Motor Current Oscillation 3

[Masbibe]

Hi to you all.

I am interesting does anyone have expirience in commissioning big induction motors supplied via VSD that workink in scalar control mode (V/f).
I had commissioning 2.85 MW motor, 660 V, 50 Hz. Drive is ABB ACS800.
Uncoupled motor was worked very bad in scalar with some high oscillation of current and in DTC everything was OK.
You can see in attach graph of uncoupled motor in scalar.
Does anyone know what could be reason of oscillations.

Best regards,

Milovan Milosevic

 

[Masbibe]

"Yes, I agree. And it argues that it is not strictly the dc component causing the torque oscillations as you said. Instead there are subtle changes in phase angles going on that contribute to these torque oscillations. I believe it is better viewed and explained in the synchronous d-q reference frame, where we see that the d and q variables oscillate at 60hz because it is a resonant frequency of the stator circuit. I still plan to post a little more on that.
"

I notice also. Torque oscillations are because stator and rotor angle oscillate. This is main factor for such high torque oscillations.
So only question is why angle between rotor and stator current oscilate.

Milovan Milosevic

 

[Masbibe]

Sorry not stator and rotor angle I meant to say angle between stator and rotor current oscilate.

Milovan Milosevic

 

[Masbibe]

sorry,
One more mistake. Frequency of pulsation in 1 phase motor is 100 hz or 120 not as I said before. You was right. I was think a little bit.

Milovan Milosevic

 

[electricpete]

I am not sure when I look on this graphs what is with frequency, because oscillations ends on very small speed.Probably you have better zoom so I only can trust you. But on my simulations frequency of torque oscillations are same as rotor currents. Can you try simulate some flystart (starting motor that is already running for example on half of synchronous speed). Then you will be sure what is with frequency.

I stand corrected. You are absolutely right, the torque oscillations occur at a frequency of LF*(1-slip). This is demonstrated attached using lower inertia (100 kg*m^2 changed to 10). I'm definitely learning something here, thanks!

=====================================
(2B)+(2B)' ?

 

[electricpete]

It's been a good discussion, I've learned a lot, I hope it continues.

I'd like to summarize some of the features of the DOL start that we've talked about, and add a little bit of an attempt at explanation/analysis.

First there are two frequencies of oscillation we'll mention: LF*s and LF*(1-s)

1 - The oscillation at LF*s occurs in the torque and is obvious in the d and q currents. (see slides 1 and 2)

2 - The oscillation at frequency LF*(1-s) occurs in the DC offset of the currents when viewed as the a./b/c stationary variables (as we expect to measure them in the field). This is shown in slides 3 and 4. Some notes to add about this oscillation:
2A – It starts off very closely resembling "classical" simple decaying dc component (as discussed in thread237-283053 ) perhaps because initially the frequency LF*(1-s) is zero.... but as LF*(1-s) increases the oscillation becomes evident (as expected).
2B – For locked rotor condition, LF*(1-s) remains zero throughout, and the dc offset remains a classical simple decaying dc throughout, with no oscillation in that dc offset (as expected) as shown in slide 6. Locked rotor torque is shown in slide 7 and of course frequency LF*(1-s) remains constant at LF. Also the oscillations seem to last much longer at locked rotor conditions.
2C – It is postulated but not confirmed that LF*(1-s) frequency oscillation in the offset is related to the LF*(1-s) frequency of the rotor current when viewed in the rotor reference frame (slide 5)

I have a vague unproven belief that if we study the synchronous-reference frame d-q transient equivalent circuit of Krause, we might gain some more understanding in a manner similar to how the steady state equivalent circuit helps us understand other things.

I think we can roughly identify where these oscillations at frequencies of s*LF and (1-s)*LF come from by studying Krause's transient circuit (slide 8).

Marked in red on Slide 8 is a path for circulating current in the rotor circuit which a resonant frequency of s*LF... I believe this is where the LF*s oscillations come from. There is analytical "proof" on slide 9. Slide 10 shows that the phase relationships in the simulation are in agreement with the proof.

Slide 12 shows a path for circulating currents in both rotor and stator circuits with a resonant frequency of (1-s)*LF. I believe this is where the LF*(1-s) oscillations come from. The analytical proof is not shown but it should be very easy to apply the exact same logic shown in slide 9 to this scenario. Slide 13 shows that the phase relationships resulting from the simulation do obey the expected phase relationships for this circulating path.

