Data-driven occupancy profile identification and application to the ventilation schedule in a school building

Citation: Lastname, F.; Lastname, F.; Lastname, F. Title. Energies 2024,1, 0. https://doi.org/ Received: Revised: Accepted: Published: Copyright: © 2025 by the authors. Submitted to Energies for possible open access publication under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/). Article Data-driven occupancy profile identification and application to ventilation schedule in a school building Kristina Vassiljeva 1,2∗ , Margarita Matson 1,3 , Andrea Ferrantelli 1,5,6 , Eduard Petlenkov 1,2, Martin Thalfeldt 1,4, Juri Belikov 1,3 1FinEst Centre for Smart Cities (Finest Centre), Tallinn University of Technology, Tallinn, 19086, Estonia 2Department of Computer Systems, Tallinn University of Technology, Tallinn, 12618, Estonia; [email protected] 3Department of Software Science, Tallinn University of Technology, Tallinn, 12618, Estonia; [email protected] 4Department of Civil Engineering and Architecture, Tallinn University of Technology, Tallinn, 19086, Estonia; [email protected] 5 Department of Mechanical Engineering, Aalto University, Espoo, 00076, Finland; andr[email protected] 6Department of Civil Engineering, Aalto University, Espoo, 00076, Finland; andr[email protected] *Correspondence: [email protected] Abstract: Facing the current sustainability challenges requires reduction of the building stock energy 1 usage towards the European Green Deal targets. This can be accomplished by adopting techniques 2 such as fault detection and diagnosis, and efficiency optimization. Taking an Estonian school as 3 case-study, an occupancy-based algorithm for scheduling ventilation operations in buildings is 4 here developed starting only from energy use data. The aim is optimizing the system’s operation 5 according to occupancy profiles while maintaining a comfortable indoor climate. By relying only 6 on electricity meters without using carbon dioxide or occupancy sensors, we use historical data of 7 a school to develop a DBSCAN-based clustering algorithm that generates consumption profiles. A 8 novel occupancy estimation algorithm, based on threshold and time series methods, then creates 12 9 occupancy schedules that are either based on classical detection with on-off method or on occupancy 10 estimation for demand-controlled ventilation. We find that the latter replaces the 60% capacity of 11 current on-off schedules by 30% or even 0%, with energy savings ranging from 3.5% to 66.4%. The 12 corresponding costs are reduced from 18.1% up to 62.6%, still complying with current national 13 regulations of indoor air quality. Remarkably, our method can immediately be extended to other 14 countries, as it relies only on occupancy schedules that ignore weather and other location-specific 15 factors. 16 Keywords: AHU; HVAC; occupancy; data clustering; DBSCAN; energy efficiency; optimization 17 1. Introduction 18 1.1. Motivation 19 Buildings account for a substantial 40% of global energy consumption and contribute 20 nearly 35% of greenhouse gas emissions [ 1 ]. In the EU, around 35% of buildings are over 21 50 years old, with almost 75% of them lacking energy efficiency [ 2 ]. Therefore, prioritiz22 ing building renovations becomes crucial to improve energy efficiency while maintaining 23 indoor climate quality through the implementation of new control and maintenance tech24 niques. Techniques such as fault detection and diagnosis, efficiency optimization, and 25 promoting sustainable changes in consumer behavior play a significant role in reducing 26 energy use in buildings, aligning with the European Green Deal by 2050 [ 3 ]. Furthermore, 27 post-occupancy evaluations have shown that work productivity strongly correlates with 28 factors like thermal conditions, indoor air quality (IAQ), humidity, and lighting [ 4 ]. Recent 29 findings have revealed a significant "performance gap" between calculated and actual 30 energy use in buildings, often exceeding 300%, primarily due to the neglect of occupants’ 31 Version October 30, 2025 submitted to Energies https://www.mdpi.com/journal/energies Version October 30, 2025 submitted to Energies 2 of 22 behavior in simulations and calculations [ 5 – 7 ]. This highlights the importance of optimiz32 ing the scheduling of air handling units (AHUs) to meet the occupants’ demand for a more 33 comfortable and flexible indoor climate while reducing energy consumption [8]. 34 AHUs, which are key components of heating, ventilation, and air conditioning (HVAC) 35 systems, can account for up to 50% of a building’s total energy consumption [ 9 ]. By adjust36 ing AHU operations to align with occupancy patterns, energy waste can be minimized. This 37 involves setting temperature and ventilation levels based on occupancy and the building’s 38 thermal properties. For example, during winter, scheduling AHUs to lower the temper39 ature in unoccupied areas during off-peak hours and restoring it to a comfortable level 40 before occupants arrive can significantly reduce energy consumption. Similarly, in summer, 41 raising the temperature in unoccupied areas and cooling them before occupants arrive can 42 lead to energy savings. 43 Effective AHU scheduling requires a comprehensive understanding of occupancy 44 patterns, thermal properties, and ventilation systems in the building [ 10 ]. The benefits 45 are multifaceted, as it not only reduces energy consumption and operating costs but also 46 enhances occupant comfort. By matching the HVAC system’s operations to occupancy 47 patterns, occupants can enjoy a comfortable indoor environment while minimizing energy 48 waste. Consistent temperature and ventilation levels in occupied areas can avoid discomfort 49 caused by temperature fluctuations and drafts. Furthermore, optimizing scheduling can 50 extend the lifespan of AHU equipment by reducing its workload, resulting in reduced 51 maintenance costs. 52 Additionally, optimizing AHU scheduling helps buildings comply with energy regula53 tions and avoid penalties or fines. Many countries and jurisdictions have energy efficiency 54 targets for buildings, and by optimizing AHU operations, building owners and managers 55 can meet these requirements while achieving energy savings and sustainability goals. 56 1.2. Background/Existing solutions 57 Connecting AHU scheduling with occupancy estimation enables energy optimiza58 tion, demand-based ventilation, comfort optimization, flexibility, and integration with 59 other smart building systems. Occupancy-centric controls are gaining traction in building 60 automation and energy management systems. Occupancy significantly impacts energy 61 consumption, with variations of 30% to 150% depending on building use and occupant be62 havior. However, many state buildings still rely on centralized systems and static schedules, 63 resulting in energy waste. Personalized occupancy profile models and machine learning 64 algorithms can optimize building system operations based on specific occupant behavior, 65 improving efficiency and adaptation over time [11]. 66 Occupancy can be classified into two categories: occupancy detection and occupancy 67 estimation. Occupancy detection determines whether someone is present in a space, while 68 occupancy estimation aims to estimate the number of occupants. Additionally, considering 69 the occupant activity level is crucial for building energy management systems. This 70 information helps estimate internal gains from occupants, optimizing building systems like 71 AHUs and lighting for energy efficiency and occupant comfort [12]. 