Transformer Differential Protection with Wavelet based Artificial Neural Networks

This paper presents a wavelet based Artificial Neural Networks (ANN) algorithm, for distinguishing betweenmagnetizing inrush currents and power system fault currents. Although, normal currents in designed networks areexamined and this method could distinguished between normal currents and fault currents and over load condition,too. Differential relays are used and processing differential current harmonics is proposed for digital differentialprotection of power transformers. The proposed technique consists of a pre-processing unit based on discrete wavelettransform (DWT) based on artificial neural network (ANN) for detecting and classifying fault, inrush and normalcurrents. The DWT acts as an extractor of distinctive features in the input signals at the relay location. Thisinformation is then fed into an ANN for classifying fault, magnetizing inrush and normal conditions. A model ofpower system was simulated using Simulink-Matlab and ATP-EMTP software. The DWT was implemented by usingof Matlab, Daubechies and Coiflet mother wavelet was used to analyze primary currents and generate training data.Although a new wavelet that named Samlet is designed and examined for this purpose and is introduced in this paper,but it should develop further for better responses. At first output signal of the DWT module is used for predistinguishingbetween magnetizing inrush currents and power system fault currents based on the use of waveletanalysis to characterize inrush currents. After that, the output signal of the DWT module is fed into a ProbabilisticNeural Network (PNN) or a feed-forward, back-propagation ANN that classifies the transient. The simulated resultspresented show that the proposed technique can discriminate between magnetizing inrush, fault and normal currents intransformer protection.

Transformer Differential Protection with Wavelet based Artificial Neural Networks Keywords:

discrete wavelet transforms , differential relay , Daubechies wavelet , Coiflet wavelet , Samlet wavelet , artificial neural network , Probabilistic Neural Network (PNN) , fault detection , inrush current