Improving Chemical-Enhanced-Oil-Recovery Simulations and Reducing Subsurface Uncertainty Using Downscaling Conditioned to Tracer Data

Chemical-enhanced oil-recovery (CEOR) mechanisms are strongly influenced by gridblock size and reservoir heterogeneity compared with conventional waterflooding (WF) simulations. In WF-simulation models, simulation grids are commonly upscaled (coarsened) on the basis of a single-phase flow to perform history matching and sensitivity analyses within affordable computational times. However, this coarse-grid resolution (typically, approximately 100 ft) is insufficient for CEOR, and hence usually fails to capture key physical mechanisms. These coarse models also tend to increase numerical dispersion, artificially increase the level of mixing, and have inadequate resolution to capture certain geological features to which EOR processes can be highly sensitive. Thus, coarse models often overestimate the sweep efficiency as a result of numerical dispersion, and underestimate the displacement efficiency as a result of the artificial dilution of chemicals. Therefore, grid refinement is necessary for CEOR simulations when the original (fine) Earth model is not available or when major disconnects occur between the original Earth model and the history-matched coarse WF model. However, recreating the fine-scale heterogeneity without degrading the history match from the coarse grid remains a challenge. Because of the different recovery mechanisms involved in CEOR, such as miscibility and thermodynamic phase behavior, the impact of grid downscaling on CEOR simulations is not well-understood. In this work, we introduce a geostatistical downscaling method that can be conditioned to tracer data, for refining coarse history-matched WF models. The proposed downscaling method refines the coarse grid and populates the relevant properties in the newly created, finer gridblocks, reproducing the fine-scale heterogeneity while retaining the fluid material balance. The method treats the values of rock properties in the coarse grid as hard data, and the corresponding variograms and property distributions as soft data. We outline a work flow that reduces uncertainties in the geological properties by integrating dynamic data such as sweep efficiency from the interwell tracers. We provide several test cases, and demonstrate the applicability of the proposed method to improving the history match of a CEOR pilot.