Abstract
As one of the renewable energy resources, wave energy has the potential to be one of the major electricity generation resources in the long term. However, the industrial-scale implementation of wave energy is still at an early stage and requires deeper research and development. Currently, only several different wave energy conversion techniques such as oscillating bodies, oscillating water columns, and over-topping modules, are available. Among these energy converters, the 6-float, multi-mode, moored wave energy converter (WEC) M4 has been developed by the University of Manchester. M4 consists of a bow float, three mid floats and two stern floats with beams hinged above the outer mid floats. The mooring to the bow float includes an elastic cable from the bed to a spherical buoy and an inelastic cable linking the buoy to the bow float. The M4 converter has been modelled numerically. However, these numerical methods often require large computational resources and therefore they can only be used offline and they cannot be implemented in real-time applications. It is highly desirable to develop a computationally efficient model for the M4 system.
In this paper, system identification methods are used to model the M4 WEC system with the experimental data. First, two linear models including autoregressive with exogenous inputs (ARX) and the Box-Jenkins (BJ) model are used to build three different types of models for 6 float M4 WEC system using three different input and output variables: wave surface wave surface elevation, the mooring forces (top and bed forces) and the bow float motions (pitch, surge and heave). Further, when constructing the model of the M4 system, the impacts of both sampling rate and model orders are considered. The modelling results are compared and analysed using three selected criteria including fitting percentage, mean squared error and final prediction error. Finally, the performance of identified models are validated by wave basin experiments which covers seven major wave conditions including different wave heights and mean wave periods. Results show that both ARX and BJ models are feasible to model the M4 WEC system, with low orders and sampling rates for high short-term prediction accuracy. These models could be potentially used for real-time applications