基于改进变分模态分解的有载分接开关机械状态监测

发布时间：2018-06-06 21:47

本文选题：有载分接开关 + 改进变分模态分解　；参考：《湖南大学学报(自然科学版)》2017年10期

【摘要】：为提高变压器有载分接开关(On-Load Tap-Changer,OLTC)机械状态智能诊断水平,提出了基于改进变分模态分解(Improved Variational Mode Decomposition,IVMD)-权重散度的OLTC机械状态特征提取方法,以及和声搜索算法(Harmony Search,HS)优化相关向量机(Relevance Vector Machine,RVM)的故障分类方法.本文进行模拟实验测得了多组不同工况下的OLTC机械振动信号,通过IVMD算法将振动信号分解为一系列有限带宽本征模态函数(Intrinsic Mode Function,IMF),计算IMF分量与原始振动信号的K-L散度(Kullback-Leibler Divergence,K-L),再乘上权重系数得到权重散度,以权重散度来表征多组OLTC机械振动信号的时频域复杂度.同时构建了RVM多分类模型,并通过和声搜索算法对RVM的核函数选择进行了优化,有效地实现了对于权重散度的分类.实验与数据分析结果表明,本文所提综合诊断方法精度较高,可准确提取机械故障特征,能够为OLTC智能故障诊断提供必要的参考.
[Abstract]:In order to improve the level of intelligent diagnosis of mechanical state of On-Load Tap-Changer-Based OLTC-based transformer with on-load tap switch, a method of feature extraction of OLTC mechanical state based on improved variational Mode decomposition (VMD) and improved Variational Mode Decomposition- weight divergence is proposed. And the harmony search algorithm (Harmony search HSV) is used to optimize the fault classification method of correlation vector machine (Relevance Vector Machine). In this paper, the mechanical vibration signals of OLTC are measured by simulation experiments. The vibration signal is decomposed into a series of intrinsic mode functions with limited bandwidth by IVMD algorithm. The K-L divergence of the IMF component and the original vibration signal is calculated and the Kullback-Leibler divergence K-Ln is obtained by multiplying the weight coefficient to obtain the weight divergence. The time-frequency complexity of multiple sets of OLTC mechanical vibration signals is characterized by weighted divergence. At the same time, the multi-classification model of RVM is constructed, and the kernel function selection of RVM is optimized by harmonic search algorithm, which effectively realizes the classification of weight divergence. The results of experiment and data analysis show that the comprehensive diagnosis method proposed in this paper has high accuracy, can extract mechanical fault features accurately, and can provide necessary reference for OLTC intelligent fault diagnosis.
【作者单位】：山东大学电气工程学院;山东省特高压输变电技术与装备重点实验室;
【基金】：国家自然科学基金资助项目(51207084) 山东省优秀中青年科学家科研奖励基金计划项目(BS2013NJ024)~~
【分类号】：TM41

【相似文献】