K. Law, R. Fruchter, V. Bazjanac, F. Grey
Building performance simulation tools model and predict building performance, but their accuracy is compromised by simplified simulated occupancy. Occupant behavior is complex yet energy modeling software represents it as deterministic and unchanging in hour-long periods of time, which leads to discrepancies between model results and measured performance. These discrepancies limit the use of the models both as a predictive tool and real time post occupancy evaluation of the building. We propose to (1) define a computational modeling framework for building and occupant cooperative sustainable performance; (2) collect correlated occupant and building performance data sets in real time (3) develop a computational spatial-temporal-physiological occupant model and a preliminary prototype Space-Mate. Real-time occupant state and building performance data feeds will generate dynamic occupancy information for building energy performance simulation and building space adjustment to respond to the evolving occupant’s energy needs and provide feedback to the occupant for potential sustainable behavior changes.
We are designing with old, limited data!
Building performance simulation tools model and predict building energy performance, but their accuracy is compromised by the approximations, assumptions and simplifications made in the model. This is equally true for building performance simulation used to predict performance of the future building and justify decisions considered in its design, as for real-time simulation used to optimize building operation. In both cases, the simulation of internal loads is a critical determinant of the results, and the basis for most of a building’s internal load is simulated occupancy.
Several research projects aim to improve the calculation of occupancy for use in building energy performance simulation are already in progress (e.g. Annex 66). Most of these projects are developing occupancy definitions based on statistics of measured and recorded occupant presence in buildings and building spaces that are varied depending on building type and use. Such methodology has a general shortcoming: It does not link the generated statistical information with its cause – the reasons that determined the defined occupancy as such. In other words, such statistics define records of what occupancy took place, with no indication why and of the involved agents of change.
Occupant behavior is complex and stochastic yet current energy modeling software represents occupant behavior as discrete, deterministic and unchanging in hour-long periods of time. Such occupancy definitions affect not only building use schedules, but also related variables such as lighting, plug loads, humidity, and airflow, which leads to questionable results and a known gap between these and the building’s actual performance. These discrepancies limit the use of the models both as a predictive tool for building performance during design – directly impacting the sustainability of the project - and as part of post occupancy evaluation of the building in the operating and maintenance phase – limiting the analysis of the usability and operability of the facility, which are three out of the four CIFE RFP goals.
What if the building space becomes a silent teammate to occupants’ activities as a continuous real-time dialogue between building energy performance and occupant state as they dynamically interact and affect each other through co-simulation?
Current building sensor technologies and biometric sensors offer new opportunities to collect data in real time from both the building’s operations as well as the occupant’s physiological state. Using this data, we propose to (1) define a computational modeling framework for building and occupant cooperative sustainable performance; (2) collect correlated occupant and building performance data sets in real time; (3) develop a computational spatial-temporal-physiological occupant model and a preliminary prototype called Space-Mate. Real-time occupant state and building performance data feeds will generate dynamic occupancy information for building energy performance simulation and building space adjustment to respond to the evolving occupant’s energy needs as well as provide feedback to the occupant for potential sustainable behavior changes. Moreover, the generated occupant models will make simulated occupancy in predictive building energy performance simulation significantly more realistic than it is now.
Original Research Proposal
Space-Mate Progress Report
There were eight major activities conducted during the first half of the Space-Mate project reporting period: (1) perform an extensive literature review, (2) collect state-of-practice building data from CIFE industry partners, (3) develop a building-occupant data collection protocol, (4) deploy IRB human subjects protocol, recruit volunteer participants, and collect signed consent forms from volunteers, (5) identify and instrument two spaces in the Y2E2 building at Stanford for building data collection, (6) launch concurrent building and occupant data collection, (7) develop preliminary data analytics and indicators using building space and occupant data towards a correlated building-occupant data set and co-simulation, and (8) participated in the definition of a concept paper that was submitted to the pre-selection phase to DoE in response of a new RFP.
Literature review. The International Energy Agency (IEA) identified occupant behavior as a driving factor for energy use in buildings, so improving occupant behavior modelling could help reduce these discrepancies. On the other hand, statistics show that people spend around 90% of their time indoors and therefore not only do occupants affect building performance, but the design and operation of the buildings affect occupant wellbeing and performance. Space-Mate targets real-time concurrent data collection and performance monitoring of occupants and building spaces. As part of the first stage of the project, we performed a literature review of research efforts on improving occupant and building simulation. We focused on three directions: 1) Occupant monitoring and data collection, 2) agent-based occupant modeling, 3) co-simulation of occupant comfort and building performance. The rise of cheaper sensor technology paired with wearable and mobile sensors allows for occupant data to be collected. Regarding the impact of occupant’s on energy performance, significant efforts have focused on determining an occupant’s movements and presence through motion sensors, infrared sensors or equipment-level energy usage monitoring. To date we have not identified any efforts from the AEC industry focused on the detection and monitoring of well-being indicators or quantifying knowledge-work-productivity, yet other industries can help as role models for data collection and analysis. For example, the Virtual Human Interaction Lab at Stanford University has used non-invasive motion detection sensors to track facial expressions and full-body non-verbal behavior of people and used the data to predict characteristics ranging from gender to human performance, group rapport, learning, and creativity. The Affective Computing Media Lab at MIT uses wearable sensor data -EDA, skin temperature, etc.-, mobile phone data, and surveys focused on several well-being indicators to estimate individual wellbeing. Such data collection and prediction could serve as input for occupant simulation tools. For convenience, occupants in commercial simulation tools are represented in terms of static schedules, which do not represent the dynamic interaction between the building and its occupants. Recently, stochastic occupant behavior models have been developed based on long-term observational studies. These models typically do not account for inter-individual variability in behaviors nor simulate multiple behaviors together, which has led to the use of agent-based models (ABMs) which do address these issues. For instance, Langevin et al.’s ABM of office occupant thermal behavior, EDF’s SMACH which models residential electricity consumption and inhabitant’s behavior and University of Pennsylvania’s ABM which models window usage of occupants. Co-simulation couples occupant and building simulations. Kashif et. Al demonstrate a co-simulation environment for smart homes. Langevin et al developed a Human and Building Interaction Toolkit (HABIT) for simulating the thermally adaptive behaviors and comfort of office occupants and building energy consumption. EDF uses a Functional Mock-up Interface (FMI) to couple SMACH and BuildSysPro. Space-Mate project builds and expands on this knowledge.
