Ground fault location on Ungrounded system

[jado85]

Hi,

I have an Ungrounded 4000A,480V 3 phase System, to comply with NEC 250.21(B) we have ground detection lights(GDL) installed on the switchgear. A few of those systems have ground faults indicated by the GDL.

1.The down time on those systems comes at a very high premium, what is the best way to locate those faults without disconnecting the power?
2. Can I use a clamp on ammeter on all 3 phases to locate the short, based on the current reading should be close to zeroA?

 

[Tony S]

Nothing I can think of. For 40 years I’ve been shutting down ungrounded systems (AC&DC) and playing hunt the fault.

You can only detect a ground fault at source as the impedance for the various circuits is so low the same fault will show everywhere.

 

[Julius Right]

There are methods-these are not so accurate, indeed-using cable radar system. See[for instance]:
http://www.electricenergyonline.com/show_article.php?mag=&article=140

 

[GoldDigger]

The first accidental or inadvertent connection between any system potential and ground will trip the ground indicator but will not carry any current except that running through the indicator bulbs.
There may be a detectable change in wire impedance (which a fault detctor/interferometer can measure) and you will get an indication from the GDI lights as to which phase conductor is faulted. If the fault is in a motor winding or resistive load, the phase indication will be ambiguous, since the fault is not in a phase conductor.

You could (whether you should or not) introduce a second limited current "fault" to allow you to trace the real first fault with an ammeter. It would be similar to introducing a single phase load, but to ground rather than to the non-existent neutral, and as such would not be code compliant even for short term testing.

 

[Phil Corso]

Jado,

The simple answer to your question is, YES! A Clamp-on can be used.

Place it around all the feeder or branch circuit's phase conductors, if accessible. Success is dependent on the magnitude of the ground-fault current compared to the expected current carried in feeder or branch circuit! Of course, it's easier if the normal (expected) current magnitudes are equal!

Do you want to know how it was done the "old-fashioned way?

Regards, Phil Corso

 

[don_resqcapt19]

...
You could (whether you should or not) introduce a second limited current "fault" to allow you to trace the real first fault with an ammeter. It would be similar to introducing a single phase load, but to ground rather than to the non-existent neutral, and as such would not be code compliant even for short term testing.

Sort of like adding an impedance grounding resistor after the fact? That might work, but if the ground fault is a very poor connection, even a limited amount of current may cause additional damage at the point of the fault.

 

[jado85]

Thank you all for responding.

Phil Corso,
Yes I would like to know how it was done the old fashioned way,
The expected fault current would be very low(based on my readings, see image attached), someone told me it could be in the Pico-A, how does that affect the feasibility of that technique ?

Jado,

The simple answer to your question is, YES! A Clamp-on can be used.

Place it around all the feeder or branch circuit's phase conductors, if accessible. Success is dependent on the magnitude of the ground-fault current compared to the expected current carried in feeder or branch circuit! Of course, it's easier if the normal (expected) current magnitudes are equal!

Do you want to know how it was done the "old-fashioned way?

Regards, Phil Corso

 

[ron]

If another ground fault occurs before finding the first may lead to regular amounts of ground fault current to flow, which would be bad.

Lots of resources on the internet. This is an example
http://www.bender.org/knowledgebase/?a=ground-fault-location-ungrounded-system

 

[jado85]

If another ground fault occurs before finding the first may lead to regular amounts of ground fault current to flow, which would be bad.

Lots of resources on the internet. This is an example
http://www.bender.org/knowledgebase/?a=ground-fault-location-ungrounded-system

Ron,

I'm actually looking at the portable system from Bender, however Connecting the pulse generator to the system requires disconnecting the power(per company policy), that's why I was asking for other methods, using a clamp on ammeter made sense to me but that could be my inexperienced opinion.

 

[Phil Corso]

Jado, several questions:

1) Why do you think it's in the pico-range?

2) What are your present ph-ph, and ph-ground measurements?

3) Are the V-measurements changing?

4) What is approximate size of the system in kVA?

5) Is it made up of cables?

6) In metallic or non-metallic conduit?

7) Apprx age?

8) Is much of the system below ground?


Phil

 

[winnie]

What you need is some sort of signal which you can reliably detect that is going through the 'ground'.

As you have already figured out, the 'natural' 60Hz signal on the ground fault will be pretty small. You calculated what the value would be on the basis of incidental capacitive coupling. The existing ground fault detectors are likely also introducing current; the classic 'three light bulbs' means that current equal to the lamp current has to be flowing through the ground fault!

You can hunt for this by looking for 'net current' on circuits. On any correctly functioning circuit, any current flowing 'out' on one circuit conductor has to be balanced by current flowing 'in' on the others. If you put the entire set of circuit conductors into a clamp meter, the circuit current will balance out to zero. Any ground fault current will read on the meter.

The problem here is that you are going to be hunting for a _small_ 60Hz current in the context of large 60Hz currents, and there are always errors.

