Development and characterisation of an AI-in-the-loop testing platform for floating wind turbines PART I: Construction, validation, and benchmark testing

Model testing is an inevitable means to verify design optimization because it is more economical than prototype testing and more reliable than numerical simulation. However, in the floating wind turbine experiment, the hydrodynamic Froude number and the aerodynamic Reynolds number cannot satisfy similar rules simultaneously, making the scale effect problem a major difficulty in the experiment. Therefore, this paper innovatively introduces AI-prediction-in-the-loop experimental technologies. The Froude similarity criterion is applied to model production and physical set-up. A Froude-similar wind turbine model (except for the blades) is placed in the wave flume and the floating platform moves. The response measurement data is input into the AI prediction module to perform real-time prediction of aerodynamic loads such as rotor thrust, output the calculation results and control the simulated load of the actuator, thereby realizing aerodynamic-hydrodynamic-structural coupling experiments under Froude's rules. Characterization benchmark and tank tests are carried out to validate the AI-in-the-loop testing methodology, and the results show good agreement between measured and predicted rotor thrust values across both high and low frequencies. Moreover, the time delay and systematic uncertainty of the proposed testing platform are identified for the first time.