Building Energy System¶

Optimal combination of segments and periods for building energy supply systems.

Author: Leander Kotzur

Import pandas and the relevant time series aggregation class

In [1]:

%load_ext autoreload
%autoreload 2

import pandas as pd
import plotly.express as px
import plotly.io as pio

import tsam
from tsam import ClusterConfig, SegmentConfig
from tsam.tuning import find_pareto_front

pio.renderers.default = "notebook_connected"
import warnings

# Added to every example notebook: silence the v3 column-order
# FutureWarning in the rendered docs (tsam v4 returns result columns in
# input order; see migration guide).
warnings.filterwarnings(
    "ignore", category=FutureWarning, message=".*sorted alphabetically.*"
)

%load_ext autoreload %autoreload 2 import pandas as pd import plotly.express as px import plotly.io as pio import tsam from tsam import ClusterConfig, SegmentConfig from tsam.tuning import find_pareto_front pio.renderers.default = "notebook_connected" import warnings # Added to every example notebook: silence the v3 column-order # FutureWarning in the rendered docs (tsam v4 returns result columns in # input order; see migration guide). warnings.filterwarnings( "ignore", category=FutureWarning, message=".*sorted alphabetically.*" )

Input data¶

Read in time series from testdata.csv with pandas

In [2]:

raw = pd.read_csv("testdata.csv", index_col=0)
raw = raw.rename(
    columns={
        "T": "Temperature [°C]",
        "Load": "Load [kW]",
        "Wind": "Wind [m/s]",
        "GHI": "Solar [W/m²]",
    }
)

raw = pd.read_csv("testdata.csv", index_col=0) raw = raw.rename( columns={ "T": "Temperature [°C]", "Load": "Load [kW]", "Wind": "Wind [m/s]", "GHI": "Solar [W/m²]", } )

In [3]:

raw = raw.drop(
    columns=[
        "Wind [m/s]",
    ],
)

raw = raw.drop( columns=[ "Wind [m/s]", ], )

Use tsam's built-in heatmap plotting for visual comparison of the time series

Plot an example series - in this case the temperature

In [4]:

# Original data heatmaps using tsam.unstack_to_periods() with plotly
unstacked = tsam.unstack_to_periods(raw, period_duration=24)
for col in raw.columns:
    px.imshow(
        unstacked[col].values.T,
        labels={"x": "Day", "y": "Hour", "color": col},
        title=f"Original {col}",
        aspect="auto",
    ).show()

# Original data heatmaps using tsam.unstack_to_periods() with plotly unstacked = tsam.unstack_to_periods(raw, period_duration=24) for col in raw.columns: px.imshow( unstacked[col].values.T, labels={"x": "Day", "y": "Hour", "color": col}, title=f"Original {col}", aspect="auto", ).show()

Tune a hierarchical aggregation with segments in combination with distribution representation¶

Use the new find_pareto_front() function to explore the Pareto-optimal combinations.

In [5]:

pareto_results = find_pareto_front(
    raw,
    period_duration=24,
    max_timesteps=100,
    cluster=ClusterConfig(
        method="hierarchical",
        representation="distribution",
    ),
    n_jobs=-1,
)

pareto_results = find_pareto_front( raw, period_duration=24, max_timesteps=100, cluster=ClusterConfig( method="hierarchical", representation="distribution", ), n_jobs=-1, )

Building Pareto front:   0%|          | 0/100 [00:00<?, ?it/s]

Building Pareto front:   2%|▏         | 2/100 [00:00<00:21,  4.63it/s]

Building Pareto front:   3%|▎         | 3/100 [00:00<00:20,  4.80it/s]

Building Pareto front:   6%|▌         | 6/100 [00:00<00:11,  8.45it/s]

Building Pareto front:   9%|▉         | 9/100 [00:01<00:08, 10.51it/s]

Building Pareto front:  12%|█▏        | 12/100 [00:01<00:07, 11.76it/s]

Building Pareto front:  15%|█▌        | 15/100 [00:01<00:06, 12.50it/s]

Building Pareto front:  18%|█▊        | 18/100 [00:01<00:06, 12.87it/s]

Building Pareto front:  21%|██        | 21/100 [00:01<00:06, 12.93it/s]

Building Pareto front:  24%|██▍       | 24/100 [00:02<00:05, 12.91it/s]

