[myself248]

It's not mysterious, it's that thermal convection is much stronger when the panels are vertical, and production is strongly correlated with lower temperature.

Angled mounts tend to have obnoxious cross-members that block the airflow that should otherwise be sliding up the back of the panel, particularly on roofs where there's basically a bed of hot air trapped underneath with no good way to escape. That boosts the panel temperatures even further than you'd assume given simply lower convection based on their angle alone. Vertical mounts cannot have framing in these places, so they don't.

[tonyarkles]

> It's not mysterious, it's that thermal convection is much stronger when the panels are vertical

I agree that, when you factor in semiconductor physics, it's not a mystery but it isn't necessarily an intuitive result for most. I've been working in aerospace for 5 years and one of the things that has been very clear to me is that peoples' intuition about things breaks down very quickly when there's non-linear factors involved in an analysis. In aero it's primarily square-law/cube-law tradeoffs; in semiconductor physics it'll be more exponential.

For this particular problem you've got an exponential (semiconductor behaviour as a function of temperature) multiplied by a trig function/dot product (cosine of the angle of the sun relative to the normal of the solar panel), with a bit of natural thermal convection thrown in for good measure. Modelling this (digital twin, as they call it) is feasible but it's not something most people are going to have a good intuition on with respect to where the sweet spot is going to be.

[schiffern]

The cited "digital twin" software[0] doesn't model convection (it just uses wind speed and an empirical factor), which is why it gives a higher predicted temperature than the physical model.

[0] https://www.sciencedirect.com/science/article/pii/S187661021...

[dboreham]

If you run PV/battery systems you pretty soon notice that in extreme cold events the controller can shut down charging because panel output becomes so high that the batteries will be overcharged.

[jacquesm]

The air near the top is 100% trapped, there is no way for it to escape and it's the hottest air under the panel. Overlap with the cells is anywhere from 1/2" to 1.5" so that's a sizeable fraction of the cells. Probably close to 25 to 30% or so.

[CyberDildonics]

The average person's intuition and 'mysterious' are two different things. One is someone with no knowledge assuming the wrong thing, the other is experts not able to figure something out.

[jacquesm]

The panels themselves are usually made from aluminum U profile and close to the edge there is a substantial amount of trapped air if the panel is at a bad angle. Given that most panels are at a bad angle this will cause the top edge cells to all be over temp and since they're all in series that drops the efficiency of the whole panel. So the cross members certainly don't help but the panel construction itself could do with some more ventilation near the top. I wonder if cutting some slots in the top members would drop the cell temperature in a way that it would show up on a measurement, this is pretty easy to test.

[madaxe_again]

Indeed - I saw the title and thought “because convection” - I actually reinstalled my PV array last year with exactly this in mind, as while it was at an optimal angle for insolation, I was finding that yield was being hampered by them getting devilishly hot - the summer before last, when we hit 47C air, really underlined the issue, as the panels were getting up to over 85C.

The increase in yield from going near vertical (80 degrees was the best I could achieve using existing mounting gear), has been about 20% - I say about as I haven’t done a scientific study of it, just looking at year on year comparisons for cloudless days, and the panels are 60 C cooler, which is far better than I had hoped for.

[schiffern]

Yes my first thoughts too.

"Oh, it's convection."

"Hey, I wonder what the best in-between angle is, balancing both temperature and cosine loss."

[myself248]

So, a lot of the recent attention is on bifacial east-west arrays because they produce a complementary duck curve throughout the day. In that case, pure vertical makes sense, and production just takes a dip at local solar noon and recovers soon after.

But for traditional south-facing panels, I'd argue that straight vertical is still optimal, at least for higher latitudes. Vertical panels are extremely good at shedding snow. They produce more in winter when you need every watt you can get, and less in summer when the sun is high in the sky and you don't need all that extra power anyway. As soon as you tilt the panels to minimize cosine loss, you open yourself up to snow buildup which can dwarf any cosine gains.

[schiffern]

I'm also a fan of vertical panels in snowy regions, but at the same time it kinda dodges the question.

