Real-world performance and module degradation impacts of robotic cleaning in PV systems

Robotic cleaning is increasingly adopted worldwide to mitigate soiling losses in utility-scale photovoltaic (PV) systems. However, long-duration, side-by-side field evaluations of different robotic cleaning technologies under real field conditions remain scarce. This study presents the first comprehensive comparative assessment of multiple cleaning robots operating under identical field conditions in a dry and dusty desert environment. In addition to extensive outdoor and indoor experiments, a key methodological contribution of this work is the development of an irradiance-independent, short-circuit current (Isc)-based protocol for robust benchmarking of cleaning efficiency in outdoor environments. At the system level, the analysis considers the combined effects of robot design, module type (monofacial and bifacial), and cleaning frequency (daily, weekly, and no cleaning) to evaluate cleaning efficiency, operational reliability, electrical performance (Pmax), electroluminescence (EL) response, anti-reflective coating (ARC) characteristics, and module vibrations during cleaning. Results show that all robots achieved high cleaning efficiency with cleanliness restored to >99% under daily cleaning schedules, whereas extended intervals lead to residual soiling (cleanliness reduced to <95%). At module level, systematic cumulative Pmax degradation was observed in monofacial modules over time, while no consistent Pmax impact was identified across robot types. In contrast, bifacial glass-glass modules remained relatively stable. Although bifacial electrical signatures suggested reduced shunt resistance under highly penetrative robot brushes, EL, reflectivity, and surface roughness measurements were inconclusive. Module vibration under robot cleaning was minor, much less than wind-induced vibration. Overall, the results indicate that robotic cleaning could be highly effective for soiling mitigation in desert environments similar to this study site, but technology-specific validation and multi-modal diagnostics to ensure long-term module reliability prior to large-scale deployment is recommended.