We interpreted a series of single-well-chemical-tracer-tests (SWCTTs) estimating residual oil (SORW) to base high salinity waterflood, low salinity waterflood and subsequent polymer flood conducted on a Greater Burgan well. Interpretation of the tests requires history matching of the back-production of partitioning and non-partitioning tracers which is impacted by differing amounts of irreversible flow and differing amounts of dispersion as well as the amount of residual oil.

We applied the state-of-the-art chemical reservoir simulator (UTCHEM) and an assisted history matching tool (BP’s Top-Down-Reservoir-Modeling) to interpret the tests and accurately quantify uncertainty in residual oil saturations post high salinity, low salinity, and polymer floods. Two optimization algorithms (i.e., Genetic algorithm (GA) and Particle-Swarm-Optimization (PSO)-Mesh-Adaptive-Direct-Search (MADS) algorithms) were applied to better address the uncertainty.

Our results show a six saturation unit decrease in SORW post low salinity with no change to the SORW post polymer. This is in-line with our expectations - we expect no change in SORW post-polymer as the conventional HPAM, which does not exhibit visco-elastic behavior, was used in the test. We demonstrate that history matching the back-produced tracer profiles is a robust approach to estimate the SORW by showing that three-or four-layer simulation model assumption does not change the SORW estimated. We accounted for the uncertainty in partition-coefficient in our uncertainty estimates.

We present several innovations that improve history matching back-produced tracer profiles; hence, better SORW estimations (e.g., different level of dispersivity for individual simulation layers to account for different heterogeneity level as opposed to assuming a single dispersion for all layers). We generate more robust estimates of uncertainty by finding a range of alternative history matches all of which are consistent with the measured data.

You can access this article if you purchase or spend a download.