Interconnection at Ungrounded Subpanel

[Garrison]

We have a small M215 based system installing on a garage that is remote from the main building. The planned point of interconnection is at an existing 100A, 240V subpanel located in this garage. The issue of concern is that this panel has only two hots and a neutral, but no EGC and no ground bus. Conduit for feeder cables is in the ground for part of the run, but then the three cables are direct buried to the home. There is a good amount of concrete in between the two buildings which makes adding an EGC challenging.

To resolve this issue my electrician suggested we install a ground rod at the existing subpanel in the garage, to ground both the existing panel as well as land our combined GEC/EGC from the M215/array.

There is no inspection required in this area, so my goal is to ensure that both our installation as well as the electrical service are safe, but not necessarily satisfy the Code (although it would be ideal if we could do both). What do you think? Should I be concerned, or could this solution be safe and functional?

Thank you for your thoughts.

 

[jaggedben]

Per your last sentence, I'll address safety without worrying too much about the letter of the code...

As far as functioning under normal operation (i.e. barring a fault), the lack of equipment grounding does not present any issue. The inverters will turn on and generate with only two hots and neutral. This is also true of the ground-fault detection features of the inverters, as long as you bond all the metal parts of the array together as detailed in the Enphase instructions.

Adding a ground rod and grounding the panel is a good idea from a personal safety standpoint and may help a bit to protect from lighting damage. You should definitely do all that, and it's also required by code.

The tricky question is whether to bond neutral to ground at the garage panel. I would say you should do it, since you have apparently do not have a metal raceway or other continuous metal between the main house and garage. ( If there is a water pipe at the garage that connects back to the house then either don't bond the water pipe or do not connect neutral to ground at the garage.) I believe this is actually allowed by the code (see 250.32), although I think I recall a lively discussion or two on this forum concerning that subject. In any case, bonding neutral to ground at the garage will ensure that a fault from a hot to a bonded metal part will trip the Enphase breaker, which otherwise might not happen.

 

[ggunn]

Per your last sentence, I'll address safety without worrying too much about the letter of the code...

As far as functioning under normal operation (i.e. barring a fault), the lack of equipment grounding does not present any issue. The inverters will turn on and generate with only two hots and neutral. This is also true of the ground-fault detection features of the inverters, as long as you bond all the metal parts of the array together as detailed in the Enphase instructions.

Adding a ground rod and grounding the panel is a good idea from a personal safety standpoint and may help a bit to protect from lighting damage. You should definitely do all that, and it's also required by code.

The tricky question is whether to bond neutral to ground at the garage panel. I would say you should do it, since you have apparently do not have a metal raceway or other continuous metal between the main house and garage. ( If there is a water pipe at the garage that connects back to the house then either don't bond the water pipe or do not connect neutral to ground at the garage.) I believe this is actually allowed by the code (see 250.32), although I think I recall a lively discussion or two on this forum concerning that subject. In any case, bonding neutral to ground at the garage will ensure that a fault from a hot to a bonded metal part will trip the Enphase breaker, which otherwise might not happen.

Isn't there a potential problem with the neutral bonded to ground in two places? Wouldn't it be better to run a ground from the subpanel back to the service?

 

[jaggedben]

Isn't there a potential problem with the neutral bonded to ground in two places? Wouldn't it be better to run a ground from the subpanel back to the service?

It would be better, yes, but the OP has stated it would be difficult to do. I believe 250.32 allows what I suggested, assuming the original installation was compliant when it was installed. In any case, unless there is a parallel conductive path to the grounded conductor - and the OP has suggested there isn't - then there will be no objectionable current resulting from bonding the neutral in two places. It wouldn't be unsafe in any way that's obvious to me, really not so different from the way each service gets a N-G bond.

 

[GoldDigger]

It would be better, yes, but the OP has stated it would be difficult to do. I believe 250.32 allows what I suggested, assuming the original installation was compliant when it was installed. In any case, unless there is a parallel conductive path to the grounded conductor - and the OP has suggested there isn't - then there will be no objectionable current resulting from bonding the neutral in two places. It wouldn't be unsafe in any way that's obvious to me.

