摘要
为解决通过油中溶解气体诊断变压器故障精确度不高的问题,提出了一种改进北方苍鹰优化(INGO)算法优化支持向量机(SVM)的故障分类模型。首先,采用主成分分析(PCA)法对油中溶解气体体积数据降维,去除冗余信息;然后,通过引入Singer混沌映射、改进的野马算法搜索机制、Lévy飞行策略多种方法改进北方苍鹰优化算法,再利用INGO算法对SVM核心参数进行优化;最后,将处理后的数据输入到INGO-SVM故障诊断模型中。结果表明,其诊断平均准确率为93.5%,与NGO、GWO、AO优化SVM相比,诊断平均准确率分别提升了3.34%、7.04%、10.12%。同时,该模型也优于极限学习机(ELM)、概率神经网络(PNN)、随机森林(RF)典型分类模型,验证了所建立的变压器故障诊断模型具有更高的精度和泛化能力。
In order to solve the problem of low fault accuracy of transformer with dissolved gas in oil,a fault classification model for improved northern eagle algorithm(INGO) optimization support vector machine(SVM) is proposed.Firstly,the principal component analysis method(PCA) is used to reduce the dimension of the dissolved gas volume data in the oil;Then,by introducing Singer chaotic mapping,improved wild horse algorithm search mechanism,Lévy flight strategy multiple methods is used to improve the Northern Goshawk optimization algorithm,and then using INGO algorithm to optimize the core parameters of SVM to improve its classification ability.Finally,the processed data is input into the INGO-SVM fault diagnosis model,and the results showed that the average diagnostic accuracy is 93.5%,which improved by 3.34%,7.04% and 10.12% compared with NGO,GWO and AO optimized SVM,respectively.At the same time,it is also better than the extreme learning machine(ELM),probabilistic neural network(PNN),and random forest(RF) typical classification model,which verifies that the established transformer fault diagnosis model has higher accuracy.
作者
陈忠华
王森
CHEN Zhonghua;WANG Sen(School of Electrical and Control Engineering,Liaoning Engineering and Technology University,Huludao 125105,China)
出处
《控制工程》
CSCD
北大核心
2024年第11期2010-2018,共9页
Control Engineering of China
关键词
变压器
故障诊断
数据处理
北方苍鹰优化算法
支持向量机
Transformer
fault diagnosis
data processing
Northern Goshawk optimization algorithm
SVM
