A Speed Prediction Method of Electric Propulsion Ship Based on GA-Elman Neural Network

Abstract

With the development of shipping industry, more and more attention has been paid to the environmental pollution of waterways and surrounding areas. Accurate prediction of ship speed plays an essential role in ship operation optimization and decision support, and is also one of the key means for ships to achieve energy conservation and emission reduction. Traditional methods of ship speed estimation are mainly based on hydrodynamics and full-size measurement, which have obvious disadvantages such as large computational workload or large measurement investment. In this study, through the analysis of historical ship speed data, a data-driven ship speed prediction approach based on the Elman neural network model is investigated. Furthermore, Genetic Algorithm (GA) is applied to optimize the weights and thresholds of Elman neural network to improve the prediction accuracy and computational efficiency of the algorithm. Finally, the proposed GA-Elman neural network model is experimentally verified by the actual navigation data of an inland electric propulsion vessel, and compared with the BP neural network model and the Elman neural network model. Experimental results show that the speed prediction method based on GA-Elman neural network proposed in this paper has higher prediction accuracy, and the prediction error is reduced by 56.67% and 48% compared with the BP neural network model and Elman neural network model respectively.