Study On Intelligent Control Technique For Induction Machine Based On Parameter Identification And Compensation Control

The speed regulating system of vector control on induction machine is chosen as the object of study in this dissertation. The main aim of study is to improve the performances of the system in dynamic responses and robust. Starting with the dynamic mathematical model study in the conditions of parameter variations, the study is deeply done on the control structures and control strategies of parameter identification, fuzzy-neural control, neural control, nonlinear control and adaptive control. In real time control, the complexity degree of online parameter identification algorithm directly affects the performances of real time system and dynamic responses. The mathematical models in the condition of parameter variations are not only used to analyze the effects of parameters on system performance, but also play the important role in improving the celerity and veracity of identification algorithms. Therefore, the dissertation presents the mathematical models considering rotor resistance and inductance variation the state increment method. Because the models have the formats of Laplace transfer function, they have the advantages of simple structure, clear concept and feasible application, which supply some new and simple methods to study the vector control of induction machine. Based on the above research, some identification methods of rotor resistance, moment of inertia, viscous coefficient and load torque are presented by using neural network. The projects for the compensations of rotor resistance variation and load torque disturbance are covered. In order to realize good tracking performance, this dissertation also establishes a non-linear controller and adaptive system based on a new identification method of rotor resistance variation and load torque disturbance with non-linear control theory. A novel neural network control strategy with torque current teacher controller and target function is proposed. In addition, the new control system including neural network controller and rotor resistance variation compensation is also built. The control algorithms of the system for both torque current target function and motor speed target function are studied. The computer simulation results parameter variations and dynamic responses show that the system has strong robustness and good dynamic performances. After the research of neural network controller, we deeply study the fuzzy-neural network controller and combine it with compensation unit of rotor resistance variation to form the intelligent control system. The controller has the regulation functions of the center point parameters and width parameters of membership functions to improve the learning ability. The simulation results demonstrate the validity and good robustness of this scheme. In order to study above controllers, the methods using torque current target function are compared with the traditional method using motor speed target function under both neural network controller and fuzzy-neural network controller. Four control algorithms are proposed and discussed. Both theoretical analysis and computer simulations prove that the intelligent controllers using torque current target function have better performances in dynamic accuracy.