Building Energy Consumption 2

The dataset consists of synthetic building-design configurations and their corresponding heating energy demand, generated using a first-principles building-energy simulation workflow. The simulation systematically varies geometric properties, envelope thermal characteristics, air permeability, glazing ratios, and internal gains to produce physically consistent building designs and their resulting heating loads. Two versions of the dataset are provided: a box-shape dataset used for causal structure discovery and a random-shape dataset used for causal effect estimation. Additionally, counterfactual datasets are generated by intervening on the treatment variable (building height) while either preserving causal dependencies among related design parameters (CATE dataset) or holding all other parameters fixed (ATE dataset). These datasets enable evaluation of causal inference methods and demonstrate how ignoring causal relationships between building parameters can lead to biased estimates of the effect of design changes on energy performance. The target variable is the annual heating load representing the building’s heating demand under simulated operating conditions.

Task: The dataset can be used to study causal discovery methods.

Summary: 

Size of collection: 6 datasets with dimensions 1000 x 41

Task: Causal Inference Problem

Data Type: Mixed Data

Dataset Scope: Collection of Datasets

Ground Truth: Known Graph

Temporal Structure: Static Data

License: MIT

Missing Values: No Missing Values

Features:

File: File name

Floor Area: Total usable floor space of the building across all floors [m²].

Height: Room height: Vertical distance from floor to ceiling within a room [m].

Volume: Total enclosed space inside the building [m³], calculated from floor area and height.

Number of Floors: Total count of building levels above ground.

External Wall Area: Total surface area of walls exposed to outdoor conditions [m²].

Ground Floor Area: Area of the building footprint at ground level [m²].

Roof Area: Total area of the building roof exposed to external conditions [m²].

Window Area: Total area of all windows in the building envelope [m²].

Total Heat Capacity: Total ability of the building materials to store thermal energy.

u-Value (Wall): Thermal transmittance. Rate of heat transfer through the wall. Lower values mean better insulation.

u-Value (Ground Floor): Thermal transmittance. Rate of heat transfer through the ground floow. 

u-Value (Roof): Thermal transmittance. Rate of heat transfer through the roof. 

u-Value (Internal Floor): Thermal transmittance. Rate of heat transfer through the internal floor. 

u-Value (Internal Wall): Thermal transmittance. Rate of heat transfer through the internal wall. 

u-Value (Windows): Thermal transmittance. Rate of heat transfer through the windows. 

g-Value (Windows): Solar heat gain coefficient; fraction of solar radiation entering through windows (ranges 0–1).

Heat Capacity (Floor Slabs): Ability of floor slabs to store heat, influencing thermal stability.

Infiltration: Uncontrolled air leakage into or out of the building through cracks and openings.

Permeability: Measure of how easily air passes through the building envelope [m³/m²·h].

Internal Mass: Amount of material inside the building (walls, floors, furniture) that stores heat.

WWR (North): Ratio of window area to wall area on the north façade.

WWR (East): Ratio of window area to wall area on the east façade.

WWR (West): Ratio of window area to wall area on the west façade.

WWR (South): Ratio of window area to wall area on the south façade.

Building:Start Time: Time when building systems begin operation each day.

Building:Operating Hours: Duration the building is occupied and systems are active per day.

Building:Light Heat Gain: Heat generated by lighting systems [W/m²].

Building:Equipment Heat Gain: Heat generated by electrical equipment [W/m²].

Building:Occupancy: Number of occupants per unit area [persons/m²].

Building:Heating Setpoint: Desired indoor temperature maintained by heating systems [°C].

Building:Cooling Setpoint: Desired indoor temperature maintained by cooling systems [°C].

Boiler Efficiency: Percentage of fuel energy converted into useful heating energy.

Heating COP: Efficiency of heating system; ratio of heat output to energy input.

Cooling COP: Efficiency of heating system; ratio of cooling output to energy input.

Heating Load: Amount of energy required to heat the building [kWh/year].

Cooling Load: Amount of energy required to cool the building [kWh/year].

Lights Load: Energy consumption due to lighting [kWh].

Thermal Energy: Total energy used for heating and cooling.

Operational Energy: Total energy consumed during building operation (HVAC, lighting, equipment).

EUI: Energy Use Intensity.

Files: 

Building_RandomShape.csv: Simulated dataset. 

Building_BoxShape.csv: Simulated dataset of box-shaped buildings. 

CATE_Building_3.csv:  Simulated interventional dataset with `Height==3`: Height = 3 for all rows; Other variables = sampled randomly from realistic distribution; Heating_Load to predict

CATE_Building_3.2.csv:  Simulated interventional dataset with `Height==3.2`. It is identical to CATE_Building.csv except in the column `Height`. 

ATE_Building_3.csv:  Simulated interventional dataset with `Height==3`: Height = 3 for all rows; Ground Floor Area = 300; Number of Floors = 3; WWR = 0.3; u-values fixed; Permeability fixed; Equipment Heat Gain fixed; Occupancy fixed; Other variables = sampled randomly from realistic distribution; Heating_Load to predict

CATE_Building_3.2.csv:  Simulated interventional dataset with `Height==3.2`. It is identical to ATE_Building.csv except in the column `Height`.

causal_graph.txt: Estimated (Greedy-Equivalent-Search) and subsequently pruned (Expert knowledge) causal graph. Compatible with the Python package `DoWhy`.