Solar Farm Drone Inspection: Thermal Mapping and Fault Validation

In This Guide

• What the thermal camera measures
• Define the job and supply the site inputs
• Set capture conditions around the smallest target
• Turn imagery into measurable deliverables
• Validate faults before assigning losses
• Questions to settle before commissioning

What the thermal camera measures

A thermal camera detects infrared radiation reaching its detector. Converting that signal into temperature requires corrections for the surface and surroundings. Emissivity describes how effectively a surface emits thermal radiation; reflected apparent temperature accounts for radiation reflected from other objects. Distance, humidity, and atmospheric temperature also affect the calculation, as explained in FLIR's thermographic measurement guidance.

That distinction matters on reflective module glass. An apparent hot or cold region may involve the surroundings as well as the module. Preserve the viewing geometry and visible photograph so the analyst can assess that possibility. Require a demonstration that the delivered files retain temperature measurements and their settings for later analysis; a screenshot of a color palette cannot substitute for that record.

Keep three conclusions separate in the report: observed thermal pattern, suspected cause, and confirmed finding. They may occupy different fields and have different confidence levels. The broader distinction between screening and diagnosis is explained in the six levels of drone-inspection evidence.

Define the job and supply the site inputs

Start the scope with the decision the inspection must support. Routine maintenance may need a prioritized list for field investigation. A commissioning baseline needs repeatable asset identification and retained original files. Troubleshooting an underperforming block needs operating records alongside imagery. A component claim needs the manufacturer's specified documentation and tests agreed before capture.

Use this intake list to assign responsibilities between the owner, flight team, and analyst:

• Asset layout: current row, table, module, string, combiner, and inverter identifiers; drawings that connect physical positions to electrical circuits; and known replacements or layout changes.
• Equipment details: module models and dimensions, mounting arrangement, tracker configuration, and relevant manufacturer inspection instructions.
• Operating context: time-stamped inverter or string data where available, alarms, planned outages, curtailment, and the intended operating state during capture.
• Site history: earlier findings, repairs, cleaning dates, vegetation work, and the locations of unresolved problems.
• Access and handoff: inspection boundary, exclusions, site contact, flight constraints, qualified electrical support, and the person who will receive urgent findings.

This is a proposed procurement checklist, not a claim that every site already has these records. Identify missing inputs before booking the flight. If the electrical drawings are unreliable, explicitly commission asset reconciliation or limit the report to physical locations. Do not silently attach an uncertain string identifier to a confidently located module.

Set capture conditions around the smallest target

The IEC TS 62446-3:2017 scope covers outdoor infrared inspection of operating PV modules and plants, including equipment, ambient conditions, procedures, reporting, and personnel qualifications. Ask the provider to identify the inspection method and document its application. A reference to the specification in a proposal does not establish that a particular dataset satisfies it.

IEA PVPS's 2021 mobile-test-equipment report, section 2.4, recommends stable illumination of at least 600 W/m² in the module plane and aerial infrared sampling no coarser than 3 cm per pixel. These are attributed planning recommendations; they do not guarantee detection of every defect. Record irradiance during collection and agree how unsuitable periods will be excluded or reflown.

Ground sample distance describes the surface distance represented by a pixel. Use the thermal detector and lens to plan it, not the RGB camera's resolution. FLIR's distance-measurement guidance recommends that a temperature target cover at least 3 × 3 pixels and overfill the measurement spot. Seeing a feature is therefore different from measuring its temperature reliably.

For scale, an assumed 6 cm-wide target at an assumed 3 cm/pixel spans only 2 pixels across: 6 ÷ 3 = 2. This illustrative calculation is not a module specification or a detection test. It shows why a site-wide mapping resolution needs a separate check against the smallest feature the provider promises to measure. Digital enlargement cannot add captured detail.

Require a short representative capture before committing the whole site. Have the analyst confirm focus, target sampling, identification, and temperature-file compatibility. The 2021 IEA PVPS report also highlights wind, shading, reflections, and changing conditions as inspection constraints. Log these alongside tracker position and plant operating state. Agree pause, restart, and reflight rules in advance rather than deciding after marginal imagery has been collected.

Turn imagery into measurable deliverables

A thermal orthomosaic joins overlapping images into a geographically organized overview. It is useful for navigation, but require access from each finding to its original thermal frame and corresponding visible image. Ask whether exported map values preserve calibrated temperature measurements or represent display colors, and how processing changes those values.

