Converting 50kW system from micro inverters to 8x SMA inverters

[pclausen]

Finally, after waiting more than a month, the SMAs finally arrived!



Unpacked them all to make sure there was no damage. They all look great:



I plan to cut over one array at a time to minimize production losses.

Should be a fun weekend (except when I get to the shop roof mounted ones, but I doubt I'll get to those this weekend).

[bcroe]

Losses are going to be there over long runs. But best to avoid lines heating
up. The daily heat/cool cycle will try to loosen your connections over time.
Bruce

[pclausen]

That makes sense Bruce. I'll need to go over several connections then both at the house and down at the shop, just to be on the safe side.

I got the inverters mounted:



Nest up with be running the #6 L1/L2 wires from the inverters to the sub-panel. I'll cut them to length but not connect them to the breakers until I take down each micro inverter string.

The 6.0, 7.0 and 7.7s all appear identical except the 7.0 and 7.7s have fans on them.

[pclausen]

I got one of the arrays converted over from micro inverters to a SMA 6.0 but I'm getting an insulation error fault (code 3501). So I shut down the inverter and measured voltages and got the following readings:



So there is clearly an issue. The readings looks ok at the array, but at the DC terminals at the inverter, something seems off, especially between B- and GND. The instant I connect my probes (Fluke 17B, I get high readings on the A terminals, but the readings instantly drop down to about 20V, and then slowly go down from there to around 12V after about 60 seconds.

B- to GND reads 385V all the time. Thinking that maybe the B- wire was missing insulation, I did a fresh pull of that wire from the array to the inverter. Made no difference at all. As I pulled out the old wire, I did notice there was water in the conduit, could that account for the odd readings?

Here is what the backside of the array looks like:



Its wired as 2 strings with 12 panels in each wired as 2 sideways "U"s. I have a single 10 AWG ground wire running back to the inverter. It is connected as follows at the array:



Along the bottom front, all the rails are bonded together with 6 AWG copper wire:



The copper wire is also connected to rebar as seen here:



The whole platform contains about 7 yards of concrete with rebar throughout it.

Here's a shot of the inverter showing the separate wire I ran for for B- and you can also make out the green ground wire towards the lower left:



The next thing I tried, was to just connect a single panel back to the inverter B inputs. I still got about 36V between B- and GND at the inverter. Tried another panel, same thing. But at the array, I was getting like 0.5V between - and GND.

Any suggestions on what to try next? Maybe pull a fresh ground wire between the array and the inverter? I'll also try removing the DC block and ground wire at the inverter and see if the reading change to match what I get at the array. That will remove the inverter from the equation completely.

My other 2 SMA inverters that are running are wired in pretty much the same way and never gave me any issue like this.

[nerdralph]

Any cracked panels I would consider probable failure points
before long, and keep them apart from the rest of the system.
Bruce Roe

I've been surprised with how well (at least initially) panels with cracked glass work. I've compared Voc and Isc with other panels and found no difference. I think the issue will be eventual moisture ingress leading to the failure.

Recent research shows cracked cells (not just cracked glass) has a minimal impact on panel performance.

https://www.pv-magazine.com/2022/07/18/small-cracks-have-negligible-effect-on-solar-cell-performance/

[nerdralph]

The copper wire is also connected to rebar as seen here:

I can't say that's your problem, but I suspect the NEC says you don't ground the array separately from your bonding connector that goes back to ground at the inverter. You could get ground loops if there is a difference in the potential at the 2 grounding points.

[pclausen]

nerdralph Good to now. In the 2 years those 4 panels have been cracked, I have not noticed any degradation in performance (this is with micro inverters). I do still plan to move those 4 panels to their own string, just as a precaution.

I was wondering about grounding the ground arrays and the pole barn. Currently they all have ground rods in addition to being bonded at the inverters, which in turn are grounded via a pair of grounding rods for the shop building. The only place ground and neutral connect, is 200' away in the 400A disconnect panel. I have ground and neutral wires running between the house and shop building.

I'll go ahead and disconnect the copper bonding between the rails from the rebar, just to see if it makes a difference.

I did a bunch of additional measurements this morning:



Still seeing the ~380V between B- and GND with the strings attached.

[pclausen]

I'm currently testing just a single panel. Here it is by itself getting the expected reading:



Positive to the frame looks good:



And negative to the frame looks good as well:



I then connect the B MPPT input wires to the panel and the ground wire runs from the frame back into the shop building:



At the inverter, the single panel voltage looks good:



And B+ to the ground wire looks good as well:



But B- to ground is showing the full voltage of the panel again:



I just don't get it...

