本計畫之目標為研製基於人工智能模糊類神經網路之智慧型高效能內藏式永磁同步馬達驅動系統，以發展每安培最大轉矩控制、弱磁控制及每伏特最大轉矩控制來提高效能。第一年首先發展利用人工智能模糊類神經網路之內藏式永磁同步馬達驅動系統，並提出利用非對稱歸屬函數之派翠機率模糊類神經網路估測轉動慣量。為了抑制永磁同步馬達因為耦合不良及機械摩擦等產生之振動，本年度亦提出基於離散小波濾波器之共振頻率偵測架構，並利用帶通濾波器掃頻找出系統之共振頻率。由於內藏式永磁同步馬達參數會因為溫度、磁飽和等外來影響產生非線性變化，故第二年將設計基於遞迴式勒壤得模糊類神經網路之每安培最大轉矩控制器，利用遞迴式勒壤得模糊類神經網路得出電流角命令，減少磁飽和造成的影響。第三年度將進一步控制直軸電流命令進而達到每安培最大轉矩控制，再進行弱磁控制及每伏特最大轉矩控制，並利用第二年發展之遞迴式勒壤得模糊類神經網路來估測交軸電感值，以代入每安培最大轉矩和每伏特最大轉矩公式中，減少磁飽的影響。此外，於速度控制迴路將採用適應性互補式滑動模式控制器以改善速度響應。 ;The objective of this project is to develop an intelligent high-performance interior permanent magnet synchronous motor (IPMSM) drive system based on artificial intelligence using fuzzy neural network. To increase the performance of the IPMSM, the maximum torque per ampere (MTPA) control, field-weakening (FW) control and maximum voltage per voltage (MTPV) control will be developed. In the first year, the IPMSM drive system based on artificial intelligence using fuzzy neural network is developed and the identification of moment of inertia using a Petri probabilistic fuzzy neural network with an asymmetric membership function (PPFNN-AMF) is proposed. Moreover, in order to overcome the vibration of PMSM due to the poor coupling and mechanical friction, a resonance frequency detection system based on the discrete wavelet filter is also proposed in this year. Then, a band-pass filter (BPF) is adopted using sweep frequency to find out the resonance frequency of the system. Since the performance of IPMSM will vary nonlinearly owing to external influences such as temperature and magnetic saturation, the MTPA controller based on the recurrent Legendre fuzzy neural network (RLFNN) will be designed in the second year. The current angle command is obtained by the RLFNN to alleviate the effect of magnetic saturation. Furthermore, in the third year, the d-axis current command will be controlled to achieve the MTPA control first. Then, the FW control and MTPV control will be presented. In addition, the q-axis inductance is estimated by the RLFNN developed in the second year, and it is substituted into the formulas of MTPA and MTPV to alleviate the saturation effect. Additionally, an adaptive complementary sliding mode controller (ACSMC) will be developed in the speed control loop to improve the speed response.
