基于LSTM-EMD模型的船舶航迹预测方法  

A Prediction Method of Ship Trajectory Based on LSTM-EMD Model

作　　者：王均刚 胡柏青[1] 高端阳 朱九鹏 WANG Jungang;HU Baiqing;GAO Duanyang;ZHU Jiupeng(College of Electrical Engineering,Naval University of Engineering,Wuhan 430033;Department of Navigation,Dalian Naval Academy,Dalian 116018)

机构地区：[1]海军工程大学电气工程学院,武汉430033 [2]大连舰艇学院航海系,大连116018

出　　处：《舰船电子工程》2023年第7期29-35,共7页Ship Electronic Engineering

摘　　要：船舶运动轨迹预测是船舶避碰、海上交通管理等领域的一个关键问题,对于提高航海安全性有着不可忽视的作用。现有成果包含了许多不同的轨迹预测模型,但很少有关于预测误差和补偿技术的研究。针对船舶轨迹预测与补偿技术,提出了一种结合长短时记忆(Long Short-Term Memory,LSTM)神经网络模型和经验模态分解(Empirical Mode Decomposition,EMD)算法的预测框架,包含数据预处理、LSTM模型预测、EMD分解、残差预测及补偿,利用AIS(Automatic Identification System,AIS)数据进行实验验证,实验结果表明:与BP神经网络模型和LSTM神经网络模型相比,该组合模型的均方根误差和平均绝对百分比误差均为最小,是一种预测精度更高的船舶运动轨迹预测方法。Prediction of ship motion trajectory is a key issue in the fields of ship collision avoidance and maritime traffic management,and plays an important role in improving navigation safety.A lot of different trajectory prediction models have been proposed,but few studies on prediction errors and compensation techniques.Aiming at ship trajectory prediction and compensation technology,a prediction framework combining LSTM(long short-term memory)neural network model and EMD(empirical mode decomposition)algorithm is proposed.The framework includes data preprocessing,LSTM model prediction,EMD decomposition,residual prediction and compensation.In this paper,the AIS(automatic identification system)data is used for experimental verification.The experimental result shows that compared with the BP and the LSTM neural network model,the LSTM-EMD model has the smallest RMSE(root mean square error)and MAPE(mean absolute percentage error).It shows that this is a better prediction model.

关 键 词：船舶运动轨迹预测 AIS 长短时记忆神经网络 经验模态分解 组合预测模型 

分 类 号：U674[交通运输工程—船舶及航道工程]