Jul 3, 2018 - occupant modelling for building codes and standards and design practice. ... modelling research, building energy codes, building design practice and ... building size, occupant domain, and building model objective; critically assessing ..... sources of uncertainty as well, such as weather, construction quality, ...
Roadmap for the advancement of occupant modelling for building codes, standards, and design practice
Date: Tuesday, July 03, 2018
Written by: William O’Brien, Mohamed Ouf, Burak Gunay, Sara Gilani, Tareq Abuimara, Aly Abdelalim Human Building Interaction Lab, Carleton University, Ottawa, Canada
Executive Summary This document provides a comprehensive roadmap for the advancement towards detailed occupant modelling for building codes and standards and design practice. It is comprised of a detailed introduction and definitions, followed by the standard roadmap components. A stakeholders’ workshop, which was held before the roadmap was created, is referred to throughout this document and is summarized in Appendix 2. First, detailed background information is provided on code compliance paths and design processes, in the context of occupant modelling. The state-of-the-art in occupant modelling methods and formalisms is introduced, with many definitions provided. Four traits of advanced occupant models are described, including dynamic, stochastic, data-drive, and agent-based. The numerous actors that influence building energy performance are described, including owners, managers, operators, janitorial staff, tenants, and finally the focus of this roadmap, occupants. Finally, the primarily occupant-related phenomena that affect building energy performance are summarized, including: occupancy (presence), lighting, office equipment, window shading devices, operable windows, thermostats, personal fans, clothing level, and metabolic rate. The International Energy Agency guide for technology roadmaps was followed closely to develop the roadmap towards advanced occupant modelling. It includes five stages: 1) goals and scope, 2) milestones, 3) gaps and barriers, 4) action items, 5) priorities and timelines. The overarching goal is to increase the accuracy of occupant modelling for improved occupant comfort, health, and productivity, and improved building performance, including energy, usability, and robustness. The scope of the roadmap was set to be limited to Canadian office buildings, with occupants as the actors. The milestones were divided into four categories, including: occupant modelling research, building energy codes, building design practice and modeller education, and building simulation tools. Seven gaps were identified, on the topics of: critical occupant-related domains, simplicity of current occupant modelling methods, outdated occupant schedules, contextual factors affecting behaviour, current building code requirements, occupant modelling in design practice, BPS tools. Key identified barriers include: unquantified value of detailed occupant modelling, challenges in communicating occupant assumptions, building codes and associated limitations, modelling complexity, and modeller education. In the context of the gaps and barriers that were identified, action items to address the milestones were established. Concrete recommendations for applied research, advancing building codes and design processes, and modeller education were made. Finally, priorities and timelines were established. Specifically, action items that are seen is particularly critical and could be achieved in the short term were identified. These include: systematically assessing the most critical and sensitive behaviours for all climate zones (and eventually all building types); making recommendations for the most appropriate occupant modelling approach as a function of building size, occupant domain, and building model objective; critically assessing the National Energy Code of Canada for Buildings (NECB) to identify weaknesses and opportunities; establishing systematic method and corresponding repository to collect occupancy, plug load, and lighting data for the purpose of updating NECB schedules; developing, demonstrating, and documenting new design workflows that use advanced occupant models; and finally, pursuing outreach and dissemination activities to communicate to industry that advanced occupant modelling is important, worthwhile, and economically justifiable.
Acknowledgements The authors of this roadmap are grateful for the generous funding and in-kind support from project partners, including Natural Resources Canada, Rowan Williams Davies & Irwin (RWDI), Autodesk, and National Research Council. The project was funded by Natural Resources Canada’s Energy Innovation Program – Clean Energy Innovation Project #812. This work would not have been possible without the input, ideas, and inspiration of the above partners. We also thank the participants of the stakeholders’ workshop for devoting over a day of their time to travel to Ottawa to provide valuable input to workshop. Participants were: Afaf Azzouz, Andrew Hicks, Azam Khan, Andrea Pietila, Daniel Heffner, Daniel Knapp, Gabriel Wainer, Guy Newsham, Ian Beausoleil-Morrison, Meli Stylianou, Phylroy Lopez, Rhys Goldstein, Ruth Urban, Sam Yip, Sebastián Carrizo, and Umberto Berardi. The following people contributed to the eSim 2018 conference paper version of this roadmap, upon which the current report is built: Aly Abdelalim, Tareq Abuimara, Ian BeausoleilMorrison, Juan Sebastián Carrizo, Ryan Danks, Sara Gilani, Burak Gunay, Ted Kesik, and Mohamed Ouf. The content of this report does not necessarily reflect the views of any of the above parties.
Contents Executive Summary ........................................................................................................................ 2 Acknowledgements ......................................................................................................................... 3 1
Background ............................................................................................................................. 1 1.1
Introduction to occupant modelling concepts .................................................................. 5
1.2
Occupant modelling strategies ......................................................................................... 7
1.3
Importance, risks, and rewards of modelling occupants accurately ............................... 10
1.4
Building agents ............................................................................................................... 13
1.5
Building modelling objectives, design phase, and spatial scale ..................................... 19
1.5.1
Building model objectives ...................................................................................... 19
1.5.2
Building design phase ............................................................................................. 22
1.5.3
Building spatial scale .............................................................................................. 23
2
Introduction to the Roadmap ................................................................................................ 25
3
Goals and Scope .................................................................................................................... 26
4
Milestones ............................................................................................................................. 28
5
4.1
Occupant modelling research ......................................................................................... 28
4.2
Building energy code ..................................................................................................... 28
4.3
Building design practice and modeller education .......................................................... 28
4.4
BPS tool development .................................................................................................... 29
Gaps and barriers .................................................................................................................. 30 5.1
Gaps ................................................................................................................................ 30
5.1.1
Gap 1: There is no consensus on which occupant-related domains to model ........ 30
5.1.2
Gap 2: Current occupant modelling methods are too simple .................................. 32
5.1.3
Gap 3: Current occupant schedule and density values are outdated ....................... 35
5.1.4
Gap 4: Contextual factors of occupant behaviour are not well understood ............ 36
5.1.5
Gap 5: Current building code procedural requirements are inadequate ................. 37
5.1.6
Gap 6: Occupant modelling in design practice is too simple.................................. 38
5.1.7
Gap 7: Occupant modelling features in BPS tools are limited ............................... 38
5.2
Barriers ........................................................................................................................... 39
5.2.1
Barrier 1: Value of occupant modelling.................................................................. 39
5.2.2
Barrier 2: Communication of occupant assumptions .............................................. 39
5.2.3
Barrier 3: Building codes and modelling complexity ............................................. 40
5.2.4 6
Barrier 4: Modeller education ................................................................................. 40
Action items .......................................................................................................................... 41 6.1
Applied research ............................................................................................................. 41
6.2
Building codes ................................................................................................................ 43
6.3
Design processes ............................................................................................................ 46
6.4
BPS tools and modeller education ................................................................................. 48
7
Priorities and Timelines ........................................................................................................ 49
8
Conclusions ........................................................................................................................... 50
Appendix 1: Glossary ................................................................................................................... 51 Appendix 2: Stakeholder workshop .............................................................................................. 54 References ..................................................................................................................................... 57
1 Background A growing body of building performance simulation (BPS) researchers view occupants as a leading cause of discrepancies between BPS predictions and reality. As such, hundreds of researchers (e.g., members of IEA EBC Annex 66) are working on occupant modelling and simulation to “close the gap” between predicted and measured building performance (Sun, Yan et al. 2014). However, another camp of BPS researchers and users has argued that the role of BPS is to predict relative performance (Hensen and Lamberts 2012; Burton 2014). The performance-based compliance path of the current National Energy Code of Canada for Buildings (NECB) (herein “the Code”), similarly to most building energy codes and standards, relies on annual building simulations to demonstrate that the proposed building achieves equal or better energy performance than a reference building (Sentence 8.4.1.2.(2)) (National Research Council Canada (NRC) 2015). See Figure 1 for summary of three code compliance paths that are typical of most building energy codes and standards. The performance compliance path implies that relative performance predictions are adequate for the Code. For instance, the Code allows great flexibility on occupant modelling assumptions, so long as they are the same for the reference and design building. For internal loads, Sentence A-8.4.2.7.(1) states “While any internal load values are permitted to be used, those default values should be used in the absence of better information.” Reasonably applied deviations from the defaults values are difficult to disprove for the modeller and code official alike. Standard 90.1-2016 (ASHRAE, 2016) takes the stricter stance that the local code authority must approve all deviations from standards; personal communication with modellers in many jurisdictions suggest that strictness varies widely in different locales. Code compliance paths
Prescriptive Strict requirements about envelope, lighting, HVAC, service hot water, electrical power systems
Trade-off Some flexibility based on demonstration that equivalents of prescriptive path are met
Performance Simulation-based approach that requires design building energy to not exceed reference building (based on prescriptive requirements)
Figure 1: Typical building energy code compliance paths
In contrast to the practice focusing on relative performance between the proposed and reference cases, a recent survey of international BPS tool users and developers (O’Brien, Gaetani et al. 2016) suggested that just over half of modeller participants do not agree with the statement “It does not matter if assumptions about occupants in BPS tools fully represent real occupants as long as occupants are represented the same way in all design variants.” Ouf and O'Brien (2018) demonstrated the consequence of allowing modellers to use customized occupant-related schedules: predicted energy savings from various energy efficiency measures could be greatly
manipulated by adjusting occupant-related schedules. For instance, energy efficient lighting is more beneficial to annual lighting electricity use reductions if lights are assumed to be on for more hours each day. Thus, the current occupant modelling approaches can have unintended consequences, such as leading design teams towards unwise decisions or operating conditions that increase absolute energy use. Gilani, O’Brien et al. (2016) further showed that simple occupant modelling assumptions may lead to significantly different optimal building designs than if occupants are more accurately characterized. Thus, absolute accuracy –rather than merely relative accuracy—is important, as is becoming evident by detailed studies (such as those above). Accuracy in this roadmap is defined as absolute (not relative) unless otherwise stated. However, it must be acknowledged that occupants and their behaviours are associated with a certain degree of uncertainty – particularly for unbuilt buildings. Of course, there are many other sources of uncertainty as well, such as weather, construction quality, and material properties. Meanwhile, occupant modelling is critical for simulation-based building design approaches because occupants are playing an increasingly important role in building energy performance. Current occupant modelling methods are unsuitable for evaluating many technologies (e.g., demand-controlled ventilation and occupancy-controlled lighting) and supporting design decisions (e.g., how many offices to put in the same control zone) that are important for designing high-performance buildings. The results of such evaluations are also critical for improving building codes, incentives, and other policy-related issues. Without accurate representation of occupant’s interactions with electric lighting, blinds, windows, equipment, and other systems in building simulation, two risks emerge: 1. 2.
