Inter-Turn Fault Detection Using Wavelet Analysis and Adaptive Neuro- Fuzzy Inference System

Abstract

The paper presents an effective on-line technique for detecting inter-turn faults in power transformers. The scheme operates by extracting negative sequence current samples from both primary and secondary line currents. Excitation currents are also extracted from the line currents. The magnitude and phase of the negative sequence current samples are separately decomposed using a 3-level wavelet decomposition. The absolute peaks of detail 3 coefficients are obtained. Magnitude and phase ratios are computed from obtained absolute peak values. The maximum values of the excitation currents are also extracted. The obtained ratios, together with the maximum excitation current values are fed into an adaptive neuro-fuzzy inference System (ANFIS) which determines whether there is an inter-turn fault. Where an inter-turn fault is detected, the level of severity is classified. Also, the side (primary or secondary) of the transformer on which the fault has occurred is also determined. Development and testing of the scheme was done through simulations on a 138/13.8 kV, 100 MVA three-phase transformer model using the Matlab software. The accuracy of the scheme was found to be high.