Solar farms: from hot cell to work order

Solar farms: from hot cell to work order

What owners, O&M providers, EPCs and commissioning agents fly drones for on utility-scale and commercial solar sites, what the standards ask of those flights, and — use case by use case — how Omniris turns those flights into graded anomalies, repair tickets, progress against the design and sealed custody records.

Summary

A solar plant is a very large number of identical parts spread over hundreds or thousands of acres, and many of its faults are too small to show up on the monitoring screen.[5][7] A string that is offline, a bypass diode that has switched on, a cracked cell slowly cooking itself — each costs a little energy, and together they add up. A radiometric thermal camera on a drone sees every one of them as heat, which is why aerial infrared thermography has become one of the two main ways, with manual electrical testing, to inspect the DC side of a plant: at commissioning, every year for preventive maintenance, at the end of a warranty and when a plant changes hands.[5][4] The same imaging finds cracked modules,[5] including front glass broken by hail,[3] and drones are flown during construction to count piles, racking and modules and to check them against the plan.[9]

The flight is the easy part. The industry's standards and guidance put the weight on what happens next: inspection conditions that must be met and documented, anomalies that must be located to the module and graded, a report that a field technician and an asset manager can both act on, and an anomaly list that leads to repairs rather than to another spreadsheet.[1][5] That is where Omniris works.

How solar uses drones today

This section summarises published practice: the standards, national-laboratory guidance, international research programmes and trade reporting. Numbers in square brackets point to the Sources. Figures that come only from a vendor are marked "vendor-reported".

Who buys the flights, and who flies them

• Asset owners and asset managers buy inspections to protect production and to hold O&M contractors, EPCs and module suppliers to their warranties. Annual plant-wide inspections are often a requirement of the plant's power purchase agreement.[7]
• O&M providers fly or subcontract the routine inspections and do the repairs. The IEA's photovoltaic research programme notes that more and more EPC and O&M companies offer regular infrared and electroluminescence inspections.[3]
• EPCs and commissioning agents fly during construction and at handover, when the plant's acceptance tests are run.[2][5]
• Specialist aerial-inspection firms sell turnkey services — the flight, automated analysis, a report and maintenance recommendations — from drones or from manned aircraft.[4] Manned aircraft cover far more ground per hour and focus on system-level faults; drones resolve faults down to the module, at lower cost.[4]
• Insurers and their adjusters need damage documented after hail, which accounts for more than half of solar insurance claim costs.[8]

The missions

How often

IEC TS 62446-3 recommends a periodic thermography inspection every four years, with the interval for a specific plant agreed with its owner or operator or set by national electrical codes, and it notes that faults of an intermittent or changing nature may or may not be captured at the time of an inspection.[1] In practice:

• A 2018 survey of utility-scale plants found that about half (52.4%) ran routine plant-wide inspections at all, and fewer than half of those did so more than once a year; many plants inspected only in reaction to a large loss, or inspected a "representative sample" of the plant.[7]
• NREL's operations-and-maintenance best-practice guide treats aerial thermal imaging as an annual preventive task, and its model statement of work lists an infrared scan of modules once a year.[5]
• An IEA photovoltaic programme presentation recommends "bi-annual (at least)" inspections, and asks whether intermittent faults can be diagnosed at an individual inspection.[4]

What a thermal inspection finds

A working cell turns part of the sunlight into electricity; a cell, substring, module or string that is not delivering power keeps that energy as heat and shows up warmer than its neighbours.[3] The common patterns:

Some faults are subtle: a disconnected string or a shunted cell may be only slightly warmer than its neighbours, and the difference grows with sunlight, which is why the inspection conditions below matter.[3]

What the standards and guidance ask for

IEC TS 62446-3:2017, Photovoltaic modules and plants — Outdoor infrared thermography, is the reference for thermal inspection of PV modules and plants in operation, including cables, contacts, fuses, switches, inverters and batteries.[1] It covers the camera, the equipment that records ambient conditions, the environmental conditions, the imaging procedure (with a section on fast carriers such as drones), evaluation, the projection of temperature differences to nominal irradiance, classes of abnormality, the inspection report, a matrix of thermal abnormalities, and the qualification of personnel.[1] It defines procedures and requirements but no pass/fail criteria, and it was not written specifically for aerial inspection.[4] It classes a drone inspection as a simplified inspection of the whole array, meant to find sub-arrays, strings and modules with obvious problems, which are then examined in a detailed inspection where that is agreed; the projection to nominal irradiance and the allocation into classes of abnormality apply in particular to the detailed inspection.[1]

