Performance evaluation of a stochastic optimization model for reservoir design and management with explicit reliability criteria

Abstract

A seasonal, chance-constrained linear programing model, which facilitates the development of reservoir system designs and operation policies and which incorporates multiple linear decision rules (LDR) conditioned upon the streamflows in other seasons, is evaluated with variations in the numbers of seasons per year and decision rules per season. The performance is also compared with the original, single-LDR model. The motivations for the expected results and the numerical solutions obtained for the examples considered are presented. The multiple-LDR model is shown to be superior to the single-LDR model: for equivalent restrictions on the reservoir operation and performance the multiple-LDR model gives a smaller capacity reservoir than the single-LDR model. Furthermore, as the number of LDR's per season is increased, the required capacity decreases. And when the operating rules defined by the models are tested in a simulation of actual reservoir operation, the multiple LDR's perform more closely to the specifications prescribed in the linear program than the single LDR.