Gap in blended values due to duplicate PV data values. This was solved in nowcasting_forecast=1.3.17 and deployed this morning on production
We should be able to solve this in general with the blending method. I.e if one forecast is not available then we load a different one. Some of this behaviour is already built in to the blend module, but clearly not enough. For example if no forecast has been made in 2 hours, then we don't use this forecast.
Note we currently mix between
0-1 hours: PVnet 1.0
2-7 hours: PVnet 2.0
8 hours: Natinoal_xg
I would propose a solution with a test, that if one forecast is not there, (or no values) then we load the closest ones.
Gap in blended values due to duplicate PV data values. This was solved in
nowcasting_forecast=1.3.17and deployed this morning on productionWe should be able to solve this in general with the
blending method. I.e if one forecast is not available then we load a different one. Some of this behaviour is already built in to theblendmodule, but clearly not enough. For example if no forecast has been made in 2 hours, then we don't use this forecast.Note we currently mix between
I would propose a solution with a test, that if one forecast is not there, (or no values) then we load the closest ones.
FYI: I would like to pull this blend method out to a different module: https://github.com/openclimatefix/uk-pv-forecast-blend/issues/1 and then use Airflow to trigger it once a forecast is complete