An NPE-PCA-based Detection Method of Distribution Transformer Closing Surge Waveform

Abstract

In order to effectively detect the waveform characteristics of the closing inrush current of distribution transformers and distinguish between the magnetizing inrush current and the fault current when closing, this paper proposes a new method based on neighborhood preserving embedding (NPE) and principal component analysis (PCA) transformer closing inrush current detection method. This method can detect and process the global and local feature information of the data. First, the NPE-PCA algorithm is used to reduce the current data to two-dimensional space, and then the fitting error σ is obtained by fitting the two-dimensional space data. The relationship between σ and a given threshold is used to identify the magnetizing inrush current when closing. Finally, a model is built on the ATP/EMTP platform to test the proposed method for detecting the waveform characteristics of the closing inrush current. The simulation results show that the NPE-PCA inrush current waveform detection algorithm proposed in this paper can effectively identify the waveform characteristics of the transformer closing inrush current, which is consistent with the second harmonic wave. Algorithm comparison analysis shows that the performance of this algorithm is better.