Wind farm 35kV collector intermittent fault 2

[ters]

System:

The facility is a larger wind farm with two 35kV collector feeders each having about twenty 1.5MW turbines. There is one step up 230/35kV transformer in the substation serving both feeders, connected as WYE/DELTA (230/35).

The turbine step up transforms use DELTA on the 35kV side, so since there is a DELTA at both ends of the collector, both feeders employ separate grounding transformers in the substation.

Both feeders are equipped with SEL351 protections. From the 230/35kV substation, both feeders use 35kV UG cables for some distance (about 1km) and then the cables connect to two overhead circuits. The OH circuits use the same poles for few km and then diverge. Each feeder also has several OH branches along which turbines connect via UG cable and one pole mounted aerial switch. Each such switch location also includes surge arresters.

The number of turbines connecting to the OH line via UG cable and the aerial switch varies, in some cases it is only one, in other cases two or max three turbines are daisy chained by UG cables before they connect to the OH line.

Turbine transformers are outdoor and use current limiting + expulsion fuses in series with a disconnect switch in-between.

Problem:

Occasionally, at high outputs, one of the feeder 51 protection trips on the phase-phase fault. The fault starts between B and C phase but it does occasionally progress to phase A too. The fault circuit current is fairly consistent, about 4kA which is about 40% of the (bolted) fault level at the substation. The SEl351 51 element is set to clear the fault within about 400ms (including breaker time). The voltage, as seen by SEL351 in the substation, does not collapse dramatically suggesting that the fault is not very close to the substation.

The nature of the problem is rather random, so sometimes on a windy day it may trip twice, but the other time on even a windier day may not trip at all. There is never a fault on restoring the grid power to the feeder. Sometimes there may not be any trips for couple weeks.

Line inspections were conducted several times and revealed nothing – there are no suspicious places at all.

Some problem with SEL351 relay is also eliminated since the fault is also seen by another feeder protection, which does pick up but does not trip (the current goes to 200% or higher).

The OH lines do employee some self resettable fault indicators for different sections and branches. One of these indicators was tripping suggesting that the fault is on a particular OH branch serving 4 turbines each connecting to the OH line individually. The indicators are single phase and are those which are installed on the conductor by a hot stick, but the make and model unknown. It does change the color from yellow to black when tripped.

However, after the last trip, that particular indicator never reset itself to normal state (to display yellow) suggesting that it is possibly defective, so any conclusion based on its operation is not particularly reliable - if it was too sensitive before it finally failed, maybe it was tripping on faults elsewhere along the feeder rather than on faults on its own branch.

Question:

Where to go from here? Any idea how to narrow down the possible location of the problem?

I sort of exclude the possibility that the problem is underground, since cables cannot behave that way, when they fail they are unlikely to repair themselves after the trip, so suspects are OH structures or possibly (but less likely) turbine transformer termination compartments, where 35kV cables connect.

Thank you for reading.

 

[ters]

Than you again pwrengrds. Forgot to comment on your previous questions - does the fault level point to a spot on the line that would have that level of fault?

According to the coordination study, the fault level at any of the OH sections is higher than 2500Amp. Minimum is about 4kA. So if the fault magnitude is only 2500-3000Amp, while what event records lists as 4kA is the peak value, then turbine transformers are definitely a suspect.

However, the current SEL351 measures in the substation is lower than the actual short circuit level along the feeder, since there is also contribution from all feeder turbines which might contribute up to 1kA (there is 25 of turbines on that feeder).

The turbine fault contribution is low, they use inverters, and the actual SC contribution may not be more 1.2 pu (don’t know the exact number), but turbine converters can push more reactive current than that if they are operating in Low Voltage Ride Through mode during the fault, which is possible.

Secondary breaker in the TX don’t exist, but there is one in the turbine, so yes, the cable from the LV cable from TX to turbine is a suspect, but how likley that it would not be permanently damaged since the short circuit level on that cable must be in the range of 15kA or more.

In any case, with 2500-3000A from the substation + up to 1000A the turbines can contribute, the possibility of having a fault at the end of several OH feeder branches still seems to be strong as well.

 

[ters]

davidbeach (Electrical) 10 Mar 12 21:03 No peak values in the SEL reports. Settings are in RMS, event reports are in RMS.

Thank you davidbeach, but how do I then interpret the fact that the same fault file shows 4kA in the event record while the charts show magnitudes of 2.5-3kA RMS? And that happens regardless if I use SEL or Siemens software to open the same COMTRADE file.

 

[ters]

Or, if 4kA in the event record is actual and real fault magnitude, in RMS, then the peak would then be 4kA*sqrt(2) ~ 5.6kA

However, since charts are plotted so that the peak is at 2.5 – 3kA, it then seems that peak is scaled not once but two times for sqrt(2) before it is ploted, i.e, 5.6/ sqrt(2)/ sqrt(2) = ~ 2.8 ?

 

[davidbeach]

I don't understand the confusion. At cycle 24 of the first event, phase A has a magnitude of 3688.85, B 4172.43, and C 3887.53. Exactly what the event summary says. Those are all RMS values. I'm not sure why you say they are only 2.5-3kA.

