Some PV questions

[electrofelon]

Might be doing some grid PV in the near future so trying to get up to speed....what do I have correct:

1) In general I can choose to have the DC system grounded or ungrounded if I meet their respective requirements.

2) Where over-current protection is required such as where strings are paralleled, if the system is grounded I only need a fuse in the ungrounded conductor. If the system is ungrounded I need a fuse in both conductors.

3) A transformerless inverter will automatically make the DC system grounded because there is no isolation from the utility system (assuming grounded utility system)

4) Does going DC grounded/ungrounded require coordination with the inverter used?

5) Im a little confused: What is the difference in ground fault detection required by 690.5 and 690.35(C)? Why not just reword 690.5 to read "grounded and ungrounded....."

Thanks!

 

[ggunn]

Might be doing some grid PV in the near future so trying to get up to speed....what do I have correct:

1) In general I can choose to have the DC system grounded or ungrounded if I meet their respective requirements.

2) Where over-current protection is required such as where strings are paralleled, if the system is grounded I only need a fuse in the ungrounded conductor. If the system is ungrounded I need a fuse in both conductors.

3) A transformerless inverter will automatically make the DC system grounded because there is no isolation from the utility system (assuming grounded utility system)

4) Does going DC grounded/ungrounded require coordination with the inverter used?

5) Im a little confused: What is the difference in ground fault detection required by 690.5 and 690.35(C)? Why not just reword 690.5 to read "grounded and ungrounded....."

Thanks!

1) not really; the inverter design supports one or the other
2) yes
3) no, transformerless inverters generally have ungrounded DC
4) yes, see 1)
5) I dunno; I don't have my code book.

 

[GoldDigger]

3) The array is not truly floating while the inverter is operating, but neither pole of the array is grounded through a full cycle.

Then there are bipolar arrays at the commercial level with their own special rules. (And never mess with an array that is off its meds!)

 

[jaggedben]

3) You have it backwards. A transformerless inverter connects only to ungrounded AC conductors because connecting an grounded DC conductor to an ungrounded AC conductor (i.e. AC-hot to ground) would result in a dead-short every half-cycle. Transformerless inverters necessarily have ungrounded arrays (with only the bipolar exception that Goldigger mentioned). Inverters with transformers have typically been grounded, but not all are and it's not required.

4) The inverter itself determines the type of grounding and the grounding connection itself is a factory configuration of the inverter. At least this is true AFAIK of all new grid-tie-only inverters, and many battery-back up systems as well. Some off-grid systems may still use a separate ground-fault protection device.

5) Perhaps the code arrangement can be better rationalized. One thing that should remain different is the required marking language. FWIW, the actual methods of GF detection are quite different between the two types of arrays. Grounded systems generally employ a fuse that needs replacing after a fault. Ungrounded systems use fancy electronics.

 

[GoldDigger]

I find the combination of an "ungrounded" array with a fused ground detector somewhat confusing.
What it actually ends up being is an array that is grounded at all times through a fuse except when there is a fault, at which point the fuse blows and removes the ground, stopping both the fault current and the operation of the inverter.
What "ungrounded" means in this sense is that you do not supply a solid external ground, but you cannot expect the array to float.
It also means that, depending on the circuitry of the particular inverter, you can still use panels that require a positive or a negative DC offset to ground for proper operation and they will work.
With ungrounded arrays with electronic GF detection, as is used with TL inverters, it is harder to predict at any given moment what the offset of either end of the array with respect to ground will be. And there will be differences between the line to line 3-phase and line to line single phase 120/240 configurations

 

[ggunn]

3) You have it backwards. A transformerless inverter connects only to ungrounded AC conductors because connecting an grounded DC conductor to an ungrounded AC conductor (i.e. AC-hot to ground) would result in a dead-short every half-cycle. Transformerless inverters necessarily have ungrounded arrays (with only the bipolar exception that Goldigger mentioned). Inverters with transformers have typically been grounded, but not all are and it's not required.

