Controlled risk management of hydroelectric reservoirs

Because of uncertainty on inflows, a hydroelectric system may at any time leave its normal zone of operation and fail. This paper proposes a way to continuously measure this risk of failure by presenting it in the form of time-varying indicators. These indicators are very flexible and could be defined according to the manager's needs. They are calculated by stochastic simulation by using inflows scenarios built according to the observed conditions and the statistical behaviour of the river. The utility of this dynamic risk is highlighted by an application on the Cabonga reservoir located in Quebec, Canada. For this reservoir, we define and evaluate six hypothetical management rules. Four of these rules are explicitly related to the instant level of risk on this reservoir. Opposition between risk and system performance was highlighted in a quantitative way.