for my master thesis I'm investigating the EMI of a 3-phase inverter, wich drives a PMSM. The inverter's structure is the common 3-Phase DC/AC Voltage Source Inverter and it uses EPC2020 GaN transistors at 100 kHz switching frequency. For my investigations the system needs to be in a stationary state with exactly periodic currents and signals, hence no closed loop control is used.
Instead I only set the desired stator voltage and frequency and let the motor do its thing. For startup I slowly increase the stator frequency and at the same time increase the voltage my inverter produces to compensate for the higher back EMF at higher frequencies.
So far so good, although this method is not practical, it lets the motor run at my desired electrical frequency of 50Hz in an idle state, but the phase currents are concerningly high.
The system currently runs with a DC rail voltage of 20V. The DC current going in the inverter is peaking at about 5A. The phase current of the PMSM on the other hand is peaking at nearly 20A. Hence my conclusion the majority of this current is reactive current between the motor and my DC link cap. So now my real question: How do I reduce that reactive current? From my understanding of synchronous machines, the reactive current is not needed for torque built up and should be controlable through the ratio of stator voltage and induced back EMF. So if I adjust the absolute value of my stator voltage it should be possible to reduce the reactive current to a minimum, while still drawing the real current needed to drive the motor.
But that's unfortunately not what I experienced until now. If I increase the stator voltage, the reactive current gets bigger. If I reduce it, the current does reduce a bit, but before it gets reduced significantly, the motor stops, so somehow also the real current got reduced, what I find counterintuitive.
I plotted one phase current and the DC current for one whole period, you can take a look. They also don't look like I expected, but I have no explanation on why that is...
So I know this may be an unusual use case, but maybe there are some machine experts out there, who may guide me in the right direction ;)