[1]邹光涛,龚石磊.基于油中溶解气体含量特征的变压器绝缘故障XGBoost识别算法[J].机械与电子,2026,44(03):96-101.
　ZOU Guangtao,GONG Shilei.XGBoost Identification Algorithm for Transformer Insulation Faults Based on the Characteristics of Dissolved Gas Content in Oil[J].Machinery & Electronics,2026,44(03):96-101.
点击复制

基于油中溶解气体含量特征的变压器绝缘故障XGBoost识别算法()
分享到：

《机械与电子》[ISSN:1001-2257/CN:52-1052/TH]

卷:
44
期数:
2026年03期
页码:
96-101
栏目:
电力控制
出版日期:
2026-03-25

文章信息/Info

Title:
XGBoost Identification Algorithm for Transformer Insulation Faults Based on the Characteristics of Dissolved Gas Content in Oil
文章编号:
1001-2257(2026)03-0096-06
作者:
邹光涛; 龚石磊
(国网江西省电力有限公司九江供电公司,江西 九江 332000)
Author(s):
ZOU Guangtao; GONG Shilei
(Jiujiang Power Supply Branch,State Grid Jiangxi Electric Power Co.,Ltd.,Jiujiang 332000,China)
关键词:
油中溶解气体; 含量特征; 变压器; 绝缘故障识别
Keywords:
dissolved gases in oil; content characteristics; transformer; insulation fault identification
分类号:
TM407
文献标志码:
A
摘要:
针对复杂工况下的变压器绝缘故障漏检率高的问题,提出了基于油中溶解气体含量特征的变压器绝缘故障极限梯度提升(XGBoost)识别算法。利用油色谱仪采集变压器的油中溶解气体含量,构造特征谱图,从中提取均值、偏斜度和突出度等特征。选取加权核主成分分析方法对所提取的油中溶解气体含量特征进行低维映射,将低维映射结果输入XGBoost算法。XGBoost算法以分类与回归树(CART)作为基分类器,通过梯度提升方法集成多棵CART树,形成高精度的分类器。该算法通过迭代拟合残差来优化分类性能,构建正则化目标函数,通过加法学习方式,输出精准的变压器绝缘故障类型。实验结果表明,该算法能够精准识别电弧放电等多种变压器绝缘故障,识别置信度稳定在0.9~1.0区间。
Abstract:
Aiming at the problems of high missed detection rate in transformer insulation faults under complex working conditions,an eXtreme Gradient Boosting (XGBoost) identification algorithm for transformer insulation faults based on the characteristics of dissolved gas content in oil is proposed.The dissolved gas content in transformer oil is collected using an oil chromatograph,and the characteristic spectrum graphs are constructed,which features such as mean value,skewness,and kurtosis are extracted.The weighted kernel principal component analysis method is selected to perform the low dimensional mapping on the extracted characteristics of dissolved gas content in oil.The low dimensional mapping results are then input into the XGBoost algorithm.Using Classification and Regression Trees (CART) as the base classifiers,the XGBoost algorithm integrates multiple CART through gradient boosting method to form a high precision classifier.This algorithm optimizes classification performance by iteratively fitting residuals,constructs a regularized objective function,and outputs accurate transformer insulation fault types through additive learning.The experimental results show that the proposed algorithm can accurately identify various transformer insulation faults,such as arc discharge,and the recognition confidence is stable in the range of 0.9-1.0.

参考文献/References:

[1] 廖才波,杨金鑫,邱志斌,等.一种基于夏普利值及油中溶解气体分析的可解释变压器故障诊断方法[J].电网技术,2024,48(4):1752 1762.
[2] 息佳琦,石晓楠,杨昭,等.基于SSA XGBoost的变压器故障诊断[J].黑龙江大学自然科学学报,2023,40(6):721 729.
[3] 王世泽,张英敏.无载调压变压器绕组匝间绝缘状态评估模型[J].计算机仿真,2022,39(3):108 112.
[4] 罗传仙,朱晔,周正钦,等.高热故障下变压器油纸绝缘系统的产气机理研究[J].绝缘材料,2024,57(5):86 94.
[5] 廖晓青,陈历,许建远,等.基于残差注意力自适应去噪网络和Stacking集成学习的局部放电故障诊断[J].电子技术应用,2024,50(11):66 73.
[6] 杨鑫,王渝红,莫思特.基于变分模态分解的变压器绝缘缺陷故障检测[J].电子设计工程,2023,31(21):60 64.
[7] 马高权,周娜,谢蒙飞,等.基于GWO KELM 模型的变压器油纸套管典型绝缘故障辨识方法[J].电网与清洁能源,2023,39(5):38 48.
[8] 赵跃,马凤翔,王安静,等.基于荧光双色比例的变压器绝缘油放电击穿故障诊断模型研究[J].光谱学与光谱分析,2022,42(4):1134 1138.
[9] 阮义,张浩天,孙建,等.基于DGA 的NRBO XGBoost变压器故障诊断方法[J].巢湖学院学报,2024,26(6):87 93.
[10] 曹铎耀,郭若琛,蔡玲珑,等.双指数及振荡型操作冲击电压下变压器匝间击穿特性[J].高电压技术,2022,48(6):2265 2275.
[11] 韩金华,马云飞,王立楠,等.电力变压器绝缘油色谱氢气超标故障的分类分析[J].变压器,2023,60(2):26 31.
[12] 龚泽威一,饶桐,王钢,等.基于改进粒子群优化XGBoost的变压器故障诊断方法[J].高压电器,2023,59(8):61 69.
[13] 张弦亮,郝建,刘丛,等.变压器绕组热缺陷温度演化规律及其辨识方法[J].高电压技术,2024,50(4):1548 1559.
[14] 张容槟,徐耀松,牛元平.基于WKPCA 与IEDO XGBoost的变压器故障诊断方法研究[J].电工电能新技术,2024,43(10):24 42.
[15] 杜江,范志远,范仲华,等.电力变压器油中溶解气体异常数据识别与含量预测研究[J].电网技术,2025,49(2):844 853.

备注/Memo

备注/Memo:
收稿日期:2025-09-01
基金项目:国网江西省电力有限公司科技项目(5218D0250004)
作者简介:邹光涛 (1977-),男,江西奉新人,高级工程师,研究方向为高电压技术、变电运检等;龚石磊 (1989-),男,湖北仙桃人,硕士,中级工程师,研究方向为高电压技术、变电运检等。
更新日期/Last Update: 2026-04-29