Integrated, scalable MBS for flow through porous media

流体在多孔材料中的流动对于理解和预测不同功能和规模的系统的行为至关重要，这些系统包括油气藏、含水层、带填料床的分离塔反应器单元、过滤器、膜分离器，甚至催化转化器。最近，为了改善油藏管理，油藏模拟中加入了更多的物理元素，并要求通过使用高性能计算（HPC）大幅提高计算能力。随着石油和天然气价格在一年内从过去20年的最低水平波动到最高水平，这种需求变得尤为重要。这个项目的目标是开发一个集成的模拟器通过非均质多孔材料的流动使用模拟层次。目前的方法包括使用具有单一物理尺度的模拟。然而，高性能计算的最新进展使得显著增加问题规模和使用更复杂的方法成为可能。挑战在于将单个模拟组合成一个集成的多尺度系统，该系统将能够包括所有物理尺度，并将根据输入数据进行自我调整。重点将放在最终产品的可移植性、可扩展性、效率和可扩展性上。该模拟器将成为油气储层管理的改进预测工具，并准备在集成网格架构中使用。多孔介质的流动是一种多尺度现象。微观尺度模拟，基于晶格玻尔兹曼方法，将用于直接模拟流动通过多孔材料。岩石样品的显微层析数字图像将被用来真实地代表空间领域受到流动，利用晶格玻尔兹曼方法的灵活性。在细观尺度上，将采用随机方法系统地隔离和研究岩石结构微观特征对流场的影响。随机方法也将用于开发岩石性质表征的方法。基于传统有限差分方法的非宏观模拟将用于测试改进的流动模型对油藏规模油气产量的影响。宏观模拟将包括生产/注水井（在油藏模型中形成奇点）在油藏整个生命周期中的行为。它还将结合流体与地质力学（孔隙度相关渗透率和非达西系数）的耦合。本项目的教育工作将着重培养本科生使用高性能计算资源的能力。本研究将：(i)提高我们对基本流动机制的理解；（ii）更新各向异性多孔材料非达西流动模型；（iii）在模拟中整合不连续的存在，如井和裂缝。拟议研究的创新之处包括：(a)在不同尺度上使用最先进的模拟；(b)使用实验测量的量来推断多孔介质的性质并更新流动模型；(c)将一套原型软件的个别组件集成为供工业使用的无缝模拟器；(d)共享内存和分布式并行的混合应用，以实现各种HPC架构的可扩展性。这个研究项目对于参与其中的研究生的教育经历将是非常有价值的。它的教育方面也将包括将高性能计算应用纳入三个系的本科课程，并为本科生的研究经验开发研究项目。因此，它将使我国的大批技术人员对高性能计算应用的有用方面和与高性能计算基础设施进行互动做好准备。

Fluid flow through porous materials is critical for understanding and predicting the behavior of systems as diverse in function and scale as hydrocarbon reservoirs, aquifers, separation tower andreactor units with packed beds, filters, membrane separators and even catalytic converters. Recently, there has been a thrust to incorporate more physics in reservoir simulations, as well as acall for substantial improvements in computational capability through the use of High Performance Computing (HPC), in order to improve reservoir management. This need hasbecome particularly critical as oil and gas prices have fluctuated within one year from the lowestlevel of the past two decades to the highest. The goal of this project is to develop an integratedsimulator for flow through heterogeneous porous materials using a hierarchy of simulations.Current approaches involve the use of simulations having a single physical scale. However,recent advances in HPC have made it possible to increase significantly the problem size and touse more sophisticated approaches. The challenge is to combine the individual simulations into anintegrated multiscale system that will be able to include all physical scales and will self-adjust inaccordance with the input data. Emphasis will be placed on the portability, scalability, efficiencyand extensibility of the final product. The proposed simulator will be an improved prediction toolfor hydrocarbon reservoir management and will be ready for use on integrated grid architectures,as they become available.Flow through porous media is a multi-scale phenomenon. Microscopic scale simulation, based onLattice Boltzmann Methods, will be used for the direct simulation of flow through porous materials. Microtomographic digital images of rock samples will be used to realisticallyrepresent the spatial domain subjected to flow, taking advantage of the flexibility of LatticeBoltzmann Methods. At the mesoscopic scale, stochastic methods will be used for the systematicisolation and study of the effects of microscopic features of rock structure on the flow field. Thestochastic approach will also be used to develop a method for rock property characterization. Amacroscopic simulation, based on conventional finite difference methods, will be used to test theimpact of modified flow models on hydrocarbon production at reservoir scale. The macroscopicsimulation will incorporate the behavior of production/injection wells (which form singularities ina reservoir model) over the life of a reservoir. It will also incorporate coupling of flow withgeomechanics (porosity-dependent permeability and non-Darcy coefficients). The educationeffort resulting from this project will emphasize the training of undergraduate students in the useof HPC resources.This research will: (i) improve our understanding of the fundamental flow mechanisms;(ii) update the model for non-Darcy flow through anisotropic porous materials; and (iii) integratethe presence of discontinuities, such as wells and fractures, in the simulation. The innovations ofthe proposed study include: (a) use of state-of-the-art simulations at different scales; (b) use ofexperimentally measured quantities to deduce the properties of the porous medium and to updateflow models; (c) integration of the individual components of a set of prototype software into aseamless simulator for industrial use; and (d) application of a hybrid of shared-memory anddistributed parallelism to achieve scalability on a variety of HPC architectures. The research project will be extremely valuable for the educational experience of the graduate students involved. Its educational aspect will also involve the incorporation of HPC applications in the undergraduate curricula of three Departments and the development of research projects for Research Experience for Undergraduates. It will, thus, prepare a large group of the technical workforce of our State to the useful aspects of HPC applications and to interact with HPC infrastructure.