How to reflect real-world occupancy into an urban-scale building energy model?

What methods, tools, and data sources can you use to reflect real-world occupancy into an urban-scale building energy model?

– Get Real

Dear Get Real,

To support cities in addressing the decarbonization goals, development of urban building energy models (UBEMs) can play a significant role. UBEMs, as crucial planning and design tools, enable the holistic optimization of energy systems and strategies for energy supply/demand at multiple scales of building, neighborhood, and city. However, developing realistic energy models at the urban scale is challenging. It is mainly associated with uncertainties from the lack of data or methodological approaches for capturing the model’s dynamic factors, such as occupancy and occupant behavior. Obtaining accurate occupancy patterns for buildings is challenging because of their multifaceted and spatiotemporal stochastic nature, which requires a systematic approach to understanding and incorporating occupancy and occupant behavior in the model. This lack can lead to a significant discrepancy between simulated and measured energy data, misleading energy prediction and management, and the improper design of building and district energy systems, resulting in excessive investment or inefficient operation.

Occupancy in UBEM can be incorporated through two sub-groups:

Occupancy and movement as defined as occupant presence/absence, occupant estimation (the number of occupants), occupant prediction (prediction in a future time window of occupancy), occupant movement (transitions between rooms/different zones), and occupant activity (identification or prediction of a specific activity); and
Occupant behavior representing an occupant’s interactions with building systems to enable meeting the occupant’s energy and comfort-related needs.

Different methods are used to model occupancy in UBEM depending on the input data, the level of complexity, and the research goals, including rule-based (schedules or profiles), deterministic (or statistical), non-probabilistic (or data mining, data-driven), probabilistic (or stochastic, machine learning), agent-based stochastic (or object-oriented) and virtual occupant behavior models. State-of-the-art UBEM mainly relies on static deterministic profiles or schedules not representative of real-world dynamic occupancy.

Some primary data sources commonly used in urban-scale occupant behavior and energy modeling include occupancy data and mobility data from surveys, especially national representative surveys, such as the American Time Use Survey (TUS) and National Household Travel Survey (HTS). Time Use Survey (TUS), which numerous countries have conducted as a part of the national census, is the major source representing occupant behavior profile and the amount of time individuals spend on different activities daily. Household Travel Survey (HTS) represents mobility patterns and travel behavior conducted by some countries that have been used for informing occupancy. Advances in digital technologies and data-driven approaches, including advances in sensing, computing, and artificial intelligence (AI) and the Internet of Things (IoT) and open data initiative, are promising in offering new opportunities for better understanding and modeling the real-world occupancy and occupant behavior representative in UBEMs with the ability to learn from the complex, dynamic and unpredicted changes and abruptly changing operational environments. Using big data analysis and machine learning approaches to predict and explain real-world occupancy and occupant energy usage-related activities and behavior is promising. IoT sensors are usually used to collect occupancy data representing indoor and outdoor mobility patterns and occupant activities. WiFi, BlueTooth, camera and electricity, and other passive data sources and wireless network data also have been used as a proxy for urban scale occupancy patterns and predicting energy use. WiFi has been used via WiFi infrastructure or in combination with datasets from count sensors or WiFi-enabled devices and other wireless network data to provide data on building occupancy. Datasets representing mobility patterns such as Geospatial trajectories and Global Positioning System (GPS) data, social media data, google trends, mobile/cell phone data, and other transportation data and Location-based services (LBS) data can potentially be a useful source for extracting occupancy and occupant behavior patterns. Mobile data has been collected via smartphone applications and Call Detail Records (CDR), and obtained from various providers. For example, the TimeGeo framework and CDR data, provided by AirSage, was used in a study for developing mobile-infrared occupancy in an urban scale energy model for the city of Boston. In another study, mobile data from Cuebiq company was used to increase the accuracy of occupancy data in UBEM. LBS data are collected through smartphone apps and can be obtained from service providers using the apps to provide a specific service. Transportation data representing various modes (e.g., car, bicycle, public transit, etc.) and location data from the public transportation system and cellular network data, representing mobility dynamics in real-time, can inform occupancy in UBEMs. In addition to the complexity associated with the occupant’s spatiotemporal stochastic nature, privacy is a main issue. Protecting privacy is addressed through anonymization and pseudonymization and defining occupancy profiles at aggregated levels. The validation of data quality and quantifying the uncertainties involved is another challenging task. Each of these data sources varies in terms of quality, resolutions, population coverage, bias, etc. A comprehensive approach is required to develop standard protocols in data gathering and guidelines on validating occupant measures and comparing results. In my view, future research on occupant-driven UBEMs will benefit from integrated modeling and real-time energy monitoring, which have yet been only marginally explored. Using Ambient Intelligent (AmI) systems enabled by AI, sensing and IoT technologies offers opportunities for supporting flexible and adaptive user-centered environments as well as elevating realistic occupancy-related data, which represent multifaceted occupant behavior and its impact on energy efficiency for modeling and designing future scenarios.

Narjes Abbasabadi, Ph.D Assistant Professor, University of Washington Contact Narjes Here