Subtask 1: Multi-aspect environmental exposure, building interfaces, and human behaviour
- Multi-aspect comfort and behaviour models
- State of the art assessment
The scope of Subtask 1 is: (1) multi-aspect environmental occupant exposure and its impact on behaviour and comfort and (2) the usability and comfort and energy implications of building interfaces. Existing models of human comfort, perception, and behaviour are commonly formulated for single-domain environmental exposure circumstances (e.g., thermal, visual, aural). We will conduct a comprehensive overview of existing theoretical and experimental work relevant to multi-domain models and provide input to Subtask 4 regarding relevant information for the development of their database (see 3.4 Activity 2b).
- A platform for multi-disciplinary human comfort and behaviour theories
Annex 66 participants and outcomes identified the key importance of interdisciplinary collaboration, especially between technical/engineering disciplines and human sciences (e.g., psychology, sociology), as a precondition for the development of next generation behavioural models. To start addressing this issue, two parallel efforts are needed: First, a structured overview of past multi-disciplinary efforts (involving both natural and human sciences) toward formulation of human comfort and behaviour models must be obtained. Second, a common conceptual framework (involving terminology and methodology discourse) must be established to facilitate the integrated consideration of both technical and social sciences in model development. This will result in a documentation of past multi-disciplinary theory building efforts in relevant fields (i.e., human comfort, occupant behaviour) as well as a common terminology/methodology document to support collaborative multi-disciplinary model developments. This subtask will aim to involve scientists from sociology/psychology and engineers focused on the indoor environment are needed for this task, particularly from the domains of acoustics, visual comfort, and indoor air quality.
- Buildings interfaces and human behaviour
- Systematic review of state of the art
To encourage user behaviour patterns that are desirable from the operational standpoint (i.e., patterns that can bring about desirable indoor environmental conditions while meeting the operational efficiency criteria), a better understanding of interfaces to control-relevant building features and systems (and corresponding occupant behaviours) is critical. There are many building interfaces that can have either a positive or negative impact on energy use or occupant comfort; however, many interfaces are poorly understood in terms of occupant behaviour and resulting energy impacts or comfort. In a systematic review, human-building interfaces will be explored and categorized. An interdisciplinary group is envisioned, including human factors experts, social scientists, designers, architects, and engineers.
- A systematic theory of the interface attributes of buildings and their systems
Based on the review and framework in 2.a and 2.b, new research studies will be initiated to explore the identified gaps of knowledge. This will be done through experimental studies in laboratories, field study campaigns, studies based on both surveys and qualitative methods (i.e. photographs, interviews, etc.), as well as virtual experiments via agent-based modeling approaches. The research will not be restricted to the physical aspects of interface but will also include aspects related to interface controls and behaviour. The data collected with these studies will also be used where possible in Subtask 2, e.g. by using data analytics to better understand occupant interactions with interfaces for internet-enabled thermostats. The outcome will be an enhanced knowledge base surrounding the acceptance and usability of interfaces and their effect on human comfort and behaviour together with a best practice guide for interface design. In addition, information will be shared with Subtask 4 to support the decision process in the development of occupant-centric control algorithms.
- Method and Studies
- New field and laboratory studies on human comfort and behaviour with respect to multi-aspect environmental variables
Based on the platform and framework developed in parts 1a and 1b, new field and laboratory studies will be conducted together with qualitative studies to close knowledge gaps identified. These studies are not restricted to Annex members; researchers in the field will be informed of the developed framework and invited to design studies based on the framework. Thereby, the framework will serve as a strong recommendation and guideline, but will not be enforced. Due to the number of influencing variables and their possible interactions, this subtask will focus on controlled experimental studies, which will permit the control of single and multiple aspects/variables, as well as with their respective interactions to establish cause-effect relationships. Such studies can be conducted in well-equipped laboratories and living labs. Moreover, the potential of an agent-based computational platform for the design and implementation of virtual experiments in occupant behaviour field well be studied and documented. Such a platform can aid the preparation of real (labor and field) studies and – in return – benefit from them in terms of model calibration. The outcome will be a compilation of new findings regarding cause-effect relations of interactions between distinct parameters of the indoor environment on human comfort and behaviour. These findings can be implemented in occupant behaviour models used for performance-based design process in Subtask 3.
