Thermal Drone Inspection of a Solar Farm: From Flight to Report

A thermal drone covers a solar farm far faster than a team on foot, but the images are only as useful as the conditions they were taken in and the notes that come with them. This guide, for pilots and O&M teams who fly their own thermal drone, follows a PV survey from planning to the finished report.

1. Planning: Choose the Window, Then the Day

A module carrying current turns part of the sunlight into heat, and a fault changes where that heat builds up; without enough sunlight and current, the patterns fade. IEC TS 62446-3:2017, a technical specification rather than a law, describes the conditions (details in our guide to IEC TS 62446-3):

Condition	What the specification describes	Practical note
Irradiance	At least 600 W/m² in the plane of the modules	Measure in the module plane; a horizontal weather-station value is a different quantity.
Stability	Stable irradiance, modules in thermal steady state	After passing clouds, wait until the modules have settled.
Wind	An upper limit, commonly cited as 4 Beaufort (about 28 km/h or 17 mph); wind is recorded	Wind cools the modules and flattens differences. The drone's own wind limits apply too.
Cloud cover	Commonly cited as at most 2 okta (2/8 of the sky)	Clouds change the irradiance and appear as reflections in the glass.
Operating state	Plant in normal operation; operating state recorded	Check inverters and curtailment; ask the O&M team which strings are off.

Time of Day and Season

Irradiance in the module plane depends on the sun's elevation, the tilt and orientation of the modules, and haze. On a clear day it peaks around solar noon for fixed-tilt arrays facing the equator; single-axis trackers keep it high for longer. In winter or at high latitudes, 600 W/m² may be reached only briefly or not at all. Let morning dew dry before you fly.

Use a Flight Planner

Our free flight planner shows, hour by hour for the next seven days, the forecast irradiance on your module plane (from the tilt and orientation you enter), wind, cloud cover and rain. It marks the windows with at least 600 W/m², wind up to 28 km/h, at most 2/8 cloud and no rain. A forecast is not a measurement: use it to choose the day, then measure on site. Whether a survey is due at all is covered in when solar thermography makes sense.

Airspace, Permissions and Site Access

A survey for a client or employer is a work flight, and local rules apply. In the United States, flying under Part 107 requires an FAA Remote Pilot Certificate; in the UK, see the CAA on flying drones for work; in the EU, the EASA open and specific categories apply. Check the current rules yourself (this is not legal advice), and agree site access and safety rules with the owner and the O&M team.

2. Equipment and File Format

Radiometric JPEG (R-JPEG) vs Plain JPEG and Video

A radiometric JPEG looks like a normal thermal picture but also stores radiometric data for every pixel, together with the capture settings. DJI's enterprise thermal cameras, such as the Mavic 3T, Zenmuse H20T and H30T, save stills as R-JPEG. In the manufacturer's software, such as DJI Thermal Analysis Tool, you can then read temperatures and adjust emissivity, reflected temperature, distance and humidity after the flight, in recent versions also for many images at once (Heliguy).

A plain JPEG or a controller screenshot stores only colours: you can mark findings, but nobody can measure them later. Unless your camera's documentation says that its video is radiometric, treat video the same way. Keep the visual photos too; they help to locate findings and to tell soiling or shading from defects.

Thermal Drone Cameras at a Glance

Compact drones have a built-in thermal camera; larger airframes carry an interchangeable payload. Typical DJI examples, with DJI's published thermal figures:

Camera	Thermal detector	Thermal lens
Mavic 3T (compact)	640 × 512, 12 µm pixels	61° diagonal, 40 mm equivalent
Matrice 30T	640 × 512, 12 µm; 1280 × 1024 in super-resolution mode	9.1 mm (40 mm equivalent), 61°
Matrice 4T	640 × 512, 12 µm; 1280 × 1024 with Super Resolution	45° diagonal, 53 mm equivalent
Zenmuse H20T (payload)	640 × 512, 12 µm	13.5 mm (58 mm equivalent), 40.6°
Zenmuse H30T (payload)	1280 × 1024	24 mm (52 mm equivalent), 45.2°

Upscaled modes give larger images but may not record temperatures: DJI's FAQ states that the H30T does not support temperature measurement in High-Res mode. Calculate the resolution with the native detector pixels unless the manufacturer says otherwise.

Resolution vs Flying Height

IEC TS 62446-3 asks for a geometric resolution of about 3 cm of module edge per pixel or finer. The flying height follows from the pixel pitch of the detector and the focal length of the lens:

Calculated from DJI's published values, for a camera looking squarely at the modules: the Matrice 30T (12 µm, 9.1 mm) gives 2.6 cm per pixel at 20 m and reaches 3 cm at about 23 m (75 ft) above the modules; the Zenmuse H20T (12 µm, 13.5 mm) at about 34 m (110 ft). DJI lists the same pixel pitch, field of view and equivalent focal length for the Mavic 3T as for the Matrice 30T. Oblique views, slopes and height errors make the real resolution coarser, so keep a margin.

