def test_get_clearsky_valueerror():
tus = Location(32.2, -111, 'US/Arizona', 700, 'Tucson')
times = pd.DatetimeIndex(start='20160101T0600-0700',
end='20160101T1800-0700',
freq='3H')
with pytest.raises(ValueError):
clearsky = tus.get_clearsky(times, model='invalid_model')
def test_get_clearsky(mocker, times):
tus = Location(32.2, -111, 'US/Arizona', 700, 'Tucson')
m = mocker.spy(pvlib.clearsky, 'ineichen')
out = tus.get_clearsky(times)
assert m.call_count == 1
assert_index_equal(out.index, times)
# check that values are 0 before sunrise and after sunset
assert out.iloc[0, :].sum().sum() == 0
assert out.iloc[-1:, :].sum().sum() == 0
# check that values are > 0 during the day
assert (out.iloc[1:-1, :] > 0).all().all()
assert (out.columns.values == ['ghi', 'dni', 'dhi']).all()
def test_get_clearsky_haurwitz(times):
tus = Location(32.2, -111, 'US/Arizona', 700, 'Tucson')
clearsky = tus.get_clearsky(times, model='haurwitz')
expected = pd.DataFrame(data=np.array(
[[ 0. ],
[ 242.30085588],
[ 559.38247117],
[ 384.6873791 ],
[ 0. ]]),
columns=['ghi'],
index=times)
assert_frame_equal(expected, clearsky)
def test_get_clearsky_simplified_solis(times):
tus = Location(32.2, -111, 'US/Arizona', 700, 'Tucson')
clearsky = tus.get_clearsky(times, model='simplified_solis')
expected = pd.DataFrame(data=np.
array([[ 0. , 0. , 0. ],
[ 70.00146271, 638.01145669, 236.71136245],
[ 101.69729217, 852.51950946, 577.1117803 ],
[ 86.1679965 , 755.98048017, 385.59586091],
[ 0. , 0. , 0. ]]),
columns=['dhi', 'dni', 'ghi'],
index=times)
expected = expected[['ghi', 'dni', 'dhi']]
assert_frame_equal(expected, clearsky, check_less_precise=2)
def test_get_clearsky_simplified_solis_dni_extra(times):
tus = Location(32.2, -111, 'US/Arizona', 700, 'Tucson')
clearsky = tus.get_clearsky(times, model='simplified_solis',
dni_extra=1370)
expected = pd.DataFrame(data=np.
array([[ 0. , 0. , 0. ],
[ 67.82281485, 618.15469596, 229.34422063],
[ 98.53217848, 825.98663808, 559.15039353],
[ 83.48619937, 732.45218243, 373.59500313],
[ 0. , 0. , 0. ]]),
columns=['dhi', 'dni', 'ghi'],
index=times)
expected = expected[['ghi', 'dni', 'dhi']]
assert_frame_equal(expected, clearsky)
def test_get_clearsky_simplified_solis_pressure(times):
tus = Location(32.2, -111, 'US/Arizona', 700, 'Tucson')
clearsky = tus.get_clearsky(times, model='simplified_solis',
pressure=95000)
expected = pd.DataFrame(data=np.
array([[ 0. , 0. , 0. ],
[ 70.20556637, 635.53091983, 236.17716435],
[ 102.08954904, 850.49502085, 576.28465815],
[ 86.46561686, 753.70744638, 384.90537859],
[ 0. , 0. , 0. ]]),
columns=['dhi', 'dni', 'ghi'],
index=times)
expected = expected[['ghi', 'dni', 'dhi']]
assert_frame_equal(expected, clearsky, check_less_precise=2)
def test_get_clearsky_simplified_solis_aod_pw(times):
tus = Location(32.2, -111, 'US/Arizona', 700, 'Tucson')
clearsky = tus.get_clearsky(times, model='simplified_solis',
aod700=0.25, precipitable_water=2.)
expected = pd.DataFrame(data=np.
array([[ 0. , 0. , 0. ],
[ 85.77821205, 374.58084365, 179.48483117],
[ 143.52743364, 625.91745295, 490.06254157],
[ 114.63275842, 506.52275195, 312.24711495],
[ 0. , 0. , 0. ]]),
columns=['dhi', 'dni', 'ghi'],
index=times)
expected = expected[['ghi', 'dni', 'dhi']]
assert_frame_equal(expected, clearsky, check_less_precise=2)
def test_get_clearsky():
tus = Location(32.2, -111, 'US/Arizona', 700, 'Tucson')
times = pd.DatetimeIndex(start='20160101T0600-0700',
end='20160101T1800-0700',
freq='3H')
clearsky = tus.get_clearsky(times)
expected = pd.DataFrame(data=np.array([
( 0.0, 0.0, 0.0),
(262.77734276159333, 791.1972825869296, 46.18714900637892),
(616.764693938387, 974.9610353623959, 65.44157429054201),
(419.6512657626518, 901.6234995035793, 54.26016437839348),
( 0.0, 0.0, 0.0)],
dtype=[('ghi', '<f8'), ('dni', '<f8'), ('dhi', '<f8')]), index=times)
assert_frame_equal(expected, clearsky, check_less_precise=2)
def test_get_clearsky_haurwitz():
tus = Location(32.2, -111, 'US/Arizona', 700, 'Tucson')
times = pd.DatetimeIndex(start='20160101T0600-0700',
end='20160101T1800-0700',
freq='3H')
clearsky = tus.get_clearsky(times, model='haurwitz')
expected = pd.DataFrame(data=np.array(
[[ 0. ],
[ 242.30085588],
[ 559.38247117],
[ 384.6873791 ],
[ 0. ]]),
columns=['ghi'],
index=times)
assert_frame_equal(expected, clearsky)
def test_get_clearsky():
tus = Location(32.2, -111, 'US/Arizona', 700, 'Tucson')
times = pd.DatetimeIndex(start='20160101T0600-0700',
end='20160101T1800-0700',
freq='3H')
clearsky = tus.get_clearsky(times)
expected = pd.DataFrame(data=np.
array([[ 0. , 0. , 0. ],
[ 258.60422702, 761.57329257, 50.1235982 ],
[ 611.96347869, 956.95353414, 70.8232806 ],
[ 415.10904044, 878.52649603, 59.07820922],
[ 0. , 0. , 0. ]]),
columns=['ghi', 'dni', 'dhi'],
index=times)
assert_frame_equal(expected, clearsky, check_less_precise=2)
def test_get_clearsky_ineichen_supply_linke(mocker):
tus = Location(32.2, -111, 'US/Arizona', 700)
times = pd.date_range(start='2014-06-24-0700', end='2014-06-25-0700',
freq='3h')
mocker.spy(pvlib.clearsky, 'ineichen')
out = tus.get_clearsky(times, linke_turbidity=3)
# we only care that the LT is passed in this test
pvlib.clearsky.ineichen.assert_called_once_with(ANY, ANY, 3, ANY, ANY)
assert_index_equal(out.index, times)
# check that values are 0 before sunrise and after sunset
assert out.iloc[0:2, :].sum().sum() == 0
assert out.iloc[-2:, :].sum().sum() == 0
# check that values are > 0 during the day
assert (out.iloc[2:-2, :] > 0).all().all()
assert (out.columns.values == ['ghi', 'dni', 'dhi']).all()
def test_get_clearsky_simplified_solis_apparent_elevation(times):
tus = Location(32.2, -111, 'US/Arizona', 700, 'Tucson')
solar_position = {'apparent_elevation': pd.Series(80, index=times),
'apparent_zenith': pd.Series(10, index=times)}
clearsky = tus.get_clearsky(times, model='simplified_solis',
solar_position=solar_position)
expected = pd.DataFrame(data=np.
