WIP: History support CSVMixin

[SGeeversAtVortech]

In GitLab by @AnneHommelberg on Oct 10, 2018, 10:18

Fixes #1010.

The optimization CSVMixin now supports history, similar to the PIMixin. The forcast date can be set through the csv_forecast_date field. If no forecast date is set, the first time of the input timeseries is used (no history).

Important note: previously the default history was used, which uses the initial_state. Now, by default the first values in the input timeseries are used. Since the initial_state is read from a separate file in the case of the CSVMixin this could result in different values. Should this be changed back to using the initial_state as a default?

[SGeeversAtVortech]

In GitLab by @TPiovesan on Oct 10, 2018, 15:21

It should be change back to using initial_state as default. @vreeken, right?

[SGeeversAtVortech]

In GitLab by @TPiovesan on Oct 10, 2018, 15:34

Commented on src/rtctools/optimization/csv_mixin.py line 324

In def seed(): you need to crop the timeseries such that it does not include the history. So

s = Timeseries(self.__timeseries_times_sec[self.__forecast_index:] , self.__timeseries[ensemble_member][variable][self.__forecast_index:] )

This will also fix the problem in def test_seed.

[SGeeversAtVortech]

In GitLab by @AnneHommelberg on Oct 12, 2018, 09:09

@TPiovesan Is that a bug in the PIMixin then? Because I haven't changed that code and there it is not cropped either.

[SGeeversAtVortech]

In GitLab by @vreeken on Dec 16, 2018, 20:07

@AnneHommelberg Can this MR be closed, as it's superseded by !227 ?

[SGeeversAtVortech]

In GitLab by @AnneHommelberg on Dec 17, 2018, 08:55

@vreeken Yes, !227 achieves the same thing.

[SGeeversAtVortech]

In GitLab by @AnneHommelberg on Dec 17, 2018, 08:55

closed