| dc.description.abstract | The expansion of offshore wind energy into floating platforms introduces unique challenges, particularly in the stability and maintenance of mooring systems, which is crucial for maintaining structural integrity. This study investigates the mooring system of a dual-spar floating wind farm and explores the application of diverse Machine Learning (ML) algorithms to predict mooring tensions accurately. Coupled time-domain numerical simulations of the floating wind farm were carried out in SIMA and further utilized to analyze four Recurrent Neural Network (RNN) algorithms, with particular emphasis on the shared mooring line and most loaded single mooring line due to their criticality. Results demonstrate that the Bidirectional Long Short-Term Memory (BiLSTM) algorithm achieved an impressive accuracy of approximately 98 % in various scenarios, making it a robust choice for predicting mooring line tensions, even under extreme weather conditions. Furthermore, a sensitivity study evaluates the impact of mooring line properties on both tension and snap loads. Nacelle acceleration inputs are incorporated to enhance predictive accuracy, revealing a minimal to no impact on the mooring line peak tension. Thus, while nacelle acceleration contributes to model accuracy enhancement, it is not crucial for achieving high prediction accuracy. By demonstrating the effectiveness of the BiLSTM algorithm and providing insights into the influence of mooring line tension, this research advances our understanding of the complex dynamics involved in offshore wind energy deployment, paving the way for more efficient and reliable floating wind farm designs. | |