Case Study: UAV Inspection at Solar Farm

TL;DR: This UAV inspection solar farm case study describes how ISS aerial-inspected a 180 MW utility-scale photovoltaic plant on the Darling Downs in Queensland, replacing a manual handheld-thermography campaign that would have taken three weeks. Using a radiometric thermal and high-resolution RGB payload flown under CASR Part 101, ISS surveyed roughly 410,000 modules across 360 hectares in two flying days, located 412 underperforming or faulty modules to GDA2020/MGA2020 Zone 56 coordinates, and delivered a georeferenced defect register that let the asset owner target O&M crews precisely rather than walking every row.

Key takeaways

• A radiometric UAV thermography survey covered 360 hectares and ~410,000 PV modules in two flying days, against an estimated 15 working days for a handheld IR camera campaign — an 85% reduction in inspection time and the removal of all hot-surface, working-near-DC manual handling.
• ISS flew DJI Matrice-class aircraft with a radiometric thermal sensor (<50 mK / 0.05 °C NETD) and a 45 MP mechanical-shutter RGB payload, holding a ground sampling distance of 3 cm/pixel thermal and 1 cm/pixel RGB at a 60 m above-ground-level flight height.
• 412 thermal anomalies were classified by IEC TS 62446-3 defect type — 168 single hot cells, 121 hotspot strings, 73 bypass-diode/substring faults, 28 fully disconnected modules, and 22 junction-box anomalies — each pinned to GDA2020 MGA Zone 56 coordinates and an inverter/string ID.
• Located defects represented roughly 0.10% of installed modules but a modelled 1.7 MW of lost capacity at the affected strings; targeted rectification recovered the bulk of it inside one O&M mobilisation.
• The work was delivered under a CASA Remote Operator Certificate (ReOC) with licensed RePL pilots, aviation-endorsed public liability cover, and full CASR Part 101 airspace management around the adjacent regional aerodrome.

The challenge

The client is the asset manager for a 180 MW single-axis-tracking solar farm commissioned in 2022 on the Darling Downs, west of Toowoomba. The plant sits on 360 hectares and carries about 410,000 bifacial modules wired into roughly 1,150 strings across 36 inverter stations. Two years into operation, the SCADA system was flagging a persistent shortfall against the modelled P50 yield — enough to matter financially, but spread thinly enough that string-level monitoring could not isolate the cause. Underperformance at the module level does not always show up at the inverter; a handful of hot cells, a failed bypass diode, or a string-stop fault can quietly erode output while the inverter still reports "healthy".

The conventional answer is handheld infrared thermography: a technician walks the rows with an IR camera during peak irradiance, photographing modules and noting faults. On a 360-hectare site that is a three-week campaign at best, it is weather- and time-of-day constrained, it puts people next to live DC strings and hot module surfaces in a Queensland summer, and it is inconsistent — coverage and image quality depend on which technician walked which row at what sun angle. Crucially, it produces photos, not coordinates. An O&M crew handed a folder of thermal images still has to find the faulty module physically, row by row.

The asset manager needed three things: complete coverage with no rows skipped, every defect located precisely enough that a crew could drive to it, and the whole survey finished inside a single short window so the data reflected one consistent plant state.

The approach

ISS scoped the job as a single-payload, two-aircraft UAV thermography survey designed around the physics of the measurement rather than the convenience of the flight. Aerial electroluminescence is not practical in daylight, so radiometric thermal imaging was the right tool: a defective cell, diode or string dissipates energy as heat under load, and that signature is unambiguous from the air when the plant is generating near capacity.

Three constraints shaped the flight plan. First, irradiance: IEC TS 62446-3 calls for thermography under at least 600 W/m² with stable, clear-sky conditions, so flying was confined to the four-to-five-hour window either side of solar noon. Second, the trackers move — single-axis tracking tilts the array through the day, so flight lines and the gimbal pitch were planned against the tracker angle at the time of each sortie to keep the sensor square to the glass and avoid sky reflection. Third, the adjacent regional aerodrome put part of the site under controlled-airspace considerations, managed through a CASR Part 101 airspace assessment before mobilisation.

ISS established a small ground-control framework first: photo-identifiable targets surveyed with a Leica GNSS receiver in GDA2020, MGA Zone 56, AHD heights, so the resulting thermal orthomosaic and every pinned defect would carry real coordinates rather than relative pixel positions. Automated grid missions were then flown at 60 m AGL, holding consistent overlap and a fixed nadir-to-glass geometry so every module across the site was imaged at the same scale and the same look angle. Two aircraft worked separate halves of the plant to compress the capture into the available irradiance windows.

Equipment used

The payload selection was driven by what the defects actually demanded — a thermal sensor sensitive enough to separate a single hot cell from ambient module temperature, paired with RGB fine enough to read junction-box labels and string IDs for the register.

