舰船摇荡预报中的双变量局域线性多步预报法

Dual-variable Local Linear Multi-step Method for Ship Sway Prediction

针对舰船各个自由度摇荡之间存在耦合的实际,介绍了一种基于双变量时间序列预报局域线性预测法,此方法可对具有一定耦合关系的二维时间序列进行综合预报。将此方法应用于陈氏吸引子仿真数据及真实海浪中某舰船摇荡时间序列数据的预测。通过与单变量时间序列预报效果的对比,发现时间序列的双变量局域线性预测法可以在一定程度上提高精度,减小误差,提高了舰船摇荡预报的有效性。对于船舶摇荡时间序列预报领域应用双变量预报方法进行了尝试。

In light of the couplings among the components of the ship sway, a dual-variable local linear multi-step method is introduced. The method is able to make a comprehensive prediction of a coupled 2--D time series. A fore- cast test is carried out with the Chen＇s attractor and the actual data of a ship moving in waves. The results show that the dual variable weighted one rank local linear multi-step method can improve prediction precision and reduce errors in predicting ship sways compared with univariate prediction method.