为解决水面单体船迎浪航行时因强耦合、非线性及长时依赖等原因导致的垂荡与纵摇大幅运动精准预报难题，提出一种水面单体船迎浪航行大幅运动预报优化算法。建立单体舰船迎浪航行运动学模型，刻画垂荡与纵摇的流体动力耦合特性及随机海浪干扰下的动力学规律，以模型数值仿真数据为输入，构建基于自注意机制与堆叠门控循环单元的（Self-Attention Stacked Gated Recurrent Unit，SAGRU）模型，通过自注意机制强化波浪砰击、甲板上浪等关键耦合特征提取，利用堆叠门控循环单元（Gated Recurrent Unit，GRU）捕捉局部瞬态与全局长时运动依赖，并引入残差连接保障梯度传播，实现高精度、轻量化预报。实验结果表明：该算法在超前3 s与10 s预报中，垂荡与纵摇预报曲线与实际值高度吻合，误差分布集中，且垂荡与纵摇运动的连续分级概率评分（Continuous Ranked Probability Score，CRPS）分别低至0.034和0.018，具有良好的预报鲁棒性与精度。
In order to solve the problem of accurate prediction of large sway and pitch motion of single surface ships sailing against waves due to strong coupling, nonlinearity, and long-term dependence, an optimization algorithm for large motion prediction of single surface ships sailing against waves is proposed. Establish a kinematic model for a single ship sailing against waves, depicting the fluid dynamic coupling characteristics of heave and pitch, as well as the dynamic laws under random wave interference. Using numerical simulation data as input, construct a Self-Attention Stacked Gated Recurrent Unit (SAGRU) model based on self attention mechanism and stacked gated cyclic units. Through self attention mechanism, enhance the extraction of key coupling features such as wave slamming and deck waves. Use stacked Gated Recurrent Unit (GRU) to capture local transients and overall long-term motion dependencies, and introduce residual connections to ensure gradient propagation, achieving high-precision and lightweight prediction. The experimental results show that the algorithm has a high degree of agreement between the heave and pitch prediction curves and the actual values in the 3-second and 10s ahead prediction, with concentrated error distribution. The continuous grading probability scores Continuous Ranked Probability Score (CRPS) of heave and pitch motion are as low as 0.034 and 0.018, respectively, demonstrating good prediction robustness and accuracy.
2025,47(23): 174-178 收稿日期：2025-6-7
DOI：10.3404/j.issn.1672-7649.2025.23.027
分类号：U661.32
基金项目：陕西省自然科学基础研究计划项目（2020JM-294）
作者简介：黄小平(1977-),女,硕士,讲师,研究方向为优化算法及数学
参考文献：
[1] 王世取, 陈静, 于欣, 等. 不同海况下破损舰船进水过程及航行轨迹研究[J]. 中国造船, 2024, 65(5): 53-66
WANG S Q, CHEN J, YU X, et al. study on flooding process and sailing trajectory of damaged warship under different sea conditions[J]. Shipbuilding of China, 2024, 65(5): 53-66
[2] 朱杰, 刘在良, 林艳, 等. 随机海浪下船舶横摇运动响应极值预报研究[J]. 中国舰船研究, 2025, 20(2): 196-202
ZHU J, LIU Z L, LIN Y, et al. Extreme value prediction of the roll motion under random seas[J]. Chinese Journal of Ship Research, 2025, 20(2): 196-202
[3] 王鑫琦, 朱齐丹. 基于输出误差模型优化的甲板运动预报算法研究[J]. 智能系统学报, 2023, 18( 1): 75-85
WANG X Q, ZHU Q D. Research on deck motion prediction algorithm based on output error model optimization[J]. CAAI Transactions on Intelligent Systems, 2023, 18(1): 75-85
[4] 张腾, 张振华, 薛蕃衍, 等. 规则波浪中船舶垂荡和纵摇运动数值预报与仿真[J]. 大连海事大学学报, 2025, 51(2): 22-31
ZHANG T, ZHANG Z H, XUE F Y, et al. Numerical prediction and simulation of ship heave and pitch motions in regular waves[J]. Journal of Dalian Maritime University, 2025, 51(2): 22-31
[5] 刘涵, 苏焱, 张国强. 基于支持向量回归的破损船舶横摇运动快速预报[J]. 上海交通大学学报, 2025, 59(7): 1041-1049.
LIU H, SU Y, ZHANG G Q. Fast Prediction for roll motion of a damaged ship based on SVR[J]. Journal of Shanghai Jiaotong University, 2025, 59(7): 1041-1049.
[6] 储纪龙, 顾民, 鲁江. 双桨双舵内倾船型的横甩失稳运动预报研究[J]. 船舶力学, 2024, 28(9): 1307-1316
CHU J L, GU M, LU J. Prediction method of the broaching of a tumblehome ship with twin propeller and double rudders[J]. Journal of Ship Mechanics, 2024, 28(9): 1307-1316
[7] 薛建胜, 高志亮. 基于TCN-BLSTM-TPA模型的不规则波中船舶运动预报[J]. 船舶工程, 2024, 46(7): 42-49
XUE J S, GAO Z L. Ship motion prediction in irregular waves based on TCN-BLSTM-TPA Model[J]. Ship Engineering, 2024, 46(7): 42-49
[8] 朱沛樵, 丁军, 耿彦超, 等. 基于GRU模型的船舶运动与载荷快速预报研究[J]. 船舶力学, 2025, 29(3): 337-350
ZHU P Q, DING J, GENG Y C, et al. Rapid prediction of ship motion and load based on GRU neural[J]. Journal of Ship Mechanics, 2025, 29(3): 337-350
