基于CEEMD-PSO-MNN的船舶航迹预测  

Ship Track Prediction Based on CEEMD-PSO-MNN

作　　者：王文标 董贵平 汪思源 孟松 杜大鹏 杜佳璐 WANG Wenbiao;DONG Guiping;WANG Siyuan;MENG Song;DU Dapeng;DU Jialu(School of Marine Electrical and Engineering,Dalian Maritime University,Dalian Liaoning 116033,China)

机构地区：[1]大连海事大学船舶电气工程学院,辽宁大连116026

出　　处：《电子器件》2021年第1期119-124,共6页Chinese Journal of Electron Devices

基　　金：国家自然科学基金项目(51079013)。

摘　　要：船舶航迹预测对保障海上交通安全具有重要意义,为进一步提高船舶航迹预测精度,提出了一种从认知神经科学和神经生理学继承的模块化设计方法用于开发神经网络,旨在通过大脑强大的功能(分而治之)来解决复杂问题。首先,利用互补集合经验模式分解算法(CEEMD)将船舶航迹时间序列分解为多个相对平稳的子序列,使其具有不同的本征模态函数及趋势项,这在一定程度上降低了航迹时间序列的复杂程度;然后,通过模糊熵(FE)量化各子序列的复杂性用于辅助模块化神经网络(MNN)任务分配;最后,将粒子群(PSO)改进的长短期记忆神经网络(LSTM)作为模块化神经网络的子网络用于解决船舶航迹时间序列预测任务。选取相关数据进行测试,验证了所提方法对船舶航迹预测的准确性和实用性。Ship track prediction plays an important role in ensuring maritime traffic safety. In order to further improve the accuracy of ship track prediction,a modular design method inherited from cognitive neuroscience and neurophysiology is proposed for the development of neural networks,aiming at solving complex problems through the powerful functions of the brain(divide and conquer).Firstly,the complementary set empirical mode decomposition algorithm(CEEMD)is used to decompose the ship track time series into several relatively stable sub-sequences,which have different eigenmode functions and trend terms,which reduces the complexity of the track time series to some extent. Then,the complexity of each subsequence is quantified by fuzzy entropy(FE)to assist the task assignment of modular neural network(MNN).Finally,the improved long and short term memory neural network(LSTM)based on particle swarm optimization(PSO)is used as a sub-network of the modular neural network to solve the task of ship track time series prediction. The accuracy and practicability of the proposed method for ship track prediction are verified by selecting relevant data for testing.

关 键 词：航迹预测 互补集合经验模态分解 模块化神经网络 粒子群 长短期记忆神经网络 

分 类 号：U675.9[交通运输工程—船舶及航道工程] TP183[交通运输工程—船舶与海洋工程]