Monte Carlo Optimization for Reservoir Operation

Abstract

A methodology that determines a probabilistic reservoir release policy by using Monte Carlo optimization is developed. For a particular streamflow sequence the methodology determines the optimal reservoir release for each time period over the operational horizon. After repeating this procedure for a large number of synthetic streamflow sequences, an operational release policy is determined by utilizing the probability mass function of the optimal releases, conditioned on observable hydrologic conditions. The methodology is applied to the Mad River Basin in northern California to develop monthly reservoir release policies that use reservoir storage and inflow as the observed conditioning hydrologic variables.