應用類神經網路於電力變壓器老化壽命損失估測

Estimation of Aging Life of Power Transformers Using Artifical Neural Networks

洪浡荏(Po-Jan Hung)

指導教授 ： 陳昭榮

國立臺北科技大學/電資學院/電機工程系研究所/碩士(2010年)

摘要

電力變壓器在電力系統中屬於重要且貴重的設備，它主要工作是為電網提供升、降壓功能。由於電力系統所要求的可靠度，不允許變壓器意外停電事故，因此絕大多數變壓器都是處於長時間的運轉狀態，這使得變壓器必須承受長時間的大電壓、大電流、高溫環境及高電磁應力的影響。因此如何維持變壓器正常運轉，減少意外事故發生，甚至進一步延長變壓器使用壽命及評估老化程度，是相當重要且值得研究的課題。
本論文主要研究內容，針對變壓器進行老化壽命損失推估，藉由變壓器老化過程中所產生之老化特徵產物數據，來量化變壓器實際老化程度，並估測可能的壽命損失，首先利用針對每台變壓器所訂定之狀況指標CI1~CI6值及變壓器之CO、CO2含量、最熱點溫度與負載率，來量化變壓器的壽命損失。並以類神經網路作為基礎，來獲得各影響因數與變壓器絕緣紙老化之關聯性。透過二輸入及四輸入變數評估模式，將所訂定之變壓器壽命損失函數，作為類神經網路的輸出，經過網路學習訓練完成後，利用已訓練完成之模式來估測變壓器老化壽命損失。經由老化壽命損失可以協助現場人員定期保養之參考。

關鍵字

電力變壓器 ； 類神經網路 ； 老化壽命

並列摘要

Power transformers as indispensable and highly expensive equipment in a power system not only have to operate for a long period of time under high temperature, high voltage, and high current but also need to withstand the impacts of electromagnetic force. Any breakdown or malfunction may trigger large-scale power outages causing drastic losses. It is therefore of crucial importance to keep power transformers in stable operation, to extend their service lives, and to monitor, predict, and prevent potential occurrence of accidents. These imperative tasks rely on solid maintenance mechanisms and effective test and diagnostic techniques.
The paper accordingly aims at estimating the aging and life loss of transformer by examining the major indicators of a transformer’s aging process so as to quantify the actual aging degree and to predict potential loss of useful life. Measurement of loss of useful life is first conducted by analyzing the condition index CI1~CI6, the value of CO and CO2, hot spot temperature and load of each transformer. Artificial neural network (ANN) is then adopted to train and examine the correlation between each variable and the aging of the transformer’s insulation paper. Based on the two-input and four-input variable evaluation models, the function of the loss of useful life caused by a transformer’s aging is used as the ANN output to facilitate necessary training. Upon the completion of training, the models are utilized to estimate the transformer’s aging-related loss of useful life. The estimation can be expected to assist on-site operators in performing regular maintenance.

並列關鍵字

Power transfommer ； Neural Networks ； Aging Life

參考文獻

[33] 林昇甫，洪成安，神經網路入門與圖樣辨識，台北：全華科技股份有限公司，2002。

[1] X. Zhang, E. Gockenbach, "Asset-management of transformers based on condition monitoring and standard diagnosis," IEEE Electrical Insulation Magazine, vol. 24, 2008, pp. 26-40.

[2] T. A. Prevost, T. V. Oommen, "Cellulose insulation in oil-filled power transformers: part I - history and development," IEEE Electrical Insulation Magazine, vol. 22, Jan.-Feb. 2006, pp. 28-35.

[3] W. J. McNutt, "Insulation thermal life considerations for transformer loading guides," IEEE Trans. Power Delivery, vol. 7, Jan.1992, pp.392-401.

[4] M. Wang, A. J. Vandermaar, K. D. Srivastava, "Review of condition assessment of power transformers in service," IEEE Electrical Insulation Magazine, vol. 18, 2002, pp.12-25.