Application of ANN to predict incipient faults in power transformer based on DGA method / Nur Diyana Mansor

• Main Author: Mansor, Nur Diyana
• Format: Article
• Language: English
• Published: Research Management Institute 2013
• Subjects: Neural networks (Computer science); Applications of electronics
• Online Access: https://ir.uitm.edu.my/id/eprint/12458/1/12458.pdf; https://ir.uitm.edu.my/id/eprint/12458/

This paper Artificial Neural Networks (ANN) arc used to predict incipient faults in power transformers oil. The prediction is performed through the Dissolved Gas Analysis (DGA) method. The function of this method is for detect and diagnose the different types of incipient faults that occur in power transformers. By interpretation of dissolved gasses in oil insulation of power transformers, this method was applied the Artificial Neural Networks (ANN) to classify the different faults by using the DGA method. In DGA method, the Roger's Ratio and International Electrotechnical Commission (IEC) Ratio were applied into ANN to see the performance of ANN's network. For assessment, two set databases are employed: Roger's ratio and IEC ratio. The data bases are collected from Tenaga Nasional Berhad (TNB) data. The results show these methods were used to predicting the fault more than 90% of accuracy in best cases.