Not every survey is flown to that limit: a DJI case study of a 181 MWdc plant flew at 40 m for about 5.5 cm per pixel. That is coarser than the 3 cm the specification describes, and small cell-level patterns are harder to see. Decide what you need and state it in the report; the trade-off is the core of drone vs. handheld thermography.

Camera Settings

• Emissivity: about 0.85 for module glass viewed close to perpendicular, lower at oblique angles (Buerhop et al.).
• Reflected temperature: the glass mirrors the sky, which on a clear day is radiatively much colder than the air. Determine it on site, for example with the reflector method of ASTM E1862, or at least note the value used.
• Gain mode: high gain measures more precisely over a narrower range; DJI lists −20 to 150 °C in high gain for the Mavic 3T and the H30T.
• Temperature span: automatic contrast rescales every frame. A fixed, narrow span keeps images comparable and shows small differences.

3. Flight Parameters and Mission Planning

Plan an Automated Mission

Fly planned missions, not by hand: an area (grid) mission per block with lines along the rows, or waypoints built from the layout. DJI's PV inspection workflow plans routes in DJI Pilot 2 or FlightHub 2. Split large sites into blocks you can finish under the same conditions, ideally following the inverter blocks. Set the height relative to the modules, not the take-off point; on slopes, use terrain following or blocks of similar elevation.

Overlap

• Image-by-image review: every module must appear completely in at least one sharp image, with a margin for drift. The DJI case study used 70% front and 20% side overlap. Seeing each module twice also helps to rule out reflections.
• Thermal orthomosaic: stitching thermal images needs far more overlap; Pix4D asks for at least 90% front and side overlap for thermal projects.

Speed and Photo Interval

The uncooled microbolometers in drone thermal cameras respond slowly, so speed smears the image; a 2025 study in Solar Energy names motion blur as a key factor limiting data quality and survey speed on PV plants (Solend, Rødningsby & Moen). The lower you fly, the worse it gets at the same speed. Fly a test line, check that module frames and cell gaps are sharp, and slow down or hover at waypoints if not.

Example: at 3 cm per pixel, the 512-pixel side of a 640 × 512 image covers about 15 m. With that side along the flight line and 70% front overlap, you need a photo every 4.6 m, so at one photo every 2 seconds the drone may fly at most about 2.3 m/s.

Camera Angle, Glare and Reflections

The specification asks for views as close to perpendicular to the modules as possible. A camera pointing straight down sees fixed-tilt modules at an angle equal to their tilt; pitching the gimbal to match the tilt, as in the DJI case study, brings the view closer to square-on. Staying a few degrees off exact perpendicular keeps the drone's own reflection out.

Glass acts as a mirror in the infrared: the sun shows up as a bright, often round spot, and clouds or warm structures can appear too. A reflection moves with the camera position, while a defect stays on its cell. Check suspicious spots in the overlapping image or from a second heading.

4. The On-Site Log

The images show what the modules looked like; the log explains why. Fill it in for every flight or battery, and keep the clocks of the drone, the irradiance logger and your phone in sync so that images can be matched to readings.

Record	Details
Site and flight	Block, flight number, date, start and end time with time zone; pilot, and who will assess the images (with qualification)
Irradiance	Value in the module plane and its source: reference cell or pyranometer at module tilt, the plant's sensor, or a forecast
Ambient temperature	Shaded sensor near the array
Wind	Speed, direction and source: anemometer, Beaufort estimate, or forecast (usually for 10 m height)
Cloud cover	In okta (eighths of the sky), and changes during the flight
Operating state	Inverters running, curtailment, strings known to be off
Module surface	Soiling, recent rain or cleaning, dew
Camera	Model, serial number, firmware, file format, gain mode, emissivity, reflected temperature, temperature span
Flight	Height above the modules, cm per pixel, gimbal pitch, speed, overlap
Exceptions	New shading, areas not flown and why

The specification also sets qualification requirements for the people who carry out and assess the inspection (Annex B). If the report will support a warranty claim, an insurance case or a sale, have the images assessed or the report reviewed by a qualified person.

5. From Images to a Report

Screen the Images

Remove take-off, landing and transit shots, then review every image of the inspected area. Mark images that cannot be assessed (blur, reflections, wrong angle, low sun, shadow) instead of deleting them; they belong under limitations.

Mark, Locate and Describe

Mark each anomaly, compare it with neighbouring modules, and give it a location a technician can find: block, row, string, module. The GPS position in the image data is where the drone was, not where the module is, so translate it with the layout plan or the visual photo. Describe the pattern: one hot cell, several cells, a warm sub-string, a whole module or string, a hot connector. Our article on solar panel hotspots explains typical patterns and causes.

Measure Where It Matters

If you report temperatures, read them from the R-JPEG in the manufacturer's software with emissivity and reflected temperature set, and give the difference (ΔT) to a healthy area on a comparable module. ΔT grows with irradiance and current, so always report it with the irradiance at capture.

