全球能源互联网（英文）2022，Vol.5Issue(6)：654-665,12.DOI:10.1016/j.gloei.2022.12.007
基于RF-组合赋权与灰色关联改进TOPSIS法的变压器油纸绝缘状态的综合评估
Comprehensive evaluation of the transformer oil-paper insulation state based on RF-combination weighting and an improved TOPSIS method
摘要
Abstract
The accurate identification of the oil-paper insulation state of a transformer is crucial for most maintenance strategies. This paper presents a multi-feature comprehensive evaluation model based on combination weighting and an improved technique for order of preference by similarity to ideal solution (TOPSIS) method to perform an objective and scientific evaluation of the transformer oil-paper insulation state. Firstly, multiple aging features are extracted from the recovery voltage polarization spectrum and the extended Debye equivalent circuit owing to the limitations of using a single feature for evaluation. A standard evaluation index system is then established by using the collected time-domain dielectric spectrum data. Secondly, this study implements the per-unit value concept to integrate the dimension of the index matrix and calculates the objective weight by using the random forest algorithm. Furthermore, it combines the weighting model to overcome the drawbacks of the single weighting method by using the indicators and considering the subjective experience of experts and the random forest algorithm. Lastly, the enhanced TOPSIS approach is used to determine the insulation quality of an oil-paper transformer. A verification example demonstrates that the evaluation model developed in this study can efficiently and accurately diagnose the insulation status of transformers. Essentially, this study presents a novel approach for the assessment of transformer oil-paper insulation.关键词
组合赋权/随机森林算法/绝缘老化评估/油纸绝缘/时域特征量Key words
Combined weight method/Random forest algorithm/Insulation aging assessment/Oil-paper insulation/Time-domain eigenvalue引用本文复制引用
宋福根,仝世超..基于RF-组合赋权与灰色关联改进TOPSIS法的变压器油纸绝缘状态的综合评估[J].全球能源互联网（英文）,2022,5(6):654-665,12.基金项目
This work was supported by the Natural Science Foundation of the Fujian Province(2021J01109). （2021J01109）
