Blade deformation prediction of floating wind turbine based on an AI-based SADA method

Abstract

Due to the aerodynamic load, the gravity load and the inertial load, the blades of floating offshore wind turbines (FOWTs) are prone to flexible deformation. However, the prediction of blades deformation is a crucial challenging task as it is very difficult to obtain in basin experiment and actual onsite measurement. This paper addresses a numerical study based on AI technology and an in-house programme DARwind to predict blade tip deformation for FOWTs. The AI module in DARwind employs deep deterministic policy gradient (DDPG) algorithm of Reinforcement Learning, to make the DARwind code intelligent for the objective FOWTs through basin experimental data training. The concept of KDPs is firstly introduced, which includes the crucial parameters in various discipline. The motions data of FOWT supporting platform from the basin experiment are selected as the target parameters to train the AI-based DARwind. The training results show that the mean value of platform motions from the AI-based DARwind match better with the basin experimental data. Through more accurate corrections to the platform motions, the blade deformation can be predicted more accurately in terms of axial, flap-wise and edge-wise.