Answer
A thermal drone flies a radiometric camera over a solar farm and shows hotspots, bypass-diode problems and inactive strings as temperature differences. IEC TS 62446-3 sets the conditions: at least 600 W/m² irradiance and a resolution of 3 cm per pixel or finer. IEA PVPS Task 13 puts the work at about 1 hour per MW of ground-mount plant.
Last updated: November 2026.
A drone survey is the fastest way to see which modules on a large array are running hot. The catch is that the survey is only valid in the right light, at the right resolution, with a report you can check. Inverter-level monitoring tells you how a plant or string is doing. A thermal camera tells you where on the array the heat is wrong. This guide covers what the standard asks for, what the equipment can do, what published inspections have found, and where the numbers are still thin. Every figure below names its source. Where we could not read a source, we left the number out.
Key Takeaways
- IEA PVPS Task 13 reports that IR inspection at the 3 cm per pixel resolution generally takes about 1 hour per MW of ground-mount plant, and that the 600 W/m² irradiance rule leaves about 4 qualifying hours a day.
- Inactive modules and strings run warmer, not cooler, because the energy they do not convert to electricity stays as heat.
- Two published sources give different wind limits for IEC-style surveys: under 4 m/s (IEA PVPS) and under 28 km/h, about 7.8 m/s (DTU slide citing the standard).
- Drone surveys are mostly qualitative. They show where a fault is, not how many kWh it costs.
- Published cost data for inspection services is thin. Ask for quotes and compare scope, not just price.
What Published Inspections Found
These two examples come from public pages you can open and check.
Nexamp, USA. Nexamp, a community solar company, wrote that its drone pilot found two potential string faults, at its Charlton Solar and Golden Hills Solar farms, that its established fault detection had not caught. It confirmed and corrected both. One was a blown fuse. The other was a loose string that was not fully disconnected. Source: Nexamp, “Using Drones for Solar Farm Inspections” (first posted 9 July 2020, updated 2 April 2026).
Sumatra, Indonesia. A University of Surabaya team studied early defects in a 24.9 MWp plant. The study did not rely on drones alone. It used string-output checks, visual inspection, measurements and thermal imaging. As reported by pv magazine, the plant has 64,140 panels of 390 W each, while the researchers’ own quote counts 64,400 modules. Both figures are printed in the article. About 678 modules showed abnormalities in the first two years, a failure rate of about 1.05%. Hotspot formation was the dominant defect. The reported counts were 350 junction box failures, 282 cases of glass cracking, 42 cases of junction box swelling and 7 cases of foggy or discolored glass. Source: pv magazine, 19 February 2025.
Check on the arithmetic: 678 ÷ 64,400 = 1.05%. 350 ÷ 64,140 = 0.55% of modules with a junction box failure.
- Junction box failures350
- Glass cracking282
- Junction box swelling42
- Foggy or discolored glass7
Source: pv magazine report on a University of Surabaya study, checked 19 February 2025.
Start From a Documented Array Layout
Inspection findings are only useful if you can match them to a string and a module position. SurgePV supports array layout and electrical workflow design, and a guided demo shows how that works on your own project.
Book a Free DemoChecking cost first? See the SurgePV pricing page
What Is Drone Thermal Inspection for Solar Farms?
Drone thermal inspection is an aerial survey method. A drone equipped with a radiometric thermal camera flies a pre-programmed grid over a solar farm. The camera captures temperature data for every panel surface. Defective cells generate excess heat instead of electricity. They appear as bright hotspots on thermal images.
The key word is radiometric.
A standard thermal camera shows relative heat patterns. A radiometric camera stores the actual temperature value of every pixel. That precision matters for fault classification, because a small temperature difference means something different from a large one. The radiometric data tells you which.
Modern setups pair the thermal sensor with a visual (RGB) camera. When the software flags a hotspot, the operator can cross-reference it with a visual image and check for dirt, shading or reflections.
The inspection happens while the panels are under load. Solar irradiance must be high enough for the thermal signatures to show. The IEA PVPS report explains why: modules that are not working convert less sunlight into electricity, and the energy they do not convert stays as heat, so the temperature gap grows with irradiance.
Pro Tip
Always request radiometric thermal data, not just visual thermal overlays. A report built only on colour images cannot show the measured temperature behind each finding.
How Thermal Drone Inspection Works
The process splits into three phases: pre-flight planning, data capture, and post-processing.