Energy and other predicted performance results can be inaccurate, and, Inaccurate simulation predictions will mislead designers to make sub-optimal design decisions and government and code committees to develop inappropriate codes and incentives programs. Sub-optimal building code, incentives, and design decisions will leave building owners, operators and society at large with a long-term liability.
These concepts are expanded on later. Scope of this document and project This document is primarily limited to a theoretical critical analysis of the current state and recommended trajectories for occupant modelling in building design practice and building energy codes and standards. This roadmap report covers the first of four phases of the project: 1. Roadmap to advance standardized occupant modelling in simulation-based design practice and the building code 2. Technical feasibility study to evaluate the practical and theoretical implementation of occupant modelling into building performance simulation tools and building codes 3. Occupant modelling for building design – best practices guidebook for Canada 4. Case study to demonstrate advanced occupant modelling in a practical industry-oriented setting
The scientific literature and research effort on occupant behaviour in buildings is overwhelmingly focused on two building types: residential (all types) and office buildings. The reasons for this are simple; they are both common, reasonable easy for researchers to access and relate to, and typologies for which occupants play a significant role in building energy consumption. The primary occupants in these two building types –unlike many building types such as restaurants, schools, and retail—have the responsibility and ability to adapt building systems to maintain comfort. Moreover, residential and office occupants spend extended periods in these buildings are likely feel significant ownership over the space and entitlement to continuously comfort conditions. Home and office occupants use considerable energy and often making purchasing decisions about plug-in appliances and equipment. A dividing feature between office and residential occupants is that office occupants are typically not responsible for energy costs, whereas residential occupants normally pay for their electricity, natural gas, and other energy use. Figure 2 provides a relative and qualitative sense of the impact of various types of building occupants on building energy use. For instance, retail buildings tend to be most greatly impacted by the building operator/owner (who may also be an employee), whereas individual customers would have little impact on energy use. In contrast, buildings with heavy process loads (e.g., manually controlled laboratory equipment, hospitals, and factories) are more sensitive to the activities of individual occupants, who are likely to be employees. Non-medical offices (herein referred to simply as offices) tend to rely on building owner/manager for building level energy management (e.g., for central HVAC) and purchasing decisions (e.g., overhead lamps), whereas the employee occupants affect room, zone, or floor-level lighting, HVAC setpoints, adaptive opportunities (such as window and window shade openings), and office equipment. Regular staff/ Employees
Medical office Non-med. office Nursing home/care Hospital School facility Warehouse Food/beverage store Retail Building owner/ manager
Motel/ hotel Visitors/guests/customers/ patients/students
Figure 2: Relative impact of occupant types on building energy use, by building type.
For the above reasons and because the diversity of behaviours in offices tends to be lower than in homes, the focus of this document (and project) is occupant modelling in office buildings. Many
of the principles apply to all building types and the authors acknowledge that accurately modelling occupants in most building types is important. Office buildings in Canada Of the 766 million square meters of floor area among Canada’s 482,300 commercial and institutional buildings as of 2009, 20% of this floor area was comprised of non-medical office buildings, as shown in Figure 3.
Figure 3: Distribution of commercial and institutional building floor areas by use in million m 2. *Other includes entertainment, leisure and recreation buildings, shopping centres, colleges, and universities. (Natural Resources Canada (NRCan) 2012)
The average office building in Canada had an energy use intensity of 1.2 GJ/m2 (333 kWh/m2), consisting primarily of electricity and often natural gas. Figure 4 shows the energy end use break-down for all commercial and institutional buildings in Canada. Notably, space cooling energy and auxiliary equipment (e.g., office equipment) for the sector have increased four and six times faster than the total sector energy since 1990 (NRCan, 2011). The impact of occupants on the various energy end uses is discussed later in this report.
Figure 4: Energy end use break-down for all commercial and institutional buildings in Canada (NRCan, 2011)
1.1 Introduction to occupant modelling concepts There are often considered three major sources of uncertainty in building performance prediction: weather, building construction quality, and occupants. Weather can cause variations in energy performance by 10% or more between years, not including the effect of climate change which will have a larger effect (Hong, Chang et al. 2013). Construction quality affects leakiness of envelopes and ducts, thermal bridging through envelopes, and other parameters of that nature. Through a case study, Wang, Augenbroe et al. (2014) found poor construction quality (including infiltration characteristics) could double heating demand in a building. Weather and workmanship can be considered stochastic (random) and modelled as such – though on very different time scales. Occupants have been shown to affect building energy performance by several hundred percent, though this is often less for large buildings (Gram-Hanssen 2010; Haldi and Robinson 2011; Parys, Saelens et al. 2011). Compared to occupants, weather is generally considered to not by affected by individual buildings. Thus, the capturing the two-way dynamic interaction between buildings and occupants is unique and critical. Occupant behaviour is often described or modelled as both stochastic and dynamic. Stochasticity refers to the degree of randomness by which individual occupants or populations of occupants can be characterized and modelled. For instance, from the perspective of modelling heat gains from occupants, the presence duration of an occupant can be described stochastically, as can the metabolic rate of that person, since the exact characteristics of that person are unlikely to be known during design. These are parameters that we may estimate, but certainly cannot be precisely known, at the time of building design. Dynamic refers to the responsive and reactive nature of occupants. In contrast to traditional ways of modelling occupants –essentially as passive recipients of the indoor environment who
generate sensible and latent heat, carbon dioxide, and odours— the next generation of occupant models should better characterize the notion that uncomfortable occupants may adapt themselves or their environment to improve their comfort. Such adaptations include strategies such as putting on a sweater, opening a window, adjusting a thermostat, and relocating within a building, but could also include purchasing decisions (e.g., buying and installing a space heater or personal fan) and complaining to facilities management. The impact of these actions on energy could range from nil (e.g., clothing change) to significant and long-lasting (e.g., prompting an operator to make major permanent overrides). In general, we can categorize occupant behaviours into two categories: non-adaptive and adaptive. Non-adaptive behaviours are those for which improving comfort is not a primary motivator. Examples of non-adaptive behaviours include turning on office equipment, turning off lights, and using kitchen appliances. Adaptive behaviours, in contrast, are motivated by real or perceived potential improvement in comfort. Examples of adaptive behaviours include turning on lights, adjusting thermostats while present, and opening windows. Behaviours are often motivated by triggers, which may include a sudden change in indoor environmental conditions, or could be a function of long-term conditions leading up to the behaviours. Both types of behaviours can be described by actions or states. Action refers to the action taken by the occupant (e.g., opening a window), while state refers to the resulting state (e.g., window position). Behaviours may be motivated by measurable triggers (e.g., indoor air temperature, illuminance levels, time of day), but they are influenced by contextual factors (e.g., perception that other occupants in a shared space may not like the results of an adaptive action). Aside from adaptive and non-adaptive actions, occupancy is another important phenomenon to model. Occupancy refers to occupant presence and may be described by number of occupants in space, occupant density, or possibly state change in occupancy (i.e., arrivals and departures). Occupancy is relevant for BPS in that occupants generate heat, moisture, and effluents (e.g., CO2), but also because occupancy is normally a necessary condition for behaviour and occupancy is important to quantify when evaluating whether a space is comfortable. Traditionally, occupancy has been considered a necessary condition for behaviours (i.e., an action cannot be taken unless an occupant is present). To a growing extent, behaviours can be performed remotely via automated systems (e.g., WiFi-enabled thermostats or a pre-scheduled appliance). Many schedules and building controls decisions (e.g., fan schedule and vacancybased lighting controls) also depend on occupancy and occupancy assumptions. Occupancy may be considered an adaptive action if an occupant relocates to achieve greater comfort (e.g., leaving an overheating office in the shoulder season), though this effect is rarely considered in research or practice. Regardless, occupancy-related events are considered important predictors of actions. For instance, it has been found that occupants are much more likely to turn on a light upon arrival versus while they are in a space for extended periods (Reinhart 2004). Figure 5 summarizes the relationships between the aforementioned concepts.
