Towards multi-timescale energy provisioning using Stochastic Dual Dynamic Programming

The operation of the electric grid in systems with a large hydroelectric component is often cast as a dynamic programming problem, in which state variables are reservoir levels. To avoid the curse of dimensionality in discretizing such states, the SDDP technique has been successfully applied. However, new demands are being placed on the optimization, with the penetration of renewable sources of faster variability, and the possible incorporation of shorter term energy storage.In this paper, we present preliminary work on extending the SDDP framework to such two time-scale problems. We apply the method to a stylized model of the Uruguayan system, relying on new open source implementations of SDDP to carry out the computations. Our results indicate that the method remains tractable despite the increased problem dimension.