Online-data-based pattern recognition of transient responses and suboptimal iteration of control parameters

Technically, parametric performance evolution of control systems can be decomposed as two sequential components: online-data-based pattern recognition of transient responses and iteration of control parameters. This paper proposes a novel reactive knowledge-based method for parametric performance evolution of control systems. First, a novel knowledge-based system is developed for online-data-based pattern recognition of transient responses of control systems. Then, a quasistate equation, with features of transient responses as quasistates and forward increments of control parameters as quasicontrol variables, is proposed and structured, and the iteration of control parameters is formulated as and solved through a linear quadratic quasistate tracking problem. Moreover, a set of heuristic rules are designed to supervise the utilization of the rolling-optimization solution to the linear quadratic quasistate tracking problem. Finally, a variety of simulations were conducted to demonstrate the reasonableness of the proposed concept of parametric performance evolution of control systems, and the feasibility and effectiveness of the developed algorithms