array([[ 131.3124497 , 1001.14754036, 1108.14147919],
[ 131.3124497 , 1001.14754036, 1108.14147919],
[ 131.3124497 , 1001.14754036, 1108.14147919],
[ 131.3124497 , 1001.14754036, 1108.14147919],
[ 131.3124497 , 1001.14754036, 1108.14147919]]),
columns=['dhi', 'dni', 'ghi'],
index=times)
expected = expected[['ghi', 'dni', 'dhi']]
assert_frame_equal(expected, clearsky, check_less_precise=2)
def test_get_clearsky_ineichen_supply_linke():
tus = Location(32.2, -111, 'US/Arizona', 700)
times = pd.date_range(start='2014-06-24', end='2014-06-25', freq='3h')
times_localized = times.tz_localize(tus.tz)
expected = pd.DataFrame(np.
array([[ 0. , 0. , 0. ],
[ 0. , 0. , 0. ],
[ 79.73090244, 316.16436502, 40.45759009],
[ 703.43653498, 876.41452667, 95.15798252],
[ 1042.37962396, 939.86391062, 118.44687715],
[ 851.32411813, 909.11186737, 105.36662462],
[ 257.18266827, 646.16644264, 62.02777094],
[ 0. , 0. , 0. ],
[ 0. , 0. , 0. ]]),
columns=['ghi', 'dni', 'dhi'],
index=times_localized)
out = tus.get_clearsky(times_localized, linke_turbidity=3)
assert_frame_equal(expected, out, check_less_precise=2)
def detect_clearsky_data():
test_dir = os.path.dirname(os.path.abspath(
inspect.getfile(inspect.currentframe())))
file = os.path.join(test_dir, '..', 'data', 'detect_clearsky_data.csv')
expected = pd.read_csv(file, index_col=0, parse_dates=True, comment='#')
expected = expected.tz_localize('UTC').tz_convert('Etc/GMT+7')
metadata = {}
with open(file) as f:
for line in f:
if line.startswith('#'):
key, value = line.strip('# \n').split(':')
metadata[key] = float(value)
else:
break
metadata['window_length'] = int(metadata['window_length'])
loc = Location(metadata['latitude'], metadata['longitude'],
altitude=metadata['elevation'])
# specify turbidity to guard against future lookup changes
cs = loc.get_clearsky(expected.index, linke_turbidity=2.658197)
return expected, cs
from pvlib import solarposition
from pvlib import irradiance
from pvlib import atmosphere
from conftest import requires_ephem, requires_numba, needs_numpy_1_10
# setup times and location to be tested.
tus = Location(32.2, -111, 'US/Arizona', 700)
# must include night values
times = pd.date_range(start='20140624', freq='6H', periods=4, tz=tus.tz)
ephem_data = solarposition.get_solarposition(
times, tus.latitude, tus.longitude, method='nrel_numpy')
irrad_data = tus.get_clearsky(times, model='ineichen', linke_turbidity=3)
dni_et = irradiance.extraradiation(times.dayofyear)
ghi = irrad_data['ghi']
# setup for et rad test. put it here for readability
timestamp = pd.Timestamp('20161026')
dt_index = pd.DatetimeIndex([timestamp])
doy = timestamp.dayofyear
dt_date = timestamp.date()
dt_datetime = datetime.datetime.combine(dt_date, datetime.time(0))
dt_np64 = np.datetime64(dt_datetime)
value = 1383.636203
class LocalRE(object):
forecast_height = 10 # for DarkSky API
def __init__(
self,
wind_turbines: list = [],
pv_arrays: list = [],
latitude: float = 57.6568,
longitude: float = -3.5818,
altitude: float = 10,
roughness_length: float = 0.15, # roughness length (bit of a guess)
hellman_exp: float = 0.2):
""" Set up the renewable energy generation
"""
# This needs to be repeated in every forecast
self.roughness_length = roughness_length
# Initialise empty forecast dataframe, just so nothing complains
self.wind_forecast = pd.DataFrame()
self.pv_forecast = pd.DataFrame()
# Wind turbine(s)
turbines = []
for turbine in wind_turbines:
turbines.append({
'wind_turbine':
WindTurbine(turbine['name'],
turbine['hub_height'],
nominal_power=turbine['nominal_power'],
rotor_diameter=turbine['rotor_diameter'],
power_curve=turbine['power_curve']),
'number_of_turbines':
turbine['qty']
})
local_wind_farm = WindFarm('Local windfarm', turbines,
[latitude, longitude])
# TODO - check for learned local data & overwrite power_curve
self.wind_modelchain = TurbineClusterModelChain(
local_wind_farm,
smoothing=False,
hellman_exp=hellman_exp,
)
# Initialise PV models
self.pv_location = Location(latitude=latitude,
longitude=longitude,
altitude=altitude)
# Now set up the PV array & system.
cec_pv_model_params = pvlib.pvsystem.retrieve_sam('CECMod')
sandia_pv_model_params = pvlib.pvsystem.retrieve_sam('SandiaMod')
cec_inverter_model_params = pvlib.pvsystem.retrieve_sam('CECInverter')
adr_inverter_model_params = pvlib.pvsystem.retrieve_sam('ADRInverter')
self.pv_modelchains = {}
for pv_array in pv_arrays:
# Try to find the module names in the libraries
if pv_array['module_name'] in cec_pv_model_params:
pv_array['module_parameters'] = cec_pv_model_params[
pv_array['module_name']]
elif pv_array['module_name'] in sandia_pv_model_params:
pv_array['module_parameters'] = sandia_pv_model_params[
pv_array['module_name']]
else:
raise RenewablesException('Could not retrieve PV module data')
# Do the same with the inverter(s)
if pv_array['inverter_name'] in cec_inverter_model_params:
pv_array['inverter_parameters'] = cec_inverter_model_params[
pv_array['inverter_name']]
elif pv_array['inverter_name'] in adr_inverter_model_params:
pv_array['inverter_parameters'] = adr_inverter_model_params[
pv_array['inverter_name']]
else:
raise RenewablesException('Could not retrieve PV module data')
self.pv_modelchains[pv_array['name']] = ModelChain(
PVSystem(**pv_array),
self.pv_location,
aoi_model='physical',
spectral_model='no_loss')
def make_generation_forecasts(self, forecast):
""" Makes generation forecast data from the supplied Dark Sky forecast
Arguments:
forecast {pandas.DataFrame} -- DarkSky originated forecast
"""
self.pv_forecast = self._make_pv_forecast(forecast)
self.wind_forecast = self._make_wind_forecast(forecast)
def _make_pv_forecast(self, forecast) -> pd.DataFrame:
"""Compile the forecast required for PV generation prediction
Uses pvlib to generate solar irradiance predictions.
Arguments:
forecast {pandas.DataFrame} -- DarkSky originated forecast
"""
# Annoyingly, the PV & wind libraries want temperature named differently
pv_forecast = forecast.rename(columns={
'temperature': 'air_temp',
'windSpeed': 'wind_speed',
})
# Use PV lib to get insolation based on the cloud cover reported here
model = GFS()
# Next up, we get hourly solar irradiance using interpolated cloud cover
# We can get this from the clearsky GHI...
if tables in sys.modules:
# We can use Ineichen clear sky model (uses pytables for turbidity)
clearsky = self.pv_location.get_clearsky(pv_forecast.index)
else:
# We can't, so use 'Simplified Solis'
clearsky = self.pv_location.get_clearsky(pv_forecast.index,
model='simplified_solis')
# ... and by knowledge of where the sun is
solpos = self.pv_location.get_solarposition(pv_forecast.index)
ghi = model.cloud_cover_to_ghi_linear(pv_forecast['cloudCover'] * 100,
clearsky['ghi'])
dni = disc(ghi, solpos['zenith'], pv_forecast.index)['dni']
dhi = ghi - dni * np.cos(np.radians(solpos['zenith']))
# Whump it all together and we have our forecast!
pv_forecast['dni'] = dni
pv_forecast['dhi'] = dhi
pv_forecast['ghi'] = ghi
return pv_forecast
def _make_wind_forecast(self, forecast) -> pd.DataFrame:
"""Creates forecast needed for wind generation prediction
Creates renamed multidimensional columns needed for the windpowerlib
system.