Perhaps I am misinterpreting what you have written (and it has been a while since I have dealt with small inverters), but transformerless inverters require connection to the AC neutral conductor for voltage reference. IIRC, you cannot, for example, connect SMA TL inverters phase to phase with no neutral connection the way you can the transformer interconnected SMA inverters.

 

[jaggedben]

Perhaps I am misinterpreting what you have written (and it has been a while since I have dealt with small inverters), but transformerless inverters require connection to the AC neutral conductor for voltage reference. IIRC, you cannot, for example, connect SMA TL inverters phase to phase with no neutral connection the way you can the transformer interconnected SMA inverters.

So... to clarify....

I was speaking of the connections to the distinct power-electronic inverter component(s) inside the device, not the 'inverter' as a engineered product comprising many different components and features which are sold in an enclosure. :blink:

Or, IOW...
Yes, most inverters require an AC neutral connection for voltage measurement, but no current actually flows from the PV conductors to this AC neutral conductor. That's typically true of both types of inverters in any case.

EDIT: I also mistyped. It should have read: "A transformerless inverter connects only to ungrounded DC conductors because..."

 

[jaggedben]

I find the combination of an "grounded" array with a fused ground detector somewhat confusing.
What it actually ends up being is an array that is grounded at all times through a fuse except when there is a fault, at which point the fuse blows and removes the ground, stopping both the fault current and the operation of the inverter.
What "ungrounded" means in this sense is that you do not supply a solid external ground, but you cannot expect the array to float.
It also means that, depending on the circuitry of the particular inverter, you can still use panels that require a positive or a negative DC offset to ground for proper operation and they will work.
With ungrounded arrays with electronic GF detection, as is used with TL inverters, it is harder to predict at any given moment what the offset of either end of the array with respect to ground will be. And there will be differences between the line to line 3-phase and line to line single phase 120/240 configurations

Your post is a little confusing... Is my fix in red correct?

I agree that a conductor 'grounded' through a fuse is not typical of 'grounded conductors' and could be misleading. Hence the required label in 690.5. (I've even considered it somewhat debatable if this conductor should be required to be colored white. But then, that's generally the easiest way to indicate which conductor is grounded through the fuse.)

 

[ggunn]

Yes, most inverters require an AC neutral connection for voltage measurement, but no current actually flows from the PV conductors to this AC neutral conductor. That's typically true of both types of inverters in any case.

Many single phase AC inverters can operate without a neutral connection. SMA transformer coupled inverters, for instance.

 

[electrofelon]

3) You have it backwards. A transformerless inverter connects only to ungrounded DC conductors because connecting an grounded DC conductor to an ungrounded AC conductor (i.e. AC-hot to ground) would result in a dead-short every half-cycle.

Oh yeah of course, dont know why i didnt see that.

Let me get this straight: So a grounded PV system with a typical GTI has an ungrounded, grounded, and EGC run back to the inverter. The EGC is connected to the grounded conductor with a fuse. If there is a fault from the ungrounded or grounded conductor to, say, the PV racking, current will flow thru the EGC and blow the fuse. Yes? Then in practice where is the GEC connection to the grounded conductor typically made (I see what 690.42 says, but does the inverter care about that connection - I dont see that it would?)

I read the article in solarpro about the Bakersfield fire (http://solarprofessional.com/articl...ield-fire-a-lesson-in-ground-fault-protection). The authors view is that we should not have grounded PV arrays because grounded conductor faults are hard to detect. I cheered to myself when I read that as the sometimes irrational obsession the industry has with connecting things to dirt is a pet peeve of mine.

 

[ggunn]

Oh yeah of course, dont know why i didnt see that.

Let me get this straight: So a grounded PV system with a typical GTI has an ungrounded, grounded, and EGC run back to the inverter. The EGC is connected to the grounded conductor with a fuse. If there is a fault from the ungrounded or grounded conductor to, say, the PV racking, current will flow thru the EGC and blow the fuse. Yes? Then in practice where is the GEC connection to the grounded conductor typically made (I see what 690.42 says, but does the inverter care about that connection - I dont see that it would?)

In a PV system with grounded DC, the connection between the (usually) negative conductor and the EGC is made inside the inverter through its ground fault detection circuitry. Any connection you make between the negative and the EGC external to the inverter will defeat the GFI function of the inverter.

