本論文將模糊概念導入至小腦模型PI控制器中以設計一模糊小腦模型PI控制器,此控制器不但保留原先小腦模型PI控制器快速學習、結構簡單和非線性之學習能力等優點外,在控制上亦具有模糊控制之特性。此外,在學習架構上加入了投影演算法,來達成線上學習及即時控制之效果,同時為了加快其誤差的收斂速度,本研究依權重的可信程度來分配誤差。 在向量控制架構上,本研究採用直接轉子磁場導向控制,並根據適應性控制法則設計出一虛擬降階型磁通估測器來調適感應馬達之變動參數。為了達到無轉速量測器控制與適應轉子電阻的變化,在此利用小腦模型PI控制器作為其轉速與轉子電阻的估測。另外,本研究加入高效能控制於穩態下之向量控制系統中以減少電流的損失及提高運轉效率。 經由實驗結果證明,此一模糊小腦模型PI控制器植入至感應馬達向量控制系統中,轉速範圍在36rpm~2000rpm,負載為8Nm時之動態響應,其各轉速皆能快速追蹤及保有良好的強健性,尤其在低轉速,亦能獲得優異之控制特性。同時由高效能控制實驗證明,系統確實有達到省能之目的。
In this thesis, a Fuzzy-CMAPIC (FCMAPIC) and the corresponding speed control of induction motor (IM) are proposed. The controller is integrated by the cerebellar model articulation PI controller (CMAPIC) and the fuzzy logic concept. The advantages of the FCMAPIC include rapid learning ability, simple structure, and non-linear function learning capability. With the aid of projection algorithm, the FCMAPIC can effectively achieve the on-line learning and real-time control. Moreover, in order to accelerate the convergence rate of FCMAPIC tuning, the credit assignment (CA) technique is adopted in this research. As to the vector control, the motor drive adopts the direct field orientation control (DFOC). To modify the control actions in presence of the IM parameter variation, the pseudo-reduced-order flux observer (APRO) is designed by using adaptive control law. In order to establish the speed sensor-less control and overcome the drift effect of rotor resistance, the CMAPICs are also used to estimate speed and rotor resistance separately. In addition, to reduce the loss of current and increase the efficiency during the steady-state operation, the high efficiency control is embedded into the system. Under the operation conditions that the speed range varies from 36rpm to 2000rpm with 8-Nm torque load, experimental results indicate that the speed response has fast robust tracking ability, especially under low-speed operations. Also, the power-saving property is examined by the experiments of high efficiency control.