Crack Detection in Photovoltaic Panel Electroluminescence Image Using Matched Filter for Performance Loss Estimation

Abstract

The long-term use of renewable energy investments which have gained importance in recent years, can be realized by tracking errors and malfunctions. It is very important to determine the cracks, microcracks and flaws that occur especially during the production and use of solar energy panels. Errors in the panels can seriously affect the energy production performance of the installed system. According to the panel arrangement in the power plant, flaws in any panel not only affect the energy production of a single panel, but also reduce the production of the entire series to which it is connected. Cracks and faults in solar panels cannot be seen with the naked eye, but can be determined by electroluminescence images. However, detecting these errors requires special expertise and a long examination time. In this study, fractures, microcracks and connection errors are detected automatically from electroluminescence images taken from photovoltaic panels. Before analyzing the images taken with the electroluminescence device, some preprocessing is required. Since the images taken from the installed power plants are not standard in terms of layout, background removal, size and perspective correction are required. In the study, the panels are not examined as a whole, but cell-based examination is carried out. For this reason, the panels are divided into cells. By applying the Matched filter to the dividing cell images, the fractures and cracks are highlighted and the features are extracted. The obtained features are evaluated statistically and the cell status is determined. An accuracy of 98.6% is achieved in determining cell damage status. The losses in the power generation performance of the panel are calculated according to the status of all cells on the panel. The calculated values are compared with the measured I-V values. As a result, efficiency loss was calculated with an accuracy rate of 96.2%.