I'm pretty sure that the system in Bakersfield was a grounded system. The problem was that the first ground fault was between the negative conductor and the conduit (EGC), which was undetectable by the inverter. It was the second ground fault from the positive conductor to the conduit that started the fire by shorting positive to negative through the conduit and there's nothing the inverter could have done about that.

 

[electrofelon]

In a PV system with grounded DC, the connection between the (usually) negative conductor and the EGC is made inside the inverter through its ground fault detection circuitry. Any connection you make between the negative and the EGC external to the inverter will defeat the GFI function of the inverter.

And where does the grounding electrode per 690.41/690.42 fit in?

I'm pretty sure that the system in Bakersfield was a grounded system. The problem was that the first ground fault was between the negative conductor and the conduit (EGC), which was undetectable by the inverter. It was the second ground fault from the positive conductor to the conduit that started the fire by shorting positive to negative through the conduit and there's nothing the inverter could have done about that.[/QUOTE]


Another question regarding that article. Here is an excerpt:

These older inverters did not interrupt the fault path; the ground fault would remain after the inverter shut down. Subsequently, however, Article 690.5 in the 2008 NEC required that most PV systems include a ground-fault protection (GFP) scheme that both detects and interrupts the fault. As a result, 3-phase inverter designs started employing a ground-fault fuse, similar to the 1 amp GFP fuse used in residential string inverters. A large inverter, however, might use a 4 amp GFP fuse, since large arrays can have over 1 amp of current flowing in the equipment ground at full irradiance under normal operating conditions.

This one amp of current normally flowing on the EGC, is that just from inductive and capacitive effects, perhaps primarily from all those singe conductors run along metallic racking?

Thanks

 

[ggunn]

And where does the grounding electrode per 690.41/690.42 fit in?



Another question regarding that article. Here is an excerpt:

The GEC does not connect directly with any of the DC conductors, it only connects the EGC to the electrode.

The problem with the Bakersfield fire was that the faults that caused the fire created a loop that did not include the inverter at all; it was actually two ground faults - one on the positive conductor and one on the negative conductor. The first one that occurred was on the negative conductor which went undetected by the inverter and defeated the GFI in the inverter so that it did not detect the second fault. If the positive conductor ground fault would have occurred first it would have been detected and the inverter would have shut down, but if no one had gone up on the roof to investigate the fault the fire would have happened anyway. Nothing that inverter or any inverter could have done would have stopped a short between the positive and negative conductors through the conduit.

The root cause of the fire was a sloppy installation whereby the conduits containing the DC conductors were installed incorrectly. Any and all safety measures built into inverters cannot overcome a bad installation.

 

[SolarPro]

It's especially difficult to overcome a bad installation when the inverter has a ground-fault blind spot. In theory, the Bakersfield Fire can't happen under the 2014 Code, which requires that the ground fault protection is able to detect ground faults on grounded conductors. But it seems like the inverter safety standard itself needs to be revised.

To the OP, if you liked the Bakersfield Fire article, you might also find this article useful: Ungrounded PV Power System in the NEC. The article explores (and hopefully clarifies) many of the topics discussed in this thread.

 

[ggunn]

It's especially difficult to overcome a bad installation when the inverter has a ground-fault blind spot. In theory, the Bakersfield Fire can't happen under the 2014 Code, which requires that the ground fault protection is able to detect ground faults on grounded conductors. But it seems like the inverter safety standard itself needs to be revised.

Even if the inverter would have faulted when the negative conductor shorted to ground the fire still would have happened if no one went up on the roof to investigate the fault; the second fault took the inverter out of the loop. I have heard plenty of stories about PV systems that were shut down for long periods of time before anyone thought to check on why they were off. Rapid shutdown at the array would have stopped the fire, of course.

 

[SolarPro]

The first fault was in a grounded PV source circuit conductor. Given the ground fault detection blind spot, that first fault could have been in the system from the day that source circuit was installed. However, if the inverter did not start up during system commissioning, the fire never would have had an opportunity to start.