72 However, achieving accurate occupancy solutions remains a challenge. Typically, 73 occupancy count estimates are derived by combining data from various sensors, including 74 CO2 , vibration, ultrasonic, infrared, floor pressure, camera-based, and audio-based sensors 75 [ 13 ]. These sensors provide valuable input for occupancy estimation, and the data fusion 76 process enhances the accuracy of the occupancy count. Despite ongoing advancements, 77 achieving a completely error-free solution is still an area of active research. 78 Traditional motion sensors like Passive Infrared (PIR) sensors have limitations in reli79 ably detecting immobile occupants and may require a direct line of sight [ 14 ], [ 15 ]. In large 80 spaces like classrooms, a system of sensors may be needed, leading to higher installation 81 and maintenance costs compared to RF-based systems [ 16 ], [ 17 ]. Wi-Fi-based occupancy 82 detection has gained popularity due to the prevalence of Wi-Fi-enabled devices [ 18 ], [ 19 ] 83 using Device-Free Passive (DfP) detection or Channel State Information (CSI) methods, but 84 Version October 30, 2025 submitted to Energies 3 of 22 careful consideration is necessary for privacy and safety concerns in settings like kinder85 gartens and schools. CO2 sensors combined with building models have been used for 86 occupancy estimation, but their accuracy can be affected by various factors, and accurately 87 determining CO2 production for different age groups is challenging [ 20 – 22 ]. Cameras offer 88 high precision for occupancy estimation, but privacy concerns, computational complexity, 89 and lighting conditions can limit their effectiveness in certain situations [23–25]. 90 Occupancy detection methods vary and are chosen based on building type, privacy, 91 cost, and accuracy needs. Using multiple sensors or hybrid approaches can improve 92 occupancy estimation, aiding building automation and energy management. 93 The advanced metering infrastructure (AMI) is key in analyzing building performance 94 and enhancing occupant service. Studying load consumption profiles helps reduce peak95 hour energy use, predict consumption, and detect anomalies. These profiles reflect electrical 96 load variations over time due to on/off-peak usage and seasonal changes [26]. 97 Electricity consumption data is increasingly used for occupancy detection in buildings 98 due to its cost-effectiveness and insight potential. Analyzing these patterns can identify 99 occupied and unoccupied areas [27–29]. 100 One commonly used technique is the threshold method, where a threshold value is set 101 for the power consumption of a specific area or zone. If the power consumption exceeds 102 the threshold, it is assumed that the area is occupied, and if it falls below the threshold, 103 it is assumed that the area is unoccupied. Determining the threshold value involves 104 analyzing power consumption patterns during known occupancy and non-occupancy 105 periods, calculating the average power consumption for each state, and setting the threshold 106 accordingly [30], [31]. 107 Using electricity data and clustering techniques, we can detect occupancy based 108 on power consumption patterns, offering a cost-effective way to understand occupancy 109 dynamics and optimize operations. Unsupervised data mining or machine learning can 110 identify similar behavioral patterns, providing insights into energy usage patterns and 111 facilitating predictions. 112 When dealing with time series data for energy usage and occupancy, it’s crucial 113 to consider seasonal changes, weather patterns, and external factors like the COVID-19 114 pandemic that can impact energy demand. 115 Various time series analysis methods can be applied to extract features and patterns 116 from individual consumer load data. Techniques such as K-means clustering, Fuzzy C117 means, hierarchical clustering algorithms, Self-organizing Maps (SOM), Support Vector 118 Machines (SVM), and Density-Based Spatial Clustering of Applications with Noise (DB119 SCAN) can help identify patterns and group similar load profiles together [ 32 – 35 ]. These 120 methods enable a more comprehensive understanding of energy consumption behavior, 121 taking into account factors such as hours of operation, weekdays versus weekends, seasonal 122 variations, and holidays [36], [37]. 123 Forecasting future occupancy patterns within a building has become a significant 124 trend, as it can inform advanced AHU controls and resource management strategies like 125 targeted energy management and demand response. Recursive algorithms, such as the 126 one developed by Schwartz et al. for daily occupancy forecasting, are used to predict 127 occupancy patterns based on historical data. These forecasts can assist in optimizing 128 building operations and energy usage [38]. 129 However, existing research in AHU scheduling often relies on CO2 sensors for oc130 cupancy detection, which can be limiting for buildings without such sensors and may 131 not account for the impact of ventilation on CO2 levels. Additionally, some approaches 132 incorporate external features like weather information, which may not be applicable as the 133 heating system is independent of ventilation, and air conditioning is not always present in 134 all buildings. 135 Version October 30, 2025 submitted to Energies 4 of 22 1.3. Problem statement and novelty of the study 136 In this paper, we propose an approach that aims to fully automate the AHU scheduling 137 process by utilizing electricity meters for occupancy detection instead of CO2 sensors, 138 making it applicable to buildings without specialized sensors. Moreover, we eliminate 139 the reliance on external features and focus on computationally efficient algorithms that 140 leverage existing information, such as electricity consumption profiles, to enable informed 141 decision-making for optimizing the AHU system operation. 142 Although the approach of this study is relatively formal, our aim is very practical. 143 For example, if the occupancy schedule indicates that the building (here, the school) is 144 unoccupied during weekends or vacations, the scheduling algorithm can adjust the settings 145 to reduce energy consumption, resulting in energy savings and cost reduction. On the other 146 hand, if the occupancy schedule indicates that the building is occupied during workdays, 147 the AHU scheduling algorithm can ensure that the building is comfortable and adequately 148 ventilated during those times. By effectively managing the AHU schedule based on the 149 occupancy timetable, the proposed method can help improve energy efficiency, comfort, 150 and cost-effectiveness of the building techno-system. Our methodology ensures that all 151 algorithms are computationally efficient as well as based on data that is readily available 152 and easily accessible, as the system needs to be able to recompute for multiple buildings. 153 2. Methods 154 2.1. Proposed solution and algorithms 155 Electricity meters // _ _ _     _ _ _ Step I collect //Historical data  Training data  _ _ _ _       _ _ _ _ Step II process oo _____       _____ Step III clustering //Consumption profiles // ______           ______ Step IV Occupancy detection algorithm ____       ____ Step VI adapt  Occupancy schedule oo _____     _____ Step V Combine ooOccupancy profiles oo AHU schedule Labelled future period OO Figure 1. Main steps of the algorithm The proposed occupancy-based ventilation scheduling method is comprised of multi156 ple steps (illustrated in Figure 1). Historical data is first collected and stored from available 157 electricity meters (Step I, see Section 2.2), then pre-processed and cleaned from outliers 158 (Step II, Section 2.3). A clustering algorithm is then trained using a subset of the historical 159 data (Step III, see Section 2.4) to generate consumption profiles for the system. These con160 sumption profiles are then fed into an occupancy detection or estimation algorithm (Step 161 IV, Section 2.5) to construct occupancy profiles. By labeling future time periods as workday, 162 weekend, or vacation, and applying the occupancy profiles in conjunction with the labels, 163 an occupancy schedule can be created (Step V, see Section 2.6) to describe the presence of 164 people in the building. Based on the occupancy schedule, adjustments to the AHU schedule 165 Version October 30, 2025 submitted to Energies 5 of 22 can be made to maintain a comfortable indoor climate without any disruption (Step VI, 166 Section 2.7). 