IMHO what would be preferable is to inject a low value of AC current that is _not_ at your normal supply frequency, so that a low current through the unintentional ground would be easily discernible from 60Hz. It looks like you are already looking at the bender system, which appears to do exactly this.

The question would be: could the bender pulse generator be installed on a branch circuit which you can turn off, such that it would inject the pulses into the system so you can locate the fault.

A second question would be: if company policy requires you to shut the system down to install the pulse generator, what are the requirements for tracing the fault assuming the pulse generator is installed?

If the pulse generator were installed, could you locate the fault without having to clamp on to circuit conductors? Could you hunt for the signal in what is supposed to be _grounded_ non-current carrying metal?

Finally, are there other signal sources that you could use? I am thinking that a VFD would be a great signal source at its switching frequency, and the _capacitively coupled_ currents at the VFD switching frequency are generally accepted flowing through the grounded metal system.

-Jon

 

[jado85]

Jado, several questions:

....

Phil

1) Why do you think it's in the pico-range? I was told so by a sales rep from Bender

2) What are your present ph-ph, and ph-ground measurements? I haven't checked any readings( the info is supplied by contractor)

3) Are the V-measurements changing? same as above

4) What is approximate size of the system in kVA? 3000KVA

5) Is it made up of cables? It feeds approximately 12 panels, that support Motors, Heaters, VFD'S ..etc

6) In metallic or non-metallic conduit? Metallic conduits

7) Apprx age? ages are 1,10,15 years( I have a total of 5 with faulted Ungrounded systems)

8) Is much of the system below ground? Mostly overhead conduit

 

[rlundsrud]

I would employ the use of a TDR. You would of course have to have the power off, but you should be able to locate the fault very quickly. The tektronix 1503C is great if you have a long cable (up to 50,000 ft) that you need to test, the 1502C is much more accurate (less than 1 inch) but is limited to less than 5000 ft. These are completely non destructive and will give distances to multiple faults. I wouldn't advise you to try it yourself, it takes some training and practice to learn how to setup and actually read the display. If you can find someone that does it in your area, it might be practical as you can minimize downtime on the system.

 

[kwired]

Jado,

The simple answer to your question is, YES! A Clamp-on can be used.

Place it around all the feeder or branch circuit's phase conductors, if accessible. Success is dependent on the magnitude of the ground-fault current compared to the expected current carried in feeder or branch circuit! Of course, it's easier if the normal (expected) current magnitudes are equal!

Do you want to know how it was done the "old-fashioned way?

Regards, Phil Corso

If there is only one fault, where is stray current going to flow to cause any imbalance in the conductors you put the clamp on?

First fault grounds the system but there is no current flow until there is a second connection to ground somewhere, second fault on same phase may result in rather low current second fault on different phase - depending on resistance of fault path causes operation of overcurrent devices.

 

[GoldDigger]

The only quantifiable current will be the indicator lamp filament, since that is phase to ground

If you have a 100% clean isolated ground/EGC, you could trace that current in the ground. But multiple ground points and building steel connections could block that method.

 

[jado85]

Thank you all, I will update this thread once I have new info or results.

 

[jado85]

I'm moving forward with the portable kit from Bender, to work around the downtime issue I think the best way to do this is to install a small panel outside the SWB that has breakers and twist lock, where we can connect the Pulse generator(Bender Device) using the lock twist then closing the breakers. In my mind that is the safest way to do this without shutting the lines down. The pulse generator uses a 16AWG banana leads and i'll modify those do connect them into a twist lock.

any inputs or thought on this idea are appreciated.

Thanks,
Jad.

 

[Jraef]

When I started at a US Steel mill 35 years ago, I was trained that if the ground light lit up, we had to run around looking for smoke! We usually found it pretty quickly, but we also had a short list of "usual suspects" to work from...

 

[jado85]

When I started at a US Steel mill 35 years ago, I was trained that if the ground light lit up, we had to run around looking for smoke! We usually found it pretty quickly, but we also had a short list of "usual suspects" to work from...

Unfortunately it's not the case at this site, the only reason we'd do that is when something isn't working. But at least now we can be more proactive having this tool in our toolbag.

A question that comes to mind, since there is not a load on the signal injecting wires will they need fusing?

 

[topgone]

If there is only one fault, where is stray current going to flow to cause any imbalance in the conductors you put the clamp on?

First fault grounds the system but there is no current flow until there is a second connection to ground somewhere, second fault on same phase may result in rather low current second fault on different phase - depending on resistance of fault path causes operation of overcurrent devices.

There is a minimal current that will flow in an "ungrounded system" if one line is accidentally grounded. That small current is capacitive and will depend on the total capacitance of the line.

To set the ground fault protection relays, we either do a "live test" or estimate the line charging current in these ungrounded systems. I tried to search the web and found this page for your info. LINK