簡易檢索 / 詳目顯示

回結果列表
研究生： 顏領呈
Yan, Ling-Cheng
論文名稱： 探討利用深度學習方法辨識太陽能電池板瑕疵問題之研究
On the Study of Solar Cell Panels Defect Recognition Using Deep Learning Methods
指導教授： 楊竹星
Yang, Chu-Sing
學位類別： 碩士
Master
系所名稱： 電機資訊學院 - 電腦與通信工程研究所
Institute of Computer & Communication Engineering
論文出版年： 2021
畢業學年度： 110
語文別： 中文
論文頁數： 62
中文關鍵詞： 太陽能板瑕疵檢測 、深度學習 、遷移學習 、影像辨識 、神經網路視覺化
外文關鍵詞： Solar Cell Panels, Deep Learning, Transfer Learning, Image Recognition
相關次數： 點閱：164  下載：42 
分享至:
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報

    • 現今世界越來越重視可再生之能源，尤其是太陽能。太陽能可以利用太陽能電
    池板去吸收獲取。然而太陽能電池板容易受到外在環境影響使得其吸收效率下降，
    且使得其吸收率下降之原因有許多，例如：太陽能電池板上裂痕損壞、受到建築物
    遮蔽而產生遮陰的太陽能電池板、太陽能電池板不平衡吸收熱能造成熱斑、附著在
    太陽能電池板上的灰塵。因此如何有效地辨識出太陽能電池板上是何種瑕疵問題，
    是本研究最主要探討的目標。過去不少關於太陽能電池板瑕疵辨識問題之研究中，
    提到關於非接觸式的辨識方法，非接觸方式指的是不需親自至太陽能電池板旁檢測
    之方式，其非接觸式之辨識方法相較接觸式之辨識方法在發現瑕疵問題之時間更
    短，且辨識率也較高，其中一種非接觸式方法為使用深度學習方法讓機器自動化辨
    識瑕疵。
    本研究透過共兩萬張之太陽能電池板影像資料集，利用深度學習方法去做太陽
    能電池板上的瑕疵影像辨識，除了使用遷移學習機制之不同神經網路辨識太陽能電
    池板瑕疵問題外，也重新提出了一個神經網路模型稱為 GHNet(Guided Block with
    Hybrid Methods Network)去做瑕疵辨識，以提升原模型之準確率，其上升約 3％的準
    確率。本研究最後也針對其神經網路模型做視覺化實驗，透過視覺化實驗使得可得
    知該神經網路模型在學習過程所擷取的特徵樣貌為何，使得原本為黑盒子之深度學
    習神經網路變得更可視化且透明清楚。

    The world pays more and more attention to renewable energy such as the solar energy Nowadays. Solar energy can be absorbed by solar panels. However, there are many reasons that cause solar panels energy absorptivity to decrease a lot. For example, cracks on solar panels, solar panels which are shaded by buildings, dust attached to the solar panels.
    Therefore, the goal of the research is to effectively recognize the defects on the solar panels. In the past, there were many researches that mentioned about non-contact method to recognize the defect on the solar panels. Non-contact method means that we don’t have to examine solar panels in person. Non-contact method which finds the defect on the solar panels costs less time and gets higher accuracy than contact methods. One of the popular methods is using deep learning method to recognize the defect on the solar panels.
    This research used a total 20,000 solar panels images with deep learning technology to identify the solar panels faulty. Besides the transfer learning technology is used in the research, this research also proposed a neural network called GHNet(Guided Block with Hybrid Methods Network) to recognize the defect on the solar panels. GHNet improved the accuracy by about 3% and the learning curve is less overfitting compared with EfficientNet used in the experiment. At the end of the research, we also did visualization experiment on GHNet model. Through the visualization experiment, we can get the features that neural network extracted during the model training. And it made the deep learning model more clearly and visible.
    Key words: Deep Learning, Transfer Learning, Solar Cell Panels, Image Recognition.

