Failure of Power Factor Correction Capacitors 1

[asymptote]

Hi Forum. I have been asked to comment on protection failures following an explosion which destroyed a 6.6kV 400kVAr PFC bank. I can explain the protection issues but would welcome some thoughts on the PFC failure. The site is connected at 33kV via a short 33kV cable from the 132/33kV BSP which has two 90MVA 132/33Yd TX's with YZ earthing TX and liquid earthing resistors. the 33kV fault Z is (100 MVA base) 0.87+j19.42 = 514MVA. On site there are 2 x 33/6.6kV 10MVA Yyo Z 10% with solid earthing operating in parallel. From the 6.6kV BB one OCB feeds via a short cable onto a bus bar cable end box, these bars have two outgoing cct's via OCB's which connect:
1) A 4.4MVA 6.6kV Auto/Star Moving Coil Regulator which is connected directly on to a 6.6/3.8/7.26 Yyo ratio changing TX Z = 1.62% with a 400kVAr Delta PFC solidly connected at the MCR primary. These PFC have now failed three times(over a period of 12 yrs) with suficient force this last time to destroy the building in which they were housed.
2) The second 6.6kV ocb feeds a similar circuit with a 1.9MVA 6600/100/550 volt Yy Moving Coil Regulator and 300kVAr PFC bank.

These MCR's feed the client test bay where a variety of motors can be tested. Due to a lack of understanding the complete system is normally left alive even though no test runs are planned. ie MCR's & PFC alive but unloaded for long periods. Simple plant failure seems to much of a coincident to be realistic. A harmonic survey is in hand, system resonance and ferroresonace need to be considered and any direction in this respect would be welcome.

 

[Baldor]

I have some experience with low (480 - 600) voltage power factor correctors. The capacitor bank failures I have encountered were caused by overvoltage or heat.
The caps should be taken off line when inductive loads are minimal, as a transient rise in voltage can occur. If your facilty is on a grid where many other facilities shut down at night or weekends, you might see a significant voltage increase from the utility.
Find out from the capacitor manufacturer the max operating temp (it could be surprisingly low!). They should also provide a normal expansion tolerance (i.e., X mm, or percentage of width ) We have employed forced-air cooling to prevent "pregnant caps".

 

[asymptote]

Hi Baldor. Thanks for that comment but we have esablished that the system volts at the time were in fact low at 32.5 kV. The PFC's were fitted with over pressure trips but they had not been connected!. The set of PFC's that did not fail, which were in fact in parallel look OK and not "pregnent"

 

[jbartos]

Suggestion: Beside mentioned system resonance and ferroresonance as worrisome culprits, the voltage surges and spikes should not be ruled out. These are detrimental to the capacitance dielectricum too. A careful power supply quality analysis should be performed.

 

[asymptote]

Hi jbartos. Thanks for the comment and yes this is an area of concern. The 132/33 SS which is very close electrcally to the PFC, supplies wide area public networks including a rail traction supply. As for spikes, having thought about this possibilty follwoing the last PFC failure (within 12 months)and the value of the MCR, surge diverters were fitted within the MCR tank on the primary terminals. This places the surge diverters just 10 metres away from the PFC. Still awaiting the harmonic results which may interesting.

 

[jghrist]

The capacitor manufacturer should be able to furnish tank rupture curves that should be coordinated with the overcurrent protection. Current limiting fuses may be required.

 

[asymptote]

Hi jghrist.
The protection failed due to the trip supply battery being O/C and there were no HRC fuses, although the manufacturer had stated there were subfuses within the tank protecting the series/parallal packs. None were evident in the remains. The tank bore no sign of any arc roots within but burnt external paint, severe over pressure and rupture suggest a rapid rise in temperature. A visual inspection of the plant some two days before gave no cause for concern.

I have no evidence of earler failures as I had no involvement in those issues but to have three sets explode in 12 years of plant one would have expected 40+ years life leads me to think we are missing a subtle point, had this event happened just two day earler then we could have been involved in an inquest.

My mind keeps working along the lines of the large ammount of windings and iron in very close coupling with what is efectively 700kVAr and no load. Has any body ideas on how to model this?

 

[jbartos]

Suggestion on the previous posting. Capacitor is modeled as C or
1/(j2pifC)=-jXc=-j/(2pifC) in Ohms
Vc(t)=(1/C)xIntegral[Ic(t)dt]+Vco
Ic(t)=CxdVc(t)/dt
inductor is modeled as L or
L or j2pifL=jXl in Ohms
Il(t)=(1/L)xIntegral[Vl(t)dt]+Ilo
Vl(t)=LxdIl(t)/dt

 

[busbar]

Longland, Hunt & Brecknell, *Power Capacitor Handbook* mentions that phase-current balance in capacitor circuits should be periodically checked.

In ANSI regions, a lot of MV-capacitor sets are parallel ungrounded-wye configuration, so one can opening or shorting in a phase may increase terminal voltage on part of the in-service caps.

Also in passing, the harmonic order for capacitor resonance is about sqrt(MVAsc/MVARcap).

 

[rbulsara]

A thought:

Parallel switching of capacitor connected with short cables and low reactance can cause serious inrush currents. If one set is already enegized and other one is switched on, the charged set will try to discharge through the new set as it sees it as a short circuit. This makes the inrush currents too high (adding to the normal inrush) and if not properly fused will rupture caps.

One cap bank failing violently while other remains in tact is indicative of such an possibility.

Two thinngs, adjustable reactor and no fuses are worrysome to me.

