非负矩阵分解方法识别电能质量扰动信号

Non-negative matrix factorization method identify power quality distur-bance signal

利用相空间重构及非负矩阵分解(NMF)相结合的方法,对电压暂降、电压暂升、电压尖峰、电压中断、暂态谐波及暂态振荡6类电能扰动信号进行分类识别研究。利用相空间重构法构造扰动信号轨迹,并将其转化为图像。从图像处理的角度出发,利用NMF在人脸、指纹图像识别应用中的基本原理,对不同的相空间重构轨迹图进行特征提取,分类识别其所对应的电能质量扰动信号。该方法可避免由于扰动信号的复杂性而难以获得扰动信号稳定特征提取的困难,具有训练时间短、所需训练样本少、识别过程可视化便于分析等特点。仿真实验结果表明其能够较好地识别电能质量扰动,是提供了扰动信号检测与分类的算法。

A combination of phase space reconstruction and Non-negative Matrix Factorization （NMF） methods is applied to recognize six disturbance signals including voltage sag, voltage swell, voltage spikes, voltage interruption, harmonics and fluctu- ation signals. The phase space reconstruction method is used to construct disturbance signal trajectories converted into images. In the view of image processing, the principle of NMF in the face and fingerprint image recognition is adopted to extract the fea- tures of different phase space reconstruction trajectories, and recognize the corresponding power quality disturbance signals. This method can avoid obtaining the difficulties of stability feature extraction which result of the complexity of the disturbance signal, with the training time is short, less training samples needed to identify the process of visualization to facilitate the analy- sis and so on. Simulation results show that it can better identify the power quality disturbance. It is disturbance signal detection and classification of possible algorithms.