Thermal capture was held at roughly 3 cm/pixel GSD and RGB at about 1 cm/pixel at 60 m AGL — fine enough to resolve a single 0.05 °C-elevated cell against its neighbours and to confirm, in the co-registered RGB frame, exactly which module and string the anomaly sat in. The thermal sensor's calibrated radiometric output meant defects could be classified by absolute temperature delta, not just visual contrast, which is what allows a hot cell to be separated from a hotspot string and a bypass-diode fault under IEC TS 62446-3.

The result

The full plant was captured in two flying days inside the solar-noon windows, against the 15-or-so working days a handheld campaign would have needed. Processing produced a single radiometric thermal orthomosaic of all 360 hectares tied to GDA2020 MGA Zone 56, with every anomaly pinned to coordinates and cross-referenced to its inverter station and string ID.

ISS identified and classified 412 thermal anomalies across the ~410,000 modules:

At roughly 0.10% of installed modules, the fault count was small in proportion but concentrated enough at specific strings to model around 1.7 MW of lost or at-risk capacity once string-stop and diode faults were weighted by their effect on the whole string. Every entry in the defect register carried a coordinate, a thermal image, the co-registered RGB frame, an absolute temperature delta, a severity rating, and the inverter/string reference — so the O&M contractor could route directly to each module instead of searching.

The outcome

Because the defects were located rather than merely photographed, the rectification campaign was planned as a single targeted O&M mobilisation: the crew drove a coordinate-ordered route, replaced or repaired the 28 dead modules and 73 diode/substring faults first (the highest-yield items), cleaned or swapped flagged hot-cell modules, and verified each fix against the register. The owner recovered the bulk of the modelled 1.7 MW inside that one mobilisation, and the georeferenced thermal orthomosaic became the baseline for the plant's annual condition monitoring — so next year's survey reports change against a fixed reference rather than starting cold.

The wider outcome was a shift in how the asset is inspected. Aerial radiometric thermography turned a three-week, weather-bound, people-near-DC manual task into a repeatable two-day survey with a coordinate-accurate output, at a fraction of the labour cost and with the safety risk of walking live strings in summer heat removed entirely. That is the recurring pattern across ISS UAV work: the aircraft is a remote-sensing tool, the value is in the georeferenced, standard-aligned data, and the payback typically lands on the first survey.

Frequently asked questions

How long does a UAV inspection of a utility-scale solar farm take?

For a 180 MW / 360-hectare site, ISS completed capture in two flying days, working only within the solar-noon irradiance windows that IEC TS 62446-3 requires. Total turnaround including the radiometric orthomosaic and classified defect register was about a week. A site of this scale would take roughly 15 working days by handheld thermography, weather permitting.

How accurate is drone thermography at finding individual faulty modules?

The radiometric thermal sensor resolves temperature differences below 0.05 °C (50 mK), which is enough to separate a single hot cell from its neighbours. Flown at 60 m AGL the thermal GSD is about 3 cm/pixel and the co-registered RGB is about 1 cm/pixel, so each anomaly is confirmed to the individual module and string. With GNSS ground control in GDA2020 MGA Zone 56, defects are located to coordinates a crew can navigate to directly.

What conditions are needed to fly a solar thermography survey?

Stable clear-sky irradiance above 600 W/m², low wind, and the plant generating near capacity so faults dissipate heat under load. On single-axis-tracking farms the flight lines and gimbal pitch are planned against the tracker angle to keep the sensor square to the glass. These constraints confine flying to the hours either side of solar noon.

Is the survey compliant with Australian aviation rules?

Yes. ISS operates under a CASA Remote Operator Certificate (ReOC) with licensed RePL pilots, registered aircraft and aviation-endorsed public liability insurance, and manages all CASR Part 101 airspace requirements — including the controlled-airspace assessment for the nearby regional aerodrome — before mobilisation.

Can the inspection be done while the solar farm is generating?

Not only can it be done while generating — it must be. The thermal signature of a faulty cell, diode or string only appears when the array is under electrical load near full output, so the plant stays online throughout. The survey is non-contact and removes O&M personnel from live DC strings and hot module surfaces entirely during data capture.

If you operate a solar farm, wind farm or any large distributed asset where manual inspection is slow, weather-bound or puts people near live equipment, a georeferenced UAV thermography survey is very likely the faster, safer and cheaper path — and the payback usually lands on the first survey. Tell ISS the site, the capacity and what you need located, and we will scope a fixed-price UAV inspection, recommend the right payload and deliverables, and manage every part of the CASA compliance. Call 0407 057 015 to request a quote. We are CASA-certified and operate across Australia.