A prediction method and device for offshore platform motions based on CEEMDAN-Transformer

Description: Patent No.: 2024118570268

Advantages and significance of the invention: The proposed CEEMDAN-Transformer architecture addresses nonlinear motion prediction of offshore platforms by combining signal decomposition and attention mechanisms. It achieves an RMSE of 0.1647, with 34% higher accuracy than LSTM, enabling real-time safety guidance. Adaptive noise injection and extreme-value-based stopping improve decomposition stability, optimize scheduling, and cut maintenance costs by over 20%. It enhances offshore safety and reduces engineering risks. The method also applies to other marine structures and, when paired with digital twins and edge computing, supports intelligent deep-sea operations. (2) Applications The CEEMDAN-Transformer architecture is applied in real-time motion prediction and safety monitoring systems for offshore platforms. This supports timely decision-making for operational adjustments, emergency responses, and maintenance scheduling, reducing engineering risks and maintenance costs. The method can also be extended to other marine structures, such as Floating Production Storage and Offloading (FPSO) vessels and offshore wind turbines, to enhance their operational safety and efficiency. For FPSO vessels, which are frequently used in deep-sea oil and gas production, the CEEMDAN-Transformer architecture can predict complex nonlinear motions caused by waves, wind, and currents, thereby improving the accuracy of cargo transfer, drilling operations, and mooring system control. In the case of offshore wind turbines, the model can forecast tower and blade vibrations due to environmental loads, supporting predictive maintenance and reducing downtime and structural fatigue. Furthermore, when integrated with digital twin systems and edge computing, the architecture enables real-time simulation and monitoring of these structures. (3) Research achievements The achievements include one published academic monograph, over 55 research papers, 10 authorized invention patents, and 4 registered software copyrights.

Organisation: Tianjin University

Innovator(s): Li Ying, Zhong Qiyuan, Zhu Yuanhao

Category: Energy

Country: China