Bifacial panel field testing?

[DanKegel]

Although bifacial panels are currently odd ducks, several forces are converging that might make them more mainstream
( see http://www.pv-tech.org/guest_blog/bi...kwh_cost_reduc ).
"ITRPV" reports http://www.solarchoice.net.au/blog/n...eleased-040515
bifacials were a percent or two of the market last year, and projects they'll hit 10% by 2017.
They wouldn't be appropriate for flush mounting on sloped roofs, they're only for use with tilted mounts that expose the reverse side to incident light.

So I went looking again for field test data about bifacial panels.

ftp://ftp.ecn.nl/pub/www/library/report/2014/m14035.pdf measured a 2% boost for bifacials in one low-albedo location with an azimuth of 170 degrees in summer, vs. a 1% *drop* in winter. (WTF? Maybe LG is right, and adding a white backpanel will give a 1% boost.) The paper didn't say how the panels were mounted, though. #fail

http://waset.org/publications/999922...ropical-region showed an 8% boost for HIT 200W bifacial panels
relative to HIT 205W monofacial panels (and shows a picture of how they were mounted on a flat roof).

http://www.photon.info/photon_lab_mo...ults_en.photon says the October 2014 issue of Photon had some test results, but it's not online.

And Panasonic has a little best-case graph on their sales page, http://panasonic.net/ecosolutions/solar/hit/ (mounted at least a foot above a flat roof with 95% albedo).

I found several other papers, but nothing great. And I didn't see anything for 300W-class panels.

So... wtf. I am tempted to substitute two GXB300 bifacial panels into my planned array of flush-mounted LG305's, just to get a baseline of whether the GXB300's are any good as monofacial panels, and to see if I can measure any back face effect at all. (I'm using solaredge, so mixing panels is ok, and I assume I can get a log of power output for each panel.) I haven't thought it through, though. I probably want to put two bifacial panels next to each other on the west edge of the array so they get the same illumination, cover the back side of one, and then compare their output. (And then repeat with the cover moved to the other module.)

To give them any chance of having backside illumination at all, I probably need to mount them a foot off the roof. I bet my installer will just love me. If anyone's interested in helping design the experiment, please PM me.

[sensij]

It is a fun idea, and since I have a 40%+ reflective roof and some extra space, one I toyed with too. (Really, I'd love to have a kickstarter in which random people on the internet gave me money to buy extra panels and experiment with them, but sadly, I don't think it works that way).

Some things about your experimental setup to keep in mind:

1) The SolarEdge optimizers don't report live data simultaneously, and there is a randomizing factor in the timing to help avoid data collisions. I have 12 panels, and it typically is takes close to 10 min for all of them to report. 10 min is enough time that detecting a performance difference on the order of a couple percent is very hard to do, because the sun position, temperature, cloud cover, etc etc can change by more than that over that amount of time.

2) There are no canned reports or API access to panel level data. You can create a chart in the portal and download it with down to 15 min resolution (resolution is a function of how much time you want to include in the chart), but the chart interface is really aggravating to me so I haven't tried to set it up to capture the data.

3) I think comparing panel level daily energy production is most likely to yield meaningful results, but the bifacial panels have to be arranged in a way that they get the same sun as the panels to which they are being compared. Shade producers locally or on the horizon can screw up a daily comparison really fast. If you have four panels in a row (A, B, C, D), you might consider having Sunpreme in positions A and C initially, then swapped to positions B and D a week later. However, shade patterns change rapidly and so even that wouldn't assure a perfectly equal comparison.

4) Don't forget about the thermal effects of panel location. A panel on the end of a string (with three exposed sides) would be expected to have better cooling (and better efficiency) than one in the middle of an array. If you choose to mount the Sunpreme's 1 ft off the roof, the LG's should be mounted that high too (for aesthetic reasons too)

5) I would focus on the Sunpreme vs LG test first before adding in factors like covering the backside. If there is no detectable difference between the Sunpreme and LG panels, there is not much point in further exploring the contribution of backside irradiance.

One other comment... a friend was looking into Sunpower panels, and according to the SP rep, the only difference between the X21-335-BLK and the X21-345 is the color of the backsheet. So, in that case (if it is true), you have an example of backsheet color affecting the output by about 3%.

[J.P.M.]

It is a fun idea, and since I have a 40%+ reflective roof and some extra space, one I toyed with too. (Really, I'd love to have a kickstarter in which random people on the internet gave me money to buy extra panels and experiment with them, but sadly, I don't think it works that way).

