Author(s)

S. Krau, M. Latraverse, D. Tremblay & A. Turgeon

Abstract

Meteorological forecasts for the next few days are usually not detailed enough to determine the exact probability distribution of the forecast precipitation, and, hence, the probability distribution of the forecast inflow into the hydrological system. However, the hydrological system manager is responsible for assessing the risk of flooding and violating constraints when operating the reservoirs. This paper deals with the problems involved in modeling forecast errors and of determining an operating policy that takes such errors into account. This is accomplished by first building an inflow scenario tree that takes forecast errors into account and then by solving the reservoir management problem with that tree. This paper describes how the tree was built. Keywords: meteorological, forecast, hydrological, management, risk, flooding, daily, stochastic, optimisation. 1 Introduction The management of a hydrological system often implies the pursuit of various divergent objectives. There is a need to optimize power plant generation but also to regulate the flow of rivers, ensure a sufficient supply of drinking water, comply with vacationing-related constraints and ensure the safety of waterside communities as well as that of the structures. Management models most often take these different objectives into account through flexible constraints, i.e. constraints for which violation is tolerated, while minimizing the overall costs incurred by such potential violations. Daily management can then be reduced to a stochastic optimization problem related to a single objective. It consists in

Keywords

meteorological, forecast, hydrological, management, risk, flooding, daily, stochastic, optimisation.