Occupancy Arrival/departure events Triggers Non-adaptive trigger
Adaptive trigger
affects
predictor
predictor Contextual factors predictor
predictor
State (present or absent; number of occupants)
predictor
predictor Behaviour Action (change of state) affects System state
Figure 5: Conceptual structure of occupant behaviour (adapted from Schweiker, Carlucci et al. 2018)
1.2 Occupant modelling strategies Occupant modelling approach is a critical aspect of capturing the influence of occupants in BPS predictions. On the basis of the previously-defined characteristics of occupant model, such models be placed in four categories, as outlined in Table 1 and Table 2. Deterministic models perform exactly the same for every simulation run; whereas probabilistic (also known as stochastic) models are associated with randomness – either at the outset of a simulation or during a simulation (e.g., at each timestep or at each month). Static occupant models do not respond to specific predicted building conditions, whereas dynamic models incorporate the two-way interaction between the occupants and the building. By far the most prevalent occupant modelling practice is to use static deterministic models in the form of fixed schedules (e.g., those in ASHRAE Standard 90.1 and NECB). Fixed schedules offer the advantage of repeatability, simplicity, and transparency, all of which are attractive for mainstream building simulation for applications such as code compliance. However, there are two limitations of static deterministic models. First, fixed schedules neglect the two-way interactions between occupants and building. This means that benefits or consequences of certain design features are not properly characterized. For instance, fixed exterior shading can reduce the frequency of daylight glare that triggers occupants to close blinds and turn on lights to compensate (O’Brien and Gunay 2015). However, NECB states that blinds should be modelled as open unless they are motorized and automated. Second, static deterministic models to not provide insight into uncertainty and risk. For instance, the potential for robust design (whereby a building is designed to be more robust against uncertainty of occupant behaviour) and the potential to take calculated risks (e.g., size HVAC such that it will be adequate for 99% of expected occupancy levels (O’Brien, Abdelalim et al. 2018)). A subtler limitation of static deterministic models is that when a single set of occupant assumptions
is chosen, it tends to be chosen as conservative. For instance, the NECB occupancy schedules for offices assume 90% occupancy midday during weekdays, whereas offices tend to peak at 50 to 80% occupancy (Duarte, Van Den Wymelenberg et al. 2013). Table 1: Properties of current and potential future occupant modelling approaches in practice
Dynamic/static
Probability
1.0
Occupant(s) Heat gains Operating conditions
Performance
Building
Building performance
Probability
Advanced
Conventional
Stochasticity
Occupant(s) Operating conditions
Comfort conditions
Building
Adaptive actions Performance
Building performance
Table 2: Four categories of occupant models and examples
Deterministic
Static
A fixed model that does not respond to conditions during simulation (e.g., a schedule corresponding to the number of occupants in a room over a day) Examples of model in practice: NECB and ASHRAE Std. 90.1 standard schedules for lighting, plug loads, and occupancy Most suitable applications of model: building codes and standards primarily at the whole-building level; early design stages
Stochastic A model that is not the same for every simulation or timestep but that does not specifically change during a simulation due to conditions (e.g., an occupancy schedule that can be multiplied by a random number to scale it upwards or downwards) Example in practice: None known Most suitable applications of model: HVAC and renewable energy system equipment sizing
Dynamic
A model that represents the occupant in a consistent way but captures the two-way interaction between occupants and buildings (e.g., a model that defines whether or not an occupant will turn on a light based on a fixed illuminance threshold). Example in practice: IES LM 83 daylight metrics for LEED uses a fixed illuminance threshold to determine whether shades are open or closed Most suitable applications of model: codes and standards at the room and zone scales, particularly related to adaptive comfort systems
A model that responds to conditions (e.g. time of day, indoor temperature) with a degree of randomness (e.g., a model used to predict the probability that occupants will turn on the lights) Example in practice: None known; this is primarily research stage, for which it is very common Most suitable applications of model: detailed architectural design and equipment sizing at the room and zone levels
Though there are no known examples in common use, an approach to introduce stochasticity to occupant modelling, while relying on convention is stochastic schedules (i.e., stochastic static models) (O’Brien, Abdelalim et al. 2018). Such schedules would be parametric and have randomized parameters that are selected at the beginning of the simulation year, day, or some other frequency. Such schedules can provide insight about performance uncertainty, though would typically not be able to characterize the two-way interaction between buildings and their occupants. However, this limitation is generally acceptable for non-adaptive occupant-related domains, such as plug loads and occupancy. Deterministic dynamic models, for where there are several established examples in use, allow the two-way interactions between buildings and their occupants to be characterized but do so consistently. This means that the model outputs depend on building-specific inputs (e.g., indoor environmental quality (IEQ) that the design that defines the IEQ). Models that only depend on time or occupied state as an input are not considered dynamic in the current context. An example of a deterministic dynamic model is in Illuminating Engineering Society (IES) of North America Standard LM 83 imposes a modelling requirement that shades are closed whenever 2% of workplane analysis points receive greater than 1000 lux of direct sunlight (IESNA 2012). This means the building designer can use this model to help inform window size, placement, and type to minimize the amount of time that the shade is closed. Finally, the state-of-the-art in occupant modelling in the current literature (Yan, O’Brien et al. 2015; Yan, Hong et al. 2017) is focused on dynamic stochastic models. These models both characterize two-way interactions and do so stochastically, such that every simulation yields a different result. Such models have primarily been applied to the single room or small building scale and have spanned domains of light use, thermostat adjustment, operable window use, window shade/blind use, beverage consumption, clothing level selection, and desk fan use (Gunay, O’Brien et al. 2013).
Another distinction between the state-of-the-art in occupant modelling and typical approach is that the new ones are often agent-based models. Agent-based models treat occupants as individuals, who experience the indoor environment, have objectives, make decisions and take actions, and interact with each other. The leading models in the literature are quite simple, but can still be considered agent-based because they represent individual occupants how interact with the building (Gunay, O’Brien et al. 2013). Stochastic models, in turn, can take on numerous forms (Gunay, O’Brien et al. 2013; Gunay, O'Brien et al. 2015). The common ones are described here. Markov Chain models predict whether an action will occur (e.g., the occupant turns on the light) on the basis of the current state (e.g., the light is off) and as a function of one or more variables (e.g., workplane daylight illuminance). Markov Chain models are among the more common and have the advantage of not only predicting resulting state (e.g., whether the light is on or off) but also the event. The number of events can be used as a proxy of discomfort (Gunay, O'Brien et al. 2018). The underlying function of Markov Chain models is commonly logistic regression. Another common form of dynamic stochastic models is Bernoulli models. Bernoulli models predict the state of a particular system (e.g., light state or window position) rather than actions. While Bernoulli models are conceptually straightforward and somewhat easier than Markov Chain models to implement, they are not suitable for predicting the number of occupant actions. The main remaining model form described here is survival models. Rather than predicting state change or state in a given simulation time step, survival models predict how long a system will stay in a certain state. It has been used to predict whether occupants will turn off lights or office equipment as a function of how long they are absent. Figure 6 shows four illustrative examples for lighting for the four occupant behaviour modelling approaches.
Uncertainty bounds for static stochastic model
0
Time of day (hours)
24
1
Probability of light being turned on
Light schedule
Static deterministic model
0
Dynamic deterministic model Dynamic stochas tic model
Illuminance
Figure 6: deterministic and stochastic models: schedule-based (static) models (left) and dynamic models (right)
1.3 Importance, risks, and rewards of modelling occupants accurately Before seeking to advance modelling complexity, it is important to ask: why should model complexity be increased and what are we attempting to achieve by increasing accuracy? As first mentioned earlier, two risks of inappropriate occupant modelling methods were previously identified as leading motivators for appropriate occupant modelling (Gilani, O’Brien et al. 2016): 1. Building performance simulation results in inaccurate results (as measured by energy, GHG emissions, occupant comfort, peak loads, equipment size, etc.). For
instance, using standard occupant density and schedule assumptions may over-predict annual cooling energy and under-predict heating energy. This affects annual energy predictions and peak loads, which may affect energy-related incentives, predictions about net-zero energy, and predictions about energy savings relative to building standards and codes. 2. Simulation users and design teams make poor design recommendations on the basis of the inaccurate results from inaccurate simulation results. For instance, a designer may conclude that energy efficient lamps are unimportant because there is ample daylighting, if a model underestimates the extent to which occupants tend to keep window shades closed. Another example is the decision of whether to include fixed or operable windows. If windows are assumed to always be open or closed, they are unlikely to be favourable with regards to annual energy predictions. However, if operable windows are assumed to be used sensibly (not necessarily optimally) by occupants, they are much more likely to yield energy savings over fixed windows. Arguably, this second risk is far graver, as design decisions –particularly those related to architecture (massing, envelope, window-to-wall area ratio, etc.)—are often 30 to 50-year or longer commitments. The notable footnote for this second risk is that building performance models only play a limited role in influencing building design in current Canadian practice. However, there are market- and policy-driven forces that are encouraging greater use of BPS in new building design (e.g., greater use of the performance path for code compliance and incentives programs that require modelling). The NECB (and other energy codes and standards) have the purpose of guiding design teams towards better performing buildings; thus, modifying the NECB to better reflect improve occupant modelling practices can have a real impact. Occupants’ simplistic treatment in the Code permits the continued narrative that occupants are a passive design constraint or boundary condition rather than active participants in the energy and comfort performance of a building. As an illustrative example, the Code requires that HVAC equipment sizes be incrementally increased if discomfort exceeds a given threshold; rather than pursue a wealth of potential occupant-centric solutions to discomfort, such as providing adaptive opportunities (e.g., operable windows and exterior solar shading). Similarly, the unmet hours discomfort metric that is used in NECB and ASHRAE Standard 90.1 ignores the past decades of advancement that show adaptive opportunities greatly increase occupants’ tolerance for a larger indoor temperature range (Brager and de Dear 1998). In general, the authors’ experience at design meetings is that occupants are insufficiently discussed, considering they are the driving purpose for buildings. Until clients demand it, one of the ways to drive better occupant-centric design is through building codes. Well-publicized researcher-led case studies that illustrate the potential for occupant modelling can also help advance standard design practice. As illustrative examples, buildings that were designed on the basis of simplistic occupant assumptions may suffer from: •
Excessively large windows to enhance daylight potential and views to the outdoors because shades are assumed open.
•
•
•
Minimal operable windows because owners fear that they will be opened in energyadverse ways, rather than incorporating occupants’ tenancy to close windows if they lead to discomfort. Limited adaptive opportunities (e.g., thermostats that cannot be adjusted) because occupants are not trusted to use them in energy efficient ways. In contrast, Gunay, Shen et al. (2018) showed that occupants prefer more energy-conserving setpoints (i.e., lower in winter and higher in summer) than standards suggest. Selection of HVAC systems that do not adapt well (e.g., reduce outdoor air supply rates) to variable and partial occupancy because the Code only tests whether a building performs well at near-full occupancy.
Assuming occupants are modelled with greater detail, there still exist some risks (and rewards), as summarized in Table 3. Table 3: List of risks and rewards associated with advanced occupant modelling
Risks
Rewards
1. Undersized equipment as a result of underestimating occupant-related heating, cooling, or ventilation loads and corresponding discomfort. 2. Undersized spaces as a result of underpredicting occupancy o If occupancy modelling is relied on for sizing spaces (i.e., architectural programming), there is a risk that unsuitable occupancy modelling assumptions could lead to undersized spaces. 3. Poor design decisions because occupant modelling assumptions were not appropriate. For instance: o Designing a building with multiple adaptive opportunities based on model, but then occupants misuse them in energy-adverse ways. o Key design decisions, such as building massing, interior finishes, window size, and window shade fabric are inappropriate because occupants use spaces or systems differently than expected.
1. Reduced equipment sizing as a result of realistic occupant modelling assumptions 2. Reduced energy use and greenhouse gas emissions as a result of right-sized equipment running in its near-optimal operating range 3. Improved architectural design (e.g., window to wall area ratio, shade materials, building massing, interior finishes, etc.) based on modelling occupants’ potentially energy-adverse adaptive behaviours 4. Greater acknowledgement of the benefit of adaptive comfort-related opportunities for occupants 5. Identified opportunities to use systems that adapt to occupancy (e.g., demandcontrolled ventilation, motion-activated lighting) 6. Improved occupant comfort and health in and usability of buildings as a result of appropriate design decisions 7. Robust design from stochastic occupant models (e.g., architectural features that inherently reduce the frequency of discomfort-triggering occupant responses) 8. Assessment of building resilience and potential to improve it through adaptive comfort opportunities
9. Smaller buildings to reflect lower real occupancy
1.4 Building agents The first part of this document –and much of the research field—is focused on individual or groups of occupants and their behaviour. However, there are many decision-making agents (or actors) involved in large buildings, including those shown in Figure 7. Their roles and impact on energy are summarized in Table 4. These tables are based on large multi-tenant office buildings and may be somewhat different for other building types or ownership structures. Building owner
Building manager
Building operators
Tenants
Janitorial staff
Individual occupants Figure 7: Hierarchy of key actors in office buildings Table 4: Building actors, their roles, and how they affect energy
Actor
Role
How they affect energy
Building owner
Owner of the building
Capital expenditures and high-level corporate strategies (e.g., behaviour change campaigns and branding)
Building manager
Performs managerial, business, and administrative duties such as leasing, receiving occupant complaints, managing janitorial services and maintenance
Sets building policies; interacts directly with tenants or occupants; develops and implement energy conservation programs
Building operator
Manages and provide technical services related to maintaining and operating services (HVAC, lighting, water appliances, elevators, etc.)
