电容器直流局部放电的统计指纹分析

Statistical Fingerprint Analysis for Capacitor PD under DC

针对电容器直流局部放电测试中的离散数据,借助统计理论,合理地对局部放电数据进行峰度、偏度、相关度、变异度等参量提取,然后利用BP网络对提取的参量进行识别,获得了较高的识别率,并发现选择不同的识别输入量,识别的结果有一定的不同,这对以后的深入研究有很大的参考价值。

According to statistical theory, this paper aims at discrete datum of PD measurement for capacitor under DC. Parameters of, kurtosis, skewness, correlation coefficient, variance coefficient etc, are extracted from the PD datum reasonably. Then the extracted parameters are recognized by using BP network, and it has a remarkable discrimination. It is found different recognition results with different inputs, which has tremendous values for future study.