Influence of irradiance and drone altitude in infrared thermography inspections of photovoltaic plants

Abstract

This study evaluates the effectiveness of drone-based infrared thermography (IRT) for detecting photovoltaic (PV) module defects under varying drone altitudes and irradiance levels. A total of 43 PV modules were artificially degraded to induce common faults, including cell cracks, potential-induced degradation (PID), cell interconnect disconnections, glass cracks, and short-circuited bypass diodes. Additionally, modules affected by different types of soiling were included in the study. The modules were characterized and deployed in two operational PV strings at the Technical University of Denmark (DTU) PV plant. IRT images were acquired under irradiances ranging from 200 W/m2 to 1000 W/m2 and at four different drone altitudes (8 m, 10 m, 14 m, and 20 m), equivalent to module spatial resolution of <1 cm/pixel to 3 cm/pixel. The thermal signatures of each fault type were characterized under the different imaging conditions. Results show that modules with cell cracks are difficult to distinguish from healthy ones and disconnected interconnects display distinct elongated thermal patterns that remain detectable at a minimum irradiance of 600 W/m2. Modules affected by PID and glass cracks did not exhibit pronounced thermal anomalies under the tested conditions confirming that early-stage or low-severity cases of such defects remain challenging to identify using IRT alone. In contrast, short-circuited bypass diodes and heavy soiling caused prominent hotspots (80 °C to over 100 °C), which remained clearly detectable even at low irradiance levels (200 W/m2) and elevated drone altitudes. The analysis also highlights that the influence of drone altitude within the tested range is generally negligible compared to irradiance effects, although temperature decreases of approximately 5–10 °C were observed in the soiling scenarios at higher flight altitudes due to the reduced spatial resolution of the thermal images. To support further research, the complete IRT image dataset generated in this study is made publicly available.