Design of Adaptive Recurrent Wavelet Cerebellar Model Articulation Controller for Induction Motor Direct Torque Control Drive Systems

• Main Authors: An-Hsiu Su, 蘇安修
• Other Authors: Shun-Yuan Wang
• Format: Others
• Language: zh-TW
• Online Access: http://ndltd.ncl.edu.tw/handle/4pxc5w

碩士 === 國立臺北科技大學 === 電能轉換與控制產業碩士專班 === 103 === In this thesis, an adaptive recurrent wavelet cerebellar model articulation controller (ARWCMAC) that integrates a diagonal recurrent neural network and a cerebellar model articulation controller (CMAC) as well as wavelet functions is proposed. The ARWCMAC was then applied as a speed controller in an induction motor by using the direct torque control scheme. Because the traditional CMAC taking only two states for mapping the input to association memory, the output is discontinuous and result in the disadvantage of non-differentiation. To improve the shortcoming mentioned above, the continuous wavelet functions were used in the diagonal recurrent cerebellar model articulation controller (DRWCMAC) architecture for mapping the input to association memory. The input layer can take an arbitrary real value between 0 and 1, which increases adaptive capability and smoothens the control amount output of the controller. According to the feedback property of the DRWCMAC, the characteristics of dynamic behavior and data storage were inherited. This dynamic and continuous mapping property can therefore improve the drawbacks of traditional CMAC. Moreover, to ensure system stability, the adaptive rules of the RWCMAC were determined using an analytical method based on a Lyapunov stability theorem, and the gradient descent was used to update the adaptive rules of the recurrent weights. Finally, for the realization of the proposed system, the ARWCMAC, fuzzy flux controller, fuzzy torque controller, and speed estimator were integrated and implemented in a direct torque control induction motor drive system to realize the proposed system. The simulation results showed that the proposed controller achieves superior speed control compared with controllers developed through other methods. After verifying the feasibility of the proposed controller through simulation, the hardware was implemented, and several experimental tests were conducted by induction motors. Experimental results at 8 Nm load torque in a speed range form 36 rpm to 2000 rpm and ±1200 rpm showed that the proposed controller has a superior speed response than traditional CMAC.