The value of artificial intelligence: Lessons learned from an Irish offshore windfarm metocean assessment

Presentation Description: This presentation will aim to demystify AI in offshore wind through application in a real life case study. A particular focus will be given to fill in the missing metocean data for potential sites. Offshore wind site characterisation needs long term datasets to draw conclusions about wind, wave, current and water levels. Potential sites have lidars and buoys deployed to obtain in-situ data. However, in-situ data collection is expensive and often suffers breakdowns, causing significant data to go unrecorded or recorded anomalously. Data imputation can fill these gaps in a statistically consistent manner and can offset the impact of missing data to a certain extent. The choice of models relate to the nature of missing data and the comparative accuracies of various techniques when applied to in-situ data from site. Here, time series analysis based imputation models are performed on wind, wave and current data gathered hourly from the Kinsale platform and the M1-M5 buoys around Ireland. The intercomparison of models markers not only provide the a quantitative estimate of the performance of the models in this sector, but also a guidance around the choice, comparison and limitations of their use, analyses and subsequent interpretation. In particular, a focus of this work has been thus on missing data in blocks, rather than those missing at random. The results are expected to inform modelers of metocean data in terms of expectation of performance from imputation models.