Because of computational advances in reservoir simulation with high-performance computing, it is now possible to simulate more than thousands of reservoir-simulation cases in a practical time frame. This opens a new avenue to reservoir-simulation studies, enabling exhaustive exploration of subsurface uncertainty and development/depletion options. However, analyzing the results of a large number of simulation cases still remains a challenging and overwhelming task. We propose a new method that enables the efficient analysis of massive reservoir-simulation results, often consisting of thousands of cases, by discovering interesting patterns of relationships among variables in large data sets. The method uses a well-known data-mining method, called association-rule mining, together with a high-dimensional visualization technique. We demonstrate the capability of the proposed method by using it to analyze the reservoir-simulation results from the Sensitivity Analysis of the Impact of Geological Uncertainty on Production (SAIGUP) project, which is an interdisciplinary reservoir-modeling project carried out earlier by Manzocchi et al. (2008a). To investigate the influence of geological features on oil recovery in shallow marine reservoirs, numerous reservoir models were built and flow-simulated in the SAIGUP project. In this paper, we analyze the simulation results from an ensemble of 9,072 models, which cover all possible combinations of several structural and sedimentological parameters individually varied to describe geological uncertainty. To be able to analyze the simulation results from such exhaustive sampling from high-dimensional model parameter space, it is crucial to efficiently decompose complex interactions between model parameters and to discover hidden impacts on flow response. By coupling the association-rule mining algorithm and high-dimensional visualization, such interactions and impacts are rapidly extracted and visualized in such a way that engineers and geoscientists can interpret meaningful sensitivities at a glance. This methodology provides a novel way to rapidly interpret flow response from a large ensemble of reservoir models without being overwhelmed by massive data.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据