    目錄 摘要………………………………………………………………………………………………… i Abstract……………………………………………………………………………………………. ii 誌謝………………………………………………………………………………………………... iii 目錄………………………………………………………………………………………………... iv 表目錄…………………………………………………………………………………………….. vii 圖目錄……………………………………………………………………………………………. viii Chapter 1 緒論..........................................................................................................................1 1.1 研究背景..........................................................................................................................1 1.2 研究動機..........................................................................................................................2 1.3 研究目的..........................................................................................................................2 Chapter 2 相關研究 ..................................................................................................................4 2.1 深度學習在太陽能電池板之瑕疵問題與研究.................................................................4 2.2 深度學習之符號定義.......................................................................................................6 2.2.1 Two Dimension Convolution (Conv2d).......................................................................6 2.2.2 Two Dimension Max Pooling (MaxPool2D) ...............................................................7 2.2.3 批量歸一化(Batch Normalization) .............................................................................8 2.2.4 Non Linear Activation Function – ReLU...................................................................10 2.2.5 Neural Network Hidden Layer Output Relation.........................................................11 2.2.6 Neural Network Loss Function..................................................................................12 2.2.7 Neural Network Gradient Descent.............................................................................12 2.3 Based Convolution Neural Network Model......................................................................14 2.3.1 VGG–16...................................................................................................................14 2.3.2 VGG–19...................................................................................................................14 2.3.3 GoogLeNet...............................................................................................................15 2.3.4 ResNet......................................................................................................................16 2.3.5 EfficientNet ..............................................................................................................16 2.4 遷移學習機制.................................................................................................................18 Chapter 3 GHNet (Guided Block with Hybrid Methods Network)..................................19 3.1 GHNet 架構....................................................................................................................19 3.2 GHNet – Hybrid Methods ...............................................................................................21 3.2.1 Gabor Network .........................................................................................................22 3.2.2 Inception Module......................................................................................................23 3.2.3 Residual Block .........................................................................................................25 3.2.4 Dilated Convolution Neural Network........................................................................27 3.3 GHNet – Guided Block ...................................................................................................29 3.4 GHNet – Two Stack of Neural Network ..........................................................................31 3.4.1 First Stack of Neural Network...................................................................................31 3.4.2 Second Stack of Neural Network ..............................................................................33 Chapter 4 實驗結果與分析.....................................................................................................35 4.1 實驗環境建置與資料集 .................................................................................................35 4.2 K-means 集群分析..........................................................................................................38 4.2.1 未使用 K-means 之訓練資料集情形 .......................................................................39 4.2.2 使用 K-means 之訓練資料集情形...........................................................................39 4.2.3 K-means 前後之準確率比較....................................................................................41 4.3 深度學習模型在辨識準確率之效能評估.......................................................................42 4.3.1 GHNet 之混淆矩陣(Confusion Matrix)分析與準確率.............................................43 4.3.2 GHNet 之 ROC Curve(Receiver Operating Characteristic Curve) ........................44 4.3.3 GHNet 之 PR Curve(Precision-Recall Curve) ...........................................................44 4.3.4 不同模型之效能評估與準確率 ...............................................................................45 4.4 深度學習模型之學習曲線(Learning Curve)...................................................................46 4.5 深度學習模型之參數量(Parameters)..............................................................................48 4.6 GHNet 模型參數調整在準確率結果之實驗 ..................................................................48 4.7 消融實驗 (Ablation Experiment)....................................................................................50 4.8 神經網路視覺化實驗.....................................................................................................51 4.8.1 視覺化操作介面......................................................................................................52 4.8.2 影像之特徵視覺化 ..................................................................................................53 Chapter 5 結論........................................................................................................................55 5.1 研究貢獻 ........................................................................................................................55 5.2 未來研究方向.................................................................................................................55 References ..................................................................................................................................57

    [1]. D. Tsai, S. Wu and W. Chiu, "Defect Detection in Solar Modules Using ICA Basis Images," in IEEE Transactions on Industrial Informatics, vol. 9, no. 1, pp. 122-131, Feb. 2013, doi:10.1109/TII.2012.2209663.
    [2]. M. Millendorf, E. Obropta, and N. Vadhavkar, “Infrared Solar Module Dataset For Anomaly Detection”, 2020 International Conference on Learning Representations(ICLR), February 2020.
    [3]. M. Garaj, K. Y. Hong, H. Shu-Hung Chung, J. Zhou and A. W. Lo, "Photovoltaic Panel Health Diagnostic System for Solar Power Plants," 2019 IEEE Applied Power Electronics Conference and Exposition (APEC), 2019, pp. 1078-1083, doi: 10.1109/APEC.2019.8721839.
    [4]. M. Sander, B. Henke, S. Schweizer, M. Ebert and J. Bagdahn, "PV module defect detection by combination of mechanical and electrical analysis methods," 2010 35th IEEE Photovoltaic Specialists Conference, 2010, pp. 001765-001769, doi: 10.1109/PVSC.2010.5615878.
