本研究之小腦模型控制器結合適應性理論之投射演算法(pojection algorithm)並以小波函數作為歸屬函數之模糊小腦模型PI控制器(Fuzzy Cerebellar Model Articulation PIcontroller, FCMAPIC),此模糊小腦控制器具線上調適PI參數值、快速學習以及結構簡單等優點,可以改善傳統PI控制器參數固定的缺點。 本研究採用直接轉矩控制(Direct Torque Control, DTC)理論實現感應電動機調速控制,以達到動態響應快速,架構及計算過程簡單等優點。為了改善傳統遲滯型轉矩及磁通控制方式因遲滯頻寬的不當選擇而產生轉矩漣波及噪音等缺點,因此直接轉矩控制方式採用空間電壓向量調變技術。另外,本研究藉由轉速估測器來實現無轉速量測器控制,除了可以節省成本之外還可以避免破壞馬達結構。 經實驗結果證明,於負載8Nm,轉速控制範圍為36rpm至1800rpm時,模糊小腦模型PI控制器相較於類神經網路PI控制器在轉速動態響應都有明顯的改善,可以達到感應馬達精密控制之目的。
In this thesis, the Fuzzy Cerebellar Model Articulation PI Controller (FCMAPIC) is proposed, which was designed with the fuzzy cerebellar model articulation controller on the basis of the wavelet function for membership functions, and the projection algorithm in adaptive theory was adopted to tune the parameters. The advantages of the FCMAPIC include on-line PI parameters adjustment, rapid learning ability, and simple structure. The drawbacks of the conventional fixed-parameter PI controller are overcome. Based on the direct torque control (DTC), the speed control algorithm was implemented for induction motors. The merits of DTC include fast system response, simple structure and less computation burden as compared with field-oriented control. In order to solve the problems of torque ripple and noise caused by the inadequately designed conventional hysteresis flux and torque control, the space voltage vector modulation (SVPWM) technique is used in this research. In addition, this research utilizes the speed estimator to realize the speed sensorless control to keep the cost-effectiveness and retain the robust of the motor structure. The experimental results show that the FCMAPIC achieved better dynamic response than NNPIC under the operation conditions that speed range from 36rpm to 1800rpm with 8-Nm load.