舰船的六自由度运动状态形成复杂的非线性过程，运动姿态会受到耦合作用、不定周期、噪声信号以及混沌特性等因素的干扰，因此很难得到精确的预测结果。为了提升舰船运动姿态的预测精度，利用舰船时间序列的特点，建立了基于长短期记忆单元（LSTM）模型，对其进行了舰船姿态预测仿真，将结果与时间序列分析法的结果进行对比。实例分析表明：基于LSTM模型的预测方法具有精确度高、易实现的特点。这为舰船运动短期预测提供了一个新的思路和方法。
The ship motion with six DOF is a complex nonlinear process. It is affected by coupling effect, variable periodicity, noise signal, and the chaos characteristic of ship motion time series so that obtaining precise forecasting results of ship motion is difficult. In order to improve the forecast precision, character of time series in ship motion is analyzed, LSTM model is established and it is used in the emulation research, and the experimental result is compared with the result based on AR model. The result analysis shows that prediction based on LSTM has a high accuracy and is implemented easily, so a new approach is presented for ship motion prediction.
2017,39(7): 69-72 收稿日期：2016-09-05
DOI：10.3404/j.issn.1672-7649.2017.07.014
分类号：TP29
作者简介：王国栋(1991-),男,硕士研究生,研究方向为智能检测、机器学习
参考文献：
[1] YIN Jian-chuan, ZOU Zao-jian,XU Feng. On-line prediction of ship roll motion during maneuvering using sequential learning RBF neural networks[J]. Ocean Engineering, 2013, 61: 139–147.
[2] E. Bal Beşikçi, O. ARSLAN, O. TURAN, A. I. Ölçer, An artificial neural network based decision support system for energy efficient ship operations[J]. Computers & Operations Research, 2016, 66: 393–401.
[3] 彭秀艳, 赵希人, 高奇峰. 船舶姿态运动实时预报算法研究[J]. 系统仿真学报, 2007, 19(2): 267–271.PENG Xiu-yan, ZHAO Xi-ren, GAO Qi-feng. Research on real-time prediction algorithm of ship attitude motio[J]. Journal of System Simulation, 2007, 19(2): 267–271.
[4] LI Ming-We, GENG Jing, HAN Duan-Feng, et al. Ship motion prediction using dynamic seasonal RvSVR with phase space reconstruction and the chaos adaptive efficient FOA[J]. Neurocomputing, 2016, 174: 661–680.
[5] PETER J. BROKWELL, 田铮. 时间序列的理论与方法(第二版)[M]. 北京: 高等教育出版社, 2001.
[6] 马洁, 韩蕴韬, 李国斌. 基于自回归模型的船舶姿态运动预报[J]. 舰船科学技术, 2006, 28(3): 28–30.MA Jie, HAN Yun-tao, LI Guo-bin. Prediction of ship pitching motion based on AR method[J]. Ship Science and Technology. 2006, 28(3): 28–30.
[7] ZACHARY C L, JOHN B, CHARLES E. A critical review of recurrent neural networks for sequence learning[M]. Computer Science, 2015.
[8] ALEX G. Supervised sequence labelling with recurrent neural networks[J], Studies in Computational Intelligence, 2012, 385.
[9] TOMASZ P. Using evolutionary neural networks to predict spatial orientation of a ship[J]. Neurocomputing, 2015, 166: 229–243.
