Wavenet Based Vector Control of a Permanent Magnet Synchronous Motor Drive

In this investigation a vector controlled permanent magnet synchronous motor (PMSM) is simulated and then a wavenet (a neural network based on wavelet functions) based speed controller is designed to make the system more robust against parameter variations. The data for training the network is extracted from a process with several PID controllers to achieve a reasonable performance in all operating regions. At the first resolution of training, the desired accuracy is obtained and finds out that going to higher resolution is not necessary. After determining the number of nodes at each resolution and the number of wavelets at each node, the location of nodes at each resolution is determined. Finally on the basis of advantages of wavenets the final results are presented. The response of the controller proves the validity of the proposed method.