Differential protection of a power transformer using ANNs

The implementation of a pattern recognizer for power system diagnosis can provide great advancement in the protection field. The work reported in this paper demonstrates the use of an ANN as a pattern classifier for differential relay operation for power transformer protection. The proposed ANN based differential relay takes care of mal-operation due to magnetizing inrush current and over-excitation and ensures operation during internal fault only. Two ANN models i.e. Feed-Forward Back-Propagation (FFBP) and Radial Basis Function (RBF) have been considered and compared as regards their performance. The off-line experimental results presented in this paper shows that RBF network is more accurate in prediction and faster in training. This model can be considered as an attractive alternative method to make the discrimination among normal, magnetising inrush, over-excitation, and internal fault currents in a digital relay implementation. Results showing the performance of the protection scheme indicate that it is fast and reliable.