Ship motion forecasting method based on intrinsic plasticity echo state network

Abstract

The invention provides a ship motion forecasting method based on an intrinsic plasticity echo state network, which comprises the following steps: acquiring historical data of swaying of a ship in a sailing process for modeling, inputting a target moment t, forecasting a motion attitude of the ship at the moment t based on an intrinsic plasticity echo state network model, and adopting correspondingmotion compensation according to a forecast value, wherein the internal plasticity echo state network is a plasticity neural network obtained by optimizing different neurons in the storage pool according to a plasticity rule; the invention is characterized by introducing IP (Intrinsic Plasticity) parameters a and b, constructing an activation function fgen (x) = tanh (ax + b) according to a plasticity rule established by KL divergence (kullback-leibler divergence, KL-divergence), so as to change a single neuron structure in a storage pool; the IP parameter optimization target is that the output of neurons in the reserve pool conforms to the maximum entropy distribution, and the purpose that the reserve pool can capture and carry more input information is achieved.

Description

technical field [0001] The invention belongs to the field of engineering signal processing, and relates to the analysis and prediction of the time series of ships swaying in sea waves, and in particular to a ship motion prediction method based on an inherently plastic echo state network. Background technique [0002] Affected by environmental factors such as waves, wind and ocean currents, ships will inevitably produce swaying motions during their navigation on the sea, which has caused many adverse effects on offshore operations. It is of great significance to model and predict the ship's motion attitude and take corresponding motion compensation measures to improve the safety, efficiency and reliability of offshore operations. [0003] Existing ship motion prediction methods are mainly prediction methods based on linear system theory, and the representative ones are auto-regressive (Auto-regressive, AR) prediction model, Kalman filter model and so on. However, under compl...

AI Technical Summary

Problems solved by technology

[0008] Therefore, the ship motion prediction method under the existing technology is difficult to achieve good prediction results under complex sea conditions, and the prediction accuracy of the ship motion prediction model is not high, which affects the safety, efficiency and reliability of the ship's offshore operations

Images