Multi-step Prediction AR Model of Ship Motion Based On Constructing and Correcting Error

作者：Xiaoyu Wang, Tong Ming, Liang Du · 发表于：2018 IEEE CSAA Guidance, Navigation and Control Conference (CGNCC) · 年份：2018 · DOI：10.1109/gncc42960.2018.9018912 · 被引用次数：3 · 研究领域：Structural Health Monitoring Techniques、Fault Detection and Control Systems、Ship Hydrodynamics and Maneuverability

A new multi-step prediction method of constructing correcting error coupled with auto-regression (AR) model is proposed in this manuscript. And it is successfully applied to 5 seconds prediction of ship motion. Prediction error in future 2.5 seconds constructed is utilized to predict future 5 seconds error so that multi-step prediction outcomes of ship motion based on AR could be effectively corrected. The outcomes agree well with the measured datum of ship motion. Compared with the original AR model, the prediction accuracy is improved by 9.15% with the method in the paper. The ratio of the prediction error to sample variance is as low as 29.31% that is obviously lower than 50% in accordance with the reference. Therefore, it is verified that the method in the manuscript is both feasible and of significance to the engineering.