海洋预报2025，Vol.42Issue(5)：86-94,9.DOI:10.11737/j.issn.1003-0239.2025.05.009
融合PSO和注意力的海上落水人员漂移轨迹预测模型研究
Research on drift trajectory prediction model of marine drowning personnel integrating PSO and Attention
摘要
Abstract
Accurate prediction of drift trajectory of marine drowning personnel plays an important role in emergency rescue for those in distress.To address the issue of excessive dependence on model parameters in traditional statistical methods,this paper proposes a data-driven deep learning prediction model.Taking into account the impact of factors such as offshore wind and ocean currents on floating targets,the differential moving average method is adopted to reduce noise in labeled data.The attention mechanism is introduced into the long short-term memory network(LSTM)to effectively capture periodical and trending motion patterns in time series.Then,the particle swarm optimization(PSO)algorithm is used to optimize the Attention LSTM model,resulting in an integrated model for predicting the trajectory of floating targets on the sea(PSO Attention LSTM).By modeling the drift trajectories of 35 pseudo human bodies over 4 850 time steps and calculating the prediction error,the results show that:the root mean square error of the predictions is 0.024 5°in the longitude direction,0.017 3°in the latitude direction,with an average displacement error of 4.844 km and a final displacement error of 7.031 km.Through ablation experiments,it has been proven that the model proposed in this paper outperforms the comparative models on three evaluation metrics:root mean square error,average displacement error,and final displacement error.关键词
海上搜救/漂移轨迹/差分移动平均/PSO-Attention-LSTM模型Key words
maritime search and rescue/drift trajectory/differential moving average/PSO-Attention-LSTM model分类
海洋学引用本文复制引用
白鹤,李钰,张默,张心如..融合PSO和注意力的海上落水人员漂移轨迹预测模型研究[J].海洋预报,2025,42(5):86-94,9.基金项目
2023年度辽宁省教育厅基本科研项目(JYTMS20230965). （JYTMS20230965）
