| 引用本文: | 黄新波,马玉涛,朱永灿.基于信息融合和M-RVM的变压器故障诊断方法[J].电力自动化设备,2020,40(12): |
| HUANG Xinbo,MA Yutao,ZHU Yongcan.Transformer fault diagnosis method based on information fusion and M-RVM[J].Electric Power Automation Equipment,2020,40(12): |
|
| 摘要: |
| 针对仅以油中溶解气体数据为主要依据的变压器故障诊断方法信息量不足以及传统证据理论的缺陷问题，研究了基于信息融合和多分类相关向量机(M-RVM)的变压器故障诊断模型。首先，将油中溶解气体分析数据与电气试验数据作为诊断模型的输入特征量向量，更真实地反映变压器的故障信息。然后，采用4个M-RVM作为分类器，对故障进行初步诊断，并将诊断结果分别转化为证据融合所需证据体，同时引入兰式距离函数与光谱角余弦函数对证据体进行修正。最后，采用改进冲突再分配策略进行决策融合，避免融合过程中出现证据互相矛盾的现象。对比分析结果表明，基于多源信息融合的变压器诊断模型相较单一特征参数诊断以及单一诊断算法具有更高的诊断准确率。 |
| 关键词: 变压器 故障诊断 证据理论 多分类相关向量机 信息融合 |
| DOI：10.16081/j.epae.202009040 |
| 分类号:TM41 |
| 基金项目:陕西省重点研发计划项目(2018ZDXM-GY-040) |
|
| Transformer fault diagnosis method based on information fusion and M-RVM |
|
HUANG Xinbo1, MA Yutao2, ZHU Yongcan1
|
|
1.College of Electronics and Information, Xi’an Polytechnic University, Xi’an 710048, China;2.Yan’an Power Supply Company of State Grid Shaanxi Electric Power Company, Yan’an 716000, China
|
| Abstract: |
| Aiming at the problems of insufficient information in transformer fault diagnosis method based only on dissolved gas data in oil and the defects of traditional evidence theory, a transformer fault diagnosis model based on information fusion and M-RVM(Multi-classification Relevant Vector Machine) is studied. Firstly, the DGA(Dissolved Gas Analysis) data and electrical test data are used as the input eigenvectors of the diagnostic model, which can reflect the fault information of transformer more truly. Secondly, four M-RVMs are used as classifiers to diagnose the faults, and the diagnostic results are transformed into the evidences needed for evidence fusion. At the same time, the blue distance function and spectral angle cosine function are introduced to correct the evidences. Finally, the improved conflict redistribution strategy is used for decision fusion, so as to avoid conflicting evidence in the fusion process. The comparative analysis results show that the transformer diagnosis model based on multi-source information fusion has higher diagnostic accuracy than single feature parameter diagnosis and single diagnostic algorithm diagnosis. |
| Key words: power transformers fault diagnosis evidence theory multi-classification relevant vector machine information fusion |
