We present a robust optimization formulation as a part of a framework being developed for the real-time operation of multi-objective and multi-reservoir river network systems. We describe our approach to the modeling of uncertainty and its propagation through the system. We discuss the utilization of uncertainty quantification within the objectives. We also develop an algorithm which additionally allows to approximate the failure probabilities associated with probabilistic constraints. This project is funded by the Bonneville Power Administration's Technology Innovation initiative.