Forecasts that sit next to your live data.
Mesoscale irradiance forecasts blended with on-site pyranometer, reference-cell, and module-temperature data — for 0–72 h power-output forecasts your dispatch desk can trade on.
The problem
Most operators get irradiance forecasts from a third party in CSV or via a separate portal, then manually adjust for module temperature in a spreadsheet. The model isn't tuned to the site, the on-site pyranometers that would calibrate it publish to a different system, and nobody recomputes the bias correction more than once a quarter. Imbalance penalties are the result.
How SolarTwin solves it
- Mesoscale forecast ingest — ECMWF (HRES & ENS), CAMS, DWD ICON, and satellite-derived nowcasts (Meteosat / GOES) for 0–6 hour horizons.
- Site-tuned bias correction using your last 30–90 days of measured vs. forecast irradiance. Recomputed continuously, not quarterly.
- Pyranometer & reference-cell fusion — on-site observations weighted into the short-term horizon (0–6 h) where they outperform the mesoscale.
- Power-output forecasts derived from your plant's IV characteristic, module temperature coefficients, and live inverter availability — so a tripped inverter doesn't poison the portfolio number.
- Curtailment-aware — known clipping limits and dispatch instructions roll into the forecast automatically.
- Forecast vs. actual scorecards updated every day, so you can prove the value to the trading and treasury teams who fund the platform.
How it works
- SolarTwin pulls forecast grids (NetCDF / GRIB2) from the configured providers on each model run.
- Irradiance is interpolated to each tracker's plane of array and bias-corrected against the site's measurement history.
- Power output is derived per inverter using the plant model, then aggregated to plant and portfolio level — accounting for live availability and curtailment.
- Outputs land in the same dashboards as live SCADA, in the same API, and in the same exports.
What you can do with it
- Submit nominations and intra-day re-nominations from the same screen you operate the plant in.
- Plan maintenance windows around forecasted low-irradiance periods — without juggling two systems.
- Defend forecast performance to the trading desk with daily MAE/RMSE scorecards.
Specs
- Sources: ECMWF HRES & ENS, CAMS, DWD ICON, satellite nowcasts (add more on request).
- Horizons: 0–72 h core; 0–10 days as supporting context.
- Granularity: per inverter, per plant, per portfolio.
- Bias correction: rolling 30/60/90 day window, automatic re-fit.
- Available via dashboard, REST API, and scheduled CSV export.
See it with your own plant
Send us a month of measured irradiance from one plant. We'll backtest a SolarTwin forecast against it and show the result.
Book a walkthrough