ProxyFL: Decentralized Federated Learning through Proxy Model Sharing

Institutions in highly regulated domains such as finance and healthcare often have restrictive rules around data sharing. Federated learning is a distributed learning framework that enables multi-institutional collaborations on decentralized data with improved protection for each collaborator’s data privacy. In this paper, we propose a communication-efficient scheme for decentralized federated learning called ProxyFL, or proxy-based federated learning. Each participant in ProxyFL maintains two models, a private model, and a publicly shared proxy model designed to protect the participant’s privacy. Proxy models allow efficient information exchange among participants using the PushSum method without the need of a centralized server. The proposed method eliminates a significant limitation of canonical federated learning by allowing model heterogeneity; each participant can have a private model with any architecture. Furthermore, our protocol for communication by proxy leads to stronger privacy guarantees using differential privacy analysis. Experiments on popular image datasets, and a pan-cancer diagnostic problem using over 30,000 high-quality gigapixel histology whole slide images, show that ProxyFL can outperform existing alternatives with much less communication overhead and stronger privacy.

Artificial Intelligence Aided Proxy Model for Water Front Tracking in Fractured Carbonate Reservoirs

Saturation mapping in fractured carbonate reservoirs is a major challenge for oil and gas companies. The fracture channels within the reservoir are the primary water conductors that shape water front patterns and cause uneven sweep efficiency. Flow simulation for fractured reservoirs is typically time-consuming due to the inherent high nonlinearity. A data-driven approach to capture the main flow patterns is quintessential for efficient optimization of reservoir performance and uncertainty quantification. We employ an artificial intelligence (AI) aided proxy modeling framework for waterfront tracking in complex fractured carbonate reservoirs. The framework utilizes deep neural networks and reduced-order modeling to achieve an efficient representation of the reservoir dynamics to track and determine the fluid flow patterns within the fracture network. The AI-proxy model is examined on a synthetic two-dimensional (2D) fractured carbonate reservoir model. Training dataset including saturation and pressure maps at a series of time steps is generated using a dual-porosity dual-permeability (DPDP) model. Experimental results indicate a robust performance of the AI-aided proxy model, which successfully reproduce the key flow patterns within the reservoir and achieve orders of shorter running time than the full-order reservoir simulation. This suggests the great potential of utilizing the AI-aided proxy model for heavy-simulation-based reservoir applications such as history matching, production optimization, and uncertainty assessment.

Delineation of discrete conduit networks in karst aquifers via combined analysis of tracer tests and geophysical data

Assessment of the karst network geometry based on field data is an important challenge in the accurate modeling of karst aquifers. In this study, we propose an integrated approach for the identification of effective three-dimensional (3D) discrete karst conduit networks conditioned on tracer tests and geophysical data. The procedure is threefold: i) tracer breakthrough curves (BTCs) are processed via a regularized inversion procedure to determine the minimum number of distinct tracer flow paths between injection and monitoring points, ii) available surface-based geophysical data and borehole-logging measurements are aggregated into a 3D proxy model of aquifer hydraulic properties, and iii) single or multiple tracer flow paths are identified through the application of an alternative shortest path (SP) algorithm to the 3D proxy model. The capability of the proposed approach to adequately capture the geometrical structure of actual karst conduit systems mainly depends on the sensitivity of geophysical signals to karst features, whereas the relative completeness of the identified conduit network depends on the number and spatial configuration of tracer tests. The applicability of the proposed approach is illustrated through a case study at the Hydrogeological Experimental Site (HES) in Poitiers, France.

Efficient adjoint-based well-placement optimization using flow diagnostics proxies

Model-based optimization of placement and trajectories of wells in petroleum reservoirs by the means of reservoir simulation forecasts is computationally demanding due to the high number of simulations typically required to achieve a local optimum. In this work, we develop an efficient flow-diagnostics proxy for net-present-value (NPV) with adjoint capabilities for efficient computation of well control gradients and approximate sensitivities with respect to placement/trajectory parameters. The suggested flow-diagnostic proxy consists of numerically solving a single pressure equation for the given scenario and the solution of a few inter-well time-of-flight and steady-state tracer equations, typically achieved in a few seconds for a reservoir model of medium size. Although the proxy may not be a particularly good approximation for the full reservoir simulation response, we find that for the cases considered, the correlation is very good and hence the proxy is suitable for use in an optimization loop. The adjoint simulation for the proxy model which provides control gradients and placement sensitivities is of similar computational complexity as the forward proxy model (a few seconds). We employ a version of the generalized reduced gradient for handling individual well constraints (e.g., bottom-hole-pressures and rates). As a result, the individual well constraints are enforced within the flow-diagnostics computations, and hence every parameter update becomes feasible without sacrificing gradient information. We present two numerical experiments illustrating the efficiency and performance of the approach for well placement problems involving trajectories and simulation models of realistic complexity. The suggested placements are evaluated using full simulations. We conclude by discussing limitations and possible enhancements of the methodology.

