Self-Tuning Controller with Model Free of Multivariable Neural Network System Identification NARMA Model
Regarding the multivariate NARMA model,it is transformed into subsystems with coupling.A multi-variate tight-format dynamic linearization method with auxiliary variables is adopted to approximate the multivariate NARMA model,and BP neural network is utilized to identify its parameters.A self-tuning controller with model free of neural network identification of multivariable NARMA model is developed by basing on generalized multivariable object function,and the self-tuning controller is nonlinear equations with control input.The nonlinear equations are solved by Newton method of nonlinear numerical analysis.Online learning of connection weight value of BP neural network is conducted by nonlinearity recursive least squares method.Simulation results indicate that the system re-sponse has excellent performance.
NETL
NSTL
万方数据
