Research on Intelligent Identification of Magnetizing Inrush Current based on Empirical Modal Decomposition

Abstract

The accurate identification of the converter transformer magnetizing inrush current can ensure the safety of the operation of the UHV DC transmission system. In the differential protection of the converter transformer, it is extremely hard to distinguish between the excitation inrush current generated when the transformer is closed and the Fault current due to internal fault, which leads to frequent misoperation of the transformer protection to the extent that the transformer access fails. This paper proposes a current identification model based on empirical modal decomposition (EMD) and probabilistic neural network (PNN) algorithm with particle swarm (PSO) optimization. The EMD algorithm decomposes the current waveform, and the eigenmode functions (IMFs) of the three highest correlation bars are selected. The energy, cliffs, and standard deviation of the IMFs are extracted as the original waveform eigenvectors.