Arguments:
forecast {pandas.DataFrame} -- DarkSky originated forecast
"""
# Easiest to build multiindexes up one by one.
columns_index = pd.MultiIndex.from_tuples([('wind_speed', 10),
('temperature', 10),
('pressure', 10),
('roughness_length', 0),
('wind_bearing', 10)])
wind_forecast = pd.DataFrame(index=forecast.index.copy(),
columns=columns_index)
wind_forecast.loc[:, ('wind_speed', 10)] = forecast['windSpeed'].loc[:]
wind_forecast.loc[:,
('temperature', 10)] = forecast['temperature'].loc[:]
wind_forecast.loc[:, ('pressure', 10)] = forecast['pressure'].loc[:]
wind_forecast.loc[:, ('wind_bearing',
10)] = forecast['windBearing'].loc[:]
wind_forecast.loc[:, ('roughness_length', 0)] = self.roughness_length
return wind_forecast
def predict_generation(self, reserved_wind_consumption=0) -> pd.DataFrame:
""" Predict electricity generated from forecast
Will use the timestamp index of the forecast property to estimate
instantaneous electricity generation. Returns table giving amounts in
kWh.
Arguments:
reserved_wind_consumption {float} - constant amount that is assumed
to be required from wind generation to meet other local need
"""
prediction = pd.DataFrame(index=self.pv_forecast.index.copy())
# First up - PV
# Create a total gen column of zeros
prediction['PV_AC_TOTAL'] = 0
for pv_array, pv_model in self.pv_modelchains.items():
pv_model.run_model(prediction.index, self.pv_forecast)
output_column_name = 'PV_AC_' + pv_array
prediction[output_column_name] = pv_model.ac
# Add to the total column
prediction['PV_AC_TOTAL'] = prediction['PV_AC_TOTAL'] + pv_model.ac
# Next - wind power.
self.wind_modelchain.run_model(self.wind_forecast)
prediction['WIND_AC'] = self.wind_modelchain.power_output
# Convert everything into kWh
prediction = prediction * 0.001
prediction['available_wind'] = prediction[
'WIND_AC'] - reserved_wind_consumption
prediction['available_wind'][prediction['available_wind'] < 0] = 0
prediction['total'] = prediction['WIND_AC'] + prediction['PV_AC_TOTAL']
prediction['surplus'] = prediction['available_wind'] + prediction[
'PV_AC_TOTAL']
prediction['surplus'][prediction['surplus'] < 0] = 0
return prediction
base_rad['null_1'] = np.where(base_rad.rad_1.isnull(), 1, 0)
#Quitar los valores que estén fuera de rango
base_rad['range'] = np.where((base_rad.rad_1 >= -1) & (base_rad.rad_1 < 1500), 0, 1)
#La diferencia de los datos no puede exceder y si es un Na tomarlo como un dato
base_rad['diff_0'] = (abs((base_rad[-base_rad.rad_1.isnull()].rad_1) - (base_rad[-base_rad.rad_1.isnull()].rad_1.shift(1))) < 555.)
base_rad['diff_1'] = np.where(base_rad.diff_0 == True, 0, 1)
#Validación de los datos que no se salgan de los valores máximos
tus = Location(float(estaciones[estaciones.cod == int(i[0:8])].LATITUD), float(estaciones[estaciones.cod == int(i[0:8])].LONGITUD), 'America/Bogota')
base_rad['sky_0'] = tus.get_clearsky(pd.DatetimeIndex(base_rad.date, tz='America/Bogota'))['ghi'].reset_index(drop=True) # ineichen with climatology table by default
base_rad['sky_1'] = np.where(((base_rad.sky_0 +10) > base_rad.rad_1), 0, 1)
##Tabla de resumen para extracción de las estadísticas
#if (n_datos - len(base_rad[base_rad.null_1 == 0])) < 0:
n_datos = len(base_rad)
rad_tabla = rad_tabla.append(pd.DataFrame([{'cod':i[0:8], #Código
'isnull':(n_datos - len(base_rad[base_rad.null_1 == 0])),
'total_isnull':n_datos,
'range':base_rad[-base_rad.rad_1.isnull()].range.sum(),
'diff':base_rad[-base_rad.rad_1.isnull()].diff_1.sum(),
class ForecastModel(object):
"""
An object for querying and holding forecast model information for
use within the pvlib library.
Simplifies use of siphon library on a THREDDS server.
Parameters
----------
model_type: string
UNIDATA category in which the model is located.
model_name: string
Name of the UNIDATA forecast model.
set_type: string
Model dataset type.
Attributes
----------
access_url: string
URL specifying the dataset from data will be retrieved.
base_tds_url : string
The top level server address
catalog_url : string
The url path of the catalog to parse.
data: pd.DataFrame
Data returned from the query.
data_format: string
Format of the forecast data being requested from UNIDATA.
dataset: Dataset
Object containing information used to access forecast data.
dataframe_variables: list
Model variables that are present in the data.
datasets_list: list
List of all available datasets.
fm_models: Dataset
TDSCatalog object containing all available
forecast models from UNIDATA.
fm_models_list: list
List of all available forecast models from UNIDATA.
latitude: list
A list of floats containing latitude values.
location: Location
A pvlib Location object containing geographic quantities.
longitude: list
A list of floats containing longitude values.
lbox: boolean
Indicates the use of a location bounding box.
ncss: NCSS object
NCSS
model_name: string
Name of the UNIDATA forecast model.
model: Dataset
A dictionary of Dataset object, whose keys are the name of the
dataset's name.
model_url: string
The url path of the dataset to parse.
modelvariables: list
Common variable names that correspond to queryvariables.
query: NCSS query object
NCSS object used to complete the forecast data retrival.
queryvariables: list
Variables that are used to query the THREDDS Data Server.
time: DatetimeIndex
Time range.
variables: dict
Defines the variables to obtain from the weather
model and how they should be renamed to common variable names.
units: dict
Dictionary containing the units of the standard variables
and the model specific variables.
vert_level: float or integer
Vertical altitude for query data.
"""
access_url_key = 'NetcdfSubset'
catalog_url = 'http://thredds.ucar.edu/thredds/catalog.xml'
base_tds_url = catalog_url.split('/thredds/')[0]
data_format = 'netcdf'
vert_level = 100000
units = {
'temp_air': 'C',
'wind_speed': 'm/s',
'ghi': 'W/m^2',
'ghi_raw': 'W/m^2',
'dni': 'W/m^2',
'dhi': 'W/m^2',
'total_clouds': '%',
'low_clouds': '%',
'mid_clouds': '%',
'high_clouds': '%'}
def __init__(self, model_type, model_name, set_type):
self.model_type = model_type
self.model_name = model_name
self.set_type = set_type
self.catalog = TDSCatalog(self.catalog_url)
self.fm_models = TDSCatalog(self.catalog.catalog_refs[model_type].href)
self.fm_models_list = sorted(list(self.fm_models.catalog_refs.keys()))
try:
model_url = self.fm_models.catalog_refs[model_name].href
except ParseError:
raise ParseError(self.model_name + ' model may be unavailable.')
try:
self.model = TDSCatalog(model_url)
except HTTPError:
try:
self.model = TDSCatalog(model_url)
except HTTPError:
raise HTTPError(self.model_name + ' model may be unavailable.')
self.datasets_list = list(self.model.datasets.keys())
self.set_dataset()
def __repr__(self):
return '{}, {}'.format(self.model_name, self.set_type)
def set_dataset(self):
'''
Retrieves the designated dataset, creates NCSS object, and
creates a NCSS query object.
