Hydropower, as one of renewable energies, has been widely used all over the world. The uncertainties such as reservoir inflows and electricity price cause random changes in the output power and the hydropower generation benefit. Thus, it is important to research on the power prediction of hydropower station considering the uncertainties. This study proposes a multi-time-scale power prediction model of hydropower station based on dynamic Bayesian network theory, considering the uncertainties of reservoir inflow, electricity price, and hydropower consumption rate. The proposed model consists of three components: a multi-time-scale coupling operation (MCO) model, a dynamic Bayesian network (DBN) model, and a probability-based prediction (PBP) model for decision making. The MCO model provides training data inputs for the DBN model, which is established based on expert knowledge and the relationships among the uncertainties. The PBP model performs power prediction of the hydropower station for decision making using the trained DBN. We apply the proposed model to the Tankeng hydropower station in China. The results show that the model not only quantitatively predicts the multi-time-scale output power and benefit of the hydropower station considering the uncertainties, but also provides the risks of power generation deficiency and power output deficiency.
Keywords: Hydropower station; Reservoir operation; Risk analysis; Uncertainty assessment.
Copyright © 2019 Elsevier B.V. All rights reserved.