The prediction of ship motion attitude in seaway based on BSO-VMD-GRU combination model

Abstract

Short-term prediction of ship attitude holds the potential to provide decision support for precise control and optimal timing selection of ship operations at sea, thereby ensuring the safety and efficiency of offshore activities. However, the time series data representing ship motion attitude exhibits complex nonlinear and non-stationary characteristics, which pose challenges to achieving accurate predictions. To improve prediction accuracy, this study employs two sequence decomposition techniques: Variational Mode Decomposition (VMD) and Adaptive Noise Complete Ensemble Empirical Mode Decomposition (CEEMDAN). These decomposition methods are integrated with a Gated Recurrent Unit (GRU) to forecast ship motion attitude. Firstly, the decomposition techniques are applied to decompose the time series data of ship motion attitude into multiple intrinsic mode function (IMF) components. Subsequently, each IMF component is predicted using the GRU. Ultimately, the final prediction results are obtained by aggregating the GRU’s predictions for each individual component. To alleviate the impact of manual parameter configuration in the VMD algorithm, this study introduces a metaheuristic optimization algorithm called the Binary System Optimization Algorithm (BSO), which is empirically validated for its effectiveness. Moreover, real ship data from the vessel named “Yukun” is utilized to experimentally verify the proposed prediction algorithm. The experimental outcomes reveal that both decomposition algorithms can enhance the prediction accuracy of the GRU, with the VMD algorithm yielding superior prediction results compared to those obtained using CEEMDAN.