What buyers expect to receive

NREL's guidance is explicit: imagery should be processed by a validated routine that identifies module defects and their location in the field, to the module level, labelled so that a technician can find and fix each one; without that, an aerial inspection "simply points out symptoms and the rough location", and troubleshooting on the ground is still needed.[5] A typical deliverable is therefore an anomaly register — each anomaly with its class, its temperatures, a thermal and a colour image, coordinates and, ideally, the block, string and module it belongs to — together with the inspection conditions, a plant-level summary and often an estimate of the power and revenue lost.[5][7]

What it replaces, and what it costs

• Manual electrical testing — open-circuit voltage, operating current or I-V curve tracing at every combiner — takes days or weeks on a large plant, requires opening enclosures under arc-flash protective equipment, is accurate to around 5% with standard field I-V tracers, and does not detect hot spots that are not yet degrading the module electrically. NREL notes that aerial thermal imaging can be performed instead of manual electrical testing as part of annual preventive maintenance.[5]
• Hand-held thermography on a 4 MWp plant takes roughly 15–20 hours; a drone, roughly 5–10.[3] Raw aerial imaging can reach 2–3 MWp per hour.[3] An IEA programme comparison puts drones at about 12 MW a day and manned aircraft at 100–150 MW an hour, at higher cost.[4]
• Price. A 2018 survey of providers found aerial inspection prices of $161–$800 per MW inspected, generally cheaper than manual inspection, noting that aerial inspections typically lead to further investigation, which is always done manually.[7] NREL's O&M cost model uses, as its worked example for a 5 MW ground-mounted plant, $4,500 every five years for infrared inspection.[6]
• The stakes. Utility-scale O&M runs as high as €8,000–15,000 per MW a year; the same IEA programme presentation reports, from experience with aerial inspection, preventive O&M more than 10% lower and recovery of an average 1–2% of PV power losses.[4]

Pain points

• Inspections done to tick a box. Annual inspections to satisfy a PPA, reactive inspections only after a large loss, and sampling instead of plant-wide surveys are all common.[7]
• Small faults left to grow. In one documented case, grass shading the bottom of a table caused two hot spots; one was fixed by mowing, the other had spread until the module had to be replaced.[7]
• Monitoring that cannot see the fault. A string offline behind a 16-string combiner barely moves the combiner's output, so monitoring alone may not prompt the cheap fix.[7]
• Conditions that are hard to prove. The thermography specification sets minimum sunlight, wind and cloud conditions, and its inspection report must state the air temperature, wind, cloud cover and plane-of-array irradiance on the day,[1] and aerial inspection practice pairs the drone with temperature and irradiance sensors to record them.[4]
• Vegetation. Vegetation management is a recurring cost; an NREL study of 54 utility-scale sites found that mowing, trimming and herbicide application cost less per acre at sites with native or pollinator-friendly ground cover than at turfgrass sites, but total vegetation costs were slightly higher, which the authors attribute to the extra work of the first three to five years of establishment.[10]

The drones themselves (United States)

Two federal actions affect which drones can be bought and flown, particularly foreign-made ones:

• FCC Covered List. On December 22, 2025 the FCC added uncrewed aircraft systems and UAS critical components produced in foreign countries to its Covered List, which bars new equipment authorizations for them.[11] On January 7, 2026 it exempted drones on the Blue UAS Cleared List of the Defense Contract Management Agency and those qualifying as domestic end products under the Buy American Standard, until January 1, 2027;[12] on July 21, 2026 that exemption was extended to January 1, 2028, and specific foreign-made drones and components can also be exempted by a conditional approval from the Department of War or Homeland Security.[14] A later proposal to prohibit continued importation and marketing of certain covered equipment states that it would not affect continued use of equipment already purchased.[13] The list changes; check it for the drone in question.
• American Security Drone Act. Since December 22, 2025, federal funds — including, for example, Department of Energy grants and cooperative agreements — may not be used to procure or operate drones made or assembled by a listed covered foreign entity.[15] Federally funded solar projects need to check which drones their pilots fly.
• Part 107. Commercial drone flights in the US are flown under 14 CFR Part 107, which limits altitude to 400 ft above ground level, with an exception near structures.[16]

Use case by use case: what Omniris does

Omniris is three things a solar team uses together: a processing stack that turns drone photos into maps, elevation models, temperature maps and 3D models; a secure vault that keeps every file with a tamper-evident custody record; and, in the cloud and data center editions, a marketplace where the flights can be ordered from pilots. The table maps each mission above to what Omniris delivers for it.