 

[ters]

OK, I get it, finally .

Several things confused me that the RMS was only 2500 Amp (and David (pwrengrds) seemed to be under a similar impression too). That is why I had problem believing that there is no discrepancy.

First, SEL chart scaling down thing to make instantaneous values for 1.41 lower then they are (why this actually is needed?).

Second, SIGRA was constantly showing a discrepancy 2.5kA RMS in the chart but 4kA in the event log (I attached a sample few posts above).

Seems Siemens software was "confused" in a similar way as I was, and for the same reason. In theory, one could use any software to analyze COMTRADE files, in practice, looks like due to this 1.41 rescaling embedded somewhere in the SEL file, SIGRA shows both, RMS currents and RMS voltages 1.41 times lower then they actually are.

Finally, the SEL chart scale was crude too, and even when I ploted Im on the top of instantaneous curve, it still sort of looked to be closer to 2.5kA. But it does not when I zoom in one phase, sample a attached, which does show the Im being about 3.7kA (this is for file No 3).

So looks like my measured fault level is consistent and about 3.7 – 4.2 kA, depending if it does progress to phase A too or nor, and when I add say at least 600Amp more (that would be 25 turbines contributing at only 120% In), the actual fault current is probably about 4.5kA average.

Thank you again for your help.

 

[davidbeach]

Maybe I'm just used to it, but I find the SEL ossilography much easier to work with the way it is. Actually the viewer doesn't scale anything, that's the way the numbers come out of the relay. If I want to compare a waveform against an instantaneous setting, or the pickup of a time overcurrent element, I don't want to have to through in a computation involving sqrt(2). If the Siemens software doesn't give exactly the same results as the SEL software, that means that the Siemens software is doing scaling all on its own, and that shouldn't be done; ever. It should just present the numbers to the engineer exactly as they are supplied by the relay. Otherwise there is no telling how to compare results against settings.

If you click on a trace in the SEL viewer, you can see the values in the upper left corner of the window, no need to guess what the values are.

 

[ters]

Siemens software does not do anything strange when opening Siemens or for example ABB or GE files.

However, when opening SEL files, it seems to be equally confused as a live person is when looking at SEL instantaneous charts – when it looks at the SEL records, it sees sinusoidal form as the SEL device records it, with, in this case, peaks ploted at about 4kA. Based on this, Siemens software will plot its own instantaneous wave form which looks exactly as SEL form does.

But, when you "ask" Siemens tool to open a SEL files and show RMS forms (there are two handy buttons to toggle between instantaneous and RMS values) it does something which in my mind is logical – it does not plot a line which sits on the top of something which looks as sine form peaks and calls that line RMS (what SEL for some reason seems to be doing), but it rather calculates the RMS with the assumption that the real sine peaks are at 4000A resulting in RMS being ~ 2800Amp (4000/1.41).

And it does exactly the same for voltage – for a 34.5kV system, healthy L-G voltage is about 20kV (RMS), sine form peaks being at 28kV. However, SEL elects to show those peaks at 20kV instead, so when opening SEL files Siemens tool sees that actual peaks being at 20kV (does not "think" that real peaks are higher at 20*1.41 = 28kV), therefore it decides to show RMS as 14kV (20/1.41) instead 20kV.

It may be possible to somehow "explain” to Siemens tool that SEL peaks are actually not peaks, and that real sine form peaks are for a 1.41 factor higher, but I don’t know how to do it.

 

[davidbeach]

OK, but why even try? Just use the SEL software with SEL event reports.

 

[jghrist]

Also consider that unless you request a RAW event from the SEL relay, what you get is filtered data (fundamental only). The relay responds to filtered data, so this is often the useful quantity to view. If the raw data has a lot of distortion, this could point to particular causes (maybe inverter problems) that get hidden when viewing filtered data.

 

[pwrengrds]

On further reflection I think that the fault current is to high for the fault to be on the secondary side(fault current would be limited to less than 1000 amps), plus the fuses make that less likely. The lack of ground fault current does not mean anything on an ungrounded system. The fault is less likely to be in the Tx because of the healing nature of the fault. It doesn't seem to be wind related due to one of the faults is with the current is just magnetizing current (20 amps which is typical for the capacitance of the cables). This leaves something like flash over on an insulator as a better guess. I think that you will want to invest in some better fault detectors to find it.

 

[magoo2]

Do you have arrester cabinets for the 35 kV cable? If so, I would check the terminations in that cabinet.

 

[ters]

Arresters are installed at all places where UG cables going to turbines connect to OH lines and also inside the termination cabinet of each turbine transformer which is the last one in a daisy chain (there is 1 - 3 turbines per each UG daisy chain).

But if arresters are a suspect, that means that at least two of them are failing or flashing over at exactly the same time, which is still possible, but can something like that come with such a long lasting and random intermittency without having a permanent fault ?

Since there are two low impedance grounding transformers in the substation (one for each feeder), relatively good ground fault detection does seem to exist and having two undetected ground faults which always occur at exactly the same time and with both fault current gradients being the same (B and C), is sort of unlikely to happen through insulators or arresters, but possible.

