Analysis of Solar Resources and Predictive Modeling for Photovoltaic System Sizing in Tinghir, Morocco

The authors of this document present an automatic pipeline that makes use of annotated images to locate faults in photovoltaic (PV) modules, which are defects that relate to their characteristics. In the suggested pipeline, it is extremely important to interpret JSON annotations to find the faulty modules and provide each module image with the binary health status label. In order to investigate the incidence of defects and the change of the module’s health status over the datasets, the labeled dataset is explored more thoroughly by applying a variety of statistical and visual methods. These include bar charts, heat maps, pie charts, etc. The results not only depict significant differences between the datasets but also suggest the pipeline’s potential to implement quality assurance at a higher level. The impulse and purpose behind this paper are to create a base for the learning of machine learning models that are capable of identifying the defects that are generated in photovoltaic (PV) systems without the intervention of the human factor.