DRNN神经网络用于船舶横摇运动的时间序列预报  

Time Series Prediction of Ship Roll Using Diagonal Recurrent Neural Networks

作　　者：王科俊[1] 李国斌 

机构地区：[1]哈尔滨工程大学自动控制系

出　　处：《哈尔滨工程大学学报》1997年第1期39-45,共7页Journal of Harbin Engineering University

摘　　要：首次将具有内部时延的内回归神经网络用于船舶横摇运动的时间序列预报，取得了良好的结果文中给出对角回归神经网络模型和学习算法，并对学习算法的收敛性进行了分析利用对角回归神经网络对船舶横摇运动的时间序列预报表明，通过离线学习内回归神经网络能掌握系统的动态特性。Diagonal recurrent neural networks (DRNN) are used to predict a ship roll motion with time series date for the first time.The model and learning algorithm for DRNN are given in the paper.The convergence of learning algorithm is analysed by Liapunov stability theorem .The experiments of ship roll time series prediction using DRNN are done with the data from ship trial.Some results are given from experiments.The inner recurrent neural networks can grasp the characteristics of ship roll by learning ,and give accurate results of prediction.

关 键 词：神经网络 船舶 横摇运动 时间序列预报 耐波性 

分 类 号：U661.321[交通运输工程—船舶及航道工程] TP18[交通运输工程—船舶与海洋工程]