def test_fuentes(filename, inoct):
# Test against data exported from pvwatts.nrel.gov
data = _read_pvwatts_8760(DATA_DIR / filename)
data = data.iloc[:24*7, :] # just use one week
inputs = {
'poa_global': data['Plane of Array Irradiance (W/m^2)'],
'temp_air': data['Ambient Temperature (C)'],
'wind_speed': data['Wind Speed (m/s)'],
'noct_installed': inoct,
}
expected_tcell = data['Cell Temperature (C)']
expected_tcell.name = 'tmod'
actual_tcell = temperature.fuentes(**inputs)
# the SSC implementation of PVWatts diverges from the Fuentes model at
# at night by setting Tcell=Tamb when POA=0. This not only means that
# nighttime values are slightly different (Fuentes models cooling to sky
# at night), but because of the thermal inertia, there is a transient
# error after dawn as well. Test each case separately:
is_night = inputs['poa_global'] == 0
is_dawn = is_night.shift(1) & ~is_night
is_daytime = (inputs['poa_global'] > 0) & ~is_dawn
# the accuracy is probably higher than 3 digits here, but the PVWatts
# export data has low precision so can only test up to 3 digits
assert_series_equal(expected_tcell[is_daytime].round(3),
actual_tcell[is_daytime].round(3))
# use lower precision for dawn times to accommodate the dawn transient
error = actual_tcell[is_dawn] - expected_tcell[is_dawn]
assert (error.abs() < 0.1).all()
# sanity check on night values -- Fuentes not much lower than PVWatts
night_difference = expected_tcell[is_night] - actual_tcell[is_night]
assert night_difference.max() < 6
assert night_difference.min() > 0
def test_fuentes_timezone(tz):
index = pd.date_range('2019-01-01', freq='h', periods=3, tz=tz)
df = pd.DataFrame({'poa_global': 1000, 'temp_air': 20, 'wind_speed': 1},
index)
out = temperature.fuentes(df['poa_global'], df['temp_air'],
df['wind_speed'], noct_installed=45)
assert_series_equal(out, pd.Series([47.85, 50.85, 50.85], index=index,
name='tmod'))