contribute to various theoretical or practical areas within nuclear physics. As computational power and experimental techniques continue to evolve, future research might benefit from this document's format by ensuring well-structured dissemination of novel findings. Furthermore, advancements in adjacent fields such as quantum computing or material science could intersect with nuclear physics research, prompting interdisciplinary studies reflected in future iterations of a paper starting from this template.

The outlined structure is primed for content that could progress theoretical physics, propose new models, or validate existing theories through innovative experiments. With ongoing developments in technology and computational methods, this format will support comprehensive and impactful communication of research findings in nuclear physics.