Real-time Optimization and Control for Oil Production under Uncertainty

Abstract

The practical use of model-based optimization depends significantly on the accuracy of the model in use. This is because the presence of uncertainty can introduce a mismatch between the model and the real plant, which may eventually lead to suboptimal or even infeasible solutions due to constraint violations in cases where fulfillment of the constraints is strictly required.

The existing robust methods for addressing the presence of uncertainty are known to be overly conservative, meaning they sacrifice optimality to a relatively great extent in order to achieve robustness. This is not ideal because the conservativeness or sacrifice of optimality implies the optimization has not been exploited to its full potential. Therefore, besides the complexity of the robust methods, the trade-off between optimality and robustness is a main challenge for this category of methods. On the other hand, the other approaches for reducing conservativeness can potentially lead to constraint violation, at least temporarily, meaning that the robust characteristics will be lost.

Accordingly, this thesis covers real-time optimization and control strategies tailored for short-term oil production processes, addressing the challenges imposed by the presence of uncertainty. Particularly, the research has been devoted to investigating the applicability of the existing methods for short-term oil production optimization under uncertainty and a further improvement in the existing method in order to reduce the conservativeness without losing the robust fulfillment of constraint and imposing extra computational complexity to the methods.

Two oil production methods, namely gas lift and electrical submersible pump lifting method, have been taken into account as case studies. The research mostly leaned toward the dynamic optimization formulated for short-term production from gas lift oil fields. The well-established methods within the domain of robust model predictive control, as well as an adaptive MPC framework with moving horizon estimation, have been investigated thoroughly. The knowledge acquired in these investigations is further employed to propose a robust method within the robust category that decreases the conservative-ness without imposing extra complexity and losing the robustness. This has been done particularly by using the output error in directly modifying the boundaries of the active constraints.

The dynamic formulation for the second case study, the ESP-lifted field, is proven to be challenging even in the deterministic case. Therefore, the scenario-based framework is used along with the steady-state model to formulate the robust approach for the ESP-lifted system. The use of the steady-state model enabled us to consider a longer horizon and assess the effect of uncertainty in the oil price as well as the uncertainty in the well parameters. The results demonstrated that the uncertainty in the oil price is not influential over the short-term horizon.

In summary, this thesis provides a comprehensive insight into the challenges associated with real-time optimization and control under uncertainty in the domain of short-term oil production. The research outcomes collectively emphasize the importance of accounting for uncertainty within the oil well characteristics. The thesis also offers insights into methods that can enhance efficiency, robustness, and operational safety in oil production processes.