Stochastic optimization for hydro‐thermal power systems

A suboptimal stochastic optimization method for the long‐term scheduling of a hydroelectric reservoir is described. The objective of the optimization is to minimize thermal station fuel costs, in a mixed hydro‐thermal power system, over a one‐year horizon. Exact stochastic and deterministic solutions, obtained by dynamic programming (DP) for a simple one‐reservoir model, are compared. The two solutions differ significantly, indicating that deterministic solutions to the problem are inadequate. Stochastic DP is not computationally feasible for a multi‐reservoir system. A suboptimal linear feedback controller with non‐Gaussian reservoir water storages is then developed. It approximates the stochastic DP solution for one reservoir very well and is capable of extension to a multi‐reservoir version.