電力變壓器主要做為電力系統電壓的轉換,需承受長時間的高電壓、 大電流、大幅溫度變化及電磁力之衝擊。一旦發生故障往往造成很大的傷 害及廣大範圍的停電損失將會非常慘重,故如何維持電力變壓器的正常功 能,甚至進一步延長使用期限及預知事故的發生以便防範於未然是相當重 要的課題,且此有賴於良好的維護並藉由試驗及診斷技術來預防事故的發 生。油中溶解氣體分析法是目前用於診斷油浸式電力變壓器潛在內部故障 現象最有效之方法及最受歡迎之技術,此方法因可靠性較佳且可在不斷電 下實施,現在已成為變壓器例行之維護項目。 本論文主要研究應用油中氣體分析的方法,對油浸式電力變壓器故障 之診斷做一探討,並結合六種診斷法建構一套快速的綜合診斷系統程式。 當變壓器內部產生故障時,將檢測出的油中氣體成份含量,經由本程式可 迅速的診斷出故障類型。更進一步再以類神經網路方法進行輔助診斷,並 與常用之氣體模式分析法、油中氣體分析法,相互比較,以驗證其可行性。 經以台灣電力公司實際之變壓器故障案例進行測試驗證,本論文所提之方 法分析確實可做為變壓器預防維護保養的參考,提升變壓器運轉安全可靠 度。
The power transformer is designed to change the voltage of power system, and have to face the burdens of long-term high voltage, large current, various temperatures and impact of electromagnetic stress. Once a breakdown happens, it always makes havoc and the outage will cause economic loss. Therefore, it is important to keep the power transformer running smoothly and longer. On the other hand, to predicate the breakdown events of the power transformer is another main subject. Dissolved gas analysis technique is the most popular and effective method for diagnosing the potential defects existing in oil-immersed power transformer. To practice the technique can keep the power supply and provide a higher reliability. The technique of dissolved gas analysis is adopted to commence routine maintenance of the power transformer. The main purpose of this thesis is to employ the dissolved gas analysis for abnormal conditions diagnose. A fast and accurate synthesis diagnosis program which based on six different existing theories is designed in this thesis. The abnormal parts and fault types can be detected when the power transformer oil dissolved gas data is applied to the program. An artificial neural network (ANN) is employed to assist diagnosis program. The approach in this thesis is compared with conventionally-used methods, gas pattern analysis (GPA) and dissolved gas analysis (DGA), to investigate its feasibility. The practical data of faulted cases from Taiwan Power Company has been utilized to test the proposed method. From the simulation results, it has been shown the proposed approach in this thesis can provide an alternative method to maintenance the power transformer operating reliability and safety.