Partial discharge data augmentation based on wavelet coefficients and variational autoencoder

doi:10.1016/j.measurement.2025.119219

Partial discharge data augmentation based on wavelet coefficients and variational autoencoder

2025-10-02

0

PRE

AI

DOI：10.1016/j.measurement.2025.119219

Original

Original request for help

Save

Abstract

En 中文

• We propose Wavelet Decomposition-Variational Autoencoder (WD-VAE) to tackle partial discharge (PD) data imbalance. • WD-VAE uses a gating mechanism to dynamically fuse latent vectors of raw signals and wavelet coefficients. • WD-VAE introduces a wavelet coefficient loss term to ensure physical constraints of generated PD data. • WD-VAE outperforms comparative algorithms in PD augmentation with better scores and efficiency.

Partial discharge data augmentation based on wavelet coefficients and variational autoencoder

Abstract

Journal

Organization