Pre-Flight Planning
The operator defines the flight path using mission planning software, usually a grid that covers every row. Altitude sets the resolution. IEC TS 62446-3 asks for a maximum of 3 cm per pixel, so the altitude has to be low enough for the camera you fly. A DTU Electro study for IEA PVPS Task 13 flew at 8, 10, 14 and 20 m and saw resolutions from under 1 cm per pixel to about 3 cm per pixel. IEA PVPS also advises a minimum flight height above 5 m so the drone does not shade the modules it is scanning.
The operator also records irradiance, wind speed and ambient temperature, because they affect the thermal readings and belong in the report.
Data Capture
The drone follows the programmed path on its own. Each thermal and visual image is geotagged. The operator watches a live feed from the ground. Flights must follow the rules of the local aviation authority.
The IEA PVPS report gives the pace. At 3 cm per pixel, an IR inspection of 1 MW of ground-mount plant generally takes 1 hour. Because the 600 W/m² rule leaves about 4 qualifying hours a day, one crew flying at that rate covers about 4 MW a day (4 hours ÷ 1 hour per MW). A 100 MW plant would then need about 25 qualifying days (100 ÷ 4). Cloud makes it worse: the standard suggests waiting 15 minutes after conditions change by more than 10% per minute. A larger area per hour is possible with manned aircraft, but at coarser resolution. IEA PVPS cites 100 MW per hour for one such service, with limits on finding cell-level faults.
Post-Processing and Analysis
After the flight, the images are processed into a map of the plant with thermal and visual layers. Software or an analyst then marks anomalies and ranks them. IEA PVPS notes that many services still depend heavily on manual or semi-manual image processing, and that human error lowers diagnostic accuracy and consistency. It also notes that these services are mostly qualitative: they show which part of the plant has a fault, not the exact yield loss.
In Simple Terms
Think of drone thermal inspection as an aerial scan of the plant. The camera records heat. The software or an analyst turns that heat map into a list of places a technician should check.
Fault Types Detected by Thermal Drones
Faults look different in thermal images. The IEA PVPS Task 13 review lists the main thermal abnormalities in PV plants as hot cells, heated bypass diodes in the junction box, heated string fuses in combiner boxes, and heated DC and AC cables and connection points.
| Fault | What it looks like in a thermal image | Source |
|---|---|---|
| Hot cell (hotspot) | One cell hotter than its neighbours. Causes include front glass breakage, external shading and internal cell problems. | IEA PVPS T13-10 |
| Active or overheated bypass diode | Warm area at the junction box. | IEA PVPS T13-10 |
| Disconnected module or string | Warmer than active neighbours, because unconverted energy stays as heat. The difference is small and grows with irradiance. | IEA PVPS T13-10 |
| PID-s (shunting) | Warmer cells close to the bottom frame, or a patchwork pattern, on modules near the negative end of the string. | IEA PVPS T13-10 |
| Heated fuse or connector | Hot spot at a combiner box fuse, clamp or cable connection. | IEA PVPS T13-10 |
| Localised dirt such as bird droppings | Can create hotspots. DTU saw a temperature difference above 10 °C even at 200 W/m². | DTU Electro, IEA PVPS T13 meeting 2025 |
| Cell cracks | Hard to see. Hotspots from cracks became hard to assess at 600 W/m² and unidentifiable below it, with differences of 1 °C at most. | DTU Electro, IEA PVPS T13 meeting 2025 |
Sources: IEA PVPS Task 13, Review on Infrared and Electroluminescence Imaging for PV Field Applications (T13-10, 2018) and DTU Electro, Frame Conditions for Thermography Inspections of Photovoltaic Plants (IEA PVPS Task 13 meeting, 2025).
The IEA PVPS failure fact sheets add background on why these faults matter. They describe junction box defects, such as a loose or burned box, as a fire and shock risk. See the IEA PVPS Task 13 failure fact sheets (T13-30, 2025). The fact sheets do not give the Sumatra plant counts quoted earlier on this page. Those come from the pv magazine report.
What Drones Cannot Detect
A drone sees the top surface of the array. It does not inspect the underside of modules, cabling out of view, or mounting hardware, and it does not replace electrical tests such as I-V curve tracing. Use it to decide where ground crews should look first.
For a site survey before construction, read solar drone site survey.
IEC 62446-3 and Inspection Standards
IEC TS 62446-3:2017 covers outdoor infrared thermography of PV modules and plants in operation. The IEC webstore description says it lays down requirements for the measurement equipment, ambient conditions, inspection procedure, inspection report, personnel qualification and a matrix of thermal abnormalities. The full text is sold by IEC, so the table below uses the figures that IEA PVPS Task 13 and DTU Electro publish from it.
