Abstract
Daylight Photoluminescence (DPL) is a relatively novel imaging technique utilized in photovoltaic (PV) system inspection, using the working principle of photoluminescence (PL) with the sun as excitation source. Filtering the luminescence signal from the strong sun irradiation is its main challenge. Images acquired at different working points (WPs) of the module, namely at open circuit (OC) and at high current (HC) such as maximum power or short circuit, allow subtraction of the background radiation while maintaining the luminescence signal. Synchronization of image acquisition and WP switching becomes particularly challenging if the camera is applied to unmanned aerial vehicles (UAVs). To overcome this challenge, an algorithm is developed to identify WP switches in a set of images taken in the field by investigating the development of image intensities. For each image, its intensity is compared to several previous images. If the variation exceeds a threshold, this step is considered as a WP switch. The identified switches are then collated to the expected switching time, considering imaging and WP switching speed. The accuracy of this method, thus, how clearly a WP is defined within, depends on imaging speed as well as on image stability. The imaging speed defines possible changes in irradiance throughout the imaging process and its possibility to cover up the WP switch. A strong improvement in imaging speed could be reached by developing a PL-ready inverter in collaboration with HUAWEI within TRUST-PV project. This inverter can toggle connected PV strings between pre-set WP values and for chosen times. This does not only allow fast toggling, but as well avoids the need of additional equipment to alter WPs of PV modules. The advantages of the PL-ready inverter will be discussed as well as best settings and necessary ambient conditions (minimum irradiance, cloud level). The image stability impacts the detection algorithm, as a pixel-wise comparison of the image sets is performed. Thus, ground-based measurements conducted with a tripod allow a direct comparison of the same pixels within each image. Instead, application of DPL on UAVs will result in fluctuations of the camera position and a direct overlap of images is not possible anymore. Image processing algorithms using OpenCV library will be used to do accurate module detection within each image and test the applicability of the developed algorithm on UAV based inspections. This will be done for ground-based simulations (with moving camera) as well as on images achieved during actual UAV based DPL inspections.