Sets building controls and schedules, addresses faulty or inefficient equipment; may provide undesirable occupant behaviours of response to complaints is slow or ineffective
Janitorial staff
Routine cleaning and maintenance of spaces
Lighting and cleaning equipment use, potentially some routine modification
to building systems (e.g., open shades or turn off task lights) Tenants
Represent and/or manage occupants and spaces
Local decisions and policies, such as lighting schedule, furnishing, lighting, and equipment selection and layout
Occupants
Perform work duties and occupy office space
Routine use of local building systems, plug-in equipment, and lighting. See Table 5 for major domains for how occupants affect energy use in offices. Note that not all occupants are equal and there are often dominant occupants who play a much greater role in decisionmaking.
On the basis that the current project is focused on occupants, we proceed with only that type of building agent. However, future research is needed to understand and potentially model the impact of all agents. Individual people from the other groups could have orders of magnitude more impact than occupants. Owners could make major lighting upgrades and janitors may turn on and leave on lights from the time they perform a cleaning until the next morning (Brown, LaRue et al. 2016). Table 5 provides a list of major occupant-related domains of interest and some corresponding characteristics. These domains are then shown graphically in Figure 8. Table 5: Office occupant-related modelling domains of interest
Domain
Definition
Key indicators
Known predictors
Impact(s) of interest
Occupancy
Presence of occupants in building
• Occupant density • Absolute number of occupants • Occupancy diversity schedule
• Time of day • Day of week • Holidays
• A necessary condition for occupant actions (except for remotely controlled systems) • Metabolic heat and moisture gains; CO2 and odour generation
• Power • Occupancy (arrival, • State/dimmed departure, level intermediate • Lighting periods) power • Indoor density and illuminance diversity from electric schedule lights or daylight • Power • Time of day/week • Power density and • Anticipated diversity occupant schedule absence duration • Office type
• Electricity use • Heat gains • Illuminance
Action on or position of window shades and blinds
• Mean state • Number of opening and closing actions per day
• Indoor illuminance • Solar gains • Heat loss
Operable windows
Action on or state of operable windows
• Mean state • Number of opening and closing actions per day
• Outdoor and outdoor illuminance • Indoor luminance • Views to outside • Indoor and outdoor temperature • Noise • Indoor air quality
Thermostats
Thermostat adjustment patterns and resulting setpoints
• Mean occupied and unoccupied setpoints • Frequency of adjustments and conditions leading to them
• Indoor temperature • Indoor relative humidity
• Heating and cooling energy • Resulting indoor temperature
Manual lighting control
Adjustment of lighting state or dimmable level
Office plug-in equipment
Use (duration, power consumption) of plug-in office equipment, such as computers and printers and kitchenette equipment
Window shades/blinds
• Electricity use • Heat gains
• Sensible and latent heat gains/losses • Contaminant transport • Resulting draughts and airflow
Personal fans
Adjustment of and state of fan (e.g., on, off, modes)
Clothing level
Clothing level of occupants
Metabolic rate
Indoor heat gains of occupants and internal heat generation
• Mean fan state • Number of fan state/speed adjustments per day • Fan power • Clo level and schedule
• Met • Heat per occupant (W/occupant) • Heat per floor area (W/m2)
• Indoor air or operative temperature
• Fan electricity and heat gains • Thermal comfort and implications of air movement
• Indoor and outdoor air temperature (with significant time lag) • Occupant presence and activity type
• Thermal comfort implications
• Heat, moisture and CO2 gains of occupants
Directly affects energy and affects heating/cooling load Affects heating/ cooling load Office plug-in equipment use
Legend
Direct physical causal effect Indirect physical causal effect Comfort related causal effect
Office equipment
Lighting use
Thermostat use
Personal fan use
Occupancy (presence)
Heating
Personal heaters use
Window shade/ blind use
Cooling
Operable window use Metabolic rate
Lighting
Indirectly affects heating/cooling load Clothing level
Figure 8: Map of key occupant behaviours, organized by how they impact building performance
This document frequently refers to performance or building performance. In the current context, performance refers to energy, peak loads, efficiency, and occupant comfort – at timescales ranging from 5 minutes to a year. The ultimate goal of occupant modelling is to support the code compliance and design processes for healthy, comfortable, usable, and resilient buildings for occupants, while also reducing energy use and costs, GHG emissions, peak loads, and equipment size and maintenance costs. In expanded form, the goals are summarized in Table 6. .
Table 6: Summary of quantification of building performance
Category
Definition
Example metric(s)
Energy use
Total building purchased energy (e.g., electricity, natural gas, biomass, etc.)
kWh/a, kWh/m2a, GJ/m2a
Energy cost
Total building energy costs, including delivered energy, and delivery and peak charges
$/a, $/m2a
GHG emissions
Effective greenhouse gas emissions (often expressed in terms of CO2 equivalent) resulting from combustion of fuels on-site as well as those associated with off-site electricity generation and fuel processing.
TCO2equiv, TCO2equiv/m2
HVAC equipment size
Capacity of boilers, chillers, cooling towers, coils, and other equipment.
kW, tons
Peak loads
The peak heating and cooling rates required by mechanical systems, given a set of assumed boundary and operating conditions
kW, MW
Comfort
A measure of occupant comfort (including thermal, visual, and acoustic domains)
Unmet hours, predicted percent dissatisfied (PPD), predicted mean vote (PMV), daylight glare probability (DGP), sound pressure level (SPL), etc.
Indoor air quality (IAQ)
A measure of the species-specific concentration of contaminants in the breathing zone of a space.
Concentration of contaminant species CCO2(t), percent of people dissatisfied (PPD)
Resilience
A measure of a building’s ability to maintain safe and habitable conditions in the event of system or energy supply failures.
Passive survivability (days of habitability after power failure), thermal autonomy (% of time when mechanical heating and cooling area not needed to maintain comfort)
Energy use and GHG emissions are not as straightforward as they appear, as there are some fundamental accounting questions (e.g., location of the spatial and temporal scope) (Marszal, Heiselberg et al. 2011). For instance, matters of inclusion of on or off-site renewable energy generation, inclusion of exterior lighting, and GHG emissions accounting are generally not trivial or standardized. These issues are beyond the scope of this project. Other potential performance
metrics are about usability (e.g., effectiveness, efficiency, and satisfaction) and additional measures of healthiness.
1.5 Building modelling objectives, design phase, and spatial scale Before discussing occupant modelling strategies, current and potential applications for building performance simulation in the new building design process must be considered. This can be represented by three axes, as shown in Figure 9. Model objective refers to the role of the BPS model to support the design process. Design phase refers to the model purpose within the design process – from conception to construction. Model spatial scale refers to the physical scale of the model – from room-level to community or larger. Technology/economic evaluation
Model objective
Occupant comfort analysis Net-zero energy design Building design Community or district Building
Standard compliance Code compliance
Storey Zone Room
Building design phase
Figure 9: Three-dimensional space of building simulation applications for design
The three axes are not necessarily independent, in that the ultimate solution for best occupant modelling practice does not likely populate the entire 3D matrix. For instance, code compliance and net-zero energy building design nearly always focus on building scale models because the performance metrics focus on whole-building annual energy. In contrast, occupant comfort analysis would likely focus on the room or zone scale, where the occupants are located. Similarly, early phases of design are likely to focus on larger scale considerations (building massing and overall design strategy), whereas detailed design also focuses on the room and zone scale. The next three sections explore the three axes of Figure 9 in much greater detail. 1.5.1 Building model objectives As reiterated by workshop participants, there are numerous objectives for creating and using BPS models for new buildings. The modelling approach for all building aspects –not only occupants— can be greatly impacted by the modelling objective and thus this is a central theme for this document. Key modelling objectives and their implications are summarized in Table 7.
Table 7: Summary of primary applications of BPS for new building construction.
Model objective
Description/ purpose
Imposed occupant modelling constraints
Possible modeller bias
Typical desired model outputs
Annual energy/GHG emissions Code and compliance
Demonstrate that proposed building model performs better than code minimum; often additional requirements for IEQ
Occupant assumptions in design are same as reference building and often use standard schedules and density values
Demonstrate code compliance by making assumptions about occupants that exaggerate energy savings from energy efficiency measures
X
Standard design process
Show impact of various design decisions to support the design process
None, though likely reliance on standard schedules to model occupants if no better information is available
Depending on liability, incentives, and client desires, modellers may make optimistic or conservative occupant modelling assumptions
X
Peak loads
Comfort
X
X
X
Absolute targetbased design (netzero energy and contractual performance requirements)
Examine an integrated set of building design features (both EEMs and renewable energy systems) to achieve a predicted, and possibly measured, netzero energy
None
Depending on expectations, modellers may be incentivized to make conservative assumptions about occupants to mitigate liability or optimistic assumptions to achieve energy target costeffectively
HVAC sizing
Using design conditions (climate, occupancy), assess the minimum permissible heating, cooling and ventilation equipment size.
Often use standardized conservative schedules and densities
Mechanical engineering consultants are likely to be riskadverse to avoid under-sizing equipment
X
X
ASHRAE suggests monthly peaks
Other uses for BPS models that are outside of the current scope include: retrofit analysis and model predictive control. 1.5.2 Building design phase Around the world, building design phases have slightly different names; the names in Figure 10 are typical in the Canadian context and were established by Athienitis and O'Brien (2015). The six phases from start to finish are defined in Table 8.
Programming
Schematic design
Design development
Construction documentation
Construction
Operation, commissioning, and monitoring
Figure 10: Design phases of new buildings, from programming to occupancy Table 8: Definition of six phases in new building design and the associated implications for occupant modelling
Phase
Objectives
Occupant modelling implications
1. Programming
Determine building and space needs, activities, and functionality; set budget, energy goals, and other environmental goals.
