電氣設備在高壓電力系統中，難以避免發生局部放電現象，通常不會立即
導致整個系統崩壞，但長期的局部放電現象會對電氣設備絕緣體造成損傷，進
而引發嚴重事故。由於過往影像辨識易因多種設備產生局部放電，可能含有兩
種以上放電型態，過往研究發現辨識結果高機率出現辨識失敗，因此，更加精
確的分析方法和技術對於提高辨識的準確性和可靠性變得至關重要。通過提升
辨識技術，可以更有效地監測和預測局部放電現象，從而減少設備故障風險，
保護電力系統的穩定運行。
本論文針對纜直線接頭與氣體絕緣開關兩種電氣設備，各三種瑕疵試驗樣
本共六種瑕疵案例。將局部放電數據經過資料處理與化簡，最終利用相位解析
圖譜，以彩色、灰階與紅藍圖譜呈現。接著，素點疊加方式和Hash相似度分
析找尋各式瑕疵中的代表性圖譜，最後將代表性圖譜拆分成0%至100%，並利
用窮舉法建立資料庫。資料庫皆會進行五種深度學習特徵提取模型，對各式瑕
疵1000 筆資料數中隨機挑選混合生成待測圖譜進行餘弦相似度分析比較，最
終利用混淆矩陣觀測結果及自訂三種策略合理解讀模型整體效能。最終氣體絕
緣開關整體模型效果有不錯的成效，MobileNet 模型及 Inception V3 模型辨識
皆達到準確率70%以上，至於電纜直線接頭待測物三種瑕疵圖譜有一定的相似
性導致有效性較低。

In high-voltage power systems, it is challenging to avoid the occurrence of partial discharge phenomena in electrical equipment. Although partial discharges do not typically lead to immediate system collapse, prolonged partial discharge can damage the insulation of electrical equipment, potentially causing severe accidents. Historical image recognition has often encountered difficulties due to partial discharges occurring in various types of equipment, which may contain more than two types of discharge patterns. Previous studies have found that recognition results frequently fail, making
more precise analysis methods and technologies crucial for improving the accuracy and reliability of recognition. By enhancing recognition technology, partial discharge phenomena can be monitored and predicted more effectively, thereby reducing the risk of equipment failure and ensuring the stable operation of power systems.
This thesis focuses on two types of electrical equipment: cable straight joints and gas-insulated switches, each with three defect test samples, resulting in a total of six defect cases. Partial discharge data is processed and simplified, and the final results are presented using phase analysis maps in color, grayscale, and red-blue spectrum formats. Subsequently, pixel stacking methods and Hash similarity analysis are used to identify representative patterns of various defects. These representative patterns are then decomposed into ranges from 0% to 100%, and a database is established using exhaustive methods. The database undergoes feature extraction using five different deep learning models. For each type of defect, 1,000 data samples are randomly selected and mixed to generate test maps, which are then compared using cosine similarity analysis. The results are observed using a confusion matrix, and three customized strategies are employed to reasonably interpret the overall performance of the models. Ultimately, the overall model performance for the gas-insulated switch shows promising results, with both the MobileNet and Inception V3 models achieving accuracy rates above 70%. However, the effectiveness is lower for the cable straight joint test samples due to the significant similarity among the three defect patterns.

[1] 台灣電力公司，「台電月刊738期」，民國一百一十三年四月。
[2] B. Kruizinga, P. A. A. F. Wouters and E. F. Steennis, “Fault Development upon Water Ingress in Damaged Low Voltage Underground Power Cables with Polymer Insulation” IEEE Transactions on Dielectrics and Electrical Insulation, vol.9, pp. 808-816, April 2017.
[3] 張建國，高電壓地下電纜接頭絕緣狀態之監測與診斷系統之研究，博士論文，
國立台灣科技大學，民國一百年。
[4] F. P. Mohamed, W. H. Siew, J. J. Soraghan, S. M. Strachan and J. McWilliam,
“Partial Discharge Location in Power Cables Using a Double Ended Method
Based on Time Triggering with GPS” IEEE Transactions on Dielectrics and
Electrical Insulation, vol.10, pp.2212-2221, December 2013.
[5] hady S. Refaat and Mohammed A. Shams, “A Review of Partial Discharge
Detection, Diagnosis Techniques in High Voltage Power Cables” 2018 IEEE
12th International Conference on Compatibility, Power Electronics and Power
Engineering, pp. 1-5, Doha, Qatar, April 2018.
[6] P. Wagenaars, P. A.A.F. Wouters, P. C.J.M. van der Wielen, E. F. Steennis,
“Accurate Estimation of the Time-of-arrival of Partial Discharge Pulses in Cable Systems in Service” IEEE Transactions on Dielectrics and Electrical Insulation, vol.10, pp.1190-1199, August 2008.
[7] Gaoyang Li, Mingzhe Rong, Xiaohua Wang, Xi Li and Yunjia Li, "Partial
discharge patterns recognition with deep Convolutional Neural Networks," 2016 International Conference on Condition Monitoring and Diagnosis (CMD), Xi'an, China, 2016, pp. 324-327, doi: 10.1109/CMD.2016.7757816.
