The wind farm is now being asked to file its own output forecast
the wind farm is now being asked to file its own output forecast: a foundation model pretrained on 126,000 american sites over seven years forecasts power zero shot, a temporal super resolution model stretches 5 minute wind speeds into 1 minute steps from 58 million minutes of observations, and a time moe fine tune adapts to a new farm with just three days of data.
the turbine now files its own schedule.
Context
The foundation model is Tyan-WP (arXiv 2606.08630), described as the first wind power foundation model for ultra-short-term probabilistic forecasting, pretrained on a wind power dataset covering more than 126,000 US sites over seven years, with static site embeddings and a power-aware meteorological fusion module to improve zero-shot forecasting. A separate model, WindFM (arXiv 2509.06311), is pretrained on about 150 billion time steps from more than 126,000 sites of the WIND Toolkit and has 8.1M parameters.
The super resolution model is WindFLOW (NREL, March 2026): a pretrained TimesFM Transformer with a three-layer head that downscales 5-minute wind speed into 1-minute sequences, evaluated on over 58 million minutes of observations from seven wind sites, and reported to beat deep learning and classical interpolation baselines.
The note's 126,000 sites matches both Tyan-WP and WindFM, and the seven years matches Tyan-WP only, so the note most likely refers to Tyan-WP, which was not confirmed beyond that match. The 58 million minutes is from seven sites. The Time-MoE fine tune on three days of data was not located: the Time-MoE repository describes fine-tuning but the excerpt read gives no three-day result, so it is unsupported here, not refuted.
Watch next
- The source of the three-day fine-tune result. Tyan-WP zero-shot error figures.
Sources
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 4 October 2026 at 08:18 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →