Neural Network Theory And Sliding Mode Control For Induction Motor System

Induction motors are a class of nonlinear multivariable uncertain strong-coupling hybrid systems subject to disturbances, which constitutes a benchmark example for nonlinear control. Many control programs have been proposed for the fast-changing nonlinear complex systems, but there are still some shortcomings in those programs. Therefore, it is necessary to develop a practical and efficient controller for the fast-changing complicated industrial systems.Based on the analysis of the mathematical model of induction motor, this paper addresses the problem of tracking control of flux linkage and angular speed.First, because sliding mode variable structure is a nonlinear control method, which is adaptive to disturbance and parameter variations. Especially, SMVS control is stable and robust against system uncertainties, this thesis proposed a finite time tracking control of induction motor based on TSM controller, we divided the multi-variable, nonlinear and strong coupled induction motor systems into two second-order nonlinear subsystems including the angular speed subsystem and the rotor flux linkage one. The non-singular terminal sliding mode (NTSM) control scheme with finite time convergence is used to the design of the controller. The global stability of the system is guaranteed and the system states can accurately track the reference signal in finite time. Simulation result shows the effectiveness of the control scheme and the good static and dynamic performance by using the proposed method.Second, because of the neural control's nonlinear imaging, self-learning tolerance, parallel processing abilities and so on, neural control strategy is widely employed. This thesis introduces an adaptive inverse RBF neural network based on control architecture, and synthesizes the PID controller, the controller is consist of two RBF neural networks, one is applied to identify the controlled object, and the other approximates the plant inverse transfer function, the neural network output suppress parameter variations and resistance perturbation (including the end effect force produced by of linear motor) greatly, this control scheme cannot only enhance track-command ability and the robustness of the system, but also has strong robustness to parameter variations and resistance perturbation.Finally, with the capability of self-learning and self-adaptability of the single neural network, an adaptive PID controller is presented, and then combines with the identification of the RBF neural network, this paper also presents a novel approach of single neuron PID model reference adaptive control based on radial basis function(RBF)neural network on-line identification. The simulation results show that this method has good performance in many cases, such as adaptability and robustness, and that its simple structure is easy to be constructed.