Closed peterdudfield closed 1 month ago
It seems like wind variables are not cumulative, so it shouldn't affect windnet too much, but definetly should retrain PVNet for india with it. And do diffs for all cumulative variables in the future, as I don't think any model will handle it very well, and normalization won't work very well for this
This is solved by #314. The only thing not covered would be to add normalisation values for the diffed ECMWF variables over India
Shall i close this now @dfulu ?
Shall i close this now @dfulu ?
Yeh sure
There are some variables that are cumulative.
This is quite tricky for the model to handle, as delayed in data can have a real impact. OF course the model should learn to take the difference but, it perhaps is easier to give the model what we want
Tasks