Detecting Incipient Faults in Transformer using Dissolved Gas Analysis by Implementing Multilayer Perceptron Model

One of the very important components of the transmission grid is the Power Transformer. The failure of transformer may cause disturbance in the distribution and transmission operation. Thus it is very important to detect the incipient faults in transformer as early as possible. There are various conventional methods for detecting these faults; Dissolved Gas Analysis is one of the reliable techniques. This paper presents the application of Artificial Neural Network (ANN) in detecting incipient faults in transformer using the Multilayer Perceptron Model. The ANN model was developed using historical data to classify three transformer faults based on amount of hydrocarbon gases. The gas ratios are based on Doernenberg’s ratio method. The test results indicate that the ANN design yields a very satisfactory result and can make a very reliable technique for detecting incipient faults in transformer.