.png){kind=icon}
.png){kind=icon}
Return
A deep learning based fatigue load predictive control method for wake wind farm
.png){kind=icon}
PRE
AI
DOI：10.1016/j.energy.2025.136123.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
This article explores the deep learning-based fatigue load predictive control approach of wake wind farm using high-fidelity wake simulations. Firstly, high-fidelity numerical wake simulation software named Simulator fOr Wind Farm Application is used to collect flow field data under the change of WT control actions (generator torque and pitch angle). Second, the bi-directional long short-term memory network is used to extract time evolution features of wake wind speed. The wind speed of each WT in multiple time steps in the future can be predicted. Then, a modified WT control model is applied to ensure that the generator torque and pitch angle are coordinated controlled. And it conforms to the control structure of SOWFA. Finally, model predictive wind farm cooperative control strategy is proposed to minimize the wake-induced mechanical dynamic load fluctuation and thus reduce the fatigue load. The early obtained predicted wake wind speed is added as previewed disturbances to predict the future dynamic behaviors of mechanical load. The simulation results under turbulent wind field reveal that the load fluctuations of WTs are reduced, and thus reduce fatigue load.
Keywords:
Fatigue load control
Model predictive control
Deep learning
Dynamic wake model
Wind farm
Journal
IF:
9.4
Papers:
4.2W
Citations:
20.2W