Troubleshooting a dodgy string

[WattsUp]

A search hasn't returned what I am looking for so I hope I can find the guidance that I seek, please.

My grid-connected domestic roof mounted system in the UK Midlands has sixteen Hyundai HiS-S245MG Silver Monosilicon panels connected as two strings of eight panels to a Sunnyboy SMA SB 3800V inverter. Installation was in early December 2011 with a Sunnybeam bluetooth data logging module added about six months later. I have a full set of generation and log files since that point. From day one of the logging module being installed I gave up trying to get the hopeless installer (long story) to fix a repeated "Insulation Failure" - though until what I describe below the system has generated about 5% ahead of my expected payback curve. It has never tripped or gone off line other than to prevent islanding during two grid failures. Other than the nag of the error message and wiring under the panels looking like a bowl of spaghetti, I've been happy.

Background:

• A few days ago I noticed that my PV system’s production was lower than expected even for the time of year
• No errors in system log file
• Checked with village friend (identical system on identical house with almost identical orientation, installed the week after mine) and found that his production is twice mine
• Suspected failed string
• I then compared my generation logs (10 minute granularity) with my friend's
• Failure happened on 22nd December at the height of a gale (confirmed by historical met data on line)

Investigation:

• Isolated AC, Isolated one DC string at breaker, reinstated AC
• Production stayed the same
• Isolated AC, Isolated the second DC string at breaker, reinstated the first DC breaker, reinstated AC
• Production fell to zero
• Isolated AC, Isolated both DC strings at their breakers, removed safety covers from breakers
• Measured input voltage of each on a grey day just before rain and with the "sun" at an oblique angle
• Suspect string: 32v DC
• Other string: 297v DC - pretty much as expected according to how I read the tech spec for the panels

What followed:

• I found that my installer has ceased trading
• Insurance Backed Guarantee won't pay for investigation or rectification until the Installer is wound up at Companies House - in process but not completed yet and no idea of timeline
• Meanwhile I am losing generation - though thankfully in the low point of the year

Question:
Before I start calling PV installers to ask for the fault to be traced and fixed (and probably paying for scaffolding too) what are the likely causes of the problem from the evidence I have provided and are there any other tests I can perform myself. The system is not currently reporting any faults despite the fact that one string is obviously poorly. The last time "insulation failure" appeared in the Sunnybeam log was a week ago so it may be a red herring.

Many thanks for your help.

[inetdog]

A search hasn't returned what I am looking for so I hope I can find the guidance that I seek, please.

My grid-connected domestic roof mounted system in the UK Midlands has sixteen Hyundai HiS-S245MG Silver Monosilicon panels connected as two strings of eight panels to a Sunnyboy SMA SB 3800V inverter. Installation was in early December 2011 with a Sunnybeam bluetooth data logging module added about six months later. I have a full set of generation and log files since that point. From day one of the logging module being installed I gave up trying to get the hopeless installer (long story) to fix a repeated "Insulation Failure" - though until what I describe below the system has generated about 5% ahead of my expected payback curve. It has never tripped or gone off line other than to prevent islanding during two grid failures. Other than the nag of the error message and wiring under the panels looking like a bowl of spaghetti, I've been happy.

Background:

• A few days ago I noticed that my PV system’s production was lower than expected even for the time of year
• No errors in system log file
• Checked with village friend (identical system on identical house with almost identical orientation, installed the week after mine) and found that his production is twice mine
• Suspected failed string
• I then compared my generation logs (10 minute granularity) with my friend's
• Failure happened on 22nd December at the height of a gale (confirmed by historical met data on line)

Investigation:

• Isolated AC, Isolated one DC string at breaker, reinstated AC
• Production stayed the same
• Isolated AC, Isolated the second DC string at breaker, reinstated the first DC breaker, reinstated AC
• Production fell to zero
• Isolated AC, Isolated both DC strings at their breakers, removed safety covers from breakers
• Measured input voltage of each on a grey day just before rain and with the "sun" at an oblique angle
• Suspect string: 32v DC
• Other string: 297v DC - pretty much as expected according to how I read the tech spec for the panels

What followed:

• I found that my installer has ceased trading
• Insurance Backed Guarantee won't pay for investigation or rectification until the Installer is wound up at Companies House - in process but not completed yet and no idea of timeline
• Meanwhile I am losing generation - though thankfully in the low point of the year

Question:
Before I start calling PV installers to ask for the fault to be traced and fixed (and probably paying for scaffolding too) what are the likely causes of the problem from the evidence I have provided and are there any other tests I can perform myself. The system is not currently reporting any faults despite the fact that one string is obviously poorly. The last time "insulation failure" appeared in the Sunnybeam log was a week ago so it may be a red herring.

Many thanks for your help.

To produce only 32V DC from the string, it is very likely that you have a short circuit to ground, probably on the - side of the panel which is at the top of the string (assuming that you have negative ground, that is.)
Shorting out the bottom 7 panels will not hurt them in any way, but will give you just the voltage of end panel on the string.
Once you (or an installer gets up on the roof and into the interconnections of the string, the problem should be obvious.

Possibly one of the panel leads was not properly secured and has been chafing through the insulation. The high winds in the storm may have been the last straw.

If you are lucky it is only a wire and not a problem inside a panel.

