An Integrated Framework for Anomaly Detection in Photovoltaic Systems Using UAV Multi-Sensor Data and SCADA Telemetry

This study proposes an integrated framework for photovoltaic (PV) anomaly diagnosis combining unmanned aerial vehicle (UAV) radiometric thermography and red–green–blue (RGB) imagery, supervisory control and data acquisition (SCADA) telemetry, meteorological measurements, solar-shading modelling, automated defect detection, and ground-level verification. The methodology was evaluated on a 26.73 kWp rooftop PV installation comprising 81 modules using 77 producing days and 3543 valid 30 min observations. UAV thermography identified an open-circuited string and shading-induced hotspots, with peak hotspot temperatures reaching 109.9 °C and temperature differences exceeding 60 °C relative to the surrounding module surface. Radiometric analysis showed uniform temperature elevations of up to 5.3 K for the open-circuited string. To isolate maintenance effects from meteorological variability, 376 pre- and post-intervention intervals were matched according to irradiance, ambient temperature, relative humidity, and wind speed. The analysis yielded a mean power increase of 0.518 kW (95% confidence interval: 0.321–0.714 kW), corresponding to 28.3% relative to the matched pre-intervention mean. Two of the four anomaly classes considered occurred on the installation (one open-circuited string and 14 shading-induced hotspot regions). Within the limits of this single-installation case study, the framework enabled thermal anomaly detection, root-cause discrimination, and quantitative assessment of PV performance changes, supporting condition-based maintenance and operational decision-making.