电工技术学报2019，Vol.34Issue(24)：5104-5114,11.DOI:10.19595/j.cnki.1000-6753.tces.181894
基于混沌序列的变压器油色谱数据并行聚类分析
Parallel Clustering Analysis of Dissolved Gas Analysis Data Based on Chaotic Sequences
李恩文 1王力农 1宋斌 1方雅琪1
作者信息
- 1. 武汉大学电气与自动化学院 武汉 430072
- 折叠
摘要
Abstract
Transformer oil chromatographic analysis (DGA) is an important technical method for transformer operation and maintenance, and Fuzzy C-Means clustering is an important intelligent algorithm for DGA. However, the objective function of the clustering algorithm is a typical non-convex function and the optimization process is a hill-climbing algorithm based on local search. Consequently, the iterative process is easy to fall into local extremum, and the algorithm cannot classify effective oil chromatographic data. Chaotic variables have randomness and ergodicity, making global optimization possible. In this paper, the chaotic sequence is used to perform "artificial mutation" on the cluster center in the process of cluster iteration, and multiple optimization trajectories are set in parallel. In the iterative process, each of the optimized trajectories is deduced according to its own gradient information, and also the optimization information of the remaining trajectories is shared. The original optimization trajectory can be changed in the process of iterative optimization, thereby avoiding the optimization process from terminating at the local extreme point and achieving global optimization. The case analysis shows that the proposed method achieves the global optimization of clustering analysis and improves the ability of fuzzy clustering algorithm to identify DGA fault patterns, which has practical application value.关键词
全局收敛/模糊聚类分析/变压器/故障诊断/溶解气体分析Key words
Global convergence/ fuzzy clustering analysis/ transformer/ fault diagnosis/ dissolved gas analysis分类
信息技术与安全科学引用本文复制引用
李恩文,王力农,宋斌,方雅琪..基于混沌序列的变压器油色谱数据并行聚类分析[J].电工技术学报,2019,34(24):5104-5114,11.
