Optimal scenario tree reduction for stochastic streamflows in power generation planning problems

摘要: The mid-term operation planning of hydro-thermal power systems needs a large number synthetic sequences to represent accurately stochastic streamflows. These are generated by periodic autoregressive model. If the is too big, optimization problem may be difficult solve. To select small set representing process well enough, this work employs two variants Scenario Optimal Reduction technique. first variant applies such technique at last stage tree defined priori for whole horizon while second combines stage-wise reduction, preserving structure, with resampling. Both approaches assessed numerically on hydrological real configurations Brazilian system.