Abstract:Simulating Darcy flows in porous media is fundamental to understand the future flow behavior of fluids in hydrocarbon and carbon storage reservoirs. Geological models of reservoirs are often associated with high uncertainly leading to many numerical simulations for history matching and production optimization. Machine learning models trained with simulation data can provide a faster alternative to traditional simulators. In this paper we present a single Fourier Neural Operator (FNO) surrogate that outperforms traditional reservoir simulators by the ability to predict pressures and saturations on varying permeability fields, well locations, well controls, and number of wells. The maximum-mean relative error of 95\% of pressure and saturation predictions is less than 5\%. This is achieved by employing a simple yet very effective data augmentation technique that reduces the dataset size by 75\% and reduces overfitting. Also, constructing the input tensor in a binary fashion enables predictions on unseen well locations, well controls, and number of wells. Such model can accelerate history matching and reservoir characterization procedures by several orders of magnitude. The ability to predict on new well locations, well controls, and number of wells enables highly efficient reservoir management and optimization.

What problem does this paper attempt to address?

### What problem does this paper attempt to solve? This paper aims to solve the problem of simulating Darcy flow in porous media, especially predicting future fluid - flow behavior in oil - gas fields and carbon storage reservoirs. Specifically, the paper proposes a surrogate model based on the Fourier Neural Operator (FNO) to replace traditional reservoir simulators. This surrogate model can handle varying permeability fields, well locations, well - control settings and the number of wells, thus achieving faster prediction speed and high accuracy in predicting pressure and saturation. ### Main research background and objectives 1. **Limitations of traditional simulators**: - Reservoir geological models usually have high uncertainty, leading to the need for a large number of numerical simulations to match historical data and optimize production. - Traditional reservoir simulators are computationally expensive, and a single simulation may take hours or even longer. 2. **Advantages of machine - learning models**: - Machine - learning models can provide faster predictions through training data, especially when dealing with large - scale data. - However, traditional data - driven models are usually difficult to extrapolate to situations outside the training parameter space and may not satisfy physical laws, resulting in inconsistent or untrustworthy prediction results. 3. **Physics - Informed Neural Networks (PINNs)**: - Physics - Informed Neural Networks directly embed physical laws into neural - network models, improving the reliability and performance of the models. - But physics - informed models usually require fully defined physical problems and are difficult to adapt to changing boundary conditions, such as well locations and well - control settings. ### Main contributions of the paper 1. **Proposing an FNO - based surrogate model**: - This model can make predictions under varying permeability fields, well locations, well - control settings and the number of wells, with a maximum average relative error of less than 5%. - By using simple data - augmentation techniques, the size of the data set is reduced and the risk of over - fitting is reduced. 2. **Efficient data processing and prediction**: - The input tensors are constructed in a binary way, enabling the model to make predictions under unseen well locations, well - control settings and the number of wells. - The model can accelerate the history - matching and reservoir - characterization processes by several orders of magnitude. 3. **Practical application value**: - This model can efficiently manage reservoirs and optimize production, and is especially suitable for scenarios that require rapid prediction and decision - making. ### Conclusion By proposing an FNO - based surrogate model, this paper overcomes the limitations of traditional reservoir simulators in dealing with changing boundary conditions and physical properties, achieving faster and more accurate predictions. This method has important application value in reservoir management and optimization.

Neural Operator-Based Proxy for Reservoir Simulations Considering Varying Well Settings, Locations, and Permeability Fields

Enhancing subsurface multiphase flow simulation with Fourier neural operator

Augmenting Deep Residual Surrogates with Fourier Neural Operators for Rapid Two-Phase Flow and Transport Simulations

Evaluating Reservoir Performance Using a Transformer Based Proxy Model

Physics-Aware Deep-Learning-Based Proxy Reservoir Simulation Model Equipped With State and Well Output Prediction

Fourier Neural Operator for Fluid Flow in Small-Shape 2D Simulated Porous Media Dataset

Multi-fidelity Fourier Neural Operator for Fast Modeling of Large-Scale Geological Carbon Storage

Predictive modeling of oil and water saturation during secondary recovery with supervised learning

Fourier Neural Operator for Solving Subsurface Oil/Water Two-Phase Flow Partial Differential Equation

Employing Deep Learning Neural Networks for Characterizing Dual-Porosity Reservoirs Based on Pressure Transient Tests

Efficient and generalizable nested Fourier-DeepONet for three-dimensional geological carbon sequestration

Accelerating Physics-Based Simulations Using Neural Network Proxies: An Application in Oil Reservoir Modeling

A Deep-Learning-Based Graph Neural Network-Long-Short-Term Memory Model for Reservoir Simulation and Optimization With Varying Well Controls

Deep Learning Based Hydraulic-mechanical Simulation of Petroleum Reservoirs

Fourier-MIONet: Fourier-enhanced multiple-input neural operators for multiphase modeling of geological carbon sequestration

A Novel A.I Enhanced Reservoir Characterization with a Combined Mixture of Experts -- NVIDIA Modulus based Physics Informed Neural Operator Forward Model

A data-driven smart proxy model for a comprehensive reservoir simulation

Fourier Neural Operator Surrogate Model to Predict 3D Seismic Waves Propagation

Reduced Geostatistical Approach With a Fourier Neural Operator Surrogate Model for Inverse Modeling of Hydraulic Tomography

Model-Parallel Fourier Neural Operators as Learned Surrogates for Large-Scale Parametric PDEs

Efficient Super-Resolution of Pipeline Transient Process Modeling Using the Fourier Neural Operator