Thermal inspection in detail

The full description of Omniris thermal processing is in the Thermal Imaging paper. This section covers what a solar inspection uses, and what the Solar site type changes.

From radiometric photos to a temperature map

Upload the flight's thermal and colour photos together. Omniris recognises the thermal photos, converts each to surface temperature and stitches them into a temperature map in which every pixel is a measured temperature. Only a map whose values pass a temperature check is labelled THERMAL, and only those are read by the tools, the report and Orion (the Omniris assistant).

Supported radiometric cameras include the DJI Mavic 3 Thermal, Matrice 4 Thermal, Matrice 30T, Zenmuse H20T / H20N and H30T and Mavic 2 Enterprise Advanced; the Skydio X10 and other FLIR radiometric cameras (those other than the X10 are labelled "camera not validated"); and the MicaSense Altum and Altum-PT thermal band.

For the DJI and FLIR cameras, the conversion uses five settings, reviewed under Temperature measurement settings and saved with the job: emissivity, distance, humidity, air temperature and reflected temperature. Humidity and air temperature are prefilled from the weather reading nearest the capture time, and the dialog names the source and the hour; distance from the drone's recorded height. Emissivity defaults to 0.95. The IEA's review gives about 0.85 for module glass viewed within 30° of perpendicular and 0.95 for polymer backsheets,[3] so for a module survey set the emissivity to suit the glass.

Flying it so it can be mapped

• Straight down for a map. A temperature map needs photos taken with the camera pointing straight down. Before a stitch, Omniris reads the camera angle from every photo; a flight shot well off vertical is not started as a map, and the dialog offers a 3D model instead.
• Tilted panels and trackers. The emissivity of glass holds only when it is viewed within about 30° of perpendicular,[3] so where rows are tilted steeply, as on tracker sites, a camera pointing straight down sees the glass outside that range; tilting the gimbal with the rows keeps it inside. Such a flight cannot become a temperature map. For DJI thermal drones whose matching colour photos are in the same folder, it can become a thermal skin on the 3D model: the model's surface coloured by temperature, viewed in 3D with an RGB / Thermal switch (see the Digital Twins paper). The skin shows temperature; it is not a surface you can click for a reading.
• Ground detail. IEC TS 62446-3's camera requirement is at most 1.2 in (3 cm) of module edge per thermal pixel, about 5 × 5 pixels on a cell.[1][3] The Omniris mission planner works out ground detail from the drone's visible camera, so for a thermal survey plan the height from the thermal camera's specification.

What the Solar site type changes

Every site in Omniris has a type: Agricultural, Solar (PV plant), Roof / building envelope, or Other. Setting a site to Solar changes where the anomaly search starts, and how the report is worded. Every value remains editable for each run.

Each anomaly is graded on the average rise of its pixels over their local background, using the specification's hot-cell figures (it gives no hot-cell figure below 10 K, so a smaller rise is shown as CoA 1):

Measuring and searching on the map

In the Site Navigator (the map view of a site's flights), a THERMAL layer is coloured with the ironbow scale over that flight's 2nd–98th percentile and gains five tools. Every number comes from the stored temperatures, °F first with °C beside it:

• Spot probe and ΔT: a temperature at a click; two clicks compared vs a similar component or vs ambient, with a severity grade from electrical-inspection practice. With two flights switched on, the second click reads the same flight as the first wherever that flight has data; when it cannot, the panel names both flights and says the ΔT compares two flights.
• Region statistics inside any outline (an outline wider than the full-resolution read covers is measured from a reduced-resolution read, and the panel says so beside the pixel count), and temperature highlights above, below or between values you type.
• Anomaly search: runs on the server, ranks anomalies by peak rise and keeps the 50 strongest of a run, saying how many it found in all; the panel lists the 15 hottest with peak, average rise, class and area and numbers them on the map. Re-run with other settings at any time; each result is saved with its settings.

Illustrative example values. A single hot cell (rank 1 and 3) and a warm substring (rank 2) as the report lists them.

