Since about 2008 I've been working on reducing CO₂ emissions enough that the climate actually notices. I'm co-founder of Open Climate Fix (a non-profit).
My main interest at the moment is to enable anyone with a laptop, an Internet connection, and a little knowledge of machine learning to train and run state-of-the-art energy forecasts (by lazily loading historical and live gridded weather forecasts from multiple weather forecast providers). We're a long way from this "dream" right now! But we're getting there! Please see this blog post for more info on this idea.
This should have a number of use-cases and benefits:
- Academics and students can easily experiment with energy forecasting (using realistic datasets).
- Companies can build state-of-the-art energy forecasts in-house (e.g. grid operators, battery optimisers, etc.).
- We can finally build a public leaderboard of different energy forecasting algorithms, measured against a standard validation dataset (again, using the types of huge datasets that are used in industry, rather than toy academic datasets). Anyone could contribute algorithms to the leaderboard.
- Maybe we could run a simple and cheap energy forecasting service with state-of-the-art performance.
My current project in this vain is hypergrib.
Once hypergrib is up-and-running, I'm planning to return to researching ML algorithms for energy forecasting, and continuing to build open-source tools for energy forecasting.
Please say "hi" on Bluesky. I'm no longer active on Twitter/X or Mastodon.
I can't take any credit for these projects but these projects all help towards the aim of making it super-easy for folks to run energy forecasts:
(alphabetical order)