coerce and rotate pvgis TMY data to desired tz and year

docs/sphinx/source/whatsnew/v0.11.1.rst

@@ -19,6 +19,10 @@ Enhancements
 * Add new parameters for min/max absolute air mass to
   :py:func:`pvlib.spectrum.spectral_factor_firstsolar`.
   (:issue:`2086`, :pull:`2100`)
+* Add ``roll_utc_offset`` and ``coerce_year`` arguments to
+  :py:func:`pvlib.iotools.get_pvgis_tmy` to allow user to specify time zone,
+  rotate indices of TMY to begin at midnight, and force indices to desired
+  year. (:issue:`2139`, :pull:`2138`)
 * Restructured the pvlib/spectrum folder by breaking up the contents of
   pvlib/spectrum/mismatch.py into pvlib/spectrum/mismatch.py,
   pvlib/spectrum/irradiance.py, and
@@ -62,5 +66,6 @@ Contributors
 * Leonardo Micheli (:ghuser:`lmicheli`)
 * Echedey Luis (:ghuser:`echedey-ls`)
 * Rajiv Daxini (:ghuser:`RDaxini`)
+* Mark A. Mikofski (:ghuser:`mikofski`)
 * Ben Pierce (:ghuser:`bgpierc`)
 * Jose Meza (:ghuser:`JoseMezaMendieta`)

pvlib/iotools/pvgis.py

@@ -18,10 +18,10 @@
 import json
 from pathlib import Path
 import requests
+import numpy as np
 import pandas as pd
+import pytz
 from pvlib.iotools import read_epw, parse_epw
-import warnings
-from pvlib._deprecation import pvlibDeprecationWarning
 
 URL = 'https://re.jrc.ec.europa.eu/api/'
 
@@ -390,9 +390,33 @@ def read_pvgis_hourly(filename, pvgis_format=None, map_variables=True):
     raise ValueError(err_msg)
 
 
+def _coerce_and_roll_tmy(tmy_data, tz, year):
+    """
+    Assumes ``tmy_data`` input is UTC, converts from UTC to ``tz``, rolls
+    dataframe so timeseries starts at midnight, and forces all indices to
+    ``year``. Only works for integer ``tz``, but ``None`` and ``False`` are
+    re-interpreted as zero / UTC.
+    """
+    if tz:
+        tzname = pytz.timezone(f'Etc/GMT{-tz:+d}')
+    else:
+        tz = 0
+        tzname = pytz.timezone('UTC')
+    new_index = pd.DatetimeIndex([
+        timestamp.replace(year=year, tzinfo=tzname)
+        for timestamp in tmy_data.index],
+        name=f'time({tzname})')
+    new_tmy_data = pd.DataFrame(
+        np.roll(tmy_data, tz, axis=0),
+        columns=tmy_data.columns,
+        index=new_index)
+    return new_tmy_data
+
+
 def get_pvgis_tmy(latitude, longitude, outputformat='json', usehorizon=True,
                   userhorizon=None, startyear=None, endyear=None,
-                  map_variables=True, url=URL, timeout=30):
+                  map_variables=True, url=URL, timeout=30,
+                  roll_utc_offset=None, coerce_year=None):
     """
     Get TMY data from PVGIS.
 
@@ -424,6 +448,13 @@ def get_pvgis_tmy(latitude, longitude, outputformat='json', usehorizon=True,
         base url of PVGIS API, append ``tmy`` to get TMY endpoint
     timeout : int, default 30
         time in seconds to wait for server response before timeout
+    roll_utc_offset: int, optional
+        Use to specify a time zone other than the default UTC zero and roll
+        dataframe by ``roll_utc_offset`` so it starts at midnight on January
+        1st. Ignored if ``None``, otherwise will force year to ``coerce_year``.
+    coerce_year: int, optional
+        Use to force indices to desired year. Will default to 1990 if
+        ``coerce_year`` is not specified, but ``roll_utc_offset`` is specified.
 
     Returns
     -------
@@ -510,6 +541,11 @@ def get_pvgis_tmy(latitude, longitude, outputformat='json', usehorizon=True,
     if map_variables:
         data = data.rename(columns=VARIABLE_MAP)
 
+    if not (roll_utc_offset is None and coerce_year is None):
+        # roll_utc_offset is specified, but coerce_year isn't
+        coerce_year = coerce_year or 1990
+        data = _coerce_and_roll_tmy(data, roll_utc_offset, coerce_year)
+
     return data, months_selected, inputs, meta
 
 

pvlib/tests/iotools/test_pvgis.py

@@ -435,6 +435,53 @@ def _compare_pvgis_tmy_basic(expected, meta_expected, pvgis_data):
         assert np.allclose(data[outvar], expected[outvar])
 
 
+@pytest.mark.remote_data
+@pytest.mark.flaky(reruns=RERUNS, reruns_delay=RERUNS_DELAY)
+def test_get_pvgis_tmy_coerce_year():
+    """test utc_offset and coerce_year work as expected"""
+    base_case, _, _, _ = get_pvgis_tmy(45, 8)  # Turin
+    assert str(base_case.index.tz) == 'UTC'
+    assert base_case.index.name == 'time(UTC)'
+    noon_test_data = [
+        base_case[base_case.index.month == m].iloc[12]
+        for m in range(1, 13)]
+    cet_tz = 1  # Turin time is CET
+    cet_name = 'Etc/GMT-1'
+    # check indices of rolled data after converting timezone
+    pvgis_data, _, _, _ = get_pvgis_tmy(45, 8, roll_utc_offset=cet_tz)
+    jan1_midnight = pd.Timestamp('1990-01-01 00:00:00', tz=cet_name)
+    dec31_midnight = pd.Timestamp('1990-12-31 23:00:00', tz=cet_name)
+    assert pvgis_data.index[0] == jan1_midnight
+    assert pvgis_data.index[-1] == dec31_midnight
+    assert pvgis_data.index.name == f'time({cet_name})'
+    # spot check rolled data matches original
+    for m, test_case in enumerate(noon_test_data):
+        expected = pvgis_data[pvgis_data.index.month == m+1].iloc[12+cet_tz]
+        assert all(test_case == expected)
+    # repeat tests with year coerced
+    test_yr = 2021
+    pvgis_data, _, _, _ = get_pvgis_tmy(
+        45, 8, roll_utc_offset=cet_tz, coerce_year=test_yr)
+    jan1_midnight = pd.Timestamp(f'{test_yr}-01-01 00:00:00', tz=cet_name)
+    dec31_midnight = pd.Timestamp(f'{test_yr}-12-31 23:00:00', tz=cet_name)
+    assert pvgis_data.index[0] == jan1_midnight
+    assert pvgis_data.index[-1] == dec31_midnight
+    assert pvgis_data.index.name == f'time({cet_name})'
+    for m, test_case in enumerate(noon_test_data):
+        expected = pvgis_data[pvgis_data.index.month == m+1].iloc[12+cet_tz]
+        assert all(test_case == expected)
+    # repeat tests with year coerced but utc offset none or zero
+    pvgis_data, _, _, _ = get_pvgis_tmy(45, 8, coerce_year=test_yr)
+    jan1_midnight = pd.Timestamp(f'{test_yr}-01-01 00:00:00', tz='UTC')
+    dec31_midnight = pd.Timestamp(f'{test_yr}-12-31 23:00:00', tz='UTC')
+    assert pvgis_data.index[0] == jan1_midnight
+    assert pvgis_data.index[-1] == dec31_midnight
+    assert pvgis_data.index.name == 'time(UTC)'
+    for m, test_case in enumerate(noon_test_data):
+        expected = pvgis_data[pvgis_data.index.month == m+1].iloc[12]
+        assert all(test_case == expected)
+
+
 @pytest.mark.remote_data
 @pytest.mark.flaky(reruns=RERUNS, reruns_delay=RERUNS_DELAY)
 def test_get_pvgis_tmy_csv(expected, month_year_expected, inputs_expected,