| Summary: | 博士 === 國立成功大學 === 工程科學系 === 88 === The study and design of induction motor controller based on a neural network is investigated in this dissertation. By means of a microprocessor, indirect field-oriented control method and some speed control strategies, the proposed robust control schemes, based on a neural network, focus on improving the dynamic behavior to achieve a high performance induction motor drive system. Our research consists of three main parts as expressed in the following.
First, this dissertation proposes a self-tuning PI speed controller with a load torque observer and feed-forward compensation based on neural network identification for an induction motor. The widely used projection algorithm is used as the learning algorithm for this network, to minimize the difference between the motors'' actual response and that predicted by the neural estimator. The proposed neural estimator uses learning to automatically adjust the PI speed control parameters on-line, thereby achieving a desired reference trajectory. The load torque observer can estimate the load disturbance and feed-forward compensation such that the speed response of the induction motor is robust against the load disturbance.
Next, we propose a model reference neural network controller for induction motor speed control based on a neural network plant estimator (NNPE) and a neural network PI controller (NNPIC). The NNPE is used to provide a real-time adaptive identification of the unknown motor dynamics. The NNPIC is used to produce an adaptive control force so that the motor speed can accurately track the model reference output. The widely used back-propagation algorithm and projection algorithm are used as the learning algorithms, respectively, in order to automatically adjust the weight parameters of the NNPE and NNPIC. To guarantee convergence and for faster learning, an approach that finds the boundaries of the adaptive learning rate was developed. The proposed robust control scheme can achieve a fairly good command tracking and regulating response to system parameter variations and external load torque disturbance.
Because the inverter current switching control, measurement and the coupling characteristics of the mechanical shaft will produce some time delay in the induction motor drives, it is therefore more significant and reasonable to consider the time delay effects, parameter variations and load torque disturbance for robust induction motor control. This dissertation proposes a robust speed control for the induction motor drive with time delay. A time delay compensator (TDC) with a neural network controller was augmented to set the time delay element be effectively moved outside the closed loop such that the stability of the robust controlled system is enhanced. Furthermore, the model reference robust speed control and self-tuning robust speed control schemes, based on a neural network, were proposed to enhance the robustness of the induction motor drive with time delay.
Finally, the proposed control scheme is superior to the conventional control method for induction motor drive speed control. The proposed control scheme was implemented with an experimental verification. Both computer simulations and experimental results have demonstrated that the proposed control scheme agrees with the theoretical analysis and can improve the induction motor drive performance with robust characteristics against the system parameter variations and external load torque disturbance.
|
|---|
