How to generate windspeed forecasts when they aren’t provided with AI model output

https://rmets.onlinelibrary.wiley.com/doi/10.1002/met.70038

Last summer a very able AI scientist joined Lake Street and together we created a way to generate forecasts for variables not included in the ECMWF AI forecast output.

Lucas’ did the majority of the work in writing this up and it’s now published as a peer reviewed paper in Meteorological Applications as Leveraging state-of-the-art AI models to forecast wind power generation using deep learning.

ECMWF have since included 100m winds in their AI forecasts. Once there’s enough data, we’ll compare our method with theirs.

A great example of cross pollination of ideas, which Lake Street thinks is essential for weather AI to fulfil its potential