Power Transformer Interruption Analysis Based on Dissolved Gas Analysis (DGA) using Artificial Neural Network

The power transformer is an important component in the power system, as it is directly related to the reliability of the electric power system operation. Therefore, the diagnosis of disturbances in power transformers is important for device safety as well as electrical system stability. Dissolved Gas Analysis (DGA) is a disruptive diagnostic technique in power transformers that has been recognized effectively, because it provides knowledge of the state of the transformer based on the dissolved gas content in the transformer oil. The DGA test results can be represented by different methods such as Doernenburg ratio, Rogers ratio, IEC ratio, Duvals triangle, and key gases. The problem presented here is that two methods namely Doernenburg ratio and Rogers ratio, for the same data inputs give two different results from the error diagnosis to know the actual state of the transformer. In this paper, a combination method is proposed to solve the problem of conflict between Doernenburg ratio and Rogers ratio by utilizing multi-layer artificial neural network perceptron to localize and identify the error on the transformer and to select the most appropriate method.