Merge branch 'main' into development

.readthedocs.yml

@@ -1,7 +1,11 @@
 version: 2
 
+build:
+   os: ubuntu-20.04
+   tools:
+     python: "3.8"
+
 python:
-    version: "3.8"
     install:
         - method: pip
           path: .

bifacial_radiance/main.py

@@ -257,7 +257,10 @@ def _subhourlydatatoGencumskyformat(gencumskydata, label='right'):
 
 
     #Resample to hourly. Gencumsky wants right-labeled data.
-    gencumskydata = gencumskydata.resample('60T', closed='right', label='right').mean()       
+    try:
+        gencumskydata = gencumskydata.resample('60T', closed='right', label='right').mean()  
+    except TypeError: # Pandas 2.0 error
+        gencumskydata = gencumskydata.resample('60T', closed='right', label='right').mean(numeric_only=True) 
 
     if label == 'left': #switch from left to right labeled by adding an hour
         gencumskydata.index = gencumskydata.index + pd.to_timedelta('1H')

bifacial_radiance/mismatch.py

@@ -172,52 +172,83 @@ def mismatch_fit3(data):
     
     Parameters
     ----------
-    data : np.ndarray
+    data : np.ndarray, pd.Series, pd.DataFrame
         Gtotal irradiance measurements. Each column is the irradiance for a module
-        at a specific time. (be carefule -- might have to .T the dataframe as input) 
+        at a specific time. 
 
     Returns
     -------
-    fit3 :  1Darray
+    fit3 :  Float or pd.Series
         Returns mismatch values for each module
     
     Equation: 1/(n^2*Gavg)*Sum Sum (abs(G_i - G_j))
     ## Note: starting with Pandas 1.0.0 this function will not work on Series objects.
     '''
     import numpy as np
+    import pandas as pd
+
+    if type(data) == np.ndarray:
+        data = pd.DataFrame(data)
 
     datac = data[~np.isnan(data)]
     mad = mad_fn(datac) /100  # (percentage)
     mad2 = mad**2
 
     fit3 = 0.054*mad + 0.068*mad2
+
+    if fit3.__len__() == 1:
+        fit3 = float(fit3)
 
     return fit3
 
 
-def mad_fn(data):
+def mad_fn(data, axis='index'):
     '''
     Mean average deviation calculation for mismatch purposes.
     
     Parameters
     ----------
-    data : np.ndarray
-        Gtotal irradiance measurements.
+    data : np.ndarray or pd.Series or pd.DataFrame
+        Gtotal irradiance measurements. If data is a pandas.DataFrame, one 
+        MAD/Average is returned for each index, based on values across columns.
+        
+    axis : {0 or 'index', 1 or 'columns'}, default 'index'   
+        Calculate mean average deviation across rows (default) or columns for 2D data
 
+        * 0, or 'index' : MAD calculated across rows.
+        * 1, or 'columns' : MAD calculated across columns.
     Returns
     -------
-    scalar :   return MAD / Average for a 1D array
+    scalar or pd.Series:   return MAD / Average [%]. Scalar for a 1D array, Series for 2D.
     
-    Equation: 1/(n^2*Gavg)*Sum Sum (abs(G_i - G_j))
-    ## Note: starting with Pandas 1.0.0 this function will not work on Series objects.
+    
+    Equation: 1/(n^2*Gavg)*Sum Sum (abs(G_i - G_j)) * 100[%]
+
     '''
     import numpy as np
     import pandas as pd
-    # Pandas returns a notimplemented error if this is a series.
-    if type(data) == pd.Series:
+    def _mad_1D(data):  #1D calculation of MAD
+        return (np.abs(np.subtract.outer(data,data)).sum()/float(data.__len__())**2 / np.mean(data))*100
+    if type(axis) == str:
+        try:
+            axis = {"index": 0, "rows": 0, 'columns':1}[axis]
+        except KeyError:
+            raise Exception('Incorrect index string in mad_fn. options: index, rows, columns.')
+
+    ndim = data.ndim
+    if ndim == 2 and axis==0:
+        data = data.T
+    # Pandas returns a notimplemented error if this is a DataFrame.
+    if (type(data) == pd.Series):
         data = data.to_numpy()
 
