Use of Model-Predictive Control for Automating SAGD Well-Pair Operations: A Simulation Study

Abstract

Summary For more than a decade, the steam-assisted-gravity-drainage (SAGD) process has been in use commercially to extract heavy oil from Athabasca oil-sand deposits located a few hundred meters below the surface. Operating challenges in the SAGD phase of operations arise not only because the process is hidden below the surface, but also because of the number of well pairs needed to drain the reserves and the number of variables that need to be considered for each well pair. Different control philosophies have been suggested and used to control SAGD well pairs, which focus on a single variable or a few out of the many variables to be considered. None of these philosophies have been found to consider the well pair as a whole, with interactions between the downhole variables, and none of the philosophies have been found to include predictive capability. In this paper, we demonstrate, for the first time, how model-predictive control (MPC) can be used to control, stabilize, and assist optimization of an SAGD well pair. This paper shows how we used proprietary MPC software to control and stabilize a typical configuration of a numerically simulated virtual SAGD well pair. Our simulation results demonstrate that MPC is suitable for use in the SAGD industry and provides insights into the use of MPC to manage actual SAGD operation.