Conditions and Equipment
| Parameter | Value | Source |
|---|---|---|
| Irradiance in the module plane | At least 600 W/m² | IEA PVPS T13-24; DTU Electro 2025 |
| Spatial resolution | Maximum 3 cm per pixel | IEA PVPS T13-24; DTU Electro 2025 |
| Wind | Not above 4 m/s (14.4 km/h) per IEA PVPS best practice. Under 28 km/h (about 7.8 m/s) per a DTU slide citing the standard | IEA PVPS T13-24; DTU Electro 2025 |
| Sky | Sunny, cloudless, clear. Wait 15 minutes after conditions change by more than 10% per minute | IEA PVPS T13-24 |
| Time of day | Morning or afternoon, when the angle of incidence is lower, as long as irradiance stays above the minimum | IEA PVPS T13-24 |
| Imager optical resolution | At least 640 × 480 pixels | IEA PVPS T13-24 |
| Thermal sensitivity (NETD) | Below 0.1 K at 30 °C | IEA PVPS T13-24 |
| Measurement error | Below ±2 K or 2% | IEA PVPS T13-24 |
| Spectral response | 2 to 5 μm or 8 to 14 μm | IEA PVPS T13-24 |
Source for the IEA PVPS rows: IEA PVPS Task 13, Qualification of PV Power Plants using Mobile Test Equipment (T13-24, 2021). That report lists the imager specifications as a recommendation for IR systems compliant with IEC TS 62446-3. It also says that, when written, there was no separate standard for aerial IR measurements, and that such inspections should still follow IEC TS 62446-3.
The two wind figures conflict. We print both rather than choose. Check the current text of the standard, or your provider’s report, before you rely on either.
Why Compliance Matters
Ask any provider whether its report follows IEC TS 62446-3 and whether it records irradiance, wind, camera details and resolution. Those are the inputs a reviewer needs to judge whether a finding is reliable. We did not find a public rule that says an insurer or manufacturer must accept or reject a claim on this basis, so check your own warranty and policy wording.
Pro Tip
When you evaluate a provider, ask for a sample report. Check that it records irradiance, wind speed, camera model, calibration and pixel size on the ground. If any of these are missing, the report may not hold up in a dispute.
Drone Hardware and Software Platforms in 2026
Most solar inspection drones are enterprise models with a radiometric thermal camera and a visual camera. The table lists three common options, using only the maker’s own specification or shop page.
| Aircraft and payload | Thermal sensor | Sensitivity and accuracy | Flight time | Price |
|---|---|---|---|---|
| DJI Matrice 350 RTK with Zenmuse H30T | 1280 × 1024, 8 to 14 μm | NETD ≤ 50 mK at f/1.0 | 55 min without payload in no wind (DJI test condition) | Not on DJI’s spec pages |
| DJI Mavic 3T | 640 × 512, 8 to 14 μm | NETD ≤ 50 mK; ±2 °C or ±2% in high gain | 45 min in no wind (36 min with the EU C1 battery) | Not on DJI’s spec pages |
| Autel EVO II Dual 640T Rugged Bundle V3 | Up to 640 × 512 | Not listed on the shop page | 38 min | USD 4,799 on Autel’s shop page, October 2026 |
Sources: DJI Zenmuse H30 series specs, DJI Matrice 350 RTK specs, DJI Mavic 3 Enterprise specs and the Autel EVO II Dual 640T Rugged Bundle V3 shop page. DJI says its flight times are lab figures for reference. A heavier payload and wind shorten real flights.
All three sensors meet the 640 × 480 minimum that IEA PVPS recommends. Resolution alone does not make a valid survey: altitude, irradiance and wind decide whether the image meets the 3 cm per pixel limit.
For context, the 2021 IEA PVPS report listed aircraft-only prices from USD 2,800 for a DJI Mavic Enterprise 2 with a FLIR camera, to USD 17,500 for a Freefly ALTA 8. It noted that the camera is a large share of the system cost. Those are 2021 figures and are not current prices.
Software and Service Providers
After the flight, images go to a processing tool or a service provider. The IEA PVPS report names several aerial inspection services, including Sitemark, Heliolytics, Scopito, AePVI, Aeroprotechnik, Above and Aerospec, and describes Raptor Maps as AI-driven software for aerial inspections. That list is from 2021 and is not a ranking. We do not repeat vendor accuracy claims here, because they are self-reported and tested on each vendor’s own data.
