| Summary: | 碩士 === 國立臺北科技大學 === 機電整合研究所 === 93 === In this thesis, the Neural Network PI Controller, designed with neural network and the projection algorithm in adaptive theory, is proposed. The controller on-line adjusts the PI parameters, and increases the adaptation ability and dynamical performance of the considered system. The drawbacks of the traditional fixed-parameter PI controller are overcome.
Based on the direct torque control (DTC), the speed control algorithm is implemented for induction motors. The merits are the fast dynamic response, simple structure and uncomplicated calculation. The flux and torque controllers of hysteresis type are not used in this thesis. In stead, the space voltage vector modulation technique is adopted to solve the torque ripples and noise problems. In addition, this research utilizes the speed estimator to realize the speed sensorless control. According to the stator flux of d-q axis estimated by the voltage type flux estimator, the stator flux angle can be found. Consequently, the electrical angular frequency of induction motor can be obtained by the speed estimator. The advantages of speed sensorless control are cost-effectiveness and retaining the robustness of the motor structure.
The experimental result shows that the NNPIC outperforms the traditional PI controller in dynamic response, when the speed is controlled in the range from 100rpm to 1800rpm with 8 Nm load.
|
|---|
