Data-driven modelling on power generation of wave-powered USV

Abstract

Unmanned surface vehicles (USVs) have been widely used in various fields credited for their high flexibility and excellent manoeuvrability. However, the limited onboard electricity of USVs restricts their applications in long-term working tasks. A promising option to address that problem is employing a wave-powered USV, where a wave energy converter is integrated into the USV. For a wave-powered USV, it is challenging to achieve a precise estimation of its power generation under complex marine environments. Although there are physics-based models of WECs used for the estimation and prediction of power generation of USVs, they cannot achieve satisfactory accuracy in practice because of the linearization and simplification of those mathematical modelling. To fill up this gap, this paper proposes a data-driven modelling method utilizing feature engineering and ensemble learning to estimate the mean power generation of a wave-powered USV, relying on time-serial measurements obtained from real physical wave-flume experiments under regular waves. Results show that the estimation achieves promising performance, where the minimum RMSEs are 3.359 mW and 6.148 mW for a harvester-to-wire model and a wave-to-wire model, respectively. Additionally, the proposed method shows great potential for real-sea applications, such as online power prediction and control.