Conference Agenda

Overview and details of the sessions of this conference. Please select a date or location to show only sessions at that day or location. Please select a single session for detailed view (with abstracts and downloads if available).

Please note that all times are shown in the time zone of the conference. The current conference time is: 5th July 2022, 13:49:29 CEST

 
 
Session Overview
Session
Session W2.6: The role of occupants
Time:
Wednesday, 01/Sept/2021:
13:00 - 14:30

Session Chair: Andrea Gasparella, Free University of Bozen - Bolzano
Session Chair: Damien Picard, KU Leuven
Location: Concert Hall - Kamermuziekzaal
't Zand 34, Bruges

External Resource: Click here to join the livestream. Only registered participants have received the access code for the livestream.
Show help for 'Increase or decrease the abstract text size'
Presentations
13:00 - 13:18

Mapping the gap in user-related building performance simulation models

Ardeshir Mahdavi, Veselina Bochukova, Christiane Berger

TU Wien, Austria

Aim and Approach

(max 200 words)

Recently, increasing attention is being paid in the building performance simulation research community to the quality and resolution of representations of building users in simulation models (Mahdavi and Tahmasebi 2019). This is reflected in a host of publications and projects with highly diverse starting points, approaches, and results (Yan et al. 2017, Berger and Mahdavi 2020). Whereas such diversity can be productive and fruitful, it may also involve redundancy and lack of strategic orientation. In this context, it is beneficial to reexamine this subject from two complementary directions. One ontological direction, characterized here as "top-down", pursues the required format and dimensions of a generalized representation of building users that could be distilled down as needed to cater for the informational requirements of specific applications. The second direction, which can be seen as "bottom-up", starts by reverse-engineering the occupant-specific input templates of common performance simulations in view of any existing shared features and structure. We suggest that the pursuit of these two directions reveals a discontinuity or gap, indicating that there is not yet a seamless path leading from occupant-centric ontologies to occupant-related model input requirements of common performance simulation tools.

Scientific Innovation and Relevance

(max 200 words)

The building information modeling research and development community has become increasingly cognizant of the following circumstance: To be truly effective, building information models must go beyond the static representation of buildings' constituent physical components (Mahdavi 2020). Rather, the time-dependent dynamics of processes associated with the design, construction, and operation of buildings must be taken into consideration. A major class of such processes involve the patterns of occupants' presence and behavior in buildings. A comprehensive solution for the respective representational challenges must go beyond ad hoc amendments to existing simulation input routines. Toward this end, a seamless transition from a comprehensive ontology of building users down to specific input schema tailored for individual simulation application would be desirable, but currently hampered to a representational discontinuity. The aforementioned concurrent top-down and bottom-up inquiries can help map this gap and hence suggest approaches to close it.

Preliminary Results and Conclusions

(max 200 words)

The work thus far has led to the definition of a general schema that captures the main dimensions of occupant-related information. These include physical data pertaining to position and movement, physiological data pertaining to state of metabolism and adaptation, cognitive data pertaining to formation of impressions and value attributions, as well as event-based data pertaining to human-building interaction processes. The result suggests that, in order to consistently structure the schema, underlying foundational theories are needed that capture building users' relevant patterns of presence and behavior. Moreover, the bottom-up reverse-engineering of existing simulation applications reveals the potential and challenges toward seamless derivation of locally tailored input information from ontologically structured sources of occupant-related information.

Main References

(max 200 words)

Berger, C., and Mahdavi, A. (2020): Review of current trends in agent-based modeling of building occupants for energy and indoor-environmental performance analysis. Building and Environment, 173; 106726.

Mahdavi, A., and Tahmasebi, F. (2019): People in building performance simulation. Building Performance Simulation for Design and Operation - Expanded Second Edition. Hensen, J., Lamberts, R. (Ed.); Routledge, New York, ISBN: 978-1-138-39219-9; pp. 117 - 145.

