Gaussian Process Regression for extending wave in situ measurements: An experimental campaign

Wave energy is recognized as one of the most promising sources of clean and abundant energy. Nonetheless, up to today this technology is still not commercially viable, due to a number of reasons, like the harshness of the sea environment, the expenses needed for the deployment and maintenance of devices in openocean and the lack of information regarding wave parameters world wide. Indeed, a proper characterization of the resource in asite is of quintessential importance for assessing the productivity of the site and dimensioning the supporting system of a device. This work wants to address the problem of the lack of data by resorting to spatial prediction techniques, using data gathered through an experimental campaign conducted at the wave basinfacility available at the Ocean and Coastal Engineering Laboratory in Aalborg University. During this campaign two months of real data from a real in situ measuring device were replicated in the basin. In the middle of the basin, underwater, some concrete blocks were deployed in order to replicate a sudden shift in the bathymetry, which should act as a disturbance to the wave propagation and arise nonlinear phenomena. 19 wave gauges were present and recorded the wave elevation for the whole time. Thena scenario where only a part of the measuring devices were workingwas replicated by considering only the data from a subsample of wave gauges and inferring the parameters in the locations ofthe other devices from them, through a Gaussian Process Regression(GPR) algorithm. The proposed algorithm was able to interpolate the parameters at the other locations, at the expense of a relatively low error, indicating that this set up could be used to increase the spatial coverage of the wave measuring buoys deployed world wide or to provide an estimate of the parameters at a buoy that is not working at the time, like for maintenance operations.