基于LSTM神经网络的随机横浪中船舶横摇运动极短期预报

Very Short-term Prediction of Ship Rolling Motion in Random Transverse Waves Based on LSTM Neural Network

基于波浪与船舶运动响应的因果关系,采用LSTM神经网络模型建立了随机波高历时与船舶横摇运动历时的映射关系,基于波高历时数据实现对船舶横摇运动的极短期预报.利用DTMB5415舰船在三、四、五级横浪作用下横摇运动势流理论计算数据,进行LSTM模型的训练和预测,验证了方法的可行性.不同提前时长条件下船舶横摇预报结果分析表明,与基于船舶横摇输入数据的模型相比,基于波高数据输入的模型在四级和五级海况下预报精度均有改善,特别是在五级海况下预报精度有大幅提高,均方根误差和最大绝对误差降幅均在40%以上.

Based on the causal relationship between waves and ship motion response,LSTM neural network model was used to establish the mapping relationship between random wave height duration and ship rolling motion duration,and the very short-term prediction of ship rolling motion was realized based on wave height duration data.The LSTM model was trained and predicted by using the theoretical calculation data of the potential flow of DTMB5415 ship rolling motion under the action of the third,fourth and fifth level transverse waves,and the feasibility of the method was verified.The analysis of ship roll prediction results under different lead times shows that,compared with the model based on ship roll input data,the prediction accuracy of the model based on wave height data input is improved under the four-level and five-level sea conditions.Especially,the prediction accuracy has been greatly improved under the five-level sea conditions,with the root mean square error and the maximum absolute error both falling by more than 40%.