Localization and Identification of Partial Discharges in High Voltage Cables Based on Improved K-means Clustering

An improved K-means clustering algorithm is proposed to address the problems of inaccurate localisation due to errors in the arrival of partial discharges in high-voltage cables, and difficulties in map recognition due to interfering signals. In the localisation of partial discharge, by eliminating the second-order term of the non-linear system of equations and converting it into a linear system of equations, at least four high-frequency current sensors are used to measure and calculate the initial value of the PD under the consideration of the arrival error, and the improved K-means clustering is used to optimise the clustering of the initial value to obtain the optimal coordinates of the source of the PD, so as to accurately locate the PD. In the identification process of local discharge signals, a detection signal processing screening system is established to collect and process the signals, generate PRPD maps and T-F maps, and the T-F maps are clustered by the improved K-means clustering to classify the PRPD maps into two clusters of signals′ PRPD maps, and accurately identify the signals according to the comparison of the features of the library of PRPD source maps. The experimental results show that the method can effectively separate the local discharge signal from the interference signal and achieve accurate identification of the type of local discharge signal.