Automated Photovoltaic Module Quality Assessment: Defect Identification and Classification from Luminescence Images Using Machine Learning

To help address the important challenge of photovoltaic module end-of-life (EoL) management, an automated module quality assessment methodology to determine the preferred EoL paths of fielded photovoltaic modules using luminescence images has been developed. First, defects within the photovoltaic modules are identified and classified through feature extraction incorporating generative representation learning models. Second, the relative module power loss is estimated and the preferred EoL path is determined based on thresholds selected by the user. A method to calibrate module performance using the quality assessment results, and perform full module circuit simulations is then developed, and initial results using this novel approach are presented.