現今工廠或屋內低壓配電線路都必須裝設過電流和過載保護裝置，確保用電安全，降低事故發生。但根據國外調查指出，某些住宅火災發生原因，可能是由電弧故障所引起，電弧故障斷路器(Arc-Fault Circuit Interrupter, AFCI)是一種偵測線路中電弧故障發生的保護裝置，在尚未發生火災前切離電源，避免電弧故障造成危險。本研究針對幾種不同特性負載以及三個混合負載的工作條件，藉由電弧故障檢測平台進行串聯電弧故障實驗，蒐集實驗數據。將實驗數據與相關文獻中時域和頻域的電弧特性做比較與驗證，並利用數位信號處理技巧，擷取串聯電弧電流特徵，做為串聯電弧故障檢測法的訓練資料。本論文提出兩個檢測法，皆利用小波包轉換和累積能量法擷取串聯電弧電流特徵，再分別搭配小波類神經網路和支持向量機，組合成兩個檢測法。再使用四個負載項目和三個混合負載條件進行正常運轉、正常電弧和串聯電弧測試，再將檢測結果和商用AFCI判斷結果做比較。本論文提出的兩種檢測方法皆可以有效檢測出串聯電弧故障，且不容易發生誤動作，若未來能開發為檢測裝置，將有助降低火災事故率。

Recent years, for the safety of utilizing electricity and reducing the occurrence of accidents, overcurrent devices and overload protection devices are needed in factories and low-voltage indoor distributed systems. But arc faults may cause fire accidents, thus, it is necessary to install arc-fault circuit interrupter(AFCI) which can switch off the power source before the occurring of fires. In this thesis, it is to investigate series arc fault detection of circuit feeding several characteristics of the load and three opration conditions of mixed load. The experiment data are collected by using the arc fault experiment platform. The characteristics of series arc faults will be verified in time-frequency domain by using experiment data, and compared with relative literature. Then, the experiment data are analyzed by using technique of digital signal process, and the feature of series arc faults are captured to make up training data of detecting methods. This study proposes two detecting methods. The first uses wavelet packet transform(WPT) and energy statistics to capture feature of series arc faults. The second combines wavelet neural network(WNN) and support vector machine(SVM) respectively. The detecting methods test each experiment data under three scenarios, including normal operation, on/off switching, and series arc fault. Finally, the test results are compared with a commercial arc-fault circuit interrupter. In this thesis, the proposed methods can detect occurring of series arc fault effectively, and the probability of malfunction is low.

[1]USFA (USA). Electrical Fire Safety: A Factsheet on Home Electrical Fire prevention. United States Fire Administration;2008. http://www.usfa.dhs.gov/citizens/all_citizens/home_fire_prev/electrical.shtm.
[2]NEMA Launches AFCI Safety Web Site [online]. Available: http://www.nema.org/pages/default.aspx
[3]內政部消防署，http://www.nfa.gov.tw/main/index.aspx。
[4]曾元超，「防範住家電氣火災的新技術」，台電月刊，549期，頁26-31，SEP. 2008。
[5]Arc-Fault Circuit Interrupters, UL Standard 1699-2008, Illinois, U.S.A., 2008.
[6]George D. Gregory, Kon Wong, Robert F. Dvorak, “More About Arc-Fault Circuit Interrupters,” IEEE Transactions on Industry Applications, Vol. 40, No. 4, pp. 1006-1011, July/August, 2004.
[7]E. Restrepo, “Arc Fault Detection and Discrimination Methods,” Proceedings of the 53th IEEE Holm Conference on Digital Object Identifier: Electrical Contacts (HOLM), Pittsburgh, USA, pp. 115-122, 16-19 Sept. , 2007.
[8]Li, Dongwei, Zhengxiang Song, Jianhua Wang, Geng, Yingsan,“A Method for Residential Series Arc Fault Detection and Identification,”Proceedings of the 55th IEEE Holm Conference on Digital Object Identifier: Electrical Contacts, Vancouver, Canada pp. 4013–4016, 14-16 Sept. , 2009.
