bugfix for high frequency time series in scaling.py, regarding issue 1257

add pull request number

bugfix for high frequency time series in scaling.py, regarding issue 1257

Author: jranalli

docs/sphinx/source/whatsnew/v0.9.0.rst

@@ -175,6 +175,10 @@ Bug fixes
 * Corrected an error affecting :py:func:`~pvlib.clearsky.detect_clearsky`
   when data time step is not one minute. Error was introduced in v0.8.1.
   (:issue:`1241`, :pull:`1242`)
+* Corrected error affecting :py:func:`~pvlib.scaling._compute_wavelet` when
+  passing a pandas time series with a sampling rate faster than 1 second.
+  (:issue:`1257`, :pull:`1258`)
+  (:issue:`1257`, :pull:``)
 
 Testing
 ~~~~~~~

pvlib/scaling.py

@@ -205,7 +205,8 @@ def _compute_wavelet(clearsky_index, dt=None):
             raise ValueError("dt must be specified for numpy type inputs.")
     else:  # flatten() succeeded, thus it's a pandas type, so get its dt
         try:  # Assume it's a time series type index
-            dt = (clearsky_index.index[1] - clearsky_index.index[0]).seconds
+            dt = clearsky_index.index[1] - clearsky_index.index[0]
+            dt = dt.seconds + dt.microseconds/1e6
         except AttributeError:  # It must just be a numeric index
             dt = (clearsky_index.index[1] - clearsky_index.index[0])
 
@@ -221,7 +222,7 @@ def _compute_wavelet(clearsky_index, dt=None):
     csi_mean = np.zeros([max_tmscale, len(cs_long)])
     # Skip averaging for the 0th scale
     csi_mean[0, :] = cs_long.values.flatten()
-    tmscales[0] = 1
+    tmscales[0] = dt
     # Loop for all time scales we will consider
     for i in np.arange(1, max_tmscale):
         tmscales[i] = 2**i * dt  # Wavelet integration time scale

pvlib/tests/test_scaling.py

@@ -37,6 +37,18 @@ def time(clear_sky_index):
     return np.arange(0, len(clear_sky_index))
 
 
+@pytest.fixture
+def time_60s(clear_sky_index):
+    # Sample time vector 60s resolution
+    return np.arange(0, len(clear_sky_index))*60
+
+
+@pytest.fixture
+def time_500ms(clear_sky_index):
+    # Sample time vector 0.5s resolution
+    return np.arange(0, len(clear_sky_index))*0.5
+
+
 @pytest.fixture
 def positions():
     # Sample positions based on the previous lat/lon (calculated manually)
@@ -51,6 +63,18 @@ def expect_tmscale():
     return [1, 2, 4, 8, 16, 32, 64, 128, 256, 512, 1024, 2048, 4096]
 
 
+@pytest.fixture
+def expect_tmscale_1min():
+    # Expected timescales for dt = 60
+    return [60, 120, 240, 480, 960, 1920, 3840]
+
+
+@pytest.fixture
+def expect_tmscale_500ms():
+    # Expected timescales for dt = 0.5
+    return [0.5, 1, 2, 4, 8, 16, 32, 64, 128, 256, 512, 1024, 2048, 4096]
+
+
 @pytest.fixture
 def expect_wavelet():
     # Expected wavelet for indices 5000:5004 for clear_sky_index above (Matlab)
@@ -109,6 +133,25 @@ def test_compute_wavelet_series_numindex(clear_sky_index, time,
     assert_almost_equal(wavelet[:, 5000:5005], expect_wavelet)
 
 
+def test_compute_wavelet_series_highres(clear_sky_index, time_500ms,
+                                         expect_tmscale_500ms, expect_wavelet):
+    dtindex = pd.to_datetime(time_500ms, unit='s')
+    csi_series = pd.Series(clear_sky_index, index=dtindex)
+    wavelet, tmscale = scaling._compute_wavelet(csi_series)
+    assert_almost_equal(tmscale, expect_tmscale_500ms)
+    assert_almost_equal(wavelet[:, 5000:5005].shape, (14, 5))
+
+
+def test_compute_wavelet_series_minuteres(clear_sky_index, time_60s,
+                                         expect_tmscale_1min, expect_wavelet):
+    dtindex = pd.to_datetime(time_60s, unit='s')
+    csi_series = pd.Series(clear_sky_index, index=dtindex)
+    wavelet, tmscale = scaling._compute_wavelet(csi_series)
+    assert_almost_equal(tmscale, expect_tmscale_1min)
+    assert_almost_equal(wavelet[:, 5000:5005].shape,
+                        expect_wavelet[0:len(tmscale), :].shape)
+
+
 def test_compute_wavelet_array(clear_sky_index,
                                expect_tmscale, expect_wavelet):
     wavelet, tmscale = scaling._compute_wavelet(clear_sky_index, dt)