Extension and Verification of a Simple Model for Vertical Sweep in Foam Surfactant-Alternating-Gas Displacements

Abstract

Summary Foam is a promising method to improve sweep in gas-injection enhanced-oil-recovery (EOR) projects. For cases in which well-to-well distances are large, the question arises whether foam can prevent gravity segregation over these distances. For such cases, theoretical studies suggest that the best process is to inject one large slug of surfactant followed by one large slug of gas. Shan and Rossen (2004) present a simple model for such a process that provides an initial assessment whether foam can prevent segregation over large distances. They did not extend their calculations to the large distances between wells in some potential applications, and they treated only the case in which vertical permeability kv equals horizontal permeability kh. Here, we extend the model to cases of kv < kh. We derive an analytical solution for the limit as kv approaches zero, which could serve as a quick first estimate of feasibility for other cases. Surprisingly, the model predicts that gravity segregation is worse as kv decreases; the reason is that, with large kv, foam pushes downward in response to the pressure difference across the tilted foam front. We illustrate the use of the model with an example from a North Sea field, in which the issue was whether foam could prevent gravity segregation over a distance of several kilometers. The simple model said this was feasible, and more-detailed 2D cross-section simulation then confirmed that segregation was insignificant over an interwell distance of 6 km. One can fit the model parameters to foam parameters derived directly from laboratory data. We illustrate with a model fit by Rossen and Boeije (2013) to data of Persoff et al. (1991), a fit specifically designed for a hypothetical surfactant-alternating-gas (SAG) application. The fit of simulations to the simple model is closer than in the North Sea example, and the prediction that sweep is better with kv = kh than for kv = 0 is confirmed. Moreover, the idealized model predicts the injectivity in this case very accurately.