Ultrafast tripping times

[deltawhy]

Hello,
I am wondering if anyone has run across any whitepapers or IEEE articles relating to ultrafast tripping times. I am trying to find, given our current technology, what the fastest time a relay can initiate a correct trip signal is.
As of right now I cannot find anything where the required sample size is less than one cycle. Current relays would sample for a minimum of one full cycle (0.01667s), convert the analog signal to digital, put it through a peak detection filter, and most likely a cosine filter to get rid of the harmonics and DC components. Does anyone know if it is possible to render a useful signal (where the relay can send a trip signal with high confidence that the sampled signal is in fact a fault) with a sample size less than this (perhaps 1/2 cycle)?

Thanks

 

[davidbeach]

What kind of element? How much above pickup? There are distance relays that can issue a trip in 0.8 cycles and there are also sub-cycle differential relays.

 

[deltawhy]

This would be for medium and low voltage systems, instantaneous element (50), and probably between 1-10 times pickup. For high multiple values where CT saturation poses an issue, I'm also unsure how quickly the relay can recognize the actual rms current.

 

[davidbeach]

If the 50 is not terribly far above load current it could trip as fast as a quarter cycle after fault inception if the total fault current is high enough. If the relay is set at 10A and the relay sees 60A of fault current it may well take a full cycle to register 60A, but it will cross 10A in the first 1/4 cycle or certainly in the first half. On the other hand if set at 15A and the fault current is 15.25A then it may even take 1.25 cycles to get there.

 

[deltawhy]

Okay, thanks for the responses. I now realize my question requires further info.

1. I have only seen relays which initiate a tripping signal based on the filtered rms value (ie. the signal sent by the CT and filtered for DC and harmonic components). Is this for the most part correct?
a. If so, does anyone know the typical time interval used to calculate the rms current (ie. the signal is integrated over 1/4 cycle continuously)?
davidbeach, by what you have said above, it is logical to conclude that the time to register an instantaneous trip signal is a function of the multiple of prospective fault current (ie. pickup @ 1/2 of available) and not the multiple above continuous load current. Would this be correct?

2. Has anyone seen a di/dt method used in conjunction with rms current magnitude for locating fault events?

Thank you

 

[collies99]

Our line protection relay all have similar speed for zone 1 type faults, close in, high current.

Using common relay test set, from the time if current initiation to the output trip contact closure is 10 mses for SEL-321 / 421 or ABA REL series relays.

The event record has a sple rate of only 4 times per cycle, but I do not kniw what the on-biard A/D clock cycle is. The output trip relay,ga a 4 mse response time in the spec ? And tha's from memory.

our line breakers are 1.5 or 2 cycle breakers.

 

[odlanor]

deltawhy

1-It is 1 cycle. In numerical relays;- RMS value at reference point tc is calculated by measuring window situated on the left of tc and its length corresponds to ONE period of the nominal frequency (60Hz=> 16.66ms)

 

[dpc]

If you search the SEL website, you should find some technical data regarding their sampling algorithm. IIRC, they can get a reasonable estimate of the waveform with only a few sample points - maybe only two. You can probably just contact the relay's tech support and ask the question. The relay specifications may give a response time. If the current magnitude is way above the setpoint, it doesn't take as long to make a trip decision.

 

[jghrist]

An SEL-351S has a nondirectional instantaneous overcurrent element pickup time of 1.15 - 1.45 cycles at 1.2 times pickup and 0.22 - 0.52 cycles at 10 times pickup. See Fig 3.5 in SEL-351S Instruction Manual 20130620.

 

[deltawhy]

Excellent, thanks very much for the responses.

 

[mikekilroy]

di/dt is used today in many high end vfd drives to detect output shorts in very much less than 1/4 cycle - probably in usec ranges. Just depends on your circuit and impedances. With old (1980's) short circuit protection in vfd drives that DID require 1/2-1-2 cycle times to trip, the vfd often failed before the trip; if it didn't, then it often did the 2nd or 3rd try - by the tech who thought just powering down and back up and try again would clear the fault.

If your system has inherent Z that gives a dependable worst case max current rise that is probably >10x short circuit rise time, this 'rate of rise' method should be useful.

http://www.KilroyWasHere<dot>com

 

[mikekilroy]

googled:

rate of current rise in drives

got a couple possible hits?

http://www.galeon.com/tapioca/shortcircuit.pdf

http://www.evdl.org/docs/igbt_shortoff.pdf

not sure if it helps give ideas for you but may be interesting reading and a direction to turn?

http://www.KilroyWasHere<dot>com