基于特征提取和AHC的变压器数据异常检测方法
CSTR:
作者:
作者单位:

(东华大学信息科学与技术学院 ,上海 201620)

作者简介:

通讯作者:

齐金鹏(1977—),男,博士,副教授,主要从事大数据分析、模式识别研究;E-mail:qipengkai@dhu.edu.cn

中图分类号:

TP311

基金项目:

国家自然科学基金(61305081,61104154);上海市自然科学基金(16ZR1401300,16ZR1401200)


Transformer data anomaly detection method based on feature extraction and AHC
Author:
Affiliation:

(College of Information Science and Technology , Donghua University , Shanghai 201620, China)

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    油中溶解气体分析 (dissolved gas analysis,DGA)是评估变压器运行状态的重要指标,其异常变化可预示存在潜在故障。针对变压器监测数据信息冗余、缺失及孤立噪声等问题,提出一种基于特征提取和凝聚式层次聚类(agglomerative hierarchical clustering,AHC)的变压器数据异常检测方法。首先,对DGA中的气体浓度数据中的缺失数据采用均值插补修正,进行 Z-score归一化处理;其次,采用改进滑动窗口策略与时间序列变换 (time series transformation,TST)算法对数据进行特征提取并形成特征矩阵;最后,采用基于密度的 AHC方法进行聚类,根据聚类结果对变压器运行状态进行综合分析。实验结果显示,该方法对变压器运行状态异常识别的准确率可达98.91%,相较于 Fixed-TST 算法和 k-近邻(k-nearest neighbor,kNN)算法分别高出 11.06个百分点与 8.50个百分点,表明该方法能够有效提取关键特征,降低数据的复杂度,为变压器故障预警提供一种分析方法。

    Abstract:

    Dissolved gas analysis (DGA) is an important indicator for evaluating the operating status of transformers,and its abnormal changes can indicate potential faults.To solve the problems of redundancy,missing data,and isolated noise in transformer monitoring data,a transformer data anomaly detection (AD) method based on feature extraction and agglomerative hierarchical clustering (AHC) is proposed.Firstly,the missing data in the DGA ’s gas concentration data are supplemented and corrected by mean interpolation,followed by Z-score normalization.Secondly,the improved sliding window strategy and time series transformation (TST) algorithm are adopted to extract features from the data and construct the feature matrix.Finally,a density-based AHC method is employed for clustering,and the operating status of transformers is comprehensively analyzed based on the clustering results.The experimental results show that the accuracy of this method in identifying abnormal transformer operating status can reach 98.91%,which is 11.06 percentage points and 8.50 percentage points higher than that of the Fixed-TST algorithm and k-nearest neighbor (kNN) algorithm,respectively.This indicates that this method can effectively extract key features,reduce data complexity,and provide an analytical approach for transformer fault early warning.

    参考文献
    相似文献
    引证文献
引用本文

黄莉娜,齐金鹏,戴理,等.基于特征提取和AHC的变压器数据异常检测方法[J].电力科学与技术学报,2025,40(5):14-23.
HUANG Lina, QI Jinpeng, DAI Li, et al. Transformer data anomaly detection method based on feature extraction and AHC[J]. Journal of Electric Power Science and Technology,2025,40(5):14-23.

复制
分享
相关视频

文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2024-04-25
  • 最后修改日期:
  • 录用日期:
  • 在线发布日期: 2025-12-18
  • 出版日期:
文章二维码