A Dual Decomposition Based DMPC for Networked Systems with Varying Topology

Abstract

Economical and technological advancement contributes to the emergence of large-scale systems in industrial field, which are considered as networked systems consisting of several interacted subsystems. For this class of industrial processes, when factors such as variation of materials in production exist, subsystems may be required to be removed from or inserted into the current processing system, which leads to the change in topology of the networked system. This can bring about the in-feasibility of the controller which is designed for previous system. For this problem, this paper proposes a distributed model predictive control (DMPC) algorithm to allow the immediate removal or plugging-in of specific subsystems, where decision variables steady state and steady input are included in the optimization problems to expand the feasibility region and will be optimized along with the predictive input values. At each time instant, the computation for the actual input is conducted by solving max-min problems based on dual decomposition, which ensures that the optimization problem is computed in a distributed manner with neighbor-to-neighbor communication. Here the optimization problem is solved in an iterative manner for the global performance. A numeric example is provided to illustrate the effectiveness of the algorithm.