Applying Deep Learning Algorithms to EL-images for Predicting the Module Power

Buerhop-Lutz C, Hoffmann M, Reeb L, Pickel T, Hauch J, Maier A (2019)

Publication Language: English

Publication Type: Conference contribution, Abstract of lecture

Publication year: 2019

Pages Range: 858 - 863

Conference Proceedings Title: Proceedings of the 36th European Photovoltaic Solar Energy Conference and Exhibition

Event location: Marseille FR

ISBN: 3-936338-60-4

DOI: 10.4229/EUPVSEC20192019-4CO.1.2

Open Access Link: https://www.photovoltaic-conference.com/participation/publications-2/conference-proceedings.html

Abstract

EL-images nicely visualize the cell quality of PV-modules in terms of existing failures like cell cracks and cell fractures. Nonetheless, the benefit of EL-images recorded in the field increases if a direct prediction of the module power is enabled. For the first time, deep learning algorithms are applied to predict the module power using EL-images. The dataset consists of more than 600 EL-images from different crystalline PV module manufactures. Regarding the network architecture, ResNet18 was identified to perform best with respect to computational time and accuracy. The predicted powers are quite accurate: the mean absolute error over the entire data set is 4.57 W. In addition, 72% of the module powers are predicted with a high accuracy of less than 2% error. Finally, it should be emphasized that even the EL-images recorded in the field with perspective distortion and motion blur are suitable for excellent predictions

Authors with CRIS profile

Mathis Hoffmann Lehrstuhl für Informatik 5 (Mustererkennung) Jens Hauch
Andreas Maier Lehrstuhl für Informatik 5 (Mustererkennung)

Related research project(s)

Laboranalyse von Degradationsmechanismen unter beschleunigter Alterung und Entwicklung geeigneter feldtauglicher bildgebender Detektionsverfahren und Entwicklung und Evaluation eines Algorithmus zur Fehlerdetektion und Prognostizierung der Ausfallwahrscheinlichkeit (iPV4.0) Aug. 1, 2018 - July 31, 2021 Analysis of Defects on Solar Power Cells Laboranalyse von Degradationsmechanismen unter beschleunigter Alterung und Entwicklung geeigneter feldtauglicher bildgebender Detektionsverfahren und Entwicklung und Evaluation eines Algorithmus zur Fehlerdetektion und Prognostizierung der Ausfallwahrscheinlichkeit (iPV4.0) Aug. 1, 2018 - July 31, 2021

Involved external institutions

Helmholtz-Institut Erlangen-Nürnberg für Erneuerbare Energien (HI-ERN) DE Germany (DE)

How to cite

APA:

Buerhop-Lutz, C., Hoffmann, M., Reeb, L., Pickel, T., Hauch, J., & Maier, A. (2019, October). Applying Deep Learning Algorithms to EL-images for Predicting the Module Power. Paper presentation at 36th European Photovoltaic Solar Energy Conference and Exhibition, Marseille, FR.

MLA:

Buerhop-Lutz, Claudia, et al. "Applying Deep Learning Algorithms to EL-images for Predicting the Module Power." Presented at 36th European Photovoltaic Solar Energy Conference and Exhibition, Marseille 2019.

BibTeX: Download