Pipeline

1 day ahead forecast

if we want to build mode with high performance we should include these features from ENSTO-E :

• load/demand
• power generation
• power generation from wind & solar (already train a model for each ) However, ENTSO-E only provides forecasts for these features up to 1 day ahead. Therefore, to forecast beyond that, we shift to using alternative features such as weather data. This comes with a trade-off of lower performance. Below is the pipeline that would have been used if these features were available for longer-term forecasts:

7 days ahead forecast

this pipeline use weather data like

• temperature
• precipitation
• cloud cover
• wind speed (10m)
• sunshine duration The other features remain unchanged.

Here is the current pipeline :

other folders

prices model

• fetch is done (each script run would concatenate the existing data with the new ones)
• still train need to be tweaked

wind model

• smart fetch is done

PV model

• Smart fetch done

Smart fetch

is used in both wind and solar to fetch wind speed and GHI you can find it in each folder under the name smart_fetch.py

Smart fetch : Automatically picks the right data endpoint for a given date range.

• Past‑only window → calls the archival API.
• Future‑only window → calls the forecast API.
• Mixed window → calls both, then stitches the results chronologically.

The caller just supplies coordinates and the desired period; the helper handles API choice, merging, and ordering behind the scenes.

TO DO

• add hourly observatory formulas ( need to be tweaked)
• add wind PV formulas ( using new formulas of weights )
• add power generation from wind solar models ( this wont be done for now cuz we are using weather data in pipeline 2 instead of original features)