Objectionable current is not the only potential problem. If ground is bonded to neutral and the neutral carries current, then interruption of the neutral wire could cause grounded metal to rise to a potential limited only by the OCPD and the resistance of the local earth bond. (I won't call it a ground. )

 

[Smart $]

Objectionable current is not the only potential problem. If ground is bonded to neutral and the neutral carries current, then interruption of the neutral wire could cause grounded metal to rise to a potential limited only by the OCPD and the resistance of the local earth bond. (I won't call it a ground. )

That'd be two concurrent faults: 1) open feeder neutral, and 2) a line-to-ground fault.

So what would be the difference if the neutral wasn't bonded to ground at the subpanel?

 

[GoldDigger]

That'd be two concurrent faults: 1) open feeder neutral, and 2) a line-to-ground fault.

So what would be the difference if the neutral wasn't bonded to ground at the subpanel?

It would only take one fault: open feeder neutral.
There can be current (although probably small) applied to the neutral conductor inside the inverter even though the neutral is not involved in the high power inverter output circuit. Also will be a problem if a convenience outlet is fed using the same neutral line, or other similar wiring added later.

If the neutral is not bonded to ground at the subpanel, there is no way that voltage appearing on the neutral will be applied to nominally grounded conductive surfaces.

I admit that it is a stretch.

 

[GoldDigger]

....I admit that it is a stretch.

Extending the reasoning one step farther. (Have to keep picking at it....)

Given that as long as the inverter does not under normal conditions source any current at all into the neutral, and that there is no way for a line-to-neutral load to be easily connected inadvertently, this is definitely a two-failure scenario.

But what makes this somewhat different from the typical double fault is that if the neutral truly does not carry current, an interrupted neutral will never be detected. (The inverter will be happy with the reference "neutral" voltage supplied by the ground.) From there on, as the first fault remains undetected indefinitely, it would only take the addition of a line to neutral fault to heat up the ground.

 

[jaggedben]

Objectionable current is not the only potential problem. If ground is bonded to neutral and the neutral carries current, then interruption of the neutral wire could cause grounded metal to rise to a potential limited only by the OCPD and the resistance of the local earth bond. (I won't call it a ground. )

Grounded metal could rise in potential with either one of two things:
- if ground is bonded to neutral, an open feeder neutral and a path from hot to panel neutral through a load
or
- a hot to ground fault if neutral isn't bonded to ground.

Which is more likely to happen? I'd say probably the latter. Especially with AC cables exposed to sunlight on a solar array.

 

[Garrison]

Thanks all for your thoughts. So it sounds like if we do not bond ground to neutral in the subpanel we risk compromising PV OCPD functionality, and if we do bond ground to neutral we run the risk of parallel fault paths. Hmm. 250.32(B) seems pretty clear that an EGC is required to be run along with the feeders. I see a concrete saw in my future.

It is a good question JaggedBen posed...to paraphrase, "how would bonding ground to neutral in the subpanel differ from a standard service off the utility?"

Thanks all

 

[jaggedben]

(The inverter will be happy with the reference "neutral" voltage supplied by the ground.) From there on, as the first fault remains undetected indefinitely, it would only take the addition of a line to neutral fault to heat up the ground.

Hmmm... I wonder. The voltage would probably drop due to the high resistance through the earth and would quitely likely fall out of the inverter's required voltage range.

 

[GoldDigger]

It is a good question JaggedBen posed...to paraphrase, "how would bonding ground to neutral in the subpanel differ from a standard service off the utility?"

Answer in two double parts, and in the form of questions:
1. What would the current return path be in the case of a fault on the premises side of the service bond? What would the consequences of an interrupted neutral on the service side be?
2. Do you have control over what is done on the POCO side of the service? Do you have control over what is done between the subpanel and the main panel?

