基于改进SVM的特高压变压器局部放电故障自动检测方法

Automatic Fault Detection Method of Partial Discharge in UHV Transformer Based on Improved SVM

针对特高压变压器局部放电故障的自动检测问题,传统方法通常独立针对异常区域进行实时核验,效率较低且误识率高。因此,提出了基于改进SVM的特高压变压器局部放电故障自动检测方法。该方法通过多层级方式提取变压器局部放电故障特征,强化检测效率,并设计多层级检测故障树。随后,构建改进SVM模型,结合迁移学习实现自动检测。测试结果显示,本方法具有更低的检测误识率,证明其具有高效性,可有效提升检测精度。

In view of the problem of automatic detection of partial discharge faults in UHV transformers,the current methods often carry out real﹣time verification independently for abnormal areas,which has low efficiency and high error rate.Therefore,an automatic fault detection method for partial discharge of UHV transformer based on improved SVM is proposed.Firstly,the fault characteristics of transformer partial discharge are extracted by multi﹣level method,the detection efficiency is strengthened,and the multi﹣level detection fault tree is designed.Then,an improved SVM model is constructed and automatic detection is realized with transfer learning.The test results show that this method has a lower detection error rate,which proves its high efficiency and significant improvement of detection accuracy.