摘要
考虑变压器故障诊断的不确定性,构建了变压器模糊聚类模型,提出了用矩阵特征值分析方法得出样本集最佳分类数,实现了无监督的故障诊断。针对模糊C均值算法用于变压器故障诊断存在的问题,提出用人群搜索算法(SOA)得到较优的初始聚类中心。SOA算法是一种新型的启发式智能算法,克服了粒子群算法(PSO)和遗传算法(GA)等智能优化算法收敛性差、局部寻优的缺陷。仿真结果表明,该算法收敛速度更快,且具有更好的全局搜索能力,比传统的智能算法具有更高的有效性和鲁棒性,为变压器故障诊断聚类分析提供了参考。
A fuzzy clustering model is presented considering the uncertainty of transformer fault diagnosis, and a matrix eigenvalue analysis method is proposed to estimate the correct number of clusters which can implement the unsupervised fault diagnosis. Aiming at the problem existed in fuzzy c - means clustering algorithm which is applied to transformer fault diagnosis, seeker optimization algorithm （ SOA ） is introduced to obtain the optimized initial clustering center. SOA simulates human randora search behavior and overcomes the defects of particle swarm optimization （PS0） and genetic algorithm （ GA ） with local search and poor convergence. Simulation results show that SOA has a higher convergence speed and a better global searching ability. Comparing with the traditional intelligent optimization algorithms, SOA is more effective and robust, which can give a reference for transformer fault diagnosis.
作者
陶飞达
吴杰康
曾振达
梁浩浩
邹志强
张丽平
黄智鹏
杨夏
Tao Feida;Wu Jiekang;Zeng Zhenda;Liang Haohao;Zou Zhiqiang;Zhang Liping;Huang Zhipeng;Yang Xia(Heyuan Power Supply Bureau,Guangdong Power Gild Corporation,Heyuan 517000,Guangdong,China;School of Automation,Guangdong University of Technology,Guangzhou 510006,Guangdong,China)
出处
《四川电力技术》
2018年第3期1-5,24,共6页
Sichuan Electric Power Technology
基金
国家自然科学基金项目(51567002
50767001)
广东省公益研究与能力建设专项资金项目(2014A010106026)
广东电网有限责任公司科技项目(031600KK52160004)
关键词
变压器故障诊断
无监督聚类
矩阵特征值分析
人群搜索算法
transformer fault diagnosis
unsupervised clustering
matrix eigenvalue analysis
seeker optimization algorithm