 

[asymptote]

Hi Forum.
Thanks for the valuable contributions. It's nice to be able to bounce an issue around the table and see if any similar cases have been solved. We need to resolve this and when complete I shall report back.

 

[njengr1]

Having just finished investigating the destruction of a 21MVA bank operating at 27kV it is apparent caps are very sensitive to overvoltage. The stated max voltage for continuous duty is 110%, but no responsible vendor will sell a design rated to operate at the level.
IEEE Standard 18 has a table covering times that various overvoltage levels can be sustained beyound this expect breakdown
Duration Times Rated volts
6 cycles 2.20
15 cycles 2.00
1 second 1.70
15 seconds 1.40
1 minute 1.30
30 minutes 1.25
Additional info about permissible transient discharge overcurrents is included.
The root of our problem was bad fuses that took too long to clear and permitted can ruptures to se the bank afire.
A few papers are good to review. See "A Primer On Capacitor Bank Protetion" Martin Bishop, Tim Day et al. IEEE Transactions on Industry Applicatons, Vol, 37,No.4 July/August 2001

 

[asymptote]

Hi njengr1
I shall take a look at the reference paper you indicate and see what light this shows on this matter. As for over voltage we know from SCADA that the 33kV system volts were a tad low at the time of our issue which is not to say where they were during the previous 48 hours which might be worth asking the utility.

 

[jghrist]

You say the 33kV system volts were a tad low, but how about the voltage to the capacitors. Are there taps on the 33-6.6kV transformers? I'm not familiar with moving coil regulators. What is the output voltage? What is the effective impedance in series with the capacitors?

 

[asymptote]

Hi jghrist.
The two 33/6.6 10MVA TX's have two off-load taps and it's no easy to determin what tap there are on without an outage. It's a guess that they will have been on the same tap for 30 years. The effective Z in series with the PFC's is very low 0+j69 on 100 MVA base and without detuning reactors. As for your comment about not being very familiar with MCR, me also. My background is in distribution and we had the odd small single phase one LV/LV but this 4.4MVA 6.6kV is a big boy and I'm now hearing that this has also suffered faulty windings during the span of these PFC failures. As I said up front, my task was investigate protection mal-operation which was the easy bit and a side issue to what I think is a much deeper design problem.

 

[electricpete]

Just some thoughts of logical things to check:

Forensic investigation of failed caps.

Is it considered unusual that no fuses blew to protect the caps prior to failure? Does this suggest the failure was not overload from overvoltage/or harmonic content, but rather dielectric failure? Other reports of fuses blowing without failure?

Since you have recurring problem, the same degrading influences are likely present on in-service caps. Investigate:

Infrared survey. – Possible high resistance connections?
Doble test of caps. Is it possible moisture was getting in?
Harmonic survey. - Start at the caps and see if any high frequency present there.
Voltage – You said it’s not a concern – how about balance?

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

 

[asymptote]

Hi to electricpete. No fuses fitted and as local and back up protection failed, the arc which resulted from PFC tank exploding, burnt until it was too long for system volts to sustain and the cable was alive until the fire crew ask for it to be demostraited dead. I'm not familiar with Doble test but I assume that it a test to establish tan delta (loss angle of dielectric) but as the bank has gone to a beter place! As for moisture again no tank to test but as they were hermertically sealed and in a dry enclosed environment it's a low probabilty. I had not considered voltage balance but as there are two 33/6.6 TX's in parallal I can't imagin what sort of problem would cause unbalanced volts. The PFC were connected in delta and the 6.6 star pt is solid. In addition this network feeds a motor test bay and I hope they would have noticed unbalanced voltages. I am awaiting the harmonic survey results which may have some points of interest.

 

[electricpete]

Yes Doble test is power factor, practically the same thing as tan delta.

You say you have no caps to test for Doble or infrared. But you said it has failed three times. So at some point it will be restored and at that point you can test. Presumably unless you have solved the problem, the degradation influences will be present in the replacement capacitors, same as the last three sets. In ideal world you solve it before you replace them, but not always so lucky.


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

 

[busbar]

asymtote — There will always be regional differences and localized customs in protection philosopy, but partial rationale for ungrounded-wye cap banks in IEEE 1036 asserts:

In grounded-wye and delta-connected applications, the maximum current is the system available fault current at the capacitor location. The withstand capability of each capacitor varies with design and size, so the available fault current at each location should be compared with the capacitor case rupture curve, which is available from each manufacturer. If the short-circuit current is excessive, other options available include the utilization of current-limiting fuses to limit the fault current, connecting the bank in ungrounded-wye which will significantly reduce the fault current, or moving the bank to another location with an acceptably lower available fault current.

(It should be noted that when the bank is connected in an ungrounded-wye configuration, a phase-to-ground fault may still occur if the capacitor tank is grounded. However, the failure of the major insulation from the internal capacitor pack to the capacitor case is rare.)

 

[Skogsgurra]

Asymptote,

We had a similar problem with 6 kV capacitors and the root cause was that the power factor regulator was a bit too "eager" to switch capacitor sections in and out. So fast, in fact, that the capacitors did not have enough time to discharge in the off time. This sometimes led to capacitors with "wrong" polarity (negative w/ respect to actual polarity of the 6 kV) being switched back to the 6 kV. The first thing that happened was that the breaker got stuck (welded). We than introduced time delays (we had VTs connected and needed less than 30 seconds to discharge) in the PFC controller's output and that seemed to solve our problem.

What about your moving coil PFC controller? Does it have the necessary delays?