Scholarlink

On-site transformer partial discharge (PD) signals are susceptible to white noise and narrowband interference lead to low signal/noise ratio extracting functions is difficult. Solve the limitations of traditional wavelets. Transformation and empirical modal decomposition, such as strong parameter dependence and mode mixing, this paper proposes a hybrid denoising method combining Ensemble Empirical Mode Decomposition (EEMD) and Particle Swarm Optimization (PSO)-optimized wavelet thresholding. First, the noisy signal is decomposed via EEMD, where Gaussian white noise is introduced to suppress mode mixing, and effective intrinsic mode functions (IMFs) are selected based on kurtosis and correlation coefficient. Subsequently, The PSO algorithm is used to optimize the wavelet threshold function, enabling adaptive adjustment of denoising parameters and avoiding the blindness of traditional threshold selection. Finally, the denoised signal is reconstructed, and its performance is evaluated using metrics including SNR and root mean square error (RMSE). Modeling and measured data verification show that this method effectively suppresses noise while maintaining local unloading characteristics, significantly enhancing the reliability of insulation condition monitoring in power systems.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Partial Discharge Signal Denoising Method Based on Ensemble Empirical Mode Decomposition and Particle Swarm Optimization-Optimized Wavelet Thresholding

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Zhang Han,
  • Liu Weidong,
  • Liu Yaofeng

摘要

On-site transformer partial discharge (PD) signals are susceptible to white noise and narrowband interference lead to low signal/noise ratio extracting functions is difficult. Solve the limitations of traditional wavelets. Transformation and empirical modal decomposition, such as strong parameter dependence and mode mixing, this paper proposes a hybrid denoising method combining Ensemble Empirical Mode Decomposition (EEMD) and Particle Swarm Optimization (PSO)-optimized wavelet thresholding. First, the noisy signal is decomposed via EEMD, where Gaussian white noise is introduced to suppress mode mixing, and effective intrinsic mode functions (IMFs) are selected based on kurtosis and correlation coefficient. Subsequently, The PSO algorithm is used to optimize the wavelet threshold function, enabling adaptive adjustment of denoising parameters and avoiding the blindness of traditional threshold selection. Finally, the denoised signal is reconstructed, and its performance is evaluated using metrics including SNR and root mean square error (RMSE). Modeling and measured data verification show that this method effectively suppresses noise while maintaining local unloading characteristics, significantly enhancing the reliability of insulation condition monitoring in power systems.