Time- and Frequency-domain Comparisons of the Wavepiston Wave Energy Converter

Analysis of wave-energy converters is most frequently undertaken in the time-domain. This formulation allows the direct inclusion of nonlinear time-varying loads such as power take-off (PTO) reactions, mooring forces, and viscous drag. However, integrating the governing equations of motion in the time domain is relatively computationally expensive, and requires a simulation to be conducted for each incident-wave state. In contrast, calculating the linearised performance of a wave energy converter (WEC) in sinusoidal waves of a given frequency is relatively inexpensive, albeit with the lower accuracy associated with the assumption of linearity. Combining this frequency-domain information with aspectral characterisation of the sea state therefore offers an opportunity to predict the power-capture performance of a WEC with less computational expense than a direct time-domain approach. In this regard, methods such as spectral domain linearisation (Folley, 2016) and nonlinear frequency domain analysis using a basis of trigonometric functions (Mérigaud and Ringwood, 2018) have been proposed to provide a compromise between speed and accuracy in assessments of WEC performance.This paper will compare time- and frequency-domain analyses of the Wavepiston surging-plate WEC. This device consists of a surging plate close to the free surface that drives a staged telescopic hydraulic PTO. Modelling this system in the frequency domain presents challenges associated withthe signiffcant nonlinear forces arising from both the PTO reactions and the non-negligible viscous drag acting on the plate. Equivalent linear damping coeffcients are used to model these forces in the frequency domain, while they are included explicitly in the time domain. The main idea of this paper is to quantify, for this device, the errors associated with linearising these two nonlinear processes. Ouraim here is to assess the trade-offs between speed and accuracy when using a fully-linear frequency-domain approach compared to a partially-nonlinear time-domain method