统计参数用于局部放电模式识别的研究

Statistical Parameter Method for PD Pattern Recognition

通过研究局部放电二维谱图的统计特性 ,对局部放电脉冲幅值谱图进行 Weibull分析 ,获得尺度参数 α和形状参数 β,从局部放电脉冲相位谱图中估计出统计算子 Sk、Ku ,用以上统计参数作为人工神经网络的输入 ,实现局部放电的模式识别。对设计的 5种典型人工绝缘缺陷进行了大量模拟实验 。

A new PD characteristic vector generation method using of statistical parameters for the artificial neural network input is presented in this paper. According to the statistical properties of PD 2-D spectrums, scale parameter α and shape parameter β of Weibull parameter estimated from PD pulse is amplitude distribution, together with statistical operators S k and K u estimated from PD pulse phase distribution, are used to generate the artificial neural network inputs which is applied to the PD pattern recognition. A lot of simulation experiments of 5 typical artificial defects have been done, and the results of statistical parameter method and data array method have shown that statistical parameter method is feasible to recognize PD patterns