Your Solar Array Passed Its Last Inspection. Here's What That Inspection Couldn't See.

A solar panel with a hotspot looks exactly like a healthy panel. The surface is intact. The glass is clean. The module is installed correctly and connected to the string. In a standard visual walkthrough inspection, it passes without comment.

Meanwhile, a single affected cell inside that module is forcing neighboring cells to dissipate excess power as heat rather than convert it to electricity. The panel is underperforming. Depending on the defect type, it may be degrading further with every sun-hour. And in cases where the temperature differential exceeds 20°C, it is a documented fire risk, still producing nothing visibly alarming to a person walking the row.

This is not an edge case. A thermography study covering more than 12 gigawatts of solar PV plants, more than 30 million modules assessed across five continents, found a median hotspot prevalence of 1.4% of installed modules, with some individual plants reaching 30%. Energy loss attributable to undetected hotspots ranges from 0.5% to 3% of total generation per affected system, according to the same dataset.

At utility scale, those percentages are not small numbers. And the panels causing them are not waving a flag.

What Visual Inspection Was Built to Find, and What It Wasn't

A competent visual inspection catches what's visually wrong: physical damage, broken glass, soiling visible from normal inspection distance, obvious connection problems, panels that have separated from their mounting. For a newly commissioned site, it confirms the installation looks correct. For an operating site, it identifies problems that have progressed far enough to show on the surface.

What it structurally cannot detect is a thermal anomaly beneath an intact-looking module surface.

A hotspot, a string fault, or potential-induced degradation don't announce themselves to the naked eye. They show up as temperature, a specific cell running hotter than its neighbors, a string running cooler because it's offline, a pattern of degradation spreading across a module that looks visually identical to a healthy one. Detecting these conditions requires capturing temperature, and capturing temperature across an entire array at meaningful resolution requires thermal imaging from above.

Visual inspection, however thorough, is answering a different question than the one thermal inspection answers. "Does this panel look damaged?" and "Is this panel performing correctly?" are not the same question.

What a Thermal Survey Actually Finds

Radiometric thermal imaging captures an actual temperature value at every pixel, not a general warm-versus-cool impression, but a quantified temperature reading at every point across the module surface. The resulting thermal map makes the following categories of defect visible in a way no other non-invasive inspection method can match.

Hotspots appear as bright warm zones in the thermogram. Causes include cracked cells, manufacturing defects, partial shading, or elevated internal resistance. A single affected cell can force heat dissipation across neighboring cells, creating a cascade effect that compounds energy loss and accelerates material degradation over time. The IEA-PVPS 2025 failure analysis of a 24.9 MW plant identified hotspots as the most common module abnormality, affecting approximately 1.05% of the 64,140 modules analyzed.

String faults appear as a distinctive dark row in the thermal image, a string or substring running cooler than neighbors because it's generating less power or has gone entirely offline. A single blown fuse can take 20 to 30 panels offline simultaneously, and the fault is invisible to ground-level visual inspection while immediately apparent in a thermal survey.

Potential-induced degradation (PID) develops gradually and silently over operating years. Studies have found PID-affected modules can lose 20 to 30% of their rated output within 2 to 5 years, a yield loss that would be immediately obvious if it appeared on day one, but develops quietly enough that it often goes undetected through multiple annual inspection cycles, according to analysis of deployed utility-scale systems.

Junction box and connection faults — where the thermal signature appears not on the module surface but at the electrical connection points, represent some of the most fire-prone fault types in an operating array. The 2025 IEA-PVPS failure fact sheet identified 350 junction box failure cases in a single plant dataset. Junction boxes running at elevated temperature are both a performance problem and a genuine safety risk.

The Sampling Problem

Most utility-scale solar inspection programs don't inspect every panel. They inspect a sample, commonly 10 to 25% of the array, and use that sample to infer the condition of the full installation.

The logic is economically understandable: at utility scale, a complete manual ground inspection of every module is prohibitively slow, and traditional single-operator drone thermal inspection of a large array took long enough that comprehensive coverage was practically difficult. Neither condition applies at the same degree anymore.

The problem with the sampling approach is the same one that affects every sampling methodology: it can only find defects in the portion of the array that was actually checked. The 75 to 90% of the array that a 10-25% sample leaves uninspected contains whatever it contains, including, with near certainty, some share of the hotspots and string faults present across the full installation.

A fleet of three to five drones can now inspect a 200 MW farm in a single day. Full-coverage thermal inspection is no longer a theoretical alternative to sampling, it's an operational reality at the scale where the sampling argument was strongest.

The Financial Stakes Across a 20-Year Asset Lifecycle

Solar projects are financed on 20- to 25-year project economics. A power purchase agreement prices anticipated energy output over that lifetime. Degradation that begins in years 1 to 3, from manufacturing-origin soldering defects, from PID establishing itself in the first operating seasons, from hotspots compounding cell damage each year they go unaddressed, shapes the entire project revenue curve from the outset.