    [5]. X. Gong, Q. Yao, M. Wang and Y. Lin, "A Deep Learning Approach for Oriented Electrical Equipment Detection in Thermal Images," in IEEE Access, vol. 6, pp. 41590-41597, 2018, doi: 10.1109/ACCESS.2018.2859048.
    [6]. R. Pierdicca, E.S. Malinverni, F. Piccinini, M. Paolanti, A. Felicetti, and P. Zingaretti, “Deep Convolutional Neural Network For Automatic Detection Of Damaged Photovoltaic Cells”, ISPRS TC II Mid-term Symposium, May 2018.
    [7]. M. Bhattarai and M. MartíNez-Ramón, "A Deep Learning Framework for Detection of Targets in Thermal Images to Improve Firefighting," in IEEE Access, vol. 8, pp. 88308-88321, 2020, doi: 10.1109/ACCESS.2020.2993767.
    [8]. I. Zyout and A. Oatawneh, "Detection of PV Solar Panel Surface Defects using Transfer Learning of the Deep Convolutional Neural Networks," 2020 Advances in Science and Engineering Technology International Conferences (ASET), 2020, pp. 1-4, doi: 10.1109/ASET48392.2020.9118382.
    [9]. K. Simonyan, and A. Zisserman, “Very Deep Convolutional Networks For Large-Scale Image Recognition”, International Conference on Learning Representations(ICLR) 2015, April 2015.
    [10]. L. Wen, X. Li, X. Li and L. Gao, "A New Transfer Learning Based on VGG-19 Network for Fault Diagnosis," 2019 IEEE 23rd International Conference on Computer Supported Cooperative Work in Design (CSCWD), 2019, pp. 205-209,doi:
    10.1109/CSCWD.2019.8791884.
    [11]. O. Ronneberger, P. Fischer, and T. Brox, “U-Net: Convolutional Networks for Biomedical Image Segmentation”, Computer Vision and Pattern Recognition(CVPR), May 2015.
    [12]. T. Y. Lin, P. Dollár, R. Girshick, K. He, B. Hariharan, and S. Belongie, “Feature Pyramid Networks for Object Detection”, Computer Vision and Pattern Recognition(CVPR), April 2017.
    [13]. A. Chaurasia, and E. Culurciello, “LinkNet: Exploiting Encoder Representations for Efficient Semantic Segmentation”, Computer Vision and Pattern Recognition(CVPR), June 2017.
    [14]. M. Tan, and Q. V. Le, “EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks”, Computer Vision and Pattern Recognition(CVPR), September 2020.
    [15]. R. Pierdicca, M. Paolanti, A. Felicetti, F. Piccinini, and P. Zingaretti, “Automatic Faults Detection of Photovoltaic Farms: solAIr, a Deep Learning-Based System for Thermal Images”, Energies 2020,December 2020.
    [16]. C. Szegedy, W. Liu, Y. Jia, P. Sermanet, S. Reed, D. Anguelov, D. Erhan, V. Vanhoucke, and A. Rabinovich, “Going Deeper with Convolutions”, Computer Vision and Pattern Recognition(CVPR),September 2014.
    [17]. K. He, X. Zhang, S. Ren, and J. Sun, “Deep Residual Learning for Image Recognition”, Computer Vision and Pattern Recognition(CVPR), December 2015.
    [18]. L. Du, "How Much Deep Learning does Neural Style Transfer Really Need? An Ablation Study," 2020 IEEE Winter Conference on Applications of Computer Vision (WACV), 2020, pp. 3139-3148, doi: 10.1109/WACV45572.2020.9093537.
    [19]. B. P. Salmon, W. Kleynhans, C. P. Schwegmann and J. C. Olivier, "Proper comparison among methods using a confusion matrix," 2015 IEEE International Geoscience and Remote Sensing Symposium (IGARSS), 2015, pp. 3057-3060, doi: 10.1109/IGARSS.2015.7326461.
    [20]. Y. Wu, S. Fu, J. Zhao and C. Bryan, "Powering Visualization With Deep Learning," in IEEE Computer Graphics and Applications, vol. 41, no. 5, pp. 16-17, 1 Sept.-Oct. 2021, doi: 10.1109/MCG.2021.3102711.
    [21]. B. Zhou, A. Khosla, A. Lapedriza, A. Oliva, and A. Torralba, “Learning Deep Features for Discriminative Localization”, Computer Vision and Pattern Recognition(CVPR), December 2015.
    [22]. R. R. Selvaraju, M. Cogswell, A. Das, R. Vedantam, D. Parikh, and D. Batra, “Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization”, Computer Vision and Pattern Recognition(CVPR), December 2019.
