Detailed operational building data for six office rooms in Denmark: Occupancy, indoor environment, heating, ventilation, lighting and room control monitoring with sub-hourly temporal resolution

The operational building data presented in this paper has been collected from six office rooms located in an office building (research and educational purposes) located on the main campus of Aalborg University in Denmark. The dataset consists of measurements of occupancy, indoor environmental quality, room-level and system-level heating, ventilation and lighting operation at a 5 min resolution. The indoor environmental quality and building system data were collected from the building management system. The occupancy level in each monitored room is established from the computer vision-based analysis of wall-mounted camera footage of each office. The number of people present in the room is estimated using the YOLOv5s image recognition algorithm. The present dataset can be used for occupancy analysis, indoor environmental quality investigations, machine learning, and model predictive control.

Dataset link: A Danish high-resolution dataset for six office rooms with occupancy, indoor environment , heating, ventilation, lighting and room control monitoring (Original data) a b s t r a c t The operational building data presented in this paper has been collected from six office rooms located in an office building (research and educational purposes) located on the main campus of Aalborg University in Denmark.The dataset consists of measurements of occupancy, indoor environmental quality, room-level and system-level heating, ventilation and lighting operation at a 5 min resolution.The indoor environmental quality and building system data were collected from the building management system.The occupancy

Value of the Data
• The dataset has a high time resolution of five minutes and spans almost a full year (including three different seasons, with both heating and cooling periods), which is currently rare in the building sector.• The dataset covers most room-level control and indoor environmental variables typically found in building management systems (BMS) of office buildings, along with all control and measurement values for the connected heating, ventilation, and air conditioning (HVAC) central systems connected to the six monitored rooms.Room-level artificial lighting activation and presence detection from passive infrared (PIR) sensors are also included.
• The ground truth on the occupancy of the rooms (number of people present in each room at a given time) is accurately established from computer vision-based analysis of camera footage from each monitored room.This information is very rarely present in building datasets.• Besides the dataset, a detailed description of each room and the building systems is provided, thus leaving no missing information for most building applications.• Researchers focusing on occupant detection through BMS data, building indoor environmental analysis, air handling unit (AHU) performance analysis, and model predictive control (MPC) could benefit from this dataset due to its high resolution and completeness.

Data Description
The dataset is comprised of one full dataset (dataset__2023_02_27__2023_12_31) and four subsets covering a winter case (dataset__2023_03_08__2023_03_21), a winter/transition period case (dataset__2023_04_01__2023_04_13), a summer case (dataset__2023_06_01__2023_07_05) and a summer/transition period case (dataset__2023_09_02__2023_10_04).
Only the full dataset is presented here.The datasets are available in two file types: either .xlsxor .csv.All .xlsxfiles contain a metadata sheet with all the data variables' descriptions and the corresponding number of missing data points in each file.The .csv files only contain the dataset.The .csv file uses a semicolon as separator between columns (variables) and a period as decimal separator.Missing values are indicated by #N/A.Both file types contain a starting index in the first column, which can be used to easily find where the subsets are positioned in the full dataset.The first row in both file types is the header, with the naming of each variable explained in the following subsection.

Dataset__2023_02_27__2023_12_31
This dataset [1] comprises the following parts: -Room-level indoor environmental quality, presence detection from PIR sensor, and occupancy measurements, along with artificial lighting, radiator valve, and variable air volume (VAV) damper operational data in six different office rooms.
• These always start with the label "RoomX:" where X is a letter from A to F. -Measurements of the central AHU connected to the six office rooms.
• These always start with the label "Ventilation:" • One should note that the AHU supplies more than just the six rooms of this dataset.-Measurements of the central heating system supplying the radiator to the six office rooms.
• These always start with the label "Heating:" • One should note that the central heating system supplies more than just the six rooms of this dataset.-Measurements of the outdoor conditions.
• These always start with the label "Outdoor:" The timestamp in the file is in the format "YYYY-MM-DDThh:mm:ss + hhmm" according to ISO 8601 and is showing the local Danish time (time zone Europe/Copenhagen), which in standard time is UTC + 1 (CET) and daylight-saving time is UTC + 2 (CEST).Transitions between standard time and daylight-saving time were on March 26th, 2023, at 02:00 (CET) and back to standard time on October 29th, 2023, at 03:00 (CEST).
Tables 2-11 contain all the variables for each room, system and outdoor condition.The "Limits on operating range" column indicates any natural limits for the respective variables.The following options can be found: -"-" means it is unrestricted.
-"0-" means it cannot be lower than 0. -"0/1" means it is a Boolean, such as on or off.-"0-1" means an averaged Boolean value, such as on or off, but it can be any decimal value between 0 and 1 due to data treatment and averaging, thus indicating the share of state 1 between the current and previous timestamp.-"0-100" means it can be any integer between 0 and 100, but due to data treatment and averaging, it can become any decimal value between 0 and 100.-"1/10/14" means that only these specific values can occur, but due to data treatment and averaging, decimal values other than these can occur when transitioning between the values.