Building Pareto front:  27%|██▋       | 27/100 [00:02<00:05, 12.78it/s]

Building Pareto front:  30%|███       | 30/100 [00:02<00:05, 12.59it/s]

Building Pareto front:  33%|███▎      | 33/100 [00:02<00:05, 12.43it/s]

Building Pareto front:  36%|███▌      | 36/100 [00:03<00:05, 12.21it/s]

Building Pareto front:  48%|████▊     | 48/100 [00:03<00:02, 22.91it/s]

Building Pareto front:  52%|█████▏    | 52/100 [00:03<00:02, 20.56it/s]

Building Pareto front:  56%|█████▌    | 56/100 [00:03<00:02, 18.71it/s]

Building Pareto front:  70%|███████   | 70/100 [00:04<00:01, 28.82it/s]

Building Pareto front:  75%|███████▌  | 75/100 [00:04<00:00, 25.42it/s]

Building Pareto front:  80%|████████  | 80/100 [00:04<00:00, 22.95it/s]

Building Pareto front:  85%|████████▌ | 85/100 [00:05<00:00, 21.11it/s]

Building Pareto front:  90%|█████████ | 90/100 [00:05<00:00, 17.40it/s]

Building Pareto front: 100%|██████████| 100/100 [00:05<00:00, 18.21it/s]

And determine the pareto optimal aggregation up to 100 total time steps. This may take some time...

In [6]:

# Show the last result in the Pareto front
last_result = pareto_results[-1]
print(
    f"Final configuration: {last_result.n_clusters} periods, {last_result.n_segments} segments"
)

# Show the last result in the Pareto front last_result = pareto_results[-1] print( f"Final configuration: {last_result.n_clusters} periods, {last_result.n_segments} segments" )

Final configuration: 20 periods, 5 segments

And show the results for the last aggregation

In [7]:

# Reconstruct the data from the last Pareto result
reconstructed = last_result.reconstructed

# Reconstruct the data from the last Pareto result reconstructed = last_result.reconstructed

In [8]:

# Reconstructed data heatmaps from last tuned aggregation
unstacked_recon = tsam.unstack_to_periods(reconstructed, period_duration=24)
for col in reconstructed.columns:
    px.imshow(
        unstacked_recon[col].values.T,
        labels={"x": "Day", "y": "Hour", "color": col},
        title=f"Reconstructed {col}",
        aspect="auto",
    ).show()

# Reconstructed data heatmaps from last tuned aggregation unstacked_recon = tsam.unstack_to_periods(reconstructed, period_duration=24) for col in reconstructed.columns: px.imshow( unstacked_recon[col].values.T, labels={"x": "Day", "y": "Hour", "color": col}, title=f"Reconstructed {col}", aspect="auto", ).show()

In [9]:

last_result.n_segments

last_result.n_segments

Out[9]:

5

In [10]:

last_result.n_clusters

last_result.n_clusters

Out[10]:

20

In [11]:

# Example with specific configuration using distribution_minmax representation
result = tsam.aggregate(
    raw,
    n_clusters=14,
    period_duration=24,
    cluster=ClusterConfig(
        method="hierarchical",
        representation="distribution_minmax",
    ),
    segments=SegmentConfig(n_segments=8),
    preserve_column_means=False,
)

# Example with specific configuration using distribution_minmax representation result = tsam.aggregate( raw, n_clusters=14, period_duration=24, cluster=ClusterConfig( method="hierarchical", representation="distribution_minmax", ), segments=SegmentConfig(n_segments=8), preserve_column_means=False, )

In [12]:

# Reconstructed data heatmaps with 8 segments and 14 periods
recon = result.reconstructed
unstacked_recon2 = tsam.unstack_to_periods(recon, period_duration=24)
for col in recon.columns:
    px.imshow(
        unstacked_recon2[col].values.T,
        labels={"x": "Day", "y": "Hour", "color": col},
        title=f"Reconstructed {col} (8 seg, 14 periods)",
        aspect="auto",
    ).show()

# Reconstructed data heatmaps with 8 segments and 14 periods recon = result.reconstructed unstacked_recon2 = tsam.unstack_to_periods(recon, period_duration=24) for col in recon.columns: px.imshow( unstacked_recon2[col].values.T, labels={"x": "Day", "y": "Hour", "color": col}, title=f"Reconstructed {col} (8 seg, 14 periods)", aspect="auto", ).show()