Merge pull request #103 from BruisVanVlijmen2020/update_DiagnosticPro…

…perties_nx

Initial commit to discharge throughput normalization

beep/features/featurizer_helpers.py

@@ -867,6 +867,7 @@ def get_fractional_quantity_remaining(
     """
     Determine relative loss of <metric> in diagnostic_cycles of type <diagnostic_cycle_type> after 100 regular cycles
 
+
     Args:
         processed_cycler_run (beep.structure.ProcessedCyclerRun): information about cycler run
         metric (str): column name to use for measuring degradation
@@ -886,3 +887,59 @@ def get_fractional_quantity_remaining(
     )
     summary_diag_cycle_type.columns = ["cycle_index", "fractional_metric"]
     return summary_diag_cycle_type
+
+
+def get_fractional_quantity_remaining_nx(
+    processed_cycler_run, metric="discharge_energy", diagnostic_cycle_type="rpt_0.2C"
+):
+    """
+    Similar to get_fractional_quantity_remaining()
+    Determine relative loss of <metric> in diagnostic_cycles of type <diagnostic_cycle_type>
+    Also returns value of 'x', the discharge throughput passed by the first diagnostic
+    and the value 'n' at each diagnostic
+
+    Args:
+        processed_cycler_run (beep.structure.ProcessedCyclerRun): information about cycler run
+        metric (str): column name to use for measuring degradation
+        diagnostic_cycle_type (str): the diagnostic cycle to use for computing the amount of degradation
+
+    Returns:
+        a dataframe with cycle_index, corresponding degradation relative to the first measured value, 'x',
+        i.e. the discharge throughput passed by the first diagnostic
+        and the value 'n' at each diagnostic, i.e. the equivalent scaling factor for lifetime using n*x
+    """
+    summary_diag_cycle_type = processed_cycler_run.diagnostic_summary[
+        (processed_cycler_run.diagnostic_summary.cycle_type == diagnostic_cycle_type)
+    ].reset_index()
+    summary_diag_cycle_type = summary_diag_cycle_type[["cycle_index", metric]]
+
+    # For the nx addition
+    if 'energy' in metric:
+        normalize_qty = 'discharge' + '_energy'
+    else:
+        normalize_qty = 'discharge' + '_capacity'
+
+    normalize_qty_throughput = normalize_qty + '_throughput'
+    regular_summary = processed_cycler_run.summary.copy()
+    regular_summary[normalize_qty_throughput] = regular_summary[normalize_qty].cumsum()
+    indices = summary_diag_cycle_type.cycle_index
+    initial_throughput = regular_summary.loc[
+        regular_summary.cycle_index == indices.iloc[0]
+    ][normalize_qty_throughput].iloc[0]
+    x = regular_summary.loc[
+            regular_summary.cycle_index == indices.iloc[1]
+        ][normalize_qty_throughput].iloc[0] - initial_throughput
+
+    summary_diag_cycle_type['x'] = x
+    summary_diag_cycle_type['n'] = (
+            (1/x) * regular_summary[regular_summary.cycle_index.isin(indices)][normalize_qty_throughput]
+    ).values
+
+    # end of nx addition
+    summary_diag_cycle_type[metric] = (
+        summary_diag_cycle_type[metric]
+        / processed_cycler_run.diagnostic_summary[metric].iloc[0]
+    )
+
+    summary_diag_cycle_type.columns = ["cycle_index", "fractional_metric", "x", "n"]
+    return summary_diag_cycle_type

beep/featurize.py

@@ -1099,12 +1099,14 @@ def validate_data(cls, processed_cycler_run, params_dict=None):
     @classmethod
     def features_from_processed_cycler_run(cls, processed_cycler_run, params_dict=None):
         """
+        Generates diagnostic-property features from processed cycler run, including values for n*x method
         Args:
             processed_cycler_run (beep.structure.ProcessedCyclerRun): data from cycler run
             params_dict (dict): dictionary of parameters governing how the ProcessedCyclerRun object
             gets featurized. These could be filters for column or row operations
         Returns:
-            pd.DataFrame: cycles at which capacity/energy degradation exceeds thresholds
+            pd.DataFrame: with "cycle_index", "fractional_metric", "x", "n", "cycle_type" and "metric" columns, rows
+            for each diagnostic cycle of the cell
         """
         if params_dict is None:
             params_dict = FEATURE_HYPERPARAMS[cls.class_feature_name]
@@ -1113,9 +1115,10 @@ def features_from_processed_cycler_run(cls, processed_cycler_run, params_dict=No
         X = pd.DataFrame()
         for quantity in params_dict["quantities"]:
             for cycle_type in cycle_types:
-                summary_diag_cycle_type = featurizer_helpers.get_fractional_quantity_remaining(
+                summary_diag_cycle_type = featurizer_helpers.get_fractional_quantity_remaining_nx(
                     processed_cycler_run, quantity, cycle_type
                 )
+
                 summary_diag_cycle_type["cycle_type"] = cycle_type
                 summary_diag_cycle_type["metric"] = quantity
                 X = X.append(summary_diag_cycle_type)

beep/tests/test_features.py

@@ -198,7 +198,7 @@ def test_feature_generation_list_to_json(self):
             self.assertIsInstance(features_reloaded, DiagnosticProperties)
             self.assertListEqual(
                 list(features_reloaded.X.iloc[2, :]),
-                [143, 0.9753520623934744, "rpt_0.2C", "discharge_energy"],
+                [141,0.9859837086597274,91.17758004259996,2.578137278917377,'reset','discharge_energy'],
             )
 
             # Workflow output
@@ -357,8 +357,8 @@ def test_DiagnosticProperties_class(self):
             folder = os.path.split(path)[-1]
             dumpfn(featurizer, featurizer.name)
             self.assertEqual(folder, "DiagnosticProperties")
-            self.assertEqual(featurizer.X.shape, (10, 4))
+            self.assertEqual(featurizer.X.shape, (30, 6))
             self.assertListEqual(
                 list(featurizer.X.iloc[2, :]),
-                [143, 0.9753520623934744, "rpt_0.2C", "discharge_energy"],
+                [141,0.9859837086597274,91.17758004259996,2.578137278917377,'reset','discharge_energy']
             )