Weighting matrices design for LQG controller based on data-driven economic performance PSO

Aiming at the difficulty of weighting matrices design for LQG controller, a data-driven economic performance PSO algorithm is presented in this paper. The method introduces a subspace approach to obtain the process characteristics and the explicit functional relationship between the manipulation variance and the output variance, which solves the LQG benchmark problem directly. The PSO algorithm based on the results of economic performance assessment subsequently realizes the accurate optimization of the LQG weighting matrices. The proposed approach is illustrated and verified through a simulation example.