'''
keys = list(self.model.datasets.keys())
labels = [item.split()[0].lower() for item in keys]
if self.set_type == 'best':
self.dataset = self.model.datasets[keys[labels.index('best')]]
elif self.set_type == 'latest':
self.dataset = self.model.datasets[keys[labels.index('latest')]]
elif self.set_type == 'full':
self.dataset = self.model.datasets[keys[labels.index('full')]]
self.access_url = self.dataset.access_urls[self.access_url_key]
self.ncss = NCSS(self.access_url)
self.query = self.ncss.query()
def set_query_latlon(self):
'''
Sets the NCSS query location latitude and longitude.
'''
if (isinstance(self.longitude, list) and
isinstance(self.latitude, list)):
self.lbox = True
# west, east, south, north
self.query.lonlat_box(self.latitude[0], self.latitude[1],
self.longitude[0], self.longitude[1])
else:
self.lbox = False
self.query.lonlat_point(self.longitude, self.latitude)
def set_location(self, time, latitude, longitude):
'''
Sets the location for the query.
Parameters
----------
time: datetime or DatetimeIndex
Time range of the query.
'''
if isinstance(time, datetime.datetime):
tzinfo = time.tzinfo
else:
tzinfo = time.tz
if tzinfo is None:
self.location = Location(latitude, longitude)
else:
self.location = Location(latitude, longitude, tz=tzinfo)
def get_data(self, latitude, longitude, start, end,
vert_level=None, query_variables=None,
close_netcdf_data=True):
"""
Submits a query to the UNIDATA servers using Siphon NCSS and
converts the netcdf data to a pandas DataFrame.
Parameters
----------
latitude: float
The latitude value.
longitude: float
The longitude value.
start: datetime or timestamp
The start time.
end: datetime or timestamp
The end time.
vert_level: None, float or integer
Vertical altitude of interest.
variables: None or list
If None, uses self.variables.
close_netcdf_data: bool
Controls if the temporary netcdf data file should be closed.
Set to False to access the raw data.
Returns
-------
forecast_data : DataFrame
column names are the weather model's variable names.
"""
if vert_level is not None:
self.vert_level = vert_level
if query_variables is None:
self.query_variables = list(self.variables.values())
else:
self.query_variables = query_variables
self.latitude = latitude
self.longitude = longitude
self.set_query_latlon() # modifies self.query
self.set_location(start, latitude, longitude)
self.start = start
self.end = end
self.query.time_range(self.start, self.end)
self.query.vertical_level(self.vert_level)
self.query.variables(*self.query_variables)
self.query.accept(self.data_format)
self.netcdf_data = self.ncss.get_data(self.query)
# might be better to go to xarray here so that we can handle
# higher dimensional data for more advanced applications
self.data = self._netcdf2pandas(self.netcdf_data, self.query_variables)
if close_netcdf_data:
self.netcdf_data.close()
return self.data
def process_data(self, data, **kwargs):
"""
Defines the steps needed to convert raw forecast data
into processed forecast data. Most forecast models implement
their own version of this method which also call this one.
Parameters
----------
data: DataFrame
Raw forecast data
Returns
-------
data: DataFrame
Processed forecast data.
"""
data = self.rename(data)
return data
def get_processed_data(self, *args, **kwargs):
"""
Get and process forecast data.
Parameters
----------
*args: positional arguments
Passed to get_data
**kwargs: keyword arguments
Passed to get_data and process_data
Returns
-------
data: DataFrame
Processed forecast data
"""
return self.process_data(self.get_data(*args, **kwargs), **kwargs)
def rename(self, data, variables=None):
"""
Renames the columns according the variable mapping.
Parameters
----------
data: DataFrame
variables: None or dict
If None, uses self.variables
Returns
-------
data: DataFrame
Renamed data.
"""
if variables is None:
variables = self.variables
return data.rename(columns={y: x for x, y in variables.items()})
def _netcdf2pandas(self, netcdf_data, query_variables):
"""
Transforms data from netcdf to pandas DataFrame.
Parameters
----------
data: netcdf
Data returned from UNIDATA NCSS query.
query_variables: list
The variables requested.
Returns
-------
pd.DataFrame
"""
# set self.time
try:
time_var = 'time'
self.set_time(netcdf_data.variables[time_var])
except KeyError:
# which model does this dumb thing?
time_var = 'time1'
self.set_time(netcdf_data.variables[time_var])
data_dict = {key: data[:].squeeze() for key, data in
netcdf_data.variables.items() if key in query_variables}
return pd.DataFrame(data_dict, index=self.time)
def set_time(self, time):
'''
Converts time data into a pandas date object.
Parameters
----------
time: netcdf
Contains time information.
Returns
-------
pandas.DatetimeIndex
'''
times = num2date(time[:].squeeze(), time.units)
self.time = pd.DatetimeIndex(pd.Series(times), tz=self.location.tz)
def cloud_cover_to_ghi_linear(self, cloud_cover, ghi_clear, offset=35,
**kwargs):
"""
Convert cloud cover to GHI using a linear relationship.
0% cloud cover returns ghi_clear.
100% cloud cover returns offset*ghi_clear.
Parameters
----------
cloud_cover: numeric
Cloud cover in %.
ghi_clear: numeric
GHI under clear sky conditions.
offset: numeric
Determines the minimum GHI.
kwargs
Not used.
Returns
-------
ghi: numeric
Estimated GHI.
References
----------
Larson et. al. "Day-ahead forecasting of solar power output from
photovoltaic plants in the American Southwest" Renewable Energy
91, 11-20 (2016).
"""
offset = offset / 100.
cloud_cover = cloud_cover / 100.
ghi = (offset + (1 - offset) * (1 - cloud_cover)) * ghi_clear
return ghi
def cloud_cover_to_irradiance_clearsky_scaling(self, cloud_cover,
method='linear',
**kwargs):
"""
Estimates irradiance from cloud cover in the following steps:
1. Determine clear sky GHI using Ineichen model and
climatological turbidity.
2. Estimate cloudy sky GHI using a function of
cloud_cover e.g.
:py:meth:`~ForecastModel.cloud_cover_to_ghi_linear`
3. Estimate cloudy sky DNI using the DISC model.
4. Calculate DHI from DNI and DHI.
Parameters
----------
cloud_cover : Series
Cloud cover in %.
method : str
Method for converting cloud cover to GHI.
'linear' is currently the only option.
**kwargs
Passed to the method that does the conversion
Returns
-------
irrads : DataFrame
Estimated GHI, DNI, and DHI.
"""
solpos = self.location.get_solarposition(cloud_cover.index)
cs = self.location.get_clearsky(cloud_cover.index, model='ineichen',
solar_position=solpos)
method = method.lower()
if method == 'linear':
ghi = self.cloud_cover_to_ghi_linear(cloud_cover, cs['ghi'],
**kwargs)
else:
raise ValueError('invalid method argument')
dni = disc(ghi, solpos['zenith'], cloud_cover.index)['dni']
dhi = ghi - dni * np.cos(np.radians(solpos['zenith']))
irrads = pd.DataFrame({'ghi': ghi, 'dni': dni, 'dhi': dhi}).fillna(0)
return irrads
def cloud_cover_to_transmittance_linear(self, cloud_cover, offset=0.75,
**kwargs):
"""
Convert cloud cover to atmospheric transmittance using a linear
model.
0% cloud cover returns offset.
100% cloud cover returns 0.
Parameters
----------
cloud_cover : numeric
Cloud cover in %.
offset : numeric
Determines the maximum transmittance.
kwargs
Not used.
Returns
-------
ghi : numeric
Estimated GHI.