Knowledge of typical occupant presence as a function of design occupancy can facilitate space-saving strategies and better balancing occupant densities to improve space utilization and occupant comfort.
2. Schematic design
Assess and determine design strategies related to energy and comfort (e.g., passive solar strategies, building form, daylighting, window to wall area ratio, natural ventilation, etc.); set approximate envelope properties and HVAC concept.
Analysis of occupant comfort and behaviour can and should greatly inform key design decisions such as window sizing and placement, use of operable or fixed windows, fixed solar and daylight shading, opaque envelope design, adaptive opportunities for occupants, outdoor air requirements, etc. For HVAC, knowledge of occupant activities, densities, and corresponding internal gains from lighting and office equipment can facilities right-sizing HVAC equipment – from plant down to terminal units and distribution.
3. Design development
Established detailed level of design for envelope design, HVAC components and control strategies, and window technologies.
Occupant modelling can be used to support major decisions such as: whether or not to include operable windows, providing HVAC controls to adapt to variable occupancy levels, etc. Ideally, infrastructure to monitor occupants should be installed. This serves two purposes: 1) to later
compare actual occupancy to design conditions in order to track building performance and 2) to facilitate greater understanding and modelling of occupants for future building design. 4. Construction documentation
All design specifications are documented in sufficient detail that the contractor can implement the design and can communicate with tradespeople.
None regarding occupant modelling; however, care should be taken to document subtleties regarding occupants (e.g., positioning and nature of building interfaces and reflective surfaces that cause daylight glare).
5. Construction
Building is constructed as per the design drawings and other documentation.
None regarding occupant modelling, however, care should be taken to ensure the building is constructed to be comfortable and usable (e.g., care to seal leaks to prevent draughts).
6. Operation, commissioning and monitoring
Prior to occupancy (ideally), the building is tested, commissioned, and continuously monitored.
All occupant-related controls should be tested and commissioned. If the BAS is capable of measuring and archiving occupant-related phenomena, this system should be tested to ensure proper functionality.
7. Occupancy
Normal building operation
Using the building automation infrastructure, data about occupancy and occupant behaviour should be continuously collected and archived to compare design intent to real performance and detect anomalies.
1.5.3 Building spatial scale Building scale is one of the axes on Figure 9 that is an important consideration for occupant modelling strategy because the design and operating questions and decisions vary greatly by scale. The system size, system type, and degree of connection with the occupant are affected by system scale. Refer to Figure 11 for various scales and the current scope.
Corresponding energy systems
Scale
Power plant
City
Local electricity distribution and/or district thermal network
Neighbourhood or district
Local thermal plant and air-handling (AHU) unit
Building Wing, quadrant, section of building Zone
Room
Storey
Air-handling unit, overhead lighting Variable-air volume (VAV) units, etc., overhead lighting
Terminal HVAC units, task lighting, personal heaters and fans, user interfaces Current project scope
Figure 11: Building scale and corresponding energy service systems
Aside from the obvious above effects of scale on occupant modelling and related decisions, diversity and uncertainty also vary greatly by scale. For instance, at the room level, two individual occupants may cause energy performance to vary by a factor of 10 as a result of the occupants’ energy-related behaviours. However, as the number of occupants increases (e.g., within a whole building), there is a tendency for the aggregated population’s energy use to converge toward a mean (Gilani, O’Brien et al. 2018). This is a powerful concept with regards to sizing equipment at the building scale. The diversifying effect of occupants is somewhat implicitly considered in HVAC sizing guidelines (Nall 2015), which recommends different safety factors for different levels (i.e., terminal units in individual rooms, distribution equipment, and plant equipment). However, with knowledge of occupant diversity, these factors could be much more precise and tailored. For instance, there is greater risk of having numerous higherenergy-consuming occupants or tenants in a two-story building with 20 occupants than a highrise multi-tenant office tower (O’Brien, Abdelalim et al. 2018).
2 Introduction to the Roadmap The objective of this roadmap is to lay a foundation and framework for bringing occupant modelling from primarily research-grade, as it is now, to advanced occupant modelling for building energy code compliance and simulation-aided building design practice. The roadmap follows guidelines laid out by the International Energy Agency (IEA, 2014), which recommends a stakeholder workshop prior to roadmap development. Thus, a stakeholder workshop involving about 30 diverse experts was held in Ottawa on May 1, 2017. The workshop outcomes are described in Appendix 2. The remaining stages of the roadmap are summarized in this roadmap and shown in Figure 12.
Goals and scope
Milestones
Gaps and barriers
Action items
Priorities and timelines
Figure 12: Roadmap outline
One notable deviation between this roadmap and traditional roadmaps is that considerable elaboration and recommendations for implementation is provided in some sections, on the basis that the authors have broken significant ground in some areas.
3 Goals and Scope The overarching goal of this roadmap is to guide (1) building codes and (2) design practice to place greater emphasis on occupant-related considerations to achieve: • • • • • •
More accurate energy predictions based on more realistic occupant modelling; Improved occupant comfort, health, and productivity; Reduced building environmental impact (e.g., energy consumption, GHG emissions, water use); Improved building usability; Greater building robustness (e.g., certainty of energy performance under variable and uncertain conditions); and, Buildings’ improved ability to adapt to variable occupancy and occupant behaviour.
The endpoint of this roadmap is to reach the point of making concrete recommendations based on current knowledge (e.g., from the literature and authors’ experience), while identifying future research, policy, and development needs. In the current scope, office occupants are building users whose primary purpose is to carry out workplace activities; they are not building operators. Also, occupant comfort modelling for its own sake is not a priority unless it is a predictor for energy-related behaviours. The current project is focused on occupant modelling and simulation to support the design –from conception to detailed design and code/standard compliance—of new medium and large (1,000 to 100,000 m2) Canadian office buildings. This covers spaces –open plan, private, and mixed offices— that are primarily used for office work (e.g., paper, computer, and other professional work) rather than process loads (e.g., manufacturing, laboratory testing). Many of the findings are to some extent applicable to other space types (e.g., libraries and laboratories) and phases of the building life cycle (e.g., retrofit analysis), but this is not the primary objective of this roadmap and the project at large. This report is also focused on regular occupants’ presence and behaviours and not on building owner, building manager, or office manager behavior. With the above in mind, the authors of this report in no way suggest that these are not important topics; rather we wish to ensure adequate depth on an already-complex topic. With regards to building codes and standards, the emphasis is on the National Energy Code of Canada for Buildings (NECB), herein often referred to as “the Code”. The following additional domains are outside of the scope of this document and project (with justification): •
•
Direct behaviour change of occupants. While energy-related behaviour change through messaging, gamification, and other means is an active area of research and development, the focus of this project is on modelling occupant behaviour. However, the effect of building design and operation on behaviour is within the scope. Building faults. To reduce scope, buildings are assumed to be functioning properly in this project. However, unexpected occupant behaviours, which could be considered faults, are within the scope.
•
•
• •
•
Water use and related behaviours. Water use and its environmental impacts in office buildings is relatively low compared to other building types such as residential buildings, hotels, educational facilities, and hospitals. Occupant behaviour and energy in common and specialized spaces (e.g., hallways, kitchens, elevators, washrooms, lobbies, restaurants, stairwells and elevators, mechanical rooms, and parking garages). The focus of this project is on occupant in regular workspaces. Transportation (e.g., commuter mode of transportation, electrical vehicle charging). The focus of this project is on activities and presence within buildings; not between them. Occupant mobility (e.g., tracking individual occupants between spaces). Focus of this project is on whether occupants are in a given space or not rather than tracking individual occupants between spaces. That is, the focus on presence is on a building-centric perspective rather than on an occupant-centric perspective. Indoor environmental quality (IEQ). The focus of this project is on occupant behaviour and occupant presence. The scope of IEQ is limited to where it is necessary to understand and model to predict energy-related occupant behaviour.
4 Milestones This section identifies major milestones required to elevate occupant modelling practice from current levels to an elevated, but practical, state. Milestones are intermediate performance targets for achieving the primary roadmap goal. The Action Items section elaborates on solutions to achieve the milestones, while the ordering and logistics of key action items are left for the Priorities and Timelines section. Key milestones to achieve the above goals are divided into four categories: research, building energy code, design practice, and BPS tool development. The target completion date for all milestones is ten years.
4.1 Occupant modelling research 1. Evaluate the impact of occupancy and occupant behaviours across all building types and climates covered by the NECB, with a range of scales from private office to large building (100,000 m2). 2. Determine the most appropriate occupant modelling approach (previously defined as a balance of accuracy in energy prediction and practicality) by building size, occupantrelated domain, and modelling purpose. 3. Collect extensive Canadian building field data from at least 100 different buildings (particularly related to occupancy, lighting and receptacle schedules and power densities) to update schedules and develop new models.
4.2 Building energy code 1. For each occupant-related domain, compile a complete list of current specifications required by the NECB. 2. Assess the relative energy impact of occupancy and occupant behaviours across all major Canadian climatic regions in order to prioritize their incorporation into the Code and to identify future research needs. 3. For each occupant-related domain, assess the most suitable method(s) to incorporate said domain into building codes (e.g., updated schedules, new prescriptive requirements, new modelling procedures). 4. For Code modifications, recommend specific implementation details such as wording and values). 5. Test new Code requirements and their implications and modify as required. 6. Submit proposed code changes to the Canadian Commission on Building and Fire Codes (the committee responsible for writing the NECB).
4.3 Building design practice and modeller education 1. For each occupant-related domain, compile a comprehensive list of current modelling approaches on the basis of survey and workshop results. 2. Systematically apply different occupant modelling approaches for each occupant modelling domain (e.g., lighting, blinds, occupancy) to a building to understand the strengths and weaknesses for different applications in the context of informing design at each design stage.
3. Develop a best practices guidebook that serves as an educational resource for advanced building simulation practitioners. 4. Develop and publish detailed building case studies to demonstrate the application of advanced occupant models and modelling procedures and quantify and illustrate their benefits (economic, comfort, environmental) with regards to the influence it has on building design decisions.
4.4 BPS tool development 1. Systematically evaluate current prevalent tools (e.g., EnergyPlus and front-end tools that use it, IES VE, and eQUEST) to determine current occupant modelling capabilities for each domain. Determine potential for existing tools to be expanded on ways this could be achieved (e.g., via source code, new schedules, etc.) 2. On the basis of survey and workshop results and recommended building code and design practice (above), identify BPS tool modelling approaches and procedures (e.g. parametric analysis) that need to be updated.
