Monitoring tells you a string is underperforming. Thermal imaging tells you which module, and why. For a plant past its second year, it is the highest-value inspection available.
Cell-level hotspots. A cell that is shaded, cracked or mismatched becomes a resistive load rather than a generator, and heats up. Left alone, hotspots degrade encapsulant and eventually damage the module permanently.
Bypass diode failure. Shows as a whole substring – typically a third of a module – running noticeably warmer or producing nothing.
Connector and junction box heating. The important one. High-resistance connections generate heat well before they fail, and this is the primary early-warning route for DC arc risk. Nothing else finds it reliably.
Cracked cells. Often invisible to the eye, frequently caused by mishandling during installation or by walking on modules. They show as irregular warm regions.
PID-affected modules. Potential induced degradation tends to appear in a pattern related to string position and system voltage, and thermal patterns often reveal it.
This is where surveys go wrong, and a survey done in the wrong conditions is worse than none because it produces false confidence.
Practically, this means a clear day, mid-morning to mid-afternoon, outside the windiest months.
Drone surveys cover large arrays quickly and reach roofs safely. They suit ground mounts and big rooftops well. Resolution is the limitation – you need enough pixels on each module to resolve a single cell, which constrains flight height.
Handheld surveys give better resolution and let the inspector examine junction boxes and connectors directly, which drones generally cannot. On a commercial rooftop with accessible walkways, handheld often finds more.
A reasonable compromise on larger sites: drone sweep to locate anomalies, handheld follow-up to diagnose them.
Temperature differences are what matter, not absolute readings.
An initial survey at commissioning gives you a baseline and catches installation damage while it is still someone else's liability – that alone often justifies the cost. After that, every two to three years for a typical commercial plant, or annually where the array is large, critical, or in a harsh environment.
Always survey after a hail event, after any work on the array, and before a warranty period expires.
Georeferenced or clearly located anomalies – string and module position, not just a photograph. Both thermal and visible-light images of each finding. Temperature differential stated numerically. A severity classification with a recommended action and timeframe.
A folder of thermal photographs with no location referencing is not a survey. You cannot act on it, and I have seen more than one client pay for exactly that.