Assessment of parameter uncertainty in autoregressive streamflow models for stochastic long-term hydrothermal scheduling

Hydrothermal systems optimal scheduling requires the representation of uncertainties in future inflows in order to hedge against adverse future low inflows by committing thermal plants, and also to store water in reservoirs while avoiding spillage when high future inflows occur. Stochastic optimization technics has been widely used as a tool for long-term hydrothermal scheduling. These models rely on Monte Carlo simulation in order to capture the inflow uncertainty during the planning horizon. Since the parameters of these models are typically estimated from historical data, it is not surprising that the actual performance of a chosen reservoirs strategy often significantly differs from the designer´s initial expectations due to unavoidable modeling ambiguity. The objective of this work is to assess the impact of inflow parameter uncertainty on the stochastic hydrothermal scheduling. The results presented in this work may be useful for the improvement of stochastic optimization techniques. The results presented show that the uncertainty on the parameters of the stochastic model consists on a supplementary source of risk that should be taken into account in the scheduling model.