变压器油纸绝缘多参量诊断与失效概率研究

Study of Multiple Parameters Diagnosis and Failure Probability for Transformer

为了准确诊断油纸绝缘的老化状态,对多组试样进行加速老化实验,并测量多个非破坏性参量。通过显著性检验,筛选出与绝缘纸聚合度最相关的参量,进而建立绝缘纸聚合度的多元线性回归预测模型,并分析聚合度与变压器失效概率之间的关系。结果表明:参量中糠醛含量的对数、一氧化碳、二氧化碳、水分、氢气与绝缘纸的聚合度相关性较大,利用上述参量建立了预测聚合度的多元线性回归模型,拟合优度为0.93。当聚合度大于400时,变压器的失效概率几乎为零,而当聚合度小于400时,变压器的失效概率随聚合度的降低而不断增加。

In order to evaluate the ageing status of oil-paper insulation accurately, we conducted accelerat- ed ageing experiments on the multi-group samples and measured their several non-destructive parameters. The most relevant parameters to degree of polymerization（DP） of insulation paper were selected through significance test, and a multiple linear regression forecast model was built to analyze the relationship between DP and failure probability of transformer. The results show that the logarithmic of furfural content, CO, CO2, H2O, and H2 are more correlative with DP. The multiple linear regression model based on the above parameters was built, and the goodness of fit is 0.93. When the DP is above 400, the failure probability of transformer is nearly zero. When the DP is below 400, the failure probability increases with the decrease of DP.