Adaptive Inverse Deep Reinforcement Lyapunov learning control for a floating wind turbine

Document Type : Article


1 University of Tabriz

2 University of Tabriz, Department of Mechanical Engineering


Offshore floating wind turbines (FWT) decrease adverse climate change effects without occupying significant land and harvesting fields. Owing to the earth planet unexpected climate, online adaptive feedback control of FWTs will be effective in the sense of optimal and uniform energy capture. In this paper, a deep reinforcement learning (DRL)-based control system is proposed to offset both the disturbance and noise effects. Large variations of wind and water waves generate enormous information give rise to convergent learning of deep neural networks model of the wind turbine. As a result of the disturbance and wind abrupt changes, an adaptive inverse control equipped with DRL could easily cope with the inherent drawback of DRL i.e., tracking error. Furthermore, received rewards in the DRL algorithm are passed through the newly designed training algorithm to predict control actions such that the loss function is decreased. The attenuation of disturbance and noise on the tracking performance of closed-loop FWT is clarified through software implementation tests while the weight’s convergency and update rules are proved by the direct Lyapunov theorem.


Main Subjects