Parameter identification for fault detection of power transformer using artificial neural network

Abstract

Fault diagnosis is a challenging problem because there are numerous fault situations that can possibly occur to an electrical transformer. There are a lot of previous works done by researchers on fault diagnosis in power transformer but all of them used data from Dissolved Gas Analysis (DGA) as input for detection. This study will focus on parameter identification that is electrical measurement, which is voltage and current for fault detection due to several limitations of data from DGA that can lead to wrong diagnosis of fault in power transformer. The transformer that been used in this power system model is 132/20 kV with 250 MVA rating. The simulation of nine types of possible fault has been done by MATLAB R2013a Simulink software. To recognize the pattern of fault data, ANN was chosen because of it was easy to apply in power system network and it will work as pattern classifier with the ability to identify fault types accurately. The ANN programming has been done by ANN Pattern Recognition Tool that also in MATLAB R2013a software. It is found that the fault of power transformer can be detected by measuring electrical parameter such as voltage and current and with ANN, detection and classification of fault can be done to diagnose fault in power transformer. After the fault data had been trained for a few times, ANN will learn how to classify it accurately and then it is able to properly resolve new situations which are different from those fault data presented in the learning process.