To address the operational challenge of offshore wind turbines in the high-temperature and high-humidity environment of the southeastern coast
which are prone to overheating faults
a key temperature measurement point virtual perception and application method based on finite difference domain-deep neural networks (FDD-DNN) is proposed. Firstly
a high-dimensional abnormal operational data identification method integrating mechanism analysis and Isolation Forest algorithm
as well as a high-dimensional similar condition-based missing data imputation method
is proposed. Then
a finite difference regression vector that can characterize the operational conditions and time-delay characteristics of the wind turbine is defined
and a soft margin support vector machine (SSVM) is introduced to partition the differential dynamic operational domain. On this basis
a domain-based temporal dynamic modeling method based on DNN is proposed to achieve refined characterization and virtual perception of key temperature measurement points under full operational conditions. Finally
an engineering method for the deployment and application of the virtual perception model is proposed. Taking the main shaft temperature of offshore wind turbines as an example
the results show that the proposed method improves all evaluation metrics.
ZHANG S H, ZHANG T, LIU Y J.Early warning model of wind turbine front bearing based on Conv1D and LSTM[C]//2021 IEEE International Conference on Industrial Engineering and Engineering Management (IEEM). Singapore, Singapore, 2021: 1029-1034.
MANSOURI M, DHIBI K, HAJJI M, et al.Interval-valued reduced RNN for fault detection and diagnosis for wind energy conversion systems[C]//IEEE Sensors Journal. 2022: 13581-13588.
WANG Z, CHU X N.Operating condition identification of complete wind turbine using DBN and improved DDPG-SOM[C]//2022 IEEE 11th Data Driven Control and Learning Systems Conference (DDCLS). Chengdu, China, 2022: 94-101.