quiz - can a DOL-start unloaded induction motor "overshoot" sync speed 3

[electricpete]

If I were to perform a direct-on-line start a large motor with no connected load, would you expect the speed to overshoot synchronous speed?

(note the word quiz - that's a clue that I know the answer - just asking for fun)

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[Marke]

If the load inertia is very low, then the rate of acceleration can be very high.
The rotor circuit has a relatively long time constant and for a short period of time (relative to the rotor time constant), the rotor appears to have constant excitation.
If the rate of acceleration is higher than the rotor time constant, then as the rotor approaches synchronous speed, there will be a period where the rotor is "over excited" relative to the speed it is operating. It is thus able to behave like an over excited synchronous motor and ocsillate around synchronous speed.

Best regards,

Mark Empson
L M Photonics Ltd

 

[Skogsgurra]

Yes, Marke. That is what we saw. An oscillating system where there shouldn't possibly be one.

Oscillation is undamped ringing. Put some velocity dependent losses in the system (first derivative) and you have a system that overshoots when subject to high acceleration.

Pete is the one most versed in math. Looking forward to an analytical explanation.

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[electricpete]

I agree with Mark's explanation - it is the same basic phenomenon but explains the oscillation a little better. To the extent the current is slowly decaying it acts like dc. And if you have dc current on the rotor it acts like a sync machine.

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[electricpete]

Actually the sync machine is not a perfect analogy since it is not oscillation in phase but oscillation in frequency. We would be slipping poles many times on a sync machine and that difference in speed doesn't generate any net restoring torque on a sync machine but it does on induction.

Another way to look at it is that we have a control loop with a target speed of 30hz. The rotor current is part of the feedback comparing actual speed to target speed. Since there delay (due to inductance) in the rotor current feedback, we have overshoot and oscillation.

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[edison123]

Very interesting posts. I assume this does not happen in slipring motors or motors which have controlled start like wye/delta starting, VFD, soft starters etc ?

 

[electricpete]

From my perspective reading Krauss, it applies to large squirrel cage motors since they have high L/R ratio, especially in the rotor circuit. It also applies during rapid changes (in addition to DOL start, he also shows example of slight ringing on the large motor during rapid load change and during external fault which is quickly cleared).

It would not apply to smaller motors since they have lower L/R (for example see the 50hp startup curve in attachment above - no overshoot). It would not apply to vfd start since the speed is slowly ramped. It would not apply to slip ring motor if: 1 - speed is increased slowly, or 2 - motor is small. If there is large slip ring motor started DOL, I suppose it's a possibility although I think rotor resistance is higher for slip ring motors than squirrel cage motors.

I was able to do a simulation which matched the results posted. I will posts some results from there to show a slightly different view: speed vs time and torque vs time. The thing that confuses me a little bit is the time constants. For this 2250hp motor with R2' ~ 0.02, XL2' ~ 0.2, Xm ~ 13, which has L/R on the high (long) end of the spectrum, it still seems to have a fairly small/short time constant. Using the the rotor leakage inductance:
Rotor L/R = L'/R' = Xl' / R' /(2*PI*60) = (0.2/0.02) / 377 ~ 0.025 sec.
That seems relatively short in comparison with the transient and the oscillation a..........

Now if for some reason you included Xm in the ratio the time constant would be a lot longer (but it's not obvious that should be included)
Rotor L/R = L'/R' = Xl' / R' /(2*PI*60) = (13/0.02) / 377 ~ 1.5 sec.

At this point it doesn't quite make sense to me - I can't reconcile the time constant with the transient. Maybe my question will be clearer to readers when I post the plots vs time (maybe tonight).


=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[electricpete]

Correction in bold:

Now if for some reason you included Xm in the ratio the time constant would be a lot longer (but it's not obvious that should be included)
Rotor +Magnetizing L/R ~ (L2'+Lm)/R' = [(X2'+Xm) / R'] /(2*PI*60) = (13/0.02) / 377 ~ 1.5 sec

[/quote]

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[Skogsgurra]

Looking forward to that, Pete. I knew you would do it.

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[ijl]

Interesting subject! I made some simulations with ATP. Results are shown in the attached picture. Speed is given in rad/s as a function of time. Motor has some "typical" parameters (that I picked up somewhere..), and two pole pairs. The frequency is 60Hz. The synchronous speed is 188.5 rad/s.

First picture: nominal parameters, normal start-up.
Second picture: inertia reduced by a factor of ten.
Third picture: Additionally, the rotor resistance reduced by a factor of ten. Has this kind of growing oscillations been observed with real motors?

