Acceptable spread in panel performance

[Jon Fisher]

We just got a new solar system installed on a flat roof in Arizona, the system consists of 14 Sun Power 327W panels installed in two rows of 7 on a fixed rack, with SolarEdge optimizers and one SolarEdge SE7600H inverter. The system produces the output expected for our environment of around 30 kWh/day, however we have noticed marked differences in individual panel performance. Three panels are running up to 18% below the calculated average for all 14 panels and the spread between the highest and lowest producing panel is 33%. There is no shading on the panels, the three that are lower in production are not located on the outer edge, and the panels are all clean.

We contacted the installer to get information about the normally observed spread in panel output, however we did not get an answer other than to be content that the expected daily output is achieved.

Any insight what spread to expect would be appreciated.

[J.P.M.]

I've got 16 ea. Sumpower 327's and their output probably varies by maybe some fraction of 1 % one to another.

As a max. variance, panels of the same type running side/side in the same location ought to be putting out +/- the stated tolerance as on the data sheet.
My panel's stated tolerence is +5/-3%. but they're going on 7 yrs. old. Newer 327's tollerence is tighter. Check your data pansl sheet.

When was your array installed ?
When was the last time the array was cleaned ?

The array is on a "flat" roof. I'll assume that means the roof is close to horizontal. Is the plane of the array mostly parallel to the plane of the roof it sits on ?

If the array has a mostly horizontal orientation, it'll foul more quickly, particularly in the desert. Horizontal or shallow pitch panels/arrays I'm familiar with in Borrego Springs (so.CA desert w/climate quite similar to Phoenix/other desert areas) turn into shallow sandboxes in about a week or two and performance rolls off rather quickly. I'm a bit surprised the bobcats don't use them for cat boxes.

FWIW, closer to the coast (zip 92026) and at a 20 deg. tilt, my array fouls at a rate such that the performance decreases ~ 0.75 - 0.9 %/week if it doesn't rain or without cleaning.

Besides fouling more quickly and hard to determine, horizontal desert arrays may well foul in an irregular pattern from local wind patterns. The few arrays I'm familiar with in the desert load up in a fairly regular and repeating pattern as f(wind pattern, location on a roof) for any one location, particularly for horizontal roof "Santa Fe style buildings w/ 2-3 ft. parapets w/ arrays close to the parapets. BTW : Any possibility that parapet/other things are shading some of the array early/late in the day ? Just wonderin'.
What's the location of the 3 underperforming panels relative to the other panels. That is, what's their location in the array ?
What's the location of the panels with the higher output(s) ?
Has the reporting system been checked ? Seems to me folks on this forum report problems w/SolarEdge stuff, including the monitoring system with some regularity and more often than other mfg. stuff.
Has the system wiring been checked ? Optimizers ?

I'd start by cleaning the array - in the morning BEFORE the panels get a chance to warm up to avoid panel damage from thermal shock - to eliminate to influence of fouling, and then keeping an eye on output. If the decreased output pattern persists on a clean array, it ain't dirt.

If someone put a gun to my head and made me guess, I'd go for a monitoring/reporting problem or a wiring problem. After that, the optimizers would be on my radar. But all that's no better than a dart throw whole wearing a blindfold without being there.

You paid for performance. That includes performance. It appears - to you at least - you're not getting it. I'd ask the installer to explain what you see as anomalies in a way you can understand or fix the system.

Welcome to the neighborhood.

[bcroe]

Your array might be capable of 4.6KW peak power, I would not worry if the inverter limits
or clips below that level. But if operating within inverter capacity, I think 33% below average
is way outside the proper range. I suppose you can identify the panels, check them for any
obvious fault. A couple things to check, is the DC current at the panel lead (not the optimizer
output) normal, and go over those panels with an infrared temp gauge for a hot spot. Find
one, get rid of that panel.

If the equipment can report individual panel voltage and current while at peak, it might
reveal something. good luck, Bruce Roe

[bcroe]

The other question is, with no shade, why are you using optimizers? Bruce Roe

[Jon Fisher]

Thanks Bruce. We used the optimizers as this seems to be the standard offered and used by the installer, the nice thing about the optimizers is that you can easily get individual panel data. I followed what you suggested and looked at the various output values, tabled below:

Panel #
9: 3.87 (A), 183.77 (W), 47.50 (V)
8: 4.01, 191.80, 47.88
7: 4.14, 212.05, 48.12
13: 4.15, 199.72, 48.12
10: 4.19, 198.68, 47.38
11: 4.26, 202.17, 47.50
12: 4.37, 210.79, 48.25
4: 4.39, 214.20, 48.75
6: 4.44, 210.78, 47.50
14: 4.53, 216.93, 47.88
12: 5.41, 260.85, 48.25
1: 5.44, 261.68, 48.12
5: 5.46, 264.63, 48.50
3: 5.51, 264.30, 48.00

Any insights you could share would be great. Regarding checking for hot spots I am unclear how to do that as our ambient temperatures are so high (~110F) during the day. I could try with an infrared temp gauge early in the morning. Any advise?

