一种改善HHT端点效应的新方法及其在电能质量中的应用

利用Hilberr-Huang变换(HHT)方法进行电能质量检测分析,可以得到准确的瞬时频率和瞬时幅值,但该方法在应用中存在严重的端点效应,会影响分析结果。为了改善其端点效应问题,提出了一种基于人工神经网络和镜像延拓相结合的新方法对短时间序列进行延拓。采用三层BP神经网络对信号两端进行延拓,用带镜像延拓程序的经验模态分解(EMD)方法对延拓后的信号进行边分解边延拓,得到具有原始信号长度的固有模态函数(IMF);为了改善Hilbert变换中的端点效应,再次利用BP神经网络对各个IMF分量进行延拓。最后对延拓后的IMF分量进行Hilbert变换,从而得到精确的瞬时频率和瞬时幅值。将其应用到电力系统的谐波分析中,取得了较好效果。

End effect restraint of Hilbert-Huang transform and its application in power quality monitoring

HHT(Hilbert-Huang Transform)is used to acquire accurate instantaneous frequency and amplitude in power quality monitoring,but its end effect influences the detection results seriously. A method combining artificial neural network and mirror extension is proposed to extend short signal series for end effect restraint.A three-layer BP neural network is used to extend both ends of signal series and the EMD(Empirical Mode Decomposition)with mirror extension procedure to decompose the extended signal,by which IMFs(Intrinsic Mode Functions)with same length as original signal are obtained.To restrain the end effect of Hilbert transform,BP neural network is used again to extend the obtained IMFs and the Hilbert transform of the extended IMFs is then carried out to get accurate instantaneous frequency and amplitude.Its application in the harmonic analysis of power system shows its effectiveness.