Hydrological Scenario Reduction for Stochastic Optimization in Hydrothermal Power Systems

Most of the methods developed for hydrothermal power system planning are based on scenario-based stochastic programming and therefore represent the stochastic hydro variable (water inflows) as a finite set of hydrological scenarios. As the level of detail in the models grows and the associated optimization problems become more complex, the need to reduce the number of scenarios without distorting the nature of the stochastic variable is arising. In this paper we propose a scenario reduction method for discrete multivariate distributions based on transforming the moment-matching technique into a combinatorial optimization problem. The method is applied to hydro inflow data from the Chilean Central Interconnected System (SIC) and is benchmarked against results for the optimal operation of the SIC determined with the selected subsets and the complete set of historical hydrological scenarios. Simulation results show that the proposed scenario-reduction method could adequately approximate the probability distribution of the objective function of the operational planning problem.