Clouds, Rain, and Solar Production in Florida

The question usually arrives in July, at 3 in the afternoon, when the app falls off a cliff and the rain hits the pool cage. The system is probably fine. Central Florida’s wet season is built into the way solar works here. This page is what that drop is, what it is not, and which number is worth worrying about.

What the panel is actually using

A solar panel is rated at a laboratory condition called Standard Test Conditions. One piece of that test is a brightness of 1,000 watts of sunlight per square meter, which is roughly a clear noon. Your roof does not live there. Brightness changes all day, and clouds change it faster than the clock does.

Two kinds of light reach the roof. Direct light is the beam, the hard shadow on the driveway. Diffuse light is what the clouds and the sky scatter, the brightness that is still there when you cannot see a shadow. Panels use both. They prefer the beam. They do not go to zero when the beam disappears.

How far they fall depends on the cloud, not on a single “cloudy day” number.

• Thin high cloud or a bright haze often leaves half to three quarters of the clear-sky power. The app looks soft, not broken.
• A typical building cumulus, the puffy afternoon clouds, takes a bite out of the beam and then gives it back when the gap passes. Over a whole day the loss is real and it is not a wipeout.
• A thick thunderstorm, the dark anvil that sits on the house, can cut the power to something like a tenth to a third of the clear-sky rate for as long as it sits there. Half an hour is common. A slow storm can be two hours. Output below about 200 watts per square meter, against that 1,000-watt clear noon, is the kind of sky this is.
• A stalled front or the outer bands of a tropical system are the ugly version. The cloud does not leave. A day like that can produce a small fraction of a clear day, and two or three of them in a row will show up in the monthly total.

Partly cloudy days have a trick that scares people who are watching the app too closely. When the sun is out and a bright cloud next to it reflects extra light onto the panels, power can jump above the clear-sky number for a few minutes. The inverter then limits the output at its rated size. That spike is sunlight, not a fault. Clipping on those minutes is covered on the inverter page.

Rain is mostly the cloud

Raindrops in the air block a little light. The dark cloud that produced them blocks far more. If you stand outside in a sun shower, you already know the difference. The production chart follows the cloud, not the sound on the roof.

After the cell passes, two small things happen. The panels are wet, which can change the reading for a few minutes, and the glass is cleaner than it was at noon. Central Florida pollen, oak dust, and a dry spell leave a film. A normal rain rinses a lot of it. In the standard production models, dirt is a loss of around a couple of percent over a year, folded into a larger bucket of losses that totals about 14 percent before the weather is even counted. Rain is not a maintenance plan, and it is not the reason a stormy afternoon looks dead. If the glass is coated after a long dry stretch and production on clear days is sagging, that is the maintenance page, not the weather.

Hail and wind are a different subject. A hard storm can crack glass or lift a loose mount. Production going soft for an hour is weather. Production staying soft the next bright morning is damage, or a tripped breaker. The hurricane page is the storm you prepare for. This page is the ordinary Tuesday thunderstorm.

The year you were quoted already has the clouds in it

Installers do not estimate a Central Florida roof by assuming 365 clear noons. The usual tool is PVWatts, from the National Renewable Energy Laboratory, now part of the National Laboratory of the Rockies. It uses a typical weather year built from satellite measurements of sunlight, cloud, and temperature. The clouds are in the file.

For a south-facing array around Orlando, tilted modestly, that model commonly lands near 1,500 kilowatt-hours a year for each kilowatt of panels. Published runs for the same city sit roughly in a band from about 1,490 to about 1,550, depending on tilt, direction, and the exact weather file. A 7-kilowatt array in that band is on the order of 10,500 to 11,000 kilowatt-hours a year. The capacity factor, which is actual output divided by what the panels would make if they ran at full rated power every hour of the year, is about 17 to 18 percent. That sounds low if you think the rating is a promise. It is what a real roof does, nights included, winters included, and July thunderstorms included.

Those figures are models of a typical year, not a meter on your house. Shade, a west roof, a steep tile hip, and a dirty array move your number. A quote that shows a much higher year, with no shade study and a perfect south roof you do not have, is the number to challenge. A quote in this band is already telling you the truth about clouds.

Summer is stormy. It is not the worst season.

