One system, read out at three levels

This is one system rather than three products. The same terrain, the same flow model and the same validation loop stand behind all three read-outs below; what changes between them is how far the answer is carried — to the wind at each machine, to the power that wind makes, or to what that power earns at your settlement point.

Each level answers the same four questions in the same order, so they can be read against each other: what you get, at what resolution and horizon, how it is delivered, how it is validated. They are numbered because the order is real — the generation forecast is computed from the wind field, and the revenue forecast from the generation — and taking the read-out at one level does not require taking the ones above it.

Common to all three

Coverage: [COVERAGE_REGION]
Horizons: [HORIZONS]
Update cadence: [UPDATE_CADENCE]
Refresh latency: [REFRESH_LATENCY]
Delivery format: [DELIVERY_FORMAT]
Coordinates to first forecast: [ONBOARDING_TIME]

The layers

Meter-level wind forecasting

Hub-height wind at each turbine, resolved over the site's real elevation with the operational forecast setting the conditions at the edges of the domain — not interpolated from forecast-model output.

What you get: [SERVICE_1_WHAT]
Resolution & horizon: [SERVICE_1_RESOLUTION] · [HORIZONS]
Delivery: [SERVICE_1_DELIVERY]
Validation: [SERVICE_1_VALIDATION]

Turbine generation prediction

That wind converted to power through a curve calibrated on observed generation, with a learned correction for turbine-to-turbine wake losses trained across multiple wind farms.

What you get: [SERVICE_2_WHAT]
Resolution & horizon: [SERVICE_2_RESOLUTION] · [HORIZONS]
Delivery: [SERVICE_2_DELIVERY]
Validation: [SERVICE_2_VALIDATION]

Pricing

Predicted generation carried through to the ERCOT nodal price at your settlement point, and reported as revenue on the same horizons as the two layers above.

What you get: [SERVICE_3_WHAT]
Resolution & horizon: [SERVICE_3_RESOLUTION] · [HORIZONS]
Delivery: [SERVICE_3_DELIVERY]
Validation: [SERVICE_3_VALIDATION]

Standard and tuned

Two ways to run a site

These are alternatives, not steps. Standard is the finished product: it runs on ground the model has never seen, and the numbers in Results are what it does in exactly that condition. Tuned is opt-in, for sites where the measured error says there is margin worth narrowing — a model that is right about turbulence still knows nothing about one machine's particular history, the soiling, the curtailment behaviour, the derates, the way a row of turbines performs in a south-easterly. Those are not fluid dynamics, so no simulation contains them and only measurement carries them. Which of the two a site is remains a measured question rather than a matter of appetite: every site is validated against ERCOT 60-day SCED generation data before anything goes into production, and any number reported for a tuned site is measured on a period held out of the tuning.

Standard

Every site

What you get: The whole forecast, on ground the model has never seen — nothing is fitted to you. Your terrain and the operational forecast are imposed on a model trained on turbulence, so there is no training run and no record to supply first. Coordinates, hub heights and a settlement point are all it takes.
Coordinates to first forecast: [ONBOARDING_TIME]

Tuned

Opt-in

What you get: Opt-in, and the one place anything is fitted to your site: the flow model is tuned on the site's own measured record, and the generation and pricing layers are recalibrated on that same record. Not on more simulation — on what the site actually did.
Lead time: [TUNED_LEAD_TIME]
What it asks of you: Measured data, and enough of it: observed generation, plus whatever is measured on the site where it exists — met mast or nacelle wind speed and direction, availability and curtailment records. At whatever resolution you keep them; nothing has to be prepared first.
Record needed: [TUNED_HISTORY]
Uplift (held-out period): [TUNED_UPLIFT]
Candidate when: The validation says so — error worse than the standard forecast reaches elsewhere, in a regime the record can explain: unusually complex ground, a wake regime the general model handles less well, a horizon where the margin is worth narrowing. Read off the measurement, not off appetite. Where the validation does not show it, standard is the answer and we say so.

How it runs

These are engagements, not plans

What a new site needs is the turbine coordinates, the hub heights and the settlement point. The ground and the operational forecast do the rest, and coordinates to first forecast is [ONBOARDING_TIME]. Historical generation is welcome and is what the validation is run against, but it is not a precondition: a site with no usable history still gets a forecast.

Every site is validated against ERCOT 60-day SCED generation data before anything goes into production, and you see that error first. What lands in your systems is described above under Delivery, at the cadence under Update cadence. If it does not beat what you have on your own ground, you will see that in the validation.

[CTA_LABEL] [CONTACT_EMAIL]