Abstract

The extensive integration of renewable generation in electricity systems is significantly increasing the variability and correlation in power availability and the need for energy storage capacity. This increased uncertainty and storage capacity should be considered in operational decisions such as the short-term unit commitment (UC) problem. In this work, we formulate a day-ahead UC problem with energy storage, considering multistage correlated uncertainty on renewables' power availability. We solve this multistage stochastic unit commitment (MSUC) problem with integer variables in the first stage using a new variant of SDDP that can explicitly deal with temporal correlations. Our computational results on the IEEE 118-bus system demonstrate the significance of considering multistage uncertainty and correlations, comparing our solution with other multistage solutions, two-stage solutions, and deterministic solutions typically used by industry. We also solve the MSUC problem for a representation of the Chilean power system, finding superior UC solutions for scenarios where adapting generation to the unfolding uncertainty is costly. Finally, we demonstrate that the MSUC approach can be used to define a more efficient deterministic UC solution, outperforming the current industry practice.