Improved Waterflooding Performance using Model Predictive Control

Hydrocarbon resources around the world are limited, and to satisfy future energy demands more efficient recovery solutions are needed. In oil production, a common method to bring the oil to the surface is to flood a reservoir by injecting water. After properly locating the injection and production wells, injection continues as long as it is profitable, and the wells are shut in afterwards (refer to as reactive control). In order to increase the recovery factor, this task has been cast into an integrated and more structured approach called Closed Loop Reservoir Management (CloReM). In this methodology, optimal production settings are determined based on reservoir models in a feedback loop. These models are highly non-linear and have large number of parameters and states, which are usually expressed with a great deal of uncertainties. Therefore the performance of CloReM is highly dependant on quality of the models, that are not updated quite often. To circumvent the unwanted effects of uncertainties and disturbances, a reference tracking framework using Model Predictive Control (MPC) is investigated, in order to provide more rapid corrective responses. This framework acts like a secondary loop (in addition to the original feedback loop) that has a supervisory task, and reoptimizes the settings based on low order linear models. Validity (i.e. the prediction horizon) of such models is relatively short, but they can get updated rapidly in several working points. In this assignment we have studied the possibility of using system identification methods using Prediction Error Identification (PEI) and Subspace Identification (SubID) to derive Linear Time Invariant (LTI) reservoir models to be implemented later in MPC. Furthermore, the benefits of the proposed loop are examined in an open loop life-cycle optimization of the reservoir, i.e. with fixed optimal references during the whole production life. The MPC has been implemented on a 5-spot homogeneous, and a 3D multi layered heterogeneous reservoirs. It has been shown that manipulating the injection rates and Bottom Hole Pressure (BHP) according to an MPC controller, can increase the profit up to 6.3% in heterogeneous reservoir and up to 18% in the homogeneous case, in compare to the conventional control scheme. M.Sc. thesis Amin Rezapour