Florida’s wet season runs from about late May into October. Inland, the pattern is famous. The morning is bright. The sea breeze builds. By midafternoon a thunderstorm dumps on one neighborhood and misses the next. The day’s solar production is often a strong morning plus a dented afternoon, not a lost day.

A published location model using NASA sunlight data for the Orlando area puts the seasons in this order, as average daily output per kilowatt of panels: spring about 6.7 kilowatt-hours, summer about 5.9, fall about 4.8, winter about 4.1. Spring wins. Summer, storms and all, still beats fall and winter. Winter is the low season because the sun is lower and the days are shorter, even on the blue days that follow a cold front.

That is the opposite of the story people import from the Northeast, where summer is the obvious peak and winter is snow. Here the painful month on the app is often a dark week in July, and the painful month on the annual chart is still more likely to be January. One awful Tuesday does not set the season. The season is already in the 5.9, not in the hour the anvil was overhead.

Heat works the other way, and it is easy to mix up with clouds. Panels lose output as they get hot, on the order of a few tenths of a percent for each degree above their test temperature. A clear August afternoon can be 95 degrees in the yard and far hotter on the glass. A cloudy afternoon is dimmer and the glass is cooler, so each bit of light is used a little better. You still get less energy, because the missing light is a much bigger cut than the heat savings. The heat page is that other penalty. Do not blame July’s clouds for what is actually a hot, clear roof, or the reverse.

What a normal bad day looks like

Take a 7-kilowatt array that peaks around 5 to 6 kilowatts on a clear spring noon, after the inverter and the heat have taken their share. A dark storm overhead can show something like 0.5 to 2 kilowatts until it moves. The energy for that hour might be a couple of kilowatt-hours instead of five. The rest of the day, if the morning was clear, can still be a decent day. Ten days like that in July change the month. They do not erase it.

A week of thick overcast from a tropical system is the case that does show up in the year. Those weeks happen. A typical year already contains some of them. A year with an extra one is a bit lower. It is not evidence that the panels failed, and it is not a reason to add panels in November so that next July feels better. Extra panels in this climate run into the utility’s sizing rules and into cheap year-end export, which the Duke, FPL, and OUC page explains. You do not buy a bigger array to insure a thunderstorm.

A battery does not manufacture the missing afternoon either. It can only store what the roof made earlier in the day. If the whole day was dark, the battery starts the evening less full. That can still be useful. It is not a cloud refund. The battery page is the decision. Weather is a bad reason to make it.

When the low number is not the weather

Use the sky, not the app alone.

• If the neighbors are in the same storm and your production is a small fraction of a clear day, that is the cloud.
• If the sky is bright, the ground has hard shadows, and the array is near zero, that is not weather. Check the breaker, then the inverter, then the monitoring. The replacement page starts with those checks.
• If one section of a string system dies whenever a cloud shades part of the roof, and the rest of the roof is fine, that can be design, not the storm. A single shaded panel can pull a whole string down. Microinverters and optimizers limit that to the shaded panel. Clouds make the weakness obvious. They do not create it.
• If clear days this year are well below clear days last year, look at shade that grew, a failed panel, or dirt, not at “it has been cloudier.” Compare the same kind of day.
• Judge the system on a season or a year against the estimate, not on a week. PVWatts is a typical year. Your July is one July.

What to do with this when you are buying

Ask the installer for the production estimate in kilowatt-hours per year, the direction and tilt they used, and whether shade is in the number. A Central Florida south roof near 1,500 kilowatt-hours per kilowatt is a sane starting point. A west roof will be lower. A shaded roof should be lower still, and if the quote did not drop, the shade was ignored.

Do not pay extra for a “cloudy-day panel” pitch that is really just a more expensive panel with a similar response to diffuse light. Ordinary silicon panels already use diffuse light. The expensive fixes for shade are the electronics on each panel, and they matter when something solid blocks the sun, a vent or a tree, more than they matter when the whole sky is gray.

Do not expect the first July to match a clear week in April. If the contract’s savings assume the high month all year, the contract is wrong. The worth-it page is where a typical year belongs, storms included, federal credit not included.

Watch the annual total. If a year lands within a modest band of the estimate, the clouds did what clouds do. If it lands far below, on a roof that was not newly shaded, then it is time to look at hardware. The weather is allowed to have a bad month. It is not required to explain a bad year.

This page is general information, not a production guarantee for your roof. A typical-year model is not a promise that next July will match the last one.