Some things about your experimental setup to keep in mind:

1) The SolarEdge optimizers don't report live data simultaneously, and there is a randomizing factor in the timing to help avoid data collisions. I have 12 panels, and it typically is takes close to 10 min for all of them to report. 10 min is enough time that detecting a performance difference on the order of a couple percent is very hard to do, because the sun position, temperature, cloud cover, etc etc can change by more than that over that amount of time.

2) There are no canned reports or API access to panel level data. You can create a chart in the portal and download it with down to 15 min resolution (resolution is a function of how much time you want to include in the chart), but the chart interface is really aggravating to me so I haven't tried to set it up to capture the data.

3) I think comparing panel level daily energy production is most likely to yield meaningful results, but the bifacial panels have to be arranged in a way that they get the same sun as the panels to which they are being compared. Shade producers locally or on the horizon can screw up a daily comparison really fast. If you have four panels in a row (A, B, C, D), you might consider having Sunpreme in positions A and C initially, then swapped to positions B and D a week later. However, shade patterns change rapidly and so even that wouldn't assure a perfectly equal comparison.

4) Don't forget about the thermal effects of panel location. A panel on the end of a string (with three exposed sides) would be expected to have better cooling (and better efficiency) than one in the middle of an array. If you choose to mount the Sunpreme's 1 ft off the roof, the LG's should be mounted that high too (for aesthetic reasons too)

5) I would focus on the Sunpreme vs LG test first before adding in factors like covering the backside. If there is no detectable difference between the Sunpreme and LG panels, there is not much point in further exploring the contribution of backside irradiance.

One other comment... a friend was looking into Sunpower panels, and according to the SP rep, the only difference between the X21-335-BLK and the X21-345 is the color of the backsheet. So, in that case (if it is true), you have an example of backsheet color affecting the output by about 3%.

A comment/two on the 345/335 diff: If the only diff. is the backsheet color, and the temp. coef. of eff. is ~ .004%/deg. C., that would seem to imply that the 345's run ~ (.03/.004)~ 7-8 deg. C cooler. From my experience, that's a pretty large temp. diff.

The highest delta T. I've measured between backsheet temp. and amb. on my array is about 30/31 deg. C. under "full sun" and low wind vector mag., and usually about 22-27 deg. C or so. Seems to me, to decrease that delta T. - backsheet to amb. - by say, 5 to 7 deg. C. would require a change in the module energy balance beyond what such a change in color by itself could accomplish except under some pretty unusal or contrived operating conditions. Maybe I'm wrong, but I'd guess the 345/335 diff. in STC rating is at least partially due to tweaking other stuff, including claims, tolerances and parameters.

For example, I suspect, but can't prove, and have no way to verify, my 327's actually started out as 335's or maybe even 345's that didn't flash as expected - something lower, but still > 327 W.

If I was making panels, I might build a bunch the same, lower the tolerances a bit to save mfg./Q.C. costs, sell the less efficient to more eff. as different "models" as they flashed, and try to garner a premium for the ones that flashed higher. More economies of scale, less problems w/ Q.C. and a marketing opp. showing/implying constant product improvement among other things. Seems all upside in that sense.

[DanKegel]

If the cells are bifacial as in lg305's, a high backsheet albedo could allow a little light to come in from the side and reflect onto the cell back, capturing a bit of power that would otherwise be lost. Are they, I wonder?

Thanks for the tips!

[sensij]

Sunpower's bifacial cell is known as "Pegasus", the cells used in the panels mentioned above are Maxeon Gen III. J.P.M. is probably right... the effect is not likely to be thermal alone, but also some cell sorting. Here is a bit more about that particular cell.

[DanKegel]

Pegasus was their old one from 2003, wasn't it?

btw found a great rant about bifacial measurement at http://pvinnovation.ca/bifacial-phot...0th-ieee-pvsc/

[sensij]

I should actually read the links I post, eh? Yeah, Pegasus was old. Looking closer at the Maxeon III write up, it talks about a textured trench on the back for improved bi-facial response. So, there is at least a mechanism for backsheet color to contribute more than just cooler temps to the efficiency of the panel, although it is hard to say how significant it is.

[DreInMD]

A comment/two on the 345/335 diff: If the only diff. is the backsheet color, and the temp. coef. of eff. is ~ .004%/deg. C., that would seem to imply that the 345's run ~ (.03/.004)~ 7-8 deg. C cooler. From my experience, that's a pretty large temp. diff.

The highest delta T. I've measured between backsheet temp. and amb. on my array is about 30/31 deg. C. under "full sun" and low wind vector mag., and usually about 22-27 deg. C or so. Seems to me, to decrease that delta T. - backsheet to amb. - by say, 5 to 7 deg. C. would require a change in the module energy balance beyond what such a change in color by itself could accomplish except under some pretty unusal or contrived operating conditions. Maybe I'm wrong, but I'd guess the 345/335 diff. in STC rating is at least partially due to tweaking other stuff, including claims, tolerances and parameters.

