Effect of Inflow Class Selection on Multi-Objective Reservoir Operation Using Stochastic Dynamic Programming

Abstract

A combined simulation optimization model was developed to derive optimal operational policies for a multi-objective reservoir in a semiarid environment. Stochastic dynamic programming was selected as the optimization technique. Different scenarios were considered to optimize the current operational policy as well as different possible future uses of the reservoir. The first scenario was used to maximize hydropower production within the framework of the current operational policy. Other scenarios were used to address multiple objective operation of the reservoir including hydropower, agricultural, and domestic uses. Generated policies for all the scenarios were simulated in real time using historical inflow data for the Qar’awn reservoir within the Litani Basin in Lebanon. Sensitivity analysis on number of inflow classes was performed. Results showed that the newly derived policies decrease failure by a range of threefold to sixfold and improve hydropower production by more than 15 %. The model was able to derive policies that decreased system failure and shortages to less than 10 %. Best inflow classes were found to be in the range of 3–5.