Fault Diagnosis of Transformer Based on SAMME and DGA Method

HUANG Xinbo; LI Wenjunzi; SONG Tong; WANG Yanmei

doi:10.13296/j.1001-1609.hva.2016.02.003

您当前的位置：

首页 >

文章列表页 >

Fault Diagnosis of Transformer Based on SAMME and DGA Method

更新时间：2025-04-26

- Fault Diagnosis of Transformer Based on SAMME and DGA Method
- High VoltageApparatus Vol. 52, Issue 2, Pages: 13-18(2016)
- 作者机构：
  
  西安工程大学电子信息学院
- 作者简介：
- 基金信息：
- DOI：10.13296/j.1001-1609.hva.2016.02.003
  CLC： TM407
- Published：2016
- 稿件说明：
移动端阅览
[1]黄新波,李文君子,宋桐,王岩妹.基于DGA技术和SAMME的变压器故障诊断[J].高压电器,2016,52(02):13-18.

HUANG Xinbo, LI Wenjunzi, SONG Tong, et al. Fault Diagnosis of Transformer Based on SAMME and DGA Method[J]. High voltageapparatus, 2016, 52(2): 13-18.
[1]黄新波,李文君子,宋桐,王岩妹.基于DGA技术和SAMME的变压器故障诊断[J].高压电器,2016,52(02):13-18. DOI： 10.13296/j.1001-1609.hva.2016.02.003.

HUANG Xinbo, LI Wenjunzi, SONG Tong, et al. Fault Diagnosis of Transformer Based on SAMME and DGA Method[J]. High voltageapparatus, 2016, 52(2): 13-18. DOI： 10.13296/j.1001-1609.hva.2016.02.003.

摘要

油中溶解气体分析（dissolved gas analysis

DGA）技术可以有效发现充油变压器内部的早期潜伏性故障

是对充油设备进行诊断的常用检测方法。CART（classification and regression trees）分类回归树是一种可处理连续属性的不平衡算法。SAMME作为Ada Boost二分类算法的延伸

可将多个略好于随机猜测的弱分类器组合提升为分类精度更高的强分类器

完成多分类任务。针对单一算法往往不能满足实际工程对分类精度的需求且高精度算法难以获得的问题

采用SAMME算法对CART弱学习器的权重根据误差不断调整

再通过加权投票将其组合提升为强分类器

并采用V折交叉验证确定SAMME-CART模型的最优迭代次数。以此提高故障诊断模型的泛化能力。实验结果表明

运用该模型结合DGA技术对变压器故障进行诊断

相比于单一CART算法

诊断准确率整体提高了18.7%

表明该方法是可行的。

Abstract

Latent faults of oil immersed transformers can be discovered efficiently by means of DGA（dissolved gas analysis）method. This method has been very popular for oil immersed equipment. CART（classification and regression trees）is an unbalanced algorithm which can deal with continuous attributes. Ada Boost algorithm can be used to deal with two-class problem. As one of the expansions of the Ada Boost algorithm

SAMME can be used in multi-class problem by combining a number of weak classifiers whose performance is a little better than random guessing into a stronger classifier. In many cases

using single algorithm for the transformer failure classification often can’t meet the requirements of practical engineering. SAMME constantly adjusted the weight of each CART weak learner according to the error. Then

these weak learners were promoted to a strong classifier by weighted voting. Meanwhile

V-fold cross validation has been used to establish optimal number of iteration for SAMMECART model. In this way

the generalization ability of the fault diagnosis model has been improved. The results of simulation have shown that this model which based on DGA method is available in improving the accuracy rate of transformer faults diagnosis. Compared to single CART algorithm

the diagnostic accuracy rate has proved to be increased by 18.7%. Therefore

this method is available.

关键词

Keywords

references

Views

213

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Fault Diagnosis Model of DGA for Power Transformer Based on Kernel Fisher Discriminant Analysis Technology

Fault Diagnosis Model of Transformer Based on GJO Feature Optimization and AO-RF

Research on Diagnosis Method of Transformer Winding Deformation Fault Based on Dynamic Time Warping

New Approach to Transformer Fault Diagnosis Based on ANN

APPLICATION OF INFORMATION FUSION TECHNIQUE TO TRANSFORMER FAULT DIAGNOSIS

Related Author

WU Xiao-hui1

WANG Song1

FANG Xiao-ming1

LI Yan-mu2

LI Yan-ming1

YE Yulin

LIU Sen

HAN Xiaohui

Related Institution

School of Electrical Engineering, Xi’an Jiaotong University

School of Mechanical and Electrical,Xi’an University of Architecture and Technology

China Nuclear Power Engineering Co., Ltd.

School of Electrical Engineering, Hebei University of Science and Technology

Hebei Provincial Key Laboratory of Electromagnetic & Structural Performance of Power Transmission and Transformation Equipment, Baoding Tianwei Baobian Electric Co., Ltd.

⁰