The report

The stitch report is generated once and saved with the results. For a thermal flight on a Solar site it contains the temperature map with its scale, the ten highest-ranked anomalies with their class column and the basis of the classes, the weather during the flight, or on its day when the flight window is not known (air temperature and wind), from a weather archive, and a solar note. The note cites the conditions IEC TS 62446-3 calls for — at least 600 W/m² of plane-of-array irradiance and wind below 17 mph (28 km/h) — asks the reader to check the recorded wind against that limit, and states that irradiance was not recorded, so the grades are indicative rather than a standard-compliant classification. If detection is re-run after the report was made, Refresh Report brings it up to date.

Asking Orion, and sharing

Orion, the Omniris assistant, reads the same saved results. Ask it for the hottest spot or for the anomalies on a flight and it answers with the temperature map, a numbered marker and the coordinates; it never estimates a temperature, and it sees only what the person asking is allowed to see. A read-only Site Navigator link lets someone without an Omniris account — an owner, an O&M contractor, a module supplier — read point temperatures, area statistics, temperature highlights and the latest anomaly search on the shared flights.

From anomaly to repair ticket

An anomaly table tells you where the plant is hot. A finding is what gets it fixed: a work item with a place on the map, a category, a severity, a status, a person responsible, evidence images and a complete history. Findings are described in full in the Findings paper; this is how they serve a solar O&M programme.

Lists, groups and exports for the O&M team

The Findings view lists and maps every finding with filters and bulk updates. Named groups ("Inverter 3", "Work order 112") organise them by block or campaign. Exports are GeoJSON, CSV and KML, up to 20,000 at a time; signed webhooks and an API with read-only or read-write tokens keep a work-order system in step.

Construction, commissioning and as-built

On a solar build, drones are flown to count what has been installed — piles, racking, modules — and to catch piles, fencing and trenches that are not where the plan puts them.[9] Omniris covers this through its stitching, comparison and design-progress tools. The Stitching paper and the Design and Progress paper describe them in full.

Weekly maps of the build

• Maps and elevation. Each flight is stitched into an orthomosaic, a surface model, a bare-ground terrain model and a point cloud, filed under the site by flight date. Name the site when the stitch starts and every week's flight lines up in the Site Navigator.
• Accuracy stated. RTK flights and ground control points tie the map to survey coordinates; checkpoints left out of the fit give measured horizontal and vertical error on the delivered map. The site's RTK settings state the correction frame, and there is no default, so positions are never silently placed in the wrong frame.
• Earthworks. Compare two flights' elevation models for material added and removed, each with its uncertainty, or compare a flight with the design surface for cut and fill. A saved volume is re-measured on every new flight that covers it.
• Change alerts. An elevation comparison can become a standing watch that emails you, and optionally files findings, when a new flight changes an area by more than you set.

Progress against the design

Import the design — IFC, DWG, DXF or DGN — and each element becomes a named object on the site, drawn over every flight on the map and in 3D. Import the Primavera P6 or Microsoft Project schedule and link activities to design objects: pick them on the map, draw around an area, or match by a rule such as layer name against activity code. Each object then counts by number, length, area, volume, mass or percent, and after each flight it is measured one of three ways:

Measured values are proposals until a person approves them; one approval covers a flight and is sealed into the custody record. Approved progress rolls up to percent complete against plan, a forecast finish and earned value, and Export to schedule writes physical percent complete and actual dates into a copy of the P6 or Project file. Findings, files and issues can be kept in step with Procore and Autodesk Forma.

Counting with AI

Open-vocabulary AI detection finds, outlines and counts objects you name in plain words, on photos or full-resolution orthomosaics. Each result is a draft a person checks, corrects and marks verified, with a sealed record of the changes (see the AI Detection paper). Omniris publishes no accuracy figure for counting solar components; the count is as good as the check.

Commissioning and handover

The commissioning thermal survey runs through the same workflow as any other: a temperature map, a graded anomaly list and findings that become the EPC's punch list, with the EPC's pilots and the owner's team working the same findings in their own roles. A delivery from a pilot to a client carries the findings filed in the delivered folders, with their history and evidence. The electrical side of commissioning — insulation, polarity and I-V curve results under IEC 62446-1 — stays in your test equipment's records.

As-built

The as-built map, elevation models and point cloud export as GeoTIFF, LAZ and CAD files or publish to ArcGIS. A drone survey is not a licensed land survey; where a stamped survey is required, a licensed surveyor is still needed.