5 Gaps and barriers In this section, gaps are the knowledge- and technical-related obstacles that must be overcome to achieve the milestones. Meanwhile, the barriers are focused on market-, education-, and policyrelated obstacles.
5.1 Gaps The gaps regarding occupant modelling requirements in the building energy codes and design practice are investigated spanning the following topics. 1) 2) 3) 4) 5) 6) 7)
There is no consensus on which occupant-related domains to model Current occupant modelling methods are too simple Current occupant schedule and density values are outdated Contextual factors of occupant behaviour is not well understood Current building code procedural requirements are inadequate Occupant modelling in design practice is too simple Occupant modelling features in BPS tools are limited
Following discussion of the identified gaps, barriers are discussed. The gaps and barriers are about problems and challenges, whereas the Action Items section recommends solutions. 5.1.1 Gap 1: There is no consensus on which occupant-related domains to model To date, there has not been an exhaustive study to quantify the relative importance of the listed behaviours across building types, climates, and technologies. A limited number of specific simulation and in situ-based studies have revealed the impact of occupants (e.g., Clevenger and Haymaker 2006; Haldi and Robinson 2011; Parys, Saelens et al. 2011; Clevenger, Haymaker et al. 2013). However, prioritizing modifications to building codes and informing design processes will require a comprehensive understanding of the modelling assumptions that are most sensitive to occupants. Put differently, we need to understand the consequences of making poor occupant modelling assumptions and prioritize modification of occupant modelling assumptions accordingly. While workshop participants expressed most interest in understanding how occupants affect energy use, impact on other aspects of performance (e.g., peak loads, comfort, etc.) is also important. Moreover, the impact of occupant modelling approaches on resulting design decisions is also critical (as per the second risk explained in Section 1.3). The relative importance of each domain of occupant modelling across climates is non-trivial. In general, predicting internal gains accurately is most important for mild climates (e.g., Victoria, BC) where envelope and ventilation thermal loads are modest. They are generally least important where the heating and cooling loads are dominated by weather. However, in climates with similar heating and cooling loads, higher internal gains somewhat shift the energy use from heating to cooling. Predicting operable window positions is most critical for extreme climates (e.g., Edmonton) because of the potentially high temperature gradient between indoors and out. With that said, occupants are much more likely to open a window in the winter in milder climates (e.g., Victoria, BC and Toronto, ON) where the daytime temperature may be mild (e.g., 10-15°C). These examples illustrate that the current matter is non-trivial and requires extensive
parametric simulations to fully grasp. Beyond basic trends, there remains a major gap in knowledge about the relationship between climate and occupant modelling priorities. Table 9 contains a comprehensive list of energy-related occupant domains from (Schweiker, Carlucci et al. 2018). The table divides the domains into those covered and not covered by NECB and ASHRAE Standard 90.1. Note that coverage does not mean the assumptions are adequate, such as for blinds, which are assumed to be open unless motorized and automatically controlled. Not all listed items directly impact energy use (e.g., clothing level). However, for instance, clothing is an important predictor for thermal comfort, which in turn can be used to predict window opening and thermostat adjustment actions (refer to Figure 8). Table 9: Summary of occupant domains covered and not covered by NECB and ASHRAE Standard 90.1
Covered
Not covered
• Occupancy (schedule) • Receptacle (plug) loads (schedule) • Elevators and escalators (schedule; included in receptacle loads) • Lighting (schedule; with rules about controls) • Thermostat setpoints (schedule with rules about controls) • Blinds/shades (always open, unless motorized and automated) • Hot water use (schedule)
• • • • • • • •
Clothing level Manual operable window Occupant metabolic rate Occupant position and orientation within a room Consumption of hot/cold drinks/food Use of portable heaters/air conditioners Use of ceiling/desk fans Interior/exterior door position
Understanding the interrelationships between domains and the corresponding modelling approach is a significant gap in the literature. For instance, it is widely understood that occupants are much more likely to take adaptive actions upon arrival (Haldi and Robinson 2010). Thus, modelling individual occupants’ behaviours requires detailed predictions of arrival events, rather than the standard occupancy schedules. Current modelling approaches often implicitly acknowledge correlations by using similarly shaped schedules. However, this has not been extensively explored in the scientific literature. For instance, is an energy-conscious light user also an energy-conscious air-conditioning user? If correlations exist, there are implications on generating synthetic occupants or groups of occupants because these correlations should be maintained. The implication is that correlations may lead to more extreme occupant types (e.g., a copious user of electricity, may also be a copious user of fuel and water). These fundamental issues are summarized in Figure 13.
Office plug-in equipment use Lighting use
Thermostat use
Personal fan use
Energy Building performance simulation
Personal heaters use Window shade/ blind use
Peak loads Indoor environmental quality
Operable window use Metabolic rate Interactions?
Feedback loop?
Clothing level
Figure 13: Schematic illustrating the role of occupant modelling in understanding building performance uncertainty
Similarly, order of adaptive actions is not well understood. For instance, if an occupant faces overheating, will they take off clothes, open a window, turn on a fan, or lower the thermostat setpoint first? The answer is likely complex and depends on effort, effectiveness, cost, permanent, and other contextual factors (O'Brien and Gunay 2014), as further discussed in Section 5.1.4. With the trend of increasing building automation – often with remote operators – the importance of operator behaviour and automation systems relative to occupant behaviour also need to be quantified. This importance will vary by building type and size. In general, larger buildings are more likely to have a dedicated operator and a higher degree of automation. For some domains, the above recommended occupant impact study may also reveal that the Code should mandate automation of some systems. 5.1.2 Gap 2: Current occupant modelling methods in the Code are too simple As noted in the Section 1.2, the current building codes’ and standards’ occupant models are mostly limited to static-deterministic (i.e., fixed schedules), with a few exceptions (e.g., IES LM 83). The static nature of current schedules means that building models fail to explicitly penalize uncomfortable buildings that frequently trigger energy-adverse behaviours and fails to reward passive features that would reduce such triggers. For instance, O’Brien and Gunay (2015) showed that daylight availability can be improved by providing fixed shading that reduces window shade-triggering glare. The deterministic nature of current schedules means that the
Code is unable to provide ranges of predicted performance resulting from occupant uncertainty and assess a building design’s robustness against this uncertainty. Across the occupant-related domains, there is a gap in knowledge about the ultimate consequences (i.e., poorly designed buildings) of using fixed schedules rather than dynamic models to describe occupants and their behaviour. Table 10 compares key occupant modelling domains to the leading approach in the literature. Table 10: Key occupant related domains and how they required to be modelled in NECB
Domain
NECB
Leading approaches in the literature or proposed methods
Occupancy (presence)
Standard hourly schedule by day and building type; nominal occupant density
Stochastic occupant schedule with explicit arrival and departure events to serve as inputs for models below
Lighting
Standard hourly schedule by day and building type; nominal lighting power density; multipliers for occupancy control for baseline; lighting schedules in the proposed building should reflect automatic controls; the baseline building should have vacancy-off controls as required by the Code
Explicit modeling of controls, considering daylight and occupancy-based controls (as will be implemented in the real building) and manual adjustment based on advanced occupant model
Office plug-in equipment
Standard hourly schedule by day and building type; nominal power density
Updated schedules that are a function of modeled occupancy, and with potential to add stochasticity to account for diversity as a function of building size
Window shades/blinds
Always open (non-existent) unless motorized and automated
For manual shades: stochastic shade use model that considers occupant presence, indoor and outdoor daylight levels, and tendency for occupants to leave shades closed for extended periods For automatic shades: the control algorithm explicitly implemented, but with a model for manual overrides
Operable windows
Operable window schedules between baseline and proposed building models cannot differ if they are manually controlled.
For manual operable windows: stochastic window use models that considers occupant presence and indoor and outdoor air temperatures
For automatic operable windows: the control algorithm explicitly implemented, but with a model for manual overrides Thermostats
Standard hourly schedule by day and building type
Stochastic occupant manual thermostat adjustment model or operator model at occupants’ request
Clothing level
Not considered for standard simulations
Adaptive clothing model that accounts for outdoor temperature, but is based on field data for similar building types, considering social norms (e.g., business wear for offices)
Recent research and building occupant comfort standards on adaptive comfort (Brager and de Dear 2000; American Society of Heating Refrigerating and Air Conditioning Engineers (ASHRAE) 2017) have brought to light and standardized the notion that occupants often adapt themselves and their environment under uncomfortable circumstances. In contrast, the current predominant method to model occupants using schedules fails to recognize the two-way dynamic behaviour between occupants and buildings. Accordingly, the explicit benefit of providing adaptive opportunities is not recognized by BPS models; worse, adaptive opportunities in buildings are often penalized by designers because occupants are modelled as making load and energy-adverse decisions (e.g., window shades open even in the event of overheating). The vast majority of occupant models that have been developed in the literature in the past few decades are agent-based models with dynamic traits. However, it is unclear whether this high degree of resolution is necessary in all building simulation applications (e.g., code compliance, design support for large office buildings, etc.). If the primary modelling objective is estimating annual comfort and energy performance of building-scale, agent-based models may be excessive and impractical. More research is needed to understand the added benefits of these models and the limitations of simpler approaches in the context of code compliance and design practice. In addition to dynamic occupant models, the second common property of research-grade occupant models is stochasticity. However, stochastic occupant models are uncommon and have only been developed for very limited contexts and domains. Gunay, O’Brien et al. (2013) provided a comprehensive review of such models. Moreover, model validation –both internal (i.e., with respect to the same dataset) and external (i.e., with respect to other datasets from other buildings and climates)—has been very limited and inconsistent (Mahdavi and Tahmasebi 2017). It is unclear from the current literature whether the benefit of the stochastic nature of occupant models would sufficiently benefit building codes to justify the added complexity. If the Code required stochastic occupant models, multiple simulation runs (on the order of 50 to 100) would be necessary to estimate the predicted distribution of simulation results. Notably, O’Brien, Gaetani et al. (2016) found that 57% of surveyed modellers find this acceptable if the process was automated. A potential hybrid solution to achieve assessment of robustness without tens of simulations is suggested in the Section 6.2.
