粒子群算法用于局部放电小波阈值去噪

Application of Particle Swarm Optimization to Wavelet Threshold De-noising for Partial Discharge

小波阈值去噪在局放信号监测分析方面应用效果较好,而小波系数阈值的选取是决定局放信号去噪后的失真和误差的关键因素。针对局放脉冲频谱特征,提出了一种基于粒子群算法的小波最优阈值选择方法,用于局放脉冲信号去噪。采用小波对局放信号进行分解,在估计最优阈值时以广义交叉验证为标准,利用粒子群算法进行全局搜索,使阈值寻优效果大大提升。对人工模拟加噪信号和典型局放脉冲仿真信号进行去噪处理和定量分析,结果表明与标准软阈值法以及Donoho阈值计算法相比,对局放信号的去噪效果更好,有着非常好的应用前景和价值。

Wavelet threshold de- noising produces a good effect when it is applied to the monitoring and analysis of partial discharge （ PD } signals, and selection of the threshold for the wavelet coefficient is a key factor for determining the distortion and error of the de- noised PD signal. Under consideration of the characteristics of PD pulse spectrum, this paper proposed an optimal wavelet threshold selection method based on the particle swarm optimization （ PSO ） for de- noising of PD pulse signals. The wavelet was used to decompose PD signal, and generalized cross validation （ GCV ） was adopted as the criterion in the estimation of the optimal threshold. PSO was used for global search to greatly improve the threshold optimization effect. De- noising and quantitative analysis was conducted on the artificially simulated noise- adding signal and the typical PD pulse simulation signal. The results show that, compared with standard soft threshold method or Donoho threshold calculation method, better de-noising effect is produced with PD signals and that has a good application prospect and value.