ELM-based sensorless speed control of permanent magnet synchronous machine

Abstract

This paper deals with Extreme Learning Machine (ELM) based sensorless speed estimation and speed control of Permanent Magnet Synchronous Machines (PMSMs). ELM, first proposed by G.B. Huang as a new class of learning algorithm for Single-Hidden Layer Feedforward Neural Networks (SLFNs), is extremely fast and accurate, and has better generalisation performance than the traditional gradient-based training methods. To implement Field-Oriented Control (FOC) in PMSMs, the stator magnetic field is always kept 90 degrees ahead of the rotor. This requires rotor position information all the time. This information is accurately obtained with an ELM-based observer without the position sensor for PMSMs, and hence, the cost of the system is reduced, while the problems associated with the sensors are minimised.