Mahdavi, A. (2020): Bringing HIM closer to HER. Keynote. Proceedings of SIMAUD: Symposium on Simulation for Architecture and Urban Design. Vienna (Online), 25-26 May 2020. ISBN: 978-1565553712.

Yan, D., Hong, T., Dong, B., Mahdavi, A., D´Oca, S., Gaetani, I., and Feng, X. (2017): IEA EBC Annex 66: Definition and simulation of occupant behavior in buildings. Energy and Buildings, 156; pp. 258 - 270.



13:18 - 13:36

The impact of occupancy prediction accuracy on the performance of model predictive control (MPC) in buildings

Tao Yang, Fisayo Caleb Sangogboye, Krzysztof Arendt, Konstantin Filonenko, Jonathan Dallaire, Mikkel Baun Kjærgaard, Christian Veje

Center for Energy Informatics, University of Southern Denmark, Denmark

Aim and Approach

(max 200 words)

This paper aims to investigate the impact of occupancy accuracy on the performance of MPC-based building control. As an experimental setup, a grey-box model representing the heating, ventilation, and air conditioning (HVAC) system in a case study building is developed and calibrated. Subsequently, a number of 3D stereo-vision cameras is deployed to obtain accurate occupancy counts. Based on the obtained measurements, a data-driven model for predicting occupancy count in simulation is developed. In the evaluation, two MPC-based building controllers with different occupancy accuracy are compared based on multiple shooting optimization algorithm. The first of the compared models uses occupancy estimates from the deployed camera, while the second uses occupancy prediction from the data-driven model. The two MPC-based building controllers are further compared with a conventional rule-based controller in terms of energy consumption and indoor thermal discomfort.

Scientific Innovation and Relevance

(max 200 words)

Globally, buildings are responsible for nearly 40% of total energy consumption among other sectors [1]. MPC is a promising and widely investigated strategy employed in buildings to reduce energy consumption while maintaining thermal comfort. While there exists a wide range of parameters for MPC-based building controls, the accurate estimation of occupancy in buildings constitute a major factor for achieving considerable ambient comfort and energy saving within buildings [2-5]. This paper goes beyond simple analysis of the occupant presence and applies a data-driven approach to accurately predict occupant count and subsequently studies the impact of occupancy accuracy on MPC performance, which contributes to better analyze relationship between occupancy information and MPC performance in buildings.

Preliminary Results and Conclusions

(max 200 words)

The developed grey-box model yields good accuracy of capturing thermal dynamics of the system. Using occupancy estimates from the deployed cameras, an MPC-based controller in contrast to a rule-based controller demonstrates better indoor thermal comfort and higher energy consumption due to that MPC prioritizes thermal comfort (hard constraint) over energy. Comparing MPC with two different occupancy accuracy, occupancy predictions with low accuracy can lead to lower energy consumption at the expense of thermal comfort violations. When increasing optimization horizon, MPC-based controller with more accurate occupancy prediction shows larger energy consumption and improved thermal comfort. Besides, the negative influence of prediction error can be partially mitigated by adopting longer optimization horizons.

Main References

(max 200 words)

[1] Cao, Xiaodong, Xilei Dai, and Junjie Liu. "Building energy-consumption status worldwide and the state-of-the-art technologies for zero-energy buildings during the past decade." Energy and buildings 128 (2016): 198-213.

[2] Amin Mirakhorli and Bing Dong. Occupancy behavior-based model predictive control for building indoor climate - a critical review. Energy and Buildings, 129: 499–513, 2016.

[3] Justin R Dobbs and Brandon M Hencey. Model predictive hvac control with online occupancy model. Energy and Buildings, 82: 675–684, 2014.

[4] Frauke Oldewurtel, David Sturzenegger, and Manfred Morari. Importance of occupancy information for building climate control. Applied energy, 101: 521–532, 2013.

[5] Siddharth Goyal, Herbert A Ingley, and Prabir Barooah. Occupancy-based zone climate control for energy-efficient buildings: Complexity vs. performance. Applied Energy, 106: 209–221, 2013.

