| 引用本文: | 李继方,庞 鑫,曾小平,等.一种混合噪声的局部放电信号去噪方法[J].电力系统保护与控制,2026,54(08):24-36. |
| LI Jifang,PANG Xin,ZENG Xiaoping,et al.A denoising method for partial discharge signals with mixed noises[J].Power System Protection and Control,2026,54(08):24-36 |
|
| 摘要: |
| 电力电缆的绝缘状态对供电系统的稳定运行至关重要，而局部放电 (partial discharge, PD) 在线监测则成为评估绝缘状况的重要手段。然而，电力电缆的 PD 信号易受到现场噪声污染，导致监测信号失真，进而影响故障诊断的准确性和供电系统的稳定性。为此，提出了一种综合去噪方法，将模态分解与分类降噪相结合，逐步提高去噪效果。首先，采用常春藤算法 (Ivy 算法) 实现变分模态分解中模态数与惩罚因子的自动寻优，从而准确分解含噪 PD 信号。其次，通过计算各个模态分量的相关系数来区分 PD 主导分量和噪声主导分量，利用维纳滤波对 PD 主导分量进行去噪，对噪声主导分量则采用准则进行反向去噪，并对去噪后的信号进行重构。最后，构建自适应小波阈值函数对重构后的 PD 信号进行去噪处理，得到最终的 PD 去噪信号。仿真与实测信号分析表明，所提综合去噪方法去噪效果理想。与 Spearman 变分模态分解 (Spearman variational mode decomposition, SVMD) 等去噪方法相比，该方法信噪比提高了 66.16%，均方根误差降低了 0.03337%，波形相似系数提高了 1.138%，展现了优良的去噪性能。 |
| 关键词: 局部放电 常春藤算法 变分模态分解 维纳滤波 小波阈值 |
| DOI：10.19783/j.cnki.pspc.250789 |
| 分类号: |
| 基金项目:国家自然科学基金项目（U1804149） |
|
| A denoising method for partial discharge signals with mixed noises |
|
LI Jifang，PANG Xin，ZENG Xiaoping，WANG Feiyang，LI Jinheng
|
|
1. North China University of Water Resources and Electric Power, Zhengzhou 450045, China；2. Lankao Vocational College of Sannong, Kaifeng 475300, China
|
| Abstract: |
| The insulation condition of power cables is crucial for the stable operation of power systems, and online monitoring of partial discharge (PD) has become an important means of assessing insulation status. However, PD signals in power cables are highly susceptible to contamination from on-site noise, leading to signal distortion and consequently affecting the accuracy of fault diagnosis and system reliability. To address this issue, this paper proposes a comprehensive denoising method, which combines modal decomposition with classification-based noise reduction to progressively improve denoising performance. First, the Ivy algorithm is used to automatically optimize the modal number and penalty factor in variational mode decomposition, enabling accurate decomposition of noisy PD signals. Second, correlation coefficients of each modal component are calculated to distinguish PD-dominant components from noise-dominant components. The PD-dominant components are denoised using Wiener filtering, while the noise-dominant components are inversely denoised according to the criterion, followed by signal reconstruction. Finally, an adaptive wavelet threshold function is constructed to further denoise the reconstructed PD signal, yielding the final denoised output. Simulation and experimental results demonstrate that the proposed comprehensive denoising method achieves excellent performance. Compared with methods such as Spearman variational mode decomposition (SVMD), the proposed approach improves the signal-to-noise ratio by 66.16%, reduces the root mean square error by 0.03337%, and increases the waveform similarity coefficient by 1.138%, exhibiting superior denoising performance. |
| Key words: partial discharge Ivy algorithm variational mode decomposition Wiener filtering wavelet threshold |
