Botched installation

[ggunn]

We sent our master electrician and a helper out to a site where we had a PV system installed by a contractor (who is long gone, of course) a couple of years ago to investigate why the system wasn't running. They determined that the problem had to be in the 150' underground DC run from the array combiner to the inverter. Our master stood next to the inverter and had the helper close the DC disco at the array, and he began to hear a sound like frying bacon emanating from the conduit. They pulled the wire and found that the contractor had somehow managed to slit the insulation on both the DC+ and DC- conductors. Apparently when the contractor meggered the run they found something wrong, pulled the wire, found the compromised insulation, taped it up, and pulled the wire back into the conduit, with predictable results over time when water got into the conduit. Both conductors were badly burned; the DC- only had 2 or 3 strands left.

Since it was a + to - fault and the insulation on the EGC was intact, the GFI on the inverter never saw it.

 

[Dennis Alwon]

Yep, he got his money and ran

 

[SolarPro]

Arc fault detection would have caught that. Good argument in favor of the new Code requirements, I suppose. This application might not technically need an arc-fault capable inverter, but at come point in the future most inverters on the market will have this capability by default.

 

[ggunn]

Arc fault detection would have caught that. Good argument in favor of the new Code requirements, I suppose. This application might not technically need an arc-fault capable inverter, but at come point in the future most inverters on the market will have this capability by default.

I don't think so. Most arc detection I have seen is sensitive to series faults but not parallel faults, as this was. Regardless, shutting down the inverter would not have cleared the fault.

 

[GoldDigger]

I don't think so. Most arc detection I have seen is sensitive to series faults but not parallel faults, as this was. Regardless, shutting down the inverter would not have cleared the fault.

Interesting. I have heard exactly the opposite argument, that parallel faults are more reliably detected than series faults. I have seen demos of repeated series fault production with no trip.
Hmm. I guess it depends in part whether you think the signature of an intermittent, arc-connected, resistive load is that of a series or a parallel fault.

 

[GoldDigger]

I don't think so. Most arc detection I have seen is sensitive to series faults but not parallel faults, as this was. Regardless, shutting down the inverter would not have cleared the fault.

Interesting. I have heard exactly the opposite argument, that parallel faults are more easily detected than series faults. I have seen demos of repeated series fault production with no trip.
Hmm. I guess it depends in part whether you think the signature of an intermittent, arc-connected, resistive load is that of a series or a parallel fault.

There may also be some differences between DC AFCIs and AC ones. I know little if anything about DC AFCIs.

 

[ggunn]

Interesting. I have heard exactly the opposite argument, that parallel faults are more reliably detected than series faults. I have seen demos of repeated series fault production with no trip.
Hmm. I guess it depends in part whether you think the signature of an intermittent, arc-connected, resistive load is that of a series or a parallel fault.

Look at 690.11 in the 2011 NEC. It's obvious they are talking about series arc fault detection. A DC+ to DC- fault is neither a series arc fault nor a ground fault. Even the new SMA's wouldn't have caught it.

 

[SolarPro]

Good point. Parallel arc fault detection requirements were proposed for NEC 2014, but I don't think they were accepted.

Based on discussions with manufacturers, I suspect that detection isn't really a problem. The challenge probably has more to do with the fact that the proper response to a series arc-fault situation (open the circuit) is fundamentally different than the appropriate response to a parallel arc-fault situation (short the circuit). Arguably, the fastest way to eliminate either problem is at the source.

I've heard some inverter manufacturer representatives complain that they are tired of being told to fix safety problems that they feel are the module manufacturers responsibility to solve. They point out that the inverter is not the power source in the circuit or the source of potential fault currents. While it's hard to argue with that point of view, its also a dangerous argument for an inverter manufacturer to make. After all, module-level power electronics could make centralized inverters obsolete.

 

[ggunn]

Good point. Parallel arc fault detection requirements were proposed for NEC 2014, but I don't think they were accepted.

Based on discussions with manufacturers, I suspect that detection isn't really a problem. The challenge probably has more to do with the fact that the proper response to a series arc-fault situation (open the circuit) is fundamentally different than the appropriate response to a parallel arc-fault situation (short the circuit). Arguably, the fastest way to eliminate either problem is at the source.

I've heard some inverter manufacturer representatives complain that they are tired of being told to fix safety problems that they feel are the module manufacturers responsibility to solve. They point out that the inverter is not the power source in the circuit or the source of potential fault currents. While it's hard to argue with that point of view, its also a dangerous argument for an inverter manufacturer to make. After all, module-level power electronics could make centralized inverters obsolete.

I can see their point. How would an inverter differentiate a parallel fault between DC+ and DC- from a normal response to darkness? Even if the inverter could detect it and were to go to a dead short, a short farther up the line would still present a lower impedance to the array.

 

[jaggedben]

Parallel arc faults are probably at least as likely to happen through metals parts that should be grounded as they are directly from the DC+ conductor to DC- conductor. Thus GFDI provides a measure of protection against them, because it should set an alarm when the first half of the fault occurs, before the second half does.

Series arc faults (e.g. loose connectors) aren't detected by GFDI. So I think that's why the 690.11 requirement was put in there.