167 2.2. Step I: Data sources and data collection 168 Metered data was obtained from a school in Estonia that underwent renovations and 169 was completed in August 2020. Smart electricity meters were used to collect data at an 170 hourly resolution scale, which can be accessed through a Building Management System 171 (BMS). The measuring points include the Main Meter1, Kitchen Main Meter2, as well as 172 ventilation units (SV01–SV11). 173 To gain a better understanding of the occupants’ behavior at the school, the total 174 lighting and plug loads were also added. To determine the electricity usage of non-air175 conditioning loads such as plugs and lighting in a building, a simple yet effective method is 176 to subtract the total energy usage of all AHU systems from the readings of the main meter. 177 This approach provides a convenient way to estimate the energy consumption of all other 178 electrical devices and appliances within the building, including lighting, computers, and 179 other office equipment. 180 2.3. Step II: Data cleaning 181 Firstly, missing data, outliers, and noise were removed to ensure accurate clustering 182 analysis. Additional data engineering was needed to correct errors, like the significant 183 peak in August 2021 due to a sensor disconnection and subsequent Main Meter calibration. 184 This outlier was replaced with the last accurate value. For comparing consumption across 185 buildings, data scaling/normalization was also necessary to capture temporal variations 186 over absolute values [33]. 187 2.4. Step III: Creating usage profiles: clustering 188 The choice of clustering method is crucial for a smart platform for educational build189 ings, impacting accuracy and reliability. K-means is common but has drawbacks like 190 pre-determined cluster numbers and initial centroids, leading to suboptimal results with 191 diverse load profiles. Outliers also influence centroids, requiring data preprocessing [ 36 ], 192 [39]. 193 Incorrect cluster numbering can affect the platform’s effectiveness in optimizing 194 energy usage and ventilation schedules. To address these issues, DBSCAN, an alternative 195 clustering method, should be considered. It identifies clusters of different shapes and sizes 196 and isolates outliers, which is relevant for educational buildings with various load profiles 197 and anomalies [40], [41]. 198 In general, a minimal number of points nmin to core point identification can be set as 199 nmin ≥D+1, (1) where Dis the number of dimensions in the dataset and 200 (p∈Nϵ(q), |Nϵ(q)|>nmin, 201 Nϵ(p) = {q∈D|dist(p,q)≤ϵ}, (2) where Nϵ(p) is the ϵ -neighbourhood of the point p , q is a dataset point, and dist(p , q) is the 202 according distance function [42]. 203 Selecting the appropriate values for ϵ and nmin according to the specific dataset is now 204 crucial. To start with, we typically set nmin to a value between 3 and 5, depending on the 205 size of the dataset. For larger datasets, we recommend selecting a higher value for nmin.206 Version October 30, 2025 submitted to Energies 6 of 22 2.4.1. Electricity usage profiles 207 Our approach defines a consumption profile PF as an hourly time series for a single 208 day, with each hour denoted by a value PF(h) , measured in kilowatt-hours (kWh), represent209 ing the typical usage of “Plugs and lighting” for that day. We distinguish between separate 210 profiles for workdays – PF(wd) , weekends – PF(we) , COVID period – PF(cov) , and school 211 vacation – PF(vac) . To obtain any profile PF , we apply the clustering method described 212 above to train data that includes examples of previous behavior for the same types of days. 213 For instance, to obtain the workday profile PF(wd) , the training data should only contain 214 workdays. The resulting profile will be the centroid of the cluster. 215 In cases where there are additional clusters with more centroids, the centroid that 216 most accurately describes the training data should be selected based on the Mean Absolute 217 Percentage Error (MAPE): 218 M=100% n n ∑ t=1 At−Ct At , (3) where Atis the explicit day consumption value and Ctis the centroid value. 219 2.5. Step IV: Creating occupancy profiles: occupancy detection and estimation 220 Occupancy detection (or estimation) based on load monitoring uses smart meter 221 data and load monitoring technology to determine occupancy patterns in buildings. By 222 analyzing the power consumption of appliances and devices within a building, it can 223 provide valuable insights into when and how spaces are being used [43]. 224 To create occupancy profiles, we propose two methods: occupancy detection (Section 225 2.5.1), involves using algorithms to detect the presence or absence of occupants in a building 226 based on the so-called threshold approach. The second method, occupancy estimation 227 (Section 2.5.2), involves using algorithms to estimate the number of occupants in a building 228 based on the power consumption data. 229 Both methods can be applied to the previously computed “Plugs and lighting” con230 sumption profiles (see Section 2.2). 231 2.5.1. Occupancy detection 232 The process of determining whether people are present or absent in a space is com233 monly referred to as occupancy detection. In our proposed approach, we utilize a threshold234 based method for occupancy detection , which involves setting a predetermined threshold 235 for the “Plugs and lighting” electrical consumption. For each single day, an occupancy 236 profile is defined as a time series with binary hourly values O(h) . Each value in this series 237 indicates whether the space is occupied or unoccupied during a specific hour, and it is com238 puted by using a threshold-based method, where we compare the power consumption value 239 Qe(h) at a given hour to a calculated threshold value defined as Qemin + (Qemax −Qemin )·δ . 240 Here, Qemax and Qemin represent the maximum and minimum values of the consumption 241 time series over a predetermined period (such as daily or weekly). The user-defined thresh242 old value δ is determined through a process of experimentation and analysis. It is essential 243 to set δ at a level that is sufficiently high to distinguish between occupied and unoccupied 244 states but low enough to avoid false positives. After rigorous testing, we found that δ= 0.4 245 provided the best fit for the consumption curve. 246 A monthly dataset was used to create two electricity usage profiles: weekdays and 247 weekends. The occupancy detection algorithm analyzed these daily profiles, effectively 248 capturing daily occupancy variations. This method was found to be well-suited for the 249 application. 250 The binary values in the resulting occupancy profile O(h) are straightforward: when 251 the power consumption at a given hour exceeds the calculated threshold value, O(h) is set 252 to 1, indicating occupancy during that hour. Conversely, if the consumption falls below the 253 threshold, O(h)is set to 0, signifying that the space is unoccupied at that time. 254 Version October 30, 2025 submitted to Energies 7 of 22 O(h)=(1, if Qe(h)>Qemin + (Qemax −Qemin )·δ 0, otherwise (4) 2.5.2. Occupancy estimation 255 Occupancy estimation involves estimating the number of people that are present in a 256 given space or area. However here we do not aim to estimate the exact number of people. 