A UAV thermal payload inspects a utility-scale solar site in hours instead of weeks, detecting panel hotspots, bypass-diode faults and string outages as distinct thermal patterns — each mapping to a specific maintenance action. The skill that separates a useful survey from a folder of warm pictures is reading the pattern, not finding the heat.
- 3–5 cm/pxTarget GSD, IEC 62446-3
- >600 W/m²Minimum irradiance
- 640×512Recommended thermal array
- <5 m/sPractical wind ceiling
Key takeaways
- Thermal finds a pattern, not a cause. The pattern determines the work order; a paired visible image confirms the cause.
- Below 600 W/m² irradiance the temperature differences you are measuring do not exist. A survey flown under cloud produces a complete-looking report that found nothing.
- GSD is the hard constraint: 3–5 cm/px is what cell-level judgement under IEC 62446-3 requires.
- The value of a survey is comparability — fixed altitude, fixed angle, fixed time window, or next year’s data cannot be compared with this year’s.
On this page
Thermal signatures and what they mean
Every thermal pattern on a PV site corresponds to a specific failure mechanism and a specific action. Misread the pattern and you dispatch the wrong crew.
The table below is the field key we give inspection teams. Temperature rises are typical values under clear sky at full irradiance; module type and mounting shift them, so always calibrate against your own site baseline.
| Thermal pattern | Typical rise | Failure mechanism | Action |
|---|---|---|---|
| Single hot cell | 10–30 K | Cell crack, local shading or soiling | Clean and re-measure; if still hot, flag for replacement |
| Patchwork across module | 5–15 K | Potential-induced degradation (PID) | Check system grounding and inverter configuration |
| One third of module uniformly hot | 15–40 K | Bypass diode failure or substring open | Replace junction box |
| Whole string uniformly warm | 2–6 K | String offline / fuse open | Check combiner box and fuses |
| Hot band along module edge | 8–20 K | Frame grounding fault or moisture ingress | Inspect sealing and grounding |
| Localised hot spot at junction box | 20–60 K | Oxidised or poorly crimped connection | Act immediately — fire risk |
| Whole row running cool | — | Tracker not actuating in that zone | Inspect tracking system, not the modules |
The most common misread: soiling and a cracked cell produce almost identical thermal signatures. The way to separate them is to clean and re-measure — which is why surveys flown before washing routinely produce unusable conclusions.
Flight and capture parameters
PV thermography is a measurement task, not an aerial photography task. Get the parameters wrong and the flight itself has no value.
- Confirm irradiance first. Below 600 W/m², reschedule. Low irradiance suppresses both healthy and faulty module temperatures, and the difference you came to find disappears with it.
- Choose the window. 10:00–14:00 solar time. Earlier or later and modules have not reached thermal equilibrium, while low sun angles introduce specular reflection.
- Compute altitude from the GSD target. Aim for 3–5 cm/px. GSD = pixel pitch × distance ÷ focal length. Run the arithmetic; do not reuse an altitude from another payload.
- Shoot near-nadir with a slight offset. Perfectly vertical puts the sun’s reflection into the lens; 5–15° off avoids it while keeping effective coverage.
- Hold altitude and angle across the whole site. Comparability matters more than any single frame — it is what makes year-on-year trending possible.
- Capture visible simultaneously. Thermal locates the anomaly, visible confirms the cause, one flight delivers both.
Wind is routinely underestimated. Above about 5 m/s, convective cooling flattens the temperature differences you are trying to detect and weak anomalies are missed. This is signal loss at the physical level, not an image-stability problem.
From anomaly to work order
The deliverable is not a thermal image; it is an actionable work order. A report the O&M team can accept needs three things: module-level location, severity graded against a stated criterion, and a recommended action with a timeframe.
Location is the step most often skipped. “Hot spot in the north-east corner of array 3” is useless to a crew; a coordinate or a module ID can be dispatched directly. Payloads with laser ranging stamp each anomaly with a coordinate — though it is worth understanding the error budget behind that coordinate.
Grade against a published criterion and name which one in the report. The full report structure is in what a thermal inspection report should contain.
Cost and what the survey actually saves
Handheld ground thermography on a 50 MW site typically takes two people several weeks. Drone survey of the same area completes in one to two days — an order-of-magnitude difference.
But the cost is not in the flying. The real expense is analysis: a single survey of a 50 MW site produces tens of thousands of thermal frames, and reviewing them costs far more than the flight. That is why “does it export radiometric data and feed automated analysis” affects total cost more than any resolution figure — unpacked in the five-year cost of ownership.
On the return side, lost generation is the easiest number to defend. An offline string may have produced nothing for months before anyone noticed; at typical capacity and tariff, finding two or three offline strings in one survey usually covers the entire annual inspection budget.
Where this goes wrong
PV thermography fails in a small number of highly predictable ways.
- Flying under cloud. The report looks complete and found nothing — worse than not flying, because it creates false confidence.
- Insufficient GSD. Cell-level anomalies need 3–5 cm/px. Fly higher and you only see string-level faults.
- Emissivity and reflection left unhandled. Glass reflects sky and sun, producing false hot spots. See thermal reflections and false hot spots.
- Thermal without a visible pair. Soiling cannot be separated from a cracked cell, so every work order reads “confirm on site”.
- Parameters changing between annual surveys. Different altitude, angle or time window destroys trending — every year starts from zero.
Recommended UAVThermal payloads
MV-2P — 640×512 radiometric thermal at 130 g for compact mapping platforms. The default choice for large-area sweeps.
MV-4X — thermal plus starlight and NIR, so one payload covers daytime inspection and night-time site security.
OP-90A — zoom plus thermal plus laser ranging, for targeted re-inspection of flagged strings and coordinate tagging.
Related reading
Technology: radiometric temperature measurement and how uncooled VOx microbolometers work.
Field practice: PV inspection patterns and workflow, emissivity in drone thermography and isotherms and temperature alarms.
Adjacent missions: substation inspection and power line inspection.
FAQ
When is the best time to fly a PV thermal inspection?
Clear sky, irradiance above roughly 600 W/m², typically 10:00–14:00 solar time. Low irradiance suppresses the very temperature differences you are measuring, so a survey flown under cloud produces a complete-looking report that found nothing.
Can thermal alone identify the fault cause?
No. Thermal finds and classifies the anomaly pattern; a paired visible image is what distinguishes soiling from a cracked cell. Dual-sensor payloads solve this in one pass, which is why single-sensor thermal surveys tend to produce work orders that say confirm on site.
What ground sample distance do I need?
3–5 cm/px to resolve cell-level anomalies and satisfy IEC 62446-3 judgement requirements. Compute the altitude from pixel pitch, focal length and distance rather than reusing a figure from another site, because the same altitude gives different GSD on different payloads.
Does wind affect a thermal PV survey?
Yes, more than most operators expect. Above about 5 m/s, convective cooling flattens the temperature differences you are trying to detect and weak anomalies are missed. This is signal loss at the physical level, not an image-stability problem.