充油電纜油中氣體分析診斷之研究

A Study in Dissolved Gases Analysis diagnosis of Oil-Filled Cable

葉志峰(Chih-Feng Yeh)

指導教授 ： 田方治

國立臺北科技大學/管理學院/工業工程與管理研究所/碩士(2007年)

https://doi.org/10.6841/NTUT.2007.00358

摘要

隨著輸變電設備日益增加，台灣早期建設之充油地下電纜設備已逐漸邁入「中年期」，維護管理問題將成為國內輸電工程未來首要的挑戰。充油電纜絕緣物質性質與負載特性有別於變壓器系統，其絕緣劣化預測模式也有所分別；而國內外目前有關於充油電纜劣化預測之相關探討文獻所見不多，惟對變壓器絕緣油劣化及油中溶解氣體分析診斷，有諸多詳盡之論述及研究，因此，對於不同運轉條件之充油電纜設備，如何運用其油中溶解氣體預測故障劣化之研究，仍有很大研究討論空間。
本研究主要研究對象為台灣地區運轉20年以上之輸電充油電纜線路及接續匣，分析其可能之異常原因，電纜線路之老化因素、機制及其相關效應；運用對數圖形法及主成份分析法，降低特徵氣體資料維度，藉由找出主成分變異較大之特徵氣體，推測電纜線路潛在故障型態；對電纜接續匣之油中氣體分析數據，設定其運轉溫度，轉換特徵氣體之生成速率，並將搜集之數據資料分不同逾年限之時期與現況將資料預作分類（A類為已逾管理建議值，即故障已存在，應及時處理；B類為逾管理建議值年限在二年之內；C類為逾管理建議值年限在二~五年之內；D類為五年以上），並以類神經網路進行電纜接續匣診斷分類。
本研究結果發現運用數圖形法及主成分分析法，比台灣電力公司目前所使用之判別基準更能提供明確之資訊，而類神經網路分類結果準確率可達86.33％以上，足見在充油電纜接續匣逾年限之分類上，確能提供較台電公司目前所使用之評定方法為佳之參考訊息。

關鍵字

油中溶解氣體分析 ； 類神經網路 ； 充油電纜

並列摘要

Early constructed Oil-filled cable apparatuses have been into「mid-life period」by transmission and substation apparatuses increasing rapidly, maintenance management issue will be risen to the first challenge for transmission and substation engineering. In Taiwan, underground power cables maintenance management still stay in eyes-watching and routine works, and for those function’s consumed also taking a passive strategy which is a strategy of fix until it out of control. For instant and non-anticipative situations could become a huge cost loss. Thus, the maintenance strategies of underground cable transmission lines will replace tradition routine work with possessiveness maintenance management.
This study analysis probably unusual reasons, aging factors, mechanisms and relative effects for underground transmission lines which operating over 20 years in Taipei, and according to those oil-filled cables dissolved gases analysis data setting operation temperature to find out its characteristic gases rising rate and potential failure model of cable joints and lines, and predict its aging degree. So that can provide useful relative messages of life cycle maintenance management for maintenance sections. To use logarithm drawing the graphs and principal component method finding the potential failures out, and getting sensitivity rate by probability neural network, the results of this study provide more useful information than existing to diagnose methods in Taipower company.

並列關鍵字

Dissolved Gases Analysis ； Neural Network ； Oil-filled Cable

參考文獻

[18] Zhenyuan Wang，Artificial Intelligence Application in the Diagnosis of Power Transformer Incipient Faults，Virginia，2000，Chapter2-4。

[22] Endrenyi. J. Albrecht, P.F. Billinton, R. Bulk,” power system reliability assessment-Why and How?” IEEE paper 82WM 147-7, presented at the winter Power meeting, New York, 1982.

[23] J. Densley, “Ageing Mechanisms and Diagnostics for Power Cables-An overview,” IEEE Electrical Insulation Magazine, vol.17, no.1, 2001, pp.14-22.

[24] I. A. Basheer, M. Hajmeer, ”Artificial neural networks : fundamentals, computing, design, and application,” Journal of Microbiological Methods,vol.43, no.1, 2000, pp.3-31.

[25] V. G.. Arakelian, ”Effective Diagnostics for Oil-Filled Equipment,” IEEE Electrical Insulation Magazine, vol. 18, no. 6, 2002, pp.26-38.