257 Instead, we define the occupancy profile ˆ N as a time series with values ˆ N(h) for each hour, 258 where ˆ N(h)∈[ 0, 1 ] represents the estimated occupancy level for that hour. The value of 259 ˆ N(h)is calculated via the formula 260 ˆ N(h)=Qe(h)−Qemin Qemax −Qemin , (5) where Qemax and Qemin represent the maximum and minimum values of the consumption 261 time series over a predetermined period (daily). The resulting value of ˆ N(h) is between 0 262 and 1, representing the ratio of full occupancy at that particular hour. 263 2.6. Step V: Creating an occupancy schedule 264 To generate an occupancy schedule, we need to consider two main components: occu265 pancy profiles and a designated labelled time period for which the schedule will be gener266 ated. The set of labels used to label the time period is denoted as L={wd , we , cov , vac} , 267 representing workday, weekend, COVID, and vacation, respectively. 268 The labelled time period for which we are scheduling is defined as: 269 SchL={L0,L1, . . . , Ln}, (6) where 270 •nis the length of the scheduled period in days, and 271 •L0 , L1 , . . . , Ln are labels from the set L used to denote each day of the scheduled period. 272 The occupancy schedule is then defined as the concatenation of occupancy profiles 273 with respect to the labels of the period: 274 OSch ={OL0,OL1, . . . , OLn}, (7) where OL0 , OL1 , . . . , OLn are occupancy profiles obtained from the occupancy detection 275 algorithm. Similarly, for the occupancy estimation schedule, we have the following defini276 tion: 277 ˆ NSch ={ˆ NL0,ˆ NL1, . . . , ˆ NLn}, (8) where ˆ NL0 , ˆ NL1 , . . . , ˆ NLn are occupancy profiles obtained from the occupancy estimation 278 algorithm. 279 2.7. Step VI: Creating ventilation schedules 280 The proposed air exchange timetables cover two cases: classical—based on occupancy 281 detection, and demand control ventilation (DCV) strategy—based on occupancy estimation. 282 The choice between those two methods for ventilation scheduling depends on the specific 283 needs of the building, available resources, desired level of control and legislation applied 284 in specific regions. 285 2.7.1. Classical approach based on occupancy detection 286 When using occupancy detection with simple presence detection, the AHU schedule 287 can be based on turning the ventilation on or off depending on the presence or absence of 288 occupants as indicated in the occupancy schedule OSch.289 Version October 30, 2025 submitted to Energies 8 of 22 Additionally, it is possible to consider maximum and minimum air flows for schedul290 ing, so the OSch values will correspond to a specific airflow. In occupancy detection, “0” 291 represents the minimum airflow rate required for the basic ventilation and “1” represents 292 the maximum airflow rate that will be set for ventilation, 293 AOSch ={AOL0,AOL1, . . . , AOLn}, (9) where AOL0 , AOL1 , . . . , AOLn are the corresponding airflow rates as a percentage from the 294 maximum airflow qmax associated with these occupancy levels. 295 However, the AHU scheduling is also influenced by compliance with legal require296 ments or regulations related to ventilation, like minimum ventilation rates, fresh air intake 297 regulations or specific guidelines for indoor air quality [ 44 ], or by the building’s design, 298 location, and orientation. Adjustments may be needed for climate factors or features like 299 large windows that allow natural ventilation. 300 2.7.2. Demand-controlled ventilation strategy based on occupancy estimation 301 In demand-controlled ventilation (DCV), the airflow rates are dynamically adjustable 302 based on number of occupants or demand of the space. It aims to provide ventilation in 303 proportion to the number of occupants or the level of pollutants that are present, rather 304 than relying on fixed ventilation rates. 305 Occupancy estimation can be one method used within a DCV to estimate the number 306 of occupants in a room. To generate an AHU schedule based on occupancy estimation 307 within a DCV system, the air flow of the AHU can be adjusted as a percentage, denoted by 308 ˆ NSch, of its maximum airflow, namely 309 Aˆ NSch ={Aˆ NL0,Aˆ NL1, . . . , Aˆ NLn}, (10) where Aˆ NL0 , Aˆ NL1 , . . . , Aˆ NLn are airflow adjustments (as a percentage of the maximum 310 airflow) for a specific day. 311 To comply with legal regulations or requirements, the AHU schedule may need to be 312 further adjusted. This could include setting a minimum airflow rate to meet ventilation 313 standards (ANSI/ASHRAE Standard 62.1-2019, EN 16798-1:2019 and Estonian regulation 314 RT I, 10.01.2023, 12). 315 2.7.3. Determination of minimum airflow rate for scheduling procedure 316 Even with no occupants ( qmin = 0 [l/s] ), maintaining minimum ventilation is recom317 mended for air quality. This depends on local building rules and guidelines. Adherence to 318 these standards when unoccupied prevents issues related to poor air quality. [45]. 319 Studies suggest that an effective approach is to operate the continuous fan airflow at a 320 fraction of the baseline rate, typically between 13%–40%. Implementing this strategy during 321 unoccupied periods can strike a balance between maintaining air quality and reducing 322 energy consumption [46]. 323 2.7.4. Preand post-occupancy flush out 324 Pre-occupancy flush-out periods can be a beneficial strategy to enhance indoor air 325 quality, reducing the recovery period and potentially saving energy. This approach involves 326 rapidly ventilating the rooms with fresh outdoor air before occupants return, effectively 327 purging any stagnant or potentially polluted air [46]. 328 Occupant peak exposure reduction, shorter recovery periods, and energy savings can 329 thus be realised, particularly when no ventilation air is provided during unoccupied hours. 330 During the flush-out period, the ventilation system is typically set to a maximum airflow 331 rate to achieve a high air exchange rate within the space and remove the contaminants 332 that may have accumulated during unoccupied periods. It is noteworthy that the Estonian 333 regulation on indoor climate in non-residential buildings [ 47 ] requires the application of 334 this precise strategy. 335 Version October 30, 2025 submitted to Energies 9 of 22 2.7.5. Schedule scenarios 336 Based on the techniques that were proposed in Sections 2.7.1 and 2.7.4, a variety 337 of ventilation scheduling scenarios was created. These scenarios encompass different 338 techniques and combinations to calculate general energy consumption. 339 Table 1. Schedule scenarios Scenario Base method Minimal allowed airflow Flush Schedule 1 Classical Unoccupied – Schedule 2 Classical Reduced – Schedule 3 Classical Reduced 30% – Schedule 4 Classical Unoccupied Flush Schedule 5 Classical Reduced Flush Schedule 6 Classical Reduced by 30% Flush Schedule 7 DCV Unoccupied – Schedule 8 DCV Reduced – Schedule 9 DCV Reduced by 30% – Schedule 10 DCV Unoccupied Flush Schedule 11 DCV Reduced Flush Schedule 12 DCV Reduced by 30% Flush The approaches include (Table 1): 340 • Classical occupancy detection with on-off method (Schedule 1) 341 • Classical occupancy detection with reduced minimum airflow corresponding to the 342 value used in a specific AHU, see Schedule 2. 343 • Classical occupancy detection with reduced minimum airflow set to 30% of the maxi344 mum level, see Schedule 3. 345 • Schedules 4 to 6 involve combining the corresponding Schedules 1 to 3 with a pre346 and post-occupancy detection flush strategy. 