"""
transmittance = ((100.0 - cloud_cover) / 100.0) * 0.75
return transmittance
def cloud_cover_to_irradiance_liujordan(self, cloud_cover, **kwargs):
"""
Estimates irradiance from cloud cover in the following steps:
1. Determine transmittance using a function of cloud cover e.g.
:py:meth:`~ForecastModel.cloud_cover_to_transmittance_linear`
2. Calculate GHI, DNI, DHI using the
:py:func:`pvlib.irradiance.liujordan` model
Parameters
----------
cloud_cover : Series
Returns
-------
irradiance : DataFrame
Columns include ghi, dni, dhi
"""
# in principle, get_solarposition could use the forecast
# pressure, temp, etc., but the cloud cover forecast is not
# accurate enough to justify using these minor corrections
solar_position = self.location.get_solarposition(cloud_cover.index)
dni_extra = extraradiation(cloud_cover.index)
airmass = self.location.get_airmass(cloud_cover.index)
transmittance = self.cloud_cover_to_transmittance_linear(cloud_cover,
**kwargs)
irrads = liujordan(solar_position['apparent_zenith'],
transmittance, airmass['airmass_absolute'],
dni_extra=dni_extra)
irrads = irrads.fillna(0)
return irrads
def cloud_cover_to_irradiance(self, cloud_cover, how='clearsky_scaling',
**kwargs):
"""
Convert cloud cover to irradiance. A wrapper method.
Parameters
----------
cloud_cover : Series
how : str
Selects the method for conversion. Can be one of
clearsky_scaling or liujordan.
**kwargs
Passed to the selected method.
Returns
-------
irradiance : DataFrame
Columns include ghi, dni, dhi
"""
how = how.lower()
if how == 'clearsky_scaling':
irrads = self.cloud_cover_to_irradiance_clearsky_scaling(
cloud_cover, **kwargs)
elif how == 'liujordan':
irrads = self.cloud_cover_to_irradiance_liujordan(
cloud_cover, **kwargs)
else:
raise ValueError('invalid how argument')
return irrads
def kelvin_to_celsius(self, temperature):
"""
Converts Kelvin to celsius.
Parameters
----------
temperature: numeric
Returns
-------
temperature: numeric
"""
return temperature - 273.15
def isobaric_to_ambient_temperature(self, data):
"""
Calculates temperature from isobaric temperature.
Parameters
----------
data: DataFrame
Must contain columns pressure, temperature_iso,
temperature_dew_iso. Input temperature in K.
Returns
-------
temperature : Series
Temperature in K
"""
P = data['pressure'] / 100.0
Tiso = data['temperature_iso']
Td = data['temperature_dew_iso'] - 273.15
# saturation water vapor pressure
e = 6.11 * 10**((7.5 * Td) / (Td + 273.3))
# saturation water vapor mixing ratio
w = 0.622 * (e / (P - e))
T = Tiso - ((2.501 * 10.**6) / 1005.7) * w
return T
def uv_to_speed(self, data):
"""
Computes wind speed from wind components.
Parameters
----------
data : DataFrame
Must contain the columns 'wind_speed_u' and 'wind_speed_v'.
Returns
-------
wind_speed : Series
"""
wind_speed = np.sqrt(data['wind_speed_u']**2 + data['wind_speed_v']**2)
return wind_speed
def gust_to_speed(self, data, scaling=1/1.4):
"""
Computes standard wind speed from gust.
Very approximate and location dependent.
Parameters
----------
data : DataFrame
Must contain the column 'wind_speed_gust'.
Returns
-------
wind_speed : Series
"""
wind_speed = data['wind_speed_gust'] * scaling
return wind_speed
def test_get_clearsky_valueerror(times):
tus = Location(32.2, -111, 'US/Arizona', 700, 'Tucson')
with pytest.raises(ValueError):
clearsky = tus.get_clearsky(times, model='invalid_model')
def find_clearsky_poa(df,
lat,
lon,
irradiance_poa_key='irradiance_poa_o_###',
mounting='fixed',
tilt=0,
azimuth=180,
altitude=0):
loc = Location(lat, lon, altitude=altitude)
CS = loc.get_clearsky(df.index)
df['csghi'] = CS.ghi
df['csdhi'] = CS.dhi
df['csdni'] = CS.dni
if mounting.lower() == "fixed":
sun = get_solarposition(df.index, lat, lon)
fixedpoa = get_total_irradiance(tilt, azimuth, sun.zenith, sun.azimuth,
CS.dni, CS.ghi, CS.dhi)
df['cspoa'] = fixedpoa.poa_global
if mounting.lower() == "tracking":
sun = get_solarposition(df.index, lat, lon)
# default to axis_tilt=0 and axis_azimuth=180
tracker_data = singleaxis(sun.apparent_zenith,
sun.azimuth,
axis_tilt=tilt,
axis_azimuth=azimuth,
max_angle=50,
backtrack=True,
gcr=0.35)
track = get_total_irradiance(tracker_data['surface_tilt'],
tracker_data['surface_azimuth'],
sun.zenith, sun.azimuth, CS.dni, CS.ghi,
CS.dhi)
df['cspoa'] = track.poa_global
# the following code is assuming clear sky poa has been generated per pvlib, aligned in the same
# datetime index, and daylight savings or any time shifts were previously corrected
# the inputs below were tuned for POA at a 15 minute frequency
# note that detect_clearsky has a scaling factor but I still got slightly different results when I scaled measured poa first
df['poa'] = df[irradiance_poa_key] / df[irradiance_poa_key].quantile(
0.98) * df.cspoa.quantile(0.98)
# inputs for detect_clearsky
measured = df.poa.copy()
clear = df.cspoa.copy()
dur = 60
lower_line_length = -41.416
upper_line_length = 77.789
var_diff = .00745
mean_diff = 80
max_diff = 90
slope_dev = 3
is_clear_results = detect_clearsky(measured.values,
clear.values,
df.index,
dur,
mean_diff,
max_diff,
lower_line_length,
upper_line_length,
var_diff,
slope_dev,
return_components=True)
clearSeries = pd.Series(index=df.index, data=is_clear_results[0])
clearSeries = clearSeries.reindex(index=df.index, method='ffill', limit=3)
return clearSeries
# TEST year 2015
times = pd.DatetimeIndex(start='2015-01-01', end='2016-01-01', freq='1min',tz=bvl.tz) # 12 months
# TEST year 2016
#times = pd.DatetimeIndex(start='2016-01-01', end='2017-01-01', freq='1min',tz=bvl.tz) # 12 months
# Test year 2017
#times = pd.DatetimeIndex(start='2017-01-01', end='2018-01-01', freq='1min',tz=bvl.tz) # 12 months
# In[512]:
if run_train:
# TRAIN set
times2010and2011 = pd.DatetimeIndex(start='2010-01-01', end='2012-01-01', freq='1min',
tz=bvl.tz) # 24 months of 2010 and 2011 - For training
cs_2010and2011 = bvl.get_clearsky(times2010and2011) # ineichen with climatology table by default
cs_2010and2011.drop(['dni','dhi'],axis=1, inplace=True) #updating the same dataframe by dropping two columns
cs_2010and2011.reset_index(inplace=True)
cs_2010and2011['index']=cs_2010and2011['index'].apply(lambda x:x.to_datetime())
cs_2010and2011['year'] = cs_2010and2011['index'].apply(lambda x:x.year)
cs_2010and2011['month'] = cs_2010and2011['index'].apply(lambda x:x.month)
cs_2010and2011['day'] = cs_2010and2011['index'].apply(lambda x:x.day)
cs_2010and2011['hour'] = cs_2010and2011['index'].apply(lambda x:x.hour)
cs_2010and2011['min'] = cs_2010and2011['index'].apply(lambda x:x.minute)
cs_2010and2011.drop(cs_2010and2011.index[-1], inplace=True)
print(cs_2010and2011.shape)
cs_2010and2011.head()
from conftest import (requires_ephem, requires_numba, needs_numpy_1_10,
pandas_0_22)