- New research on building interfaces
Based on the review and framework in 2.a and 2.b, new research studies will be initiated to explore the identified gaps of knowledge. This will be done through experimental studies in laboratories, field study campaigns, studies based on both surveys and qualitative methods (i.e. photographs, interviews, etc.), as well as virtual experiments via agent-based modeling approaches. The outcome will be an enhanced knowledge base regarding the acceptance and usability of interfaces and their effect on human comfort and behaviour.
Subtask 2: Data-driven occupant modeling strategies and digital tools
The focus of Subtask 2 is to investigate and develop methodologies and tools for data-driven Occupant Presence and Action (OPA) modelling. The developments in sensing modalities and computing platforms enable many new sensing technologies and data sources on OPA. The wealth of data opens new opportunities for extracting knowledge through data-driven modeling of OPA. In terms of data-driven methods, Subtask 2 will focus on the many opportunities with machine learning techniques including supervised and unsupervised learning for both classification, regression and clustering problems. Utilizing these opportunities creates new models and information relevant for generating new knowledge on multi-aspect environmental exposure, building interfaces, and human behaviour (Subtask 1), occupant-centric building design (Subtask 3) and operation (Subtask 4). Thereby the data-driven outcomes of this subtasks will provide recommendations for the occupant modeling methodologies of Subtask 1, 3 and 4, e.g., for calculating new schedules or models based on the real conditions observed in buildings, data-driven analysis of the performance design versus the built, model predictive controls for buildings and fault detection and diagnostics.
- Develop a Novel Occupant Data Collection Approach for OPA
- Assessment of the state of the art
Work in Annex 66 reviewed the properties of sensor modalities and general protocols for collecting OPA data using sensors. The general focus was on optimizing the accuracy of the collected data. However, in any deployment trade-offs have to be made among accuracy, privacy invasiveness and the cost of data collection (e.g. building installed versus tenant/occupant owned sensors). Therefore, this activity will conduct a literature review on the above-mentioned trade-offs made by current state-of-the-art sensor-based methods.
- Collect data examples and improve methods for ground truth annotation
The activity will collect data examples with sensors of different qualities including established and novel sensor modalities. The data collection will consider the trade-offs in previous work on accuracy, privacy invasiveness and cost and compare the trade-offs effect on the data. To support the collection the activity will develop novel means for collecting annotation labels for in situ OPA data that enable more accurate and fine grained labeling.
- Develop novel data collection methods
The activity will develop examples of data collection methods that go beyond state-of-the-art data-driven methods in how they are able to balance data quality, privacy invasiveness and cost of data collection. The methods will cover opportunities for fusing sensing modalities and online data streams. For evaluation, the activity will apply techniques for benchmarking data-driven methods including cross-validation and real-time validation.
- Investigate new Data-driven Methods for OPA
- Systematic review of existing methods
Previous work has documented the challenges in selecting the best data processing methods and knowing when data-driven methods are the right tool. To address this challenge, this activity will distill the elements of a data processing tool chain for data-driven models for OPA modeling. The work will cover different types of methods including: classification, regression, clustering and preprocessing of data, e.g., cleaning of data from faulty, misplaced or tampered sensors. This also includes metrics for quantifying the properties of the methods and identifying the limitations of data-driven methods.
- Develop anonymization methods for handling privacy invasiveness of occupant data
The wealth of OPA data creates new threats to the occupant in terms of the violation of the individual's right to privacy. The activity will address this challenge by developing new anonymization methods and guidelines for privacy handling in the processing of occupant behaviour data (e.g. based on the Findable, Accessible, Interoperable and Reusable (FAIR) principles and to implement the EU General Data Protection Regulation (GDPR)).
- Demonstrate novel approaches for using data-driven data processing for OPA modeling.
The aim of this activity is to research new approaches for data-driven models and associated data processing for OPA data. It will demonstrate, with other subtasks, how to apply approaches for extracting new insights from data that can contribute in new findings on OPA and how these new findings may support occupant-centric building design, simulation and controls. Distill application guidelines for data-driven modeling of occupant behaviour from in situ OPA sensing data. This also includes how to visualize the data.
