會議論文
| 學年 | 95 |
|---|---|
| 學期 | 1 |
| 發表日期 | 2006-10-21 |
| 作品名稱 | 多重電力品質干擾事件之之偵測與評估 |
| 作品名稱（其他語言） | |
| 著者 | 蕭瑛東; 塗家銘; 張志翰; 陳建澤 |
| 作品所屬單位 | 淡江大學電機工程學系 |
| 出版者 | |
| 會議名稱 | 國科會電力學門九十四年度研究計畫成果發表會 |
| 會議地點 | 苗栗縣, 臺灣 |
| 摘要 | 本計畫針對多重電力品質事件之偵測與分析，進行相關研究與發展。現今機電設備，乃至於電子儀器與零件，因為運作精密程度的提升，對於電力品質優劣之敏感度與要求皆日益增加，而且某些電器設備有可能產生電力品質相關問題，使得其他電器設備亦遭到干擾。基於成本與效益的考量，電力系統中不同之電力品質事件之干擾，皆有許多不同的解決措施。而欲解決電力品質干擾問題，則必須首先具有高性能之電力品質事件之辨識與偵測系統。有鑑於過去許多研究，皆僅針對電力品質事件中眾多干擾因素之單一因素進行辨識與分析，而然在實際系統上，經常發生一電力品質事故同時具有多重電力品質干擾因素，如諧波、電壓閃爍與電壓過高等同時存在於一事件之情況，因此，他們的方法可能會受到限制。故本計畫將針對常見之電力品質事件，發展一套可同時辨識多重電力品質事件之線上即時偵測與評估系統，並以系統單晶片之方式實現之。本計畫第一年將針對幾種典型之電源干擾問題(如電壓突升、突降、中斷，或諧波、閃爍等)之電壓、電流與頻率之特性等進行相關數據之量測，並且以統計歸納方式進行電力品質事件波形資料庫之建立。此外，為補充量測數據之不足，本計畫亦發展一具圖形介面之電力品質模擬系統，可模擬產生電力品質事件含有多種電力品質干擾因素，將之以圖形及相關數據表現，以提供方便使用之人機介面。本計畫所建立之電力品質事件波形資料庫，將做為發展電力品質辨識與分析系統之測試資料庫，亦可提供作為教學及實驗之工具。本計畫第二年將發展一套新型動態結構類神經網路演算法，藉以改善傳統類神經網路之缺點以強化辨識效能，同時發展一套以小波理論為基礎的資料萃取演算法，以方便處理資料數量龐大而複雜之電力品質事件資料。最後將這兩個演算法結合，發展出具有高辨識率、高擴充性與快速之小波類神經網路電力品質事件辨識之演算法與視窗應用程式。本計畫第三年將以第二年計畫中所發展之小波類神經網路辨識系統架構為基礎，進行系統晶片硬體架構之設計與實現，以提供經濟、準確且快速之線上電力品質事件之偵測與辨識。此外，本計畫亦發展並實現一套具人工智慧之電力品質嚴重程度判別系統，以提供電力系統上各種電力品質干擾事件之嚴重等級之訊息，並依相關管制(限制)標準提出警訊，可供使用者或自動監控系統採取適度之因應措施，以避免運作中之電子儀器遭受危害。本報告為第一年計畫之精簡報告。 This project is contemplated to realize a new technique to perform diagnostics and assessment on the multiple power quality events. The quality of electricity supplies has become a major concern of electric utilities and end-users. The newly developed and widely used electric devices, while themselves are often the sources responsible for producing variant disturbance, are becoming more and more sensitive to power quality variations. Considering cost and performance, different power quality disturbance in the power system requires solving method with differential approaches. However, for diagnosing power quality problems, the causes of the disturbances should be understood before appropriate action can be taken. Attempts to solve the power quality problem from different perspectives have been considerable. It is noted that, most of these method treated the power quality problem as a single event problem. However, the presence of multiple power quality events (such as simultaneously existing harmonics, voltage flicker, and voltage swell) is natural in many power systems and makes the diagnostics of power quality problem interesting to solve. Thereby, those methods are inconvenient to classify the multiple power quality events. Therefore, this project is aimed to develop a power quality disturbances classification system, which is capable of classifying multiple power quality disturbances in a measured waveform or a PQ event, and results in an online real-time measurement and analysis IC. This project is divided into 3 years to proceed. In the first year, the test system will be setup to measure the characteristic on power quality of electric systems under typical operating and disturbance modes. These data will be analysis by statistic method for building database for the next-two-years project. Also, this object will develop the virtual power system for the power quality analysis. The virtual power system is a time domain simulation tool, which can simulate the multiple events due to disturbances and determine the effects of a specific disturbance on specific loads. The second year project will develop a novel dynamic structural neural network algorithm for improving the disadvantage of the traditional neural networks to diagnose the power quality problems. In the meantime, this project will develop a wavelet-based algorithm for extracting the critical data from the huge measurement power quality waveforms. Finally, the two algorithms will be combined for implementing as a solution algorithm with GUI PC-based program. The third year project is to develop a power quality index meter algorithm by utilizing AI techniques to calculate the power quality related index and derive the severe grade of power quality. This project is also contemplated with the system-on-chip design to realize general-purpose portable power quality device with diagnostic and alarm function. Since system-on-chip makes the microprocessor, memory, control logic and algorithm integrally-designed in one chip, thus it is provided with the characteristics of low cost, small size, fast operation speed, etc. This report is for the first year project. |
| 關鍵字 | 類神經網路;小波分析;電力品質;量測;偵測;Neural network;Wavelet analysis;Power quality;Measurement;Diagnostics |
| 語言 | zh_TW |
| 收錄於 | |
| 會議性質 | 國內 |
| 校內研討會地點 | |
| 研討會時間 | 20061021~20061021 |
| 通訊作者 | |
| 國別 | TWN |
| 公開徵稿 | Y |
| 出版型式 | 紙本 |
| 出處 | 國科會電力學門九十四年度研究計畫成果發表會論文集，21頁 |
| 相關連結 |
機構典藏連結 ( http://tkuir.lib.tku.edu.tw:8080/dspace/handle/987654321/96003 ) |