 

[Smart $]

It would only take one fault: open feeder neutral.
There can be current (although probably small) applied to the neutral conductor inside the inverter even though the neutral is not involved in the high power inverter output circuit. Also will be a problem if a convenience outlet is fed using the same neutral line, or other similar wiring added later.

If the neutral is not bonded to ground at the subpanel, there is no way that voltage appearing on the neutral will be applied to nominally grounded conductive surfaces.

I admit that it is a stretch.

Extending the reasoning one step farther. (Have to keep picking at it....)

Given that as long as the inverter does not under normal conditions source any current at all into the neutral, and that there is no way for a line-to-neutral load to be easily connected inadvertently, this is definitely a two-failure scenario.

But what makes this somewhat different from the typical double fault is that if the neutral truly does not carry current, an interrupted neutral will never be detected. (The inverter will be happy with the reference "neutral" voltage supplied by the ground.) From there on, as the first fault remains undetected indefinitely, it would only take the addition of a line to neutral fault to heat up the ground.

To sum it up, there is more unsafe potential with neutral isolated from ground at the subpanel than with them bonded. After you get done theorizing you'll see I'm correct...

 

[GoldDigger]

To sum it up, there is more unsafe potential with neutral isolated from ground at the subpanel than with them bonded. After you get done theorizing you'll see I'm correct...

I will grant you that. And safer than any of them is to provide a real EGC to the panel.

 

[BillK-AZ]

Seems to me that if a second EGC/neutral bond is made in the remote panel, but an EGC is not installed to the remote panel, and the primary ground is opened due to some event, the phase unbalance current will then flow to ground via the remote panel neutral and may cause failure and possible burns at that location. Regardless of the PV system connection.

 

[petersonra]

The thing is I don't believe there is a compliant way to add an EGC. If you have UF cable for instance, the EGC has to be part of it. you can't just did a trench and run a separate EGC.

 

[GoldDigger]

Seems to me that if a second EGC/neutral bond is made in the remote panel, but an EGC is not installed to the remote panel, and the primary ground is opened due to some event, the phase unbalance current will then flow to ground via the remote panel neutral and may cause failure and possible burns at that location. Regardless of the PV system connection.

The way the micros are wired, there should be no neutral current unless there is a fault, regardless of any imbalance between the inverter outputs.
If you are proposing neutral current caused by a fault, we are right back to my reasoning which has been pretty nicely countered.

 

[GoldDigger]

The thing is I don't believe there is a compliant way to add an EGC. If you have UF cable for instance, the EGC has to be part of it. you can't just did a trench and run a separate EGC.

Make that no compliant way to ONLY add an EGC. You could always replace the UF.

 

[BillK-AZ]

The way the micros are wired, there should be no neutral current unless there is a fault, regardless of any imbalance between the inverter outputs.
If you are proposing neutral current caused by a fault, we are right back to my reasoning which has been pretty nicely countered.

I am not concerned about a fault in the PV system, but rather in the main service. Over the years I have come across two residences where the main neutral connection to the utility opened, allowing unballanced phase currents to place a voltage on the neutral because the main grounding was essentially open. With a remote neutral/EGC bond as being discussed, this would have the unballance current flowing through this remote bond without any OCPD protecting the conductors.

 

[GoldDigger]

I am not concerned about a fault in the PV system, but rather in the main service. Over the years I have come across two residences where the main neutral connection to the utility opened, allowing unballanced phase currents to place a voltage on the neutral because the main grounding was essentially open. With a remote neutral/EGC bond as being discussed, this would have the unballance current flowing through this remote bond without any OCPD protecting the conductors.

In the situation the OP described, there would still be no imbalance in the current delivered by the inverters. They would not care what the neutral voltage was, and if there were no ground/neutral bond the inverter ground would be just fine. If the service tried to send primary current through the secondary neutral to ground, then there would be neutral current, but only if there were a bond, or if the insulation in the inverters broke down.

In your situation there will be a damaging voltage on the neutral because there are unbalanced loads in the house. Or in other houses on the same transformer.