基于SVMD-CMSEE与GSA-SVM的新型电力系统变压器故障状态智能诊断方法

李峰; 陈皖皖; 李晓华; 夏能弘

李峰,陈皖皖,李晓华,夏能弘.基于SVMD-CMSEE与GSA-SVM的新型电力系统变压器故障状态智能诊断方法[J].电测与仪表,2024,61(12):17-25.
LI Feng,CHEN Wanwan,LiI Xiaohua,XIA Nenghong.An intelligent fault diagnosis method for transformer in novel power system based on SVMD-CMSEE and GSA-SVM[J].Electrical Measurement & Instrumentation,2024,61(12):17-25.

基于SVMD-CMSEE与GSA-SVM的新型电力系统变压器故障状态智能诊断方法

An intelligent fault diagnosis method for transformer in novel power system based on SVMD-CMSEE and GSA-SVM

DOI：10.19753/j.issn1001-1390.2024.12.003

中文关键词: 变压器振动信号故障诊断复合多尺度能量熵逐次变分模态分解支持向量机

英文关键词:transformer vibration signal, fault diagnosis, CMSEE, SVMD, SVM

基金项目:国家自然科学基金资助项目（ 51607110）、国网上海市电力公司科技项目（B3094020000L）

作者中文名	作者英文名	单位
李峰	LI Feng	上海电力大学电气工程学院，上海 200090
陈皖皖	CHEN Wanwan	上海电力大学电气工程学院，上海 200090
李晓华	LiI Xiaohua	上海电力大学电气工程学院，上海 200090
夏能弘	XIA Nenghong	上海电力大学电气工程学院，上海 200090

摘要点击次数: 425

中文摘要:

新型电力系统在促进“碳中和，碳达峰”的目标实现的同时，对电力系统中变电设备的可靠运行提出了新的挑战。为进一步提高变压器机械故障的识别精度，文中从变压器的振动特性出发，提出一种基于SVMD-CMSEE与GSA-SVM的新型电力系统变压器故障状态智能诊断方法。采用逐次变分模态分解(successive variational modal decomposition, SVMD)算法自适应性地分解出变压器振动信号的各模态分量，联合复合多尺度能量熵(combining compound multi-scale energy entropy,CMSEE)提取了振动信号的时频分布变化特征，并引入类间区分度确定了特征中的最优特征子集，通过引力搜索算法(gravity search algorithm,GSA)对支持向量机(support vector machine,SVM)的关键参数进行优化，构造了基于GSA-SVM的变压器故障识别模型。对某10 kV油浸式变压器振动信号的计算结果表明：基于SVMD-CMSEE算法得到的变压器振动信号复合特征能有效估计时间序列的动态变化，所提出的GSA-SVM诊断模型具有较高的识别精度和计算效率，准确率可达98%，从而为基于振动信号的变压器状态监测提供了技术支撑。

英文摘要:

The novel power system not only promotes the realization of the goal of carbon neutrality and carbon emissions peak, but also poses new challenges to the reliable operation of substation equipment in the power system. In order to further improve the identification accuracy of transformer mechanical faults, an intelligent fault diagnosis method for transformer in novel power system based on SVMD-CMSEE and GSA-SVM is proposed. Each modal component of transformer vibration signal is decomposed adaptively by successive variational modal decomposition (SVMD) algorithm. The time-frequency distribution characteristics of vibration signals are extracted by combining compound multi-scale energy entropy (CMSEE), and the optimal feature subset is determined by introducing inter-class discrimination. Finally, the key parameters of support vector machine (SVM) are optimized by gravity search algorithm (GSA), and a transformer fault identification model based on GSA-SVM is constructed. The calculation results of vibration signal of a 10 kV oil-immersed transformer show that the composite features of transformer vibration signals based on SVMD-CMSEE algorithm can effectively estimate the dynamic changes of time series. The proposed GSA-SVM diagnosis model has high recognition accuracy and computational efficiency, and the accuracy can reach 98%, which provides technical support for transformer state monitoring based on vibration signals.

查看全文查看/发表评论下载PDF阅读器

关闭