GFI on ungrounded MV system 1

[peebee]

I'm currently reviewing a submittal that indicates residual-type ground fault detection to be provided for an existing ungrounded 12,470-volt system that currently has no ground fault indication. I'm rejecting this because I do not believe this can work.

My first preference would be detection of zero-sequence voltage. Unfortunately, it is probably not possible to install PT's anywhere on this system due to space restraints.

Therefore, I'm considering zero-sequence current detection. I'm still not convinced this will work, but have not been able to find any real documentation either way.

Any thoughts?

 

[dpc]

I agree with you that trying to monitor zero sequence current probably isn't going to work. There just won't be enough ground fault current.

 

[busbar]

pb--you have an intersting but not unsolvable dilemma. Neither zero-sequence overvoltage or overcurrent relaying is directional, so it can be kinda’ tricky what should trip [or alarm.] A limitation of residual-current sensing in ungrounded systems is the lack of reliable, predictable fault current.

Voltage-based fault detection is well understood and accepted in petrochem operations. It is most often applied on 5kV-class systems, but it has been studied for 15kV-class systems and would seem a phenominal improvement for modification of ungrounded MV systems IIRC, JR Dunki-Jacobs studied it in the 1970s; published in IEEE/IAS transactions.

There are an inherent set of three phase-to-ground-connected capacitors in the form of cable shields and winding-to-core insulation. To avoid damage, capacitance has to be damped, an it can be done most cost-effectively with effectively parallel resistive means. A starting point is ~2-5 amperes per MVA of serving transformer capacity.

One approach is to install transformers {rated for continuous phase-to-phase voltage} connected in a grounded-wye/broken-delta configuration with a damping resistor on the “break” in the delta-wired PT secondaries.

Relay is usually a '59G' device that responds to relatively low fundamental voltage [id est, ~0.02-0.15 p-u] but is third-harmonic voltage restrained.

An alternaitive for a wye-source arrangement may be an appropriately rated wyepoint-to-ground resistor, directly connected, or via a 1ø transformer.

 

[jwerthman]

It is not generally feasible to utilize zero sequence current detection for ground fault relaying on ungrounded systems. However, there will be some current flow during a ground fault due to the distributed capacitance of the system. Schweitzer Engineering Labs (SEL) has an interesting approach to this problem that is discussed in a very recent application guide on their website. Go to

http://www.selinc.com

and look for Application Guide AG2002-06. I don't think this idea is really new, but I haven't seen it for awhile and it's much more feasible with the sensitivity of the newer digital relays than it was with the old electromechanical relays.