 

[Skogsgurra]

Hey! The third picture is what we got back in the early seventies when our induction motor started oscillating for no apparent reason (se post 5 Jul 09 0:08 in this thread).

Yes, it does exist IRL. Only, our oscillation didn't grow limitless. It stayed within a few percent amplitude.

This thread is getting better and better!

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[electricpete]

I got busy last night, never got to post my simulation. Will try again tonight. It was a damped oscillation.

The oscillation will certainly be larger if you decrease rotor resistance and/or inertia. But continually growing oscillations seem impossible to me (I'll bet if you continue the simulation it will level out and eventually start decreasing and if long enough return to steady state).

This conclusion is based on looking at the motor start event in the synchronous ref frame (which is a very interesting and informative excercize). In the sync reference frame, we have similar a pair of stationary (sync frame) magnets representing the stator field. Initial conditions: at the moement of start we know in our normal reference frame the rotor is at rest, but in the sync ref frame the rotor is rotating at 3600rpm (2-pole motor) the moment before start. Then the magnetic fields induce circulating bar currents which will slow down the motor. Conservation of energy tells us that by the time the rotor comes to rest in the sync frame (up to sync speed in our normal ref frame), the I^2*R energy dissipated by the rotor must equal the initial kinetic energy of the rotor (correspondin to 3600rpm). (This also correctly predicts the amount of energy dissipated in an unloaded motor during start by the way). Once the energy is dissipated in the rotor resistance, there is nothing to sustain the oscillation and it must eventually disappear (otherwise we have a perpetual motion machine).

By the way, what is ATP?

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[electricpete]

I should clarify my conclusion is based on DOL start of unloaded motor connected to a pure sinusoidal power supply. More complicated power supply or load can of course cause sustained oscillations.

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[Skogsgurra]

ATP means Alternative Transient Program. It is similar to ETAP, only a lot cheaper (free).

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[electricpete]

Attached are results of a simulation of the 2250hp motor example discussed above. Simulated in Matlab using d-q variables (current, voltage, flux linkage) in the synch reference frame as outlined in Krauss page 150.

Slide 1 are machine parameters.
Slide 2 is torque vs speed – same as was shown in Krauss.
Slide 3 is speed vs time
Slide 4 is Torque vs time
(oscillations in speed and torque occur at 7-8hz and decay quickly)
Slide 5 is stator flux linkage (lambda) along the q axis in sync ref frame
Slide 6 is stator flux linkage (lambda) along the d axis in sync ref frame
Slide 7 is rotor flux linkage (lambda) along the q axis in sync ref frame
(This is the only one of the 4 flux linkages that shows the oscillation!)
Slide 8 is rotor flux linkage (lambda) along the d axis in sync ref frame
Slide 9 is stator current along the q axis in sync ref frame
Slide 10 is stator current along the d axis in sync ref frame
Slide 11 is rotor current along the q axis in sync ref frame
Slide 12 is rotor current along the d axis in sync ref frame
(oscillation shows in both stator and rotor current in q axis only)
Slides 13-15 are stator currents transformed to the normal (stationary) reference frame

I think there are at least three interesting things to talk about.

ITEM 1: OVERSHOOT OF SPEED AND SUBSEQUENT OSCILLATION
(the main subject of the thread).
Krauss makes the statement that these are the result of a rotor electrical transient. He "proves" this in the chapter on "reduced order models" by doing a reduced-order simulation which NEGLECTS stator transients but includes rotor transients. The overshoot and subsequent speed oscillation is still there in the results of that simulation and therefore not a result of stator transient and must be a result of rotor transient. The reduced order model can be solved for eigenfrequency of about 7-8hz.

If I look at the results of my simulation slides 5-8 are 4 flux linkages in the q and d axes for the stator and rotor. The oscillation only shows up in one of the four – the rotor q axis flux linkage. I guess this is consistent with the fact that the overshoot originates from rotor transient. I'm not sure what we learn from the fact that it's associated with the q axis as opposed to d axis. When we look at stator currents (slides 9 through 12), we see the oscillation shows up in the q axis current of both stator and rotor – I guess the stator current oscillation is a result of the transient rather than an initiator. Although when we look at the normal stationary-frame abc stator currents in slides 13-15 it's very tough to see the oscillation there.