[6] K. Arendt and C. Veje, “MShoot: an Open Source Framework for Multiple Shooting MPC in Buildings,” in 16th IBPSA International Conference and Exhibition Building Simulation 2019, Rome, 2-4 September, 2019, 2019.

[7] Antoine Garnier, Julien Eynard, Matthieu Caussanel, and Stéphane Grieu. Predictive control of multizone heating, ventilation and air-conditioning systems in non-residential buildings. Applied Soft Computing, 37: 847–862, 2015.



13:36 - 13:54

Consequences-based graphical model for contextualized occupants’ activities estimation in connected buildings

Huynh Phan1, Thomas Recht1, Laurent Mora1, Stéphane Ploix2

1I2M Bordeaux, University of Bordeaux, CNRS, Arts et Metiers Institute of Technology, Bordeaux INP, F-33400 Talence, France; 2G-SCOP, Grenoble Institute of Technology, UMR CNRS 5272, 46 Avenue Felix Viallet, 38031 Grenoble Cedex 1, France

Aim and Approach

(max 200 words)

Occupants' activities are addressed as important factors resulting in the discrepancy between simulated and actual energy consumptions in residential buildings. Many models with statistical data are proposed to better estimate the activities of occupants than conventional approaches. However, these models are hybrid of the statistical data, which contains the information of many dwellings with different habits and characteristics, and not entirely sufficient to represent contextualized activities in a particular household. To solve this problem, data-driven approaches with machine learning techniques are proposed to estimate the activities based on the data measured from on-site sensors. However, they are black-box approaches and difficult to understand. It is preferable to propose an understandable model, which is able to estimate and evaluate the contextualized activities in a particular household. In this contribution, a general approach model taking into account a particular context is proposed to estimate and predict the occupants’ activities in a specific household. Specifically, many sensors (CO2, temperature, motions, etc.) are installed to capture the data in the household. Then, combining with the context information, the necessary features are extracted from measured data and are used to build a consequences-based Bayesian Network, which is understandable, and flexible for the contextualized activities estimation.

Scientific Innovation and Relevance

(max 200 words)

Recently, statistical approach is the most popular approach for occupants’ activities estimation in buildings. [1,6] used statistical models to estimate the activities using the characteristics of both buildings and households. Though, their models are hybrid of statistical data of different dwellings and not sufficient to estimate and evaluate the contextualized activities in a particular household. To deal with it, [2,7] applied data-based models with machine learning techniques to estimate activities using the measurement data. However, they do not provide understandable outputs and are not adaptable to context’s changes. Otherwise, [3,4] proposed agent-based models to simulate the activities of windows/doors. An agent perceives the space’s conditions and executes activities to achieve its predefined comfort. However, agent-based models are too complex with many agents. Besides, [5] built Bayesian Network (BN) to estimate the activities of doors based on the measurement data and expert knowledge. BN is an understandable, easy adaptable model for activities estimation in a particular household. However, all studies focus on simple activities (actions of windows/door, presence, etc.) and the structure of the BN is difficult to define. This study proposes a BN with an expert structure for contextualized activities estimation from measurement data, in a specific household.

Preliminary Results and Conclusions

(max 200 words)

A general approach has been proposed in this contribution to estimate the daily activities in a residential building. Dynamic Time-Series Clustering and expert discretization techniques are used to extract features from variables while Information Gain is used for the necessary features selection. Consequences-based Bayesian Network is used as an estimation model based on an expert structure, which is determined by the consequences of the activities. Finally, cross-validation and F1-score are used to validate the proposed model in the testing data. The proposed model was used to estimate some activities such as cooking breakfast, cooking lunch, washing dishes, etc. The testbed is a detached house in France, which includes 5 household members and 11 rooms. The data was collected from numerous installed sensors of ambient, power, motion, windows/doors contacts, etc. The labels of activities were collected for two months by a self-developed mobile application. In this contribution, the activity cooking lunch in the kitchen is presented. Data set covers 45 weekdays from 09/12/2019 to 30/02/2020 (labels collection period). Results show that cooking lunch is mainly linked to the patterns of the usages of frequently involved appliances (microwave, toaster). Cross-validation is used and F1-score is approximately 91%.

