Influence of Large-Scale Climate on Offshore Wind-Farm-Scale Environment: A Generative Deep Learning Approach

Characterization of the variability of local, wind farm scale environment is necessary for better developing offshore wind energy infrastructure. On the one hand, spatiotemporal variability at such scales is dominated by the variability of small-scale processes such as coastal low level jets (CLLJ), mesoscale fronts, and internal boundary layers. On the other hand, because of the complex multiscale nature of the climate system, large scale climate phenomena such as marine heatwaves and atmospheric blocking and modes of large-scale climate variability such as El Nino Southern Oscillation, Pacific Decadal Oscillation and North Atlantic Oscillation are likely to contribute to the variability of the local environment as well. Such influences could be through changing land-ocean temperature contrast, changing boundary-layer stratification, etc.,---aspects of the local environment that in turn condition smaller scale processes. Since high resolution simulations of the global climate system are computationally expensive and since high resolution data is sparse, we investigate the use of deep-learning to increase the signal to noise ratio and facilitate the examination of the influence of large-scale climate on the wind farm-scale environment. We present preliminary results from a study that uses a reanalysis-driven, regional, high resolution Weather Research Forecast (WRF) simulation as reference and develops a generative model---a conditional Variational AutoEncoders (cVAE)---to augment wind farm scale data conditioned on large-scale climate data.