[9]Wen-Jun Li and Yuan Chun Li, “Arc Fault Detection Based on Wavelet Packet,” Proceedings of 2005 International Conference on Machine Learning and Cybernetics, Guangzhou, China, pp. 1783-1788, 18-21 Aug. , 2005.
[10] P. Müller, S.Tenbohlen, R. Maier, and M. Anheuser, “Artificial Low Current Arc Fault for Pattern Recognition in Low Voltage Switchgear,” Proceedings of the 55th IEEE Holm Conference on Digital Object Identifier: Electrical Contacts, Vancouver, Canada pp. 15-21, 14-16 Sept. , 2009.
[11] P. Müller, S. Tenbohlen, R. Maier, and M. Anheuser, “Characteristics of Series and Parallel Low Current Arc Faults in the Time and Frequency Domain,” Proceedings of the 56th IEEE Holm Conference on Digital Object Identifier: Electrical Contacts (HOLM), Charleston, South Carolina, pp. 1-7, 4-7 Oct. , 2010.
[12]Zhang Ruixiang, Song Zhengxiang, “Arc Fault Detection Method Based on Signal Energy Distribution in Frequency Band,” Proceedings of Conference on the Power and Energy Engineering (APPEEC), Shanghai, China, pp. 1-4, 27-29 March, 2012.
[13]Shaohua Ma, Lina Guan, “Arc-Fault Recognition Based on BP Neural Network,” Proceedings of International conference on Measuring Technology and Mechatronics Automation (ICMTMA), Shanghai, China, pp. 584-586, 6-7 Jan, 2011.
[14]王奕捷，「低壓線路串聯電弧故障時域及頻域分析與檢測」，碩士論文，國立台灣科技大學碩士論文，民國101年。
[15]洪晨軒，「低壓屋內配線串聯電弧故障檢測及辨識」，碩士論文，國立台灣科技大學碩士論文，民國102年。
[16]G.D. Gregory, G.W. Scott, “The Arc-Fault Circuit Interrupter：An Emerging Product,” IEEE Transactions on Industry Applications, Vol. 34, no. 5, pp. 928 - 933, Sep/Oct, 1998.
[17]羅華強，訊號處理：matlab的應用，全華科技圖書股份有限公司，民國97年。
[18]張德丰，「詳解MATLAB數字信號處理」，北京，電子工業出版社，民國99年。
[19]傅祖勳，「電力品質資料分析與壓縮」，博士論文，國立台灣科技大學博士論文，民國91年。
[20]高志、余嘯海著，MATLAB小波分析工具箱處理與應用，國防工業出版社 民國93年。
[21]Charles K. Chui, “Orthogonal Wavelet and Wavelet Packet ,” in An Introduction to Wavelets, 1st edition, Harcourt Brace Jovanovich, Texas: Academic Perss, 1992, pp.215-240.
[22]Fredric M. Ham, Ivica Kostanic, “Mapping Network,” in Principles of Neurocomputing for Science & Engineering, International Edition 2001, New York: McGraw-Hill, 2001, ch. 3, pp. 96-164.
[23]高張斐章，張麗秋，和黃浩倫，「類神經活用理論與實務」，東華書局，民國93年。
[24]王進德，「類神經網路與模糊控制理論入門與應用」，全華圖書股份有限公司，民國100年。
[25]Simon O. Haykin, “Support Vector Machine,” Neural Network and Learning Machine, Third edition, USA: Prentice Hall, 2008, ch. 6, pp. 168-193.
[26]Whei-Min Lin, Chien-Hsien Wu, Chia-Hung Lin, Fu-Sheng heng “Detection and Classification of Multiple Power Quality Disturbances With Wavelet Multiclass SVM,” IEEE Transactions on Power Delivery, Vol. 23, No. 4, pp. 2575-2578, Octorber, 2008.
[27]Yang Liu, Rui Wang, Ying-Sheng Zeng, “An Improvement of One-against-One Method for Militi-class Support Vector Machine,” Proceedings of International conference on Machine Learning and Cybernetics, Hong Kong, China, pp. 2915-2920, August, 2007.