Kashif, A., Ploix, S., Dugdale, J., & Le, X. H. (2013). Simulating the dynamics of occupant behaviour for power management in residential buildings. Energy and Buildings, 56, 85-93.
Langevin, J., Wen, J., & Gurian, P. L. (2015). Simulating the human-building interaction: Development and validation of an agent-based model of office occupant behaviors. Building and Environment, 88, 27-45.
State-of-practice building data collection and management. As part of our CIFE sponsored Space-Mate seed project we designed and deployed an online survey to collect state-of-practice building data collection and management practices of CIFE members . The survey was launched in February 2016 and data collection is ongoing. The response rate to date is 30% and we continue to contact CIFE member companies inviting them to share their current energy and sustainability strategies. The survey invites any building facility manager in the company to fill out this survey in case companies have multiple nationally or globally distributed offices (i.e. it would be ideal to receive input per building, or a representative sample of the corporate building stock). The link to the survey is: https://stanforduniversity.qualtrics.com/SE/?SID=SV_6YhcuawnXREACs5 All input/data will be treated confidentially and is only used for the research project. The survey has in total 11 questions and it would take you 5 minutes to fill it out. We plan to aggregate the data by company, as well as synthesize facility and building management state-of-practice to determine benchmarks. We plan to share with the overall survey results and confidentially enable participants to position their company with respect to these benchmarks.
Data collection protocol. Current occupant behavior focuses only on how the occupant impacts the building performance as a function of their presence in specific spaces, opening windows and doors. Such occupancy definitions affect not only building use schedules, but also related variables such as lighting, plug loads, humidity, and airflow, which leads to questionable results and a known gap between these and the building’s actual performance. These discrepancies limit the use of the models both as a predictive tool for building performance during design – directly impacting the sustainability of the project - and as part of post occupancy evaluation of the building in the operating and maintenance phase – limiting the analysis of the usability and operability of the facility. Building energy consumption and occupant comfort and well-being need to be addressed in an integrated approach. The knowledge productivity depends on the human physiological state, the activity or task being performed, the location where the activity takes place, and the technology infrastructure used to perform the activity. These spatial-temporal-physiological occupant characteristics are hard to characterize due to their temporal and contextual variable nature in and thus do not align with the simplified, deterministic assumptions of simulated occupant models used in energy simulation tools.
A key objective of the project is to develop a protocol for concurrent building spaces and occupant data collection that leverages innovative sensor technologies to collect building space data and spatial-temporal-physiological occupant data. These two data sets will yield a longitudinal correlated space and occupant database. Such correlated databases are not currently available. Using this database, we will develop the spatial-temporal-physiological occupant response probability distributions.
We developed a concurrent building and occupant data collection protocol. The target units of analysis for the building analysis were enclosed rooms in the Y2E2 building at Stanford to scope the number of external variables that impact space energy performance (e.g. weather, season). The two spaces are – the PBL Lab room 280 and the CIFE iRoom 292. In addition, since the researchers have 24X7 access to these spaces and interact and know the people who typically use these space it enabled them to instrument them and collect data.
Instrument the Y2E2 building rooms and occupants. For the building space data we aimed to collect accurate information about room temperature, CO2, and humidity. Thanks to discussions with Mr. Patrick Shiel, visiting scholar at CIFE we decided to purchase two novel mobile data collection instruments from Rotronics that measure, track, and store room temperature, CO2, and humidity – minute-by-minute. Kill a Watt device for real time power monitoring used to track the energy consumption per device per hour and quantitatively evaluate the plug load in the specific room. For the spatial-temporal-physiological occupant data we leveraged the PBL Lab sensor infrastructure: GoPro to track presence, location, proximity, space and ICT use; biometric and physiological sensors – Kinect for body movement, FirstBeat for HRV, Mindwave for brainwave activity; a virtual world cloud-based collaboration platform to capture and track activities and communication channel use of occupants; RescueTime application to track occupant productivity.
Concurrent building and occupant data collection started in February 2016 and is ongoing. We engaged 18 Stanford participants from the AEC Global Teamwork program who volunteered to assist with the biometric and physiological data collection during their weekly AEC project global meetings in the PBL Lab. The 18 participants are members of six globally distributed AEC project teams. The six AEC global project team meetings are held weekly at different times of the day – between 8:00am PST and midnight PST.
Devloping data analytics. We started to develop preliminary data analytics and indicators using building space and occupant data towards a correlated building-occupant data set and co-simulation.
Seeking further funding for broader project scope. Vlado Bazjanec, Renate Fruchter, and Flavia Grey participated in a multi-institution collaborative effort engaging Stanford University, U Penn, and LBL in the definition of a concept paper that was submitted to the pre-selection phase of a RFP. The concept paper received mixed reviews and the team continues to explore other funding opportunities.