Distributed Model Predictive Control Based on Cooperation

Aimed at a class of large networked systems consisted of some subsystems which are not completely decoupling, with arbitrary bounded noise and constraints, this paper proposes a cooperation-based distributed model predictive control algorithm. Firstly, composite models combined of complete decoupling models and all interaction models are established, according to the interaction of subsystems. Afterwards, the strong convex combination of all the subsystems' objective function found system-Aide objective function to guarantee system-wide coordinated operation, by using composite models. Than, a distributed model predictive control state estimation method is presented, which obtain optimal state estimation by making full use of the known measurement information and interaction of subsystems. This algorithm is implemented in parallel among subsystems, which improves the computational efficiency greatly. The computer simulations verify this algorithm is system-wide feasibility and this estimation method has high accuracy, through comparison with centralized model predictive control.