Study on Neural Network-Based Prediction Methods for Ship Mooring Forces

LNG vessels are significantly affected by external environmental conditions during mooring operations, and there are risks such as mooring line failure in severe sea states. In order to solve this problem, this study employs back propagation neural network (BP), long short-term memory network (LSTM) and gated recurrent unit (GRU) to predict the mooring force of LNG vessels. Comparative analysis of the models shows that GRU outperforms the BP and LSTM models, achieving lower mean absolute error (MAE), root mean square error (RMSE), and mean absolute percentage error (MAPE), while obtaining higher coefficient of determination ($\mathrm{R}^{2}$). These results indicate that the GRU model provides higher prediction accuracy and reliability in predicting mooring forces. In addition, the GRU model exhibits excellent performance in terms of prediction accuracy, convergence speed and generalization ability. The experimental results highlight that the GRU model is able to effectively capture the nonlinear dynamic characteristics of mooring force time series data and significantly reduce the prediction error compared to other models.