Prediction of plume migration using injection data and a model selection approach

During a geological sequestration project, realistic assessment of geological heterogeneity of the storage formation is needed for predicting the migration of CO2.A single deterministic model of the aquifer would not correctly represent the underlying uncertainties (geological, petrophysical, fluid properties etc.), nor would it address the issue of uncertainty in predictions based on uncertain subsurface models. Recently we have explored the idea of creating groups of models that share common flow characteristics and subsequently selecting the group exhibiting flow characteristics closest to the observed field data. This idea conveys two key advantages: it provides an estimate of uncertainty in the results of the history matching, and it helps identify those heterogeneities large enough to affect plume movement significantly. In this paper, we propose and demonstrate certain improvements in the workflow. This approach holds promise as an inexpensive method of monitoring plume migration.