基于Transformer的船舶运动预测模型研究

Research on Ship Motion Prediction Model Based on Transformer

为准确而有效地预测船舶运动,提出一种基于Transformer的船舶运动预测模型。鉴于传统的Transformer模型为自然语言领域所运用,论文首先改进Transformer模型使其能够适用于船舶运动预测领域。其次,通过改进模型分析出船舶过去一段时间的运动规律,据此规律预测船舶未来一段时间的运动。研究表明,该模型不仅在训练工况下有着极高的预测精度,而且在与训练工况相近的工况下也有很高的预测精度。本研究结果可以作为补偿船舶运动的基础。

To predict the ship motion accurately and effectively, a transformer based ship motion prediction model is proposed. As the traditional transformer model is the model used in the natural language field, in the paper the Transformer model is first improved to make it applicable to the field of ship motion prediction. Secondly,the motion rule of the ship in the past period is analyzed by the improved model, and the motion of the ship in the future period is predicted according to this rule. It is shown that the model has very high prediction accuracy not only in the training conditions, but also in the conditions similar to the training conditions. The results of this study can be used as a basis for compensating the ship motion.