基于SVDD和改进K-Means的变压器故障诊断模型
Transformer Fault Diagnosis Model Based on SVDD and Improved K-Means
谢旭钦 1刘泉辉 1赵湘文 1张清松 1林剑雄 1张帆1
作者信息
- 1. 中国南方电网有限责任公司 广东广州增城供电局,广州 广东 510000
- 折叠
摘要
变压器状态对于智能配电房的安全稳定运行具有重要意义.为实现对变压器故障的准确诊断,在变压器油中溶解气体分析(DGA)的基础上,提出了一种联合使用支持向量数据描述(SVDD)和改进K-Means聚类的变压器故障诊断方法.首先利用 SVDD构造闭合分类曲面实现"正常"和"故障"两类判断,然后对"故障"类样本进行K-Means聚类分析,自动将其划分为低能放电、中低温过热、高能放电、高温过热和局部放电 5 种故障类型,同时针对 K-Means 初始聚类中心选取难题,提出局部密度概念自动确定 K-Means初始聚类中心,提升聚类性能.最后利用变压器故障真实数据开展实验,结果表明,相较于支持向量机(SVM)和BP神经网络模型,所提方法的故障诊断准确率分别提升 9.8%和 8%.
Abstract
The operation status of transformer is of great significance to the stability and reliability of intelligent distri-bution room.In order to realize the accurate diagnosis of transformer faults,based on the analysis of dissolved gases in transformer oil,a multi-classifier joint fault diagnosis method based on the combined use of support vector data description(SVDD)and improved K-Means clustering is proposed.First,SVDD is used to construct a closed classification surface to realize"normal"and"fault"judgments.Then K-Means clustering analysis is carried out on the"fault"samples,which are automatically divided into five types:low energy discharge,medium and low temperature overheat,high energy discharge,high temperature overheat and partial discharge.At the same time,the concept of local density is proposed to automatically determine the initial clustering center of K-Means to improve the clustering performance.Finally,the transformer fault data of the intelligent distribution room is used to carry out the verification experiment.The results show that compared with the traditional support vector machine(SVM)and BP neural network model,the fault diagnosis accuracy of the proposed meth-od is improved by 9.8%and 8%,respectively.
关键词
智能配电房/变压器故障诊断/油中溶解气体分析/支持向量数据描述/多分类器联合Key words
intelligent distribution room/transformer fault diagnosis/analysis of dissolved gas in oil/support vector da-ta description/multi-classifier association引用本文复制引用
基金项目
中国南方电网有限责任公司科技项目(082900KK52190001)
出版年
2024
国家科技期刊平台
NSTL
万方数据