SurgePV does not sell inspection drones or drone analysis software. If you are comparing monitoring tools for the plant itself, see the best solar monitoring software list.
Keep inspection findings tied to your design
Thermal anomalies are easier to act on when every string and module position is documented. See how SurgePV handles array layout and string design on your own project.
Book a DemoChecking cost first? See the SurgePV pricing page
ROI and Cost Analysis
We could not find a public price list for thermal inspection services, so this page gives no price per MW. A fair cost comparison needs written quotes that state the plant size, the resolution, the report format and whether the provider follows IEC TS 62446-3.
What Providers Claim
IEA PVPS Task 13 reports that service providers justify a quick return on aerial inspection with two claims: preventive O&M costs more than 10% lower than with manual and ground-based inspection, and recovery of an average 1% to 2% of PV power capacity losses. These are provider claims as reported by IEA PVPS, not independent measurements.
How to Build Your Own Case
Use your own numbers:
- Estimate the energy lost to the faults a survey could find: annual generation × loss fraction.
- Multiply by your tariff or PPA price to get revenue at risk.
- Compare it with the quoted survey cost and the repair cost.
Time is the other input. At about 1 hour per MW and 4 qualifying hours a day, a 20 MW plant needs about 5 survey days for one crew (20 ÷ 4). Plan around that, not around the very high daily figures that appear on some vendor pages.
What We Left Out
Earlier versions of this page carried service prices per MW, ROI percentages and a Gujarat case study with no source. We removed them. If you have written quotes from providers in the USA, the Philippines or elsewhere, a regional cost table can be added once the quotes and their dates are on file.
Common Drone Inspection Mistakes
Treating the Drone as a Complete Solution
Drones detect and locate. IEA PVPS describes aerial IR services as mostly qualitative diagnosis, not reliable quantitative yield-loss estimates. A drone tells you which panel is hot. It does not tell you why. Ground crews with electrical testers answer that.
Flying Outside the Conditions
Irradiance below 600 W/m² hides faults. The DTU study found that hotspots from cell cracks became hard to assess at 600 W/m² and were unidentifiable below that. Cloud passing during a flight breaks the steady state, and the standard suggests waiting 15 minutes. A weather forecast with irradiance and wind helps you pick the window.
Judging a Camera by Pixel Count Alone
A 640 × 512 camera flown low can give a finer picture than a larger sensor flown too high. The test is whether the survey meets the 3 cm per pixel limit at the altitude you fly. Altitude, overlap and calibration matter as much as the sensor.
Trusting an Automated Report Without Review
Vendors publish their own accuracy figures. Independent numbers vary. One 2025 Scientific Reports paper on a research prototype reports 92.3% fault-detection accuracy in its abstract and “consistently above 95%” in its body text, which are two different figures from the same paper. Treat any single percentage with care and have an engineer review the borderline calls. Source: Scientific Reports, 2025.
Ignoring Dirt and Shading
Dirt and shade create thermal patterns that look like faults. DTU saw bird droppings produce hotspots with temperature differences above 10 °C. Clean and clear what you can before the flight and compare thermal images with the visual layer before sending a crew.
Common Limitations and How to Work Around Them
Weather Dependency
Thermal data needs stable, high irradiance and low wind. In regions with a monsoon season or regular afternoon storms, the window is short. Plan surveys for the dry season and keep spare days in the schedule.
Battery Life and Coverage
Flight time is limited. DJI quotes 55 minutes for the Matrice 350 RTK without payload and 45 minutes for the Mavic 3T, both in no wind. Real flights are shorter. Bring spare battery sets and plan the mission in blocks.
False Positives
Soiling, vegetation shadows, bird droppings and reflections create patterns that resemble faults. Clean and clear where you can, and check each anomaly against the visual image before dispatching a repair crew. IEA PVPS also lists reflections from module surfaces as an external effect to watch for.
Regulatory Barriers
Flights must follow the local aviation authority. In the USA, flying a drone for work under the FAA’s Small UAS Rule (Part 107) requires a Remote Pilot Certificate, according to the FAA. Large plants can exceed visual line of sight, so ask your provider how it handles that in your country.
Cannot Inspect Underside or Balance of System
A drone sees the top surface of panels. For cabling, inverters, transformers and mounting hardware, schedule ground checks.
Pro Tip
Start with a baseline survey soon after commissioning. Later surveys then have something to compare against, and you can tell a new fault from a feature of the original installation.
