Robust Optimization of Reservoir Operation Considering Ensemble Inflow Uncertainty

The application of optimization model to tackle reservoir operation problems has been extensively studied since the Harvard Water Program in the 1960s. However, most of the studies considered deterministic optimization models in which some variables such as future inflows are treated deterministically. Robust Optimization (RO), a more recent approach to optimization under uncertainty, seeks for a solution that is optimal for any realization of the uncertainty. The characteristic of RO is that it does not need to know the probability distribution of uncertain parameters, but only the uncertain set of uncertain parameters.

This study developed an RO model for reservoir operation and applied to the short-term operation of Qingshitan reservoir in the Lijiang River of China. Inflow to Qingshitan reservoir, the main source of uncertainty, is described by a set of inflow forecast, i.e., ensemble inflow. The ensemble inflow is obtained by using ensemble precipitation prediction as input of a rainfall-runoff model. Because of the complexity of ensemble forecasting information, the spectral optimization model (SOM) is used to reduce the dimension of the uncertainty. Then the RO model of Qingshitan reservoir operation is developed considering inflow as uncertain parameter, and the appropriate size and shape of the uncertain set are selected according to the characteristic of the uncertain parameter. The RO model controls the distribution of objective function with mean and variance while satisfying operational and physical constraints. In order to make the optimization tractable and easy to calculate, the robust optimization is transformed into the robust counterpart. The model is solved using genetic algorithm. The results show that robust optimization can greatly improve the robustness of Qingshitan reservoir operation compared with the traditional optimization model.