347 • Schedules 7 to 12 replicate the approaches 1 to 6, but instead of using classical occu348 pancy detection, they utilize occupancy estimation techniques for DCV. 349 2.7.6. AHU demand/air flow relation 350 To calculate the AHU demand by taking into account the air flow, several equations 351 can be utilized. First, if PAHU and Pfan are, respectively, the powers of the air handling unit 352 and of the fan, we can write 353 PAHU =Pfan ·2. (11) The AHU is in fact usually composed of two ventilation systems, one for supply and one 354 for exhaust. Typically, the energy consumption of both systems is kept at the same level, 355 meaning that the total power consumption of the AHU can be divided by two. However, if 356 there is a difference in the consumption of the inflow and outflow systems, the total power 357 consumption can be proportionally split based on the energy consumption of each system, 358 namely 359 Pfan max =SFPfan ·qfan max, (12) where SFP is a Specific Fan Power. On the other hand, the maximum consumption of the 360 fan can be found by using the cube law for fans, stating that the power required to operate 361 Version October 30, 2025 submitted to Energies 16 of 22 2021-02-08 2021-02-09 2021-02-10 2021-02-11 2021-02-12 2021-02-13 2021-02-14 2021-02-15 0 2 AHU#7: calculated current airflow vs airflow based on schedule 7 2021-02-08 2021-02-09 2021-02-10 2021-02-11 2021-02-12 2021-02-13 2021-02-14 2021-02-15 0 2 AHU#7: calculated current airflow vs airflow based on schedule 8 2021-02-08 2021-02-09 2021-02-10 2021-02-11 2021-02-12 2021-02-13 2021-02-14 2021-02-15 0 2 AHU#7: calculated current airflow vs airflow based on schedule 9 2021-02-08 2021-02-09 2021-02-10 2021-02-11 2021-02-12 2021-02-13 2021-02-14 2021-02-15 0 2 AHU#7: calculated current airflow vs airflow based on schedule 10 2021-02-08 2021-02-09 2021-02-10 2021-02-11 2021-02-12 2021-02-13 2021-02-14 2021-02-15 0 2 AHU#7: calculated current airflow vs airflow based on schedule 11 2 0 210 20 8 2 0 210 20 9 2 0 210 2-1 0 2 0 210 2-11 2 0 210 2-12 2 0 210 2-1 3 2 0 210 2-14 2 0 210 2-1 5 0 2 AHU#7: calculated current airflow vs airflow based on schedule 12 Figure 11. Comparison of current schedule (black) and proposed Schedules 7–9 (blue) based on occupancy estimation and comparison of current schedule (black) and proposed Schedules 10–12 (dashed blue) based on occupancy estimation with flush out. tion. The “0” centroid (outliers) and “1” centroid (consumption) are quite distinct during 480 weekdays, portraying the typical large variance of different daily occupants’ schedules. 481 Weekends, as they have nearly zero or very little occupancy, exhibit little variance thus 482 overlapping curves. 483 A clear advantage of using Schedules 1 to 3, which are based on occupancy detection, 484 is portrayed in Figure 10. The 60% capacity of the current schedule during unoccupied 485 hours is replaced by respectively, 0% and 30% for Scenarios 1 and 3, with evident energy 486 savings. Taking this a bit further, and in order to comply with the 2023 Estonian regulations 487 about indoor air quality, one can apply Schedule 4 in Figure 10 that prescribes pre-and 488 post-occupancy flush-out with one hour shift. In addition to these energy considerations, 489 the above results and figures prove that the proposed algorithm for occupancy detection 490 is capable of reproducing an actual scheduling profile in the school. It is therefore well 491 validated. 492 Moving on to the occupancy estimation, this represents one more degree of difficulty 493 compared to the bare (binary) detection. One can see that in Figure 11 the same validation 494 of our method is present, together with some energy saving potential capabilities from the 495 calculated curves (Schedules 7 to 12). Qualitatively, there is no difference about validation 496 whether one chooses AHU#7 or AHU#8. 497 Tables 2and 3constitute our main result from the point of view of applications. First, 498 Table 2quantifies in absolute values how our scenarios will affect expenses by different 499 prices (which have increased). One only needs to keep in mind that this consumption is 500 only based on one week since the idea is to create schedules for the week or the month 501 ahead, taking into account profiles from the closest weeks and similar ones from previous 502 years. 503 An even more pristine view of the benefit that can be reached through a more flex504 ible schedule is given in Table 3. For instance, Scenario 1 provides 10.4% energy saving 505 Version October 30, 2025 submitted to Energies 17 of 22 2021-02-08 2021-02-09 2021-02-10 2021-02-11 2021-02-12 2021-02-13 2021-02-14 2021-02-15 0 2 4 AHU#8: calculated current airflow vs airflow based on schedule 1 2021-02-08 2021-02-09 2021-02-10 2021-02-11 2021-02-12 2021-02-13 2021-02-14 2021-02-15 0 2 4 AHU#8: calculated current airflow vs airflow based on schedule 2 2021-02-08 2021-02-09 2021-02-10 2021-02-11 2021-02-12 2021-02-13 2021-02-14 2021-02-15 0 2 4 AHU#8: calculated current airflow vs airflow based on schedule 3 2021-02-08 2021-02-09 2021-02-10 2021-02-11 2021-02-12 2021-02-13 2021-02-14 2021-02-15 0 2 4 AHU#8: calculated current airflow vs airflow based on schedule 4 2021-02-08 2021-02-09 2021-02-10 2021-02-11 2021-02-12 2021-02-13 2021-02-14 2021-02-15 0 2 4 AHU#8: calculated current airflow vs airflow based on schedule 5 2 0 2 1 - 0 2 - 0 8 2 0 2 1 - 0 2 - 0 9 2 0 2 1 - 0 2 - 1 0 2 0 2 1 - 0 2 - 1 1 2 0 2 1 - 0 2 - 1 2 2 0 2 1 - 0 2 - 1 3 2 0 2 1 - 0 2 - 1 4 2 0 2 1 - 0 2 - 1 5 0 2 4 AHU#8: calculated current airflow vs airflow based on schedule 6 Figure 12. Comparison of current schedule (black) and proposed Schedules 1–3 (green) based on occupancy detection and comparison of current schedule (black) and proposed Schedules 4–6 (dashed green) based on occupancy detection with flush out compared to the measured consumption. However, the fact that in 2021 it costs 4.1% 506 more shows how establishing a dynamical schedule that would i) closely reflect occupancy 507 changes, ii) comply with national regulations, and iii) save energy at the same time is not 508 an easy task. 509 The relative percentages over the actual schedule’s cost in Table 3are manifestly 510 showing that Scenarios 1 to 6 (occupancy detection with or without flush-out) exhibit larger 511 costs than the reference actual profile from the school. The reason for increased energy 512 consumption in these cases is the utilization of maximum airflow during occupied periods 513 and the same minimum airflow as in the original schedule. 514 On the other hand, Scenarios 7 to 12 are systematically and substantially cheaper 515 than the actual schedule, reaching 61.2% savings for AHU#8 for the reference week. It 516 therefore seems that a more sophisticated algorithm that is based on occupancy estimation 517 is more rewarding. This is another pristine confirmation of the cost effectiveness of DCV 518 ventilation. 519 We should remark that the above considerations are drawn from somewhat crude 520 estimations that are based only on one single week. Over the whole year, the percentages 521 in Table 3might thus change. As anyway the week here considered for testing is typical, 522 therefore representative of the school’s normal operation, we do not expect substantial 523 deviations from these results, which are still valid, at least qualitatively. 524 It is also worth noticing that Scenario 6, although consuming circa 1 kWh more energy, 525 ensures compliance with both ASHRAE 62.1 and EN 16798 standards, along with the 526 Estonian regulation (RT I, 10.01.2023, 12) for IAQ. Conversely, strictly adhering to the 527 Estonian legislation, where ventilation is completely turned off during unoccupied hours, 528 will result in lower energy consumption. However, this approach may not lead to significant 529 cost savings as market prices tend to be higher during the beginning of the working day 530 and evenings when the flush out strategy is applied. 