From 742c762f19b0f9313310f275b1a7d8c88c273e72 Mon Sep 17 00:00:00 2001
From: Fabian Neumann <fabian.neumann@outlook.de>
Date: Mon, 21 Aug 2023 14:15:04 +0200
Subject: [PATCH 1/4] improve performance of normalizedTypicalPeriods creation

---
 tsam/timeseriesaggregation.py | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/tsam/timeseriesaggregation.py b/tsam/timeseriesaggregation.py
index 703c3a9..f8adf0f 100644
--- a/tsam/timeseriesaggregation.py
+++ b/tsam/timeseriesaggregation.py
@@ -1066,9 +1066,10 @@ def createTypicalPeriods(self):
             )
 
         # put the clustered data in pandas format and scale back
-        self.normalizedTypicalPeriods = pd.DataFrame(
-            self.clusterPeriods, columns=self.normalizedPeriodlyProfiles.columns
-        ).stack(level="TimeStep")
+        self.normalizedTypicalPeriods = pd.concat([
+            pd.Series(s, index=self.normalizedPeriodlyProfiles.columns)
+            for s in self.clusterPeriods
+        ], axis=1).unstack("TimeStep").T
 
         if self.segmentation:
             from tsam.utils.segmentation import segmentation

From cba0523509784c917befa6e54e21c08d0d307928 Mon Sep 17 00:00:00 2001
From: Fabian Neumann <fabian.neumann@outlook.de>
Date: Mon, 21 Aug 2023 14:15:52 +0200
Subject: [PATCH 2/4] use .clip() to simplify _rescaleClusterPeriods

---
 tsam/timeseriesaggregation.py | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/tsam/timeseriesaggregation.py b/tsam/timeseriesaggregation.py
index f8adf0f..f3349c3 100644
--- a/tsam/timeseriesaggregation.py
+++ b/tsam/timeseriesaggregation.py
@@ -869,10 +869,9 @@ def _rescaleClusterPeriods(self, clusterOrder, clusterPeriods, extremeClusterIdx
                 )
 
                 # reset values higher than the upper sacle or less than zero
-                typicalPeriods[column][typicalPeriods[column] > scale_ub] = scale_ub
-                typicalPeriods[column][typicalPeriods[column] < 0.0] = 0.0
+                typicalPeriods[column].clip(lower=0, upper=scale_ub, inplace=True)
 
-                typicalPeriods[column] = typicalPeriods[column].fillna(0.0)
+                typicalPeriods[column].fillna(0.0, inplace=True)
 
                 # calc new sum and new diff to orig data
                 sum_clu_wo_peak = sum(

From 42fef503753c84e932ba2d66d6f31fccc76a1499 Mon Sep 17 00:00:00 2001
From: Fabian Neumann <fabian.neumann@outlook.de>
Date: Mon, 21 Aug 2023 14:16:10 +0200
Subject: [PATCH 3/4] use vectorized np.sum() in _rescaleClusterPeriods

---
 tsam/timeseriesaggregation.py | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/tsam/timeseriesaggregation.py b/tsam/timeseriesaggregation.py
index f3349c3..5c75a22 100644
--- a/tsam/timeseriesaggregation.py
+++ b/tsam/timeseriesaggregation.py
@@ -835,11 +835,11 @@ def _rescaleClusterPeriods(self, clusterOrder, clusterPeriods, extremeClusterIdx
         for column in self.timeSeries.columns:
             diff = 1
             sum_raw = self.normalizedPeriodlyProfiles[column].sum().sum()
-            sum_peak = sum(
+            sum_peak = np.sum(
                 weightingVec[extremeClusterIdx]
                 * typicalPeriods[column].loc[extremeClusterIdx, :].sum(axis=1)
             )
-            sum_clu_wo_peak = sum(
+            sum_clu_wo_peak = np.sum(
                 weightingVec[idx_wo_peak]
                 * typicalPeriods[column].loc[idx_wo_peak, :].sum(axis=1)
             )
@@ -874,7 +874,7 @@ def _rescaleClusterPeriods(self, clusterOrder, clusterPeriods, extremeClusterIdx
                 typicalPeriods[column].fillna(0.0, inplace=True)
 
                 # calc new sum and new diff to orig data
-                sum_clu_wo_peak = sum(
+                sum_clu_wo_peak = np.sum(
                     weightingVec[idx_wo_peak]
                     * typicalPeriods[column].loc[idx_wo_peak, :].sum(axis=1)
                 )

From 8cf3e65efa1ae0252e410da1f29a9dc7cedba2b0 Mon Sep 17 00:00:00 2001
From: Fabian Neumann <fabian.neumann@outlook.de>
Date: Mon, 21 Aug 2023 14:16:45 +0200
Subject: [PATCH 4/4] improve performance of typicalPeriods creation in
 _rescaleClusterPeriods

---
 tsam/timeseriesaggregation.py | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/tsam/timeseriesaggregation.py b/tsam/timeseriesaggregation.py
index 5c75a22..9687fc2 100644
--- a/tsam/timeseriesaggregation.py
+++ b/tsam/timeseriesaggregation.py
@@ -828,9 +828,10 @@ def _rescaleClusterPeriods(self, clusterOrder, clusterPeriods, extremeClusterIdx
         series, without changing the values of the extremePeriods.
         """
         weightingVec = pd.Series(self._clusterPeriodNoOccur).values
-        typicalPeriods = pd.DataFrame(
-            clusterPeriods, columns=self.normalizedPeriodlyProfiles.columns
-        )
+        typicalPeriods = pd.concat([
+            pd.Series(s, index=self.normalizedPeriodlyProfiles.columns)
+            for s in self.clusterPeriods
+        ], axis=1).T
         idx_wo_peak = np.delete(typicalPeriods.index, extremeClusterIdx)
         for column in self.timeSeries.columns:
             diff = 1