def hourly(daily_df, classes, parameters):
def hourly_factors(building):
# This function selects hourly factors from BGW 2006 by time and temperature class
slp = pd.DataFrame(index=classes.index, columns=classes.columns)
# Time includes the hour of the day
times = classes.index.map(lambda x: x.strftime('%H:%M'))
# For commercial buildings, time additionally includes the weekday
if building == 'COM':
weekdays = classes.index.map(lambda x: int(x.strftime('%w')))
times = list(zip(weekdays, times))
for column in classes.columns:
slp[column] = parameters[building].lookup(times, classes.loc[:, column])
return slp
buildings = daily_df.columns.get_level_values('building').unique()
results = pd.concat(
[upsample_df(daily_df, '60min')[building] * hourly_factors(building) for building in buildings],
keys=buildings, names=['building', 'country', 'latitude', 'longitude'], axis=1
)
return results.swaplevel('building', 'country', axis=1)
def temperature(input_path, year_start, year_end, mapped_population):
parameters = {'air': 't2m', 'soil': 'stl4'}
t = pd.concat([
read.temperature(input_path, year_start, year_end, parameter)
for parameter in parameters.values()
],
keys=parameters.keys(),
names=['parameter', 'latitude', 'longitude'],
axis=1)
t = upsample_df(t, '60min')
# Temperature data is filtered by country
return pd.concat([
pd.concat([
t[parameter][population.index]
for population in mapped_population.values()
],
keys=mapped_population.keys(),
axis=1) for parameter in parameters.keys()
],
keys=parameters.keys(),
names=['parameter', 'country', 'latitude', 'longitude'],
axis=1).apply(pd.to_numeric, downcast='float')
def hourly_water(daily_df, temperature, parameters):
# For water heating, the highest temperature classes '30' is chosen
# This is re-sampled to a 60-min-resolution and passed to the general hourly function
classes = upsample_df(
pd.DataFrame(30, index=temperature.index, columns=temperature.columns),
'60min'
).astype(int).astype(str)
return hourly(daily_df, classes, parameters)
def hourly_heat(daily_df, temperature, parameters):
# According to BGW 2006, temperature classes are derived from the temperature data
# This is re-sampled to a 60-min-resolution and passed to the general hourly function
classes = upsample_df(
(np.ceil(((temperature - 273.15) / 5).astype('float64')) * 5).clip(lower=-15, upper=30),
'60min'
).astype(int).astype(str)
return hourly(daily_df, classes, parameters)
def hourly_heat(daily_df, temperature, parameters):
# According to BGW 2006, temperature classes are derived from the temperature data
# This is re-sampled to a 60-min-resolution and passed to the general hourly function
# MP: For each latitude,lognitude in the grid classes contains the
# temperature in one of the 5 degree bands 5,10,15 etc to look up
# the hourly factors to multiply by.
classes = upsample_df(
(np.ceil(((temperature - 273.15) / 5).astype('float64')) * 5).clip(lower=-15, upper=30),
'60min'
).astype(int).astype(str)
return hourly(daily_df, classes, parameters)
def temperature(input_path,
year,
mapped_population,
interim_path,
country='GB',
grid='I',
hour=6):
file = os.path.join(interim_path,
'temperature_' + grid + country + str(year))
if os.path.isfile(file):
temp_data = pd.read_pickle(file)
print(
'temperature preprocessed {} already exists and is read from disk.'
.format(file))
else:
parameters = {'air': 't2m', 'soil': 'stl4'}
t = pd.concat([
read.temperature_era5(input_path, year, hour, grid, parameter)
for parameter in parameters.values()
],
keys=parameters.keys(),
names=['parameter', 'latitude', 'longitude'],
axis=1)
t = upsample_df(t, '60min')
temp_data = pd.concat([
t[parameter][mapped_population.index.tolist()]
for parameter in parameters.keys()
],
keys=parameters.keys(),
names=['parameter', 'latitude', 'longitude'],
axis=1)
# Write results to interim path
temp_data.to_pickle(file)
return temp_data
def temperature(input_path, year_start, year_end, mapped_population):
parameters = {'air': 't2m', 'soil': 'stl4'}
t = pd.concat([
read.temperature(input_path, year_start, year_end, parameter)
for parameter in parameters.values()
],
keys=parameters.keys(),
names=['parameter', 'latitude', 'longitude'],
axis=1)
t = upsample_df(t, '60min')
# Temperature data is filtered by country
return pd.concat([
pd.concat([
t[parameter][population.index]
for population in mapped_population.values()
],
keys=mapped_population.keys(),
axis=1) for parameter in parameters.keys()
],
keys=parameters.keys(),
names=['parameter', 'country', 'latitude', 'longitude'],
axis=1).apply(pd.to_numeric, downcast='float')
#%% Custom temperature
# def upsample_df(df, resolution):
# # The low-resolution values are applied to all high-resolution values up to the next low-resolution value
# # In particular, the last low-resolution value is extended up to where the next low-resolution value would be
# df = df.copy()
# # Determine the original frequency
# freq = df.index[-1] - df.index[-2]
# # Temporally append the DataFrame by one low-resolution value
# df.loc[df.index[-1] + freq, :] = df.iloc[-1, :]
# dtidx = pd.date_range(str(df.index[0]),str(df.index[-1]), freq=freq)
# df.index = dtidx
# # Up-sample
# df = df.resample(resolution).pad()
# # Drop the temporal low-resolution value
# df.drop(df.index[-1], inplace=True)
# return df
# parameters = {
# 'air': 't2m',
# 'soil': 'stl4'
# }
# t = pd.concat(
# [read.temperature(input_path, year_start, year_end, parameter) for parameter in parameters.values()],
# keys=parameters.keys(), names=['parameter', 'latitude', 'longitude'], axis=1
# )
# t = upsample_df(t, '60min')
# # Temperature data is filtered by country
# temperature = pd.concat(
# [pd.concat(
# [t[parameter][population.index] for population in mapped_population.values()],
# keys=mapped_population.keys(), axis=1
# ) for parameter in parameters.keys()],
# keys=parameters.keys(), names=['parameter', 'country', 'latitude', 'longitude'], axis=1
# ).apply(pd.to_numeric, downcast='float')