-    return (np.abs(np.subtract.outer(data,data)).sum()/float(data.__len__())**2 / np.mean(data))*100
+    if type(data) == pd.DataFrame:
+        temp = data.apply(pd.Series.to_numpy, axis=1)
+        return(temp.apply(_mad_1D))
+    elif ndim ==2: #2D array
+        return [_mad_1D(i) for i in data]
+    else:
+        return _mad_1D(data)
 
 
 
@@ -346,24 +377,19 @@ def analysisIrradianceandPowerMismatch(testfolder, writefiletitle, portraitorlan
     F.index='FrontIrradiance_cell_'+F.index.astype(str)
     B.index='BackIrradiance_cell_'+B.index.astype(str)
     Poat.index='POAT_Irradiance_cell_'+Poat.index.astype(str)
-
-    ## Transpose 
-    F = F.T
-    B = B.T
-    Poat = Poat.T
 
     # Statistics Calculatoins
     dfst=pd.DataFrame()
-    dfst['MAD/G_Total'] = mad_fn(Poat.T)
-    dfst['Front_MAD/G_Total'] = mad_fn(F.T)
+    dfst['MAD/G_Total'] = mad_fn(Poat)
+    dfst['Front_MAD/G_Total'] = mad_fn(F)
     dfst['MAD/G_Total**2'] = dfst['MAD/G_Total']**2
     dfst['Front_MAD/G_Total**2'] = dfst['Front_MAD/G_Total']**2
-    dfst['poat'] = Poat.mean(axis=1)
-    dfst['gfront'] = F.mean(axis=1)
-    dfst['grear'] = B.mean(axis=1)
+    dfst['poat'] = Poat.mean()
+    dfst['gfront'] = F.mean()
+    dfst['grear'] = B.mean()
     dfst['bifi_ratio'] =  dfst['grear']/dfst['gfront']
-    dfst['stdev'] = Poat.std(axis=1)/ dfst['poat']
-    dfst.index=Poat.index.astype(str)
+    dfst['stdev'] = Poat.std()/ dfst['poat']
+    dfst.index=Poat.columns.astype(str)
 
     # Power Calculations/Saving
     Pout=pd.DataFrame()
@@ -373,10 +399,10 @@ def analysisIrradianceandPowerMismatch(testfolder, writefiletitle, portraitorlan
     Pout['Front_Pdet']=Pdet_front_all
     Pout['Mismatch_rel'] = 100-(Pout['Pdet']*100/Pout['Pavg'])
     Pout['Front_Mismatch_rel'] = 100-(Pout['Front_Pdet']*100/Pout['Front_Pavg'])   
-    Pout.index=Poat.index.astype(str)
+    Pout.index=Poat.columns.astype(str)
 
-    ## Save CSV
-    df_all = pd.concat([Pout,dfst,Poat,F,B],axis=1)
+    ## Save CSV as one long row
+    df_all = pd.concat([Pout, dfst, Poat.T, F.T, B.T], axis=1)
     df_all.to_csv(writefiletitle)
     print("Saved Results to ", writefiletitle)
 

docs/sphinx/source/whatsnew/pending.rst

@@ -11,7 +11,7 @@ API Changes
 * Results generated with the above can be saved with the :py:class:`~bifacial_radiance.RadianceObj.exportTrackerDict`, which saves an Hourly, Monthly and Yearly .csvs in the results folder.
 *Multiple modules and rows can now be selected in a single analysis scan. ``modWanted`` and ``rowWanted`` inputs in :py:class:`~bifacial_radiance.RadianceObj.analysis1axis` can now be a list, to select multiple rows and modules for scans. (:issue:`405`)(:pull:`408`)
 *To support multiple modules and row scans for 1axis simulations, outputs like Wm2Front are now stored in ``trackerdict``.``Results``  (:issue:`405`)(:pull:`408`)
-
+* ``mismatch.mad_fn`` has new functionality and input parameter `axis`. If a 2D matrix or dataframe is passed in as data, MAD is calculated along the row (default) or along the columns by passing 'axis=1'
 