In a non-snow region (which is most of the country, a fraction which is only increasing decade by decade), what is the optimal tilt angle? 70°? 80°? More? This is quite important for large solar panel farms in the hot and sunny South/Southwest.

The cosine loss is easy to model, so to answer the question what we need is a curve that describes how the convection cooling effect varies with angle.

[beAbU]

Did you factor in possibly improving the incident angle of the sunlight? Solar benefits massively the more perpendicular the incoming sunlight is.

[madaxe_again]

The incident angle was previously optimised for insolation over the course of a year - so halfway between winter solar apogee and summer solar apogee.

[vosper]

I'm hoping to get solar on my next house, which we're designing now. Is there a way to install roof panels that improves convection / reduces temperatures, like using a different mounting system, if that's a thing?

[beAbU]

Buy 2x extra panels. Honeslty the gains you might get by optimising panel placement (beyond matching to your hemisphere and latitude) will be outweighed by the additional cost. Domestic applications is not big enough to really give you significant gains in this department.

Make sure they face the right way (south for Northern hemisphere), and match the angle with your latitude. If you have a pitched roof thats +-10 deg in the correct angle, just lay them flat on the roof.

Edit: forgot to add, a while back there was an article here about a company that proved it was viable to lay the panels flat on the ground for massive solar farm installations. The savings from less installation labour and materials went to installing more panels. And they still came out ahead. Solar is getting cheap enough that the math gets weird. Your answer is almost always "just add more panels" unless you are seriously space constrained.

[vosper]

Thanks, that makes sense.

I don't know anything about rooftop solar mounting systems, I will show my ignorance here: I was imagining that perhaps you could attach the panels to rails running vertically rather than horizontally, to allow for convection airflow below the panels.

Or if the rails are horizontal perhaps they have holes in them to allow some airflow.

I'm sure this has been thought of and doesn't work some obvious reason I just don't know about.

Edit: As I expected, this has been thought of: https://solarstone.com/blog/natural-ventilation-and-effect-o...

[beAbU]

Your installer will know best :)

Make sure you design a roof with the rigt heading, and slope. The rest will be done by the installer.

My rails run horizontally, but thats only because it required less rail and mounting hardware compared to vertically. The panels are mounted portrait, in a 5x2 square. Each panel has 2x rails under it, so 4 rails horizonfally on my roof. If you sketch it out then it'll make sense.

Had I gone with vertical rails, each column would've required 2 rails, so 10x rails vertically mounted on my roof.

[sbierwagen]

Actively cooled panels exist: https://hydrosolar.ca/products/ahtech-72sk-hybrid-photovolta... The extra cost makes them not worth it, but if you wanted the absolute maximum amount of power from a limited amount of space, you could do it.

[myself248]

I wonder how it would pencil out if you just pumped groundwater through them and returned it to an adjacent well, without even trying to harvest the heat. It cools the panels in summer, and warms them in winter when snow coverage might otherwise be an issue. Melting the snow off can turn an otherwise-nearly-zero period into a productive one.

But this is way simpler than domestic water heating. You don't care about thermostats and storage tanks, you don't worry about overheating, you don't even necessarily have to care about leaks since it's all going the same place anyway. There's only one pump and no valves. The panels themselves could be made to a lower standard since the whole thing could run at nearly-zero pressure.

[rapjr9]

In colder climates maybe it makes sense to use mirrors to increase output since the panels are already cooled by the environment?

[sbierwagen]

The term of art there is "concentrated PV solar". Big fresnel lenses concentrating light onto a postage stamp of photovoltaic material. Requires sun tracking to keep the light on the PV, so it's useless for rooftop solar.

If you look it up you'll notice all the citations are decades old. Designed for a world where PV was terribly expensive, so you'd trade a lot of mechanical and packaging complexity to use as little of it as possible.

Then silicon got 100x cheaper. Now every solar install uses fixed-angle racks or just plopping the panels flat on the ground. You lose some efficiency, but land is cheap, so who cares?