PV module power prediction by deep learning on electroluminescence images - Assessing the physics learned by a convolutional neural network
Source abstract
Abstract
Fast and reliable performance monitoring of photovoltaic modules is essential for economic forecasting in large-scale installations. Deep Learning methods, such as convolutional neural networks, have the potential to predict module power directly from electroluminescence images in an automated workflow. However, neural networks must be trained using large numbers of electroluminescence images of defective modules. Due to budgetary or technical limitations, these training images will always be biased, limiting generalization. Here, we demonstrate a transparent method to discriminate the information learned by a convolutional neural network into generally valid physics and bias. Learning of physics is assessed by providing an artificial, unbiased feature list, which is converted into synthetic electroluminescence images. Using these images, we compare the predictions of the neural network trained on the biased dataset to those of a physics-based equivalent. Bias is assessed by a closer look at the deviations between the predictions from the equivalent circuit model and the trained neural network. The assessment of physics knowledge incorporated by a Deep Learning method gives insight into how the method achieves its predictive capacity.
Source: expertdata · Quality: candidate unverified · Source record ↗
Research context
OpenAlex topics: Photovoltaic System Optimization Techniques · Solar Radiation and Photovoltaics · Solar Thermal and Photovoltaic Systems
SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.
References (25)
Show source-provided references
1. Drone-based daylight electroluminescence imaging of PV modules10.1109/JPHOTOV.2020.2978068 ↗
2. Deep learning-based pipeline for module power prediction from EL measurements10.1002/pip.3416 ↗
3. Automated pipeline for photovoltaic module electroluminescence image processing and degradation feature classificatio10.1109/JPHOTOV.2019.2920732 ↗
4. Fault Detection in aerial images of photovoltaic modules based on Deep learning10.1088/1757-899X/1012/1/012030 ↗
5. Pa M, Kazemi A, A Fault Detection Scheme Utilizing Convolutional Neural Network for PV Solar Panels with High Accuracy https://doi.org/10.48550/arXiv.2210.09226.
6. Deep convolutional neural network for automatic detection of damaged photovoltaic cells
7. Outdoor luminescence imaging of field-deployed PV modules10.1088/2516-1083/ac9a33 ↗
8. Module-power prediction from PL measurements using deep learning
9. Progress in the understanding of light- and elevated temperature-induced degradation in silicon solar cells: a review10.1002/pip.3362 ↗
10. Quantitative local current-voltage analysis and calculation of performance parameters of single solar cells in modules10.1016/j.solmat.2016.08.029 ↗
11. Quantitative assessment of the influence of camera and parameter choice for outdoor electroluminescence investigations of silicon photovoltaic panels10.1515/zna-2019-0025 ↗
12. We follow here the statistical definition of bias as a “systematic tendency which causes differences between results and facts, whereby data selection may have been skewed by the collection criteria” (https://en.wikipedia.org/wiki/Bias_(statistics)). (We distinguish this definition from the one applied in behavioral science where bias is often used in a narrower sense linked to racism).
13. Rong · 2020
14. Daw · 2017
15. Quantitative evaluation of electroluminescence images of solar cells10.1002/pssr.200903304 ↗
16. Detection of the voltage distribution in photovoltaic module by electroluminescence imaging10.1002/pip.941 ↗
17. Quantitative prediction of power loss for damaged photovoltaic modules using electroluminescence10.3390/en11051172 ↗
18. Physics-based method for generating fully synthetic IV CurveTraining datasets for machine learning classification of PV failures10.3390/en15145085 ↗
19. Site-specific assessment of mechanical loads on photovoltaic modules from meteorological reanalysis data10.1016/j.solener.2019.06.077 ↗
20. Buerhop-Lutz · 2017 · 1554
21. Evolution of cell cracks in PV-modules under field and laboratory conditions10.1002/pip.2975 ↗
22. He · 2015
23. Influence of cracks on the local current–voltage parameters of silicon solar cells10.1002/pip.2443 ↗
24. Quaschning · 1996
25. The risk of power loss in crystalline silicon based photovoltaic modules due to micro-cracks10.1016/j.solmat.2010.10.034 ↗