comparing PV panel performance

[arborlinden]

I'm not quite understanding the percentages. Are you saying that the spread in measured performance is .93/5.55=16.7%?

Based on sample size that to be approximating a 2 sigma distribution spread then you have 4.2% (1 sigma) unit to unit variation.

I don't know what panel you are talking about so I just grabbed a spec sheet from here

http://pdf.wholesalesolar.com/module%20pdf%20folder/CHSM6612P-305_Specs.pdf?_ga=1.95436063.269488771.1434675899

It shows a 46 +/-2 degC operating temperature variation (figure 2 sigma)

Based on the MPP temperature sensitivity of -0.45%/DegK you could expect +/-0.9% (2 sigma) variation per this manufacturers specs.

There is no other manufacturer specification on variability at least for this product (assuming others are similar) and there is a gap of almost 10:1 in 4.2% variation measured and .45% quoted.

If it is a temperature variation then expect a 10 degreeC variation for 1 sigma.

Go get a temperature gun and measure the variations in panel temperature to see if that would explain some portion of what is going on.

I calculated what each panel's declared output is as a percentage of total recorded cumulative output of the system. Each panel contibutes approx 1/18th of total the actual figures being given in txt filepanelpercentcontribution.txt and total pretty near 100%. The highest panel contibution panel 2 is 5.9429422% and the lowest is panel 9 at 4.9757028: a difference in contribtion of 0.9% of total.
p1 is west of top row and p9 is east of top row. p10 is east of bottom row and p18 west of bottom row. The whole array faces due magnetic south within a degree or two.
I accept that this is not a proper statistical interpretation of performance and the combined percentages do not add up to exactly 100%.
No shadowing or screening detectable and wind and cloud and tempreture variations I would have thought would have evened out over 21 days.
My panels are JA solar percium JAML-60-285/pr with optimizers

[arborlinden]

@posplayer
you might want to try reading that spec sheet again. Those panels have a +/-3% STC power tolerance. Many of the better panel companies will have +X W / -0 W tolerance.

@OP
Here is a link to my public SolarEdge page. I have about 2% difference in lifetime (~3 mo) production between my best and worst panel. I don't really have much shade right now, but for a few weeks in there I had some evening shade that would hurt a couple of the panels more than others. Panel power tolerance and thermal effects probably explain most of the observed difference, or possibly it is just calibration error in the optimizer data. SolarEdge is not really accurate better than +/-5%.

If your installer gave you the ability to run reports, you can dive off into the weeds look at the difference in panel production throughout the day. It should help you see if the production difference is consistent (suggesting calibration or mismatch), or varies as a function of time of day, temp, wind, or the other variables that might affect it.

thank you will have look at your public page when it wakes up in the morning. My installer has not got around to setting me up one yet but when he does it will obviously be beneficial to let experience PV users view and comment on it.
So far I have not been able to see anything that would explain the reported panel differences though there was one time when the lowest output panel was recording more output than the highest output panel but other than that the panel position in the output league seems pretty consistent

[arborlinden]

JPM has done some serious measurements, see

http://www.solarpaneltalk.com/showthread.php?22057-The-dirt-on-my-panels-after-6-months&p=163608&viewfull=1#post163608

He wrote, for instance:


So think about temperature differences, and look at your panels' datasheet to see how temperature affects output.

He also cautions that measuring to within 1% is more difficult than it sounds.

I will get the thermometer out and compare output and surface tempreture to see if there is a relationship

[posplayr]

I calculated what each panel's declared output is as a percentage of total recorded cumulative output of the system. Each panel contibutes approx 1/18th of total the actual figures being given in txt file[ATTACH]7255[/ATTACH] and total pretty near 100%. The highest panel contibution panel 2 is 5.9429422% and the lowest is panel 9 at 4.9757028: a difference in contribtion of 0.9% of total.
p1 is west of top row and p9 is east of top row. p10 is east of bottom row and p18 west of bottom row. The whole array faces due magnetic south within a degree or two.
I accept that this is not a proper statistical interpretation of performance and the combined percentages do not add up to exactly 100%.
No shadowing or screening detectable and wind and cloud and tempreture variations I would have thought would have evened out over 21 days.
My panels are JA solar percium JAML-60-285/pr with optimizers

It is getting late so this will be quick. I plotted the data for a clue. The Panel 9 sticks out like a sore thumb. It is 12.6% low with an standard deviation in the data of only 2.49% Something is amiss, your measurement or the panel?

Given that data point is an aberration of some sort,I did a quick analysis of what was left over.
The mean value of the remaining 17 panels is 5.69 KwHr
The standard deviation of the remaining 17 is .14 KwHr which is 2.49% of the average which is right in line with the other 2% number in the spec sheet I looked at (so did not look at yours).

