Occupant behavioral modeling: An agent-based modeling approach to building performance analysis

Abstract

Finite resources and increasing rate of consumption have made efficiency a key component in every energy consuming sector. The building industry, as a major contributor, has been the target of various initiatives and regulations aiming to lower its impact with varying degrees of success. Measurement of this impact has not always been easy, especially for buildings not yet built. Therefore, simulation tools have been heavily utilized to provide predictions.
However, the results from simulations are not always in line with the measurements. The extent of this variation is often so remarkable that it calls into question the reliability of building simulation tools and methods. A major contributor to this discrepancy has been identified as the oversimplification of occupancy inside the buildings, which neglects their impact on energy usage.
This study investigated this performance gap by comparing a normal energy model of a case study with a more realistic energy model that considers occupants and their behavior through an agent-based modeling approach while relying on real loads and set points. The normal model relied on Swedish building code and was modeled using Honeybee in Grasshopper, while the agent-based model utilized the normal model as a base case while changing different loads and set points to match the measured data. Furthermore, to model occupants, Occupancy Simulator was used to create occupant profiles as an obXML file, while a survey with a complementary code defined behavioral models for each individual; this file was then used for co-simulation via obFMU and EnergyPlus to create the agent-based model. Additionally, a comparative analysis was performed to investigate the impact of adopting an occupant-centric metric compared to energy use intensity to measure the performance of the simulations. Lastly, the accuracy of the agent-based model was evaluated.
The results demonstrated a significant gap between the total energy use of the two models, with an even larger disparity observed when using the occupant-centric metric. Furthermore, it showed that stochastic modeling of
occupant’s presence, movement, and interaction in the building had a considerable effect on energy usage. However, relying on inaccurate set points and schedules for the highest energy-consuming system that offered
no control to the occupants was the major contributor to the performance gap. The agent-based model was shown to perform correctly most of the time, although certain inaccuracies were identified.

Popular Abstract

Occupant behavioral modeling: An agent-based approach to building performance analysis
This study showed that the simulated energy consumption of an office building deviates between 20 % and 116 % from the baseline, depending on the approaches used for modeling and metrics used for measurement.
Building energy simulations are often used to predict building energy consumption and measure the effectiveness of different renovation plans. However, observations show that the simulation results vary significantly compared to real life measurements, which raises skepticism about the reliability and validity of the simulation tools.
In building energy models, specifications about the building's geometry, materials, and climate data rely on real information taken from the actual building and weather files. But when it comes to the number of occupants, their behavior and the types and number of systems used in the building, such as lights, computers, cooling and heating system along with their operation pattern, values are taken from standards.
This oversimplified representation of occupants and different systems is one of the factors contributing to the inconsistency between measured energy use and simulation outputs.
To assess this difference, an office from a university building was taken as a case study. A comparative analysis was performed between a conventional model and an improved model that is closer to reality. The conventional model relied on standards for inputs and schedules. The Improved model contained an occupant behavioral model simulating the arrival and departure of each individual and their movement in the building. In addition, interactions with different systems such as lights, windows, blinds and electric equipment were modeled based on real inputs gathered through surveys and sensors. Also, real values from the building were used for lighting and electric equipment alongside real set points for heating and cooling.
Afterward, the results were analyzed using two different metrics: energy use intensity (kWh/m²), a conventional metric used to measure a building's performance based on its total heated floor area, and (kWh/Occupant), a metric that considers the total number of occupants present in the building for a year to measure performance.

• author: Ramezani, Aryan
• supervisor: Pieter de Wilde; Niko Gentile
• organization: Department of Building and Environmental Technology; Division of Energy and Building Design
• course: AEBM01 20241
• year: 2024
• type: H2 - Master's Degree (Two Years)
• subject: Technology and Engineering
• keywords: Occupant behavioral modeling, Agent-based modeling, Performance gap
• language: English
• id: 9160953
• date added to LUP: 2024-06-11 07:50:28
• date last changed: 2024-06-11 07:50:28