Storms, hail and insurance

Hail is about 2% of solar insurance claims but more than half of claim costs, with utility-scale hail claims averaging about $58 million.[8] After an event, the owner needs to know quickly how many modules are broken and where, and the insurer needs that count documented. Omniris supports the documentation; the assessment remains a person's.

• Before and after. If the site was flown before the storm, the Site Navigator's Compare puts the two dates side by side, as a fade, a swipe curtain or a blink, for any layer — colour or temperature.
• Damage drawn or detected. Draw findings directly on the photographs or the orthomosaic; a shape drawn on an angled photo is placed on the ground for you, with a stated margin of error. AI detection can be asked to look for damage in plain words, but condition words ("broken", "cracked") are the weakest kind of prompt and the results need close checking by a person before they count.
• Dead modules in thermal. A module that has stopped producing runs warmer than its neighbours,[3] and the anomaly search finds and ranks it like any other.
• A report for someone who will never log in. The situation report is one PDF for the site between two dates: the area, the two dates side by side, the findings in the area, and for every image where it came from, its capture date, its custody history and fingerprint, and a certificate anyone can verify.
• Held for the claim. A legal hold on the site, the job or the folder stops the photos, maps, reports and findings from being changed or deleted for as long as the claim runs; signed certificates show a file is the one the report was made from.

Accuracy and standards

Omniris follows the published standards where it uses them, says which ones, and does not claim that a flight or a report complies with them. Whether an inspection meets IEC TS 62446-3 depends on the conditions on the day, the camera, the flight and the qualifications of the person evaluating it.

Reading a temperature map with care

• Differences are more reliable than absolutes. The tools lead with ΔT for that reason, and the Solar site type compares each spot with its own neighbours.
• A cell smaller than a pixel is averaged. When a map pixel covers more ground than a cell, the pixel's temperature mixes the hot cell with what surrounds it, and the rise is understated. Fly lower for cell-level work.
• Large maps. Anomaly search on a very large temperature map works from a reduced-resolution read and says so; so do region statistics on a wide outline. Single-pixel extremes may be missed in either; draw a smaller outline for a full-resolution reading.
• Anomalies are contrast, not causes. Cells overheated by bird droppings or other local shading look much like cells overheated by internal defects; the colour map and a site visit tell them apart.[3]

Evidence and records

Solar inspection results end up in warranty claims against module suppliers, in disputes with EPCs over punch-list items, in insurance claims and in due diligence when a plant is sold. In each case the question is whether the file in front of you is the one that was produced on the day. Omniris keeps every file in a vault with a tamper-evident custody record, described in the Evidence Integrity paper:

• Sealed from creation. The temperature map, the orthomosaic, the elevation models and the report each get a custody record and a file fingerprint (a SHA-256 hash) when they are created, linked to the job that made them. The temperature settings used for the conversion are saved with the job.
• Findings too. Every step of a finding — created, assigned, commented, given evidence, resolved — is sealed in the custody record, naming the person, automatic run or API token that did it. A signed finding certificate lists the complete history and the fingerprints of its evidence, and can be checked online or offline.
• Legal holds and retention. A hold on a file, folder, job or site stops anything under it from being changed or deleted; a retention policy keeps a folder's or site's files for a set time and then releases them, or disposes of them with a signed certificate.

The custody record proves what happened to a file in Omniris. It does not prove that the inspection conditions were met on the day; the report's conditions note and the recorded weather are there so a reader can judge that.

Getting the flights flown

Ordering the work (cloud and data center editions)

In the Omniris cloud and data center editions, an owner or O&M manager can order a flight as a job on the Omniris marketplace. Marketplace, payments and payouts are not part of on-premises installations, where the same processing and analysis run on flights your own team uploads.

• Define the job. Draw the plant on the map, mark no-fly areas (a substation, a neighbouring property), choose deliverables — orthomosaic, thermal, 3D model and others — and the dates. Drop numbered pins with notes where you want specific ground-level photos, such as an inverter pad.
• Choose a pilot. Set a fixed price or ask pilots for proposals, send the job directly to a pilot who has delivered for you before, or let approved pilots whose travel radius covers the site be notified. A browsing pilot sees only the city and an approximate pin until assigned.
• Make it recurring. A job can repeat weekly, every two weeks, monthly, quarterly, or every N days, weeks or months — quarterly thermal flights, or weekly construction flights — for a fixed number of occurrences or indefinitely. Each occurrence has its own flight window and is billed when it is completed.
• Delivery. The pilot delivers results into the job's folder; the client accepts or rejects them with notes, and the files and findings arrive in the client's vault with their custody records.