- Develop a Platform for sharing Data-Driven Methods
- Community building for data-driven methods on OPA
To support building engineers and architects in successfully applying data-driven methods requires better access to methods and knowledge. This activity will develop an online platform for sharing data and methods for OPA data based on existing repository and data science platforms (e.g. via github and Jupyter Notebook). The activity will share examples of data and method implementations on the online platform based on the outcomes of the other activities. The activity will also analyze the limitations and good practices for reuse of data via the collaboration platform including ethics review issues across countries.
- Develop a metadata schema for OPA data
The activity will create a metadata schema for OPA data to enable consistent sharing and reuse of OPA data. This activity builds on the outcomes of Annex 66 in terms of ontologies.
- Data competition on OPA data
This activity will organize a data competition with open and industry data sets to foster community establishment based on the developed platform and benchmark methods under identical conditions. Benchmarking for data-driven methods under identical conditions is difficult if method implementations and data is not made available.
Subtask 3: Applying occupant behaviour models in performance-based design process
The focus of Subtask 3 is occupant-centric building design process. While occupant modeling and implementation of occupant models in simulation have matured considerably (particularly through Annex 66), application of occupant models into the design process is in its infancy – particularly in practice. Many opportunities are available to not only improve energy prediction but also better understand how occupants will respond to various architectural and controls concepts. It follows that building design can be greatly improved with regards to energy performance and occupant comfort. The advent of advanced occupant modelling opens many new opportunities to pursue novel design procedures, such as: robust design, design optimization and generative design, and integration with building information modeling (BIM). New methods will be explored to integrate occupant behaviour models in modern digital planning environments.
The goal of Subtask 3 is to develop systematic methods to better apply existing occupant models to achieve high-performance building designs with regards to comfort, usability, and energy performance. Moreover, Subtask 3 will make policy recommendations through both prescriptive and performance paths of building energy codes and standards. It will leverage relationships with organizations and associations such as the ASHRAE (American Society for Heating, Refrigerating, and Air-conditioning Engineers) Multidisciplinary Task Group on Occupant Behaviour in Buildings (MTG.OBB). To this end, ST3 will take inputs from ST1 and ST2 on the state-of-the-art multi-aspect occupant behaviour models and data-driven modeling methods. However, given the large body of existing research on occupant behaviour modeling, ST3 activities are not dependent on the outcomes of the aforementioned subtasks.
Subtask 3 will also include approximately three to four focused case studies to evaluate and disseminate new occupant modeling approaches in design practice. In this regard, ST3 researchers will adopt two approaches depending on the possible involvement in the buildings’ design process: a) Contributing to a future building design by providing the design teams with insights on the needs and behaviours of occupants; b) Investigating existing buildings (ideally with detailed measurements) in a retrospective manner to identify alternative design possibilities, which could have been resulted from the use of detailed occupant models in the design process. ST3 will document the case study buildings (e.g., drawings, occupant interfaces, occupancy data, sensors, etc.) and the process and outcome of the occupant-centric performance-based design space explorations (e.g., building models, associated parametric and optimization-based studies, performance indicators, design alternatives, etc.).
Because there are strong linkages between building design and operation and controls, ST3 will work closely with ST4 to ensure that the output (e.g., design guidelines) from ST3 properly encourage such integration. The case studies are expected to include aspects of both design and controls and with thus reinforce the integration of building design and controls.
- Review of codes/standards, practices, and exemplary projects involving performance-based simulation-assisted design, with a focus on representation of occupants
- Develop methods and guidelines to choose the most appropriate occupant modeling approaches across scales, building types, design stages, climates, and modeling objectives
- Develop, test, and document new simulation-based design procedures, including robust design, parametric performance-based design, design optimization, etc.
- Make recommendations for occupant-related prescriptive and performance paths of building energy codes and standards
- Develop effective ways for communicating occupant-related assumptions between stakeholders (owner, modeler, designer, architect, engineers, etc.) in new buildings
- Test the developed methods in a small number (3-4) of real building design projects with stakeholder involvement
Subtask 4: Development and demonstration of occupant-centric building controls
The focus of Subtask 4 is to develop and demonstrate occupant-centric building controls (OCC). The activities within Subtask 4 will reveal practical challenges regarding the implementation of OCC in existing buildings. The results from focused case studies will highlight and quantify potential improvement in occupant comfort and energy savings reductions through OCC. These case studies will identify future research and development needs in OCC. They will guide practitioners in implementing occupant-centric indoor climate control strategies. They will help identify appropriate building automation system sensor/actuator configurations. The case study buildings will be selected starting with office buildings. However, the focus may extend to other building types such as academic, retail, and health care buildings.
