Optimal SAG Design in Heterogeneous Reservoirs

Abstract

Foam can divert flow from higher- to lower-permeability layers and thereby improve vertical sweep efficiency in gas-injection enhanced oil recovery. Recently, Kapetas et al. (2015) measured foam properties in cores from four sandstone formations ranging in permeability from 6 to 1900 md, and presented parameter values for a foam-model fit to those data. Permeability affects the limiting capillary pressure at which foam collapses in the "high-quality regime". Kapetas et al. showed how foam would divert injection between layers of these formations if all layers were full of foam at a given quality (gas fractional flow). Here we examine the effects of SAG (surfactant-alternating-gas) injection method on diversion in a dynamic foam process using fractional-flow modelling and the model parameters derived by Kapetas et al. We consider a hypothetical reservoir containing non-communicating layers with the properties of the four formations in that study. The effectiveness of diversion varies greatly with injection method. In a SAG process, diversion of the first slug of gas depends on foam behaviour at high foam quality. Foam mobility in the foam bank during gas injection depends on the nature of a shock front that bypasses most foam qualities usually studied in the laboratory. The foam with the lowest mobility at fixed foam quality does not necessarily give the lowest mobility in a SAG process. In particular, diversion depends on how and whether foam collapses at low water saturation; this property varies greatly among the foams reported by Kapetas et al. Moreover, diversion depends on the size of the surfactant slug received by each layer before gas injection. This of course favours diversion away from high-permeability layers that receive a large surfactant slug, but there is an optimum surfactant slug size: too little surfactant and diversion from high-permeability layers is not effective; too much and mobility is reduced in low-permeability layers, too. Using a model based directly on laboratory data, this study shows how diversion between layers differs with injection method.