The ideal reservoir characterization is achieved from a closed-loop system that integrates geological modeling, upscaling/downscaling, reservoir simulation, history matching, and production forecasts associated with unbiased uncertainty quantification. The initial step of reservoir characterization is to identify reservoir models from all plausible geologic scenarios reflecting prior uncertainty in reservoir description. CSM has been developing model selection that implements a proxy to image flow and geomechanical responses of geologic models for CO2 sequestration (Figure 1). A fast approximation utilizes a partial coupling scheme to obtain pressure and stress fields sequentially. A particle tracking algorithm mimics flow paths of the models, and a stress-field solver calculates displacements of the models caused by injected CO2.
The prior models showing similar proxy responses are grouped into clusters by multi-dimensional scaling and k-means clustering. Full physics simulations are run for cluster representatives to select the best-fit cluster, which is the group whose representative exhibits the smallest discrepancy between observed and estimated responses. The models in the best-fit cluster constitute the posterior model set. In summary, the posterior ensemble incorporating geophysical time-lapse observations is more representative of the reservoir than the larger prior ensemble.
Figure 1. Reservoir characterization coupled a connectivity-based proxy within a model selection framework (above)
Figure 2. Posterior ensemble mean of the permeability field for the Krechba field, Algeria: (a) the uppermost layer after matching bottom-hole pressure and (b) the uppermost layer after matching bottom-hole pressure and InSAR (Interferometric Synthetic Aperture Radar) data (above)