Distributed model predictive control based on the alternating directions method of multipliers applied to voltage and frequency control in power systems

Abstract

We present a straightforward way to solve a model predictive control problem for a power network system given as a nonlinear differential-algebraic equation (DAE) in a distributed way using the consensus alternating directions method of multipliers (consensus ADMM) algorithm. While no convergence- or stability results are available for fully nonlinear DAE models, this gives unprecedented experimental evidence that power network systems of the presented structure allow to be controlled in this way, unlocking the numerous combined advantages of distributed and predictive control schemes in the context of energy distribution networks, as well as broadening the field of use for the consensus ADMM algorithm to nonlinear DAE models.