fcas_semi_scheduled may be empty

[andrewhn]

Hi,

Great library, thanks for writing it and making it accessible to everyone.

I think there may be an issue in running earlier years that don't feature semi scheduled gens offering FCAS.

To reproduce what I think is the error, you can run the example code in README and substitute references to 2019 with 2015. I get an error in units._scaling_for_uigf which seems to arise because the fcas_semi_scheduled data frame is empty.

My workaround was to add a check for whether the data frame was empty:

    if not fcas_semi_scheduled.empty:
        # Scale high break points.
        fcas_semi_scheduled['HIGHBREAKPOINT'] = \
            fcas_semi_scheduled.apply(lambda x: get_new_high_break_point(x['UIGF'], x['HIGHBREAKPOINT'],
                                                                         x['ENABLEMENTMAX']),
                                      axis=1)

        # Adjust ENABLEMENTMAX.
        fcas_semi_scheduled['ENABLEMENTMAX'] = \
            np.where(fcas_semi_scheduled['ENABLEMENTMAX'] > fcas_semi_scheduled['UIGF'],
                     fcas_semi_scheduled['UIGF'], fcas_semi_scheduled['ENABLEMENTMAX'])

        fcas_semi_scheduled.drop(['UIGF'], axis=1)

        # Combined bids back together.
        BIDPEROFFER_D = pd.concat([energy_bids, fcas_not_semi_scheduled, fcas_semi_scheduled])
    else:
        # Combined bids back together.
        BIDPEROFFER_D = pd.concat([energy_bids, fcas_not_semi_scheduled])

The tests I have done so far are very close to what NEMDE produces so all seem to be working.

Happy to look at it more and submit a PR if helpful.

[nick-gorman]

Hi @andrewhn

No worries, glad your finding it interesting.

Thanks for pointing out this problem, investigating, and suggesting a fix.

Interesting that it's never come up as an issue before, I would have thought there'd be times in 2019 when no semi scheduled units were bidding into FCAS.

If you want to investigate further and submit a PR that would be really appreciated, otherwise I will try and look at it after addressing #6

Cheers,
Nick

[nick-gorman]

@andrewhn this should be fixed in the latest release, update with pip, see notes here: https://github.com/UNSW-CEEM/nempy/releases/tag/v1.1.3