The control and regulation of power grids has historically relied upon large-scale prescheduled generation and relatively stable load demand profiles. With the advent of local renewable energy generation technologies as well as the incorporation of load demand response (DR) methodologies, it has become imperative that new distributed control strategies are developed to better regulate the increasingly volatile nature of modern generation and load profiles. In this thesis, I introduce a distributed algorithm called Proximal Atomic Coordination (PAC) to solve for optimal control strategies in distributed power grids, a problem called Optimal Power Flow (OPF). Using a convex relaxed variant of OPF, we show that PAC exhibits sublinear convergence to the optimal ergodic cost, and linear convergence to the OPF solution in the absence of local constraints. We demonstrate our results on various power grid topologies with large levels of renewable energy penetration and DR, and show that PAC converges to optimal control profiles in these scenarios. We further show that in certain regimes PAC outperforms the standard distributed 2-Block ADMM algorithm, and we discuss the benefits of using PAC over 2-Block ADMM and other standard distributed solvers.
Presenter’s Affiliation: Active Adaptive Control Laboratory
Thesis Supervisor: Dr. Anuradha Annaswamy, Senior Research Scientist