Decomposition and Distributed Control of Integrated Lumped and Distributed Parameter Process Networks

This work considers the problem of achieving model-based plant-wide control of a prototypical process network comprising interconnected lumped and distributed parameter systems. A community detection algorithm is used to obtain an optimal decomposition of this network for distributed control. The community detection is performed on a novel graph representation of the dynamics of the network, which accounts systematically for the interconnections among the variables of the process systems, including the different types of variables of the distributed parameter systems. The resulting distributed model predictive control implementation is shown to be computationally tractable, with performance close to that of centralized model predictive control.