A Plant Down 8% Looks Exactly Like a Plant Working Fine.
A rooftop plant losing eight percent looks identical to one performing correctly. Nothing is visibly broken, no alarm has fired, and the monitoring portal reports the inverters as running. The loss shows up only against what the plant should be generating — which is why it so often persists for years.
Inverter-level data tells you a string is underperforming. It does not tell you which module, or why. Thermography does, because a failing cell dissipates energy as heat instead of exporting it.
What the thermal signature actually distinguishes
| Pattern | Typical cause |
|---|---|
| Single hot cell | Cell damage, crack, or localised shading/soiling — a hot spot, and a fire risk if severe |
| A whole substring warmer than the rest | Bypass diode conducting — commonly a failed diode |
| An entire module uniformly warm | Open circuit, or a connection fault at the junction box |
| A whole string cool and inactive | String disconnection — a wiring or fuse fault, not a module fault |
| Patchy warmth following a pattern | Soiling, bird fouling, or shading — cleanable, not a defect |
That last row matters commercially: it separates a maintenance action from a warranty claim, and the two have very different owners.
Why aerial rather than handheld
Coverage. A ground survey with a handheld camera on a large array is slow, requires roof access to every row, and is done at an angle that produces poor thermal readings. An aerial survey covers the whole plant in a single flight, at a consistent angle and time of day, which is what makes the images comparable across the array and across years.
The comparability is the underrated part. A survey repeated annually turns a snapshot into a trend, and a module degrading slowly is visible long before it fails.
Conditions the survey depends on
Thermography needs sufficient irradiance to load the modules — a survey flown under cloud produces images with no useful contrast. It needs the plant generating, not shut down. And it wants low wind, since convective cooling flattens the very differences being measured. A survey flown in the wrong conditions produces a clean report and a false sense of health.
Aerial operations near industrial sites also carry airspace and permission requirements that need checking before mobilisation rather than on the day.
Where it fits in an O&M programme
Annually is a reasonable default for an industrial rooftop, and worth doing at the end of the first year while the workmanship warranty still has time to run — installation defects surface early. Also worth doing after hail, a severe storm, or any work that involved walking on the array. Related: why industrial solar underperforms.
What we do differently
Thermographic inspection is part of our solar AMC scope rather than an extra, and findings are separated into cleanable, repairable and warranty-claimable — because those three go to different budgets and different counterparties. See Solar EPC.
More insights
Rooftop Solar Payback in India: What Businesses Should Expect
For commercial and industrial rooftops, solar typically pays back in 3–5 years and then delivers free power for two decades. Here's how the maths works.
Solar EnergyThree Vendors, One Roof, Three Different Generation Figures.
Solar yield estimates for the same roof routinely differ by double digits. The gap is rarely the panels — it is what each model assumed about shading, soiling, temperature and losses, and almost none of it is visible in the headline number.