Data Set for Probabilistic Indoor Temperature Forecasting

1. Dataset Manifest

This text provides a description of the dataset used for model training and evaluation in our study "A Tutorial on Deep Learning for Probabilistic Indoor Temperature Forecasting". The dataset consists of various simulated thermal and environmental parameters for different room configurations. Below, you will find a table detailing each column in the dataset along with its description and unit of measurement. 

1.1. Columns Description

Column Name
Description
Unit

time
Time stamp of the measurement
-

ZweiPersonenBuero.TAir
Air temperature inside a two-person office
°C

heatStat.Heat.Q_flow
Heating rate in the room
W

weaDat.AirPressure
Atmospheric pressure
Pa

weaDat.AirTemp
Outside air temperature
°C

weaDat.SkyRadiation
Longwave sky radiation
W/m²

weaDat.TerrestrialRadiation
Terrestrial radiation
W/m²

weaDat.WaterInAir
Absolute humidity
g/kg

VAir
Air volume in the room
m³

AExt0
Exterior wall area facing the south
m²

AExt1
Exterior wall area facing the north
m²

AInt
Total interior wall area
m²

AFloor
Floor area of the room
m²

AWin0
Window area facing the south
m²

AWin1
Window area facing the north
m²

azi0
Azimuth (direction) of the first exterior wall
rad

azi1
Azimuth (direction) of the second exterior wall
rad

id
Unique identifier for the room configuration
-

is_holiday
Indicator whether the day is a holiday (1 for yes, 0 for no)
-

1.2. Note on Multi-Value Columns

For rooms with multiple exterior walls (rooms 15-30):

AExt: {Exterior wall 1 area, Exterior wall 2 area}

AWin: {Window area on exterior wall 1, Window area on exterior wall 2}

azi: {Azimuth of exterior wall 1, Azimuth of exterior wall 2}

Example:

AExt = {10, 15}

AWin = {2, 0}

azi = {0, 3.1415}

This indicates two exterior walls with areas of 10 m² and 15 m² facing south (0 rad) and north (3.1415 rad), respectively. The south-facing wall has a window of 2 m², while the north-facing wall has no window.

1.3. Data Sources

Room Model: Simulated using the reduced-order package of the Modelica Buildings Library.

Weather Data: Provided by the German Meteorological Service (DWD) in Test Reference Year (TRY) format.

This comprehensive dataset provides crucial parameters required to train and evaluate thermal models for different room configurations. The simulation data ensures a diverse range of environmental and occupancy conditions, enhancing the robustness of the models.

1.4. Data scaling

The data set contains the raw data as well as the scaled data used for training and testing the model. The scaling was carried out using the StandardScaler package.

1.5. Weather data license

This data set contains weather data recorded by the DWD under license „Datenlizenz Deutschland – Namensnennung – Version 2.0" (URL). The data is provided by "Bundesinstitut für Bau-, Stadt- und Raumforschung". The data can be downloaded from here. We use data from the year 2015 from Heilbronn. We have added the weather data to the data set unchanged.