A Robust Method for Adaptive Model Uncertainty Reduction, 20-R9704

Printer Friendly Version

Principal Investigators
Alexander Y. Sun
Alan P. Morris
Sitakanta Mohanty

Inclusive Dates:  04/01/07 – Current

Background - Effective management of modern oil and gas reservoirs and ground water resources calls for adaptive decision-making frameworks that are facilitated by efficient data fusion methodologies and high-fidelity geologic and fluid flow models. Natural geologic systems are formed by complex geologic processes and exhibit heterogeneities at multiple scales. Consequently, model uncertainty, arising from lack of measurements and system simplification, is an inescapable aspect of geologic modeling. Recent advances in innovative sensing technologies allow engineers to harness a great amount of dynamic production data in real-time which, in turn, can be used to infer geologic features and identify inaccuracies in the a priori system representation. This project is motivated by the strong need to integrate production data into management models to support real-time or near real-time model updating and, thus, improve model reliability and decision-making effectiveness.

Approach - Sequential Bayesian updating algorithms based on the Kalman filter and its variants are widely applied in the data assimilation community to update model states and model parameters. However, a major technical challenge in our case stems from the high dimensionality of simulation models. The sample size required to estimate a multivariate probability distribution function with a given accuracy increases exponentially with dimensionality. To circumvent this so-called curse-of-dimensionality dilemma, we have developed a grid-based, cluster ensemble Kalman filter (GCEnKF) for updating non-Gaussian parameter distributions in high-dimensional models via dynamic data assimilation. Our GCEnKF algorithm combines a grid-based local ensemble analysis approach with a Gaussian mixture model data-clustering technique to effectively decrease the problem dimension and reduce the chance of filter degeneracy. We are demonstrating the algorithm for two application areas: multifacies alluvial systems and faulted reservoirs.

Accomplishments - We have demonstrated the GCEnKF numerically for identifying both the facies geometries and hydraulic conductivities in multifacies alluvial aquifers, where both intra- and interfacies heterogeneities are incorporated. Our results indicate that GCEnKF is a promising tool for assimilating high-dimensional, multimodal parameter distributions. Current work includes applying GCEnKF to a carbonate faulted reservoir to identify subseismic faults.

2008 Program Home