Thermal imaging of PV arrays: method and fault interpretation

Infrared thermography turns an invisible electrical fault into a visible temperature anomaly, which is why it has become the standard first-pass inspection method for large PV fleets. A cracked cell, a failed bypass diode, or a poor solder joint all dissipate energy as heat rather than converting it to electricity, and that heat shows up on a thermal camera before it shows up as a measurable yield loss in most monitoring systems. This article is for O&M teams and inspection contractors who need to run thermal surveys that produce actionable, standard-aligned reports rather than a folder of pretty infrared images with no diagnostic value. We cover the physics of what a thermal camera actually measures, the fault signatures that correspond to specific defect classes, the requirements of IEC TS 62446-3, and the practical trade-offs of drone-based versus handheld survey workflows.

What a thermal camera measures

An infrared camera measures radiant emittance from a surface and converts it to an apparent temperature using an assumed emissivity, typically 0.90-0.95 for glass and polymer backsheet surfaces. It does not measure electrical current directly; temperature differences arise because localized resistive or recombination losses convert electrical energy to heat at the fault location rather than useful output. Absolute temperature readings from handheld or drone thermal cameras carry meaningful uncertainty, commonly +/-2 degC or +/-2 percent of reading for uncalibrated radiometric cameras, and are strongly affected by reflected sky temperature, wind speed, and viewing angle. For this reason, fault classification should rely primarily on relative temperature difference (delta-T) between a suspect area and a known-healthy reference area on the same module or string, not on absolute temperature alone.

Conditions required for a valid survey

IEC TS 62446-3 specifies minimum irradiance of 600 W/m2 for a valid thermal survey, since fault-related temperature differentials scale with current flow and are difficult to distinguish from camera noise at low irradiance. Surveys should avoid periods of variable cloud cover (which cause transient heating/cooling unrelated to faults), avoid high wind speeds above roughly 5-7 m/s (which convectively cool hotspots and mask smaller anomalies), and be conducted with the array under normal operating load, not open-circuited, since some fault types (e.g. cell cracks) only manifest a clear thermal signature under current flow.

Fault signatures

Single hotspot cells

An isolated cell running several degrees warmer than its neighbors, often within an otherwise uniform module, typically indicates a cracked cell, a cell-to-cell soldering defect, or localized cell degradation. The severity is graded by delta-T: IEC TS 62446-3 defines indicative severity categories, commonly cited as delta-T of 3-9 degC (minor, monitor), 10-19 degC (moderate, schedule replacement), and 20 degC or more (severe, replace promptly, risk of accelerated degradation or fire in extreme cases).

Full-module heating

An entire module running uniformly warmer than surrounding modules in the same string usually points to a module-level issue such as delamination, internal circuit degradation, or a failed bypass diode conducting continuously even without shading. This differs from a shading-induced hotspot, which will move or disappear as the sun angle changes during the survey.

Substring or diode-pattern heating

A rectangular patch corresponding to exactly one-third (or one-half, depending on module design) of a module's cell area, with a relatively uniform but elevated temperature, is the classic signature of a failed or short-circuited bypass diode, since that substring is being forced into reverse-bias operation continuously rather than only during partial shading events.

String and combiner-level patterns

A hotspot at a connector, junction box, or combiner terminal, rather than on the module glass itself, generally indicates a poor electrical connection with elevated contact resistance - a loose MC4 connector, corroded terminal, or undersized crimp. These faults carry fire risk and are typically prioritized for immediate correction regardless of the delta-T reading, because contact resistance faults can escalate faster than cell-level defects.

Diffuse or patchy heating across many modules

Widespread, irregular warm patches across many modules in a zone, without the sharp boundaries of cell or diode faults, often indicates soiling or bird droppings creating localized shading and partial hotspotting rather than a manufacturing or electrical defect - visual cross-check against the same-time visible-light image usually confirms this.

Comparing survey delta-T against baseline

Worked example

A drone survey of a 2.1 MWp site flags 14 modules with cell-level anomalies out of 5,320 total modules. Delta-T readings: 6 modules at 5-8 degC (minor), 5 modules at 12-16 degC (moderate), 3 modules at 22-27 degC (severe). Under a risk-based O&M budget, the 3 severe modules are replaced within the current site visit (cost: 3 modules x approximately EUR 180 each plus labor), the 5 moderate modules are scheduled for the next quarterly visit, and the 6 minor anomalies are logged for re-inspection at the next annual survey. This tiered approach, driven directly by the delta-T severity bands, avoids the cost of blanket module replacement while addressing the highest fire and yield risk first.