A 2026 report analyzing factory audits across the US domestic solar manufacturing base found that 70% of American factories fell into the lowest quality tiers. Soldering flaws, cold solder joints, grid breaks, oversoldering, were among the most prevalent defect types identified through electroluminescence imaging, and field research indicates these flaws can reduce project output by 10 to 30% in affected areas.

These defects ship from the factory. They're present at installation. A pre-commissioning thermal survey, timed after installation but before the warranty period begins, is the opportunity to identify them while the installer is still responsible for the work. A defect discovered after the warranty period expires is the project owner's problem.

An annual survey after commissioning catches the defects that develop over operating time, PID establishing itself, mechanical stress producing cell cracks, connection degradation creating thermal anomalies that weren't present at year one. Each year a defect goes undetected is a year of compounding yield loss that cannot be recovered.

How CropCopters Would Execute This

Mission planning: Site layout, module count, string configuration, and inverter zones inform flight path design. Arrays are divided into inspection zones matched to drone battery and flight time parameters. Pre-commissioning and annual inspection are distinct mission types with different documentation objectives.

Data acquisition: The Zenmuse H30T on the M400 RTK platform delivers radiometric thermal imaging at the resolution required for cell-level anomaly detection across the full array. Radiometric capture means every pixel carries an actual temperature value, not a false-colour representation, which is what makes quantified anomaly detection possible. RGB imaging runs simultaneously for georeferenced visual documentation.

Timing matters: Thermal inspection of solar arrays is most effective when irradiance is high and modules are operating under load, typically between 10am and 2pm in clear conditions, when the temperature differential between healthy and defective cells is most pronounced. Inspections conducted under poor irradiance conditions produce less reliable thermal contrast and higher rates of false negatives.

Processing: DJI Terra processes thermal and RGB capture into georeferenced orthomosaics. Each module is identifiable by its position in the array layout. Thermal anomalies are extracted by temperature differential from the median module temperature in the same string or zone.

AI audit: AI-assisted classification screens the thermal dataset for hotspot patterns, string fault signatures, and junction box anomalies, flagging findings by defect type and temperature differential. Findings above the 20°C differential threshold, the level associated with direct safety risk, are flagged as priority. AI classification narrows what requires engineering review; a qualified solar O&M engineer confirms finding significance and determines repair priority.

Deliverable:

• Georeferenced thermal orthomosaic of the full array

• Anomaly report with each flagged module identified by location, defect type, and temperature differential

• Priority findings list sorted by severity, safety-risk (>20°C ΔT) flagged separately from performance-degrading findings

• String fault identification with affected panel count

• Pre-commissioning version: formatted as installer defect documentation for warranty remediation

• Annual version: year-over-year comparison with previous inspection data

What the operator can do next: For the first time, know which specific panels in the array are underperforming and why, not as an estimated proportion of a sampled section, but as a georeferenced inventory of every defective module in the installation, ranked by severity and ready for repair crew dispatch.

Technical Reality

What this does well: Identifies thermal anomalies across the full array in a single survey day at utility scale. Catches defects that are invisible to visual inspection and that sampling programs structurally miss in unchecked portions of the array. Provides a quantified, georeferenced record suitable for warranty claims, insurance documentation, and year-over-year performance comparison.

What this doesn't replace: Electrical testing, IV curve tracing and electroluminescence imaging catch defect types that thermal imaging can miss, including certain cell crack patterns that don't produce a thermal signature under normal operating conditions. A comprehensive O&M program typically combines thermal drone inspection with periodic electrical testing rather than treating either as a complete substitute for the other.

Conditions that matter: Thermal inspection requires adequate solar irradiance, overcast days produce insufficient temperature differential for reliable defect detection. Wind above certain thresholds can cool defective cells enough to suppress the thermal signature. Pre-inspection weather conditions and time-of-day planning are not optional details; they determine whether the resulting data is actually reliable.

Radiometric vs. visual thermal: Not all drone thermal inspection is equal. A radiometric sensor records actual temperature values at every pixel. A non-radiometric thermal camera produces a false-colour image that looks similar but cannot support quantified anomaly detection. Confirming that any thermal inspection service uses radiometric capture is a basic qualification question worth asking before engaging a provider.

The Human Story

The clean energy transition depends on solar assets performing close to their design specifications across their full operating lives. A utility-scale project with 3% undetected hotspot-driven yield loss isn't delivering the clean generation that financed it, the grid reliability that depends on contracted output, or the carbon displacement that justifies its land use and investment.

At the scale the industry is growing, the US alone projected to add 211 gigawatts of utility-scale capacity between 2026 and 2031, the aggregate impact of systematically undetected underperformance is not a marginal issue. It's the difference between the clean energy transition delivering what it's supposed to deliver and delivering somewhat less, year after year, in ways that don't show up in a visual walkthrough.

The ratepayers under clean energy contracts. The grid operators managing contracted capacity. The investors whose project returns depend on actual versus modelled performance. None of them benefit from a 10% or 20% yield loss that's been silently compounding since year two of a twenty-year asset.