> Details

Hoffmann, Mathis [Author]; Buerhop-Lutz, Claudia [Author]; Christlein, V. [Author]; Reeb, Luca [Author]; Pickel, Tobias [Author]; Winkler, Thilo [Author]; Doll, Bernd [Author]; Wuerfl, T. [Author]; Peters, Ian Marius [Author]; Brabec, Christoph [Author]; Maier, A. [Author]

Deep-learning-based pipeline for module power prediction from electroluminescense measurements

Reference
management

Bookmarks

Remove from
bookmarks

Share this on Whatsapp

Close

Bookmarks



You can manage bookmarks using lists, please log in to your user account for this.
  • Media type: E-Article
  • Title: Deep-learning-based pipeline for module power prediction from electroluminescense measurements
  • Contributor: Hoffmann, Mathis [Author]; Buerhop-Lutz, Claudia [Author]; Christlein, V. [Author]; Reeb, Luca [Author]; Pickel, Tobias [Author]; Winkler, Thilo [Author]; Doll, Bernd [Author]; Wuerfl, T. [Author]; Peters, Ian Marius [Author]; Brabec, Christoph [Author]; Maier, A. [Author]
  • Published: Wiley, 2021
  • Published in: Progress in photovoltaics 29(8), 920-935 (2021). doi:10.1002/pip.3416 ; Seminar, Deutschland
  • Language: English
  • DOI: https://doi.org/10.1002/pip.3416
  • ISSN: 1062-7995; 1099-159X
  • Origination:
  • Footnote: Diese Datenquelle enthält auch Bestandsnachweise, die nicht zu einem Volltext führen.
  • Description: Automated inspection plays an important role in monitoring large-scalephotovoltaic power plants. Commonly, electroluminescense measurementsare used to identify various types of defects on solar modules but have notbeen used to determine the power of a module. However, knowledge of thepower at maximum power point is important as well, since drops in thepower of a single module can affect the performance of an entire string.By now, this is commonly determined by measurements that require todiscontact or even dismount the module, rendering a regular inspectionof individual modules infeasible. In this work, we bridge the gap betweenelectroluminescense measurements and the power determination of a module. We compile a large dataset of 719 electroluminescense measurementsof modules at various stages of degradation, especially cell cracks andfractures, and the corresponding power at maximum power point. Here,we focus on inactive regions and cracks as the predominant type of defect.We set up a baseline regression model to predict the power from electroluminescense measurements with a mean absolute error of 9.0 ± 3.7W (4.0 ± 8.4 %). Then, we show that deep-learning can be used to traina model that performs significantly better (7.3±2.7W or3.2±6.5 %).With this work, we aim to open a new research topic. Therefore, wepublicly release the dataset, the code and trained models to empowerother researchers to compare against our results. Finally, we present athorough evaluation of certain boundary conditions like the dataset sizeand an automated preprocessing pipeline for on-site measurements showingmultiple modules at once.
  • Access State: Open Access