# setup times and location to be tested.
tus = Location(32.2, -111, 'US/Arizona', 700)
# must include night values
times = pd.date_range(start='20140624', freq='6H', periods=4, tz=tus.tz)
ephem_data = solarposition.get_solarposition(times,
tus.latitude,
tus.longitude,
method='nrel_numpy')
irrad_data = tus.get_clearsky(times, model='ineichen', linke_turbidity=3)
dni_et = irradiance.extraradiation(times.dayofyear)
ghi = irrad_data['ghi']
# setup for et rad test. put it here for readability
timestamp = pd.Timestamp('20161026')
dt_index = pd.DatetimeIndex([timestamp])
doy = timestamp.dayofyear
dt_date = timestamp.date()
dt_datetime = datetime.datetime.combine(dt_date, datetime.time(0))
dt_np64 = np.datetime64(dt_datetime)
value = 1383.636203
def get_pvlib_data(latitude, longitude, tz, altitude, city, start_time,
end_time):
# getting turbidity tables
pvlib_path = os.path.dirname(os.path.abspath(pvlib.clearsky.__file__))
filepath = os.path.join(pvlib_path, 'data', 'LinkeTurbidities.h5')
def plot_turbidity_map(month, vmin=1, vmax=100):
plt.figure()
with tables.open_file(filepath) as lt_h5_file:
ltdata = lt_h5_file.root.LinkeTurbidity[:, :, month - 1]
plt.imshow(ltdata, vmin=vmin, vmax=vmax)
# data is in units of 20 x turbidity
plt.title('Linke turbidity x 20, ' + calendar.month_name[month])
plt.colorbar(shrink=0.5)
plt.tight_layout()
plot_turbidity_map(1)
plot_turbidity_map(7)
# getting clearsky estimates
loc = Location(latitude, longitude, tz, altitude, city)
times = pd.date_range(start=start_time, end=end_time, freq='H', tz=loc.tz)
cs = loc.get_clearsky(times)
# getting pvlib forecasted irradiance based on cloud_cover
#irrad_vars = ['ghi', 'dni', 'dhi']
model = GFS()
raw_data = model.get_data(latitude, longitude, start_time, end_time)
data = raw_data
# rename the columns according the key/value pairs in model.variables.
data = model.rename(data)
# convert temperature
data['temp_air'] = model.kelvin_to_celsius(data['temp_air'])
# convert wind components to wind speed
data['wind_speed'] = model.uv_to_speed(data)
# calculate irradiance estimates from cloud cover.
# uses a cloud_cover to ghi to dni model or a
# uses a cloud cover to transmittance to irradiance model.
irrad_data = model.cloud_cover_to_irradiance(data['total_clouds'])
# correcting timezone
data.index = data.index.tz_convert(loc.tz)
irrad_data.index = irrad_data.index.tz_convert(loc.tz)
# joining cloud_cover and irradiance data frames
data = data.join(irrad_data, how='outer')
# renaming irradiance estimates
cs.rename(columns={
'ghi': 'GHI_clearsky',
'dhi': 'DHI_clearsky',
'dni': 'DNI_clearsky'
},
inplace=True)
data.rename(columns={
'ghi': 'GHI_pvlib',
'dhi': 'DHI_pvlib',
'dni': 'DNI_pvlib'
},
inplace=True)
# joining clearsky with cloud_cover irradiances
data = data.join(cs, how='outer')
return (data)
def test_get_clearsky_valueerror(times):
tus = Location(32.2, -111, 'US/Arizona', 700, 'Tucson')
with pytest.raises(ValueError):
tus.get_clearsky(times, model='invalid_model')
class ForecastModel(object):
"""
An object for querying and holding forecast model information for
use within the pvlib library.
Simplifies use of siphon library on a THREDDS server.
Parameters
----------
model_type: string
UNIDATA category in which the model is located.
model_name: string
Name of the UNIDATA forecast model.
set_type: string
Model dataset type.
Attributes
----------
access_url: string
URL specifying the dataset from data will be retrieved.
base_tds_url : string
The top level server address
catalog_url : string
The url path of the catalog to parse.
data: pd.DataFrame
Data returned from the query.
data_format: string
Format of the forecast data being requested from UNIDATA.
dataset: Dataset
Object containing information used to access forecast data.
dataframe_variables: list
Model variables that are present in the data.
datasets_list: list
List of all available datasets.
fm_models: Dataset
TDSCatalog object containing all available
forecast models from UNIDATA.
fm_models_list: list
List of all available forecast models from UNIDATA.
latitude: list
A list of floats containing latitude values.
location: Location
A pvlib Location object containing geographic quantities.
longitude: list
A list of floats containing longitude values.
lbox: boolean
Indicates the use of a location bounding box.
ncss: NCSS object
NCSS
model_name: string
Name of the UNIDATA forecast model.
model: Dataset
A dictionary of Dataset object, whose keys are the name of the
dataset's name.
model_url: string
The url path of the dataset to parse.
modelvariables: list
Common variable names that correspond to queryvariables.
query: NCSS query object
NCSS object used to complete the forecast data retrival.
queryvariables: list
Variables that are used to query the THREDDS Data Server.
time: DatetimeIndex
Time range.
variables: dict
Defines the variables to obtain from the weather
model and how they should be renamed to common variable names.
units: dict
Dictionary containing the units of the standard variables
and the model specific variables.
vert_level: float or integer
Vertical altitude for query data.
"""
access_url_key = 'NetcdfSubset'
catalog_url = 'https://thredds.ucar.edu/thredds/catalog.xml'
base_tds_url = catalog_url.split('/thredds/')[0]
data_format = 'netcdf'
units = {
'temp_air': 'C',
'wind_speed': 'm/s',
'ghi': 'W/m^2',
'ghi_raw': 'W/m^2',
'dni': 'W/m^2',
'dhi': 'W/m^2',
'total_clouds': '%',
'low_clouds': '%',
'mid_clouds': '%',
'high_clouds': '%'
}
def __init__(self, model_type, model_name, set_type, vert_level=None):
self.model_type = model_type
self.model_name = model_name
self.set_type = set_type
self.connected = False
self.vert_level = vert_level
def connect_to_catalog(self):
self.catalog = TDSCatalog(self.catalog_url)
self.fm_models = TDSCatalog(
self.catalog.catalog_refs[self.model_type].href)
self.fm_models_list = sorted(list(self.fm_models.catalog_refs.keys()))
try:
model_url = self.fm_models.catalog_refs[self.model_name].href
except ParseError:
raise ParseError(self.model_name + ' model may be unavailable.')
try:
self.model = TDSCatalog(model_url)
except HTTPError:
try:
self.model = TDSCatalog(model_url)
except HTTPError:
raise HTTPError(self.model_name + ' model may be unavailable.')
self.datasets_list = list(self.model.datasets.keys())
self.set_dataset()
self.connected = True
def __repr__(self):
return '{}, {}'.format(self.model_name, self.set_type)
def set_dataset(self):
'''
Retrieves the designated dataset, creates NCSS object, and
creates a NCSS query object.
'''
keys = list(self.model.datasets.keys())
labels = [item.split()[0].lower() for item in keys]
if self.set_type == 'best':
self.dataset = self.model.datasets[keys[labels.index('best')]]
elif self.set_type == 'latest':
self.dataset = self.model.datasets[keys[labels.index('latest')]]
elif self.set_type == 'full':
self.dataset = self.model.datasets[keys[labels.index('full')]]
self.access_url = self.dataset.access_urls[self.access_url_key]
self.ncss = NCSS(self.access_url)
self.query = self.ncss.query()
def set_query_time_range(self, start, end):
"""
Parameters
----------
start : datetime.datetime, pandas.Timestamp
Must be tz-localized.
end : datetime.datetime, pandas.Timestamp
Must be tz-localized.
Notes
-----
Assigns ``self.start``, ``self.end``. Modifies ``self.query``
"""
self.start = pd.Timestamp(start)
self.end = pd.Timestamp(end)
if self.start.tz is None or self.end.tz is None:
raise TypeError('start and end must be tz-localized')
self.query.time_range(self.start, self.end)
def set_query_latlon(self):
'''
Sets the NCSS query location latitude and longitude.
