针对当前水陆两栖飞机在波浪中运动时历有效预报时长较短的问题，提出对水陆两栖飞机在一段时间内的运动统计值进行预报，并基于长短期记忆神经（long short-term memory, LSTM）神经网络构建了水陆两栖飞机运动统计特性预报模型。以NACA TN 2929水陆两栖飞机为例，基于其数值仿真数据分别对三、四、五级海况下水陆两栖飞机的垂荡、横摇和纵摇三自由度运动统计值进行预报并对其预报效果进行详细分析。结果表明：基于LSTM神经网络的水陆两栖飞机运动统计特性预报模型具有较好的预报精度，在实际工程应用中可以通过该模型对水陆两栖飞机未来一段时间内的运动统计值实现精确预报，为海上作业提供辅助决策信息。

Aiming at the problem of short effective prediction time for the movement history of amphibious aircraft in waves, the statistical values of amphibious aircraft movement over a period of time were proposed to predict, and a prediction model for the statistical characteristics of amphibious aircraft movement was constructed based on long short-term memory neural networks(LSTM). Taking the NACA TN 2929 amphibious aircraft as an example, based on its numerical simulation data, the statistical values of the three degrees of freedom motion of heave, roll, and pitch of amphibious aircraft under sea conditions of level 3, 4, and 5 were predicted, and their prediction effects were analyzed in detail. The results show that the LSTM neural network-based model for predicting the statistical characteristics of amphibious aircraft motion has good prediction accuracy. In practical engineering applications, this model can accurately predict the statistical values of amphibious aircraft motion in the future, providing auxiliary decision-making information for offshore operations.