While diversity is recognized in simple ways (normally a multiplier on standard schedules), there is significant room for improvement in understanding the range of possible performance levels. Two ways to model diversity include: describing population using continuous distributions of occupant traits or by using occupant types (O’Brien and Gunay 2015). For the latter category, it has often been suggested that there are active and passive occupant types (Reinhart 2004). In brief, current occupant modelling methods are generally too simple for many BPS applications. However, advancing them (particularly where specific methods are mandated, such as for code compliance) requires careful consideration of the added challenges (complexity, education of modellers, simulation time, tool capability, building energy code enforceability, etc.). There is a major gap in understanding these trade-offs across the dimensions of model purpose, design phase, spatial scale, and occupant domains (as defined in Section 1.5). 5.1.3 Gap 3: Current occupant schedule and density values are outdated Building codes, BPS tools, and common practice reinforce the practice of modelling occupantrelated domains using schedules. As such, it is likely that schedules will likely be the predominant modelling approach for many years to come. Accordingly, it is still important to review and evaluate current schedules and the corresponding densities (e.g., lighting power density and occupant density), while acknowledging their major limitations. The current NECB performance path allows modellers to use the standard schedules listed in Sentence 8.4.3.2.(1) or modify them using “reasonable professional judgement”. However, the same modified schedules must be used in both proposed and reference buildings. The occupantrelated schedules include: occupancy, lighting, receptacle equipment (which includes elevators and escalators), thermostat setpoints, and service hot water. However, workshop participants noted that modellers are incentivized to be biased (e.g., pessimistic occupancy assumptions drive up predicted savings from EEMs). Thus, providing free reign on schedules could also be problematic. Enforcing a requirement to properly justify (with evidence) adjustments to the standard schedules should be considered. However, there is a gap in research on how this might be implemented – particularly for buildings with considerable uncertainty (e.g., new builds). Despite the flexibility afforded by the Code, O’Brien, Gaetani et al. (2016) found that the majority of surveyed modellers use default schedules; this was reinforced at the stakeholder workshop. Thus, these schedules have the potential to have a profound effect on model predictions and building design (Ouf and O'Brien 2018). Numerous workshop participants expressed eagerness for new schedule and density values. They had the following comments for current values: • Assumptions are usually very conservative • Schedules are outdated and do not reflect new office equipment and worker schedules • While current values may be conservative, relying on standardized and widely-accepted values (e.g., in NECB) allows them to avoid liability • Despite the apparent obsolesce current standard occupant-related schedules, their purpose –then and now—must be considered in the context of current and future occupant modelling applications. The original intent of the schedules was primarily for (1) sizing HVAC equipment, for which conservative schedules were used to reduce the risk of
undersizing and (2) serving the performance path of building energy codes, whereby emphasis is on having a standard set of assumptions about occupants (and all energyrelated) by which the predicted performance of the proposed design and code-minimum equivalent can be compared. Tracing schedule and density values back to their origin often reveals that they are several decades old (e.g., Abushakra, Haberl et al. 2004) and based on a surprisingly small sample of data. Between then and now, office space utilization has profoundly changed with regards to technology (Sarfraz and Bach 2017). Personal computer equipment is often restricted to laptops that rely on heavy computing power outside of the conditioned space and lighting efficacy has significantly improved. Meanwhile, the requirement that employees are tied to their office space is significantly diminished, though new space management strategies (e.g., hotelling) may lead to higher occupant densities (Morrison and Macky 2017). In brief, the workshop participants mostly felt that standard schedules are obsolete and need updating. Given the popularity of schedule-based approaches, updating is certainly a worthwhile endeavour. A major gap requiring significant coordination is to collect new data to form the basis of new schedules. 5.1.4 Gap 4: Contextual factors of occupant behaviour are not well understood Regardless of the data source to building an occupant model, contextual factors have been shown to be highly-influential in some cases (O'Brien and Gunay 2014). For instance, a model developed from data for private (single-occupant) offices may not be applicable to double offices or open plan offices because of the social dynamics at play in the multi-occupant spaces. Most published occupant models include a note about lack of external validation (i.e., from other contexts). Key categories of contextual factors include: •
• • •
• • •
Culture and climate: certain cultures are accustomed to relying more on adaptive opportunities, such as natural ventilation and other strategies, compared to others. Of course, this depends in part on local climate. Availability of personal control: the occupant’s availability of a particular adaptive opportunity (e.g., flexibility to dress down or turn on a light switch) Accessibility of personal control: the occupant’s ability to access a building system (e.g., a thermostat that is not locked in a case) Complexity and transparency of the automation system: the ability for the occupant to understand and properly use a system (e.g., whether or not the thermostat logic is intuitive to an occupant) Presence of mechanical and electrical systems: the availability of electrical and mechanical means to improve one’s comfort (e.g., air conditioning) Views to the outside: particularly in the context of window shade/blind use, the opportunity to have views if shading devices are opened Interior design: ability for the occupant to relocate, rotate, and access building systems to improve comfort, in the context of furniture and interior layout
•
•
Visibility of energy use: the availability of information on personal or otherwise disaggregated energy use (e.g., availability of a dashboard that displays current energy use) Social constraints and dynamics: presence of other occupants and the impact that an adaptive opportunity to improve one’s comfort has on other occupants.
Thus, the scope of applicability of occupant models remains an open question. Systematic documentation and external validation should be prioritized for all current and future occupant models. Beyond the above contextual factors, it should also be noted that much of the emphasis on recent occupant model development is absolute accuracy (e.g., predicting individual occupant actions). The models tend to be agent-based (representing one occupant at a time) and may not be suitable or practical for other scales. Thus, a gap in knowledge is to understand how to develop future models to be suitable for broader applications (e.g., HVAC sizing, whole building simulation, etc.) or whether the current type of model is adequate. 5.1.5 Gap 5: Current building code procedural requirements are inadequate The current NECB (and ASHRAE Standard 90.1) performance path mandates a single simulation for each of the design and reference building models. This requires a single set of occupant assumptions (e.g., schedules) so that all occupant assumptions are consistent across both models. As noted previously, this neglects the fact that building design can affect behaviour. It may also obscure the benefits of buildings that adapt well to variable occupancy. However, the current agent-based dynamic stochastic models yield a different energy prediction every time they are run (in general) and thus it would not be possible to run just two simulations –for the reference building and proposed building—as is currently standard. In fact, even a single stochastic occupant model in the current code would be problematic in this context since such models yield a different result with each run. ASHRAE Standard 90.1-2016 has set a precedent in building codes to mandate sensitivity analysis procedural requirements in the performance path. Table 11.5.1 states that the reference model energy consumption should be the average energy result of four model rotations (90 degrees each) if the east or west fenestration area exceeds one quarter of the building’s total fenestration area. More research is needed to explore the potential for developing procedural changes to accommodate more detailed occupant modelling. Beyond the specific schedules, there is a gap in knowledge about the impact of the schedules regarding energy predictions and impact of energy efficiency measures (Ouf and O'Brien 2018). The schedules are not limited to direct energy impacts but also impact heating and cooling loads. In fact, heating and cooling loads and resulting part-load HVAC performance are the primary reason that the NECB requires internal heat gains to be explicitly modelled. Meanwhile, technologies that improve part-load performance or adapt to different occupancy levels (e.g., demand-controlled ventilation and finer lighting control zone resolution) may be overlooked because simple schedules do not suggest they would be beneficial. For instance, demandcontrolled ventilation’s full performance benefits are not exhibited with repeating occupancy schedule.
5.1.6 Gap 6: Occupant modelling in design practice is too simple So far, much of the focus has been on building energy codes; however, occupant modelling in design practice could be considerably more flexible. The ultimate question is: for a given modelling objective (new building design to retrofit analysis), model scale (single room to community or city), design phase (programming to detailed design), and domain (e.g., occupancy, lighting, and window opening), what is the ideal occupant modelling approach? Workshop participants stated that current building models are highly-dependent on the application (e.g., code compliance, standard building design, net-zero energy building design, etc.). It follows that occupant modelling strategies are also expected to vary considerably for different building modelling applications. Until now, this remains a major gap which was also identified in the final report of IEA EBC Annex 66 (Yan, Hong et al. 2017). The above question goes far beyond choosing one of the four categories that were laid out in Section 1.2. For instance, there are many procedural strategies and design questions that could be explored, such as: • • •
How do occupant modelling assumptions affect the rank of most cost-effective energy efficiency measures? What is the relationship between design robustness (e.g., standard deviation of energy predictions) versus mean energy prediction? What is the optimal window area given that occupants actively adjust window shades and lighting according to indoor daylight illuminance and glare?
Moreover, a new set of metrics needs to be developed considering the additional insight that advanced occupant models can yield (O'Brien, Gaetani et al. 2017). In building models with advanced occupant models incorporated, insights that are much richer than monthly or annual energy can be derived. For instance, metrics available as a result of advanced occupant models could include: • • • • •
Number of occupant actions (as a proxy for discomfort) Robustness Mean shade occlusion Number of hours when windows are left open when the outdoor temperature is below 10°C Potential in reduction number of workstations could be provided if hotelling office management were implemented
5.1.7 Gap 7: Occupant modelling features in BPS tools are limited Generally, BPS tool occupant modelling procedures are driven by building code modelling requirements and as such are quite simple (e.g., schedules and densities). For the most part, they do not include dynamic or stochastic occupant modelling features, though exceptions exist. There are two ways to incorporate occupant modelling in BPS tools: 1. Use existing capabilities (e.g., custom schedules) 2. Develop new features in BPS tools to accommodate more advanced occupant models or related procedures.
For the most part, current commercial-grade BPS tools are only currently suitable for the former approach. Some recent efforts to incorporate occupant modelling into BPS tools have taken both approaches. Gunay, O'Brien et al. (2015) demonstrated that dynamic stochastic models could be implemented into EnergyPlus using the Energy Management System (EMS) functionality, which allows some custom functions to be called during simulation runtime. Hong, Sun et al. (2016) developed obFMU (occupant behaviour functional mock-up) to interface external occupant models with a variety of common BPS tools. The following features and traits were suggested by workshop and survey participants. • • • • • •
Features to provide users with sensitivity to occupant assumptions More facilities to provide and visualize uncertainty-based results Better modelling of comfort; tools are most mature in IAQ and thermal comfort; but all domains require more rigour Higher spatial and temporal resolutions of simulation outputs, rather than the current emphasis on energy and loads at monthly or annual scales More occupant-centric outputs in addition to the focus on energy Greater transparency of occupant-related inputs as well as improved outputs
However, while recent progress has been made, there remains a gap in knowledge regarding which features should be included in BPS tools, let alone how to implement them (Ouf, O’Brien et al. 2018). And ultimately, there is a gap in the ability of current BPS tools that would prevent some of the more ambitious solutions to the gaps above from being pursued.
