MIMAH Logo

Solar Thermography: IEC 62446-3 Inspections Explained

Author

Hisham Abdalla

Date Published

Illustration of a solar array in rows with one warm gold panel standing out, a small figure holding a thermal camera and a drone hovering above

Disclaimer: Research and analysis by the engineering team. Sources referenced below.

Solar thermography is an infrared survey of a PV plant while it is generating. A thermal camera finds cells, modules, strings and connections running hotter than their neighbours, and the pattern points to the fault. The reference document is IEC TS 62446-3, which asks for at least 600 W/m² of sunlight in the plane of the modules, wind no stronger than 4 on the Beaufort scale (28 km/h), no more than 2 okta of cumulus cloud (an okta is an eighth of the sky), and a camera and distance that give no more than 3 cm of module edge per pixel.

A survey flown under patchy cloud, or on a gusty afternoon, produces patterns that look like faults and can hide the real ones.

What IEC 62446-3 Covers

IEC 62446-3 is the third part of the IEC 62446 series on testing, documentation and maintenance of PV systems. Its full title ends in Outdoor infrared thermography, and the current text is edition 1.0 from June 2017. It is a Technical Specification rather than a full International Standard, a form IEC uses when a subject is still developing. IEC's page lists a stability date of 2026, and the Serbian standards institute's catalogue shows a second edition as a project in development. Clause numbers here are from the 2017 text, so check which edition your tender names.

The scope is passive imaging of plants in operation, under natural sunlight, and it reaches past the modules to cables, contacts, fuses, switches, inverters and batteries.

Part 1 of the series sets out the commissioning pack, and its Category 2 tests include infrared inspection for larger or more complex plants. Part 3 says how that inspection is done. Our guide to solar PV commissioning covers Part 1.

What Conditions Does IEC 62446-3 Require?

The conditions sit in a table in clause 5.3. The fullest public view of the text we could find is the free preview on iTeh's catalogue page, a standards reseller's listing that runs from clause 1 to part of clause 7. Five conditions matter.

Irradiance. At least 600 W/m² in the plane of the modules when the modules are the target. For cables, connectors and other electrical parts the test is current instead: at least 30 percent of the rated current in the path being inspected, typically reached above 300 W/m², with more than 600 W/m² still recommended. The reading comes from a sensor calibrated to within 5 percent.

Wind. No more than 4 on the Beaufort scale, or 28 km/h. Wind cools modules unevenly, and a gust moving across an array paints a pattern of its own.

Cloud. No more than 2 okta of cumulus, which reflects off the glass and misleads the camera. When load or irradiance moves by more than 10 percent a minute, for example under passing cirrus, the specification recommends waiting 15 minutes before imaging. The IEA PVPS Task 13 review puts the settling time of a module at 5 to 15 minutes.

Soiling. None or low, which the specification puts at under 10 percent loss of operating current, with cleaning recommended if there are bird droppings. A dirt-shaded cell heats like a broken one.

Plant state. The plant must be operating, and its DC load recorded through the survey, so that a string left open or short-circuited by a grid event is not mistaken for a fault.

The 600 W/m² floor comes down to signal size. The Task 13 review of infrared and electroluminescence imaging explains that faults which leave part of a module inactive, such as a disconnected string, shunted PID cells or a short-circuited bypass diode, warm that area only slightly, and the difference is often proportional to irradiance. Suppose a dead substring runs 4 K warmer than its neighbours at 600 W/m². At 300 W/m² the same fault shows 4 × 300 ÷ 600, or 2.0 K.

The review recommends weather that gives a difference above 2.5 K, the line it draws to rule out an artefact, and it asks for care with differences of 1 to 2 K because convective cooling and emissivity effects can outweigh them. So that 2.0 K reading is weak evidence. A healthy module also carries natural gradients of up to 8 K, but those are smooth ramps, and a defect shows as a step at a cell, string or module edge.

Straight-line scaling suits area faults like these. The review says point defects such as solder joints behave less predictably, and clause 7.4 of the specification has its own projection method.

The conditions IEC TS 62446-3 sets for imaging modules: at least 600 W/m2 in the plane of the modules, wind no more than 4 on the Beaufort scale or 28 km/h, no more than 2 okta of cumulus, low soiling, and an operating plant with the DC load recorded

The conditions for a valid thermography survey of modules. Below 600 W/m2 a real fault can shrink to a difference too small to trust. Source: IEC TS 62446-3:2017, clause 5.3; IEA PVPS Task 13 report T13-10:2018.

What Camera Does a Survey Need?