Planning the flight

The mission planner, described in the Mission Planning paper, turns an area into a flight plan: a grid for a straight-down map, a crosshatch, corridor, perimeter or orbit for 3D, with overlap, speed and motion blur worked out from the camera, optional terrain following, and no-fly zones enforced. It exports to DJI Pilot 2, DJI Fly, Litchi, QGroundControl and others; Omniris writes the file, and the pilot flies it in the aircraft's own app. For a thermal survey, the planner sizes ground detail from the visible camera, so set the height from the thermal camera's specification.

Airspace, weather and the checklist

• Airspace. The whole route is checked against FAA airspace at the height it will actually be flown, including the floors of controlled airspace, and Omniris works out whether a LAANC authorization is needed.
• Weather. A flyability score for every hour up to 16 days ahead, against the aircraft's limits, including wind and cloud cover, with the best window for the job. Irradiance is not part of the forecast; a clear-sky window is the closest proxy for the 600 W/m² the inspection needs.
• Pre-flight. A Part 107 checklist that cannot skip the regulatory items, saved with the job and frozen once complete.

A marketplace job can require a drone on the Blue UAS list: a pilot with no aircraft on that list among the aircraft registered to their account cannot apply for it. Federally funded projects should still confirm that the aircraft flown is not from a covered foreign entity under the American Security Drone Act.[15]

Integrations and delivery

Site boundaries can carry custom fields — text, whole numbers or decimals, for the whole site or per flight date — so a block polygon can hold its inverter number, capacity or an inspection score you track from flight to flight. Values are entered by hand. Sites, boundaries and the Site Navigator are described in the Site Management paper.

Cloud, data center and on-premises

Omniris runs in four ways, with the same thermal processing, anomaly search, findings, reports and custody records in each:

A self-hosted installation with Hybrid enabled can send a thermal stitch to the cloud when its own machines are busy or too small; the result lands in the local vault like any other run. Utilities and owners whose plants count as critical infrastructure can keep imagery of them on their own servers.

What the results are, and what they are not

A solar inspection result is used for repairs, warranty claims and sales, so Omniris states what its numbers mean where they appear:

• Class grades are indicative. Plane-of-array irradiance is not recorded with the flight, so measured temperature differences are graded as flown, not projected to 1,000 W/m², and the panel, the report and Orion label every grade indicative. Each anomaly is graded on its temperature rise, not pattern by pattern against the standard's full abnormality matrix.
• Anomalies are temperature contrast, not diagnoses. The search finds and ranks warm areas; whether a hot cell is a defect, soiling or shading, and what it costs in energy, is the inspector's call, helped by the colour map and a site visit.
• Locations are map positions. Anomalies and findings carry coordinates, an outline and a stated position uncertainty, not a string or module ID or serial number; on densely packed rows, confirm the module on site.
• Temperature maps come from straight-down photos. Angled, tracker-following flights become a 3D thermal skin to view, not a temperature map to measure or search.
• Not a certificate or a licensed survey. Omniris does not certify that an inspection meets IEC TS 62446-3 or IEC 62446-1, and does not qualify thermographers. Electrical commissioning results stay in your test equipment's records, and a drone survey is not a licensed land survey.
• The pilot flies. Omniris plans missions and connects owners with pilots; the pilot flies the aircraft. A Blue UAS requirement is checked against the aircraft registered to a pilot's account, not against the aircraft actually flown.

Sources

Industry facts in this paper come from the sources below, read in October 2026. Omniris product behaviour is described from the product itself.