Room-Level Measurements
An overview of the room-level measurement variables available for each room, along with a short description of the meaning of each variable, can be found in Table 1 .The distribution overview of all the different room measurements can be seen in Fig. 1 .Some room measurement visualizations in Figs.2-7 were generated using the Python script in Ref. [1] .VAV damper position overview in the six rooms over the entire monitoring period.0% opening means that the damper is at minimum presetting of the damper; see Table 1 for more information (white indicates missing data).

Occupancy Measurements
The real-time occupancy for each room is the current number of occupants detected in that room at a given time.The six rooms have the following number of desks (potential fixed working spaces for occupants): -Room A: 5 desks -Room B: 4 desks -Room C: 3 desks -Room D: 6 desks -Room E: 4 desks -Room F: 4 desks The actual number of occupants over the measurement period can be seen in Fig. 8 , while the periods where the occupancy data is available are shown in Fig. 9 .

Air Handling Unit measurements
An overview of the variables can be seen in Table 9 .

Heating System Measurements
An overview of the variables can be seen in Table 10 .

Outdoor Measurements
An overview and visualization of the variables can be seen in Fig. 10 and Table 11 .

Experimental Design, Materials and Methods
A general description of the case study building can be found in Johra, 2023 [2] .All the data, except for the occupancy, was collected from the BMS and either resampled or realigned to a 5 min resolution.For the BMS data, all measurements from the rooms originally had a sampling rate of five minutes and were therefore only realigned by shifting the logged timestamp to the aligned timestamp, this was done as the logged timestamp was within 1 min of the aligned timestamp, thus it was deemed close enough.The HVAC and outdoor measurements (except for the outdoor temperature) originally had a sampling rate of one minute and, therefore, were downsampled to 5 min resolution using the mean value.This process will remove extremes to some degree, but as the fluctuation in most cases is low, it will only have a minor impact.Any missing data has been labelled as #N/A, as no imputation of data was performed."Outdoor:Temperature_air" was linearly interpolated from 15 min values to 5 min values.In the case when one of the 15 min values was missing, no interpolation was done between this point and its neighboring datapoints.
The supply airflow and pressure across the supply fan were recalculated in the period between 2023-02-27T0 0:0 0:0 0 + 010 0 and 2023-06-13T08:0 0:0 0 + 020 0, as an improper connection in the sensor was found to cause too high pressure difference and thus too high flow measurements (details of the correction can be seen in Table 9 ).
The number of occupants in each room at a given time was determined by analyzing the footage of a wall-mounted camera installed in each monitored room with a computer visionbased algorithm that accurately detects humans.An image was taken every one minute during extended work hours (07:00 -18:00 during standard time (before 2023-03-26T02:0 0:0 0 + 010 0 and after 2023-10-29T02:0 0:0 0 + 010 0), otherwise 08:0 0 -19:0 0 during daylight savings time).Outside of the work hours the pictures were only taken every five minutes.All the images were processed using a pre-trained YOLOv5s algorithm with default settings [3] to identify the number of occupants in each image.The accuracy of the prediciton model is discussed in the Limitations section.The images were aligned with the BMS data by using the image with the closest timestamp for each BMS data point.If no images were found within ± 10 min of the BMS datapoint, the image was regarded as missing, and an #N/A value was recorded for the occupancy level measurement of this data point.
All rooms are equipped with VAV dampers for the ventilation distribution system, along with radiators for the heating system.The overall schematic of the rooms can be seen in Figs.11 , 12 , 13 .The rooms all have balanced ventilation, with the VAV dampers in each room being controlled by the same signal for supply and extraction.It is important to note that 0% opening of the dampers corresponds to roughly 30% of the maximum airflow to ensure the base minimum ventilation rate.The measured relationship between damper opening and airflow rate for each room can be seen in Table 1 .When a room window is opened, the VAV dampers and the radiator control valve are turned to 0%.The dampers and heating system in each room are controlled with a PI controller according to a temperature setpoint with a deadband.This deadband varies depending on the room and time of day.It can be found directly for rooms C-F while it must be calculated from the heating and cooling limits for rooms A-B.An illustration of the relationship between control variables used for the heating and cooling control can be seen in Fig. 14 .The opening of the damper depends on the highest value of the opening signal for temperature, and CO 2 setpoint.The characteristic of the temperature and CO 2 opening signals can be seen in Fig. 15 .
The AHU is a VAV unit with a rotating wheel heat recovery unit and a water-based heating coil supplied by district heating (DH).During normal operation, the AHU is controlled to maintain a specific temperature setpoint for the supply air, and a specific air pressure setpoint in both the supply and extraction ducts.All controllers are PI-controllers with a small deadband of either 0.1 °C or 1 Pa.The AHU supplies air to one seminar room, two meeting rooms, 21 offices, six toilets, eight auxiliary rooms, and hallways/open areas on three floors.A schematic of the AHU and the locations of the measurement points can be seen in Fig. 16 .The heating system provides heat for roughly half of the building.It is a direct DH-based system with a mixing control, meaning that depending on the heating need of the building, the return water from the building will be recirculated and mixed with the DH water supply to ensure that the temperature setpoint of the supply to the building is met.The mixing ratio is controlled by the control valve located on the DH return, as the pump is only controlled with an on/off controller.A schematic of the heating system can be seen in Fig. 17 .
The outdoor measurements come from sensors located on the building's rooftop.The sensors used in the rooms and systems, along with their accuracy has been summarized in Table 12 .

