短时奇异值分解用于局放信号混合噪声抑制

Mixed Noises Suppression of Partial Discharge Signal Employing Short-Time Singular Value Decomposition

电缆终端局部放电检测是诊断电缆终端绝缘状态的有效手段。为了有效抑制局放信号中的多种噪声源并保留局放信号的细节,提出了一种基于短时奇异值分解的局放信号混合噪声抑制方法。该方法首先利用短时滑动数据窗截取含噪局放信号片段进行奇异值分解,然后利用最优奇异值阈值对周期性窄带干扰进行甄别重构,并进行混合噪声的抑制。对含有混合噪声的局放仿真信号和实验室及现场实测局放信号进行去噪,并将去噪结果与自适应奇异值分解、形态学小波综合滤波器去噪结果进行对比。结果表明:所提去噪方法相比于自适应奇异值分解、形态学小波综合滤波器去噪能取得更好的去噪效果,去噪后波形相似度更高,误差更小,且当数据量较大时,该方法相比于自适应奇异值去噪能显著提高执行效率,具有较好的应用价值。

Partial discharge(PD)detection for cable termination is a useful method in cable termination diagnosis.To suppress mixed noises in partial discharge signals and preserve more features,a mixed noises suppression method based on short-time singular value decomposition(STSVD)is proposed.Firstly,the singular value decomposition(SVD)is performed on the noisy PD segment obtained by a sliding window along the time axis of analyzed PD signal,and then the mixed noises are eliminated by periodic narrowband noise reconstruction and optimal singular value threshold.The proposed method has been examined on simulated,laboratory obtained and field-detected noisy PD signals,and its results have been compared with adaptive singular value decomposition(ASVD)denoised results and morphology-wavelet filter(MWF)de-noised results.The results show that:compared with ASVD and MWF de-noising method,the proposed method has better performance,and this method can improve the execution efficiency obviously compared with ASVD when the analyzed data is large,which shows the good application value in practical PD detection.