- Current state and future challenges in occupant-centric operating strategies
Current building automation systems are mostly operating in static and inefficient ways that mainly neglect the occupants’ preferences, habits, and schedules. It is a consequence of a variety of aspects such as lack of sensor equipment, coarse spatial control zones or lack of building interfaces. Consequently, the results of this activity will be a review report/paper critically assessing the existing OCC algorithms. The review paper will reflect the state-of-the-art in this area and highlight possible ways of integrating OCC algorithms into the building automation systems.
- Survey of occupant sensing technologies
A comprehensive review of the common occupant sensing technologies in existing buildings will be conducted. This contrasts activities of ST1 and Annex 66, where specific technologies will be/were reviewed. Through interviews with facility and energy managers, and building operators, we will identify common occupant sensor configurations in automation and control networks – lighting, HVAC, information technology (IT), access, security, etc. Interviews will be administered in Europe, North America, and Asia. Through this exercise, the transferability of existing OCC algorithms (reviewed in 1a) to these sensor configurations will be identified.
- Development of occupant-centric control algorithms
- Selection of case study buildings and corresponding input and output parameters
The OCC algorithms from the literature and from ST2 will be modified so that their inputs/outputs would be compatible with common sensor configurations identified in 1.a. Moreover, recommendations for sensor configuration upgrades will be developed to ensure that the state-of-the-art OCC algorithms will become compatible with existing buildings. At least three case study buildings will be selected with different densities of occupant sensing technologies (high, medium, low). Here, buildings with high sensor density can have CO2 sensors, motion detectors, etc. and provide BAS-integrated interfaces so that occupants have a high level of control over their indoor environments. In contrast, buildings with medium sensor density may have submeters and only a subset of occupant actions can be monitored. Buildings with low sensor coverage may have bulk utility meters, and occupants may change their indoor climates through complaint calls only. Regarding the target function, the outcomes of 1b will be relevant.
- Database for OCC evaluation and data exchange
The sensor density of the case study buildings determines the type of control algorithms that can be applied for OCC. A structured database is needed, which provides detailed information on the building type, building automation system, applied OCC algorithms (e.g. stochastic, agent-based, etc.), sensor/occupant data required for the development and application of the algorithms as well as further relevant information that address multi-domain influences on the occupant’s comfort, behaviour and their linkage to the building interfaces defined within the scope of ST1. The platform must provide objective feedback on energy and comfort related outputs to be able to evaluate the effectiveness of the applied OCC algorithms. Furthermore, the database shall continue to exist beyond this Annex. It shall give all community members the opportunity to continuously improve their knowledge on OCC strategies.
- Demonstration and verification of the occupant-centric control strategies
- Integration of occupant-centric control strategies in the building automation system
The developed algorithms will be integrated to the building automation systems of the case study buildings. Moreover, working with ST3, the algorithms will be implemented within the building performance simulation models of various building archetypes and climatic conditions. The purpose of this simulation-based investigation is to explore the robustness of OCC algorithms outside of the case study buildings. Inputs from researchers and practitioners will be sought who have relevant experience regarding technical and non-technical obstacles that appeared upon integrating their algorithms. Results shall be summarized within a technical report and shall be made accessible via the data exchange platform developed in 2b.
- Testing phase
To be able to evaluate the new algorithms with respect to their target function (energy efficiency, human comfort etc.) the entire system shall be tested for a minimum period of one year in the case study buildings. This allows us to draw conclusions on the influence of seasonal fluctuations on the algorithms, their robustness and generalization of the results.
- Evaluation phase
This activity will start in the second half of 3b. Its main objective is to summarize the results of the entire subtask, make the “lessons learned” available for others and highlight promising types of algorithms. To assess the impact of the algorithms on the energy performance, baseline energy models will be trained with energy use data collected before and after the implementation of the OCC algorithms. To evaluate the impact of the OCCs on occupant comfort, changes in the frequency of thermal complaints will be studied. Alternatively, online surveys can be used to analyze changes in occupant satisfaction after implementation of the algorithms. The summary will be documented in a report/paper, and a guideline document for the implementation of occupant centric controls will be developed. Relevant information shall be also included in the database developed in activity 2b.