Coordinating occupant behavior for building energy and comfort management using multi-agent systems
Highlights
► We simulate energy and comfort in a building under different control strategies. ► Simulation uses actual occupant preferences and behaviors from test bed building. ► Multi-objective MDP used for energy-conscious occupant meeting rescheduling. ► Proactive-MDP control achieves energy savings during peak occupancy. ► Proactive and proactive-MDP control achieve improved comfort over baseline control.
Introduction
There is growing interest in reducing building energy consumption through increased sensor data and increased computational support for building controls. This interest is largely motivated by the significant percentages of global energy consumption attributed to buildings. In the US, buildings account for over 40% of national energy consumption – greater than the consumption of either the transportation or industrial sectors [1]. Buildings' contribution to national energy consumption remains high globally as well, 39% in the UK, 37% in the EU [2], and 25% in China [3]. As people today are spending more and more time indoors, buildings also have a significant impact on human productivity, learning, health and happiness [4], [5]. Building energy and occupant comfort are therefore two critical parameters by which the performance of indoor environments can be assessed and improved. For both energy and comfort, the most critical systems include heating, ventilation, and air conditioning (HVAC) systems, lighting systems, and electrical appliances and devices, which account for 36%, 18% and 10% of building energy usage, respectively [1].
Today, traditional building management systems (BMS) lack real-time input of dynamic factors including occupancy, occupant preferences, occupant actions and decisions. Even with this information, current systems still lack intelligent reasoning to deal with such dynamic and distributed input. New building energy and comfort management strategies must be generated and implemented to adjust both device control and occupant behaviors. Occupant behaviors are defined as actions and decisions taken by building occupants that impact the energy use of their building. These include actions taken on objects within an occupant's personal control such as doors, windows, lights, and computers, as well as actions taken on the occupants themselves such as changing clothing, locations, or schedules. Consequently, the optimization of building energy and comfort becomes a complex problem requiring computational support and a real world interface. To meet these demands, this paper presents and implements a multi-agent comfort and energy system (MACES) to simulate alternative management and control of building systems and occupants. MACES relies on intelligent agents to manage and coordinate input from human and building system devices in a distributed fashion to achieve the goals of reduced building energy and increased occupant comfort.
MACES is implemented with data from a real-world building including actual thermal zones, zone temperatures, occupant preferences, and occupant schedules. The operations of this building are then simulated according to four distinct control strategies, Baseline, Reactive, Proactive, and Proactive with MDP (Markov Decision Problems), involving varying levels of intelligent coordination of devices and occupants. Finally, the energy and comfort results of these four strategies are compared and opportunities for further energy savings are assessed. The goals of the presented MACES are to (1) provide a realistic and replicable simulation model that interfaces easily with real-world input and uncertainty for customized building energy management, (2) assess and predict how changes to the building, occupant behavior, and operational policies will affect energy consumption and occupant comfort, and (3) develop a framework through which the MAS simulation (“cyber-world”) can interact directly with occupants in the “real-world” through proxy agents to provide suggestions for reducing building energy consumption.
Section snippets
Occupant-driven control
Current BMS generally operate according to fixed schedules and maximum design occupancy assumptions. While temperature, airflow, and lighting set points for most commercial facilities can be controlled digitally through the centralized BMS, these set points must still be determined and scheduled manually. Typically, operational settings are dictated according to assumed occupied and unoccupied periods of the day and do not consider when buildings are only partially occupied. Observations of
Multi-agent systems
In artificial intelligence, agents are physical or virtual entities that intelligently interact in an environment by both perceiving and affecting it. In multi agent systems (MAS), agents can additionally communicate and coordinate with each other as well as with their environment. Multi-agent frameworks are therefore used to model complex problems with multiple cyber agents in simulations or physical agents (proxies) that act in the real world. The architecting of a MAS requires first dividing
Multi-Agent Comfort and Energy System (MACES)
Building on the previous work involving MAS and MDPs for building energy and comfort management, MACES is developed as a system to affect both occupant behaviors and the operation of building devices. Building on the authors' previous work [62], MACES allows for full-building simulation of occupants, devices, and consequential building conditions under different control strategies. These strategies include reactive and proactive control in which HVAC and lighting systems are automatically
Evaluation of results
Evaluations of both energy and comfort are performed for the four simulated control strategies, Baseline, Reactive, Proactive, and Proactive-MDP. The results generated by MACES for a 24-hour period in the spring season are used to assess the effectiveness of the three alternative control strategies over the Baseline control strategy and to specifically assess the energy and comfort impacts of coordinating occupant behavior in addition to controlling building devices.
Future work and real world deployment
Goals for future work with MACES include increased complexity in MDP coordination and experimentation with real-world proxy agents. The MDP coordination of occupant behaviors will be improved to add further complexity in generating optimal policies and behavioral alternatives. In addition to relocating meetings, meeting coordination will focus on scheduling meetings out of perimeter zones as well as negotiating classroom schedules to further reduce energy consumption in high demand zones.
Conclusions
The results of the presented MACES provide motivation for investigating multi-agent systems for reduced building energy consumption and improved occupant comfort. The simulation of building control showed significant energy savings as compared to more traditional control approaches including those that considered both occupancy and occupant preferences. Specifically, the Proactive-MDP control strategy resulted in reduced building energy consumption during times of peak occupancy, whereas the
Acknowledgments
Authors would like to acknowledge the Department of Energy's support of this project through funding for the Building Level Energy Management (BLEMS) project (DOE Award DE-EE0004019). Any opinions, findings, conclusions, or recommendations presented in this paper are those of the authors and do not necessarily reflect the views of the Department of Energy.
References (67)
- et al.
Impacts of lighting density on heating and cooling loads in different climates in China
Energy Conversion and Management
(2006) - et al.
Towards a sensor for detecting human presence and characterizing activity
Energy and Buildings
(2011) - et al.
A comprehensive multi-factor analysis on RFID localization capability
Advanced Engineering Informatics
(2011) - et al.
Application of WiFi-based indoor positioning system for labor tracking at construction sites
- et al.
In-situ implementation and validation of a CO2-based adaptive demand-controlled ventilation strategy in a multi-zone office building
Building and Environment
(2011) Demand controlled ventilation a case study for existing Swedish multifamily buildings
Energy and Buildings
(2004)- et al.
Smart occupancy sensors to reduce energy consumption
Energy and Buildings
(2000) - et al.
An information technology enabled sustainability test-bed (ITEST) for occupancy detection through an environmental sensing network
Energy and Buildings
(2010) - et al.
Adding advanced behavioural models in whole building energy simulation: a study on the total energy impact of manual and automated lighting control
Energy and Buildings
(2006) - et al.
Occupant comfort in UK offices-How adaptive comfort theories might influence future low energy office refurbishment strategies
Energy and Buildings
(2007)