If we are looking for physcial insight, we observe that we have currents when represented as d-q variables in sync frame are sinusoids of constant magnitude. Oscillating between d and q axes. This represents phase change of the sinusoidal currents which is what we had already speculated.

Maybe (hopefully), someone can chime in if there are more or different conclusions we can draw on all of this.


=====================================
(2B)+(2B)' ?

 

[electricpete]

I think the slide 8/9/10 explanation for the loop that creates s*LF was good and it shows up in the q and d currents in sync ref frame.

However the slide 11/12 discussion of a loop creating (1-s)*LF was incorrect. This frequency of (1-s)*LF does not show up in the sync ref frame currents. Maybe instead it has something to do with the rotor reference frame rotor currents shown in slide 5.

=====================================
(2B)+(2B)' ?

 

[electricpete]

I had another thought about the physical interpretation of the LF*s oscillation which shows up in the torque and in the synch ref frame d-q currents.

Focus on the rotor d-q currents in synch ref frame. They represent components of current which are physically 90 degrees apart on a ficticous rotor that is rotating at sync speed. What we see in the simulation (slide 12) is that the oscillating component of these currents is equal magnitude and are 90 time degrees apart (90 time degrees as defined by the frequency s*LF). What does that tell us?

It doesn't tell us that the resultant synchronous field is just oscillating back and forth about some equilibrium. It is in fact rotating (the sum of two sinusoidal fields displaced by 90 degrees in both time and space produces rotation). They are rotating backwards with respect to the sync field. It is somewhat analogous to a phenomonon of pole slipping, but it occurs continuously as this component of the rotor field moves steady backwards with respect to the stator field. No wonder there are large torque oscillations predicted.

=====================================
(2B)+(2B)' ?

 

[Masbibe]

electricpete,
Thank you so much for such effort you make to explain all off this. I need few days to think a little about what you wrote and to make final conclusions.

Milovan Milosevic

 

[electricpete]

Thanks. I still have a few errors that I've got to fix. I reversed s*LF and (1-s)*LF in some places. Also I calculated rotor current in rotor ref frame wrong (plugged wr*t into the transformation from sync to rotor instead of wr*t-we*t), but that doesn't change the picture much. All of that is corrected in the new attachment labeled "Rev2"

One more thought I have had about these torque oscillations at s*LF is that when we transform a rotating field (like the one we see in the Iq & Id oscllation at s*LF in sync frame), it's frequency changes by the same amount as the change in ref frame speed. So if we transformered that s*LF oscillation to the rotor ref frame, maybe we would see dc. That would lead to a much simpler explanation: the resonant frequency in the rotor ref frame transforms to zero which is what we expect out of simple R/L circuit... this is simply a dc current in the rotor ref frame that is dieing away slowly due to rotor inductance.
At first glance, that seems logical. If I calculate the time constant associated with that loop, I get
(LM+L2)/R2 =[(XM+X2)/<2*Pi()*50>] / R2 =[(0.4789909
+0.0197605110515279)/<2*Pi*50>] / 0.000978488734217218 = 1.62247673 seconds. That looks like roughly the same constant that is evident in decay of torque oscillations slide 7. So it seems like it might lead to a simple explanation that the dc in the rotor tends to remain at the same location with respect to rotor and causes torque oscillations at s*LF as the synch field passes by it at a rate of s*LF.

But unfortunately I still have one fly in the ointment. The rotor current in rotor ref frame does not seem show any dc component (slide 5). I have a suspicion that maybe I still have some error in my transformation of rotor currents from sync frame to rotating frame. After all we see what appears to be frequency getting faster... but it should be getting slower. Will keep working on checking that transformation.

=====================================
(2B)+(2B)' ?

 

[electricpete]

On attached slides 11 I added corrected rotor currents in rotor reference frame (Now I integrated wr instead of multiplying by t as discussed in other thread). The main frequency is now s*(1-LF) as expected (starts slow and speeds up). There is superimposed a high frequency ripple, either LF or s*LF... I can't tell (tried a few things on slides 12-14 to try to figure it out but doesn't provide any clues). There is no sign of the dc that I thought would be there. Back to the drawing board in terms of an explanation for this torque oscillation at s*LF.

=====================================
(2B)+(2B)' ?

 

[electricpete]

I suspect the rotor current plot in rotor frame is still wrong - I will try to correct that in the next few days.

=====================================
(2B)+(2B)' ?