 

[magoo2]

Although I mentioned arrester cabinets, it's the terminations in those cabinets that I would be checking.

 

[acog]

ters,

Why do you have two earthing transformers at the substation?

Why not just one earthing transformer connected to the delta of the larger 230/35kV transformer which is utilised by both collector circuits?

 

[davidbeach]

A grounding transformer per circuit means that the circuit continues to have a ground source even with the collector breaker open. Worst case, breakers is tripped open for a ground fault on the collector circuit. Before the generation all goes away, assuming grounding transformer on the system side of the collector breaker, the voltage on the unfaulted phases jumps to 34.5kV to ground. Not pleasant for arrestors rated 20kV or anything else on the circuit. Don't plan on using 34.5kV equipment on an ungrounded system, it isn't really rated for that use.

Even with wye-delta-wye 230-34.5kV transformers, collector circuits we have with delta high-side on the tower transformers all have grounding transformers at the switchyard on the collector side of the breakers.

 

[acog]

Thanks for your thoughts davidbeach.

My understanding is that all normal equipment should be factory tested for power frequency withstand voltages in excess of the phase to phase voltage.

For example, 70kV for 60 seconds is applied to all equipment with a 33kV voltage rating plate in IEC land as a part of factory acceptance tests.

So, although the stress on the equipment and cable insulation is greater than that which would be seen with an earthing transformer for each collector circuit, some may argue that the equipment IS rated for the short overvoltage seen after the fault is disconnected by the collector circuit breaker and before the turbine converters disconnect.

Surge arrestors could be specified to have a turn on voltage in excess of the phase to phase voltage such that the insulation co-ordination of the system is correct and the surge arrestors do not operate.

Grounding transformers are expensive.

 

[ters]

Use of grounding transformers is required on delta-delta systems by the power utility but I don’t know if it was them imposing one per feeder (I was not involved in the design). In any case, as davidbeach explained, they are there to also suppress voltage on healthy phases in case one phase is faulted to ground.

35kV equipment is designed for maximum operating voltage of 38kV, equipment is factory hipot tested at 80kV 60Hz withstand and BIL 150kV (or more in some cases).

Re arresters, the way how they are rated seems to be rather different in IEC and ANSI worlds. I have recently worked on an unusual case where I had to choose IEC arresters for a 27.6kV system in Ontario, since wind turbines came from Europe with a 36kV GIS IEC spec switchgear requiring IEC cable elbows (which are not interchangeable with similar ANSI 35kV elbows) and hence IEC arresters which plug piggyback to elbows were required.

So while searching for those arresters, I concluded that the IEC does not offer more than one MCOV arrester rating within the same equipment voltage class. Seems to be as simple as this: equipment voltage class rating L-L = arrester MCOV rating L-G. For example, for 36kV class L-L, the only available arrester is 36kV MCOV (obviously L-G). For 24kV class, the only arrester is 24kV MCOV, etc.

However, the IEEE world offers several arrester MCOV ratings for the same equipment voltage class what allows for a bit better insulation coordination. For example, 35kV IEEE equipment class offers several arresters with different MCOV to choose from, such as 22kV, 24kV, 27kV and 29kV. This luxury to choose appear not to exist in the IEC world.

This difference is probably closely related to the fact that IEEE/ANSI/CSA allow for about 17 different system voltages in the medium voltage category (anything from 2.4kV to 46kV in steps of 2-3kV), while Europe has just few, seem to be not more than about 5, systems voltages in the MV category.

 

[ters]

Would you gents agree with me that the CT phasing is reversed here?
Please look at the attached phasor diagram This is a pre-fault condition, some 3 cycles before the fault, with 3 symmetrical currents RMS 650Amp, which represents normal operating conditions at full load (which is ~ 37MW).

The voltage measurement is via three PTs L-G.

Current vs voltage angle (IA vs VB) is ~ 200 DEG and is consistent for all phase. This cannot be right for normal operating conditions, so I suspect that most likely CT wiring is reversed on the SEL351 device so A current is actually C, C is actually B and B is actually A.

This would make much more logical sense, the angle between voltage and current would be much more “normal”, about 45 DEG, which is still a bit wide, but rather possible depending what the PF correction system at the substation was doing at that time.

 

[pwrengrds]

No, I would not agree, its just back feeding and near unity power factor. The towers are near unity PF, the var load of the padmounts TX make it look lagging (needs some capacitance to achieve unity). The feeder is set up to look out at the towers and the current is feeding toward the Main TX so it looks backwards.

 

[ters]

Thank you David, I overlooked that, I was confused with a recent similar application of SEL351 which had to look the other way (to also detect faults on the line), where when generating the angle was not 200 DEG but more like 10 DEG. In this case it does not matter, they do not use any directional elements, but still at low outputs that angle is nowhere near as per the attached chart – if the first one is near unity , then this PF is sort of near zero which is again puzzling but possible given the collector length and the fact that turbine reactive capability is such at low kW it can deliver or absorb very little kVARs.