Optimal Design of Distributed Controllers for Large-Scale Cyber-Physical Systems

Large Cyber-Physical Systems (CPSs) can consist of thousands of components and can span very large geographical areas. Centralized control of such systems is in general challenging due to the necessity for reliable, fast, and long-distance communication links. Additionally, centralized control raises issues about privacy of data, as well as issues with robustness. For these reasons, controllers which require a smaller number of communication links have been developed, such as distributed and decentralized controllers. However, the design of distributed controllers is in general challenging if system-wide specifications need to be achieved. We review three (sub)optimal controller design methods for distributed controller design of large-scale systems with a special focus towards electric power networks. The first method uses \(\mathcal {H}_{\infty }\) optimization to tune controller parameters of structured local controllers. Furthermore, \(\mathcal {H}_2\) and \(\mathcal {H}_{\infty }\) optimization are considered, respectively, to create distributed controllers focusing on the reduction of communication links while achieving satisfying system performance. We apply these methods to damp oscillations in power systems, one of the most complex existing CPSs today. For this purpose, we first build up the model of the power system and its state-of-the-art controllers. Then, we compare the three methods with respect to the transients after a load step, achieved system norms, as well as the necessary communication structures for the control schemes.