Non-anticipative risk-averse analysis with effective scenarios applied to long-term hydrothermal scheduling

We deal with long-term operation planning problems of hydrothermal power systems by considering scenario analysis and risk aversion. This is a stochastic sequential decision problem whose solution must be non-anticipative, in the sense that a decision at a stage cannot use knowledge of the future. We propose strategies to reduce the number of scenarios in such way that the decision obtained by solving the non-anticipative risk-averse problem considering the subset of effective scenarios is as reliable as the decision from the whole set of scenarios. Numerical experiments are presented for validation of the strategies proposed by solving the problem for two test systems with real data extracted of the Brazilian interconnected system.