A Physics-Based Proxy for Surface and Subsurface Coupled Simulation Models

Full-physics subsurface simulation models coupled with surface network can be computationally expensive. In this paper, we propose a physics-based subsurface model proxy that significantly reduces the run-time of the coupled model to enable rapid decision-making for reservoir management. In the coupled model the subsurface reservoir simulator generates well inflow performance relationship (IPR) curves which are used by the surface network model to determine well rates that satisfy surface constraints. In the proposed proxy model, the CPU intensive reservoir simulation is replaced with an IPR database constructed from a data pool of one or multiple simulation runs. The IPR database captures well performance that represents subsurface reservoir dynamics. The proxy model can then be used to predict the production performance of new scenarios – for example new drilling sequence – by intelligently looking up the appropriate IPR curves for oil, gas and water phases for each well and solving it with the surface network. All necessary operational events in the surface network and field management logic (such as facility constraints, well conditional shut-in, and group guide rate balancing) for the full-coupled model can be implemented and honored. In the proposed proxy model, while the reservoir simulation component is eliminated for efficiency. The entirety of the surface network model is retained, which offers certain advantages. It is particularly suitable for investigating the impact of different surface operations, such as maintenance schedule and production routing changes, with the aim of minimizing production capacity off-line due to maintenance. Replacing the computationally intensive subsurface simulation with the appropriate IPR significantly improves the run time of the coupled model while preserving the essential physics of the reservoir. The accuracy depends on the difference between the scenarios that the proxy is trained on and the scenarios being evaluated. Initial testing with a complex reservoir with more than 300 wells showed the accuracy of the proxy model to be more than 95%. The computation speedup could be an order of magnitude, depending largely on complexity of the surface network model. Prior work exists in the literature that uses decline curves to replicate subsurface model performance. The use of the multi-phase IPR database and the intelligent lookup mechanism in the proposed method allows it to be more accurate and flexible in handling complexities such as multi-phase flow and interference in the surface network.

A Proxy Peng-Robinson EOS for Efficient Modeling of Phase Behavior

Equation-of-state (EOS) compositional simulation is commonly used to model the interplay between phase behavior and fluid flow for various reservoir and surface processes. Because of its computational cost, however, there is a critical need for efficient phase-behavior calculations using an EOS. The objective of this research was to develop a proxy model for fugacity coefficient based on the Peng-Robinson EOS for rapid multiphase flash in compositional flow simulation. The proxy model as implemented in this research is to bypass the calculations of fugacity coefficients when the Peng-Robinson EOS has only one root, which is often the case at reservoir conditions. The proxy fugacity model was trained by artificial neural networks (ANN) with over 30 million fugacity coefficients based on the Peng-Robinson EOS. It accurately predicts the Peng- Robinson fugacity coefficient by using four parameters: Am, Bm, Bi, and ΣxiAij. Since these scalar parameters are general, not specific to particular compositions, pressures, and temperatures, the proxy model is applicable to petroleum engineering applications as equally as the original Peng-Robinson EOS. The proxy model is applied to multiphase flash calculations (phase-split and stability), where the cubic equation solutions and fugacity coefficient calculations are bypassed when the Peng-Robinson EOS has one root. The original fugacity coefficient is analytically calculated when the EOS has more than one root, but this occurs only occasionally at reservoir conditions. A case study shows the proxy fugacity model gave a speed-up factor of 3.4% in comparison to the conventional EOS calculation. Case studies also demonstrate accurate multiphase flash results (stability and phase split) and interchangeable proxy models for different fluid cases with different (numbers of) components. This is possible because it predicts the Peng-Robinson fugacity in the variable space that is not specific to composition, temperature, and pressure. For the same reason, non-zero binary iteration parameters do not impair the applicability, accuracy, robustness, and efficiency of the model. As the proxy models are specific to individual components, a combination of proxy models can be used to model for any mixture of components. Tuning of training hyperparameters and training data sampling method helped reduce the mean absolute percent error to less than 0.1% in the ANN modeling. To the best of our knowledge, this is the first generalized proxy model of the Peng-Robinson fugacity that is applicable to any mixture. The proposed model retains the conventional flash iteration, the convergence robustness, and the option of manual parameter tuning for fluid characterization.

Efficiency of Using a Proxy Model for Modeling of Reservoir Pressure

The presented paper is devoted to the development and testing of a computational tool for assessment of the reservoir pressure and prompt generation of the pressure maps of collectors. The tool is based on a proxy model that allows to solve the two-dimensional diffusion equation for unsteady liquid filtration using the boundary element method. To expand the applicability of the proxy model, an algorithm for automated parameter adaptation was developed. This algorithm allows to exclude knowingly unreliable data or low-quality data from modeling. This is achieved due to analyzing the correlation between the injection, production and bottom-hole pressures for the entire well stock over the history of the reservoir development. In addition, this paper describes an approach to modeling two-phase oil and gas filtration based on the use of pseudofunctions. This approach considers the influence of gas on the oil filtration process. The use of pseudofunctions allows us to linearize the diffusion equation for two-phase filtration and to solve it using the boundary element method in the same manner as for the case of oil filtration without gas. To demonstrate the results of the proxy model validation, examples of its use for generating the pore pressure maps for two real collectors are given. The average values of the reservoir pressure at the wells calculated using the proxy model are compared with the results of the corresponding well tests and with the traditional isobar maps. The analysis showed that the average deviation of the proxy model from the real reservoir pressures is less than 10%.

