Research on Multiple Partial Discharge Signals Separation and Reconstruction Algorithm

The reconstruction of partial discharge (PD) signals in transformer is helpful for monitoring the operation performance of transformer, and it increases the stability and reliability of quality of power supply. The PD signals are reconstructed by under-determined blind source signal analysis methods. Firstly, the number of PD sources is analyzed by the detected time sequences and the singular spectrum analysis. Then, according to the sparsity of PD signals, K-means clustering algorithm and the number of PD signals, a separation matrix is analyzed. Finally, based on the detected time sequences and separation matrix, the PD signals are reconstructed by using the shortest path method. And the results of an example shows the effectiveness of the proposed method.