Multi-purpose reservoirs often have to be managed according to conflicting objectives, which requires efficient tools for trading-off the objectives. This paper proposes a multi-objective simulation-optimisation approach that couples off-line rule curve optimisation with on-line real-time optimisation. First, the simulation-optimisation framework is applied for optimising reservoir operating rules. Secondly, real-time and forecast information is used for on-line optimisation that focuses on short-term goals, such as flood control or hydropower generation, without compromising the deviation of the long-term objectives from the optimised rule curves. The method is illustrated for optimisation of the Hoa Binh reservoir in Vietnam. The approach is proven efficient to trade-off conflicting objectives. Selected by a Pareto optimisation method, the preferred optimum is able to mitigate the floods in the downstream part of the Red River, and at the same time to increase hydropower generation and to save water for the dry season. The real-time optimisation procedure further improves the efficiency of the reservoir operation and enhances the flexibility for the decision-making. Finally, the quality of the forecast is addressed. The results illustrate the importance of a sufficient forecast lead time to start pre-releasing water in flood situations.
Keywords: flood control, hydropower, multi-objectives, real-time optimisation, reservoir optimisation, rule curves