[Active Front End Rectifier] Leading current issues

[burnttoastzuko]

Hey everyone!

I am currently working on an active front end rectifier control. More specifically, I am using dq control, setting Id_ref to dc link capacitor Pi controller output and Iq_ref = 0. And switching controlled by space vector modulation. The issue that I face is that, my line current seems to be leading my line voltage by a fixed phase difference. And also that there is a negative dc offset to my current, which again is fixed throughout the voltage range (testing from 240V-480V).
Some fixes I have tried - 1. Implementing a clamped svpwm, which reduces this offset, but increases harmonics (expected). 2. Feed iq_ref = const value or a lag compensator, which brings me unity power factor. 3. Implemented the same control using spwm, which while does give better peaks for current, has similar phase and offset issues.

However, my understanding of AFE is that irrespective of what reactive load the AFE faces, it should be able to drive the system to unity power factor. I do not think my Pi controllers are the issue, because they drive the control variables to the references just fine, with very little oscillations. I have also verified for any offsets in sensor ADC measurements, which is not the case.

I would love to hear from you guys as to what I am missing or could be improved. My hard constraints are - no increase in switching freq (currently 5kHz ISR freq), no reduction of capacitance and no changes to switching deadtimes (constraints levied by other issues, resolved by other team). Please also let me know if I could provide you with any other constraints in order to paint a better picture of the issue.

Thanks in advance!

 

[crshears]

Welcome to the forums, burnttoast.

 

[iop95]

Maybe an error in frames reference transformation and/or PLL low pass filter frequency set.

 

[burnttoastzuko]

Maybe an error in frames reference transformation and/or PLL low pass filter frequency set.

I am not using PLL to determine the angles, instead using arctan of Clarke transforms; gives me more accurate angles, with less oscillations and less tuning effort.

I have had my doubts with the axes transformation being the source of my errors, but my colleague insists that our instantaneous line-to-line voltage measurement to line-to-ground transforms are accurate. What we use right now:
Va = (2Vca + Vab)/3
As I mentioned earlier, the phase difference remains constant and the reactive current reference Iq_ref that I can feed is proportional to magnitude of input voltage.

 

[electricuwe]

My experience with synchronizing to the grid is that solutions that somehow mimikry the behavior of a rotating mass work fine. This is robust against errors or short term disturbance. We implemented this as a three-phase PLL on 16 bit µC more than 25 years ago, but the concept was already inherited from IC-based SCR controllers available on the market at that time.

If I would have to design such systems today, I'd use a similar approach on recent hardware.