Estimating Real-Time Occupancy in Multi-Zone Offices
A Comparison Between Conventional Techniques and Deep Learning Methods
DOI:
https://doi.org/10.52825/isec.v2i.3294Keywords:
Occupancy Estimation, Physical CO2 Mass Balance, Hybrid Deep Learning, Multi-Zone Office BuildingAbstract
It is well known that a proper estimation of the real-time occupancy in offices is key for efficient control of the technical systems as well as for security aspects. This work presents an in-depth comparison between two contrasting methods for real-time occupancy estimation in multi-zone offices: from conventional techniques, such as the physical CO2 mass balance, to more cutting-edge techniques based on Deep Learning (DL) methods. The aim is to compare the applicability and accuracy of both methods for estimating real-time occupancy based on non-intrusive sensors measurements. The comparability is performed both from quantitative, based on a wide range of error indicators, and qualitative point of views. The results indicate that both methods estimate real-time occupancy with remarkable accuracy, although the DL method offers slightly better results. Nevertheless, the qualitative analysis can provide additional insights when deciding whether to use one method or the other.
Downloads
References
[1] J. Brooks, S. Kumar, S. Goyal, R. Subramany, and P. Barooah, ‘Energy-efficient control of under-actuated HVAC zones in commercial buildings’, Energy Build., vol. 93, pp. 160–168, 2015, doi: 10.1016/j.enbuild.2015.01.050.
[2] F. Felgueiras, Z. Mourão, A. Moreira, and M. Fonseca Gabriel, ‘Indoor environmental quality in offices and risk of health and productivity complaints at work: A literature re-view’, J. Hazard. Mater. Adv., vol. 10, p. 100314, 2023, doi: 10.1016/j.hazadv.2023.100314.
[3] T. Li et al., ‘A systematic review and comprehensive analysis of building occupancy pre-diction’, Renew. Sustain. Energy Rev., vol. 193, p. 114284, Apr. 2024, doi: 10.1016/j.rser.2024.114284.
[4] X. Lu, Z. Pang, Y. Fu, and Z. O’Neill, ‘Advances in research and applications of CO2-based demand-controlled ventilation in commercial buildings: A critical review of control strategies and performance evaluation’, Build. Environ., vol. 223, p. 109455, Sep. 2022, doi: 10.1016/j.buildenv.2022.109455.
[5] E. A. B. Maldonado and J. E. Woods, ‘A method to select locations for indoor air quality sampling’, Build. Environ., vol. 18, no. 4, pp. 171–180, Jan. 1983, doi: 10.1016/0360-1323(83)90025-2.
[6] R. Zhang, K. P. Lam, Y. S. Chiou, and B. Dong, ‘Information-theoretic environment fea-tures selection for occupancy detection in open office spaces’, Build. Simul., vol. 5, no. 2, pp. 179–188, 2012, doi: 10.1007/s12273-012-0075-6.
[7] J. Yang et al., ‘Comparison of different occupancy counting methods for single system-single zone applications’, Energy Build., vol. 172, pp. 221–234, Aug. 2018, doi: 10.1016/j.enbuild.2018.04.051.
[8] M. S. Zuraimi, A. Pantazaras, K. A. Chaturvedi, J. J. Yang, K. W. Tham, and S. E. Lee, ‘Predicting occupancy counts using physical and statistical Co2-based modeling method-ologies’, Build. Environ., vol. 123, pp. 517–528, Oct. 2017, doi: 10.1016/j.buildenv.2017.07.027.
[9] A. Zhao, M. Zhang, W. Quan, and W. Sun, ‘A hybrid prediction model for heating load of buildings within residential communities considering occupancy rates’, Energy Build., vol. 329, p. 115220, Feb. 2025, doi: 10.1016/j.enbuild.2024.115220.
[10] I. Khan, E. Greco, A. Guerrieri, and G. Spezzano, ‘Occupancy Prediction in Buildings: State of the Art and Future Directions’, in Device-Edge-Cloud Continuum: Paradigms, Ar-chitectures and Applications, C. Savaglio, G. Fortino, M. Zhou, and J. Ma, Eds, Cham: Springer Nature Switzerland, 2024, pp. 203–229. doi: 10.1007/978-3-031-42194-5_12.
[11] Y. Shi and P. Chen, ‘Energy retrofitting of hospital buildings considering climate change: An approach integrating automated machine learning with NSGA-III for multi-objective optimization’, Energy Build., vol. 319, p. 114571, Sep. 2024, doi: 10.1016/j.enbuild.2024.114571.
[12] J. Zhu et al., ‘A learning-based model predictive control method for unlocking the potential of building energy flexibility’, Energy Build., vol. 330, p. 115299, Mar. 2025, doi: 10.1016/j.enbuild.2025.115299.
[13] P. Hernandez-Cruz, N. Vicente-Gómez, Z. Azkorra-Larrinaga, R. Pérez-Orozco, and A. Erkoreka-Gonzalez, ‘Applicability and accuracy of physical CO2 mass balance in multi-zone mechanically ventilated office buildings to estimate real-time occupancy’, Build. En-viron., vol. 289, p. 114099, Feb. 2026, doi: 10.1016/j.buildenv.2025.114099.
[14] M. Cordeiro-Costas, R. Pérez-Orozco, P. Hernandez-Cruz, F. Troncoso-Pastoriza, and E. Granada-Álvarez, ‘Hybrid LSTM-MLP model with NSGA-II-based hyperparameter opti-mization for non-invasive occupancy estimation’, Energy AI, p. 100643, Oct. 2025, doi: 10.1016/j.egyai.2025.100643.
[15] D. Calì, P. Matthes, K. Huchtemann, R. Streblow, and D. Müller, ‘CO2 based occupancy detection algorithm: Experimental analysis and validation for office and residential build-ings’, Build. Environ., vol. 86, pp. 39–49, Apr. 2015, doi: 10.1016/j.buildenv.2014.12.011.
[16] F. Banihashemi, M. Weber, F. Deghim, C. Zong, and W. Lang, ‘Occupancy modeling on non-intrusive indoor environmental data through machine learning’, Build. Environ., vol. 254, p. 111382, Apr. 2024, doi: 10.1016/j.buildenv.2024.111382.
[17] C. Jiang, M. K. Masood, Y. C. Soh, and H. Li, ‘Indoor occupancy estimation from carbon dioxide concentration’, Energy Build., vol. 131, pp. 132–141, Nov. 2016, doi: 10.1016/j.enbuild.2016.09.002.
[18] ASTM Standard D5157-97, Standard Guide for Statistical Evaluation of Indoor Air Quality Models, ASTM D5157-97, West Conshohocken, PA, USA., 1997.
Published
How to Cite
Conference Proceedings Volume
Section
License
Copyright (c) 2026 Pablo Hernandez-Cruz, Moisés Cordeiro-Costas, Raquel Pérez-Orozco, Zaloa Azkorra-Larrinaga, Pablo Eguía-Oller, Aitor Erkoreka-Gonzalez
This work is licensed under a Creative Commons Attribution 4.0 International License.
Funding data
-
Agencia Estatal de Investigación
Grant numbers This publication is part of the R+D+i project PID2024-156054OB-C21 and PID2024-156054OB-C22, financed by MICIU/AEI/10.13039/501100011033 and FEDER, UE.