The turbine is now being asked to announce its own output
the turbine is now being asked to announce its own output: a wind power foundation model pretrained on 126,000 u.s. sites over seven years forecasts power zero shot from terrain and weather, xgboost reaches an r squared of 0.99 across multiple farms, and a fourier neural operator fuses scada and numerical forecasts for site level predictions.
the turbine now files its own power budget.
Context
The Tyan-WP paper describes a wind power foundation model pretrained on a dataset covering more than 126,000 U.S. wind turbine sites over seven years, aimed at zero-shot probabilistic forecasting at new or data-scarce sites, with static site embeddings. The PLOS One study compares XGBoost, Random Forest and support vector regression across four geographically different datasets and reports R-squared values of 0.99 for XGBoost and linear-kernel SVR in all locations.
The Fourier Geometric paper presents a multimodal framework that joins historical point-based SCADA turbine data with grid-based numerical weather prediction forecasts, using a Fourier Neural Operator architecture, for short-term wind power forecasting.
The note joins three separate studies. The 126,000 sites and seven years belong to Tyan-WP. The 0.99 R-squared comes from the PLOS One paper, and that paper's task is forecasting wind speed from four datasets, as the pages read describe it, not site power from a foundation model. The note reads it as wind power across multiple farms, which is a looser description.
The Fourier Neural Operator and SCADA fusion are from a third paper. 'Zero shot from terrain and weather' fits the Tyan-WP abstract's description of site attributes and zero-shot use. The paper's own accuracy figures were not compared here, so how it does against other models is unsupported here, not refuted.
'The turbine now files its own power budget' is the author's line. These are forecasting models, and none of them is a turbine reporting on itself.
Related work
- Tyan-WP: A Wind Power Foundation Model for Ultra-Short-Term Probabilistic Forecasting (arXiv 2606.08630) ↗Source for the 126,000 sites, seven years and zero-shot design.
- A robust multi-location evaluation of a machine learning framework for wind power forecasting (PLOS One, April 30, 2026) ↗Source for the XGBoost and SVR R-squared of 0.99 and the four datasets.
- Fourier Geometric Wind Power Forecasting with Numerical Weather Prediction (arXiv 2607.17095, ACM 2026) ↗Source for the Fourier Neural Operator, SCADA and weather forecast fusion.
Watch next
- Check what the PLOS One study forecasts and its validation split. Read Tyan-WP's reported errors on unseen sites.
Sources
- Tyan-WP: A Wind Power Foundation Model for Ultra-Short-Term Probabilistic Forecasting (arXiv 2606.08630)arxiv.org
- A robust multi-location evaluation of a machine learning framework for wind power forecasting (PLOS One, April 30, 2026)journals.plos.org
- Fourier Geometric Wind Power Forecasting with Numerical Weather Prediction (arXiv 2607.17095, ACM 2026)arxiv.org
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 3 October 2026 at 19:55 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 →