ITEM2: SLOW OSCILLATING DC ADDED TO LRC ENVELOPE DURING THE FIRST SECOND
See slides 13 – 15. Look at the top and bottom of the envelope and you see they move up and down together indicating some added dc component. The envelope of the LRC slowly varies up and down.
Krauss says it is due to interaction of stator electrical transient and rotor electrical transient (whatever that means... hard to visualize... don't see it in the flux linkages). If anyone can provide a better explanation I'd be interested to hear.

The interesting thing is I have observed similar oscillations in recorded motor start information on a very similar motor (2500hp, 1800rpm, different voltage=13.2kv). (If anyone wants me to post that let me know). I have wondered about it before - now I know it is nothing abnormal - it is expected. This particular item is apparently not limited to large / high-efficient motors – it also shows up in Krauss' simulation of a 3hp motor start.

ITEM 3: 60HZ TORQUE OSCILLATION IMMEDIATELY AFTER START (NEAR ZERO SPEEED).
This is shown in slides 2 and 4 (also a slight speed ripple shows up in slide 3). Krauss says this is due to the transient dc decaying offset of the stator currents. In the chapter on reduced order models he simulates the start with stator transients nelgected, and this initial 60hz torque oscillation is mostly absent (which sort of proves it is caused by stator electrical transient). Now it's interesting to observe that the initial stricly-decayind dc offset is gone within the first 0.1 seconds (slides 13-15), but yet the 60hz torque oscillations last to at least 0.7 seconds (slide 5). I guess the relevant dc is associated with that weird slowly-oscillating dc of item 2? But it's strange because the envelope of the 60hz oscillating torque doesn't show any of the similar oscillation as shown in item 2... the 60-hz oscilating torque envelope just decays very smoothly to 0. Pretty bizarre.

I can't say I have ever seen evidence of 60hz oscillating torque during start although I wouldn't know what to look for. If there were a 60hz torsional resonance it could be briefly excited. I guess this is not so bad as the sync motor startup torque oscillations which vary in frequency from 120hz down to 0hz and can excite every torsional resonance in between.


=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[electricpete]

I should clarify what is meant by stator and rotor transients and neglecting stator and rotor transients.

The induced voltages in any reference frame can be expressed as a sum of a speed voltage and a d/dt(Lambda). Transients are associated with the d/dt(Lambda) term. Reduced order simulation discards the d/dt term - sort of like a quasi-steady state analysis. Good enough for most purposes and significantly reduces the compuation time.

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[Skogsgurra]

Great that you did this, Pete!

I have been looking at the frames and your comments. The reason why you do not see any oscillation in Id is, I think, that there is no energy exchange involved between Id and the pole angle oscillation. Simply because there is no torque produced by Id, only EMF (remember our discussion about Back EMF?).

The oscillations shall only show in Iq, because that is where the "spring", i.e. elastic coupling, between rotor and stator flux is and that "stretching/relaxing" the spring influences the current.

A great thread, this is. I will look further and try to understand what is happening. Did you ever try ATP? I think you should. I registered and received the package. But it takes some time to get started with it. And I don't ever seem to get that time :-(

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[electricpete]

Thanks for those comments Gunnar

The applied stator voltage in my simulation was on the vq axis in the sync ref frame. So it would make sense to me that Iq is in-phase with applied voltage and therefore associated with real power (associated with non-zero average torque) and Id is associated with changes in stored magnetic field energy. Is that correct?

Even if that is the case I don't see that this necessarily proves why Id is not involved in transient torque. Look at the instantaneous torque equation:
Te = (3/2) (Poles/2) * (Lamba_q*I_d - Lambda_d*I_q)
I believe the steady state torque is associated with Lambda_d*I_q. But the transient torque also includes contributions from Lambad_q*Id. So it is not immediately obvious that Idr would be necessarily be excluded from torque oscillations.

For my present purposes, I am happy using Matlab or excel for solution of O.D.E. initial value problems. It is a very general solution method applicable to a very wide variety of problems and I am already familiar with it. I may eventually look at ATP if I find the need for a very large simulation or if I need to double check results of a model.

I noticed ijl's simulation also shows small speed ripples during the early acceleration - evidence that his model also predicts those line frequency torque oscillations immediately after start (item 3 above).


=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[electricpete]

I recreated ijl's scenario of increasing oscillations by decreasing R2 by factor of 10 and decreasing J by a factor of 10. Unexpectedly, the oscillations came to steady state and never showed any sign of decreasing even though the simulation ran for 300 sec. That's a surprise to me - I have no explanation.

=====================================
Eng-tips forums: The best place on the web for engineering discussions.