The following is a proposed buyer handoff specification. It translates the reporting scope in IEC TS 62446-3 and the localization and follow-up needs described in the IEA PVPS 2022 O&M guidelines, section 4.1.1, into checks an owner can request. It is not a standardized set of pass/fail limits.

Define coverage with an explicit denominator. For example, if 9,500 of an agreed 10,000 modules have usable inspection records, assessed coverage is 9,500 ÷ 10,000 × 100 = 95%. The other 500 remain unassessed, not healthy. These are hypothetical counts illustrating a reporting metric, not a detection-rate claim.

Also agree how duplicates, revisits, and replacement modules will be handled. A stable asset identifier should survive a second inspection even when the image filenames change. Specify data retention and export access so closing a work order does not depend on keeping one supplier's dashboard subscription indefinitely.

Validate faults before assigning losses

IEA PVPS recommends cross-checking aerial findings with ground observations and additional measurements, including current-voltage testing and electroluminescence imaging, to improve root-cause analysis and reduce false positives. Current-voltage, or I-V, testing characterizes electrical output across a range of operating points. Electroluminescence uses electrically stimulated light emission to investigate cell condition. These are complementary investigations, not extra labels an analyst can infer from a thermal color.

For each selected finding, use the same asset and finding IDs through this sequence:

1. Confirm the observation. Review the original thermal and visible frames, capture conditions, and any repeat views. Record whether the pattern is reproducible or ambiguous.
2. Choose a discriminating field check. Have the qualified PV technician select the visual or electrical investigation that can distinguish the suspected causes. Record the question the check is intended to answer.
3. Record the diagnosis and action. Keep unsupported causes provisional. Assign maintenance only with the appropriate technical assessment and site authorization.
4. Verify closure. Attach the intervention, date, follow-up measurement or image, remaining uncertainty, and responsible reviewer to the original finding.

Keep urgency separate from diagnostic confidence. A finding may need prompt technical attention while its cause remains uncertain. Agree the escalation route with the site's responsible technical team; do not let an automated confidence score decide whether to notify them.

Ask the inspection provider how it samples apparently normal assets as well as flagged ones. Checking only detected anomalies can help assess false alarms, but cannot establish how many faults the method missed. A claimed detection rate needs an independently checked reference population, defined fault classes, and stated capture conditions.

Do not convert a temperature difference directly into a universal percentage of lost production. The 2022 IEA PVPS report describes limits to quantitative yield-loss assessment from aerial infrared inspection. For a proposed energy-recovery estimate, require the electrical measurements or model, operating period, assumptions, and treatment of overlapping module, string, and inverter findings. Keep modeled loss, measured loss, and post-repair recovery separate.

Questions to settle before commissioning

Ask competing providers the same questions so the proposals describe comparable work:

• Which fault classes and smallest targets are included, and which are outside the inspection's capability?
• What conditions stop collection, who decides to resume, and who pays for required repeat visits?
• Can the owner inspect original radiometric files and export findings without a proprietary viewer?
• Who reviews automated classifications, and how are uncertain findings and missed-fault checks reported?
• Does the quoted scope include field validation, electrical testing, asset reconciliation, and closure visits, or only aerial capture and analysis?
• Who receives potentially urgent findings, and how soon must the provider communicate them?

Choose the scope that closes the owner's actual maintenance decision. For broad screening, a well-documented survey with an agreed follow-up plan may be sufficient. For a specific fault diagnosis or a quantified loss claim, commission the additional investigation at the outset. The practical test is whether a field team can move from a finding to the right asset, an appropriate check, and a recorded resolution.

Source notes

• IEC TS 62446-3:2017: official technical-specification summary establishing the scope of outdoor PV thermography; the public summary does not reproduce its full requirements.
• IEA PVPS, Qualification of PV Power Plants using Mobile Test Equipment (2021): institutional report, hosted by PV Lab Australia; sections 2.4–2.5 discuss aerial capture recommendations and measurement uncertainty.
• IEA PVPS, Guidelines for Operation and Maintenance of PV Power Plants in Different Climates (2022): institutional guidance on aerial findings, complementary testing, and limits to yield-loss assessment.
• FLIR, Thermographic measurement techniques: manufacturer documentation explaining emissivity, reflected apparent temperature, and atmospheric corrections.
• FLIR, How Far Can You Measure with a Thermal Camera? (2023): manufacturer guidance distinguishing visible targets from targets large enough for temperature measurement.

Last checked: September 6, 2026.