To solve the waveform distortions and uncertainties of traditional threshold methods and blind separation methods for periodic narrowband noise rejection of partial discharges (PD), this paper presents a noise rejection method based on improved FastICA algorithm. Firstly, in this paper, the number of periodic narrowband noise is determined by information entropy theory combined with mixed partial discharge DFT power spectrum, and the power spectrum "two steps" method is proposed to estimate the interference frequency accurately. Then the simplified Simpson-Newton iteration method and calibration signal are introduced to improve the FastICA algorithm, which can reduce iterations and modify the separated PD signal effectively. The simulation and experiment results show that: the proposed method can reject periodic narrowband noise and keep the original features of partial discharges effectively with low time complexity. Compared with frequency slice wavelet and FFT denoising method, the denoised PD signals using the method of this paper are better without edge effect.%为了解决传统阈值法和盲分离法在局部放电周期性窄带干扰抑制过程中存在的波形畸变与不确定性问题,提出利用改进的FastICA算法对局部放电(PD)信号进行窄带干扰抑制.首先结合PD混合信号离散傅里叶变换功率谱的特点,引入信息熵理论对窄带干扰个数进行确定,并提出利用功率谱"两步法"对窄带干扰频率进行精确估计;然后引入简化的Simpson-Newton迭代公式对FastICA算法进行改进,减少算法的迭代次数,并提出利用校准信号对分离信号进行校准,修正分离PD信号的幅值和相位.仿真及实测结果表明,该方法能够有效抑制周期性窄带干扰并能有效保留信号原始特征,且算法的时间复杂度较小,同时与频率切片小波去噪方法和傅里叶变换滤波相比,该方法去噪效果更明显,且能有效避免边缘效应的影响.
展开▼
