Development of lean, efficient, and fast physics-framed deep-learning-based proxy models for subsurface carbon storage. (February 2022)

• Record Type: Journal Article
• Authors: Omosebi, Omotayo A.; Oldenburg, Curtis M.; Reagan, Matthew
• Is Part Of: International journal of greenhouse gas control. Volume 114(2021)
• Publication Date: 2022-02
• Subjects: Fast proxy model -- Deep-learning -- Machine-learning -- Physics-guided -- Carbon storage -- Carbon sequestration; Greenhouse gases -- Environmental aspects -- Periodicals; Air -- Purification -- Technological innovations -- Periodicals; Gaz à effet de serre -- Périodiques; Gaz à effet de serre -- Réduction -- Périodiques; Air -- Purification -- Technological innovations; Greenhouse gases -- Environmental aspects; Periodicals; 363.73874605
• DOI: 10.1016/j.ijggc.2021.103562
• Languages: English
• ISSNs: 1750-5836

Abstract

Highlights: An approach is proposed for physics-based formulation of deep-learning problems relevant to CO2 injection. Deep-learning-based surrogate models were developed using the proposed methodology and are shown to be capable of substituting as computationally efficient surrogates for traditional numerical simulators during history matching. Reservoir pressure, CO2 saturation plume, and water extraction rate can be rapidly predicted given formation properties, reservoir mesh, well operating conditions, and initial and boundary conditions. ABSTRACT: We present deep-learning-based surrogate models for CCUS developed with four different algorithms and a physics-framed two-phase flow problem involving displacement of water by CO2 . The deep-learning models were trained using 3D datasets describing the pressure plume, CO2 saturation plume, and water extraction rate generated by numerical simulation. The hyperparameters defining the architecture of the neural networks were optimized to determine the slimmest network size and training parameters that give the most efficient performance at the least training cost. To develop a robust model that closely mimics the governing physical laws, the discretized form of the two-phase fluid transport equation was used to formulate the supervised deep-learning task. The algorithms investigated in this study predicted the data to above 95% accuracy, with the multi-layer perceptron model demonstrating the best performance by balancing training speed, prediction time, and prediction accuracy with lean network capacity. Furthermore, the surrogate models simultaneously predict reservoir pressure and CO2 saturation in every grid block, including the surface well extraction rate and bottomhole pressure, at all simulation times for a given static model realization in just a few seconds on a standard desktop computer. A key outcome of this study is that limits can be placed on network design parameters to avoid over designing neural networks, with associated efficiencies in training and prediction times. This is very useful because large volumes of data may be generated in CCUS projects and over-design of neural network architectures imposes penalties that are antithetical to the goal of near-real time forecasting.