| Summary: | 碩士 === 國立臺北科技大學 === 電機工程系 === 107 === Traditional direct torque control (DTC) generally operates with three-phase voltage using the measured three-phase currents and the IGBT switching state of the inverter . Then the DSP is considered to transform three-phase into two-phase operation . After the measured stator flux and torque of the motor induction have been passed through the hysteresis controller , the corresponding stator voltages are captured according to the values in switch table . Note that the traditional DTC hardware is implemented in DSP .It results in various delays in both calculation time and sampling time and generates a large ripple response. To resolve permanent problem ,this study completes the descried function with Verilog hardware description language (HDL). It shorten the calculation time and sampling time considerably, Furthermore fuzzy control, predictive control and neural network PID, are integrated into the motor controller to improve the performance of IM. The fuzzy rule base and defuzzification are considered to have good flux command value in the hysteresis controller and the PID controller with revival driven neural network filtering is used to deal, and the torque is sent to the PID controller by the revival-driven neural network filtering is used to deal with the speed to generate a torque commend value. Both flux and torque commend are used to select an appropriate stator voltage to drive the induction motor smoothly and reduces the ripples considerably.
After verifying with FPGA development board , an ASIC is implemented in TSMC 0.18-um CMOS process by passing through the optimized software , which is provided by Synopsys and Cadence , to complete the layout and winding automatically . Finally a predictive DTC ASIC is developed with fuzzy control and neural network PID speed control.
|
|---|
