Session: 15-05-01 Mooring, Riser and Pipelines

Submission Number: 181966

Multi-Head Attention-Enhanced Gate Recurrent Units for Mooring Tension Prediction of TLP FOWTS 

Tension-Leg Platform (TLP) floating offshore wind turbines (FOWTs) are a promising solution for deep-water wind energy development due to their economic efficiency and excellent station-keeping performance. Moored by taut tension legs, TLPs exhibit high global stiffness and elevated natural frequencies in heave and pitch, leading to more complex dynamic responses under combined wind, wave, and current loading compared to semi-submersibles or spars. Given that tendon integrity governs overall structural safety, accurate and real-time prediction of extreme mooring tensions is essential for engineering design, operational monitoring, and risk assessment. Although recurrent networks such as RNNs, LSTMs, and GRUs are widely used for time-series forecasting, they struggle to capture long-range temporal dependencies in lengthy, dynamically rich sequences—limiting their ability to model critical correlations across distant time steps in mooring tension responses. To address this, we propose a novel model that augments GRUs with a multi-head attention (MHA) mechanism to explicitly resolve long-range dependencies while preserving temporal fidelity. Using the MIT 5 MW TLP FOWT as a case study, high-fidelity mooring tension data are generated via fully coupled aero-hydro simulations under various environmental conditions and formatted into training samples via a sliding-window approach. The MHA-GRU model is trained and evaluated on a dedicated test set. Results show that MHA significantly improves the model’s ability to capture long-range dynamics. Compared to a baseline GRU, the proposed architecture achieves higher predictive accuracy with low inference latency, demonstrating its effectiveness for real-time mooring load forecasting in TLP-based offshore wind systems.

Presenting Author: Maokun Ye Shanghai Jiao Tong University

Presenting Author Biography: Dr. Ye is a Post-Doc researcher in CMHL at SJTU.

Authors:

Maokun Ye Shanghai Jiao Tong University
Yisheng Yao Shanghai Jiao Tong University
Shiyuan Zhang Offshore Oil Engineering Co. Ltd
Shu Dai Shanghai Investigation, Design and Research Institute Co. Ltd.
Decheng Wan Shanghai Jiao Tong University