make it easier to manually pick weights based on cluster numbers in p…

…aper, also to extract load profiles of individual simulated sessions

HowToExtractIndividualSessions.py

@@ -28,7 +28,7 @@
 config = SPEEChGeneralConfiguration(model)
 scenario = Scenarios('BaseCase')
 config.change_pg(new_weights = scenario.new_weights)
-config.change_pg(new_weights = {7:0.2, 10:0.2})#scenario.new_weights)
+config.change_pg(new_weights = {7:0.2, 10:0.2}, dend=True) # If you are making your own weights, use "dend=True"
 
 
 # Run a version of "config.run_all":
@@ -53,3 +53,24 @@
 for key in individual_session_parameters_all.keys():
     if individual_session_parameters_all[key] is not None:
         individual_session_parameters_all[key].to_csv('IndividualSessionsOutputData/'+key+'.csv')
+
+
+# Extract the individual load profiles, not just the session parameters
+categories = ['Home', 'MUD', 'Work', 'Other Slow', 'Other Fast']
+total_load_profiles = {}
+for segment in categories:
+    if individual_session_parameters_all[segment] is not None:
+        individual_session_parameters_all[segment] = individual_session_parameters_all[segment].reset_index(drop=True)
+        n_segment = len(individual_session_parameters_all[segment])
+
+        individual_profiles = {}  # collect the individual load profiles
+        for index_value in np.arange(0, n_segment):  # calculate for each session
+            tmp1, tmp2 = model.end_times_and_load(individual_session_parameters_all[segment].loc[index_value, 'Start'].reshape(1,), individual_session_parameters_all[segment].loc[index_value, 'Energy'].reshape(1,), individual_session_parameters_all[segment].loc[index_value, 'Rate'].reshape(1,))
+            individual_profiles[index_value] = tmp2
+        individual_profiles = pd.DataFrame(individual_profiles)
+        individual_profiles.to_csv('IndividualSessionsOutputData/'+segment+'_individual_load_profiles.csv', index=None)  # save the individual load profiles in the same folder
+        total_load_profiles[segment] = individual_profiles.sum(axis=1)
+
+total_load_profiles = pd.DataFrame(total_load_profiles)
+total_load_profiles['Total'] = total_load_profiles.loc[:, categories].sum(axis=1)  # calculate the total load profile as the sum of all sessions in all segments
+total_load_profiles.to_csv('IndividualSessionsOutputData/total_load_profiles.csv', index=None)  # save the total load profile

SimpleExample.py

@@ -30,7 +30,8 @@
 # Demonstration of changing group weights and behaviour weights:
 model = SPEECh(data)
 config = SPEEChGeneralConfiguration(model)
-config.change_pg(new_weights={0: 0.5, 1: 0.5})  # Adjust distribution over driver groups so that 0 and 1 have each 50%
+config.change_pg(new_weights={3: 0.5, 4: 0.5}, dend=True)  # Adjust distribution over driver groups so that 0 and 1 have each 50%
+# Note counting of groups starts at 0 and ends at 15 (subtract 1 from the numbering in the paper)
 # (^ down-weights the others to 0 accordingly)
 
 # Could also uncomment these two lines to use a pre-set weighting: 
@@ -41,7 +42,7 @@
 config.groups()
 # Give weights 40% and 60% to behaviors 0 and 1 in the group 1 workplace segment:
 # (down-weights the others to 0 accordingly)
-config.change_ps_zg(1, 'Work', 'weekday', {0: 0.4, 1: 0.6})
+config.change_ps_zg(3, 'Work', 'weekday', {0: 0.4, 1: 0.6})
 config.run_all(weekday=weekday_option)
 plots = Plotting(model, config)
 plots.total(save_str='simple_example_plot_adjusted.png')

speech.py

@@ -263,7 +263,7 @@ def change_ps_zg(self, g, cat, weekday, new_weights):
 
             self.group_configs[g].segment_gmms[weekday][cat] = gmm
 
-    def change_pg(self, new_weights, new_col='pg'):
+    def change_pg(self, new_weights, new_col='pg', dend=False):
         """Change the distribution over driver groups, P(g).
 
         Parameters:
@@ -273,6 +273,12 @@ def change_pg(self, new_weights, new_col='pg'):
         The input need only specify new weights for a subset of the components;
         the remaining weight will be distributed among the remaining components proportionately.
         """
+
+        if dend: # If inputs are given using cluster numbers from the dendrogram
+            new_weights_old = copy.deepcopy(new_weights)
+            new_weights = {}
+            for key, val in new_weights_old.items():
+                new_weights[self.speech.data.cluster_reorder_dendtoac[key]] = val
 
         self.speech.pg[new_col] = self.speech.pg['pg'].copy()
         all_inds = np.arange(0, self.speech.data.ng)