Two layer optimal control for a class of "gray-box" system, theory and experiment

This paper presents a two layer optimal control scheme with a control layer and an optimize layer. The control layer realize the close-loop control while the optimize layer is used to optimize the control parameters. The optimize layer identify the model of plant on-line and optimize the control parameters based on the identified model. Because the optimize layer is running parallel with controller therefore, the two layer controller is able to adapt plant's slow time-variant. The real system will stable and reliable due to the optimization process is only acting on the identified model and only stable and reliable control parameter will update to the real controller. This study only consider the class of "gray-box" system which plant model is clear except the model parameter is unknown. Both of model parameters identification and control parameters optimization are realized by Particle Swarm Optimization (PSO) algorithm. An experiment study of output voltage control of a second order electronic filter is presented. The experiment results demonstrate that the two layer optimal control scheme has ability of control parameter optimization.