Is Artificial Intelligent the Future for AHM? Decision Making Between an Automated Optimisation of ANN Proxy Versus Full Numerical Simulation Optimisation Technique

Over the past decades, Assisted History Matching has been the new norm for history matching that leverages the rapid advancement in digital computational performance. Continuous advancements such as parallel computing and GPU accelerates numerical simulation which overcomes the cumbersome experience of working with large fine-scaled model that mainly concerns the simulation time and intervention of engineers. As more interest emerges around artificial intelligence in the optimisation process, this paper explores the Artificial Intelligence algorithm to optimize two proxy modelling techniques: Quadratic Polynomial and Artificial Neural Network proxy model. These techniques are compared with stochastic optimisation method known as Differential Evolution algorithm on their efficiency of optimizing the objective functions, time taken, and knowledge investment needed by engineer, given today's hardware technology. This paper starts off by using Latin hypercube experimental design to generate first ensemble of simulation cases to generate proxy models to match the historical cumulative oil and water production by well level. The quality of both proxy modelling techniques is evaluated using R2 coefficient and proxy plot. Proxy models are then further validated by creating real simulation models from variants generated via Monte Carlo Analysis. The history matching quality and practicality were compared between the AI algorithm that runs optimizer on top of existing proxy models, and Differential Evolution algorithm in optimizing the regional porosity and permeability multipliers. The ANN proxy model prevailed over quadratic proxies to mimic the numerical reservoir model output with high degree of accuracy. The black-box nature of the ANN proxies limits the interpretability of predicted model when compared quadratic proxies where the formula for the proxy model can be obtained. Quadratic approximations are more flexible, simplistic in nature, and requires less computational cost to be constructed. Despite that, its prediction quality maybe subjected to the degree of non-linearity in the simulation model. The use of AI algorithm vastly reduces the number of full reservoir simulation required to achieve the minimum objective function at a shorter timeframe, which is proved to be the strength of such method. However, AI optimisation is highly susceptible to be trapped in local minimum. This paper proved the superiority of Differential Evolution algorithm over AI, that it may avoid being trapped in local minimum to achieve high degree of prediction accuracy for the history matching given the larger number of iterations required. This paper provides a preliminary understanding of optimisation workflow and how to go about each optimisation strategies: quadratic polynomial proxy, ANN proxy, stochastic optimisation, artificial intelligence techniques, and a novel approach of converting proxy predicted variants into real simulation cases to evaluate proxy quality. Hence establishes engineers’ expectation by appraising the pros and cons of each optimisation strategies.

Robust Screening Criteria for Foam-Assisted Water-Alternating Gas FAWAG Injection

Foam-Assisted-Water-Alternating-Gas (FAWAG) injection has been proposed to improve the inherent unfavorable mobility ratio of gas and liquid in WAG process. The foam reduces gravity override and gas channeling as to improve volumetric sweep efficiency and thus oil recovery. There are still a lot of uncertainties yet to be understood in foam dynamics, surfactant adsorption, and foam stability when contacting oil, which impact the actual foam propagation into the reservoir. Although some insights are gained from laboratory and field experiments, the performance, and design of the injection strategy and facilities as part of the field development of FAWAG is not trivial and field data is sparse. Extensive laboratory experiments and simulation studies are necessary to de-risk enhanced oil recovery (EOR) application, but these processes are time consuming and expensive. For this reason, a screening study is normally conducted to increase the possibility of selecting high potential candidates prior to embarking on the detailed feasibility studies. Unfortunately for FAWAG, the screening criteria are not readily established nor commonly available in commercial screening tools unlike for other matured EOR methods, largely contributed by the limited database on FAWAG field implementations worldwide. This paper presents a robust FAWAG screening tool which accounts for important reservoir properties, uncertainties in foam model parameters, as well as various reservoir conditions of oil and gas production and injection plans. The FAWAG process is modelled from the assumption of local equilibrium of foam creation and coalescence using an Implicit Texture model. Relevant foam scan experiments/steady state coreflood data were analyzed to derive parameters that characterize foam dynamics. The sensitivity study in this paper ranks and identifies the main risks and opportunities for the FAWAG process, quantifies the reliability of the model and increases the understanding of the effective dynamic behaviour. The sensitivity study was the basis for the development and validation of a proxy model by design of experiments. The screening tool employs this proxy model to generate immediate screening results without the need to run additional simulations. The screening tool was further validated with upscaled experimental data. A set of prediction results on the range of oil recovery for numerous plausible field scenarios was established; these screening criteria will be used as the basis for high-level decision making.