All of that is notwithstanding the example that started this thread. It's hard to see how an inverter could do anything to stop such a parallel arc fault once it occurs, which makes it almost a moot point whether it detects it or not.

 

[hurk27]

AFCI's and GFCI's do not detect faults in a DC circuit, the only thing that would have removed the DC power to these DC lines would have been fuses at the supply end of the conductors and only if the current reached the tripping curve of the fuse, many times a high resistance fault like this will not reach this rating of current, GFCI's and AFCI's operate on AC principles, and these devices can not protect conductors ahead of them, I have never heard of any AFCI that will operate on DC, you could use an electronic detection circuit that might detect a ground fault but if the fault was line to line as th eOP said, and in PVC conduit it would most likly not detect any ground fault, also it would have to be located at the supply end of the DC circuit to remove the power in the circuit.

 

[jaggedben]

AFCI's and GFCI's do not detect faults in a DC circuit,

Well, of course, devices designed for AC will probably not work in DC. No one is saying they will...

the only thing that would have removed the DC power to these DC lines would have been fuses at the supply end of the conductors

And no one designing a solar system would ever put such fuses at the supply end. That would defeat the goal of putting through the max power to the grid (or batteries).

GFCI's and AFCI's operate on AC principles

I'd say you're thinking of devices designed for AC. There are reliable ground-fault interruption devices for DC and solar PV specifically. They do not typically interrupt the circuit close the supply, but there's no technical reason they couldn't. The 2014 code will require interruption capability near PV arrays and thus probably change what is available on the market as listed devices.

I have never heard of any AFCI that will operate on DC,

I have. One is included in the new SMA TL-12 inverters. I have no idea how exactly it works or how reliable it is, but I have heard of it. In fact it has been mentioned in this thread.

if the fault was line to line as th eOP said, and in PVC conduit it would most likly not detect any ground fault, also it would have to be located at the supply end of the DC circuit to remove the power in the circuit.

That's all true, although I have heard of a couple cases of ground faults being detected in PVC conduit when it became flooded with water. Not line-to-line faults, just line to ground via the water.

 

[GoldDigger]

I can see their point. How would an inverter differentiate a parallel fault between DC+ and DC- from a normal response to darkness? Even if the inverter could detect it and were to go to a dead short, a short farther up the line would still present a lower impedance to the array.

Because a panel array is a strictly current limited source, a dead short at the inverter would not draw much more current than the normal load, but would drop the voltage along the whole wiring network to the point that the other fault would either stop conducting or would in any case not be dissipating enough energy to be dangerous. If a short closer to the panel continued to conduct with a short at the inverter, it would not be a hazard.
(Ok, if the array voltage is 500 volts and the DC conductors are sized for a 2% voltage drop, there could be as much as 10 volts across the other short, but nothing like the same current at 500 volts would do.)

 

[BillK-AZ]

......
And no one designing a solar system would ever put such fuses at the supply end. That would defeat the goal of putting through the max power to the grid (or batteries).
....

Large systems use DC sub-combiners to combine the outputs of regular DC combiners. These sub-combiners have fuses to protect the cables to the regular DC combiners from other parallel regular DC combiners.

The second quoted sentence does not make any sense, properly selected over current protection does not waste power.

 

[ggunn]

AFCI's and GFCI's do not detect faults in a DC circuit, the only thing that would have removed the DC power to these DC lines would have been fuses at the supply end of the conductors and only if the current reached the tripping curve of the fuse.

That cannot happen. Any fuse that would interrupt the current flow in the event of a DC short would also trip under operational full load conditions unless it was set to trip between Isc and the maximum "legal" Imp it would ever see. Fuses aren't that precise. Fuses on the DC side are not there to protect conductors, anyway, but to protect strings in parallel from backfeeding each other. The conductors are sized to handle well over Isc.

 

[jaggedben]

Large systems use DC sub-combiners to combine the outputs of regular DC combiners. These sub-combiners have fuses to protect the cables to the regular DC combiners from other parallel regular DC combiners.

I think you didn't read through the thread carefully.

Suppose the faulted circuit that the OP described was going to a fused subcombiner instead of directly to the inverter. Even if all the fuses in the subcombiner opened, power could still flow through the fault.

The second quoted sentence does not make any sense, properly selected over current protection does not waste power.

The sentence made perfect sense. See ggunn's last response to this thread stated it a little more clearly. Properly selected overcurrent protection won't trip or blow in a line to line PV fault. "Such fuses" that would stop such a fault would not be properly selected fuses.

 

[SolarPro]

I have never heard of any AFCI that will operate on DC,

See 690.11 in the 2011 NEC, which requires dc arc-fault circuit protection. Eaton was quick to adapt its existing ac AFCI technology for dc applications. SMA and Fronius both have listed inverters with this capability. Some combiner box manufacturers have products with this capability as well.

I suspect enforcement of 690.11 is spotty right now, just because some of the biggest solar markets are still operating under the 2008 Code. Also, when the 2011 Code was released, there were no listed dc AFCI devices, which meant that the requirement was moot. But there are definitely jurisdictions where it is enforced now.