Dynamic Self-Optimizing Control for Uncertain Oil Reservoir Waterflooding Processes

Abstract

Oil production is commonly enhanced via a process called waterflooding that simply involves the injection of water into an underground reservoir to increase productivity. Optimal production of oil through waterflooding has to consider oil revenue as well as costs of water injection and processing afterward. This is a dynamic optimization problem and can be solved numerically through traditional optimal control theory, which relies entirely on the accuracy of a process model and can only give open-loop solutions. However, reservoir properties are highly uncertain, and therefore, the traditional method is not adequate for real reservoir applications. To counteract the effects of such uncertainties, a feedback control approach has been recently proposed based on flow measurements. However, this approach is not realistically implementable when initial flow measurements are not available. This article presents a novel method based on self-optimizing control technique for optimal waterflooding operation. The approach finds appropriate controlled variables (CVs) from the combination of measurement histories and manipulated variables through regression. A feedback control law for control offset in the form of linear combination of measurement histories was then obtained from the CVs. Initial control signals in the absence of flow measurements are computed from the estimates of reservoir size and field configuration of wells. The robustness of the feedback solution was shown through case studies where it was found to be better than open-loop solutions by 95.06% but only 1.48% worse than that achieved through true optimal control (where systems properties are assumed to be known a priori).