Artificial Intelligence-based Sensorless Control of Non-Sinusoidal Seven-Phase PMSM

This paper proposes a novel sensorless control method based on the artificial neural network (ANN) for a non-sinusoidal seven-phase permanent magnet synchronous machine (NSPMSM). This method exploits different harmonics in different fictitious machines to estimate the positions of back-EMF vectors. Due to the limited hardware resources, the number of neurons in neural networks cannot be increased infinitely. To reduce the requirements for the number of neurons in neural networks, an adaptive linear neural network (Adaline NN) is exploited to extract the fundamental components of voltages for training. With three artificial neural networks, position estimation can be performed without the machine parameters. Finally, the performance of the sensorless control method is compared to the Luenberger Observer (LO) and Sliding Mode Observer (SMO) in simulation.