1. International Electrotechnical Commission. IEC TS 62446-3:2017, Photovoltaic (PV) systems — Requirements for testing, documentation and maintenance — Part 3: Photovoltaic modules and plants — Outdoor infrared thermography, Edition 1.0, June 2017, the current edition (clauses 4–8 and Annexes B and C cited; publisher's preview of the scope and contents linked). vde-verlag.de (PDF)
2. International Electrotechnical Commission. IEC 62446-1:2016, Photovoltaic (PV) systems — Requirements for testing, documentation and maintenance — Part 1: Grid connected systems — Documentation, commissioning tests and inspection, Edition 1.0, January 2016 (contents, publisher's preview). vde-verlag.de (PDF)
3. IEA PVPS Task 13. Review on Infrared and Electroluminescence Imaging for PV Field Applications, Report IEA-PVPS T13-10:2018, March 2018. iea-pvps.org (PDF)
4. Tsanakas, I. (CEA-INES). Aerial Inspections and Diagnostics of PV Power Plants, IEA PVPS Task 13 Focus Workshop on Operation & Maintenance, Fraunhofer ISE, September 30, 2021. iea-pvps.org (PDF)
5. Walker, H. A. Best Practices for Operation and Maintenance of Photovoltaic and Energy Storage Systems; 3rd Edition, NREL/TP-7A40-73822, National Renewable Energy Laboratory, December 2018 (§9.1 DC Array Inspection; model statement of work). docs.nlr.gov (PDF)
6. Walker, A., et al. Model of Operation-and-Maintenance Costs for Photovoltaic Systems, NREL/TP-5C00-74840, National Renewable Energy Laboratory, June 2020. docs.nlr.gov (PDF)
7. Ng-Cordell, L., et al. A Snapshot of an Emerging Industry: Aerial Inspections of Utility Scale Solar Plants, UCLA Institute of the Environment and Sustainability, August 2018. ioes.ucla.edu (PDF)
8. pv magazine. "New tool aims to cut costly hail losses in utility-scale solar", April 22, 2025 (VDE Americas figures). pv-magazine.com (page)
9. PCL Construction. "Leveraging AI and drones for tracking solar project progress" (case study; vendor-reported). pcl.com (page)
10. McCall, J., Macdonald, J., Burton, R., Macknick, J. "Vegetation Management Cost and Maintenance Implications of Different Ground Covers at Utility-Scale Solar Sites", Sustainability 15(7), 5895, 2023. doi.org (page)
11. Federal Communications Commission, Public Safety and Homeland Security Bureau. Public Notice DA 25-1086, December 22, 2025: addition of UAS and UAS critical components produced abroad to the FCC Covered List. docs.fcc.gov (PDF)
12. Federal Communications Commission, Public Safety and Homeland Security Bureau. Public Notice DA 26-22, January 7, 2026: exemption of certain UAS and UAS critical components from the Covered List. docs.fcc.gov (PDF)
13. Federal Communications Commission. Public Notice DA 26-832, August 10, 2026: comment sought on prohibiting importation and marketing of certain covered UAS and components. docs.fcc.gov (PDF)
14. Federal Communications Commission, Public Safety and Homeland Security Bureau. Public Notice DA 26-1063, October 2, 2026: conditional approval and exemption of certain UAS, routers and advanced robotic devices from the Covered List (summarising the UAS exemptions in force, including the July 21, 2026 extension to January 1, 2028). docs.fcc.gov (PDF)
15. U.S. Department of Energy. Acquisition Letter AL 2026-01 / FAL 2026-04 / OTL 2026-02, Prohibition on Acquiring UAS from Covered Foreign Entities under Financial Assistance and Other Awards (American Security Drone Act of 2023, §§ 1822, 1825), January 2026. energy.gov (PDF)
16. 14 CFR § 107.51, Operating limitations for small unmanned aircraft. law.cornell.edu (page)

Glossary

Bypass diode

A diode in the module's junction box that lets current pass around a substring of cells that cannot carry it; when it switches on, that substring stops producing and warms up.

CoA (class of abnormality)

The IEC TS 62446-3 grading of a thermal anomaly: CoA 1 no abnormality, CoA 2 thermal abnormality, CoA 3 safety-relevant thermal abnormality.

EPC

Engineering, procurement and construction contractor: the company that builds the plant and hands it over.

I-V curve tracing

An electrical test that measures a string's current against its voltage to reveal faults; done at the combiner box.

Irradiance (plane-of-array)

The sunlight striking the panel surface, in watts per square metre (W/m²). IEC TS 62446-3 asks for at least 600 W/m² for a module inspection.

O&M

Operations and maintenance: running the plant day to day and repairing it.

PID

Potential-induced degradation: power loss caused by voltage stress between cells and frame, often seen as a patchwork of warmer cells.

Radiometric

A thermal camera that records measured heat at each pixel, from which temperature can be calculated.

String

A series-connected chain of modules feeding a combiner or inverter input.

Temperature map

A map-aligned image in which every pixel is a measured surface temperature (a thermal orthomosaic), labelled THERMAL in Omniris once verified.

ΔT

A temperature difference — here, how much hotter an anomaly is than the panels around it. Shown °F first and converted as a difference.