APP下载

基于DBN-PID模型的舰船运动姿态短期预测及仿真

2021-05-09迟鑫鹏郑旺辉

现代信息科技 2021年20期

迟鑫鹏 郑旺辉

摘  要:舰船在海浪中航行受到海浪、海风等环境因素干扰,不可避免地产生摇摆,给舰船的海上航行造成很大的安全隐患。舰船航行过程中,运动姿态会受到耦合作用、不定周期等诸多因素的干扰,因此很难精确地预测舰船的短期运动姿态。文章提出一种新的预测方法,将机器学习DBN网络与PID网络相互结合,建立DBN-PID模型对舰船的姿态变化进行短期预测。模型有效地解决了舰船运动姿态预测时间短、误差大的问题,具有良好的工程应用价值。

关键词:舰船运动;非线性;短期预测

中图分类号:TP183    文献标识码:A 文章编号:2096-4706(2021)20-0071-07

Short-term Prediction and Simulation of Ship Motion Attitude Based on DBN-PID Model

CHI Xinpeng, ZHENG Wanghui

(Beijing Institute of Machinery Equipmenth, Beijing  100854, China)

Abstract: Ship navigation in the waves is disturbed by the waves, sea breeze and other environmental factors, inevitably swaying, causing great security risks to the ship navigation. In the process of ship navigation, the motion attitude will be disturbed by many factors, such as coupling effect, indefinite period and so on. Therefore, it is difficult to accurately predict the short-term motion attitude of a ship. In this paper, a new prediction method is proposed, which combines machine learning DBN network with PID network to establish DBN-PID model for short-term prediction of ship attitude changes. The model effectively solve the problems of short time and large error of ship motion attitude prediction, and has good engineering application value.

Keywords: ship motion; nonlinearityl; short term forecast

0  引  言

目前对舰船姿态的极短期的预测方法的研究中,主要是从频域和时域进行探究。归纳起来主要有:统计预报方法、卷积方法、卡尔曼滤波法、周期图法、艏前波法、谱估计方法、投影寻踪方法等。这些上述方法应用起来都有一定的局限性,如采用滤波方法,需要提供系统自身的状态方程,由于舰艇的运动形式十分复杂,想要获得其相对精确的状态方程十分困难。时间序列分析法则计算量较大。而且对于舰船的运动的姿态来说,单纯地采用一个线性的数学模型是无法准确地逼近所要预测的舰船的运动姿态,此外艏前波法是要基于舰艏波的信息参数,这些在实际计算中是难以实现的[1]。本文建立的DBN-PID舰船运动姿态短期预测模型可以降低计算量,延长预测时间达到更加良好的预测效果。

1  DBN-PID模型建模原理

1.1  DBN深度……

登录APP查看全文