[bcroe]

I am going to assume, you have a fault to ground in either an
array or the wiring. Here is how I would check that. A super
high impedance voltmeter tends to be mislead by static charges,
and normal, very small, insulation leakage. The tool I use is 4
75W 120V bulbs in series, which can respond to a fraction of an
amp at up to array voltage. The impedance is low enough to put
some load on the panels, while instantly draining static charges
and ignoring normal tiny insulation leakage. It operates at system
voltage, in case that is needed to reveal faults. I cannot show a
picture since the new software blocks my uploads.

With the arrays disconnected from the feed wires and some sun,
first use the bulb test from ground, to either a +, then a - array
output lead. Both should give no response, otherwise that array
has a ground. The voltage on either side should give an indication
of how far down the string, the fault is. Test connection to the
array leads should light the bulbs.

Next connect the feed wires to the array, but not connected to the
inverter. Do the same test from ground, if one wire lights bulbs,
the other wire has the ground fault.

I had a momentary ground fault, that blew a GFI fuse in the inverter.
Replaced that, and all was well.

Looks like your panels are strung horizontally, like mine. Unless your
panel leads are longer than mine, that requires an extention jumper
between panels going across. I turned half my panels around, to
eliminate half the jumpers. Another way is wire them top to bottom,
then back around to the top again, before a jumper might be needed.
I could not do that, because there is a snow drop through gap which
should have no wiring crossing it. good luck, Bruce Roe

[pclausen]

Thanks for that advice! I made up a tool like you suggested and the stray readings went away completely. Here's what I get when connected between B+ and B-:



And between GND and any of the positive or negative DC terminals:



I went ahead and brought the inverter online, but it is still not producing:



Note how the insulation resistance is 1,311 kOhm on the working inverter and 433 kOhm on the one with the fault. Maybe I just need to lower the sensitivity on the 6.0 inverter?

It is currently set for 500 kOhm:



So I can lower it to like 300 kOhm and the issue should go away I assume. Need to go dig up my grid guard code since that parameter appears to be locked.

[pclausen]

I made some tweaks to the parameters:



Status went to ok but I'm still not producing any power:



And my insulation resistance actually increased all by itself:



Maybe I need to power cycle it before it starts making power?

[pclausen]

Just had to be a little more patient:


And now the insulation resistance is above 1 MOhm. I didn't change a thing. Go figure...



So I pretty much chased my tail for the last 2 days for nothing it would seem. At least I learned something and got a cool new tool for testing against ground faults!

Also, to answer an earlier question, yeah I had to make 3.5 foot extensions to go between the panels as those Ubiquiti SunMax panel have really short leads and they were custom made for use with their inverters.



And I have started my pile of inverters and connection cables:

[bcroe]

So system verified and working now? 1 Meg ohm is probably too high
for that equipment to measure with any accuracy. My own stuff with 60
panels running at 400VDC, plus a 250 foot feed, in conduit, measured
more in the area of 100 meg ohm. Either is high enough not to present
a hazard to man or machine.

Those expensive digital meters are fine for precision lab measurements,
not so good for finding simple field faults. Bruce Roe

[pclausen]

Yep all good now. Wish I had kept my old analog multimeter. I don't recall throwing it away, but I haven't seen it in years.

I had the sun peek out briefly between showers and maxed out the 6.0 inverter. Those old (circa 2016) 255 W Ubiquiti panels still got it. 24 x 255 = 6,120 W. I was not expecting them to be able to max out a 6.0, but was pleasantly surprised.



So with that finally out of the way, I completed wiring the #6 L1/L2 to each inverter, along with #10 Neutral and GND. I also did Ethernet homeruns to each inverter. Yes I know you can daisy chain them all together, but I always homerun each wired device back to the nearest switch.

[pclausen]

Got the 60 panels on the pole barn re-wired and spread across a pair of 7.7s. As mentioned before, for now I'm doing 4 strings with 15 panels in each.

I'd like to build a nice dashboard that graphs AC and DC voltages, current and power. The Sunny Portal leaves a lot to be desired. Right now with just 4 of 8 inverters running, I keep 4 windows open like this:



Full rez link since the forum really scales down the image:

https://i.imgur.com/FNxseFP.jpg

But that takes up way too much screen real-estate, even on a 49" 4K monitor, and doesn't look pretty at all. I run Home Assistant, so maybe I can log all this stuff there and build a nice dashboard? Will have to look into that.

My SMA 7.7-40 is on firmware 2.4.98.R which is the latest for the 40 series. The new 41 series I got last week are on 3.2.15.R, which newer than the latest on the SMA website, which is 3.2.13.R. Go figure. 3.2.13.R supposedly added DC string stats to Sunny Portal, but I'm not seeing it anywhere.