| dc.description.abstract | Uncertainty poses a major concern in reservoir management. To maximize production, engineers and geoscientists use numerical models of the reservoir to simulate the production process in advance, and mathematical optimization is used to find an optimal recovery strategy. However, acquiring an accurate description of a subsurface hydrocarbon reservoir is impossible, and hence these models are highly susceptible to uncertainty. With uncertainties present in the model, it is hard to know what the optimal production configuration is, as the actual outcome might be different from the prediction. The standard approach to this issue is to use the expected values of the uncertain parameters when solving the optimization problem, which basically ignores the uncertainty.
This thesis addresses the uncertainty problem in reservoir management. A methodology for handling optimization problems containing uncertain parameters,called stochastic programming, is introduced. The reservoir control optimization problem is formulated using this framework, and it is argued for why it is necessarily better than a regular deterministic formulation. A case study regarding a realistic reservoir model with uncertainties is carried out, showing how the stochastic solution yields higher expected return and fewer constraint violations than the deterministic solution. Moreover, it is explained how stochastic programming
can be used to control risk when making decisions based on uncertain information. | |
