Why heat costs production, and by how much
Solar modules are rated at 25 degrees Celsius cell temperature. Real cells on a Florida roof run far hotter than the air, because the module sits in direct sun with limited airflow behind it. Every degree above 25C costs a fraction of the module's rated power.
The fraction is the temperature coefficient of Pmax, published on every module datasheet. Across current modules the range runs from about -0.24 percent per degree on premium heterojunction modules to about -0.36 percent on older poly and PERC products. Mainstream N-type TOPCon modules, which dominate 2025 and 2026 installations, sit around -0.29 to -0.30 percent per degree.
Formula used below: loss percent = (cell temperature - 25) x |coefficient|. The coefficient is from the module datasheet.
| Cell temperature | Coefficient | Instantaneous loss |
|---|---|---|
| 45C, warm spring day | -0.29 %/C | About 5.8 percent |
| 55C, typical summer roof | -0.29 %/C | About 8.7 percent |
| 65C, hot afternoon | -0.29 %/C | About 11.6 percent |
| 70C, very hot roof | -0.29 %/C | About 13 percent |
| 70C, older PERC module | -0.36 %/C | About 16 percent |
What a normal heat derate looks like
A useful sanity check is the performance ratio: actual production divided by what the system should produce at the measured irradiance and temperature. A system running near its expected ratio in hot weather is derating normally. A system well below it, especially in cool conditions, is not.
- The whole array loses output together, not one panel or one string.
- The drop tracks the weather: worst in the early afternoon, better on cloudy days and in the evening.
- The shape of the daily curve stays normal, just lower at the peak.
- The same month across two years looks similar when the weather is similar.
- Production recovers in cool, clear winter conditions without any repair.
The signs that point at a fault instead
Two conditions deserve special attention because they look like heat derate but are not. The first is one underperforming module while its neighbours are normal, which points at a device, a connection, or a shading change. The second is a system that keeps underperforming in cool months, which is a fault that has nothing to do with temperature.
Heat derate is array-wide and reversible. Faults are not.
| Pattern | Normal heat derate | Worth a diagnostic |
|---|---|---|
| Scope | Whole array moves together | One panel, one string, or one microinverter |
| Timing | Worst at peak heat, recovers after | Persists in cool weather and in the evening |
| Curve shape | Normal shape, lower peak | Flat segments, dropouts, or a hard cutoff |
| Comparison | Matches neighbours with similar systems | Clearly below a neighbour's identical system |
| History | Similar year over year | Progressive decline across several months |
| Warnings | None | Device alerts, ground fault, arc fault, or inverter codes |
What a homeowner can check, safely
- Compare the same month across two years, not two consecutive months.
- Note whether the drop is array-wide or device-specific in the monitoring app.
- Look for any alert or code in the app or on the inverter display.
- Check for new shading: a tree that grew, a new structure, or debris.
- Photograph the inverter display during peak afternoon heat.
When to book a diagnostic
Book a diagnostic if the loss is not array-wide and smooth, if it persists in cool weather, if any device reports an alert, or if the same month across two years is clearly worse. Otherwise, treat a modest peak-hour drop in July and August as the cost of a Florida roof and compare again in cooler months.
If the technology is older, ask the service provider to check the module datasheet coefficient against what the system is actually doing. A 2015 PERC array and a 2026 TOPCon array should not be judged against the same expectation.