[8] W. Wang and N. Yu, "Partial Discharge Detection with Convolutional Neural Networks," 2020 International Conference on Probabilistic Methods Applied to Power Systems (PMAPS), Liege, Belgium, 2020, pp. 1-6, doi: 10.1109/PMAPS47429.2020.9183469.
[9] M. Karimi, M. Majidi, H. MirSaeedi, M. M. Arefi and M. Oskuoee, "A Novel
Application of Deep Belief Networks in Learning Partial Discharge Patterns for
95 Classifying Corona, Surface, and Internal Discharges," in IEEE Transactions on Industrial Electronics, vol. 67, no. 4, pp. 3277-3287, April 2020, doi: 10.1109/TIE.2019.2908580.
[10] O. Aldosari, M. A. Aldowsari, S. M. Batiyah and N. Kanagaraj, "Image-Based Partial Discharge Identification in High Voltage Cables Using Hybrid Deep Network," in IEEE Access, vol. 11, pp. 50325-50333, 2023, doi: 10.1109/ACCESS.2023.3278054.
[11] Shirsendu Sikdar, Dianzi Liu, Abhishek Kundu, Acoustic emission data based deep learning approach for classification and detection of damage-sources in a composite panel,Composites Part B: Engineering,Volume 228, 2022, 109450, ISSN 1359-8368
[12] J. Guo et al., "Identification of Partial Discharge Based on Composite Optical Detection and Transformer-Based Deep Learning Model," in IEEE Transactions on Plasma Science, doi: 10.1109/TPS.2024.3382320.
[13] T. Sukumar, G. Balaji, B. Vigneshwaran, A. Prince and S. P. Kumar, "Recognition of Single and Multiple Partial Discharge patterns using Deep Learning Algorithm," 2021 International Conference on Artificial Intelligence and Smart Systems (ICAIS), Coimbatore, India, 2021, pp. 184-189, doi:
10.1109/ICAIS50930.2021.9395881.
[14] T. Liu, J. Yan, Y. Wang and Y. Du, "Research on GIS Partial Discharge Pattern
Recognition Based on Deep Residual Network and Transfer Learning in
Ubiquitous Power Internet of Things Context," 2020 5th Asia Conference on
Power and Electrical Engineering (ACPEE), Chengdu, China, 2020, pp. 207-211,
doi: 10.1109/ACPEE48638.2020.9136170.
[15] Iwata, S., Kitani, R. Phase-resolved partial discharge analysis of different types of electrode systems using machine learning classification. Electr Eng 103, 31893199 (2021). doi： 10.1007/s00202.021.01306.5
[16] Wang, Y., et al.: GAN and CNN for imbalanced partial discharge pattern
recognition in GIS. High Volt. 7(3), 452– 460 (2022). https://doi.org/10.1049/hve2.12135
[17] J. He, C. Liu, Q. Tang, C. Zhang, M. Cao and X. Xu, "Partial Discharge Pattern Recognition Algorithm Based on Sparse Self - coding and Extreme Learning 96 Machine," 2018 2nd IEEE Conference on Energy Internet and Energy System Integration (EI2), Beijing, China, 2018, pp. 1-6, doi: 10.1109/EI2.2018.8582500.
[18] S. Lu, H. Chai, A. Sahoo and B. T. Phung, "Condition Monitoring Based on Partial Discharge Diagnostics Using Machine Learning Methods: A Comprehensive State-of-the-Art Review," in IEEE Transactions on Dielectrics and Electrical Insulation, vol. 27, no. 6, pp. 1861-1888, December 2020, doi:
10.1109/TDEI.2020.009070.
[19] D. Ahmad and S. Wang, "Bidirectional LSTM Based Partial Discharge Pattern Analysis for Fault Detection in Medium Voltage Overhead Lines with Covered Conductors," 2020 IEEE 18th International Conference on Industrial Informatics (INDIN), Warwick,United Kingdom, 2020, pp. 70-73, doi:
10.1109/INDIN45582.2020.9442162.
[20] Y. Wang et al., "Detection and Recognition for Fault Insulator Based on Deep Learning," 2018 11th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI), Beijing, China, 2018, pp. 1-6, doi: 10.1109/CISP-BMEI.2018.8633245.
[21] J. Long et al., "A Comprehensive Review of Signal Processing and Machine
Learning Technologies for UHF PD Detection and Diagnosis (II): Pattern
Recognition Approaches," in IEEE Access, vol. 12, pp. 29850-29890, 2024, doi:
10.1109/ACCESS.2024.3368866.
