基于改进HHT+WT复合特征提取的电能质量扰动识别方法

Power Quality Disturbance Identification Method Based on HHT+WT Composite Feature Extraction

单一变换方法处理扰动信号提取的特征并不显著.将希尔伯特-黄变换(Hilbert-Huang Transform,HHT)与小波变换(Wavelet Transform,WT)相结合进行复合特征提取,对于非平稳部分和含谐波分量的扰动信号利用HHT变换提取特征,提取频率成分、持续时间、瞬时幅值标准差和瞬时频率标准差作为一部分特征;对于电压闪变等扰动利用小波变换提取特征,引入电压闪变能量作为参考向量,取输入扰动能量,与参考向量做差,作为另一部分特征.将提取的特征组成特征向量,作为一维卷积神经网络输入,卷积层进行卷积处理(即二次特征提取),全连接层进行分类.仿真结果表明:与传统小波变换相比,本方法提取的特征较为显著,且分类效果较好.

The single transformation method processes the disturbance signal,and the extracted features are not significant.Combining Hilbert-Huang Transform(HHT)and wavelet transform(WT)for extracting features.For the non-stationary part and the disturbance signal containing harmonic components,HHT transform is used to extract features,and frequency components,duration,standard error of transient amplitude and standard error of transient frequency as part of the characteristics.For disturbances such as voltage flicker,wavelet transform is used to extract features,and voltage flicker energy is introduced as a reference vector.The energy of the input disturbance is taken and the difference with the reference vector is taken as another part of the feature.The extracted features are input into the one-dimensional convolutional neural network,the convolutional layer performs secondary feature extraction,and the fully connected layer performs classification.The simulation results show that the features extracted by this method are more obvious and the classification effect is better comparing with the traditional wavelet transform.