油纸绝缘局部放电与油中产气规律的典型相关分析

Canonical Correlation Analysis of Dissolved Gas Production With Partial Discharges in Oil Paper Insulation

油中溶解气体分析方法(dissolved gas analysis,DGA)是对局部放电进行检测和诊断的重要手段之一,但是目前其准确度有一定局限性,主要原因是缺乏对油中溶解气体和变压器实际故障之间关系的研究。该文基于3种常见的油纸绝缘局部放电模型模拟实验,将典型相关分析方法应用于局部放电与油中产生气体的相关性研究。结果表明:典型相关分析法能定量地评估局部放电发展过程中的统计特征参数与油中溶解气体含量2组数据之间的相关性,以及不同参数向量对相关性的贡献,挖掘出每组参数中的主要特征向量,如沿面放电特征统计参数为偏斜度sk和峰值个数pk,特征气体为乙烯、乙烷。实验结果为局部放电产气机制研究和提高DGA法的准确度提供参考。

Dissolved gas analysis （DGA） is one of the important methods to measure and diagnose partial discharge （PD） defects. Its accuracy is limited because of the lack of correlation of oil dissolved gases with actual faults found in transformers. Based on the experiments of three typical PD models in oil paper insulation, a study on the correlation between statistical parameters of partial discharges and oil dissolved gas products was presented using canonical correlation analysis （CCA）. The result shows that the correlation between the two groups of parameters can be evaluated in a quantitative manner by this approach. In addition, the contribution of each parameter to the correlation was also calculated and the main parameters of each group were extracted, e.g. characteristic statistical parameters of surface discharge are skewness （Sk） and peak numbers （Pk） while characteristic parameters of dissolved gases are ethylene and ethane. The experimental result provides a reference for further research of gassing mechanism of partial discharges and the accuracy enhancement of DGA method.