Distributed Optimization Based Control on the Example of Microgrids

Abstract

Model Predictive Control (MPC) is nowadays one of the most successful advanced process control methodologies and is used in a wide range of applications. While originally limited to processes with slow dynamics and a limited number of states, the applicability of MPC schemes increased dramatically over the past years due to the performance of modern microchips and the concurrent advancements of mathematical optimization, in particular, distributed optimization. In this paper, we outline the ideas of distributed optimization schemes embedded in MPC implementations on the example of the dual ascent algorithm and the alternating direction method of multipliers. The performance and the properties of the resulting distributed optimization based control schemes are illustrated on the example of a network of distributed energy systems. In particular, the overall power demand of the network is optimized by using flexibilities resulting from distributed storage devices and controllable loads.