Application of Dynamic Team Theory to the Scheduling of Reservoirs in the Presence of Random Inflows

Abstract

The management of water in a multiple reservoir system constitutes an important element of the operational scheduling problem in electric power systems. The presence of random inflows and system nonlinearities would require stochastic dynamic programming to provide an optimal feedback solution. However, the dimensionality of the hydro-scheduling problem rules out the direct use of stochastic dynamic programming, and leads to the consideration of decomposition methods requiring less computation. This paper uses concepts from dynamic team theory to characterize and develop stochastic spatial decomposition methods to systems of (a) parallel reservoirs, and (b) series reservoirs. These methods can be combined to generate the stochastic spatial decomposition algorithm for the general reservoir configuration. The computational features of these algorithms are discussed and numerical results will be presented at the conference.