[solarix]

On a flat roof, the rows of panels will almost certainly be shading themselves to some degree in early morning or late afternoon as the shadows will get very long.

[bob-n]

The SE7600H should be capable of 8000+ watts, so I don't think this is a inverter limitation issue (sometimes called clipping).

It could be that some of the panels are defective. It could also be that you are seeing awful variation in accuracy of the optimizers. Or it could be that your particular roof has different levels of sunlight, such as shading from a power pole or chimney. Or it could be that something is really wrong.

The solar world talks of something called "standard test conditions" or STC, which is an exact level of light and exact environmental conditions like temperature. Your panels are rated to produce 327 watts at STC. Sunlight at midday in Arizona should be relatively close to STC. Your air temperature is higher than STC, which means less power.

The Sun Power E20-327-COM is rated -0.35% per degree C. If roof air is 110F, then panel temperature is roughly 130F due to sunlight warming the panel. That's roughly 30 degree C above STC, so you should expect ~11% less power due to heat. Could be 9%, could be 13%, but something in the range of 290 watts per panel at noon on a clear day.

Seeing power levels 180 watts to 260 watts really concerns me.

Many of us like a solar modeling tool called PVWatts. You can run it on the web at pvwatts.nrel.gov With this tool, you enter your panel, your roof orientation and angle, your location, and get the expected production power from your array. It would be good to run this calculation to confirm that something is wrong. If it tells you 30kWh per day in June, then perhaps the only issue is the accuracy of the optimizers. But I'll bet it tells you that you should be getting significantly more.

With that calculation and your measurements, you will be in a good position to get things corrected by the installer for free.

[bcroe]

Your table shows consistent voltages, which means no panel sections are going into bypass.
Next thing I would do is take my clamp on DC ammeter to the PANEL wires (not optimizer
output wires) and see if your readings are consistent with the table. As was pointed out, an
optimizer could be reading current badly. If that really is the correct current, I would be ready
to swap panels. Bruce Roe

[PVAndy]

Looks like shade to me. Ask the installer for the shade report at different locations on the roof.

Andy

[PugPower]

Just curious as to why you have a 7,600kW inverter hooked to a system totaling 4,578kW (14x327)? From what I understand SolarEdge inverters work most efficient when run in the upper end of the stated power range. I might also assume that you are producing less power in low light conditions because the inverter is so over sized for the panels.

[J.P.M.]

Thanks Bruce. We used the optimizers as this seems to be the standard offered and used by the installer, the nice thing about the optimizers is that you can easily get individual panel data. I followed what you suggested and looked at the various output values, tabled below:

Panel #
9: 3.87 (A), 183.77 (W), 47.50 (V)
8: 4.01, 191.80, 47.88
7: 4.14, 212.05, 48.12
13: 4.15, 199.72, 48.12
10: 4.19, 198.68, 47.38
11: 4.26, 202.17, 47.50
12: 4.37, 210.79, 48.25
4: 4.39, 214.20, 48.75
6: 4.44, 210.78, 47.50
14: 4.53, 216.93, 47.88
12: 5.41, 260.85, 48.25
1: 5.44, 261.68, 48.12
5: 5.46, 264.63, 48.50
3: 5.51, 264.30, 48.00

Any insights you could share would be great. Regarding checking for hot spots I am unclear how to do that as our ambient temperatures are so high (~110F) during the day. I could try with an infrared temp gauge early in the morning. Any advise?

Since you seem to be asking:

A couple of questions first.
What date and what time of day was that data taken ?
What's your zip ?
What's the panel tilt and azimuth ?
I
As Bruce notes, looks like voltages for that data don't vary much.

For a simple string inverter setup (not using optimizers), since panel voltage is mostly f(cell temp.) and cell temp. is f(irradiance input), uniform panel temps tend to indicate no or little shading.

But, for optimizer setups, that fairly uniform temp. is mostly due to the way the optimizers want to maintain voltages to the inverter and vary the current from a panel that's shaded or has reduced output - see the SolarEdge application note relating to fixed string voltage. The information in that note would seem to confirm the idea of maintaining fixed voltage from a panel to the inverter and varying the current of that panel, which the data you provide seems to indicate.

Seems to me you either have some shade you don't know about or something is amiss in the material or assembly of the array, including the wiring of the array.

IMO only, but from experience as well, if there is no shade on the array and all is OK, the output from each 327 will be within a percent or two of the others. If the output is as you report, it's either shade or a material/assembly problem.

Also as Bruce (somewhat rhetorically it may seem) asks, with no shade, too bad/unfortunate you got optimizers and what alos looks like an oversized inverter.

In such situations of no shade, the advantage of optimizers (such as it is spun by peddlers) is not needed and the probability of problems by virtue of violating the KISS principle by using unnecessary stuff is increased - as may be happening now.

