Data-Based Estimation of Excitation Force in Wave Energy Converters

Wave energy conversion technology inherently requires tailored control system technology to improve energy capture performance, leading to a reduced cost of energy and allowing commercialisation in the short term. It is wellknown that optimal energy absorption requires instantaneous knowledge of the resource, i.e. the so-called wave excitation force. To date, most of the techniques available for estimating the wave excitation are model-based, often requiring a precise system and resource description, which is significantly difficult to achieve due to the inherent uncertainty introduced by standard practices in hydrodynamic modelling. Motivated by this issue, this paper presents a data-based approach to wave excitation force estimation for wave energy converters. Based on experimental data obtained from a comprehensive wave tank campaign designed to simulate diverse maritime conditions, this paper evaluates the performance of various neural network architectures, including static models and those incorporating temporal considerations. The analysis focuses on two input groups: kinematic variables alone and kinematic variables augmented with surrounding wave height measurements. The findings reveal that architectures incorporating temporal considerations and supplementary input features achieve superior performance, particularly in wide-banded sea states. These results highlight the capability of data-based models to achieve accurate estimation performance, reducing the need for precise analytical modelling while leveraging the ability of neural networks to extract information from auxiliary input signals.