An Automated Framework for Drone-based Solar Panel Soiling Detection

The installation of both residential and commercial photovoltaic cells (PVs) has been growing rapidly. In either situation maximising the efficiency of the PVs is critical this can be hindered by soiling, cracks, and other defects. Detecting such faults through regular inspections is an essential yet costly task for small and large scale PVs deployments. The research presented in this paper proposes a drone-based system that navigates a PVs installation, capturing an image and the location of the panels. Machine Learning is then used to detect any instances of soiling or visual faults, these locations can then be used by a cleaning robot or a maintenance team depending on the type of damage or soiling. The proposed approach is evaluated using real deployed PV cell arrays, demonstrating it as an effective solution for detecting PV soiling.