[Naptown]

Sounds like a ground fault
Disconnect both sides of affected string
take voltage readings from - to ground and + to ground and report what they are.

[WattsUp]

Thank you both sincerely.
I'll perform the requested readings and report back.
It is dull and raining at the moment and the Sunnybeam reports just 38 watt from the good string (the duff one is isolated at its breaker) so I will wait awhile.

[WattsUp]

Ok, here are my findings.

My inverter is hung on the outside wall of the house so I stuck a clean galvanized steel stake about 60cm / 24" in to the wet soil and then used it to take my ground leakage readings.
The positive DC cable registers 23 volt and the negative 15 volt.

[bcroe]

Fault

My inverter is hung on the outside wall of the house so I stuck a clean
galvanized steel stake about 60cm / 24" in to the wet soil and then used it to take my ground
leakage readings. The positive DC cable registers 23 volt and the negative 15 volt.

I don't believe your primary fault is a short. First, you were getting power at 297V from one
string. The other string was connected in parallel (you didn't say any breaker or fuse was open),
so the defective string was at the same potential. A short would have dragged down the good
string as well.

Instead I suggest there is an open circuit. The low voltages you read were very high impedance
sources (negligible current capability) read thru the very small leakage currents. To see if this
is correct, connect a dummy load (couple 240V 70W bulbs in series) in parallel with the voltmeter.

If the voltages at the isolated bad string drop to zero, confirmation. You will be looking for an
open circuit. The bulbs should glow nicely on the good string. Bruce Roe

[WattsUp]

Many thanks for the suggestion, but just to clarify, I did say that the breakers were open. Per the fifth bullet in the section for investigation in my original post; both strings were isolated from each other so one could not drag down the other. Each string has its own breaker so I set them to open which as well as breaking each circuit also allowed me to remove the safety covers and get to the terminals inside (there is a safety interlock). I obtained the 297v for one string and 32v for the other by measuring the potential across the INPUT terminals to each breaker. Each breaker was open circuit thus ensuring that the inverter was disconnected and each measurement was in isolation of the other.

Cheers,
WattsUp

[bcroe]

Many thanks for the suggestion, but just to clarify, I did say that the breakers were open. Per the fifth bullet in the section for investigation in my original post; both strings were isolated from each other so one could not drag down the other. Each string has its own breaker so I set them to open which as well as breaking each circuit also allowed me to remove the safety covers and get to the terminals inside (there is a safety interlock). I obtained the 297v for one string and 32v for the other by measuring the potential across the INPUT terminals to each breaker. Each breaker was open circuit thus ensuring that the inverter was disconnected and each measurement was in isolation of the other. Cheers, WattsUp

I was referring to when you first noticed REDUCED output. At that time the strings were in
parallel, and a short on one would have killed the other also. Bruce

[WattsUp]

Thank you for explaining, Bruce.
It's just gone midnight here so I'll make up the test leads later in the day - kids permitting!

[WattsUp]

A short update to say that family duties mean there has been no chance today to try the trick with resistive load.
With the sun shining from a clear blue sky though (no clouds = consistent generation) I tried breaking and making the defective string several times using its DC breaker.
On each occasion there was no effect; throughout the hour involved generation never varied outside the range 725-780 Watt regardless of whether the defective string was in circuit or not.

As an aside; generation of this order is a good 400 Watt below what I would expect from a single string for the time of year and prevailing weather, so it looks like the "good" string isn't in full health either.

[bcroe]

Tests & Failures

A short update to say that family duties mean there has been no chance today to try the trick with resistive load.
With the sun shining from a clear blue sky though (no clouds = consistent generation) I tried breaking and making the defective string several times using its DC breaker.
On each occasion there was no effect; throughout the hour involved generation never varied outside the range 725-780 Watt regardless of whether the defective string was in circuit or not.

As an aside; generation of this order is a good 400 Watt below what I would expect from a single string for
the time of year and prevailing weather, so it looks like the "good" string isn't in full health either.

This is the tool I use for sanity checks on my strings, up to 450VDC and 8A. Just 4 bulbs in series, 75W 120V.
If your bulbs are 240V, only 2 needed; pick a wattage to throw 5 to 10% load on your string. This throws
enough load on to eliminate any very high Z leakage or static that a DVM might read, and the bulbs give a
direct indication of power. Sensitive enough to read current from a string end to a ground. Leave it
connected and use your DVM if precision is needed.

If some cells are bad, a single string might continue to operate. It would have a lower MPPT voltage, because
current would be flowing through a bypass diode instead of through the bad cell and those near it. My panels
have bypass diodes that would drop 1/3 of the panel cells if operating (pretty common), and near another volt
for the diode. If you had cells knocked out I'd expect to see fractional panel loss of voltage, current about the
same as before. Power down 1/3 ought to be a big voltage change. Bruce Roe

[WattsUp]

Thank you Bruce. A very helpful elaboration for when I shortly do the test on my installation.

[WattsUp]

A courtesy update rather than silence. No chance to get anywhere near further troubleshooting. It'll be at the weekend if I am lucky with anything earlier a bonus.

[WattsUp]

Yes, I have an open circuit on the duff string.
Thanks for your suggestion and clear explanations, Bruce.
Most helpful.

All panels to be removed, tested, and refitted properly with tidy clipped-on wiring during the week of 10th February.

[russ]

Glad to hear of your success!