implemented mode -1

docs/examples/snow-detection/snow-mode.py

@@ -52,6 +52,7 @@
 import numpy as np
 import re
 import pvlib
+from matplotlib import colormaps as cm
 import matplotlib.pyplot as plt
 from matplotlib.dates import DateFormatter
 import matplotlib.patches as mpatches
@@ -187,7 +188,7 @@
 
     # Data where inverter is clipping based on AC power
     mask1 = data[a] > max_ac_power
-    mask2 = clipping.geometric(ac_power=data[a], freq='15T')
+    mask2 = clipping.geometric(ac_power=data[a], freq='15min')
     mask3 = np.logical_or(mask1.values, mask2.values)
 
     data.loc[mask3, v] = np.nan
@@ -595,7 +596,11 @@ def wrapper(voltage, current, temp_cell, effective_irradiance,
     loss = np.maximum(data[name_modeled_power] - data[i_col]*data[v_col], 0)
     data[name_loss] = loss
 
-# %%
+# %% Plot measured and modeled power, color by mode
+
+N = 6
+alpha = 0.5
+cmap = cm.get_cmap('plasma').resampled(N)
 
 loss_cols = [c for c in data.columns if "Loss" in c]
 mode_cols = [c for c in data.columns if "mode" in c and "modeled" not in c]
@@ -606,13 +611,6 @@ def wrapper(voltage, current, temp_cell, effective_irradiance,
 mod = mode_cols[col]
 pwr = modeled_power_cols[col]
 
-# Color intervals by mode
-cmap = {0: 'r',
-        1: 'b',
-        2: 'yellow',
-        3: 'cyan',
-        4: 'g'}
-
 fig, ax = plt.subplots(figsize=(10, 10))
 date_form = DateFormatter("%m/%d")
 ax.xaxis.set_major_formatter(date_form)
@@ -628,15 +626,16 @@ def wrapper(voltage, current, temp_cell, effective_irradiance,
     ax.plot(temp_grouped[pwr], c='k', ls='--')
     ax.plot(temp_grouped[pwr] - temp_grouped[los], c='k')
     ax.fill_between(temp_grouped.index, temp_grouped[pwr] - temp_grouped[los],
-                    temp_grouped[pwr], color='k', alpha=0.2)
+                    temp_grouped[pwr], color='k', alpha=alpha)
 
     chng_pts = np.ravel(np.where(temp_grouped[mod].values[:-1]
                                  - temp_grouped[mod].values[1:] != 0))
 
     if len(chng_pts) == 0:
         ax.axvspan(temp_grouped.index[0], temp_grouped.index[-1],
-                   color=cmap[temp_grouped.at[temp_grouped.index[-1], mod]],
-                   alpha=0.05)
+                   color=cmap.colors[temp_grouped.at[temp_grouped.index[-1],
+                                                     mod]],
+                   alpha=alpha)
     else:
         set1 = np.append([0], chng_pts)
         set2 = np.append(chng_pts, [-1])
@@ -645,29 +644,22 @@ def wrapper(voltage, current, temp_cell, effective_irradiance,
             my_index = temp_grouped.index[start:end]
             ax.axvspan(
                 temp_grouped.index[start], temp_grouped.index[end],
-                color=cmap[temp_grouped.at[temp_grouped.index[end], mod]],
-                alpha=0.05)
+                color=cmap.colors[temp_grouped.at[temp_grouped.index[end],
+                                                  mod]],
+                alpha=alpha, ec=None)
 
 # Add different colored intervals to legend
 handles, labels = ax.get_legend_handles_labels()
 
 modeled_line = Line2D([0], [0], label='Modeled', color='k', ls='--')
 measured_line = Line2D([0], [0], label='Measured', color='k')
-
-# TODO don't mark offline periods as Mode 0
-red_patch = mpatches.Patch(color='r', alpha=0.05, label='Mode 0')
-blue_patch = mpatches.Patch(color='b', alpha=0.05, label='Mode 1')
-yellow_patch = mpatches.Patch(color='y', alpha=0.05, label='Mode 2')
-purple_patch = mpatches.Patch(color='cyan', alpha=0.05, label='Mode 3')
-green_patch = mpatches.Patch(color='g', alpha=0.05, label='Mode 4')
-
 handles.append(measured_line)
 handles.append(modeled_line)
-handles.append(red_patch)
-handles.append(green_patch)
-handles.append(blue_patch)
-handles.append(yellow_patch)
-handles.append(purple_patch)
+
+for i in range(-1, N-1):
+    my_patch = mpatches.Patch(color=cmap.colors[i], label=f'Mode {i}')
+    handles.append(my_patch)
+
 # handles.append(gray_patch)
 
 ax.set_xlabel('Date', fontsize='xx-large')
@@ -724,4 +716,6 @@ def wrapper(voltage, current, temp_cell, effective_irradiance,
 ax.set_ylabel('[%]', fontsize='xx-large')
 ax.set_xticks(xvals, days)
 ax.xaxis.set_major_formatter(date_form)
-ax.set_title('Losses incurred in modes 0 -3', fontsize='xx-large')
+ax.set_title('Losses incurred in modes -1, 0, 1, 2, 3', fontsize='xx-large')
+
+# %%

