Design of Adaptive Function Coupling Recurrent Cerebellar Model Articulation Controller for Switched Reluctance Motor Drive Systems

• Main Authors: Wu, Jia Jhen, 吳佳甄
• Other Authors: 王順源
• Format: Others
• Language: zh-TW
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/5jzda3

碩士 === 國立臺北科技大學 === 電機工程研究所 === 105 === This thesis proposes an adaptive functional coupling recurrent cerebellar model articulation controller (AFCRC). The AFCRC system contains an integrated error function, a compensating controlle, and a novel cerebellar model articulation controller (CMAC), which is developed according to the concept of recurrent neural network (RNN) and functional coupling NN (FCNN).The proposed controller comprises a functional coupling recurrent CMAC (FCRCMAC) and a Tagaki–Sugeno–Kang (TSK) fuzzy compensator. The AFCRC introduces the Gaussian function into a CMAC; therefore, the control strategy can provide a relatively smooth output control quantity. Furthermore, a RNN structure is used to convert the originally static CMAC into a dynamic controller, and a FCNN structure is used to improve the learning speed of CMAC. Subsequently, the compensating controller compensates for the error between the AFCRC-controlled variable and ideal controlled variable. An analytical method based on the Lyapunov theory is proposed for determining the adaptive learning algorithms of recurrent weights, FCNN weights, Gaussian function mean parameter, and Gaussian function standard deviation to ensure the stability of the AFCRC control system. To verify the performance and effectiveness of the proposed AFCRC system, the study uses the proposed AFCRC scheme for controlling the direct torque control drive system of a switched reluctance motor (SRM), as well as compared it with traditional CMAC and FCMAC. The simulation and experimental results reveal that the system is operated at 100 rpm, 800 rpm, 1000 rpm, 1600 rpm and ±800 rpm with 1 Nm torque load, respectively. The speed error in the transient response speed is less than 20 rpm (in the steady state, the speed error was within ±2 rpm). The root mean square error (RMSE) , maximum speed error and the average speed error are used as a performance indices for comparing the traditional CMAC, FCMAC and AFCRC system. The results show that the proposed AFCRC system has improved robustness against external disturbances. These experimental results reveal that the proposed control strategy is advantageous at various speed commands and improved dynamic responses.