Dissipativity Based Distributed Model Predictive Control for Process Network Reconfiguration

This article presents a reconfigurable distributed model predictive control strategy based on dissipativity theory to deal with online process network topology change. We mainly address the situation when process units want to join or leave the process network during closed-loop operation, while maintaining the plant-wide stability. The capacity of capturing the effects of process network topology on plant-wide stability is one key feature of dissipative systems theory, which can be utilized to deal with the challenge of performing online process network reconfiguration. The redesign is restricted to subsystems whose input-output properties are directly influenced by the topology change to allow quick operation. When the request of process units joining or leaving the process network is sent, feasibility of the process network topology change has to be assessed first. If permitted, the controllers of the influenced subsystems have to be updated for the modified dynamics to maintain the stability constraint of the plant-wide system, which is translated into dissipative trajectory condition for each local controller. The main advantage of the proposed dissipativity based method is that the design procedure is simple and straightforward compared to some existing algorithm based on invariant set, in which a transition phase is required to be designed in advance to ensure recursive feasibility before process network topology change.