Lebensdauer- und Ertrags-Analysen von PV-Modulen und -Systemen mittels Leistungsbewertung und bildgebender Verfahren (ReliaREN-Pro-D)

Abstract

The ReliaREN-Pro-D project aimed to enhance the reliability assessment of thin-film photovoltaic (PV) systems (PV parks) through advanced imaging techniques and automated analysis methods. The German consortium, comprising Kiwa PI Berlin AG, Forschungszentrum Jülich, and Aerial PV Inspection GmbH, focused on developing and automating electroluminescence (EL) and infrared (IR) imaging analysis for thin-film technologies, particularly CdTe modules. The project employed drone-based high-throughput inspection methods to capture imaging data from multiple PV installations in Italy and France. A key innovation was the development of an AI-powered image processing pipeline using generative neural networks for automated module and cell detection, capable of processing over 105,000 images with high accuracy. The system achieved processing speeds of 0.72 seconds per image on GPU, representing a significant improvement over existing methods. Laboratory investigations revealed critical findings regarding seasonal performance variations in thin-film modules. The research demonstrated that stabilization methods prescribed by manufacturers may not accurately reflect long-term field performance, leading to discrepancies between laboratory measurements and actual field output. Temperature coefficients were found to vary seasonally and by location, challenging current standardization practices.