Fault early warning method of wind turbine generator based on WOA-GRU
In order to realize the early capture of wind turbine generator fault and improve the accuracy of fault warning, a fault warning method of wind turbine generator based on WOA-GRU model was proposed. Firstly, outlier data of wind turbine generator temperature was removed through box plot analysis, an...
Saved in:
Published in | Dianji yu Kongzhi Xuebao = Electric Machines and Control Vol. 29; no. 6; p. 54 |
---|---|
Main Authors | , , , , |
Format | Journal Article |
Language | Chinese English |
Published |
Harbin
Harbin University of Science and Technology
01.01.2025
|
Subjects | |
Online Access | Get full text |
ISSN | 1007-449X |
DOI | 10.15938/j.emc.2025.06.006 |
Cover
Summary: | In order to realize the early capture of wind turbine generator fault and improve the accuracy of fault warning, a fault warning method of wind turbine generator based on WOA-GRU model was proposed. Firstly, outlier data of wind turbine generator temperature was removed through box plot analysis, and the grey correlation analysis method was applied to extract feature parameters with high correlation with wind turbine generator temperature from high-dimensional SCADA data as model inputs. Subsequently, by using the whale optimization algorithm to optimize the hyperparameter group of the gated recurrent unit neural network, the obtained optimal parameter gated recurrent unit neural network model was used to predict the temperature of the wind turbine generator. Based on setting the alarm threshold through adaptive threshold algorithm, fault warning of wind turbine generators was carried out accordingly. Finally, taking the SCADA data of a wind turbine unit in a domestic wind farm as an example, the WOA-GRU model was compared with BP, ELM, RF, GRU, and LSTM models. The results show that the WOA-GRU model had higher prediction accuracy than other models and could more accurately capture early faults of wind turbine generators. |
---|---|
Bibliography: | ObjectType-Article-1 SourceType-Scholarly Journals-1 ObjectType-Feature-2 content type line 14 |
ISSN: | 1007-449X |
DOI: | 10.15938/j.emc.2025.06.006 |