To visualize this look at the plot. It includes #9 but the statistics are as stated above for the "normal" 17(excluding #9).
You will note that +/- 2 sigma (about +/-10% of 5.69 KwHr is 95.4% of the population). That would be from 5.40 KwHr to 5.98 Kwhr.

To generate the plot I ordered the Kw-Hrs and plotted against the % they contributed to the total (again excluding #9). You can see that given the sample size it is a pretty "normal" looking plot (that means Gaussian ).

UK_array.jpg

[J.P.M.]

We still have not heard back from the OP if he is really seeing 16.7% min to max spread, but the binning performance would imply much lower unit to unit variation (10:1) and indicating either a combination of measurement errors or environmental to get to 4.2% 1 sigma.

OP says all panels are on one string; not sure how he determined individual panel performance.

OP says there are no other discernible physical differences, don't know if he took a IR thermometer to the panels.

If it is not unusual, and there are no marked thermal differences then it would appear to be erroneous measurements?

BTW the Binning resolution indicates the same NOCT variation through the specified sensitivity. -%/DegK

Posplayr: I'm of the opinion that you are out of your depth with respect to solar energy. Except to flag what I believe to be errors in your posts that can cause problems and that involve things I think I may know something about, I'll not be commenting to your posts in the future.

Arborlinden: More opinion and to repeat myself, from what you report, I think your array is performing nominally.

[DanKegel]

the actual figures being given in txt file [ATTACH]7255[/ATTACH] and total pretty near 100%. ...
My panels are JA solar percium JAML-60-285/pr with optimizers

The data sheet for your panels, http://www.jasolar.com/uploads/files...500_C3xPXj.pdf,
says "temperature coefficient of Pmax is -0.36% / degree C" and "power tolerance -0, +5%".

Per-panel monitoring isn't meant to be accurate enough to diagnose minor
problems; I think it's meant to find big problems. So I doubt it's more than 5% accurate.
It's possible that your best performing panel reads 5% hot, and the worst 5% cold,
simply due to monitoring inaccuracies. Add the 5% tolerance of the
panel itself, and you'd need a 15% variation between best and worst panel
before worrying. (But I have no idea which brand of optimizer you have,
nor how accurate its monitoring really is, nor much about that data you sent,
and I've been known to make math errors, so take that with a grain of sun-dried salt.)

That text file indicates that one panel is putting out 8% less than the 2nd least-performing panel, and
17% less than the best-performing panel, so this might be tiptoeing into the area where
something actually is slightly amiss.

To account for a 5% difference in output with temperature, the panel would
have to be 15 degrees C hotter, which seems just barely possible.

So you might have a slightly weak panel. What does your installer say?

[sensij]

The data sheet for your panels, http://www.jasolar.com/uploads/files...500_C3xPXj.pdf,
says "temperature coefficient of Pmax is -0.36% / degree C" and "power tolerance -0, +5%".

I don't know what data sheet you are looking at. The one that I think belongs to these panels is this one, showing tolerance in W, not %. Also, the P coefficient is -0.39%/deg C.

So you might have a slightly weak panel. What does your installer say?

If what the OP is stating is accurate, agreed. The warranty allows 3% degredation in the first year. Proving it is difficult, and I think a closer look at the SolarEdge data available is warranted before initiating some kind of warranty service request.

[DanKegel]

[QUOTE=sensij;164536]I don't know what data sheet you are looking at. ... tolerance in W, not %. Also, the P coefficient is -0.39%/deg C.

Same data sheet. Different eyes. I made two typos/thinkos while copying info into my post (pathetic, isn't it?). 5W tolerance on a 285 watt panel is about 2%. But that plus the 3% first-year degradation Sensij mentioned brings us back to 5% tolerance, maybe.

Anyway, a single slightly weak panel is fine as long as it isn't an indicator of future worse degradation. I have no idea whether the right thing to do is watch and wait for a worse change, or to think about replacing it. As Sensij says, a closer look at the data seems appropriate. And getting your installer's opinion might be useful.

[arborlinden]

It is getting late so this will be quick. I plotted the data for a clue. The Panel 9 sticks out like a sore thumb. It is 12.6% low with an standard deviation in the data of only 2.49% Something is amiss, your measurement or the panel?

Given that data point is an aberration of some sort,I did a quick analysis of what was left over.
The mean value of the remaining 17 panels is 5.69 KwHr
The standard deviation of the remaining 17 is .14 KwHr which is 2.49% of the average which is right in line with the other 2% number in the spec sheet I looked at (so did not look at yours).

To visualize this look at the plot. It includes #9 but the statistics are as stated above for the "normal" 17(excluding #9).
You will note that +/- 2 sigma (about +/-10% of 5.69 KwHr is 95.4% of the population). That would be from 5.40 KwHr to 5.98 Kwhr.

