I had a great trip to the Wind Energy Science Conference 2017 last week (28 June 2017) in DTU, Copenhagen, where I saw lots of interesting presentations about wind power forecasting.
I gave a talk about our ESIPP research, comparing wind and solar observation data to reanalysis data and forecast data. I made the point that reanalysis data are not always in good agreement with local observations, and that sometimes forecasts at a higher resolution can outperform reanalysis at a lower resolution, even though the latter are driven by observations. This shows the importance of surface-driven (high-resolution) data on skill.
Presentation slides are available here: WESC2017 presentation