Abstract A distributed model predictive control (DMPC) framework is proposed. The physical plant structure and the plant mathematical model are used to partition the control duties over self-sufficient estimation and control nodes. Linear models and local measurements at the nodes are used to estimate the relevant plant states. This information is then used in the model predictive control calculations. Communication among relevant nodes during estimation and control calculations provides improvement over the performance of completely decentralized controllers. The DMPC framework is demonstrated for the level control of an interacting four-tank system. The performance of the DMPC system for disturbance rejection is compared with other control configurations. The results indicate that the performance of the proposed framework provides significant improvement over completely decentralized MPC controllers, and approaches the performance of a fully centralized design.
[1]
Arthur G. O. Mutambara,et al.
Decentralized Estimation and Control for Multisensor Systems
,
2019
.
[2]
Francis J. Doyle,et al.
Model based control of a four-tank system
,
2000
.
[3]
Bruce H. Krogh,et al.
Distributed model predictive control
,
2001,
Proceedings of the 2001 American Control Conference. (Cat. No.01CH37148).
[4]
Sigurd Skogestad,et al.
Control structure design for complete chemical plants
,
2004,
Comput. Chem. Eng..
[5]
Francis J. Doyle,et al.
A distributed state estimation and control algorithm for plantwide processes
,
2003,
IEEE Trans. Control. Syst. Technol..
[6]
P. Varaiya,et al.
Control and coordination in hierarchical systems
,
1982,
Proceedings of the IEEE.