关键词: 电能质量, 相空间重构, 非负矩阵分解, 图像识别

Abstract: 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 fluctuation 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 features 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 analysis 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.

Key words: power quality, phase space reconstruction, Non-negative Matrix Factorization（NMF）, image recognition