 Enhancements
 ~~~~~~~~~~~~
@@ -21,6 +21,7 @@ Enhancements
 
 Bug fixes
 ~~~~~~~~~
+* Fixed  Pandas 2.0 errors by re-factoring ``mismatch.mad_fn``  (:issue:`449`)
 * Fixed typo on Opacity calculation factor (:issue:`426`)
 
 Documentation
@@ -31,4 +32,5 @@ Documentation
 Contributors
 ~~~~~~~~~~~~
 * Silvana Ayala (:ghuser:`shirubana`)
-* Chris Deline (:ghuser:`cdeline`)
+* Chris Deline (:ghuser:`cdeline`)
+* Kevin Anderson (:ghuser:`kandersolar`)

tests/test_mismatch.py

@@ -49,18 +49,29 @@ def test_setupforPVMismatch():
 
 
 def test_MAD():
+    ans_T = pd.Series([ 70.892019,  69.953052, 69.014085,  68.075117,
+                      67.136150,  66.197183, 65.258216,  64.319249,
+                      63.380282,  62.441315, 61.502347,  60.563380])    
+    ans = pd.Series([72.222222, 22.698413, 13.465160,  
+                       9.571620,  7.424714,   6.064461])
 
     assert bifacial_radiance.mismatch.mad_fn(TEST_ARRAY) == \
-        pytest.approx(2433.333,abs = 0.001)
+        pytest.approx(ans.to_numpy(), abs = 0.001)
+
+    assert bifacial_radiance.mismatch.mad_fn(TEST_ARRAY, axis='columns') == \
+        pytest.approx(ans_T.to_numpy(), abs = 0.001)        
 
     temp = bifacial_radiance.mismatch.mad_fn(pd.DataFrame(TEST_ARRAY))
-    ans = pd.Series([15706.061,4936.190,2928.249,2081.526,1614.642,1318.8295])
-    pd.testing.assert_series_equal(temp,ans,check_less_precise=True)
+    temp2 = bifacial_radiance.mismatch.mad_fn(pd.DataFrame(TEST_ARRAY), axis=1)
+    pd.testing.assert_series_equal(temp, ans)
+    pd.testing.assert_series_equal(temp2, ans_T)
     # test pd.Series objects are correctly handled
-    assert bifacial_radiance.mismatch.mad_fn(ans) == \
-        pytest.approx(96.491,abs = 0.001)
-#    assert temp == \
-#        pytest.approx(2433.333,abs = 0.001)    
+    assert bifacial_radiance.mismatch.mad_fn(ans_T) == \
+        pytest.approx(5.674, abs = 0.001)
+    assert bifacial_radiance.mismatch.mad_fn(ans_T.to_numpy()) == \
+        pytest.approx(5.674, abs = 0.001)
+
+
 
 def test_analysisIrradianceandPowerMismatch():
     #analysisIrradianceandPowerMismatch(testfolder, writefiletitle, 
@@ -77,4 +88,11 @@ def test_analysisIrradianceandPowerMismatch():
     df_all = pd.read_csv(writefiletitle)
     assert df_all.Mismatch_rel[0] == pytest.approx(0.376, abs = 0.001)
     assert df_all["MAD/G_Total"][0] == pytest.approx(1.987, abs = 0.001)
+
+
+def test_mismatch_fit3():
+    ans = pd.Series([0.074469, 0.015761, 0.008504,  0.005792,  0.004384, 0.003525])
+    pd.testing.assert_series_equal( bifacial_radiance.mismatch.mismatch_fit3(TEST_ARRAY),  ans, atol=1e-6)
+    pd.testing.assert_series_equal( bifacial_radiance.mismatch.mismatch_fit3(pd.DataFrame(TEST_ARRAY)),  ans, atol=1e-6)
+    assert bifacial_radiance.mismatch.mismatch_fit3(TEST_ARRAY[:,0]) == pytest.approx(ans[0], abs = 0.001)