(170f) Multi Variable Model Predictive Control to Improve Oil Production for Steam-Assisted Gravity Drainage (SAGD)

Steam-Assisted Gravity Drainage (SAGD) is a major thermal recovery technique in Western Canada. It involves injection of superheated steam through an upper injection well to mobilize and drain the highly viscous bitumen to form a liquid pool above a bottom producer well before its subsequent production. Existing research has shown that controlling the temperature difference between the injected steam and the produced oil-water emulsion, which is also called the subcool temperature difference, can regulate the height of the liquid pool such that it is favorable for production. This technique, called steam trap control, prevents the production of live steam and also protects the formation from fracturing, by keeping the steam injection rates within allowable limits. However, major drawbacks of this technique are the absence of a method to optimally control the production of the fluids from the liquid pool and the lack of guidelines to determine the ideal subcool temperature difference setpoint.

In this study, we address these issues through the development of a multi input multi output (MIMO) model predictive controller (MPC) that incorporates adaptive oil rate and steam trap controls. The addition of the adaptive oil rate control helps improve the oil production rate from the liquid pool and drainage rate of oil into the liquid pool. Furthermore, the steam trap control setpoint was turned into an optimization variable, such that the optimizer determined the optimal subcool setpoint to satisfy the adaptive oil rate constraints. It should be noted that oil rate control by itself leads to formation damage due to high injection rates of steam and live steam production due to high production rates of the oil-water emulsion from the liquid pool. Therefore, it was used in conjunction with steam trap control, which not only reduces the live steam production, but also protects the formation from fracturing. The novel MIMO control strategy was compared with a multi input single output (MISO) MPC steam trap control strategy. The real time control study for both strategies were enabled using a bidirectional communication that was established between CMG STARS (virtual oil reservoir) and MATLAB (onsite controller) software. The results show a 171% improvement of oil recovery for the novel MIMO controller as compared to the MISO controller over a period of 6 simulated years. Moreover, there was a 35% improvement in the cumulative steam-to-oil ratio (a performance metric for SAGD) for the MIMO controller over the MISO controller.