基于多维信息的变压器绝缘在线监测数据分析方法  

Research on data processing method of transformer insulation on-line monitoring based on multidimensional information

作　　者：宋建[1] 董俊[1] 单节杉[1] 常勇[1] SONG Jian;DONG Jun;SHAN Jie-shan;CHANG Yong(Faculty of Electric Power Engineering,Kunming University of Science and Technology,Kunming 650500,China)

机构地区：[1]昆明理工大学电力工程学院,云南昆明650500

出　　处：《云南民族大学学报（自然科学版）》2022年第4期458-464,共7页Journal of Yunnan Minzu University:Natural Sciences Edition

基　　金：国家自然科学基金(51667010);云南省重大科技专项计划(202002AF080001);昆明理工大学虚拟仿真实验教改(2020年).

摘　　要：针对变压器绝缘老化问题进行过程性分析,提出基于数据适应性的绝缘变化过程建模.以多维度绝缘变化信息为主要变量,考虑数据驱动下,不同维度的数据之间的适应性变化,建立统一评价体系下的在线测量模型.采用核主成分分析和支持向量机做数据整体特征挖掘,典型相关分析算法做局部特征分析,融合变压器绝缘状态信息,分别辨识聚合模型的全局及局部特征.基于实测数据条件对模型进行校验,从而实现变压器绝缘状态评估的在线分析.The insulation aging process of transformer is analyzed,and the insulation change modeling based on data adaptability is proposed.Taking multi-dimensional insulation information as the main variable and data-driven as the core,an on-line measurement model based on unified evaluation system is established.The kernel principal component analysis(KPCA)and Support vector machine are used to do the global feature mining,the canonical correlation analysis(CCA)algorithm is used to do the local feature analysis,and the insulation state information of transformer is fused to identify the global and local features of the aggregated model.Based on the measured data and running several algorithms to verify the model,the on-line analysis of transformer insulation state evaluation is realized.

关 键 词：变压器 多维度信息 在线监测 数据驱动 特征挖掘 

分 类 号：TM855[电气工程—高电压与绝缘技术] TM41