531 Version October 30, 2025 submitted to Energies 18 of 22 2021-02-08 2021-02-09 2021-02-10 2021-02-11 2021-02-12 2021-02-13 2021-02-14 2021-02-15 0 2 4 AHU#8: calculated current airflow vs airflow based on schedule 7 2021-02-08 2021-02-09 2021-02-10 2021-02-11 2021-02-12 2021-02-13 2021-02-14 2021-02-15 0 2 4 AHU#8: calculated current airflow vs airflow based on schedule 8 2021-02-08 2021-02-09 2021-02-10 2021-02-11 2021-02-12 2021-02-13 2021-02-14 2021-02-15 0 2 4 AHU#8: calculated current airflow vs airflow based on schedule 9 2021-02-08 2021-02-09 2021-02-10 2021-02-11 2021-02-12 2021-02-13 2021-02-14 2021-02-15 0 2 4 AHU#8: calculated current airflow vs airflow based on schedule 10 2021-02-08 2021-02-09 2021-02-10 2021-02-11 2021-02-12 2021-02-13 2021-02-14 2021-02-15 0 2 4 AHU#8: calculated current airflow vs airflow based on schedule 11 2 0 2 1 - 0 2 - 0 8 2 0 2 1 - 0 2 - 0 9 2 0 2 1 - 0 2 - 1 0 2 0 2 1 - 0 2 - 1 1 2 0 2 1 - 0 2 - 1 2 2 0 2 1 - 0 2 - 1 3 2 0 2 1 - 0 2 - 1 4 2 0 2 1 - 0 2 - 1 5 0 2 4 AHU#8: calculated current airflow vs airflow based on schedule 12 Figure 13. Comparison of current schedule (black) and proposed Schedules 7–9 (blue) based on occupancy estimation and comparison of current schedule (black) and proposed Schedules 10–12 (dashed blue) based on occupancy estimation with flush out To summarise, all the applied DCV schedules with occupancy estimation (Scenarios 7 532 to 12) lead to decreased energy consumption and lower total bills, considering both energy 533 consumption and cost savings. It is therefore important to strike a balance between meeting 534 ventilation standards and optimizing energy efficiency. By selecting an appropriate DCV 535 strategy, we have shown that it is possible to achieve both reduced energy consumption 536 and lower the overall expenses. 537 The method described above has several limitations that should be taken into consid538 eration. First, if the system has only one main meter for cumulative readings and lacks 539 individual readings for the AHUs at a given measurement frequency, it becomes challeng540 ing to accurately aggregate the “plugs and lighting” feature. This can lead to less accurate 541 occupancy detection and estimation. 542 Secondly, the method assumes predictable behavior patterns. However, significant dis543 ruptions like the COVID-19 pandemic can compromise the accuracy of occupancy detection 544 and estimation. Unforeseen events or behavior shifts can challenge occupancy prediction, 545 especially initially. The pandemic-induced shift to remote work and school closures altered 546 electricity usage, rendering models based on historical patterns less effective. 547 Third, the clustering procedure that is used for occupancy estimation requires some 548 minimum amount of data. If the system was just installed, thus it lacks sufficient data, 549 performing clustering becomes almost impossible, resulting in lower overall accuracy. 550 It is also important to note that behavioral patterns and schedules derived from data 551 are typically effective for shorter periods, ranging from one week to one month. This is 552 because behavior can be influenced by seasonal shifts, the occurrence of seasonal diseases, 553 and other factors. As a result, it is not recommended to rely on longer-term schedules 554 without regular updates to account for changing patterns and circumstances. Under555 standing and acknowledging these limitations is crucial for implementing accurate and 556 Version October 30, 2025 submitted to Energies 19 of 22 Table 2. Weekly energy consumption and costs as induced by the proposed schedules in comparison with those measured (energy) or estimated (costs) with the current school scheduling Scenario AHU 7 AHU 8 Cons. 2021 2022 2023 Cons. 2021 2022 2023 kWh e e e kWh e e e 1, CL 627.9 59.85 90.22 91.96 583.7 55.64 83.87 85.48 2, CL 843.17 71.3 106.2 110.37 699.06 61.77 92.43 95.35 3, CL 654.68 61.27 92.21 94.25 608.42 56.95 85.7 87.6 4, CL 676.2 63.65 96.97 98.15 628.6 59.17 90.15 91.24 5, CL 881.02 74.28 111.49 115.22 738.36 64.87 97.93 100.39 6, CL 701.68 64.97 98.78 100.27 652.12 60.39 91.81 93.2 7, DCV 235.55 25.42 35.21 35.23 219.1 23.64 32.75 32.77 8, DCV 496.72 41.18 57.58 60.44 352.36 31.47 43.8 45.28 9, DCV 262.33 26.85 37.19 37.52 352.36 31.47 43.8 45.28 10, DCV 329.05 32.81 48.13 47.26 306.0 30.51 44.76 43.95 11, DCV 572.42 47.17 68.04 70.19 430.96 37.69 54.66 55.4 12, DCV 354.53 34.14 49.94 49.39 329.52 31.73 46.43 45.91 Measured 700.78 57.51 85.67 89.55 650.56 56.12 84.25 87.49 Table 3. Weekly energy consumption and prices for different years: comparison of the proposed schedules with the actual school schedule in terms of percentage difference Scenario AHU 7 [%] AHU 8 [%] Cons. 2021 2022 2023 Cons. 2021 2022 2023 1, CL -10.4 4.1 5.3 2.7 -10.3 -0.9 -0.5 -2.3 2, CL 20.3 24.0 23.9 23.2 7.5 10.0 9.7 9.0 3, CL -6.6 6.5 7.6 5.2 -6.5 1.4 1.7 0.1 4, CL -3.5 10.6 13.2 9.6 -3.4 5.4 7.0 4.3 5, CL 25.7 29.1 30.1 28.6 13.5 15.5 16.2 14.7 6, CL 0.1 12.9 15.2 11.9 0.2 7.5 8.9 6.5 7, DCV -66.4 -55.8 -58.9 -60.7 -66.3 -57.9 -61.2 -62.6 8, DCV -29.1 -28.5 -32.8 -32.6 -45.8 -44.0 -48.1 -48.3 9, DCV -62.6 -53.4 -56.6 -58.1 -45.8 -44.0 -48.1 -48.3 10, DCV -53.0 -43.0 -43.9 -47.3 -53.0 -45.7 -46.9 -49.8 11, DCV -18.3 -18.1 -20.7 -21.7 -33.8 -33.0 -35.2 -36.8 12, DCV -49.4 -40.7 -41.8 -44.9 -49.4 -43.6 -45.0 -47.6 effective occupancy-based ventilation strategies, and it highlights the need for continuous 557 monitoring and adaptation to maintain optimal indoor air quality and energy efficiency. 558 Lastly, we should point out that the present investigation is currently focused on 559 control algorithm optimization and computer simulations. The actual experiments are 560 planned to be conducted in different settings such as offices or shopping malls. Whilst 561 occupancy inference and the subsequent AHU control algorithm are trained on real-world 562 data from the school, we have certain limitations in conducting experiments in facilities 563 primarily used by children due to ethical considerations. As our estimates in Tables 2and 564 3are however encouraging, an immediate development of this study will involve a full 565 experimental assessment to validate and fine-tune such estimates against a school’s AHU 566 automation control that integrates our algorithm. 567 5. Conclusions 568 Defining a ventilation schedule in schools that would guarantee good IAQ, satisfy 569 national standard’s requirements, save energy and be cost-effective at the same time is 570 a rather involved task. Demand-controlled ventilation seems to be a valid tool to this 571 aim; nevertheless, its implementation requires knowledge of the amount of occupants in 572 Version October 30, 2025 submitted to Energies 20 of 22 the enclosures. More often than not, installing in classrooms the necessary sensors for 573 occupancy detection or estimation is not possible. 574 In this paper, we have attempted to circumvent this difficulty by establishing a novel 575 statistical method that is based on advanced data clustering techniques and measured 576 energy use from a school in Estonia during the years 2020 and 2021. Using only energy 577 consumption data of the air handling units, the algorithm is able to either simply predict 578 whether occupants are present in the space, or to estimate their number. Such predictions 579 were used to implement 12 schedules for the operation of ventilation systems in the 580 school. These schedules cover different techniques and combinations to calculate general 581 energy consumption, and vary from Classical occupancy detection with on-off method to 582 occupancy estimation for demand controlled ventilation. 