Data-driven modeling and extreme-event analysis of a floating power system with mooring dynamics in random waves

Abstract

This study examines the performance of a data-driven model in predicting the motions and loads of a floating power system (FPS) in various irregular waves, considering nonlinear wave–structure–mooring interactions. A surrogate model, constructed using a long short-term memory network, is trained and tested on robust datasets obtained from physical tank tests. The model's accuracy is initially validated through the temporal responses of the moored FPS and the associated extreme events. Its effectiveness is further assessed under varying random sea states, including different peak periods and wave steepness. A detailed harmonic analysis of extreme events indicates that while the model accurately captures the linear components of FPS's motions, it tends to underpredict higher-order harmonics, especially for pitch motions and mooring line tensions. Our results highlight the potential of data-driven models in marine applications, offering substantial computational savings for complex physical problems and the possibility of creating digital twins of real offshore structures. However, their limitations in capturing extreme events and higher-order nonlinearities must be carefully addressed when applying this methodology in strongly nonlinear sea states.