Modeling approaches for addressing enigmatic migration patterns for aqueous- and nonaqueous-soluble tracers in an enhanced oil recovery field

Abstract

A series of six aqueous-soluble and four nonaqueous-soluble tracer experiments and corresponding numerical simulations were executed for the Farnsworth Field in Ochiltree County, Texas, USA, a field which is undergoing tertiary enhanced oil recovery with water-alternating-gas (WAG) production. The combination of field experiments and numerical simulations was designed to identify flow pathways between injectors and producers and potential short circuiting of injected fluids. Field recoveries of aqueous-soluble tracers were dependent on the WAG stages of the tracer injection well, with shorter arrival times for strictly waterflooding and delayed arrival times for alternating injection stages. Aqueous-soluble tracer (i.e., 1,3,6-naphthalene trisulfonate, 1,5-naphthalene disulfonate, 1,6-naphthalene disulfonate, 2-naphthalene sulfonate, 2,6-naphthalene disulfonate, and 2,7-naphthalene disulfonate) arrivals for WAG injectors indicated water bypass was occurring during gas injection stages. Nonaqueous-soluble tracer (i.e., perfluoro-1,2-dimethylcyclohexane, perfluoroethylcyclohexane, perfluoromethylcyclohexane, and perfluoromethylcyclopentane) experiments revealed faster migration velocities than for the aqueous-soluble tracers and flow heterogeneities that resulted in the tracers bypassing nearer production wells. Base-case numerical simulations of the tracer experiments used a geologic model of the Morrow B sandstone production interval with parameters calibrated from history matching simulations, with the Morrow B sandstone sub-divided into hydrologic flow units (HFUs). Alternative simulation scenarios investigated HFU-dependent three-phase relative permeability models and dynamic intrinsic permeability enhancement with exposure to aqueous-dissolved CO₂. Compositional petroleum models with four components were shown to be sufficient for tracer modeling compared against a nine-component model, with a factor of four difference in simulation execution time. HFU-dependent relative permeability models and dynamic intrinsic permeability modifications influenced arrival times and production concentrations of both aqueous- and nonaqueous-soluble tracers but did not yield unique flow pathways compared to those observed in the base-case scenario.