Main References

(max 200 words)

1. Dorien Aerts. 2015. Occupancy and Activity Modelling for Building Energy Demand Simulations, Comparative Feedback and Residential Electricity Demand Characterisation. PhD Thesis, Vrije Universiteit Brussel.

2. Alaa Alhamoud, Pei Xu, Frank Englert, et al. 2015. Extracting Human Behavior Patterns from Appliance-level Power Consumption Data. Wireless Sensor Networks, Springer International Publishing, 52–67.

3. Yoon Soo Lee and Ali M. Malkawi. 2014. Simulating multiple occupant behaviors in buildings: An agent-based modeling approach. Energy and Buildings 69: 407–416.

4. Khadija Tijani, Ayesha Kashif, Stéphane Ploix, Benjamin Haas, and Julie Dugdale. 2015. Comparison between purely statistical and multi-agent based ap-proaches for occupant behaviour modeling in buildings. arXiv:1510.02225 [cs].

5. Khadija Tijani, Quoc Dung Ngo, Stéphane Ploix, Benjamin Haas, and Julie Dugdale. 2015. Towards a General Framework for an Observation and Knowledge based Model of Occupant Behaviour in Office Buildings. Energy Procedia 78: 609–614.

6. Urs Wilke. 2013. Probabilistic Bottom-up Modelling of Occupancy and Activities to Predict Electricity Demand in Residential Buildings. .

7. Suyang Zhou, Zhi Wu, Jianing Li, and Xiao-ping Zhang. 2014. Real-time Energy Control Approach for Smart Home Energy Management System. Electric Power Components and Systems 42, 3–4: 315–326.



13:54 - 14:12

Multi-Agent based simulation of human activity for building and urban scale assessment of residential load curves and energy use

Mathieu Schumann1,5, Quentin Reynaud2, Nicolas Sabouret3, François Sempé6, Benoit Charrier1,5, Jérémy Albouys1,3,4,5, Yvon Haradji1, Christian Inard4,5

1EDF R&D, Palaiseau, France; 2QRCI, Clermont-Ferrand, France; 3LIMSI-CNRS, Univ. Paris-Sud, Univ. Paris-Saclay, Orsay, France; 4LaSIE, CNRS, La Rochelle Université, La Rochelle, France; 54evLab, EDF R&D, CNRS, LaSIE, La Rochelle Université, La Rochelle, France; 6FSCI, Paris, France

Aim and Approach

(max 200 words)

Because of the crucial impact of human activity on energy consumption (Janda, 2011), many building and urban energy models integrate human activity modeling. An increasing number of research works use Time Use Surveys (Chenu et al, 2006) and stochastic person-based approaches to generate occupancy profiles, window uses, or activity chronograms (Baetens et al, 2016). The goal of this paper is twofold: 1/ to discuss the limits of the existing state-of-the-art approaches on human activity modeling, and 2/ to present our modeling approach: an agent-based, individual-centered simulation of human activity, taking individual decisions and interactions between individuals into account to produce activity diagrams that are consistent at the household scale. Associated with a population generator, and by linking the simulated activity with appliance use, one of the main applications is the calculation of residential households load curves.

Our work puts the simulation of human activity and everyday life decision-making processes at the center of energy consumption, thermal comfort or indoor air quality assessments in households. This approach has its roots in ergonomics studies, which demonstrated the ability of multi-agent systems to simulate a realistic human activity, using innovative validation methodologies such as participatory simulations with real households (Haradji et al, 2012).

