Neural network-based super-twisting control for floating wind turbines: Design and real-time validation

• A Radial Basis Function Neural Network-based Super-Twisting Algorithm (RBFNN-STA) is developed for collective blade pitch control in Floating Offshore Wind Turbines (FOWTs). • The control law is derived from a reformulated nonlinear model, capturing rotor speed and platform pitch rate dynamics. • A Hardware-in-the-Loop (HIL) framework is implemented to validate the controller, ensuring real-time feasibility and robustness. • Comparative analysis with the Super-Twisting Algorithm (STA) alone and the GSPI baseline controller confirms superior rotor speed regulation and improved platform stability under varying wind and wave conditions. This paper presents a novel model-based control strategy for collective blade pitch regulation in Region III of a 5 MW semi-submersible FOWT. The proposed approach combines a Radial Basis Function Neural Network (RBFNN) with a Super-Twisting Algorithm (STA) to enhance robustness against disturbances and improve dynamic performance. The control design is based on a reformulated nonlinear control-oriented model to support the synthesis of adaptive control laws. In addition to the control strategy, this work introduces a real-time Hardware-in-the-Loop (HIL) platform as a key contribution, specifically designed for the validation of FOWT control strategies. This platform provides a strictly synchronized execution framework for accurate real-time emulation and integrates OpenFAST for high-fidelity dynamic simulations. By enabling real-time validation, the HIL platform plays a crucial role in assessing control performance under realistic operating conditions. Experimental results confirm both the real-time feasibility of the proposed RBFNN-based STA controller and the effectiveness of the HIL platform in evaluating its performance, particularly in regulating generator speed and mitigating platform pitch motion under varying wind conditions.