Model Selection Coupled with a Particle Tracking Proxy Using Surface Deformation Data for Monitoring CO2 Plume Migration

Abstract:

This study formulates a framework of a model selection that refines geological models for monitoring CO₂ plume migration. Special emphasis is placed on CO₂ injection, and the particular techniques that are used for this study including model selection, particle tracking proxies, and partial coupling of flow and geomechanics. The proposed process starts with generating a large initial ensemble of reservoir models that reflect a prior uncertainty in reservoir description, including all plausible geologic scenarios. These models are presumed to be conditioned to available static data. In the absence of production or injection data, all prior reservoir models are regarded as equiprobable. Thus, the model selection algorithm is applied to select a few representative reservoir models that are more consistent with observed dynamic responses. A quick assessment of the models must then be performed to evaluate their dynamic characteristics and flow connectivity. This approach develops a particle tracking proxy and a finite element method solver for solving the flow equation and the stress problem, respectively. The shape of CO₂ plume is estimated using a particle-tracking proxy that serves as a fast approximation of finite-difference simulation models. Sequentially, a finite element method solver is coupled with the proxy for analyzing geomechanical effects resulting from CO₂ injection. A method is then implemented to group the models into clusters based on similarities in the estimated responses. The posterior model set is chosen as the cluster that produces the minimum deviation from the observed field data. The efficacy of non-dominated sorting based on Pareto-optimality is also tested in the current model selection framework. The proposed scheme is demonstrated on a carbon sequestration project in Algeria. Coupling surface deformation data with well injection data enhances the efficiency of tracking the CO₂ plume. Therefore, this algorithm provides a probabilistic approach to assessing the migration of the CO₂ plume in the aquifer and quantifying the risk associated with sequestration projects. Iterating the selection of clusters can further refine the forecasts of CO₂ plume migration.