Analysis of Photovoltaic Cell Defect Classification Using Residual Neural Network
The photovoltaic (PV) system has seen significant advancements in recent years. Accurate defect classification is essential for maximizing the energy output of PV cells throughout their lifespan. With the rapid progress in deep learning and convolutional neural networks (CNN), PV images can now be l...
| id |
Assoli A.R.; Shahbudin S.; Yusof Y.W.M.; Zolkapli M. |
|---|---|
| spelling |
Assoli A.R.; Shahbudin S.; Yusof Y.W.M.; Zolkapli M. 2-s2.0-85219521304 Analysis of Photovoltaic Cell Defect Classification Using Residual Neural Network 2024 2024 IEEE 22nd Student Conference on Research and Development, SCOReD 2024 10.1109/SCOReD64708.2024.10872644 https://www.scopus.com/inward/record.uri?eid=2-s2.0-85219521304&doi=10.1109%2fSCOReD64708.2024.10872644&partnerID=40&md5=3d0a0ea6ce466a1417be5682ffe1b58a The photovoltaic (PV) system has seen significant advancements in recent years. Accurate defect classification is essential for maximizing the energy output of PV cells throughout their lifespan. With the rapid progress in deep learning and convolutional neural networks (CNN), PV images can now be leveraged for defect classification, offering improvements over traditional methods. This work proposes an analysis of PV cell defect classification using two different types of Residual Neural Networks (ResNet), namely ResNet-18 and ResNet-50, to determine which architecture delivers the best performance. The results are further compared with other CNN architectures, including deep CNN, GoogLeNet, and VGG-16. Various performance metrics were evaluated and compared across these models. The findings reveal that ResNet-18 outperforms the other architectures, achieving the highest performance with an accuracy of 97.96 %, specificity of 99.22%, sensitivity of 96.59%, precision of 97.42 %, and an F1-score of 96.80 %. This analysis will help to improve the defects classification system for PV cells. © 2024 IEEE. Institute of Electrical and Electronics Engineers Inc. English Conference paper |
| author |
2-s2.0-85219521304 |
| spellingShingle |
2-s2.0-85219521304 Analysis of Photovoltaic Cell Defect Classification Using Residual Neural Network |
| author_facet |
2-s2.0-85219521304 |
| author_sort |
2-s2.0-85219521304 |
| title |
Analysis of Photovoltaic Cell Defect Classification Using Residual Neural Network |
| title_short |
Analysis of Photovoltaic Cell Defect Classification Using Residual Neural Network |
| title_full |
Analysis of Photovoltaic Cell Defect Classification Using Residual Neural Network |
| title_fullStr |
Analysis of Photovoltaic Cell Defect Classification Using Residual Neural Network |
| title_full_unstemmed |
Analysis of Photovoltaic Cell Defect Classification Using Residual Neural Network |
| title_sort |
Analysis of Photovoltaic Cell Defect Classification Using Residual Neural Network |
| publishDate |
2024 |
| container_title |
2024 IEEE 22nd Student Conference on Research and Development, SCOReD 2024 |
| container_volume |
|
| container_issue |
|
| doi_str_mv |
10.1109/SCOReD64708.2024.10872644 |
| url |
https://www.scopus.com/inward/record.uri?eid=2-s2.0-85219521304&doi=10.1109%2fSCOReD64708.2024.10872644&partnerID=40&md5=3d0a0ea6ce466a1417be5682ffe1b58a |
| description |
The photovoltaic (PV) system has seen significant advancements in recent years. Accurate defect classification is essential for maximizing the energy output of PV cells throughout their lifespan. With the rapid progress in deep learning and convolutional neural networks (CNN), PV images can now be leveraged for defect classification, offering improvements over traditional methods. This work proposes an analysis of PV cell defect classification using two different types of Residual Neural Networks (ResNet), namely ResNet-18 and ResNet-50, to determine which architecture delivers the best performance. The results are further compared with other CNN architectures, including deep CNN, GoogLeNet, and VGG-16. Various performance metrics were evaluated and compared across these models. The findings reveal that ResNet-18 outperforms the other architectures, achieving the highest performance with an accuracy of 97.96 %, specificity of 99.22%, sensitivity of 96.59%, precision of 97.42 %, and an F1-score of 96.80 %. This analysis will help to improve the defects classification system for PV cells. © 2024 IEEE. |
| publisher |
Institute of Electrical and Electronics Engineers Inc. |
| issn |
|
| language |
English |
| format |
Conference paper |
| accesstype |
|
| record_format |
scopus |
| collection |
Scopus |
| _version_ |
1828987861538439168 |