Scientific Innovation and Relevance

(max 200 words)

The analyzed approaches appear to be insufficient to account for the dynamics of individual and collective activity and its impact on energy consumption (Happle, 2018), particularly as households and energy communities become more actively integrated in smart grids and involved in flexibility, collective self-consumption or energy exchange schemes. Such configurations involve individual and collective decision-making, as well as the management of new appliances such as electric vehicles.

We suggest that the co-simulation of autonomous agents, whose activities are built upon statistical data (e.g. Time Use Surveys), with building and appliance models (Plessis, 2014) can effectively tackle these new challenges. We propose a fine-grained, 1 minute time step household simulation, able to perform population-wide yearly simulations. Our agent-based model can reach a satisfying level of diversity in human activities, and it supports interactions between autonomous agents and a changing environment (e.g. activity priorities, incentives, energy prices). This model accounts for emerging, reactive, adaptive, and collective behaviors. Moreover, it allows to deal with many human behavioral aspects (e.g. presence, heating adjustment, housing ventilation) within the same decisional process, bringing consistency to the role of occupant in energy assessments.

Preliminary Results and Conclusions

(max 200 words)

We present the agent-based SMACH platform (SiMulation of human Activity and Consumption in Households) (Haradji et al, 2018). We show how the modeling of occupants as intelligent autonomous agents accounts for the individual and collective dynamics of daily life and the related energy consumption, modeled as the result of the interactions between agents and the household’s electrical appliances. We demonstrate how the latest developments of the SMACH simulation platform, combining population synthesis, simulation of individual and collective activities as well as electric mobility, answer some of the challenges related to the role of occupants in energy consumption (Happle, 2018). Especially, we show that multi-agent modeling offers a high degree of modularity due to the internal capabilities of the agents to organize themselves, plan their day, or react to events or changes in the environment. This approach is also explicit, and makes it possible to explain why, when, how and with whom the simulated energy consumption is performed. We illustrate the validation approaches and variety of uses of this model in energy questions and engineering applications such as load curve calculation, at an individual household scale as well as at the population scale.

Main References

(max 200 words)

Janda, Kathryn B. (2011). Buildings Don’t Use Energy: People Do. Architectural Science Review 54 (1): 15–22

Chenu, A. and Lesnard L. (2006) Time Use Surveys: A Review of Their Aims, Methods, and Results. European Journal of Sociology 47 (3): 335–59.

Baetens, R. and Dirk S. (2016). Modelling Uncertainty in District Energy Simulations by Stochastic Residential Occupant Behaviour. Journal of Building Performance Simulation 9 (4)

Haradji Y., Poizat G., Sempé F (2012) Human Activity and social simulation. Proceedings of the 4th Advances in Human Factors and Ergonomics Conference, San Francisco, California, USA

Plessis, G., Edouard A., Haradji, Y. (2014). Coupling Occupant Behaviour with a Building Energy Model - A FMI Application. 10th International Modelica Conference 96:321–26 Lund, Sweden

Haradji Y. et al, (2018) From modeling human activity to modeling for social simulation: between realism and technological innovation. Activités 15-1

Happle G., Fonseca J. A. and Schlueter A. (2018) A review on occupant behavior in urban building energy models. Energy and Buildings, vol. 174. Elsevier Ltd, pp. 276–292



14:12 - 14:30

The formulation of a reference load curve to measure energy flexibility

Muhammad Salman Shahid1, Benoît Delinchant1, Béatrice Roussillon2, Frédéric Wurtz1, Daniel Llerena2

1G2Elab, 21 Rue des Martyrs, CS 90624, 38031 Grenoble CEDEX 1, France; 2GAEL, 1241 Rue des Résidences, 38400 Saint-Martin-d'Hères, France

Aim and Approach

(max 200 words)

