基于SA-APSO及降维投影-动态加权统计的局部放电定位方法

A Positioning Approach Based on SA-APSO and Reduced Dimension-Dynamic Weighting Statistics for Partial Discharge Source

在进行变压器局部放电(partial discharge, PD)源的定位过程中,存在时延估计测量误差、定位算法近似估计及电磁波信号在传播过程中存在绕射现象等各级误差,导致难以准确定位PD源。为此,提出一种基于模拟退火-自适应混合粒子群优化(simulated annealing-adaptive particle swarm optimization,SA-APSO)及降维投影-动态加权统计的PD定位方法。利用标准测试函数对所提APSO算法进行测试,验证了其在稳定度及精度上均优于普通PSO算法。在变压器PD模拟实验平台,通过SA-APSO方法提取多组PD数据。将三维空间点投影在二维平面进行聚类处理,利用圆内点密度设定阈值排除偏离点,并通过动态加权统计的方式获取平面PD源,进行均值处理后回归空间量。实验结果表明,经过降维投影-动态加权统计处理后的PD定位精度明显提高,定位误差降至5.86 cm。

Due to the measurement errors of time delay estimation,approximate estimation of location algorithm and the diffraction problem of electromagnetic wave in signal propagation,it is hard to accurately locate partial discharge(PD)source in the transformer.Therefore,this paper proposes a PD location method based on simulated annealing-adaptive particle swarm optimization(SA-APSO)and reduced dimension-dynamic weighting statistics.By using the standard functions to test the proposed APSO algorithm,it verifies the proposed method is superior to the ordinary PSO algorithm in stability and accuracy.Based on a transformer experimental platform for PD,multi-group PD data are extracted by this method.Three-dimensional space points are projected on the two dimensional plane for clustering process,and the interior point density is used to set the threshold values and exclude the deviation points.Meanwhile,by means of the dynamic weighting,it can obtain the plane PD source for determining the mean value as the final local source.The experimental results show that the positioning accuracy of the method is significantly improved after reduced dimension and dynamic weighting statistics,and the positioning accuracy reduces to 5.86 cm.