# Global AI Challenge for Building E&M Facilities 25 — Data

来源：https://www.kaggle.com/competitions/global-ai-challenge-trial-2025/data

采集时间：2026-10-20T00:53:25Z

The dataset consists of historical cooling load of buildings from chilled water systems deployed in a building and related chiller operational information. Various types of sensor readings with short reading intervals are also available for use. You may also use external open data (e.g. Hong Kong Observatory Open Data, The Building Data Genome Project 2 etc.) to build your own model.


*   **Building_X.csv** - The training data for model development. It contains chiller sensor data from one building that contains year-round cooling data from 2023, recorded at 15-mins intervals in Hong Kong Time in csv format.

   **Data Used for Building X**
   Date & Time Range: 1/1/2023 00:00 to 31/12/2023 23:00;
   
     **Data Provided**
| Raw Chiller Sensor Data | Description | Unit |
| :--- | :---: | :---: |
|`record_timestamp`	|Date and Time of the record in Hong Kong Time	|YYYY-MM-DD HH:MM:SS|
|`CHR-01-KW`	|Chiller 1 Power Consumption	|kW|
|`CHR-01-CHWSWT`	|Chiller 1 Chilled Water Supply Water Temperature	|°C|
|`CHR-01-CHWRWT`	|Chiller 1 Chilled Water Return Water Temperature	|°C|
|`CHR-01-CHWFWR`	|Chiller 1 Chilled Water Flowrate	|L/s|
|`CHR-02-KW`	|Chiller 2 Power Consumption	|kW|
|`CHR-02-CHWSWT`	|Chiller 2 Chilled Water Supply Water Temperature	|°C|
|`CHR-02-CHWRWT` |Chiller 2 Chilled Water Return Water Temperature	|°C|
|`CHR-02-CHWFWR` |Chiller 2 Chilled Water Flowrate	|L/s|
|`CHR-03-KW` |Chiller 3 Power Consumption	|kW|
|`CHR-03-CHWSWT`	|Chiller 3 Chilled Water Supply Water Temperature	|°C|
|`CHR-03-CHWRWT`	|Chiller 3 Chilled Water Return Water Temperature	|°C|
|`CHR-03-CHWFWR`	|Chiller 3 Chilled Water Flowrate	|L/s|




*   **test.csv** - This is the competition file required for prediction. Participants need to predict the total hourly cooling load for the next 30 days.
The file contains hourly timestamps for the 30-day prediction period, and each timestamp serves as the unique identifier for your prediction target. 
Predictions must be organized according to the format provided in test.csv.

      **Prediction Range**
Building: X; Date & Time Range: 1/1/2024 00:00 to 30/1/2024 23:00;

      **Submission Format**
Submit a CSV file with 2 columns:
`predicted_time`: hourly timestamps from 1/1/2024 00:00 to 30/1/2024 23:00
`predicted_load`: Predicted total cooling load in **kW** for that hour

      **Example Submission**
```
prediction_time,predicted_load
1/1/2024 0:00:00 am,100
1/1/2024 1:00:00 am,200
1/1/2024 2:00:00 am,300
1/1/2024 3:00:00 am,150
1/1/2024 4:00:00 am,50
etc.
```