Artificial Neural Network Modeling of a Synchronous Reluctance Motor

Artificial neural networks are applied in different application areas, and electrical machine implementation is one of them. The structures, resembling the motor behavior like motor controllers, observers, parameter prediction units, can also be modeled with artificial neural networks. Choosing network structure, layer, and neuron numbers play a critical role to make the model successful. In this study, multilayer feedforward and recurrent neural networks are trained with various layer and neuron numbers to compare their speed and torque tracking performances. The average training time for each combination and optimal window for neuron and layer numbers are also evaluated.