Premium
A transfer learning fault diagnosis model of distribution transformer considering multi‐factor situation evolution
Author(s) -
Yang Zhichun,
Shen Yu,
Zhou Renfei,
Yang Fan,
Wan Zilin,
Zhou Zhiqiang
Publication year - 2020
Publication title -
ieej transactions on electrical and electronic engineering
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.254
H-Index - 30
eISSN - 1931-4981
pISSN - 1931-4973
DOI - 10.1002/tee.23024
Subject(s) - transformer , computer science , fault (geology) , data mining , engineering , algorithm , electrical engineering , voltage , seismology , geology
Aiming at the problem of limited fault data and data expiration of distribution transformers, a transfer learning fault diagnosis model of the distribution transformer considering multi‐factor situation evolution is proposed in this paper. First, the state quantities that influence the distribution transformer operation are constructed, and fuzzy binary quantification is used for the state quantities. The association between the state quantities and the fault is explored by the fuzzy Apriori algorithm, and the key state variables that induce the transformer fault are extracted. The Tanimoto coefficient is introduced for the limited fault data of distribution transformers, and the effective auxiliary fault data is transferred to the target distribution transformer. On this basis, the fault diagnosis model based on information transfer is proposed. A health index is introduced to describe the distribution health levels, and the auxiliary fault data in different health levels is transferred because data has expired. On this basis, the fault diagnosis model for data expiration is proposed. The weights of the target and auxiliary fault data in the above model are iteratively solved by using the TrAdaBoost algorithm. Finally, an example analysis is carried on the basis of distribution transformer fault data, and the simulation results show that the fault diagnosis accuracy of the model is high and that it has stronger generalization ability than traditional diagnosis models. © 2019 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.