Clause 4 sets the minimum in a table.

Spectral range. 8 to 14 micrometres for modules. The 2 to 5 micrometre band is allowed only for electrical parts such as fuses, because glass is transparent near 3 micrometres.

Sensitivity and accuracy. Thermal sensitivity (NETD) of 0.1 K or better at 30 °C, and an absolute measurement error under plus or minus 2 K.

Calibration and files. A traceable calibration at least every two years, so ask for the certificate. Images are stored with full radiometric data, because a non-radiometric picture shows pattern and, at best, differences.

Resolution. No more than 3 cm of module edge per pixel, which the specification equates to 5 by 5 pixels on a 6 inch cell. Cell sizes have moved on since 2017, so work from the 3 cm. The camera should view the module as square-on as possible and always at more than 30 degrees to the module plane, because the emissivity of glass falls at shallow angles. A camera of at least 320 × 240 pixels is recommended, with a separate photo camera for the visual record.

The footprint of one pixel is the camera's instantaneous field of view, in milliradians, multiplied by the distance. At 3.0 mrad that is a 30 mm pixel at 10 m, which passes, and a 42 mm pixel at 14 m, which does not. A drone works the same way, with its height above the modules as the distance.

What Do the Common Anomalies Look Like?

Seven patterns are worth learning first.

Hot cell. One cell clearly hotter than those around it, from broken cells or front glass, external shading or internal defects. In extreme cases hot spots can smoulder.

Hot substring. A block of cells, normally one bypass section, runs evenly warmer than the rest of the module. That substring has stopped delivering power and the current is going around it, usually through its bypass diode.

Bypass diode. A warm spot over the junction box. A diode runs hotter when it conducts, which it should do only when cells are shaded, mismatched or damaged. Treat a hot diode as a symptom and ask what it is bypassing.

Junction box. An overheated terminal or solder joint inside the box. That is a resistance fault, and the diode may be fine.

String out. A whole string, or one inactive module, warmer than its neighbours, because the energy that would have left as electricity stays in it as heat. Check the DC load log first: an open isolator, a blown fuse and an inverter input that is not tracking all look the same.

PID pattern. Potential-induced degradation shows as warmer cells near the bottom frame, or a patchwork, in modules near the negative end of the string. The review notes that the pattern disappears at open circuit, so the plant must be loaded.

Connector and combiner faults. A hot cable lug, string fuse clamp, loose screw, inverter DC connector or AC termination. We would deal with these first, because heat here has a short route to fire, and the specification requires a visual photo of every safety-relevant abnormality.

Allow for false alarms. Sun, clouds, buildings and trees reflect off the glass, so move the camera and see whether the pattern moves too.

Seven common thermal patterns in a PV survey: hot cell, hot substring, bypass diode, junction box, string out, PID pattern, and connector and combiner faults, each with what to check next

Seven thermal patterns to learn first in a PV survey, and what to check for each. Source: IEA PVPS Task 13 report T13-10:2018; MIMAH engineering practice.

Simplified or Detailed, Drone or Handheld?

The specification defines two levels of inspection.

Simplified. A limited check that modules and electrical parts are working, for example at a basic commissioning. It asks less of the surveyor's qualifications, reads thermal patterns rather than exact temperatures, and cannot support firm conclusions about module quality.

Detailed. Absolute temperatures may be measured alongside temperature differences, for periodic inspections and for troubleshooting an underperforming plant. An authorised PV expert carries it out with thermography experts.

Drone imaging is classed as simplified. The specification says drones speed up large-area detection but can lack the resolution to identify specific failure modes, and that smearing may appear at about 3 m/s with common bolometer detectors. A note proposes a drone pass over the whole array to find the problem modules and strings, then a detailed inspection of those modules. It does not say how, and we would do that step on foot with a handheld camera.

A 2024 review of drone-based plant diagnosis in Heliyon by Michail and colleagues summarises a comparison on a 3 MW plant. The manual inspection recorded 1,132 anomalies, a flight at 30 m recorded 421 and a flight at 80 m recorded 275, which is about 37 percent and about a quarter of the manual count. In the most severe class, 30 K or more above normal, the manual survey recorded 92, the 30 m flight 9 and the 80 m flight none. Part of that last gap is probably the pixel and not the fault: a coarse pixel averages a small hot spot with the cooler cell around it, so the same fault reads cooler from the air. This is one plant and one set of flights, so take it as a warning about resolution and no more.

