Abstract
We present an approach for assessing the effect of using scenario tree approximations of limited size to represent the probability distribution of uncertain variables in stochastic hydropower scheduling models. The aim of reducing the size of the tree is to reduce the computation time. The scenario tree approximations are generated as reduced trees consisting of a subset of the original scenarios. The approach is applied to a profit maximization problem for a single-reservoir hydropower system, using inflow as an example of a stochastic variable. We calculate how much the expected profit is reduced when decisions for each time-step are calculated using the different scenario-reduction alternatives for the rest of the planning period. This illustrates thetrade-off between the size of the utilized scenario tree and expected profits.
