Scenario-based robust optimization of water flooding in oil reservoirs enjoys probabilistic guarantees

Model-based optimization of the water-flooding process in oil reservoirs suffers from high levels of uncertainty arising from strongly varying economic conditions and limited knowledge of the reservoir model parameters. To handle uncertainty, diverse robust optimization approaches that use an ensemble of uncertain parameter realizations (i.e., scenarios), have been adopted. However, in scenario-based approaches, the effect of considering a finite set of scenarios on the constraint violation and/or the performance degradation with respect to the unseen scenarios have not been studied. In this paper, we provide probabilistic guarantees on the worst-case performance degradation of a scenario-based solution. By using statistical learning, we analyze the impact of the number of scenarios on the probabilistic guarantees for the worst-case solution subject to both economic and geological uncertainties. For the economic uncertainty, we derive an explicit a-priori relationship between the probabilistic guarantee and the number of considered scenarios, while for the geological uncertainty, a-posteriori probabilistic upper bounds on the worst-case solution are given.