Adaptive Neural Network Control For Stochastic Nonlinear Systems And Its Application In Motor Control

In practical engineering, the systems under consideration maybe time-varying, usually contain nonlinearities, uncertainty and are subject to the influence of external disturbance. Therefore, the study of adaptive neural network control method for stochastic nonlinear systems is very important both in theory and in practice. In the design of the controller, the adaptive method can be used to solve the control design problem of the systems with uncertain parameters. The radial basis function(RBF) neural network can be used to approximate the unknown nonlinear functions, an adaptive neural network controller is thus designed in order to make the system output follow a given reference signal and to ensure that all the signals in the system are bounded. In addition, the permanent magnet synchronous motor(PMSM) is high order, multi-variable, parameter uncertain nonlinear systems with stochastic disturbance. Therefore, the study of adaptive neural control strategy of PMSM has very important practical value. Due to the input saturation often occurring in PMSM, it is very necessary to consider the input saturation problem in the control of the actual system.Compared with the existing stochastic nonlinear system control methods, the controller presented by us needs only one adaptive parameter which greatly reduce the calculation burden, in addition, the proposed control scheme is also simpler to be implemented in practical engineering. The main innovation in this thesis lies in that an adaptive neural controller is designed by taking input saturation into account, and then the existing control strategies of the motor are extended from the deterministic systems to the stochastic systems.The main research content of this thesis is divided into the following two parts:First, the adaptive neural network control for a class of stochastic nonlinear systems is studied. Adaptive method is utilized to estimate the unknown parameters, RBF neural network is used to approximate the unknown nonlinear functions and backstepping technique is utilized to construct the adaptive neural network controller. The stability analysis shows that the proposed control method can guarantee that the output of the system follow the given reference signal well and all the closed-loop signals are bounded. Finally, a simulation example is used to illustrate the effectiveness of the proposed controller.Second, based on the work in the first part, the adaptive neural network tracking control problem of stochastic nonlinear systems is further addressed for the case of PMSM with input saturation. An adaptive neural controller has been constructed. It has been proved that under the action of the proposed controller, the system output can follow the given reference signal well, and all the closed-loop signals are bounded even the input saturation occurs. The simulation results can prove the validity of this control method.