Post-processing for wind and PV power production of hectometric NWP forecasts - which Machine Learning methods are beneficial for sparse data and extreme events?

With the increasing amount of wind and solar energy fed into the European powergrid, despite slowdowns due to social acceptance and regulatory issues, plus thetransition to a fossil fuel-free energy production, accurate predictions includinguncertainties are required for grid operators. Relying heavily on renewable energysources, frequently updated and as accurate as possible predictions for bothhigh-resolution temporal and spatial scales ensure grid management and takingprevention measures in case of extreme meteorological events affecting the powerproduction. Moreover, both extreme events in weather across the nowcasting to weeks aheadtime scale and combined non-necessarily extreme weather factors can notably affect the power production grid. The latter is sometimes a combination of just-above-normal events, such as  high solar penetration, high wind penetration plusdecent hydropower combined with a reduced electricity demand. Thus, on-demandavailable predictions of the expected power production are especially needed. Post-processingmethods enable targeted forecasts of meteorological parameters at site-location andregional level, serving as a baseline for the conversion to power production,particularly a direct conversion of NWP predictions and observations to powerproduction.  However, so far most methods available use NWP forecast with spatial resolutionsbetween 2 and 9 km, and hourly output frequency thus requiring temporal interpolation andspatial downscaling or point interpolation. Recently, NWP upgradestowards sub-km scale and sub-hourly have been done to improve extreme events prediction. These NWP models are onlyavailable for selected extremes and short periods of time and often do not share thesame parametrization throughout. This poses a challenge for post-processing forrenewables, as well as the uncertainties that lie in e.g. wind farmspecifications and solar farm/PV panels. To prepare for the next phase of NWPmodels, and currently developing on-demand extreme digital twin forecasting system, fast andtransferable post-processing methods are needed. Here, we look into differentmachine learning and classical statistical methods, such as the analog method,LSTMs and Random forest, EMOS, to generate post-processed forecasts for extremeevents with a sparse training database. Furthermore, we investigate thetransferability and generalisability of these methods when pre-trained with acoarser NWP model.