[Jon Fisher]

Thank you all for sharing your insights, very helpful. There have been quite a number of questions asked that I like to answer:

1.) Why a 7,600 kW inverter for 4,578 kW system? We had mentioned to the installer that if the experience would be positive we would expand the system. Therefore he installed this larger unit from the onset

2.) When was the data depicted captured? Around 3pm on 06/26/2020

3.) ZIP code: 85262

4.) Panel tilt: 18 degrees

5.) Azimuth: 230 degrees

6.) Shading: None, no trees, chimneys or similar. The two rows of panels are mounted in parallel with around 4 feet of spacing in between on the highest section of a flat roof. However, I will go up there a number of times throughout the day to see whether there is any shading (e.g. as stated under spec. conditions from the front panel row)

7.) Cleanliness: Panels have been just installed a number of weeks ago and yes, Arizona is a dusty place almost throughout the year. I will follow the advise and wash them (early morning)

8.) Location of panels with lower performance: Nomenclature - two rows (A and B) of 7 panels each (1-7, counting left to right when looked at from the front).
- The panels showing lower output are at position B6 (panel #11), A4 (panel #5), and B4 (panel #6)
- The panels showing highest output are at position B5 (panel #2), A3 (panel #4) which are both just next to a low performing panel, A6 (panel #13)

9.) Installation: May 2020

10.) Check of reporting system: I routinely compare the reported daily output with what the local power company APS is measuring, the daily difference in recorded value is less than 1 kW

I ran a simulation on the proposed pvwatt webpage, with a system loss of 14% the system should produce around 7635 kWh/year and 814 kWh in June. So, with the observed daily output between 27 and 30 kWh we are in line with these projections.

So, the question remaining is why do we see such differences in panel performance, taking J.M.P.'s experience into consideration and assuming that there is no soiling of the panels this should not be the case. As next steps I will check cleanliness and check for any shading.

[J.P.M.]

Thank you all for sharing your insights, very helpful. There have been quite a number of questions asked that I like to answer:


So, the question remaining is why do we see such differences in panel performance, taking J.M.P.'s experience into consideration and assuming that there is no soiling of the panels this should not be the case. As next steps I will check cleanliness and check for any shading.

Thank you for the information.

Add: are you sure of that capture date of 06/26/2020 ? I'll check PVoutput for systems near you for any corrected date as well as check PVWatts hourly output for near date clear day outputs.

If you do not have shading, unless you have very unusual dust/fouling problems (which I'd suggest are probably not severe enough to cause the variation in panel output shown), I think you either have equipment problems, or monitoring problems, or both.

A bit off topic, but on array cleaning/fouling penalties:
Yours is a desert climate and dustier than mine, so take what follows FWIW. Your results and outcomes will most likely be different, but maybe have some common elements.
I'm in a fairly dry but not too dusty climate (zip 92026) and I've found my array, which is at a similar orientation to yours, fouls so as to decrease performance by ~ 0.75 to 0.9 % per week if it doesn't rain or I don't clean it. Also, and this seems contrary to what I'd have thought, and in spite of my best efforts to nail this down a bit harder, and for my location and array only, without rain or cleaning, the rate of fouling penalty seems to become asymptotic after about 6-8 weeks or so and stays at ~ 6 - 8 % performance penalty with no further cleaning. Then I thought of skylights that probably don't see water for months on end and still seem to pass what looks to be about the same amount of light. This is not an exact science.
Also, hosing with ~ 3/4 of a gal. of tap H2O per panel seems to restore ~ 3/4 of the performance lost due to dirt/fouling, big owl skrocks notwithstanding.
Lastly, I can't detect any performance improvement from D.I. rinsing and/or any squeegeing to remove any hard water spots vs. the simple water rinse and drip dry method. Such improvement may exist, I'm only saying it's small enough so that I can't measure a difference with the measurement methods I use.

[PugPower]

I live in a similar climate to JPM in S. Cali and will say that I don't detect much difference after cleaning my roof panels. Any small increase is just temporary as the panels just collect more dust/dirt during these dry months and return to the same condition in a short amount of time. Not worth it in my opinion.

I have discrepancies on my panels due to shading at different times of the day and vary by season caused by trees, chimney, etc. A small amount of shade can cause noticeable discrepancies between panels.

[J.P.M.]

Running PVWatts w/10% system losses, which is what a lot of folks are finding gives a closer match to actual outputs, at least for new(er) systems, and your inputs as supplied, and then getting the hourly output option and looking at modeled output for 1400 hrs. standard time for days in June, it looks to me that the clear sky output is somewhere between 3,300 Wh and 3,600 Wh or so for that hour. If my addition is close to accurate, your supplied output is ~ 2,828 Wh at 1400 hrs std. time.

Seems that's more weight on the side of unless there's some shading going on, something is not right, either with the system or the monitoring.