583 We have found that by considering occupancy detection, the 60% capacity of current 584 schedules can be replaced by 30% or even 0%, resulting in evident energy savings that can 585 range from 3.5% to 66.4%. While the schedules that are based on the Classical detection 586 method exhibit a wide range of energy consumption, with half of them inducing equal 587 or larger energy use than with the actual scheduling, the more sophisticated DCV-based 588 occupancy estimation methods induce systematic savings in terms of both energy and 589 costs. 590 A cost analysis with energy price estimations running from 2021 to 2023 returns 591 indeed substantial economic savings if DCV-occupancy estimation is realised, from 18.1% 592 to even 62.6%. This further supports the usage of DCV in ventilation systems’ operation. 593 Remarkably, these schedules also comply with the 2023 Estonian regulations about indoor 594 air quality, which makes them applicable and immediately ready for implementation into 595 Standards after further successful testing. 596 Some limitations of the method here introduced do however exist. Aggregating the 597 "plugs and lighting" feature could be problematic in certain buildings, due to lack of meters; 598 behavior patterns might not be easily predictable; the clustering procedure that was used 599 for occupancy estimation requires a minimum amount of data; and finally, behavioral 600 patterns and data-derived schedules are usually effective for shorter periods, from one 601 week to one month, needing dataset updates for addressing longer time periods. This all 602 constitutes some input for future work. 603 Author Contributions: Conceptualization, K.V., M.M. and A.F.; methodology, K.V., M.M., A.F. and 604 M.T.; software, K.V. and M.M; validation, K.V., M.M., A.F. and M.T.; formal analysis, J.B. and E.P; in605 vestigation, K.V., M.M., A.F; resources, J.B., E.P., M.T; data curation, K.V. and M.M.; writing—original 606 draft preparation, K.V., M.M.,A.F.; writing—review and editing, J.B., E.P. and M.T.; visualization, K.V., 607 M.M. and A.F.; supervision, J.B., E.P., M.T.; project administration, M.T.; funding acquisition, J.B., M.T. 608 and E.P. All authors have read and agreed to the published version of the manuscript. 609 Funding: This work was partly supported by the Estonian Research Council grants no. PRG1463, 610 PRG658, and PSG409, by the Estonian Ministry of Education and Research and European Regional 611 Development Fund (grant 2014-2020.4.01.20-0289), and by the European Commission through the 612 H2020 project Finest Twins (grant No. 856602), by the European Union’s Horizon Europe research 613 and innovation programme under the grant agreement No 101120657, project ENFIELD (European 614 Lighthouse to Manifest Trustworthy and Green AI) and by the Estonian Centre of Excellence in 615 Energy Efficiency, ENER (grant TK230) funded by the Estonian Ministry of Education and Research. 616 Acknowledgments: We thank Tuule Mall Parts for useful discussions. 617 Conflicts of Interest: The authors declare no conflicts of interest. 618 References 619 1. Economidou, M.; Todeschi, V.; Bertoldi, P.; D’Agostino, D.; Zangheri, P.; Castellazzi, L. Review of 50 years of EU energy efficiency 620 policies for buildings. Energy and Buildings 2020,225, 110322. https://doi.org/https://doi.org/10.1016/j.enbuild.2020.110322.621 2. European Commission. COM(2021)802 - Directive. Energy performance of buildings (recast), 2021. 622 3. European Commission. Energy efficiency in buildings, 2020. 623 Version October 30, 2025 submitted to Energies 21 of 22 4. McArthur, J.; Powell, C. Health and wellness in commercial buildings: Systematic review of sustainable building rating 624 systems and alignment with contemporary research. Building and Environment 2020,171, 106635. https://doi.org/https: 625 //doi.org/10.1016/j.buildenv.2019.106635.626 5. Hong, T.; Taylor-Lange, S.C.; D’Oca, S.; Yan, D.; Corgnati, S.P. Advances in research and applications of energy-related occupant 627 behavior in buildings. Energy and buildings 2016,116, 694–702. 628 6. Ferrantelli, A.; Kuivjõgi, H.; Kurnitski, J.; Thalfeldt, M. Office Building Tenants’ Electricity Use Model for Building Performance 629 Simulations. Energies 2020,13.https://doi.org/10.3390/en13215541.630 7. Delzendeh, E.; Wu, S.; Lee, A.; Zhou, Y. The impact of occupants’ behaviours on building energy analysis: A research review. 631 Renewable and Sustainable Energy Reviews 2017,80, 1061–1071. https://doi.org/https://doi.org/10.1016/j.rser.2017.05.264.632 8. Martinaitis, V.; Zavadskas, E.K.; Motuzien ˙ e, V.; Vilutien ˙ e, T. Importance of occupancy information when simulating energy 633 demand of energy efficient house: A case study. Energy and Buildings 2015,101, 64–75. https://doi.org/https://doi.org/10.1016/ 634 j.enbuild.2015.04.031.635 9. Maasoumy, M.; Sangiovanni-Vincentelli, A. Smart Connected Buildings Design Automation: Foundations and Trends. Foundations 636 and Trends® in Electronic Design Automation 2016,10, 1–143. https://doi.org/10.1561/1000000043.637 10. Trivedi, D.; Badarla, V. Occupancy detection systems for indoor environments: A survey of approaches and methods. Indoor and 638 Built Environment 2020,29, 1053–1069. https://doi.org/10.1177/1420326X19875621.639 11. Lu, X.; Bhattacharya, S.; Sharma, H.; Adetola, V.; O’Neill, Z. Sensor impact evaluation in commercial buildings: The case of 640 occupancy-centric controls. Energy and Buildings 2022,267, 112134. https://doi.org/10.1016/j.enbuild.2022.112134.641 12. Kjærgaard, M.B.; Ardakanian, O.; Carlucci, S.; Dong, B.; Firth, S.K.; Gao, N.; Huebner, G.M.; Mahdavi, A.; Rahaman, M.S.; 642 Salim, F.D.; et al. Current practices and infrastructure for open data based research on occupant-centric design and operation of 643 buildings. Building and Environment 2020,177, 106848. https://doi.org/10.1016/j.buildenv.2020.106848.644 13. Sun, K.; Zhao, Q.; Zou, J. A review of building occupancy measurement systems. Energy and Buildings 2020,216, 109965. 645 https://doi.org/10.1016/J.ENBUILD.2020.109965.646 14. Gul, M.S.; Patidar, S. Understanding the energy consumption and occupancy of a multi-purpose academic building. Energy and 647 Buildings 2015,87, 155–165. https://doi.org/10.1016/j.enbuild.2014.11.027.648 15. Khan, D.S.; Kolarik, J.; Hviid, C.A.; Weitzmann, P. Method for long-term mapping of occupancy patterns in open-plan and 649 single office spaces by using passive-infrared (PIR) sensors mounted below desks. Energy and Buildings 2021,230, 110534. 650 https://doi.org/10.1016/j.enbuild.2020.110534.651 16. Li, N.; Calis, G.; Becerik-Gerber, B. Measuring and monitoring occupancy with an RFID based system for demand-driven HVAC 652 operations. Automation in Construction 2012,24, 89–99. https://doi.org/10.1016/j.autcon.2012.02.013.653 17. Daniel, O.C.; Ramsurrun, V.; Seeam, A.K. Smart Library Seat, Occupant and Occupancy Information System, using Pressure and 654 RFID Sensors. In Proceedings of the 2019 Conference on Next Generation Computing Applications (NextComp). IEEE, 9 2019, 655 pp. 1–5. https://doi.org/10.1109/NEXTCOMP.2019.8883610.656 18. Wang, Y.; Shao, L. Understanding occupancy pattern and improving building energy efficiency through Wi-Fi based indoor 657 positioning. Building and Environment 2017,114, 106–117. https://doi.org/10.1016/j.buildenv.2016.12.015.658 19. Wang, W.; Wang, J.; Chen, J.; Huang, G.; Guo, X. Multi-zone outdoor air coordination through Wi-Fi probe-based occupancy 659 sensing. Energy and Buildings 2018,159, 495–507. https://doi.org/10.1016/j.enbuild.2017.11.041.660 20. O’Neill, Z.D.; Li, Y.; Cheng, H.C.; Zhou, X.; Taylor, S.T. Energy savings and ventilation performance from CO2 -based demand 661 controlled ventilation: Simulation results from ASHRAE RP-1747 (ASHRAE RP-1747). Science and Technology for the Built 662 Environment 2020,26, 257–281. https://doi.org/10.1080/23744731.2019.1620575.663 21. Franco, A.; Leccese, F. Measurement of CO2 concentration for occupancy estimation in educational buildings with energy 664 efficiency purposes. Journal of Building Engineering 2020,32, 101714. https://doi.org/10.1016/j.jobe.2020.101714.665 22. Wei, Y.; Wang, S.; Jin, L.; Xu, Y.; Ding, T. Indoor occupancy estimation from carbon dioxide concentration using parameter 666 estimation algorithms. Building Services Engineering Research and Technology 2022,43, 419–438. https://doi.org/10.1177/01436244 667 211060903.668 23. Chen, Z.; Jiang, C.; Xie, L. Building occupancy estimation and detection: A review. Energy and Buildings 2018,169, 260–270. 