In order to become economically viable and compete with other more traditional energy sources in the energy market, offshore renewable energies, such as wave energy converters (WECs), tidal energy converters and even floating offshore wind turbines, still need significant development. Key aspects for the development of these technologies include: (i) optimising the designs by reducing material use, (ii) increasing the energy generation capacity, (iii) enhancing the durability of key components, and (iv) improving accessibility and availability. All these aspects rely on accurate metocean data and the use of incomplete or inaccurate metocean data can incorporate a higher uncertainty to a design process; an area where the uncertainty level is already significant.
The uncertainty in metocean data affects the development of any stage in the chain, from the response of the system to the final energy generation capabilities. The most common sources of metocean data in the ORE sector are observation buoys and re-analysis datasets. However, long-term resource variations suggested in various recent studies suggest that long datasets are required for a better understanding of the resource. Similarly, different international organisations recommend considering relatively long periods of data. Therefore, re-analysis datasets, with decades of data since 1950, seem to be crucial, but inaccuracies of such datasets are well-known.
As a consequence, the present paper suggests different statistical bias-correction techniques in order to improve the quality of re-analysis datasets: (i) the delta method, (ii) linearly-spaced quantile-mapping (LQM) and (iii) Gumble-distribution-based quantile-mapping (GQM). In this sense, first the bias of the re-analysis ERA5 dataset is evaluated in three different locations: Bay of Biscay, off the West coast in Portugal and off the West coast in the US. That way, the capacity of the different statistical bias correction techniques will be studied, identifying the advantages and disadvantages of the different techniques, and quantifying the uncertainty in the resource re-analysis data.
Finally, the impact of this uncertainty on the design parameters is studied. Considering that one of the most relevant design parameters is power generation, differences in power generation estimations based on different resource datasets are assessed in the three locations. Figure \1 illustrates part of these results, showing the difference in the annual mean power production (AMPP) on the left and the variation in mean-to-peak ratio on the right. The underestimation of the raw ERA5 re-analysis is clear (about 40%), which demonstrates, on the one hand, the perils of using raw re-analysis data, and, on the other, the need for adequate bias-correction techniques. However, this conclusions are very dependent on the specific location and resource characteristics and, thus, the same analysis is conduction in geographical locations with significantly different resource conditions.