[NEMO-devel] use of Pandas data frames

[Ben Elliston]

Hi all

I've been making some more invasive changes to NEMO on a branch which
are now ready to be merged back into the master.  In the past, NEMO
kept a lot of its time series data (demand, generation, unserved
events) in one- or two-dimensional arrays indexed by the hour number
(eg 0 to 8760 for a one year simulation).

Pandas is a Python data analysis library that does a very nice job of
storing and manipulating time series data (http://pandas.pydata.org/).
Where it makes sense, I have switched various internal representations
from arrays to Pandas "data frames".  A data frame has an index,
typically the time/date, and columns.

Here is an example of how much better this works:

>>> import nem
>>> c = nem.Context()
>>> c.generators[0].set_capacity(13)  # produce some unserved
>>> nem.run(c)
>>> print c.unserved
Date_Time
2010-01-11 13:00:00    288.360
2010-01-11 14:00:00    445.200
2010-01-11 15:00:00    645.140
2010-01-11 16:00:00    313.530
2010-01-12 13:00:00    445.860
2010-01-12 14:00:00    378.365
2010-02-09 15:00:00    123.815
2010-02-10 13:00:00    121.550
dtype: float64

Cheers,
Ben