pvanalytics/features/snow.py

@@ -158,7 +158,7 @@ def categorize(vmp_ratio, transmission, measured_voltage, modeled_voltage,
     * Mode 0: Indicates periods with enough opaque snow that the system is not
       producing power. Specifically, Mode 0 is when the measured voltage is
       below the inverter's turn-on voltage but the voltage modeled using
-      measured irradiance is below the inverter's turn-on voltage. 
+      measured irradiance is above the inverter's turn-on voltage. 
     * Mode 1: Indicates periods when the system has non-uniform snow and
       both operating voltage and current are decreased. Operating voltage is
       reduced when bypass diodes activate and current is decreased due to
@@ -171,6 +171,12 @@ def categorize(vmp_ratio, transmission, measured_voltage, modeled_voltage,
     * Mode is None when both measured and voltage modeled from measured
       irradiance are below the inverter turn-on voltage.
 
+    An additional mode was added to cover a case that was not addressed in
+    [1]_:
+    * Mode -1: Indicates periods where voltage modeled with measured irradiance 
+      is below the inverter's turn-on voltage. Under Mode -1, it is unknown if
+      and how snow impacts power output.
+
     Parameters
     ----------
     vmp_ratio : array-like
@@ -207,23 +213,26 @@ def categorize(vmp_ratio, transmission, measured_voltage, modeled_voltage,
     umin_meas = measured_voltage >= min_dcv
     umin_model = modeled_voltage >= min_dcv
 
-    # offline if both measurement and model say that voltage is too low.
-    # if either measured or modeled is above the minimum, then system is
+    # offline if model says that voltage is too low.
+    # if modeled is above the minimum, then system is
     # possibly generating
-    offline = ~umin_meas & ~umin_model
+    # offline = ~umin_meas & ~umin_model
 
     # vmp_ratio discrimates between states (1,2) and (3,4)
     uvr = np.where(vmp_ratio >= threshold_vratio, 3, 1)
 
     # transmission discrimates within (1,2) and (3,4)
     utrans = np.where(transmission >= threshold_transmission, 1, 0)
 
-    # None if nan or system is offline
+    # None if nan
+    # if offline:
+    #   - -1 if umin_model is 0
     # if not offline:
     #   - 0 if umin_meas is 0, i.e., measurement indicate no power but
     #     it must be that umin_model is 1
     #   - state 1, 2, 3, 4 defined by uvr + utrans
-    mode = np.where(offline | np.isnan(vmp_ratio) | np.isnan(transmission),
+    mode = np.where(np.isnan(vmp_ratio) | np.isnan(transmission),
                     None, umin_meas * (uvr + utrans))
+    mode = np.where(~umin_model, -1, mode)
 
     return mode

pvanalytics/tests/features/test_snow.py

@@ -66,20 +66,24 @@ def test_get_transmission():
 
 
 def test_categorize():
-    vmp_ratio = np.array([np.nan, 0.9, 0.1, 0.6, 0.7, 0.9, 0.9])
-    measured_voltage = np.array([400., 450., 400., 420., 420., 495., 270.])
-    modeled_voltage = np.array([np.nan, 500, 4000, 700, 600, 550, 300])
-    transmission = np.array([0.5, np.nan, 0.5, 0.9, 0.5, 0.9, 0.9])
+    vmp_ratio = np.array([np.nan, 0.9, 0.1, 0.6, 0.7, 0.9, 0.9, 0.5])
+    measured_voltage = np.array([400., 450., 400., 420., 420., 495., 270.,
+                                 275.])
+    modeled_voltage = np.array([np.nan, 500, 4000, 700, 600, 550, 300, 550])
+    transmission = np.array([0.5, np.nan, 0.5, 0.9, 0.5, 0.9, 0.9, 0.9])
     min_dcv = 300
     threshold_vratio = 0.7
     threshold_transmission = 0.6
-    # ratio: np.nan, >thres, >thres, >thres, >thres, >thres, <thres
-    # measured: np.nan, >thres, >thres, >thres, >thres, >thres, <thres
-    # modeled: np.nan, >thres, >thres, >thres, >thres, >thres, <thres
-    # vr=np.nan, vr<thres, vr<thres, vr=thres, vr>thres, vr>thres, vr>thres
-    # tr<thres, np.nan, tr<thres, tr<thres, tr<thres, tr<thres, tr>thres
-    # None (vr), None (vmo), 1, 3, 3, 3, None
-    expected = np.array([None, None, 1, 2, 3, 4, 0])
+    # vmp_ratio: np.nan, >thres, <thres, <thres, =thres, >thres, >thres, <thres
+    # measured_voltage: >thres, >thres, >thres, >thres, >thres, >thres, <thres,
+    #                   <thres
+    # modeled_voltage: np.nan, >thres, >thres, >thres, >thres, >thres, <thres,
+    #                  >thres
+    # transmission: <thres, np.nan, <thres, >thres, <thres, >thres, >thres,
+    #               >thres, > thres
+    # None (vmp_ratio, modeled_voltage), None (transmission), 1, 2, 3, 4, -1,
+    # 0
+    expected = np.array([None, None, 1, 2, 3, 4, -1, 0])
     result = snow.categorize(vmp_ratio, transmission, measured_voltage,
                              modeled_voltage, min_dcv,
                              threshold_vratio, threshold_transmission)