초록 close

- This study describesneural network based dissolved gas analysis using composition patterns of gas concentrations for transformer fault diagnosis. DGA samples were gathered from related literatures and classified into six types of faults and then a neural network was trained using the DGA samples. Diagnosis tests were performed by the trained neural network with DGA samples of serviced transformers, fault causes of whichwere identified by actualinspection. Diagnosis results by the neural network were in good agreement with actual faults.