'''
if (isinstance(self.longitude, list)
and isinstance(self.latitude, list)):
self.lbox = True
# west, east, south, north
self.query.lonlat_box(self.longitude[0], self.longitude[1],
self.latitude[0], self.latitude[1])
else:
self.lbox = False
self.query.lonlat_point(self.longitude, self.latitude)
def set_location(self, tz, latitude, longitude):
'''
Sets the location for the query.
Parameters
----------
tz: tzinfo
Timezone of the query
latitude: float
Latitude of the query
longitude: float
Longitude of the query
Notes
-----
Assigns ``self.location``.
'''
self.location = Location(latitude, longitude, tz=tz)
def get_data(self,
latitude,
longitude,
start,
end,
vert_level=None,
query_variables=None,
close_netcdf_data=True,
**kwargs):
"""
Submits a query to the UNIDATA servers using Siphon NCSS and
converts the netcdf data to a pandas DataFrame.
Parameters
----------
latitude: float
The latitude value.
longitude: float
The longitude value.
start: datetime or timestamp
The start time.
end: datetime or timestamp
The end time.
vert_level: None, float or integer, default None
Vertical altitude of interest.
query_variables: None or list, default None
If None, uses self.variables.
close_netcdf_data: bool, default True
Controls if the temporary netcdf data file should be closed.
Set to False to access the raw data.
**kwargs:
Additional keyword arguments are silently ignored.
Returns
-------
forecast_data : DataFrame
column names are the weather model's variable names.
"""
if not self.connected:
self.connect_to_catalog()
if vert_level is not None:
self.vert_level = vert_level
if query_variables is None:
self.query_variables = list(self.variables.values())
else:
self.query_variables = query_variables
self.set_query_time_range(start, end)
self.latitude = latitude
self.longitude = longitude
self.set_query_latlon() # modifies self.query
self.set_location(self.start.tz, latitude, longitude)
if self.vert_level is not None:
self.query.vertical_level(self.vert_level)
self.query.variables(*self.query_variables)
self.query.accept(self.data_format)
self.netcdf_data = self.ncss.get_data(self.query)
# might be better to go to xarray here so that we can handle
# higher dimensional data for more advanced applications
self.data = self._netcdf2pandas(self.netcdf_data, self.query_variables,
self.start, self.end)
if close_netcdf_data:
self.netcdf_data.close()
return self.data
def process_data(self, data, **kwargs):
"""
Defines the steps needed to convert raw forecast data
into processed forecast data. Most forecast models implement
their own version of this method which also call this one.
Parameters
----------
data: DataFrame
Raw forecast data
Returns
-------
data: DataFrame
Processed forecast data.
"""
data = self.rename(data)
return data
def get_processed_data(self, *args, **kwargs):
"""
Get and process forecast data.
Parameters
----------
*args: positional arguments
Passed to get_data
**kwargs: keyword arguments
Passed to get_data and process_data
Returns
-------
data: DataFrame
Processed forecast data
"""
return self.process_data(self.get_data(*args, **kwargs), **kwargs)
def rename(self, data, variables=None):
"""
Renames the columns according the variable mapping.
Parameters
----------
data: DataFrame
variables: None or dict, default None
If None, uses self.variables
Returns
-------
data: DataFrame
Renamed data.
"""
if variables is None:
variables = self.variables
return data.rename(columns={y: x for x, y in variables.items()})
def _netcdf2pandas(self, netcdf_data, query_variables, start, end):
"""
Transforms data from netcdf to pandas DataFrame.
Parameters
----------
data: netcdf
Data returned from UNIDATA NCSS query.
query_variables: list
The variables requested.
start: Timestamp
The start time
end: Timestamp
The end time
Returns
-------
pd.DataFrame
"""
# set self.time
try:
time_var = 'time'
self.set_time(netcdf_data.variables[time_var])
except KeyError:
# which model does this dumb thing?
time_var = 'time1'
self.set_time(netcdf_data.variables[time_var])
data_dict = {}
for key, data in netcdf_data.variables.items():
# if accounts for possibility of extra variable returned
if key not in query_variables:
continue
squeezed = data[:].squeeze()
# If the data is big endian, swap the byte order to make it
# little endian
if squeezed.dtype.byteorder == '>':
squeezed = squeezed.byteswap().newbyteorder()
if squeezed.ndim == 1:
data_dict[key] = squeezed
elif squeezed.ndim == 2:
for num, data_level in enumerate(squeezed.T):
data_dict[key + '_' + str(num)] = data_level
else:
raise ValueError('cannot parse ndim > 2')
data = pd.DataFrame(data_dict, index=self.time)
# sometimes data is returned as hours since T0
# where T0 is before start. Then the hours between
# T0 and start are added *after* end. So sort and slice
# to remove the garbage
data = data.sort_index().loc[start:end]
return data
def set_time(self, time):
'''
Converts time data into a pandas date object.
Parameters
----------
time: netcdf
Contains time information.
Returns
-------
pandas.DatetimeIndex
'''
times = num2date(time[:].squeeze(),
time.units,
only_use_cftime_datetimes=False,
only_use_python_datetimes=True)
self.time = pd.DatetimeIndex(pd.Series(times), tz=self.location.tz)
def cloud_cover_to_ghi_linear(self,
cloud_cover,
ghi_clear,
offset=35,
**kwargs):
"""
Convert cloud cover to GHI using a linear relationship.
0% cloud cover returns ghi_clear.
100% cloud cover returns offset*ghi_clear.
Parameters
----------
cloud_cover: numeric
Cloud cover in %.
ghi_clear: numeric
GHI under clear sky conditions.
offset: numeric, default 35
Determines the minimum GHI.
kwargs
Not used.
Returns
-------
ghi: numeric
Estimated GHI.
References
----------
Larson et. al. "Day-ahead forecasting of solar power output from
photovoltaic plants in the American Southwest" Renewable Energy
91, 11-20 (2016).
"""
offset = offset / 100.
cloud_cover = cloud_cover / 100.
ghi = (offset + (1 - offset) * (1 - cloud_cover)) * ghi_clear
return ghi
def cloud_cover_to_irradiance_clearsky_scaling(self,
cloud_cover,
method='linear',
**kwargs):
"""
Estimates irradiance from cloud cover in the following steps:
1. Determine clear sky GHI using Ineichen model and
climatological turbidity.
2. Estimate cloudy sky GHI using a function of
cloud_cover e.g.
:py:meth:`~ForecastModel.cloud_cover_to_ghi_linear`
3. Estimate cloudy sky DNI using the DISC model.
4. Calculate DHI from DNI and GHI.
Parameters
----------
cloud_cover : Series
Cloud cover in %.
method : str, default 'linear'
Method for converting cloud cover to GHI.
'linear' is currently the only option.
**kwargs
Passed to the method that does the conversion
Returns
-------
irrads : DataFrame
Estimated GHI, DNI, and DHI.
"""
solpos = self.location.get_solarposition(cloud_cover.index)
cs = self.location.get_clearsky(cloud_cover.index,
model='ineichen',
solar_position=solpos)
method = method.lower()
if method == 'linear':
ghi = self.cloud_cover_to_ghi_linear(cloud_cover, cs['ghi'],
**kwargs)
else:
raise ValueError('invalid method argument')
dni = disc(ghi, solpos['zenith'], cloud_cover.index)['dni']
dhi = ghi - dni * np.cos(np.radians(solpos['zenith']))
irrads = pd.DataFrame({'ghi': ghi, 'dni': dni, 'dhi': dhi}).fillna(0)
return irrads
def cloud_cover_to_transmittance_linear(self,
cloud_cover,
offset=0.75,
**kwargs):
"""
Convert cloud cover to atmospheric transmittance using a linear
model.
