On Generation of Scenario Trees for Stochastic Optimization

Fredrik Lindell

Abstract:: This thesis is part of a project aimed towards an accurate description and construction of an optimization model for the seasonal planning problem in a hydro-thermal system for production of electrical energy. The problem is the one facing a price-taking, risk-aversive producer who wants to maximize expected profit over a finite planning horizon where the randomness of inflows and prices is considered by formulating a stochastic programming model. This rapport focuses on the generation of scenario trees used in the stochastic optimization program regarding the Swedish hydro-electric power market. The goal is to create a valid model for the inflows to the dams along Ume älv together with scenario trees containing relevant properties of this multidimensional stochastic process.