Combined SDDP and simulator model for hydropower scheduling with sales of capacity

In following work, we investigate the importance of detailed hydropower scheduling modelling when including sales of capacity, which adds complexity that is not easily incorporated in a Linear Programming (LP) problem. In the proposed approach, we use the profit-to-go function obtained from a Stochastic Dual Dynamic Programming (SDDP) scheduling-model in a Simulator Model, based on Mixed Integer Programming (MIP), and perform detailed simulations. The Simulator Model allows a more complex problem description, than by the LP formulation in the SDDP model. The Simulator Model may therefore be used to give an estimate of the LP approximation, which is used for providing the opportunity cost in short-term hydropower scheduling models or conceivably for making investment decisions. For the given case study, the expected profit from selling capacity was 29.2% higher than the linear SDDP Model to the Simulator Model. The overall profit loss was reduced by 0.93%, quantifying the overestimation of profit in the SDDP Model. This illustrates the importance of detailed modelling when considering sales of capacity.