Extracting Cell Images from Ultraviolet Fluorescence Images of Photovoltaic Modules

Abstract

With the expanse of large-scale photovoltaic power plants, the need for automatized and cost-effective inspection methods keeps increasing. Aerial PV inspection using drones and optical methods has become very popular. Ultraviolet Fluorescence (UVF) is an effective inspection technique that detects faults on the encapsulant of the PV cell. The encapsulant is a polymer material called Ethylene-Vinyl Acetate (EVA). The ageing of the PV module/cell, notably oxygen and humidity entering can create fluorescence in the encapsulant. The fluorescence pattern can indicate both the ageing process and encapsulant defects. In this paper, a low-cost experimental setup was used to acquire UVF images in the state of Santa Catarina, Southern Brazil. An automatized image-processing pipeline has been developed using contour detection, perspective correction and cropping of module and cell images. In the future, databases of UVF cell images can be created and used to for Deep Learning applications for automatized fault detection.