Getting Started: In-House vs. Outsourced
The choice between an in-house drone program and a service provider depends on plant size, inspection frequency and internal skills. Compare both before committing capital.
Outsourced Drone Inspection
Best for: occasional inspections and teams without drone expertise.
How it works: you contract a provider that brings the equipment, pilots, insurance and processing software, and you receive a report with a defect map.
Pros: no capital cost, no pilot training, access to current equipment.
Cons: scheduling depends on the provider, and you have less control over data quality and turnaround.
In-House Drone Program
Best for: operators with several sites and the technical staff to run a program.
How it works: you buy the equipment and software, train or hire a certificated pilot and fly on your own schedule.
Pros: control over timing and data, faster response to urgent issues.
Cons: upfront cost, training, equipment upkeep and your own compliance work.
Hybrid Approach
Some operators outsource the baseline survey to a specialist and fly follow-ups in-house with a smaller drone.
Questions to Ask a Provider
- What camera do you use, and is the data radiometric?
- Does your report follow IEC TS 62446-3, and can you share a sample?
- What irradiance and wind limits do you enforce, and what do you do if conditions drift?
- What resolution (cm per pixel) will the survey achieve?
- How do you classify and prioritise faults?
- Who reviews the automated findings?
- What is the turnaround from flight to report?
- How do you store and secure the data?
- Are your pilots certificated for commercial flights where the plant is?
Looking Ahead
IEA PVPS Task 13 expects aerial inspection to move from scheduled surveys toward data-driven ones, and links that shift to the commercial viability of these services. One practical step is to keep every survey on file so each new one can be compared with the last.
Design records help here. A documented array layout lets an inspection team match each anomaly to a string and module position. Cloud solar design software produces that layout, and the solar farm design guide covers how access paths and row spacing affect later O&M. The thermal drone inspection glossary entry gives a short definition. If you want to see how a documented layout works in practice, book a SurgePV demo.
Frequently Asked Questions
How much does a drone solar farm inspection cost?
We could not find a public price list for inspection services, so we give no price per MW. Ask providers for written quotes that state plant size, resolution, report format and IEC TS 62446-3 compliance. If you plan to buy your own drone, the Autel EVO II Dual 640T Rugged Bundle V3 was listed at USD 4,799 on Autel’s shop page in October 2026. DJI does not publish prices on its enterprise spec pages.
What faults can a thermal drone detect on solar panels?
IEA PVPS Task 13 lists hot cells, heated bypass diodes, heated string fuses in combiner boxes, and heated DC and AC cables and connections as the main thermal abnormalities. Disconnected modules or strings and PID-s patterns can also show, though the temperature difference for inactive parts is small. Cell cracks are hard to see in thermal images.
How accurate is AI-powered drone fault detection?
There is no single accepted figure. Vendors publish their own numbers on their own data. One 2025 Scientific Reports study of a research prototype reports 92.3% fault-detection accuracy in its abstract and above 95% in its body text. IEA PVPS notes that human error in manual image processing lowers accuracy, so keep an engineer in the loop.
What is IEC 62446-3 and why does it matter for drone inspections?
IEC TS 62446-3:2017 covers outdoor infrared thermography of PV modules and plants in operation. It sets requirements for equipment, ambient conditions, procedure, the inspection report, personnel qualification and a matrix of thermal abnormalities. IEA PVPS Task 13 reports figures from it, including at least 600 W/m² irradiance and a maximum of 3 cm per pixel.
How long does a drone inspection take compared to manual methods?
IEA PVPS Task 13 reports that an IR inspection at 3 cm per pixel generally takes about 1 hour per MW of ground-mount plant, and the 600 W/m² rule leaves about 4 qualifying hours a day. That is about 4 MW a day for one crew, or about 5 days for 20 MW. We found no source for a drone-versus-manual time saving, so we do not quote one.
Can drones fully replace manual solar farm inspections?
No. Drones are good at finding and locating anomalies. IEA PVPS describes aerial IR services as mostly qualitative. They do not see the underside of modules, hidden cabling or mounting hardware, and they do not replace electrical tests. Use the survey to decide where ground crews should go.
What weather conditions are required for valid thermal drone data?
Use clear, sunny sky and irradiance of at least 600 W/m² in the module plane. For wind, IEA PVPS says not above 4 m/s, while a DTU slide citing the standard gives under 28 km/h. If clouds change conditions by more than 10% per minute, the standard suggests waiting 15 minutes.
Where this fits
This article is part of SurgePV's Solar Technology hub, which works through the topic from first principles to the decisions a project team actually has to make.
