Open AccessA Prediction Method for Power Transformer State Parameters Based on Grid Long Short-Term Memory Network
Author(s) -
Jufeng Dai,
Hui Song,
Xiaoqi Wan,
Yingjie Yan,
Gehao Sheng,
Xiuchen Jiang
Publication year - 2019
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1346/1/012038
Subject(s) - power grid , transformer , computer science , reliability engineering , mean squared prediction error , grid , data mining , engineering , artificial intelligence , machine learning , power (physics) , mathematics , electrical engineering , physics , geometry , quantum mechanics , voltage
Power transformer state parameter prediction analysis can provide strongly technical support for equipment state assessment. The available transformer state parameter prediction models are mainly based on the very limited number of state parameters for analysis and judgment. So, the stability and the intelligence of prediction still need improvement. Based on the large amount of information of transformer equipment state, the data of environmental meteorological and grid operation, this paper proposes a transformer state parameter prediction method using deep mining of complex associations with the grid long short-term memory network (GLSTM). Association relationships captured from GLSTM are used to correct the prediction of state parameters. Finally, the method is applied to the top oil temperature trend prediction of a 500 kV transformer. The results show that the proposed method can mine and analyze the relationship between the influencing factors of equipment state. Compared with the prediction methods without considering the correlation and the traditional methods, the association relation extracted by GLSTM improves the stability of the prediction model and reduces the prediction error.