A novel approach for power transformer protection based upon combined wavelet transform and neural networks (WNN)

The role of a power transformer protective relay is to rapidly operate the tripping during fault condition and block the tripping during other operating conditions of the power transformer. This paper presents a new approach for classifying transient phenomena in power transformers, which may be implemented in digital relaying for transformer differential protection. Discrimination among different operating conditions (normal, inrush, over excitation, CT saturation and internal fault) of the power transformer is achieved by combining wavelet transform with neural network. The wavelet transform is applied for the analysis of power transformer transient phenomena, because of its ability to extract information from the transient signal simultaneously in both time and frequency domain. Neural network is used because of its self-learning and highly nonlinear mapping capability. The proposed scheme has been realized through two different artificial neural network (ANN) architectures (one is used as an internal fault detector (IFD) and another one is used as a condition monitor (CM)). These two ANN architectures were trained using BPN (back propagation algorithm) alone, and combining BPN with wavelet transforms (WNN), so that it should recognize and classify all the above operating conditions and faults. The simulation results obtained shows that the new algorithm accurately provides high operating sensitivity for internal faults and remains stable for other operating conditions of the power transformer. From that it was inferred that the combined WNN provides more accurate results and it has high-speed response and increased ability to discriminate even low-level fault signals from other operating conditions