The first fault could also have developed at some point after system commissioning. However, an inverter without a ground fault detection blind spot would have shut down as soon as the first fault occurred. Presumably, a service technician would have been dispatched to remedy that first fault before the second fault happened in the PV output circuit.

Point being, these faults did not happen at the moment same time, and there is no reason to assume that they even occurred in the same calendar year. If the two faults had somehow been related to the same "failure," then, yes, nothing could have stopped the fire. But that was not the case at the Bakersfield site.

 

[SolarPro]

(Also, this was a positively-grounded PV system, so the first fault occurred in a positive source circuit conductor. This has nothing to do with the cause but gives you some idea about whose system it was. The subsequent reports tend to use a generic set of fault analysis diagrams.)

 

[jaggedben]

Let me get this straight: So a grounded PV system with a typical GTI has an ungrounded, grounded, and EGC run back to the inverter. The EGC is connected to the grounded conductor with a fuse. If there is a fault from the ungrounded or grounded conductor to, say, the PV racking, current will flow thru the EGC and blow the fuse. Yes?

Correct. At least that's the idea. It works pretty well with an ungrounded conductor fault and sometimes also works with a grounded conductor fault.

Then in practice where is the GEC connection to the grounded conductor typically made (I see what 690.42 says, but does the inverter care about that connection - I dont see that it would?)

The fuse is the connection, and the inverter does care because the inverter also measures continuity through the fuse and shuts down if continuity is lost. That's the 'I,' for interruption, in 'GFDI.'

Practically speaking, the fuse is inside the inverter product and all you typically have to do is connect the GEC to the marked terminal, and on occasion (hopefully rarely) replace a fuse. Reading the inverter install manual is your best strategy here. Another thing to mention is that transformerless inverters have swiftly overtaken isolated inverters for new installs, just in the last year or so. With any luck if you get into the business now you may never have to replace a GFDI fuse.

I read the article in solarpro about the Bakersfield fire (http://solarprofessional.com/articl...ield-fire-a-lesson-in-ground-fault-protection). The authors view is that we should not have grounded PV arrays because grounded conductor faults are hard to detect. I cheered to myself when I read that as the sometimes irrational obsession the industry has with connecting things to dirt is a pet peeve of mine.

:thumbsup: (Bill Brooks is a smart guy.)

And where does the grounding electrode per 690.41/690.42 fit in?

See 690.47.

Another question regarding that article. Here is an excerpt:
...
This one amp of current normally flowing on the EGC, is that just from inductive and capacitive effects, perhaps primarily from all those singe conductors run along metallic racking?

I believe it's primarily due to capacitive effects of the PV cells themselves that allow leakage current through the frames. I don't believe induction is an issue because this is DC.

 

[ggunn]

The first fault was in a grounded PV source circuit conductor. Given the ground fault detection blind spot, that first fault could have been in the system from the day that source circuit was installed. However, if the inverter did not start up during system commissioning, the fire never would have had an opportunity to start.

I am not trying to be argumentative, but since the two faults that caused the fire were between DC conductors and the conduit, what has whether the inverter was operating and whether it could have detected the grounded DC conductor fault have to do with it? Other than alerting someone that something was wrong when the first fault happened, I mean. Simply shutting down the inverter would not have prevented the second fault from occurring. As I said, I know of cases where systems sat unrunning for months before someone even called for service.

Also, since the first fault was due to conduit pulling out of a coupling and abrading the insulation of the conductor, it seems unlikely to me that it would have been there since day one.

Please do not misunderstand me; I am wholly in favor of all ground faults being detectable.

 

[SolarPro]

since the first fault was due to conduit pulling out of a coupling and abrading the insulation of the conductor,

That's not the first fault; that's the second fault. We know that the second fault occurred in the PV output circuit, because that fault would have shut the inverter down. The first fault occurred in a grounded PV source circuit conductor.

Since first fault was in USE-2 run in free air, the insulation could have been pinched by the racking, or worn raw over time. If it was a pinched conductor, that is a classic ground fault that should have been detected during commissioning. If the fault had occurred in an ungrounded source circuit conductor, it would have been detected.