Sensorless driving method of permanent-magnet synchronous motors based on neural networks

In sensorless driving of permanent-magnet synchronous motors, it is difficult to deal with the nonlinear characteristic of the magnetic circuit and the harmonics of magnetic density distribution in the air gap. Using a neural network has been considered to be a powerful tool but, unfortunately, only few simulation-based works can be found. In this paper, an experiment system and successful experiment results using our proposed sensorless driving method of permanent-magnet synchronous motors based on neural networks will be presented. The method estimates the position errors from electromotive force instead of directly estimating the position using neural networks and then using the approximate algorithm to obtain the rotor position. Experiment results show that the method has excellent possibility in practical applications.