Author(s)

A. Turgeon

Abstract

This paper shows that the optimal (or near-optimal) operating policy of a reservoir feeding a hydroelectric powerplant can be determined with an optimal rule curve. The method is based on the following fact: raising the level of a reservoir feeding a powerplant is profitable as long as the gain due to the higher head is greater than the loss due to the additional spillage. There consequently exists a reservoir level t S at time t at which the expected energy generation is maximized. The optimal rule curve is the curve that links the optimal levels 1 2 , ,..., T S S S . The paper also shows how to use this curve to operate the reservoir. The problem is more complicated for several reservoirs in series since the optimal level of each reservoir is a function of the levels of all the other reservoirs. The paper shows how to solve the problem. Keywords: multi-reservoir operation, rule curves, stochastic dynamic programming, synthetic inflow scenarios. 1 Introduction The problem of determining the optimal weekly operating policy of several hydroelectric powerplants in series has been dealt with in the past, but always in an approximate way. So far, nobody has succeeded in finding a global feedback solution to the problem, i.e., an operating policy that depends on the content of each reservoir and its inflows. Arvanitidis and Rosing [2], for instance, built an aggregate model for all the installations on the river so as to have only one reservoir to manage. The optimal operating policy of the aggregate reservoir, determined with Stochastic Dynamic Programming (SDP), gives the total amount of potential energy to release from the reservoirs each week, but not the

Keywords

multi-reservoir operation, rule curves, stochastic dynamic programming, synthetic inflow scenarios.