    [23]. D. Smilkov, N. Thorat, B. Kim, F. Viégas, and M. Wattenberg, “SmoothGrad: removing noise by adding noise”, Computer Vision and Pattern Recognition(CVPR), June 2017.
    [24]. N. M. Maricar et al., "Photovoltaic solar energy technology overview for Malaysia scenario," Proceedings. National Power Engineering Conference, 2003. PECon 2003., 2003, pp. 300-305, doi: 10.1109/PECON.2003.1437462.
    [25]. J. Liu, J. Wang, Z. Tan, Y. Meng and X. Xu, "The analysis and application of solar energy PV power," 2011 International Conference on Advanced Power System Automation and Protection, 2011,pp. 1696-1700, doi: 10.1109/APAP.2011.6180758.
    [26]. A. Teal and M. Juhl, "Correcting the inherent distortion in luminescence images of silicon solar cells," 2015 IEEE 42nd Photovoltaic Specialist Conference (PVSC), 2015, pp. 1-5, doi:10.1109/PVSC.2015.7356303.
    [27]. N. Padmavathi and A. Chilambuchelvan, "Fault detection and identification of solar panels using Bluetooth," 2017 International Conference on Energy, Communication, Data Analytics and Soft Computing (ICECDS), 2017, pp. 3420-3426, doi: 10.1109/ICECDS.2017.8390096.
    [28]. P. Addabbo et al., "A UAV infrared measurement approach for defect detection in photovoltaic plants," 2017 IEEE International Workshop on Metrology for AeroSpace (MetroAeroSpace), 2017, pp.345-350, doi: 10.1109/MetroAeroSpace.2017.7999594.
    [29]. A. M. Bazzi, K. A. Kim, B. B. Johnson, P. T. Krein and A. Dominguez-García, "Fault impacts on solar power unit reliability," 2011 Twenty-Sixth Annual IEEE Applied Power Electronics Conference and Exposition (APEC), 2011, pp. 1223-1231, doi: 10.1109/APEC.2011.5744749.
    [30]. K. A. Kim, G. -S. Seo, B. -H. Cho and P. T. Krein, "Photovoltaic Hot-Spot Detection for Solar Panel Substrings Using AC Parameter Characterization," in IEEE Transactions on Power Electronics, vol. 31,no. 2, pp. 1121-1130, Feb. 2016, doi: 10.1109/TPEL.2015.2417548.
    [31]. M. Maghami, H. Hizam, and C. Gomes, “Impact of dust on solar energy generation based on actual performance”, 2014 IEEE International Conference on Power and Energy (PECon), December 2014.
    [32]. A. M. Gabor et al., "The Impact of Cracked Solar Cells on Solar Panel Energy Delivery," 2020 47th IEEE Photovoltaic Specialists Conference (PVSC), 2020, pp. 0810-0813, doi: 10.1109/PVSC45281.2020.9300743.
    [33]. H. Ziar, S. Mansourpour, E. Afjei and M. Kazemi, "Bypass diode characteristic effect on the behavior of solar PV array at shadow condition," 2012 3rd Power Electronics and Drive Systems Technology (PEDSTC), 2012, pp. 229-233, doi: 10.1109/PEDSTC.2012.6183331.
    [34]. S. Pingel et al., "Potential Induced Degradation of solar cells and panels," 2010 35th IEEE Photovoltaic Specialists Conference, 2010, pp. 002817-002822, doi: 10.1109/PVSC.2010.5616823.
    [35]. A. Paszke, S. Gross, S. Chintala, and G. Chanan, “PyTorch Get Started”, https://pytorch.org/getstarted/locally/ , 2016.
    [36]. H. Zhang, L. Zhang and Y. Jiang, "Overfitting and Underfitting Analysis for Deep Learning Based End-to-end Communication Systems," 2019 11th International Conference on Wireless Communications and Signal Processing (WCSP), 2019, pp. 1-6, doi: 10.1109/WCSP.2019.8927876.
    [37]. V. Thakkar, S. Tewary and C. Chakraborty, "Batch Normalization in Convolutional Neural Networks — A comparative study with CIFAR-10 data," 2018 Fifth International Conference on Emerging Applications of Information Technology (EAIT), 2018, pp. 1-5, doi: 10.1109/EAIT.2018.8470438.
    [38]. Y. Ho and S. Wookey, "The Real-World-Weight Cross-Entropy Loss Function: Modeling the Costs of Mislabeling," in IEEE Access, vol. 8, pp. 4806-4813, 2020, doi: 10.1109/ACCESS.2019.2962617.
