OpenEnergyPlatform¶

__copyright__ = "Reiner Lemoine Institut"
__license__   = "GNU Affero General Public License Version 3 (AGPL-3.0)"
__url__       = "https://github.com/openego/data_processing/blob/master/LICENSE"
__author__    = "oakca, Ludee"

import pandas as pd

# open df.xlsx
with pd.ExcelFile('df.xlsx') as xls:

    # save Sheet1 in df.xlsx as df
    df = xls.parse('Sheet1')

# show df
df

# Normalizing the df with melt (we keep the two columns 'Site' and 'Process'), we also add column unit, sort the values 
# and make sure no old indexes are used as columns.
# As a result we have the two kept columns 'Site' and 'Process' and all other columns are transposed into rows 
# labeled as variable and the respective values are found in value. The newly added column unit contains the respective units
norm = df.melt(['Site', 'Process']).assign(unit='').sort_values(['Site','Process']).reset_index(drop=True)

# assign values for unit
unit = {'inst-cap': 'MW', 'cap-lo': 'MW', 'cap-up': 'MW',
        'inv-cost': '€/MW', 'fix-cost': '€/MW/a', 'var-cost': '€/MWh',
        'wacc': None, 'depreciation': 'a'}

# include units in table
norm['unit'] = norm['variable'].map(unit)

# show normalized df
norm

# denormalizing the norm with pivot_table, transforms the table back into the original state (apart from column order)
denorm = norm.pivot_table(values='value', index=['Site', 'Process'], columns='variable').reset_index()

# remove the variable axis name
denorm = denorm.rename_axis(None, axis=1) 

# show denormalized df
denorm

# note: the only difference between denorm and df is that the column names are in alphabetic order

# In case of columns containing NaN values, pivot_table will omit these columns in the denormalized table.
# Hence, in this case another option is to use unstack():

# denormalizing the norm with unstack
denorm = norm.set_index(['Site', 'Process', 'variable'])['value'].unstack().reset_index()

# remove the variable axis name
denorm = denorm.rename_axis(None, axis=1) 

# show denormalized df
denorm