Transformer partial discharge signal reconstruction and feature extraction method based on time–frequency matrix

Abstract

Effectively extracting power transformer partial discharge (PD) signals is of great significance for monitoring the power transformer insulation state. However, practical and effective extraction methods have been lacking. This paper proposes a novel method for the extraction PD signals feature by the analysis of their time–frequency matrix. First, empirical mode decomposition (EMD) is carried out for raw signals to obtain the complete Hilbert time–frequency spectrum. Second, the frequency band partition is carried out. And the component of raw signals on each frequency band is constructed after filtering by a Hilbert–Huang transform (HHT) band‐pass filter. Then the time–frequency matrix is constructed by different frequency band components, and singular value decomposition is carried out. Using its singular value energy spectrum, the PD signal is reconstructed, further realizing feature extraction. Finally, the analysis result from actual examples indicates that this method can effectively extract PD signal feature frequency spectrum, and meanwhile also can eliminate the strong background interference and retain the higher time–frequency resolution. © 2017 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.