Open Access

Initial Rotor Position Detection of Induction Machines Using Feedforward Sensorless Saliency Separation


Cite

The use of induction machine spatial saliencies for sensorless vector control in the proximity of zero electrical frequency has been extensively researched over the last few decades. A robust technique to extract machine saliencies is called voltage step excitation, and it computes a saliency phasor out of phase current derivatives resulting from specific voltage steps generated by the inverter switching. Within the saliency phasor, all machine saliencies appear superposed. For some machine constructions, multiple saliencies are present, containing information about the spatial, magnetic and geometric state of the machine. Due to its direct relation with the rotor angle and its high accuracy, rotor slotting saliency is often chosen as the sensorless control signal. In order to exclusively access rotor slotting, saliency separation needs to be carried out, with the goal of eliminating all non-control saliencies from the saliency phasor. In this research, feedforward harmonic compensation based on look-up tables (LUTs) is chosen. The idea is to estimate each saliency in relation to amplitude and phase shift once, store such information in a torque-dependent LUT and use it for feedforward compensation. Yet, several saliencies are linked to the rotor position and, thus, the stored phase shift in the LUT is fixed to a defined rotor position at which the saliency estimation was performed. For the feedforward compensation to work during each sensorless start-up, an initial rotor slot detection must be carried out. This paper presents a technique to estimate initial rotor angle based only on the inherent characteristics of the induction machine multi-saliencies and an iterative feedforward compensation process that requires no extra resources and only a few PWM (Pulse Width Modulation) periods to achieve initial slot rotor angle. Experimental results measured at two different test benches prove the high accuracy of the method.

eISSN:
2543-4292
Language:
English
Publication timeframe:
Volume Open
Journal Subjects:
Computer Sciences, Artificial Intelligence, Engineering, Electrical Engineering, Electronics