    [39]. P. Baldi, "Gradient descent learning algorithm overview: a general dynamical systems perspective,"in IEEE Transactions on Neural Networks, vol. 6, no. 1, pp. 182-195, Jan. 1995, doi:10.1109/72.363438.
    [40]. Y. Lecun, L. Bottou, Y. Bengio and P. Haffner, "Gradient-based learning applied to document recognition," in Proceedings of the IEEE, vol. 86, no. 11, pp. 2278-2324, Nov. 1998, doi: 10.1109/5.726791.
    [41]. S. Bell, P. Upchurch, N. Snavely, and K. Bala, “Material Recognition in the Wild with the Materials in Context Database”, Computer Vision and Pattern Recognition(CVPR), April 2015.
    [42]. K. Chatfield, K. Simonyan, A. Vedaldi, and A. Zisserman, “Return of the Devil in the Details: Delving Deep into Convolutional Nets”, Computer Vision and Pattern Recognition(CVPR), November 2014.
    [43]. M. Cimpoi, S. Maji, I. Kokkinos, and A. Vedaldi, “Deep filter banks for texture recognition,description, and segmentation”, Computer Vision and Pattern Recognition(CVPR), November 2015.
    [44]. D. C. Cires¸an, U. Meier, J. Masci, L. M. Gambardella, and J. Schmidhuber, “Flexible, High Performance Convolutional Neural Networks for Image Classification”, IJCAI 2011, Proceedings of the 22nd International Joint Conference on Artificial Intelligence, July 2011.
    [45]. J. Deng, W. Dong, R. Socher, L. Li, Kai Li and Li Fei-Fei, "ImageNet: A large-scale hierarchical image database," 2009 IEEE Conference on Computer Vision and Pattern Recognition, 2009, pp. 248-255, doi: 10.1109/CVPR.2009.5206848.
    [46]. A. G. Howard, “Some Improvements on Deep Convolutional Neural Network Based Image Classification”, Computer Vision and Pattern Recognition(CVPR), December 2013.
    [47]. D. Erhan, C. Szegedy, A. Toshev, and D. Anguelov, “Scalable Object Detection using Deep Neural Networks”, Computer Vision and Pattern Recognition(CVPR), December 2013.
    [48]. G. E. Hinton, N. Srivastava, A. Krizhevsky, I. Sutskever, and R. R. Salakhutdinov, “Improving neural networks by preventing co-adaptation of feature detectors”, Computer Vision and Pattern Recognition(CVPR), July 2012.
    [49]. M. Lin, Q. Chen, and S. Yan, “Network In Network”, Computer Vision and Pattern
    Recognition(CVPR), March 2014.
    [50]. S. Gidaris, and N. Komodakis, “Object detection via a multi-region & semantic segmentation-aware CNN model”, Computer Vision and Pattern Recognition(CVPR), September 2015.
    [51]. R. Girshick, J. Donahue, T. Darrell, and J. Malik, “Rich feature hierarchies for accurate object detection and semantic segmentation”, Computer Vision and Pattern Recognition(CVPR), October 2014.
    [52]. K. He, and J. Sun, “Convolutional Neural Networks at Constrained Time Cost”, Computer Vision and Pattern Recognition(CVPR), December 2014.
    [53]. H. Cai, L. Zhu, and S. Han, “ProxylessNAS: Direct Neural Architecture Search on Target Task and Hardware”, Computer Vision and Pattern Recognition(CVPR), February 2019.
    [54]. E. D. Cubuk, B. Zoph, D. Mane, V. Vasudevan, and Q. V. Le, “AutoAugment: Learning Augmentation Policies from Data”, Computer Vision and Pattern Recognition(CVPR), April 2019.
    [55]. S. Han, H. Mao, and W. J. Dally, “Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Coding”, Computer Vision and Pattern Recognition(CVPR), February 2016.
    [56]. F. Zhuang et al., "A Comprehensive Survey on Transfer Learning," in Proceedings of the IEEE, vol.109, no. 1, pp. 43-76, Jan. 2021, doi: 10.1109/JPROC.2020.3004555.
    [57]. A. Alekseev, and A. Bobe, “GaborNet: Gabor filters with learnable parameters in deep convolutional neural networks”, Computer Vision and Pattern Recognition(CVPR), April 2019.
    [58]. F. Yu, and V. Koltun, “Multi-Scale Context Aggregation by Dilated Convolutions”, Computer Vision and Pattern Recognition(CVPR), April 2016.

    下載圖示 校內：立即公開
    校外：立即公開
    QR CODE