Study of a Real-time Adaptive Closed-loop Control Algorithm for Reservoir Operation

Abstract

Recent studies on the optimal control of the High Aswan Dam in Egypt have illustrated the usefulness of steady-state stochastic dynamic programming techniques for deriving optimal release policies. This work likewise uses stochastic dynamic programming for the determination of optimal High Dam releases, but a new adaptive reservoir control scheme capable of handling a nonstationary system is investigated. Using results of the stationary control problem as boundary conditions, a finite horizon optimization problem is solved in order to incorporate the multi-lead forecasts of reservoir inflows and any other nonstationarities into the solution procedure. A comparison is made of a heuristic operating policy for the High Aswan Dam with that resulting from a steady-state stochastic dynamic programming solution and that from the suggested real-time adaptive control formulation. The objective is to minimize losses due to irrigation deficits, power production deficits and damages due to flooding. It can be concluded that performance is better with the steady-state solution, and it is best using the adaptive formulation. Particularly, the use of forecasts and the adaptive formulation significantly reduces flood damages.