669 https://doi.org/10.1016/j.enbuild.2018.03.084.670 24. Petersen, S.; Pedersen, T.H.; Nielsen, K.U.; Knudsen, M.D. Establishing an image-based ground truth for validation of sensor 671 data-based room occupancy detection. Energy and Buildings 2016,130, 787–793. https://doi.org/10.1016/j.enbuild.2016.09.009.672 25. Zou, J.; Zhao, Q.; Yang, W.; Wang, F. Occupancy detection in the office by analyzing surveillance videos and its application to 673 building energy conservation. Energy and Buildings 2017,152, 385–398. https://doi.org/10.1016/j.enbuild.2017.07.064.674 26. Capozzoli, A.; Piscitelli, M.S.; Brandi, S. Mining typical load profiles in buildings to support energy management in the smart 675 city context. Energy Procedia 2017,134, 865–874. Sustainability in Energy and Buildings 2017: Proceedings of the Ninth KES 676 International Conference, Chania, Greece, 5-7 July 2017, https://doi.org/https://doi.org/10.1016/j.egypro.2017.09.545.677 27. Kleiminger, W.; Beckel, C.; Staake, T.; Santini, S. Occupancy Detection from Electricity Consumption Data. In Proceedings 678 of the Proceedings of the 5th ACM Workshop on Embedded Systems For Energy-Efficient Buildings. ACM, 11 2013, pp. 1–8. 679 https://doi.org/10.1145/2528282.2528295.680 Version October 30, 2025 submitted to Energies 22 of 22 28. Kleiminger, W.; Beckel, C.; Santini, S. Household occupancy monitoring using electricity meters. In Proceedings of the 681 Proceedings of the 2015 ACM International Joint Conference on Pervasive and Ubiquitous Computing. ACM, 9 2015, pp. 975–986. 682 https://doi.org/10.1145/2750858.2807538.683 29. Pereira, D.; Castro, R.; Adão, P. Occupancy detection and prediction from electricity consumption data in smart homes: application 684 to a Portuguese case-study. Intelligent Buildings International 2022,14, 690–709. https://doi.org/10.1080/17508975.2021.1985418. 685 30. Gunay, H.B.; O’Brien, W.; Beausoleil-Morrison, I.; Gilani, S. Modeling plug-in equipment load patterns in private office spaces. 686 Energy and Buildings 2016,121, 234–249. https://doi.org/10.1016/j.enbuild.2016.03.001.687 31. Chong, A.; Augenbroe, G.; Yan, D. Occupancy data at different spatial resolutions: Building energy performance and model 688 calibration. Applied Energy 2021,286, 116492. https://doi.org/10.1016/j.apenergy.2021.116492.689 32. Satre-Meloy, A.; Diakonova, M.; Grünewald, P. Cluster analysis and prediction of residential peak demand profiles using 690 occupant activity data. Applied Energy 2020,260, 114246. https://doi.org/https://doi.org/10.1016/j.apenergy.2019.114246.691 33. Liu, X.; Ding, Y.; Tang, H.; Xiao, F. A data mining-based framework for the identification of daily electricity usage patterns and 692 anomaly detection in building electricity consumption data. Energy and Buildings 2021,231, 110601. 693 34. Tien, P.W.; Wei, S.; Darkwa, J.; Wood, C.; Calautit, J.K. Machine Learning and Deep Learning Methods for Enhancing Building 694 Energy Efficiency and Indoor Environmental Quality – A Review. Energy and AI 2022,10, 100198. https://doi.org/https: 695 //doi.org/10.1016/j.egyai.2022.100198.696 35. Ferrando, M.; Marchenko, A.; Erba, S.; Causone, F.; Carlucci, S. Pattern Recognition And Classification For Electrical Energy Use 697 In Residential Buildings. In Proceedings of the Proceedings of the 16th IBPSA Conference, 09 2019. https://doi.org/10.26868/2 698 5222708.2019.210750.699 36. Toussaint, W.; Moodley, D. Clustering Residential Electricity Consumption Data to Create Archetypes that Capture Household 700 Behaviour in South Africa. South African Computer Journal 2020,32.https://doi.org/10.18489/sacj.v32i2.845.701 37. Wei, Y.; Zhang, X.; Shi, Y.; Xia, L.; Pan, S.; Wu, J.; Han, M.; Zhao, X. A review of data-driven approaches for prediction 702 and classification of building energy consumption. Renewable and Sustainable Energy Reviews 2018,82, 1027–1047. https: 703 //doi.org/10.1016/j.rser.2017.09.108.704 38. Schwartz, Z.; Uysal, M.; Webb, T.; Altin, M. Hotel daily occupancy forecasting with competitive sets: a recursive algorithm. 705 International Journal of Contemporary Hospitality Management 2016.706 39. Yang, J.; Ning, C.; Deb, C.; Zhang, F.; Cheong, D.; Lee, S.E.; Sekhar, C.; Tham, K.W. k-Shape clustering algorithm for building 707 energy usage patterns analysis and forecasting model accuracy improvement. Energy and Buildings 2017,146, 27–37. https: 708 //doi.org/https://doi.org/10.1016/j.enbuild.2017.03.071.709 40. Mahdi, M.A.; Hosny, K.M.; Elhenawy, I. Scalable Clustering Algorithms for Big Data: A Review. IEEE Access 2021,9, 80015–80027. 710 https://doi.org/10.1109/ACCESS.2021.3084057.711 41. Deb, C.; Frei, M.; Hofer, J.; Schlueter, A. Automated load disaggregation for residences with electrical resistance heating. Energy 712 and Buildings 2019,182, 61–74. https://doi.org/https://doi.org/10.1016/j.enbuild.2018.10.011.713 42. Ester, M.; Kriegel, H.P.; Sander, J.; Xu, V. A Density-Based Algorithm for Discovering Clusters. In Proceedings of the Proceedings 714 of the Second International Conference on Knowledge Discovery in Databases and Data Mining. AAAI Press, 1996, pp. 223–231. 715 43. Adeogun, R.; Rodriguez, I.; Razzaghpour, M.; Berardinelli, G.; Christensen, P.H.; Mogensen, P.E. Indoor Occupancy Detection 716 and Estimation using Machine Learning and Measurements from an IoT LoRa-based Monitoring System. In Proceedings of the 717 2019 Global IoT Summit (GIoTS), 2019, pp. 1–5. https://doi.org/10.1109/GIOTS.2019.8766374.718 44. Murphy, J.; Sturm, E. Demand-Controlled Ventilation. Trane Engineers Newsletter LIVE Series 2018, pp. 1–50. 719 45. Khovalyg, D.; Kazanci, O.B.; Halvorsen, H.; Gundlach, I.; Bahnfleth, W.P.; Toftum, J.; Olesen, B.W. Critical review of standards for 720 indoor thermal environment and air quality. Energy and Buildings 2020,213, 109819. https://doi.org/10.1016/j.enbuild.2020.109 721 819.722 46. Clark, J.D.; Less, B.D.; Dutton, S.M.; Walker, I.S.; Sherman, M.H. Efficacy of occupancy-based smart ventilation control strategies 723 in energy-efficient homes in the United States. Building and Environment 2019,156, 253–267. https://doi.org/10.1016/j.buildenv. 724 2019.03.002.725 47. Majandusja taristuminister. Hoolekandeasutuse hoone ja sisekliima tagamisega mitteelamu ventileerimise ja tuulutamise 726 nõuded. Technical report, Majandusja taristuminister, Riigi Teataja, 2023. 727 48. Tödtli, J. The impact of control on the energy use of fans in building ventilation systems. The REHVA European HVAC Journal 2015, 728 52, 50–57. 729 49. Ding, Y.; Brattebø, H.; Nord, N. A systematic approach for data analysis and prediction methods for annual energy profiles: An 730 example for school buildings in Norway. Energy and Buildings 2021,247, 111160. https://doi.org/https://doi.org/10.1016/j. 731 enbuild.2021.111160.732 50. Degirmenci, A.; Karal, O. Efficient density and cluster based incremental outlier detection in data streams. Information Sciences 733 2022,607, 901–920. https://doi.org/10.1016/j.ins.2022.06.013.734 51. Elering. Elering LIVE: Börsihinnad, 2023. 735 Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual 736 author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to 737 people or property resulting from any ideas, methods, instructions or products referred to in the content. 738