For example, I suspect, but can't prove, and have no way to verify, my 327's actually started out as 335's or maybe even 345's that didn't flash as expected - something lower, but still > 327 W.

If I was making panels, I might build a bunch the same, lower the tolerances a bit to save mfg./Q.C. costs, sell the less efficient to more eff. as different "models" as they flashed, and try to garner a premium for the ones that flashed higher. More economies of scale, less problems w/ Q.C. and a marketing opp. showing/implying constant product improvement among other things. Seems all upside in that sense.

When I looking into getting my system I was working with one of the Sunpower Sales Engineers and was told that it was pretty much like you stated that the modules are tested and the lower performing ones went into the 327's. Kinda like Intel with their CPUs.

[sensij]

When I looking into getting my system I was working with one of the Sunpower Sales Engineers and was told that it was pretty much like you stated that the modules are tested and the lower performing ones went into the 327's. Kinda like Intel with their CPUs.

The only flaws in that story is that the 327's have been around longer than the 335/345's, and claim to use the Maxeon Gen II cells, not Gen III.

[SunEagle]

Pegasus was their old one from 2003, wasn't it?

btw found a great rant about bifacial measurement at http://pvinnovation.ca/bifacial-phot...0th-ieee-pvsc/

Interesting article and small world. While I was long gone when Dr S Hegedus came to the U of D, I got my BSEE degree there back in the mid 70's and did some research at SES/IEC on Cadmium Sufide solar cells.

Maybe the big issue with Dr Hegedus is that he was focused on thin film solar cells which have a metal backing and can't take advantage of reflected light like the bifacial technology.

[J.P.M.]

The only flaws in that story is that the 327's have been around longer than the 335/345's, and claim to use the Maxeon Gen II cells, not Gen III.

Maybe some 327's always flashed above 345. Or, maybe the mfg./design was tweaked and newer 327's were rolled into the change w/ the higher Wattage panels.

Seems I remember the +/- tol. on the S.T.C ratings changing a bit over the last few years.

Claims and buzzwords can sometimes be just as much or more a matter of marketing as technical improvement.

In any case, to my experience it's kind of hard for the average user to measure differences as small as we're discussing here, making a good part of this discussion, while interesting, mostly moot. I think I know what I'm doing and I probably can't claim more than about ~~ 1-3% accuracy when measuring array fouling if I'm simultaneously real careful and real lucky. Precision may be a bit better only because of how/when I control testing conditions.

[DanKegel]

Maybe the big issue with Dr Hegedus is that he was focused on thin film solar cells which have a metal backing and can't take advantage of reflected light like the bifacial technology.

He asked good questions, though.

[SunEagle]

He asked good questions, though.

Yes. But there are a number of ways to get more production out of a solar cell.

One way is finding cheap materials to build a solar cell that could easily bump electrons when light is shined on it. (thin film was this direction with good results with CdTe type which Dr Hegedus worked with)

Another way is to concentrate the sunlight so there are more photons hitting the material which increases production. (using mirrors or Fresnel lens to magnify the light)

Another way is to build multi-level cell where the photons pass through the first layer hit the second or third which increases production. (this is being done in the lab with results close to getting 47% efficient cells)

Another way is to reduce the sheet resistance of the cell surface so the electrons being "freed" can be harvested faster and not lost. (i did a research paper on that principle where I found that decreasing the amount of heat during lamination of the cell could increase it's efficiency).

Anyway I feel the biafacial panel has some merit and should be researched where it could prove to be a valid manufactured process.

[DanKegel]

After some thought, I realized that it's not worth doing the measurements if it's not a real world application, so now I'm thinking of doing a reverse tilt installation of a single bifacial on the north slope of the roof. That would give the bifacial the indirect back illumination it needs to get some amount of boost.
Ideally I'd put an LG305 next to it, identically mounted, to get an apples to apples comparison.
I'd briefly considered a reverse tilt mount of a few panels anyway (I couldn't fit as many panels as I wanted on the good slopes), so now I have an excuse to.

Whether I'll actually do it depends on cost (I'll pay some premium for science, but not too much), whether the chimney to the north shades it too much, and on buy-in from management

[sensij]

Whether I'll actually do it depends on cost (I'll pay some premium for science, but not too much), whether the chimney to the north shades it too much, and on buy-in from management

I hope your appeals to management go better than mine. You seem like the kind of guy who doesn't take no for an answer though