Background and Objectives: In this paper, a constrained cooperative distributed model predictive control (DMPC) is proposed. The proposed DMPC is based on linear adaptive generalized predictive control (AGPC) to control uncertain nonlinear large-scale systems.
Methods: The proposed approach, has two main contributions. First, a novel cooperative optimization strategy is proposed to improve the centralized global cost function of each local controller. Second, using the proposed linear distributed AGPC (DAGPC), the mismatch between linearized and nonlinear models is compensated via online identification of the linearized model in each iteration of optimization.
Results: The proposed novel cooperative optimization strategy decreases the computational burden of optimization process compared to conventional cooperative DMPC strategies. Moreover, the proposed linear DAGPC decreases the satisfaction time of the terminal condition compared to conventional DMPC methods. The paper establishes sufficient conditions for the closed-loop stability. The performance and effectiveness of proposed method is demonstrated through simulation of a quadruple-tank system for both certain and uncertain situations. The imposed uncertainty changes the system from minimum phase to nonminimum-phase situation. Closed-loop stability and proper convergences are concluded from simulation results of both situations.
Conclusion: Most important advantages of proposed linear cooperative DAGPC are its less design complexity and consequently less convergence time compared to fully nonlinear DMPC methods, due to its online identification of the linearized model.