NEW CONTROL CONCEPTS FOR UNCERTAIN WATER-RESOURCES SYSTEMS .1. THEORY

A major complicating factor in water resources systems management is h andling unknown inputs. Stochastic optimization provides a sound mathe matical framework but requires that enough data exist to develop stati stical input representations. In cases where data records are insuffic ient (e.g., extreme events) or atypical of future input realizations, stochastic methods are inadequate. This article presents a control app roach where input variables are only expected to belong in certain set s. The objective is to determine sets of admissible control actions gu aranteeing that the system will remain within desirable bounds. The so lution is based on dynamic programming and derived for the case where all sets are convex polyhedra. A companion paper (Yao and Georgakakos, this issue) addresses specific applications and problems in relation to reservoir system management.