[22] 張軒豪，應用局部放電脈衝序列分析與卷積神經網路於電纜接頭之絕緣狀態
評估，碩士論文，臺北，國立臺灣科技大學電機工程系，民國111年。
[23] 林鈺翔，應用集成式學習於氣體絕緣開關與電纜直線接頭之局部放電瑕疵辨
識，碩士論文，臺北，國立臺灣科技大學電機工程系，民國112年。
[24] 鐘易樺，應用適應性類神經模糊推論系統於氣體絕緣開關瑕疵辨識之研究，
博士論文，臺北，國立臺灣科技大學電機工程系，民國102年。
[25] 李卓翰，地下電纜接頭局部放電之差動量測研究，碩士論文，臺北，國立臺
灣科技大學電機工程系，民國109年。
[26] G. C. Stone, "Partial discharge diagnostics and electrical equipment insulation condition assessment," in IEEE Transactions on Dielectrics and Electrical Insulation,vol.12,no. 5, pp. 891-904, Oct. 2005, doi: 10.1109/TDEI.2005.1522184. 97
[27] "High-voltage test techniques - partial discharge measurements," IEC 60270, 3rd ed., 2000.
[28] M. Fritsch and M. Wolter, "Saturation of High-Frequency Current Transformers: Challenges and Solutions," in IEEE Transactions on Instrumentation and Measurement, vol. 72, pp. 1-10, 2023, Art no. 9004110, doi: 10.1109/TIM.2023.3302370.
[29] Y. Hao et al., "A Measurement Method of Ultrasonic Critical Refraction
Longitudinal Wave for Thermal Stress in GIS Basin-Type Insulators," in IEEE
Transactions on Instrumentation and Measurement, vol. 71, pp. 1-9, 2022, Art no. 6001009, doi: 10.1109/TIM.2021.3137547.
[30] X. Zhang, M. Shi, J. Cai and J. Li, "A Novel Partial Discharge Detection Method for Power Transformers On Site Adopting Its Component as Ultra-High Frequency Sensor," in IEEE Transactions on Power Delivery, vol. 34, no. 6, pp. 2269-2271, Dec. 2019, doi: 10.1109/TPWRD.2019.2903927.
[31] A. Dragomir, M. Adam, M. Andruşcă and A. Munteanu, "Aspects concerning the influence of environmental factors in infrared monitoring of electrical equipment," 2016 International Conference and Exposition on Electrical and Power Engineering (EPE),
Iasi, 10.1109/ICEPE.2016.7781319. Romania, 2016, pp. 133-138, doi:
[32] 林育勳，數位局部放電量測應用於模鑄式比流器，碩士論文，臺北，國立臺
灣科技大學電機工程系，民國93年。
[33] 潘彥竹，局部放電量測系統之研製，碩士論文，臺北，國立臺灣科技大學電
機工程系，民國93年。
[34] X. Glorot, A. Bordes and Y. Bengio, "Deep sparse rectifier neural networks," Proceedings of the Fourteenth International Conference on Artificial Intelligence and Statistics, PMLR 15:315-323, 2011.
[35] N. Srivastava, G. Hinton, A. Krizhevsky, I. Sutskever, and R. Salakhutdinov,
"Dropout: A simple way to prevent neural networks from overfitting," Journal of Machine Learning Research, 2014.
[36] Simonyan, Karen, and Andrew Zisserman. "Very deep convolutional networks for large-scale image recognition." arXiv preprint arXiv:1409.1556 (2014). 98
[37] K. He, X. Zhang, S. Ren and J. Sun, "Deep Residual Learning for Image
Recognition," 2016 IEEE Conference on Computer Vision and Pattern
Recognition (CVPR), Las Vegas, NV, USA, 2016, pp. 770-778, doi:
10.1109/CVPR.2016.90.
[38] Howard, Andrew G., et al. "Mobilenets: Efficient convolutional neural networks for mobile vision applications." arXiv preprint arXiv:1704.04861 (2017).
[39] C. Szegedy, V. Vanhoucke, S. Ioffe and J. Shlens, "Rethinking the Inception
Architecture for Computer Vision." arXiv preprint arXiv: 1512.00567 (2015).
[40] A. Joshi, A. V. Shet, A. S. Thambi and S. R, "Quality Improvement of Image
Datasets using Hashing Techniques," 2023 International Conference on Intelligent and Innovative Technologies in Computing, Electrical and Electronics (IITCEE), Bengaluru, India, 2023, pp. 18-23, doi: 10.1109/IITCEE57236.2023.10091044.
[41] 吳德偉，結合改良式餘弦相似度應用於LED閱讀燈之相關專利分析研究，
碩士論文，臺北，國立臺灣科技大學機械工程系，民國102年。
[42] P. Thuan DO, T. Thu Huong Tran, V. Vajnovszki, "Exhaustive generation for
permutations avoiding a (colored) regular sets of patterns" arXiv: 1809.00742
(2018).
[43] David L. Donoho and Iain M. Johnstone. “Ideal spatial adaptation by wavelet shrinkage.” Biometrika 81 (1994): 425-455.
[44] B. K. Boyanapalli，部分放電量測週期對地下電纜接頭瑕疵識別影響之研究，
博士論文，臺北，國立臺灣科技大學電機工程系，民國111年。