Classify and Prioritise

IEC TS 62446-3 has three classes of abnormality: CoA 1 (no abnormality), CoA 2 (thermal abnormality) and CoA 3 (safety-relevant thermal abnormality). Findings are classed by comparing their pattern with the specification's examples, not by fixed ΔT thresholds, and it sets no repair deadlines. An own priority scale (high, medium, low) is fine for ordering the work if the report labels it as your own. Safety-relevant findings come first; a qualified person decides whether anything must be disconnected.

Keep Observation and Cause Apart

A thermal image shows where something is different, not why. Write what you saw (“one cell clearly warmer than its neighbours”), name the likely cause and say how far it is confirmed: not confirmed, plausible after a visual check, or confirmed by a documented test. A warm sub-string is consistent with an active bypass diode, but the reason behind it needs a check on site.

State the Limitations

Name areas not flown, images that could not be assessed, borderline conditions, and whether temperatures were measured or the images were reviewed qualitatively.

Assemble the Report

A useful report has a summary, the installation and inspection conditions, a list of findings, one page per finding (marked thermal image, close-up, location, observation, cause status, recommended action), next steps, limitations and a release. The solar thermography report template shows this structure with a sample PDF, and what an IEC 62446-3 report contains goes through it section by section.

Best Drone for Solar Inspection: What Matters

There is no single best drone, only the one that delivers radiometric images at the resolution you need from a height you can fly legally and efficiently. Check:

• Radiometric stills (R-JPEG or equivalent) that your analysis software can read.
• NETD of 0.1 K or better at 30 °C, as the specification asks. DJI quotes ≤ 50 mK at f/1.0 for the Mavic 3T, Matrice 30T and H30T; compare the test conditions.
• Long-wave (8–14 µm) or mid-wave (2–5 µm) camera; the specification accepts both.
• A detector and lens that reach about 3 cm per pixel from a practical height. More native pixels over the same field of view, or a longer lens, let you fly higher.
• Adjustable emissivity and reflected temperature.
• Automated area and waypoint missions, and terrain following for sloping sites.
• Permission to fly it where you work.

Compact drones such as the Mavic 3T or Matrice 4T are quick to set up; payloads like the H20T or H30T on larger airframes suit big sites where a longer lens or a 1280 × 1024 detector pays off (see utility-scale solar thermography). In the United States, the FCC added foreign-produced drones and drone components to its Covered List in December 2025 (Public Notice DA 25-1086), which restricts the authorization of new models; check what is available before you buy.

Common Mistakes

1. Horizontal or forecast irradiance reported as irradiance in the module plane.
2. Flying too high for the resolution you claim, or calculating it from an upscaled image.
3. Video or controller screenshots instead of radiometric stills.
4. Default emissivity and reflected temperature, with absolute temperatures reported anyway.
5. Reflections taken for hotspots, without checking the overlapping image.
6. Unknown operating state: a switched-off string or a curtailed inverter changes the pattern.
7. No log, or clocks out of sync, so images cannot be matched to the conditions.
8. A cause stated as fact, your own priority scale presented as IEC classes, or a report called “IEC compliant”.

FAQ

When is the best time for a thermal drone inspection of solar panels?

When the irradiance in the module plane is at least 600 W/m² and stable, with little wind and cloud and the plant in normal operation. On clear days that is usually around solar noon; the exact window depends on season, latitude and module orientation.

What is a radiometric JPEG (R-JPEG)?

A thermal JPEG that also stores radiometric data for every pixel, so temperatures can be read, and emissivity or reflected temperature adjusted, after the flight in the manufacturer's software. A plain JPEG or a screenshot keeps only the colours.

How high should I fly for a solar thermal inspection?

For about 3 cm of module edge per pixel, the resolution IEC TS 62446-3 asks for, the maximum height is 0.03 m × focal length ÷ pixel pitch. For a camera with 12 µm pixels and a 9.1 mm lens, such as the DJI Matrice 30T, that is about 23 m (75 ft) above the modules.

What is the best drone for solar inspection?

One that records radiometric stills your software can read, has an NETD of 0.1 K or better at 30 °C, reaches the resolution you need from a practical height, flies automated missions, and that you may fly where you work.

Does following this guide make my inspection IEC compliant?

No guide or software can do that on its own. Whether an inspection follows IEC TS 62446-3 depends on the equipment, the conditions, the procedure and the qualification of the people involved. A good report records the conditions the specification asks you to record, so readers can judge the result.

Sources

• DJI specifications: Mavic 3 Enterprise, Matrice 30, Matrice 4, Zenmuse H20, Zenmuse H30, H30 FAQ; DJI case study, 181 MWdc solar farm.
• Pix4D, image acquisition; Solend, Rødningsby & Moen (2025), Solar Energy; Buerhop et al., IR thermography of PV plants and IEC 62446-3; ASTM E1862.
• Rules: FAA, UK CAA, EASA, FCC DA 25-1086. IEC TS 62446-3:2017 values as summarised in our IEC TS 62446-3 guide.