Neural Network and Fuzzy Logic Applications to Current Regulators for Field Orientation Controlled Induction Motors

In the field oriented control (FOC) of an induction motor drive system, a precise current control is imperative since the field oriented control is based on the assumption that the real current coincides with the reference current. To reduce the torque ripple, high quality current control is also required. In fact, the use of FOC induction motor system does not grow rapidly until good current control became available [1]. The following work is involved in this thesis: 1. A BP (back propagation) multilayer neural network based adjacent principle guided CRAM (current regulated delta modulator) is simulated with the help of Simulink and a customized C program. The simulation results show that the neural network based system functions similarly to the conventional adjacent principle guided CRAM. The results show that the nonlinear mapping ability of the neural network is excellent and the off-line trained multi-layer neural network can be used to replace the adjacent principle guided CRAM. 2. A neural network based indirect adaptive control approach is proposed to on-line control the current regulators to solve the self commissioning (automatic tuning) as well as the parameter adaptation problem. Since customized neural network simulation software, such as Matlab Neural Network Tool Box, can not be used to implement the on-line training, on-line programs are designed to implement the simulation. In order to implement the on-line learning in the simulation system, on-line simulation C programs are designed with the built-in Simulink CMEX interface. The C programs then are transformed into S function modules. Finally, the S function modules are integrated with other Simulink modules to complete the whole simulation system. However, owing to the slow training speed of the BP multilayer neural network and the high band width of the current regulators, the simulation results show that the adaptive control is not successful..