Energy consumers have a degree of choice to implement indirect energy flexibility to mitigate network congestion during the intermittence of renewable production. It is essential to measure the impact of each alert for each consumer. This abstract presents a study performed to compare the different methods for formulating a reference load curve for the residential energy consumers. Hypothetically, this reference load curve gives the habitual energy consumption pattern of residential consumer. The aim of creating a reference load curve is to visualize and measure the degree of deviation of consumption load curve from habitual energy consumption load curve of a consumer. The image of reference load curve superposed on consumption load curve is sent to the consumers as part of the feedback, so that they can watch the impact of their efforts as well. For this purpose, certain statistical methods (mean, Kernel Density Distribution) and naïve methods are explored, whereas the advanced methods (RF regression, Neural networks) are in the process of study. The so forth studied methods are analyzed through an indicator that verifies the under-estimation or over-estimation of the reference load curve (to be discussed in detail in the proposed paper).

Scientific Innovation and Relevance

(max 200 words)

Hypothetically, the habitual energy consumption pattern of a residential customer can be visualized in the form of a load curve (hereafter referred as reference load curve) for a standard day. For a particular day, the deviation of consumption load curve from reference load curve gives the measure of the effort made by the residential consumer to implement indirect flexibility. In case of peak shaving, a good effort can be observed if the consumption load curve is under reference load curve. The result should be opposite in the case of load shifting for the period of time to which the load is shifted. For the purpose of experiment, two types of alerts are defined. Orange alert demands the households to implement peak shaving between 6 PM and 8 PM on alert day. Green alert demands the households to implement load shifting from evening to afternoon (between noon and 3 PM). The reference load curve is used to measure the impact of nudge signal on each household for each alert. This in consequence measures the effectiveness of nudge tool to implement indirect energy flexibility in residential sector.

Preliminary Results and Conclusions

(max 200 words)

For half-hourly sampled consumption load curve, one method is to take the mean of historical data for each of the 48 timestamps of day. Another statistical method is to take the peak value of the Kernel density estimation for each of the 48 timestamps of the day. Hypothetically, this peak value is the most probable value of energy consumption for the given half hour timestamp. However, these methods are highly susceptible to season variation and therefore gives less accurate reference load curve.

One naïve method is to take an average of the consumption load curve of day "D-1" with the consumption load curve having maximum energy consumption among the curves of days "D-2" and "D-5" (only weekdays). This reference load curve is used for orange alerts. The vice versa of this method is also formulated for green alerts. Both these reference curve keeps the effect of the near past historical consumption, therefore maintaining the effect of temperature and seasonality. These naive methods are devised to introduce a bias, so that it nudges the households makes a better effort in future. The advanced methods are still in the process of study.

Main References

(max 200 words)

Albadi, M. H., and E. F. El-Saadany. 2007. “Demand Response in Electricity Markets: An Overview.” In 2007 IEEE Power Engineering Society General Meeting, 1–5.

Bivas, Pierre. 2011. “La production d’effacement : comment offrir des économies d’électricité à des millions de foyers.” Le journal de l’ecole de Paris du management n°90 (4): 8–14.

Hatton, Leslie, and Philippe Charpentier. 2014. “Système électrique français : estimation de l’effacement des clients résidentiels,” 14.

Lesgards, Valérie, and Laure Frachet. 2012. “La Gestion de La Demande Résidentielle d’électricité: Retour Sur 30 Ans d’expérimentations Mondiales.” La Gestion de La Demande Résidentielle d’électricité: Retour Sur 30 Ans d’expérimentations Mondiales, no. 607: 162, 164, 192-210 [21 p.].

Neenan, B. 2009. “Residential Electricity Use Feedback: A Research Synthesis and Economic Framework,” 126.

Thaler, Richard H., and Cass R. Sunstein. 2008. Nudge: Improving Decisions about Health, Wealth, and Happiness. Nudge: Improving Decisions about Health, Wealth, and Happiness. New Haven, CT, US: Yale University Press.



 
Contact and Legal Notice · Contact Address:
Privacy Statement · Conference: Building Simulation 2021
Conference Software - ConfTool Pro 2.6.143+TC
© 2001–2022 by Dr. H. Weinreich, Hamburg, Germany