Day-ahead auction · Austria
Austrian electricity
price forecasts
Forecast and actual auction prices by delivery day, with forecast errors and the latest recorded job results.
01 / Prices
Forecast and actual prices
Loading…
View interval data ↗
| Interval start | Forecast €/MWh | Actual €/MWh | Fallback |
|---|
02 / Evaluation
Forecast errors
Daily mean absolute error · €/MWh
03 / Models
Model comparison
| Model | Role | Last 7 days · MAE €/MWh | Last 28 days · MAE €/MWh |
|---|
04 / Method
How forecasts are calculated
Official forecasts are made by the published model. The model behind each day is named under its chart, and section 03 compares all tracked models on the same days.
Official inputs freeze at 10:30 in Vienna on the preceding day. Publication is due by 10:45. Daylight-saving changes are handled explicitly.
05 / Operations
Latest job results
06 / MLOps loop
How a change reaches production
Every improvement travels the same circuit. A change is planned from measured error, tested, built once into a pinned image, deployed by its digest, and scored again once it runs. The measurement is what closes the loop and decides the next change.
-
01
Plan the change
Choose the next change from measured error
Design · DESIGN.md -
02
Develop
Write the code and its tests in src/
Python 3.12 · uv -
03
Test locally
Unit tests, lint and format checks
pytest · Ruff -
04
Review in CI
Run the same checks on every push and pull request
GitHub Actions -
05
Build the image
Build once in CI and push an immutable digest
Docker · ECR -
06
Deploy the image
Point the function at the pinned digest
ECR digest · Lambda -
07
Run the daily jobs
Collect, publish, settle and watch, on Vienna time
Lambda · EventBridge -
08
Measure & evaluate
Score against actual auction prices, then plan again
Scoring · MAE / RMSE