Active cable-based vibration control for wind turbine structures incorporating time-delay effects

Abstract

Various control strategies have been developed to mitigate wind turbine (WT) vibrations. Among these, active vibration control offers superior performance, although its effectiveness can be compromised by time delays stemming from the relative complexity of the control system. This study investigates the feasibility of a novel long short-term memory (LSTM)–driven active cable-based damper (LSTM–ACD) for WT vibration control, in which the LSTM neural network functions as the active control algorithm addressing the time-delay issues. The LSTM network trained using tower-top response data with given time delays as inputs, while the corresponding ideal control forces generated by the Linear Quadratic Regulator (LQR) serve as outputs. Detailed theoretical mechanisms of cable-based control and LSTM–ACD are introduced first. Subsequently, a coupled excitation–WT–LSTM–ACD model is established using an NREL 5 MW WT as a case study. The control performance of an active cable-based damper with LQR control is then evaluated under seismic and wind load scenarios, and the LSTM–ACD’s time-delay compensation capability is validated under seismic excitations. Results demonstrate that: (1) ideal active control outperforms the passive counterpart in seismic and wind load scenarios; (2) the LSTM–ACD achieves comparable vibration mitigation performance compared to the ideal LQR-based controller; and (3) the LSTM–ACD exhibits robustness against variations of time delays. These findings demonstrate that the LSTM–ACD is an effective and robust active control method for WTs, showing a strong potential to address the time-delay challenges in active vibration control systems.