The speed advantage is real, with the Task 13 review putting raw aerial imaging at 2 to 3 MWp an hour. On a plant of several megawatts we would fly first and walk the flagged strings second. On a 50 kW rooftop, walking the rows is quicker than rigging a flight.

Comparison of the simplified and detailed inspection levels in IEC TS 62446-3, with drone imaging classed as simplified and a detailed inspection of the modules it flags

The two inspection levels in IEC TS 62446-3 and where a drone fits. In one 3 MW comparison a 30 m drone flight recorded 421 anomalies and a manual inspection 1,132. Source: IEC TS 62446-3:2017; Michail et al., Heliyon, 2024.

How Do Dust, Haze and Heat Limit a Survey?

Dust is the first limit. Where soiling is strong the specification recommends cleaning first, then letting the modules settle, unless you are doing root cause work and want the plant imaged as it runs. Our guide to panel maintenance in dusty climates covers cleaning without scratching the glass.

Haze comes next. Fine dust in the air can hold plane-of-array irradiance under 600 W/m² through a day that looks bright, so trust the sensor and be ready to wait.

Heat is the least obvious. The camera table asks for operation from minus 10 to plus 40 °C, and summer afternoons on many desert sites pass 40 °C, so check the rating of your camera and drone payload before booking the day. A hot module is not by itself a finding, because a survey reads differences between neighbours.

What Should the Survey Report Contain?

Clause 8 covers the inspection report and sits outside the public preview, so check yours against a purchased copy. What follows is our own view of a usable report, built from the public clauses and the Task 13 review.

The survey. Who did the work, the camera and its calibration record, the date and time, and the scope. Clause 5.1 wants the scope agreed in writing beforehand and any known deviations or limitations noted in the report.

The conditions. The quantities in Table 2 of the specification: irradiance, air temperature, wind speed, cloud cover, soiling and module or string current. Clocks on every instrument are synchronised so that each image can be matched to irradiance and DC load.

The findings. Every finding placed on the string plan, in the plant's own string and module references. Task 13 starts its analysis by matching each image to its location.

The images. Radiometric files with the emissivity and reflected temperature used, and a visual photo for every type of abnormality.

The actions. What to do about each finding, and by when. Our solar O&M contract guide shows where thermal imaging sits in a maintenance scope.

Clause 5.1 recommends a periodic survey every four years and leaves the actual interval to agreement with the owner, or to national electrical codes. On a hot, dusty site with a performance guarantee we would fold a simplified survey into the annual visit and keep the detailed one for trouble.

What Should You Do With a Finding?

The specification sorts findings into classes of abnormality in clause 7.3, also outside the preview, and Task 13 notes that it is not meant to give modules a pass or a fail. The review's own recommendations for action are public, and they make a sensible starting point.

A single cell less than 10 °C hotter than its surroundings is normally unproblematic. From 10 to 20 °C it wants watching at later surveys. Above 20 °C the review expects lost output and says to replace the module, and it says the same for a permanently active bypass diode. One warm module, or several in series, points to the wiring. A PID pattern needs its root cause found. If more than 10 percent of modules show abnormalities, find out why, because the cause could be module quality or a planning, installation, maintenance or operating fault.

Should You Combine Part 3 With Part 1 Commissioning?

At handover, usually yes. The imaging needs the plant running under load in strong, steady sun, so one visit can run the Part 1 tests, reconnect every string, then image. Basic commissioning is the specification's own example of a simplified inspection, and Task 13 says that for a warranty claim on the modules the first survey should be done at commissioning.

Do not let the commissioning report shrink the thermography to one line reading "IR scan: OK". Thermography does not measure power, so pair it with IV curves. Our guide to solar panel degradation and lifespan explains why baselines matter years later.

How Do You Specify a Survey You Can Rely On?

We would write these into the tender or the O&M scope. The wider contract context is in our guide to solar EPC contract clauses.

Name the document, edition and level. IEC TS 62446-3 with its edition date, simplified for commissioning and whole-plant screening, detailed for periodic checks and troubleshooting.

Require evidence of conditions. Logged irradiance, wind and cloud for each image set, and the owner's right to reject sets outside the limits.

Require proof of the camera and the people. Model, serial number, a calibration certificate under two years old, the resolution calculation for the distance or flight height used, and the analyst's thermography qualification. Task 13 names ISO 9712 Level 2 as a guide.

Tie findings to action. A repair time for each level of severity, and a repeat survey after repair.

MIMAH's renewable energy team handles solar EPC and performance audits on existing arrays. If you are writing a thermography requirement into a tender, or hold a survey report you cannot interpret, get in touch.