Implementation of predictive control algorithms on networked control systems

Programmable logic controllers (PLC) with network interfaces have become standard industrial control devices. However because of their relatively low computation power they cannot be used as standalone devices for implementing complicated control algorithms such as predictive control. The work presented in this paper addresses the issue of implementing these algorithms by distributing their computations to more than one such devices which can communicate between each other over an industrial network. Specifically, the computations of the predictive control algorithms that can be executed in parallel are identified. The way these computations can be scheduled to run in parallel on different architectures of control devices are studied. The feasibility of realizing the execution of the algorithms on the studied architectures is verified by presenting the results of the experimental realization of an application on a test bed facility which is a network of 12 PLCs and a computer-based plant simulator built according to the Profibus-DP standard