基于平面相交法的敞开式变电站多源局部放电定位方法

Localization Method for Multiple Partial Discharge Sources in Air-insulated Substation by Using Plane Intersection Method

目前基于特高频（UHF）天线阵列的敞开式变电站局部放电（以下简称局放）检测与定位方法获得应用,但定位距离误差偏大,且未提供有效的多放电源时的定位方法。为此提出了基于平面相交法的敞开式变电站多源局放定位方法：采用插值互相关法计算天线信号之间的时差,对互相关函数进行插值以提高采集系统采样率较低时的时差精度;引入平面相交法求解时差方程,并统计获得多次局放信号定位结果中概率密度最高的位置,将其作为最终的定位结果。当存在多个放电源时,使用K均值算法对天线阵列的时差向量进行聚类,实现多源UHF信号的分离。之后,建立了变电站局放检测与定位实验系统,对前述算法的定位精度进行了测试。结果表明,对尺寸为4 m×2 m的菱形天线阵,当局放源与距离4 m的天线对的法线夹角〈45°且距天线阵列中心距离〈10 m时,定位误差均〈1 m,多数情况下〈20 cm;在存在多个放电源时,该方法的定位精度与单个放电源时接近。

Recently, the ultra-high-frequency-based antenna array has been applied to partial discharge（PD） detection and localization in air-insulated substations（AIS）. The range of PD source is generally estimated with high inaccuracy in most of the related instruments, and a localization algorithm which is applicable to the cases of multiple sources is not developed yet. We put forward a plane intersection method to localize multiple PD sources which may occur simultaneously in AIS. The time difference of arrival（TDOA） between antennas is estimated with the cross-correlation（CC） method, in which the CC function is interpolated with minor time interval to improve the TDOA accuracy when an acquisition instrument with low sampling rate is adopted. Thereafter, the time difference equations are solved with the plane intersection method, and the final localization results are determined as the one with highest probability among numerous signals. In the case of multiple PD sources, the K Means clustering algorithm is applied to TDOA vector of antenna array to separate UHF signals. A PD detection and localization experimental system is developed and used to test the accuracy of the proposed algorithm. The results indicate that, for diamond array with size of 4 m×2 m and PD sources within range of 10 m and 45° to normal line of antenna pair with distance of 4 m, the localization error is less than 1 m in all cases and 20 cm in most cases. Moreover, the localization accuracy for multiple PD sources are close to that of one source.