基于NGO-Bi-GRU的船舶轨迹预测模型  

Ship trajectory prediction model based on NGO-Bi-GRU

作　　者：谢海波[1] 乔冠洲 代程 丁润祯 白伟伟[1] XIE Haibo;QIAO Guanzhou;DAI Cheng;DING Runzhen;BAI Weiwei(Navigation College,Dalian Maritime University,Dalian 116026,China)

机构地区：[1]大连海事大学航海学院,辽宁大连116026

出　　处：《舰船科学技术》2025年第4期14-20,共7页Ship Science and Technology

基　　金：中国国家自然科学基金资助项目(522713602);四川省自然科学基金资助项目(2022NSFSC0891)。

摘　　要：针对传统的神经网络模型因超参数众多,在实验中比对最优参数组合效率低下导致误差较大和反应速度慢的问题。本文提出一种基于北方苍鹰优化(Northern Goshawk Optimization,NGO)算法和双向门控循环单元神经网络(Bidirectional Gated Recurrent Unit, Bi-GRU)的船舶轨迹预测模型NGO-Bi-GRU(Northern Goshawk Optimization Bidirectional Gated Recurrent Unit)。利用NGO对Bi-GRU模型的学习率、隐藏节点和正则化系数进行寻优,然后将寻优得到的网络超参数代入Bi-GRU进行船舶轨迹预测。将该模型与长短时记忆神经网络(Long Short Term Memory, LSTM)和门控循环单元神经网络模型(Gated Recurrent Unit, GRU)以及使用该算法优化的长短期神经网络模型进行实验对比,将均方误差、均方根误差、平均绝对误差作为评价标准。结果表明,NGO-Bi-GRU模型在经度和纬度预测上误差较小、精确度较高且数值波动更加稳定。In response to the challenges posed by the numerous hyperparameters in traditional neural network models,which result in significant errors and slow response times due to the inefficiency of comparing optimal parameter combina-tions in experiments,this paper introduces a novel ship trajectory prediction model based on Northern Goshawk Optimiza-tion(NGO)and Bidirectional Gated Recurrent Unit neural networks(Bi-GRU),termed NGO-Bi-GRU.The NGO algorithm is employed to optimize the learning rate,hidden nodes,and regularization coefficients of the Bi-GRU model,after which the optimized network hyperparameters are applied to Bi-GRU for ship trajectory prediction.This model is experimentally com-pared with Long Short Term Memory(LSTM)networks,Gated Recurrent Unit(GRU)networks,and LSTM networks op-timized using the same algorithm.Evaluation criteria include mean squared error,root mean squared error,and mean abso-lute error.The results demonstrate that the NGO-Bi-GRU model achieves lower errors and higher precision in predicting lon-gitude and latitude,with more stable numerical fluctuations.

关 键 词：北方苍鹰算法 船舶轨迹预测 船舶自动识别系统 神经网络 

分 类 号：U661[交通运输工程—船舶及航道工程]