5.2 Barriers There are several main market and practical barriers to adopting better occupant modelling approaches: 1) value and perceived value of occupant modelling, 2) communication of occupant assumptions in the design process, 3) building codes and modelling complexity, and 4) modeller education. This section summarizes barriers that were identified by the survey and at the workshop. 5.2.1 Barrier 1: Value of occupant modelling Numerous workshop participants challenged the premise that detailed occupant modelling is needed, by asking: what are the potential rewards to all stakeholders (modeller, building owner, architect, occupants, etc.) for modelling occupants more accurately? Moreover, they asked: what are the risks for modelling occupants more accurately and then basing design and control decisions on simulation? The stated reason for this skepticism is that clients place high value on promptness of delivering results. Thus, without monetary incentive or code requirements, modellers have little incentive to do better. Moreover, some modellers are concerned about the liability that could come from making non-standard occupant assumptions. In short, the benefit of occupant modelling needs to be clear, tangible, and monetizable. 5.2.2 Barrier 2: Communication of occupant assumptions Workshop participants noted that there is no standard way to document and communicate space uses, occupant type, etc. from owner to consultants. As such, there is little integration between
consultants for occupant-related phenomena (e.g., architectural design, programming, lighting and electrical, plumbing, HVAC, fire and egress). 5.2.3 Barrier 3: Building codes and modelling complexity The NECB Preface lays out a number of questions used to assess the validity of modifications to the Code. They focus on: • • • • •
The ability of professionals to implement requirements Enforceability of new requirements Cost of implementation Implications on policy Consensus among all stakeholders
Moreover, discussion with NEBC committee officials suggest that there is reluctance to: • • • •
Have uncertainty in model predictions (i.e., stochastic models) Drive the need for additional bps tool features Encourage technologies that rely on occupant behaviour to succeed Further deviate from ASHRAE Standard 90.1
Thus, there are some major practical considerations for modifying code to mandate more advanced occupant modelling. 5.2.4 Barrier 4: Modeller education If the code requires or the market demands more advanced occupant modelling, modellers would potentially require more education (e.g., a day-long training course). Most modellers have not received a formal education on BPS fundamentals and instead have relied on tool-specific procedures (Beausoleil-Morrison and Hopfe 2016). As for all domains of BPS, the use of the state-of-the-art in occupant models should involve fundamental education (e.g., modelling assumptions, numerical methods, limitations of models, etc.). In the case of advanced occupant models, knowledge of probability and statistics and on implanting or inputting the models is required. Several workshop participants stated that stochastic agent-based models appear too complex for the current market.
6 Action items This section provides recommendations for solution-oriented action items that can help address the gaps and barriers identified in Section 0. By and large, the literature on occupant behaviour modelling has answered the fundamental research questions necessary for incorporation into building codes and design processes (modelling forms, model validation, and data collection methodologies). The recent literature is mostly focused on developing agent-based models with increasingly resolution (e.g., predicting whether an occupant will turn off their computer at the time of departure and whether a residential building occupant is sleeping (e.g., Wilke, Haldi et al. 2013)). For the foreseeable future, such detailed models are unlikely to be accepted for practice. Thus, the main research questions and corresponding about occupant modelling for code compliance and design practice are practical in nature. A more significant gap in the research is in reliably modelling occupant comfort (thermal, visual, acoustic, indoor air quality); however, this is beyond the scope of this roadmap. Four categories of action items (continued from the Milestones) are considered: applied research (Gaps 1 through 4, Barrier 1) and implementation of more advanced occupant modelling into to the Code (Gap 5, Barrier 3), design practice (Gap 6, Barrier 2), and BPS tools (Gap 7, Barrier 4).
6.1 Applied research The key applied research action items include: 6.1.1 Evaluate the impact of occupancy and occupant behaviours across building types, building sizes, and climates This process will entail running batch simulations whereby realistic ranges of occupant-related parameters are simulated in conjunction with the office models and later the remaining of the 16 standard archetype models (US DOE, 2018). This is important for informing how to proceed with code requirements, design procedure recommendations, and BPS tool developments but also to overcome the aforementioned barrier that modellers need to be convinced of the necessity to model occupants. Many workshop participants stated that they would like to see quantitative evidence of the impact of occupants. 6.1.2 Determine the most appropriate occupant modelling approach by building size/type and occupant-related domain This action should be pursued using a similar systematic sweep as the previous action. The occupant modelling approach is likely to vary by across the three dimensions shown in Figure 14. Occupants’ role in energy consumption of a building depends greatly on building type, in approximate order from most to least in Figure 14. Thus, the optimal occupant resolution is expected to be most important for building types where they play a more active role. Building size is important because uncertainty of energy performance resulting from occupant diversity (e.g., the spectrum of high and low energy users) is significantly greater for small buildings than large buildings. Gilani, O’Brien et al. (2018) found that stochastic light use models in private offices yield as much predictive power as simple rule-based models (i.e., dynamic-deterministic models)
50m2 & 500m2 & 5000m 2
Custom MURB schedules
beyond 100 occupants because of the law of big numbers. Finally, modelling approach may vary by domain. In particular, non-adaptive domains (namely, occupancy and receptacle loads) should continue to be be modelled using static approaches; whereas adaptive domains (e.g., window shades and lighting) may be more appropriately modelled as dynamic. For non-adaptive domains, Tahmasebi and Mahdavi (2015) found that specific schedule values are more important for predicting mean energy use than whether a model is stochastic or deterministic. A small, but growing, set of occupant models that are aimed at codes and standards balance complexity and practicality. For instance, IES LM 83 (IESNA 2012) uses simple rule-based (i.e., dynamic-deterministic) control of window shades. This model penalizes large windows for daylight glare, while remaining relatively simple to understand, implement, and interpret. Since its inclusion in LEED (Leadership in Energy and Environmental Design) v4, it has been incorporated into several BPS tools, such as IES VE. Thus, we recommend exploration and development of further models of this kind.
Building type
Figure 14: 3D matrix of recommended occupant modelling approaches by building type, building size, and domain (purely for illustrative purposes). Custom schedules are those that are multiplied by some coefficient, for example, depending on certain specifications. A fourth dimension could be climate zone.
6.1.3 Collect more Canadian building field data (particularly related to occupancy, lighting and receptacle schedules and power densities) To assist in updating schedules, a target sample of 100 office buildings to obtain occupancy, plug load, and lighting profiles is suggested. The authors have encountered significant organic efforts among property managers and building owners to collect such data for purposes of corporate social responsibility reporting and billing. However, such efforts often lack adequate documentation (i.e., exact scope of metering), contextual factors (e.g., which space is leased), and measurement hardware and software details (e.g., product information, post-processing procedures). We strongly recommend that a committee be formed to coordinate a voluntary program to build a data repository. New data will become available as a result of ASHRAE 90.1-2016’s (ASHRAE, 2016) requirement that interior lighting and receptacle loads be separately monitored and recorded at 15minute intervals or higher temporal resolution. Industry could follow thermostat-maker ecobee’s lead to have an opt-in open-data policy for research purposes. Previous data collection models (Abushakra, Haberl et al. 2004) are time consuming in the context of big and open data. Additional
promising data sources are building automation systems and electricity distribution companies. Occupancy data remains rarer; however, a number of new robust and low-cost technologies, such as use of WiFi infrastructure and computer vision combined with security camera recordings (Dong, Kjærgaard et al. 2018), are greatly improving occupant measurement capabilities.
6.2 Building codes There are six methods listed below that would allow incorporation of improved occupant modelling into building codes, in approximate order of increasing complexity. 1. Update prescriptive requirements based on occupant simulation studies or the literature 2. Add prescriptive requirements based on occupant simulation studies 3. Update schedules and densities based on updated field measurements 4. Update schedules or add schedules that are developed from occupant simulation studies 5. Mandate procedural changes related to occupant modelling 6. Specify the occupant modelling approach required (i.e., in lieu of schedules) Increasing complexity does not necessarily translate to “better” and could lead to a number of drawbacks, such as misinterpretation, difficulty of enforcement, or abuse. Thus, we recommend that each occupant domain be thoroughly investigated for the most appropriate approach(es), balancing accuracy and other benefits without being cumbersome, error-prone, and otherwise too complex. Any recommended modifications to the code should be accompanied by illustrative examples of failures of the current code and how the modification would address it. For each of the above methods, Table 11 provide an example of how a particular occupant domain could be elevated in the Code, along with justification. Where possible, these are based on the authors’ research or the literature. Table 11: Illustrative examples of occupant modelling implementation into the Code
1. Update prescriptive requirements Domain: Lighting Current approach: A manual light switch should not cover an area greater than 250 m2 for spaces of 1000 m2 or less (Sentence 4.2.2.1.(3)). Proposed approach: Reduce the 250 m2 maximum lighting control zone to 100 m2 (for office spaces). Justification: Occupancy simulations based on data-driven models have indicated that occupancy may be sparse and considerable unoccupied space may be unnecessarily have lights on. 2. Add prescriptive requirements based on occupant simulation studies Domain: Operable window opening position Current approach: No requirement Proposed approach: Operable windows with an opening area greater than 0.2 m2 must include a position sensor linked to the building automation system such that the heating and cooling setpoints in the adjoined space are reduced to 10°C and 30°C, respectively, if the window is open for longer than 5 minutes.
Justification: A simulation-based study revealed that sub-optimally-operated windows could increase annual heating and cooling loads by 20%. An unintended consequence of this requirement could be a reduced number of operable windows due to the additional cost. To combat this, a rule-based model could be required for the performance path of the Code which could result in improved comfort and energy performance. 3. Update schedules from field studies Domain: Receptacle load schedule and density Current approach: Fixed schedule Proposed approach: Modify schedule to better reflect recent field data (see graph below). For simplicity, the new proposed schedule is also based on 7.5 W/m2 as per Table 8.4.3.2.(2) in the Code.
Justification: Field studies have suggested that nighttime plug loads remain high (Gunay, O’Brien et al. 2016; Bennet and O’Brien 2017). While the weekday energy for the current and proposed schedules is reasonably similar (92 W∙h/m2 versus 110 W∙h/m2), the timing of the internal gains has a significant effect on HVAC loads, part-load performance and its energy implications, and indoor temperature during unconditioned periods. 4. Update schedules from simulation studies Domain: Lighting Current approach: Static lighting schedules Proposed approach: Use a decision tree (driven by extensive parametric simulations) to mandate a certain schedule depending on design parameters and climate (e.g., see below).
WWR =0.3
Orientation =45°
>=225°
Orientation
VT glazing
=135°
=0.55
=0.55