Distributed cooperative nonlinear economic MPC ∗

Model predictive control (MPC) is a design technique that solves (typically on-line) optimization problems to determine a suitable feedback action. It is widely used because of the capability to deal with constraints and MIMO systems [3]. To treat practical industrial processes with MPC, use of so called economic cost functionals are proposed and considered for optimization in [1]. In this scheme, which is called economic MPC to differentiate from standard MPC, the economic cost function is not necessarily minimum at the desired steady state. The Lyapunov-based analysis first achieved in [2] and later extended in [1], deals with a centralized control scheme. However, implementation of economicallybased control algorithms is also of great interest in large-scale systems for which presence of a centralized optimization solver is not ideal. In [3] control methods for distributed systems are classified into 4 categories according to two criteria, the extent of sharing information and the type of cost function. These go under the name of centralized, decentralized, noncooperative and cooperative control. In this paper we focus on the cooperative distributed MPC schemes. Cooperative MPC for linear systems with convex objective functions and nonlinear systems with nonconvex objective functions are dealt in [4] and [5] respectively. Cooperative suboptimal MPC for linear distributed systems with convex economic cost functions is suggested in [6]. Similary to the standard cooperative dis-