Limitations
Due to practical limitations on the possible locations of the cameras in each room, some areas were difficult to detect people in, as the image did not capture the entire person.To mitigate this issue, the cameras were placed to cover all the fixed working stations and the entrance door of the rooms.Manual performance verification of the people detection algorithm (checking randomly picked-up images and labeled occupancy) shows that the number of people in the room is correctly determined in more than 99% of the cases (26 false occupancy assessments out of 5188 image samples).On rare occasions, people outside the room were also detected when passing in front of the open door or visible through the glazed surface next to the door.

Ethics Statement
To conduct the experiments for the generation of the present dataset, the appropriate administrative body (Aalborg University -AAU Innovation -Grants & Contracts) has been contacted in order to verify the ethical soundness of the experiment and the necessary measures that had to be taken regarding the General Data Protection Regulation (GDPR).After informing this administrative body (Aalborg University -AAU Innovation -Grants & Contracts), all the participants in the experiment have been informed about the use of the collected data and a GDPR consent form has been sent to them.The authors hereby confirm that the relevant, informed consent was obtained from all subjects who have participated in the generation of that dataset.A copy of the original consent form can be found in the appendix.Copies of the signed informed consent are retained by the authors.No additional approval from institutional review boards or local ethics committees was necessary to conduct this experiment.
The current dataset and the present dataset description do not comprise any personal or specific information, which could lead to the identification of the subjects who have participated in the generation of that dataset.

Fig. 2 .
Fig. 2. CO 2 concentration overview in the six rooms over the entire monitoring period (white indicates missing data).

Fig. 3 .
Fig. 3. Room temperature overview in the six rooms over the entire monitoring period (white indicates missing data).

Fig. 4 .
Fig. 4. Room presence detection (PIR sensor) overview in the six rooms over the entire monitoring period (white indicates missing data).

Fig. 5.
Fig.5.VAV damper position overview in the six rooms over the entire monitoring period.0% opening means that the damper is at minimum presetting of the damper; see Table1for more information (white indicates missing data).

Fig. 6 .
Fig. 6.Radiator valve position overview in the six rooms over the entire monitoring period (white indicates missing data).

Fig. 7 .
Fig. 7. Window opening overview in the six rooms over the entire monitoring period (white indicates missing data).

Fig. 8 .Fig. 9 .
Fig. 8. Occupancy overview in the six rooms over the entire monitoring period (white indicates missing data).

Fig. 10 .
Fig. 10.Outdoor conditions overview over the entire monitoring period (white indicates missing data).

Fig. 11 .
Fig. 11.Schematic of six office rooms (with geometry, window types, wall types, building systems and sensors).All dimensions are in mm.

Fig. 12 .
Fig. 12. Schematic of the wall and window types.All dimensions are in mm.

Fig. 13 .
Fig. 13.3D visualization of the rooms including the position of the fixed working spaces (office chairs) and the location of the temperature and CO 2 sensor in each room (red square near the door).

Fig. 14 .
Fig. 14.Heating and cooling control states and limits for the rooms.

Fig. 15 .
Fig. 15.Temperature and CO 2 setpoint curve for the VAV damper.A damper opening of 0% corresponds to an airflow of 30% of the maximum airflow.

Fig. 16 .
Fig. 16.Schematic for the AHU, along with an overview of the sensors providing data to the BMS.

Fig. 17 .
Fig.17.Schematic for the heating system, along with an overview of the sensors providing data to the BMS.

Table 1
General overview of data available for the rooms: X indicates that the measurement is available in the room.
If 0, all windows are closed.If 1, at least one window is open.(if the window is open, the motor valves for the radiators and the VAV dampers are closed by the

Table 2
Room A measurement variables.Room B measurement variables.

Table 8
Occupancy variables for the six rooms.

Table 11
Outdoor measurement variables.