A low-cost, data-driven, self-powered robotic system for intelligent PV panel maintenance in water-scarce regions

Soiling of photovoltaic (PV) modules remains a major obstacle to reliable solar energy generation, particularly in arid and semi-arid regions, where energy losses exceeding 40% have been reported, and global annual economic losses are estimated at €4–7 billion. Although autonomous cleaning robots have been proposed to mitigate these effects, most existing systems rely on fixed schedules or indirect sensing approaches, leading to inefficient operation, increased system cost, and limited adaptability to real-time performance degradation. In this work, a self-powered robotic cleaning system is presented that directly links cleaning actuation to in-situ PV performance degradation. The proposed platform employs a voltage-differential detection strategy based on dual 5 V reference panels, where a voltage drop of ≥ 2 V is used as a direct indicator of soiling-induced yield loss. This trigger is further supported using a particulate matter threshold (≥250 µg/m³) and a thermal safeguard (≤45°C), forming a multi-criteria decision framework that minimizes unnecessary cleaning cycles while protecting module integrity. The robot operates entirely off-grid using a dedicated 20 W solar panel and utilizes a rail-free, dry-brush mechanism for array traversal. Field testing over three months in a high-soiling desert environment demonstrated an average efficiency restoration of 86.7%, with power output increasing from 93 W to 173.7 W on a 180 W PV module, corresponding to a recovery of 93% of soiling-induced losses. The complete system is realized using commercially available components at a total cost of approximately $64, highlighting a practical and economically accessible solution for PV maintenance in water-scarce and distributed solar installations.