To generate the plot I ordered the Kw-Hrs and plotted against the % they contributed to the total (again excluding #9). You can see that given the sample size it is a pretty "normal" looking plot (that means Gaussian ).

[ATTACH]7256[/ATTACH]

I too think that panel 9 stands out like a sore thumb. The measurements are the SolarEdge monitoring portal readings and if there is not an aberration there I think the panel is in error. The question is, based on the SolarEdge monitoring, do I have sufficient evidence to ask the installer to invoke JA Solar guarantee and ask for a replacement panel?

[arborlinden]

Posplayr: I'm of the opinion that you are out of your depth with respect to solar energy. Except to flag what I believe to be errors in your posts that can cause problems and that involve things I think I may know something about, I'll not be commenting to your posts in the future.

Arborlinden: More opinion and to repeat myself, from what you report, I think your array is performing nominally.

Thank you. I appreciate your taking the time to look at my array and to give me your considered opinion

[arborlinden]

The data sheet for your panels, http://www.jasolar.com/uploads/files...500_C3xPXj.pdf,
says "temperature coefficient of Pmax is -0.36% / degree C" and "power tolerance -0, +5%".

Per-panel monitoring isn't meant to be accurate enough to diagnose minor
problems; I think it's meant to find big problems. So I doubt it's more than 5% accurate.
It's possible that your best performing panel reads 5% hot, and the worst 5% cold,
simply due to monitoring inaccuracies. Add the 5% tolerance of the
panel itself, and you'd need a 15% variation between best and worst panel
before worrying. (But I have no idea which brand of optimizer you have,
nor how accurate its monitoring really is, nor much about that data you sent,
and I've been known to make math errors, so take that with a grain of sun-dried salt.)

That text file indicates that one panel is putting out 8% less than the 2nd least-performing panel, and
17% less than the best-performing panel, so this might be tiptoeing into the area where
something actually is slightly amiss.

To account for a 5% difference in output with temperature, the panel would
have to be 15 degrees C hotter, which seems just barely possible.

So you might have a slightly weak panel. What does your installer say?

I did ask my installer to have a look at the figures and he promised to do so but he has not come back to me. I was rather hoping to get some definate opinions one way or the other from forum experts before going back to him. As I understand it he monitors all his installations each day and if SolarEdge monitoring does not present him with any red crosses all is well in his world. I think it unlikely he will act if all I can give him is the possibility that one panel may be a bit below par. what I do not know if an installer has any equipment that can determine if a panel is within or without the production spec whilst it is in situ.

[DanKegel]

Accurate and precise measurements are hard. You could swap two of your panels
(but not the optimizers) and see if the problem travels with the panel.
If not, then it's not the panel's fault. Might be cheap as a DIY test for
an experienced person, but expensive if you have to pay someone to try it.

If one panel is slightly weak, and doesn't get any weaker, the cure might be worse than the disease.

What's the loss in revenue from the weak panel? Pretty small at the moment, I bet.

If it were me, absent any knowledge about whether weak panels tend to fail later,
I'd probably just wait and watch to see if it got any worse.

[arborlinden]

If it were me, absent any knowledge about whether weak panels tend to fail later,
I'd probably just wait and watch to see if it got any worse.

I think that is fast becoming the consensus here. Looks like I need to put it alongside my prostate on my watch and wait list!

[sensij]

I think that is fast becoming the consensus here. Looks like I need to put it alongside my prostate on my watch and wait list!

I sent you a PM... you might not have seen it if your notifications aren't set up.

[posplayr]

The data sheet for your panels, http://www.jasolar.com/uploads/files...500_C3xPXj.pdf,
says "temperature coefficient of Pmax is -0.36% / degree C" and "power tolerance -0, +5%".

Per-panel monitoring isn't meant to be accurate enough to diagnose minor
problems; I think it's meant to find big problems. So I doubt it's more than 5% accurate.
It's possible that your best performing panel reads 5% hot, and the worst 5% cold,
simply due to monitoring inaccuracies. Add the 5% tolerance of the
panel itself, and you'd need a 15% variation between best and worst panel
before worrying. (But I have no idea which brand of optimizer you have,
nor how accurate its monitoring really is, nor much about that data you sent,
and I've been known to make math errors, so take that with a grain of sun-dried salt.)

That text file indicates that one panel is putting out 8% less than the 2nd least-performing panel, and
17% less than the best-performing panel, so this might be tiptoeing into the area where
something actually is slightly amiss.

To account for a 5% difference in output with temperature, the panel would
have to be 15 degrees C hotter, which seems just barely possible.

So you might have a slightly weak panel. What does your installer say?

In statistics you would add variances so standard deviations would RSS.

5% + 5% ==> 7.07%
5% + 5% + 5% ==> 8.7%

Having said that a 5% tolerance up and down as you say would still be a 5% tolerance.