Estimating two-phase reactive flow model parameters from single- and two-phase modified-salinity core flooding data

Abstract

Summary Due to the complexity of underlying physics of the modified salinity water flooding, mechanistic models are often utilized to better understand and predict its behaviour in the field scale. The mechanistic models are a combination of several submodels of different nature, each with several adjustable parameters. Adjusting these parameters by fitting the model to a limited number of recovery factors obtained from core flooding experiments is not a viable solution, often estimating highly uncertain values for the model parameters. We address this challenge by providing a framework for fitting a mechanistic two-phase reactive-transport model to a combination of low-cost single- and two-phase flow experimental data. The effectiveness of modified-salinity water flooding is often tested in (qualitative) spontaneous and (quantitative) forced imbibition tests in the lab. Several mechanisms that are suggested for explaining the observed improved oil recovery cannot be distinguished in those traditional imbibition tests. Mechanistic models with adjustable physically-meaningful parameters exist for these mechanisms, e.g. carbonate dissolution, surface charge (and force) alteration, fines migration, water weakening, etc. However, obtaining these adjustable parameters by fitting “a mechanistic model that incorporates all these mechanisms” is not a good strategy. Our mechanistic models are a combination of chemical equilibrium and kinetics models that describe the chemical reactions between the ionic species in the aqueous phase (electrolyte model with known parameters), chalk and oil surface complexation reactions (CD-MUSIC, diffuse layer models with unknown parameters), and an empirical parameter linking the surface reactions to the relative permeability and capillary pressure model parameters. When fitting the model to the core flooding data, the optimization algorithms are more sensitive to the relative permeability parameters. Moreover, the number of parameters are often too many and can result in a largely underdetermined system of equations, for which the optimization algorithm is very sensitive to the initial estimates. Our novel numerical framework optimizes the model parameters by simultaneously fitting the parameters to a set of core flooding, spontaneous imbibition, and single-phase chromatographic tests (i.e. injecting MSW to a core saturate with formation water and measuring effluent ionic concentrations with time). We obtain the initial estimates of the surface complexation model parameters by fitting the model to the zeta potential measurements performed on powdered carbonate suspensions. We finally demonstrate the capabilities of our framework by optimizing model parameters for a set of inhouse experiments performed on chalk cores from the North Sea reservoirs.