Optimizing strategies for short-term hydropower scheduling using a blackbox optimization framework

This article examines the selection of optimization models for short-term hydropower scheduling under inflow uncertainty, without solving each model in advance. Usually, deterministic or stochastic formulations are used to schedule operations over the planning horizon. The model selection problem is formulated as a blackbox optimization approach to identify the most appropriate model based on observed inflow conditions. The deterministic model is a mixed-integer linear programme using median daily inflows, whereas the stochastic model is a multistage formulation that captures inflow uncertainty through a scenario tree with stage aggregation. Models are tested over a 10-day rolling horizon for a system consisting of two powerhouses, each with five turbines. Computational results show that inflow characteristics affect model performance. When the blackbox optimization is used, the maximal difference in energy production compared to real realizations of inflows is 0.68%. This highlights the potential of model selection to improve scheduling decisions in hydropower operations.