Ship motion prediction using dynamic seasonal RvSVR with phase space reconstruction and the chaos adaptive efficient FOA

Abstract

Affected by coupling effect of 6 DOF motions, especially by the chaos characteristic, variable periodicity and noise signal of ship motion time series, it is difficult to obtain precise forecasting results of ship motion. In order to improve the forecast precision, firstly, aiming at the chaos characteristics of ship motion time series, based on the theory of the space reconstruction, this paper, employed the G-P method to determine the embedding dimension m, selected the mutual information to calculate the delay time τ, and established a chaos system reconstruction method appropriate for ship motion prediction, for rebuilding the chaotic systems of ship motion time series; Then, directing at cycle variability of the ship motion time series under different working conditions, a dynamic seasonal adjustment mechanism(namely DSAM) was designed, in view of the ship motion time series contain noise signals, robust loss function was introduced to the SVR model, and a new dynamic seasonal robust v-support vector regression forecasting model, namely DSRvSVR, was proposed and used to simulate the built chaotic systems of ship motion time series; Thirdly, in order to obtain more appropriate parameters of the DSRvSVR model, considering with shortcomings of FOA, this paper designed adaptive efficient flight guidance law (AEFGL), establish global chaos perturbation mechanism (GCPM), and established a chaos adaptive efficient fruit fly optimization algorithm, namely CAEFOA; Finally, coupling the proposed PSR method, DSRvSVR model and CAEFOA, a hybrid forecasting approach for ship motion forecasting, namely PSR–DSRvSVR–CAEFOA, was established. Subsequently, the ship heave time series under four working conditions were used to conduct numerical example, to test forecast performance of the proposed PSR–DSRvSVR–CAEFOA approach and optimize performance of CAEFOA. Analysis results showed that the proposed hybrid forecasting approach receive better forecasting performance compared with classical prediction models selected in this paper, and the CAEFOA obtain higher optimization efficiency than FOA.