Ship motion attitude prediction based on EMD-PSO-LSTM integrated model

Due to the high randomness and complexity of ships sailing at sea, a single model has limited capability and is difficult to make accurate attitude prediction. Therefore, an integrated prediction model based on EMD (empirical mode decomposition) and PSO-LSTM (particle swarm optimized long- and short-term memory neural network) is proposed to predict the ship's motion attitude. Firstly, the data of ship motion attitude measured by inertial navigation system is decoupled by EMD to obtain finite intrinsic mode function (IMF). Then, the PSO-LSTM model is used to study the short-term sequential rules of each IMF component and make predictions. The predicted values of all the IMF components are added up to get the final prediction results. Simulation results based on measured data show that, compared with the single LSTM model and the PSO-LSTM model, the proposed method reduces the mean absolute percentage errors by about 11% and 7% respectively, showing that it can effectively improve the prediction accuracy. © 2019, Editorial Department of Journal of Chinese Inertial Technology. All right reserved.