Drone-based workflows

Drone-mounted radiometric thermal cameras have largely replaced handheld ground-based surveys for utility-scale sites, since a single flight can cover 1-5 MWp per hour depending on flight altitude and camera resolution, versus a fraction of that for a ground technician walking rows. Standard practice pairs a radiometric thermal sensor with a co-registered visible-light camera, so each flagged thermal anomaly can be cross-referenced against a visible image for confirmation (distinguishing, for example, a bird-dropping shadow from a genuine cell fault). Flight altitude is a trade-off: lower altitude (10-15 m above the array) gives higher spatial resolution sufficient to resolve individual cell-level anomalies, but requires more flight time and more waypoints for full-site coverage; higher altitude (20-30 m) covers ground faster but may only reliably resolve module-level or larger anomalies. IEC TS 62446-3 provides guidance on minimum pixel resolution per module (commonly at least 6-8 pixels across the shorter module dimension) to reliably classify cell-level faults.

Data volume and automated analysis

A full-site drone survey of a utility-scale plant can generate tens of thousands of individual thermal images. Manual review at this scale is impractical, so most inspection providers now use automated anomaly detection software that flags candidate hotspots for human review, typically reducing manual review workload by 80-95 percent while requiring quality control sampling to catch false negatives from the detection algorithm.

Integrating thermal findings with other diagnostics

Thermal imaging identifies where a problem likely exists but not always its electrical root cause with certainty. Modules flagged with moderate or severe hotspots are typically confirmed with an IV curve trace before replacement is authorized under warranty, since manufacturers generally require electrical performance evidence alongside thermal images for a warranty claim. Persistent site-wide anomaly counts should also be cross-checked against the plant's performance ratio trend to confirm the thermal findings correspond to a measurable energy impact.

Common mistakes

• Conducting surveys below the 600 W/m2 irradiance threshold, producing delta-T readings too small to distinguish from camera noise.
• Relying on absolute temperature thresholds rather than delta-T relative to a healthy reference area on the same string.
• Flying drone surveys during high wind, which convectively suppresses smaller hotspots and causes under-reporting of genuine faults.
• Failing to cross-reference thermal anomalies against co-registered visible images, misclassifying shading or soiling as a cell-level electrical fault.
• Treating every flagged anomaly as requiring immediate module replacement, rather than applying a severity-tiered response that reserves urgent action for connector faults and severe cell hotspots.
• Skipping repeat surveys over time, missing the progression pattern that distinguishes a stable minor defect from one that is worsening toward failure.

Frequently asked questions

How often should a utility-scale plant undergo a full thermal survey?

Annual full-site surveys are common practice, with more frequent partial surveys (e.g. quarterly for a risk-based sample of previously flagged zones) on sites with known soiling, connector, or module-batch issues. Sites under active warranty claims investigation are often surveyed more frequently to document fault progression.

Can thermal imaging detect PID?

Not directly and not reliably on its own. PID typically manifests first as reduced power output and characteristic electroluminescence patterns rather than a strong thermal signature, so thermal surveys are a poor primary tool for PID detection; electroluminescence imaging or IV curve tracing with capacitance measurement is more appropriate.

Is rooftop residential PV worth thermal inspecting?

For small residential systems, the cost of a professional drone or handheld thermal survey often exceeds the value of the diagnostic information relative to simply monitoring string-level output and investigating on alarm. Thermal inspection becomes cost-effective at commercial scale and above, or when investigating a specific known performance complaint.

Does module glass reflectivity affect thermal readings?

Yes. Reflected sky or cloud temperature can create false warm or cool patches on glass surfaces, particularly at low viewing angles. Surveys are typically conducted with the camera close to normal incidence to the module plane to minimize this effect, and IEC TS 62446-3 recommends viewing angles within 5-60 degrees of normal depending on camera specification.

What credentials should a thermal inspection contractor hold?

Relevant qualifications include ASNT or equivalent thermography certification (commonly Level 1 or 2) for personnel, and appropriate national drone operator licensing for aerial survey providers. Neither guarantees IEC TS 62446-3 alignment on its own, so specifying the standard explicitly in the inspection scope of work is advisable.

Standards and references

IEC TS 62446-3: Photovoltaic (PV) systems - Requirements for testing, documentation and maintenance - Part 3: Photovoltaic modules and plants - Outdoor infrared thermography. IEC 60904-3: Photovoltaic devices - Part 3: Measurement principles for terrestrial photovoltaic solar devices with reference spectral irradiance data. Reference material: Fraunhofer ISE studies on infrared-based PV fault detection; NREL PV reliability and O&M best practice reports; IEA PVPS Task 13 reports on failure modes and reliability of PV systems. See also PV monitoring and performance ratio for how to prioritize which zones warrant a thermal survey, and the system sizing calculator for context on typical string and module counts referenced in the fleet-scale examples above.