0% cloud cover returns offset.
100% cloud cover returns 0.
Parameters
----------
cloud_cover : numeric
Cloud cover in %.
offset : numeric, default 0.75
Determines the maximum transmittance.
kwargs
Not used.
Returns
-------
ghi : numeric
Estimated GHI.
"""
transmittance = ((100.0 - cloud_cover) / 100.0) * offset
return transmittance
def cloud_cover_to_irradiance_liujordan(self, cloud_cover, **kwargs):
"""
Estimates irradiance from cloud cover in the following steps:
1. Determine transmittance using a function of cloud cover e.g.
:py:meth:`~ForecastModel.cloud_cover_to_transmittance_linear`
2. Calculate GHI, DNI, DHI using the
:py:func:`pvlib.irradiance.liujordan` model
Parameters
----------
cloud_cover : Series
Returns
-------
irradiance : DataFrame
Columns include ghi, dni, dhi
"""
# in principle, get_solarposition could use the forecast
# pressure, temp, etc., but the cloud cover forecast is not
# accurate enough to justify using these minor corrections
solar_position = self.location.get_solarposition(cloud_cover.index)
dni_extra = get_extra_radiation(cloud_cover.index)
airmass = self.location.get_airmass(cloud_cover.index)
transmittance = self.cloud_cover_to_transmittance_linear(
cloud_cover, **kwargs)
irrads = liujordan(solar_position['apparent_zenith'],
transmittance,
airmass['airmass_absolute'],
dni_extra=dni_extra)
irrads = irrads.fillna(0)
return irrads
def cloud_cover_to_irradiance(self,
cloud_cover,
how='clearsky_scaling',
**kwargs):
"""
Convert cloud cover to irradiance. A wrapper method.
Parameters
----------
cloud_cover : Series
how : str, default 'clearsky_scaling'
Selects the method for conversion. Can be one of
clearsky_scaling or liujordan.
**kwargs
Passed to the selected method.
Returns
-------
irradiance : DataFrame
Columns include ghi, dni, dhi
"""
how = how.lower()
if how == 'clearsky_scaling':
irrads = self.cloud_cover_to_irradiance_clearsky_scaling(
cloud_cover, **kwargs)
elif how == 'liujordan':
irrads = self.cloud_cover_to_irradiance_liujordan(
cloud_cover, **kwargs)
else:
raise ValueError('invalid how argument')
return irrads
def kelvin_to_celsius(self, temperature):
"""
Converts Kelvin to celsius.
Parameters
----------
temperature: numeric
Returns
-------
temperature: numeric
"""
return temperature - 273.15
def isobaric_to_ambient_temperature(self, data):
"""
Calculates temperature from isobaric temperature.
Parameters
----------
data: DataFrame
Must contain columns pressure, temperature_iso,
temperature_dew_iso. Input temperature in K.
Returns
-------
temperature : Series
Temperature in K
"""
P = data['pressure'] / 100.0 # noqa: N806
Tiso = data['temperature_iso'] # noqa: N806
Td = data['temperature_dew_iso'] - 273.15 # noqa: N806
# saturation water vapor pressure
e = 6.11 * 10**((7.5 * Td) / (Td + 273.3))
# saturation water vapor mixing ratio
w = 0.622 * (e / (P - e))
temperature = Tiso - ((2.501 * 10.**6) / 1005.7) * w
return temperature
def uv_to_speed(self, data):
"""
Computes wind speed from wind components.
Parameters
----------
data : DataFrame
Must contain the columns 'wind_speed_u' and 'wind_speed_v'.
Returns
-------
wind_speed : Series
"""
wind_speed = np.sqrt(data['wind_speed_u']**2 + data['wind_speed_v']**2)
return wind_speed
def gust_to_speed(self, data, scaling=1 / 1.4):
"""
Computes standard wind speed from gust.
Very approximate and location dependent.
Parameters
----------
data : DataFrame
Must contain the column 'wind_speed_gust'.
Returns
-------
wind_speed : Series
"""
wind_speed = data['wind_speed_gust'] * scaling
return wind_speed
drk = Location(36.621, -116.043, 'US/Pacific', 1010.1072, 'Desert Rock')
# In[6]:
times2009 = pd.DatetimeIndex(start='2009-01-01',
end='2010-01-01',
freq='1min',
tz=drk.tz) # 12 months of 2009 - For testing
times2010and2011 = pd.DatetimeIndex(
start='2010-01-01', end='2012-01-01', freq='1min',
tz=drk.tz) # 24 months of 2010 and 2011 - For training
# In[7]:
cs_2009 = drk.get_clearsky(times2009)
cs_2010and2011 = drk.get_clearsky(
times2010and2011) # ineichen with climatology table by default
#cs_2011 = bvl.get_clearsky(times2011)
# In[8]:
cs_2009.drop(
['dni', 'dhi'], axis=1,
inplace=True) #updating the same dataframe by dropping two columns
cs_2010and2011.drop(
['dni', 'dhi'], axis=1,
inplace=True) #updating the same dataframe by dropping two columns
#cs_2011.drop(['dni','dhi'],axis=1, inplace=True) #updating the same dataframe by dropping two columns
# In[9]:
class PVSim:
""" This PV Simulator calculates the output of a PV panel based on its parameters such as location, orientation,
and rated power using the pvlib module.
The output corresponds to the PV production in kilowatts, producers are assumed to have positive values."""
system: PVSystem
location: Location
mc: ModelChain
subscriber: Subscriber
logger: Logger
def __init__(self):
self.system = PVSystem(
module_parameters={
'pdc0': _PV_SYS_CONFIG['pdc0'],
'gamma_pdc': _PV_SYS_CONFIG['gamma_pdc']
},
inverter_parameters={'pdc0': _PV_SYS_CONFIG['pdc0']},
temperature_model_parameters=_TEMP_MODEL_PARAMS,
surface_tilt=_PV_SYS_CONFIG['surface_tilt'],
surface_azimuth=_PV_SYS_CONFIG['surface_azimuth'])
self.location = Location(latitude=_PV_SYS_CONFIG['latitude'],
longitude=_PV_SYS_CONFIG['longitude'])
self.mc = ModelChain(self.system,
self.location,
aoi_model='physical',
spectral_model='no_loss')
self.logger = Logger(_LOG_FILEPATH)
self.subscriber = Subscriber(callback_ctrl=self._on_new_meter_ctrl,
callback_data=self._on_new_meter_data)
def run(self):
self.logger.writerow(
('Datetime', 'Pac_HH[kW]', 'Pac_PV[kW]', 'Pac_sum[kW]'))
self.subscriber.run()
def get_pac_kw(self, times: pd.DatetimeIndex):
weather = self.location.get_clearsky(times=times)
self.mc.run_model(weather)
return self.mc.ac / 1e3
def _on_new_meter_data(self, timestamp, meter_pac_kw):
times = pd.date_range(start=timestamp,
end=timestamp,
tz=_PV_SYS_CONFIG['timezone'])
pv_pac_kw = self.get_pac_kw(times).values[0]
sum_pac_kw = meter_pac_kw + pv_pac_kw
data = (timestamp, meter_pac_kw, pv_pac_kw, sum_pac_kw)
# print(data)
self.logger.writerow(data)
def _on_new_meter_ctrl(self, msg):
# print(msg)
if msg == "done":
self._cleanup()
def _cleanup(self):
self.subscriber.close()
self.logger.close()
self._plot_results_to_file()
def _plot_results_to_file(self):
df = pd.read_csv(_LOG_FILEPATH, index_col=0)
df.plot()
plt.xlabel('Zeit')
plt.ylabel('Leistung in kW')
plt.setp(plt.xticks()[1], rotation=30, ha='right')
plt.grid()
plt.subplots_adjust(left=0.2, bottom=0.25)
plt.savefig(f'{_RESULTS_DIR_PATH}/plot.png', dpi=150)
