Sunteți pe pagina 1din 4

A multi agent system framework for value focused

interactions between buildings and electrical smart


grids
Aduda K.O., Zeiler W. and Boxem G.
Building Services Group, Department of Built Environment
Eindhoven University of Technology, Netherlands

Abstract—The electrical smart grid connection requires also perspective with little consideration of building side value
services from the buildings. The performance demanded during perspective and actual operational implications. Also, the
the interactions between buildings and smart electrical grids are studies have not been case specific. This in our opinion
multifaceted, involve attainment of various objectives which are deprives stakeholders of scenario details that are crucial in
sometimes in conflict with each other and require dynamic phenomenal understanding. Using a current study of an office
decisions and operational coordination. This is added to the fact building in Breda in the Netherlands, we present a framework
some of these decisions and coordination activities are required for multi-agent system approach based decision making and
at localised level. Multi-agent system approach has been favoured coordination of interactions between buildings and smart
in recent studies for realisation of these desired performance
electrical grids that is value focused. The paper makes specific
objectives. However, these studies have been mostly simulations
contribution to multi-agent framework for building energy
based and representative of the grid side perspective with little
consideration of building centred value and operational
management for smart grid connectivity for a typical Dutch
implications. This paper presents a conceptual multi-agent office. The paper is divided into sections discussing value and
framework based on a scoring rule that takes into account the performance during interactions with electrical smart grids,
building management perspective. The suggested framework is multi-agent frameworks, the case study, and conclusion
particular for ongoing tests at an office building located in Breda,
Netherlands. II. VALUE AND PERFORMANCE IN BUILDINGS
Keywords- multi-agent system; buildings; smart electrical grids; In the context of interactions with the smart grids, building
performance focused value becomes a negotiation between four main
players. These are the occupant, building manager, the network
operator and power supplier. The occupant is central as the
I. INTRODUCTION
building’s energy demand is purposely to ensure his/her
Nearly all development around smart grids are focusing on productivity, comfort and wellbeing. Thus for the occupant the
the grid side issues and hence power systems reliability, the core issue will be building comfort and usability. Comfort in
buildings which provide significant sources energy in the new buildings is a sum of characteristics mainly defining indoor air
concept are ignored. This needs to change in efforts to improve quality, acoustic, visual and thermal aspects. Lighting
performance at the building side. The performance demanded characteristics are deductible from background colour,
during the interactions between buildings and electrical smart personal preferences and fenestration, they can be measured
grids is influenced by a number of factors which appear using a single value of lux whereas carbon dioxide
complex. These include the use of a large number of concentration is often representative of indoor air quality [8].
components from multiple vendors, functional diversity in For thermal comfort in buildings, key characteristics
terms of technological mixes used and requirement for equally considered are air temperature, mean radiant temperature,
diverse demand performance characteristics. For example, humidity, air velocity, , thermal insulation of individual
pursuit of maximum comfort goes against the grain of clothing and metabolic activity level if the heat balance method
optimization for energy efficiency and decentralized control is used[9-13]. For the building manager value is all about
advocated by localised optimization may also breed delay in keeping operations cost effective and within budgetary limits
decisions and eventual loss of reliability. It is also noted that whilst also ensuring acceptable levels of productivity and
buildings and connected auxiliary infrastructure such as comfort. This implies that the manager has to try as much as
electrical smart grid are characterized by robust information possible to keep the operational utilities budgets low by either
exchange, elaborate processes involving multiple actors with restricting energy consumption, shifting demand to shave
diverse interests and equally diverse operational systems. peaks, using hybrid energy systems that may include renewable
Solution to these complexities requires use of a combination of micro-sources or providing auxiliary energy services to the
negotiation, competition and balanced approach. This makes grid. In all this, the energy manager can be termed as
multi-agent system (MAS) approach ideal as a coordination managing the expectation of the grid whilst also ensuring the
and decision tool for optimal coordination of interactions comfort of the user. These are two conflicting requirements as
between buildings and electrical smart grids [1-7]. However, maximization of one may easily lead to minimization of the
these studies have been mostly representative of the grid side
other. For the building side, it’s always desirable that comfort buildings. We specifically focus on studies that involved actual
value be weighted more than network need. implementation of multi-agent systems in comfort or energy
management. Historically there has been no study based on
The electrical grid exists to ensure accessibility to reliable multi-agent systems that involves actual interactions between
electricity supply. For electrical smart grids, the specific buildings and the grid. Studies involving multi-agent systems
functions as demand satisfaction in a manner that is cost use in buildings and the grids have been silo-structured such
effective and assures continued reliability, resilience and that actual tests are limited and where they occur the scope is
efficiency [14, 15, 16]. The network concept of value must specific to either building coordination or grid activities. For
therefore revolve around satisfaction these functions. office buildings three actual studies have been conducted
Subsequently value with respect to network’s perspective within the context of comfort and energy management, these
would span across demand connectivity, power quality, cost are the ABB experiment, SAVES experiment and the historical
effectiveness, operational and sustainability characteristics. Xerox building experiment [26, 27, 28, 29]. Analysis of these
Operational characteristics are important in that they define the three cases were done on the basis of focus of the study, agents
ability and latitude within which the grid can cope with definition, interactions with the grid and decisions framework.
changing dynamics in demand and supply. Sustainability in Three main revelations are apparent in the cases analysed ,
this case is mostly born of the increasing need to reduce energy these are:
related carbon dioxide emissions to the environment [21]. It is
implied by van Overbeeke and Vaughan [21] that ultimately, 1. The use of hierarchical structure was observable in all
the realisation of value in the interactions between buildings the cases. However in the case of Xerox and SAVES, localised
and electrical smart grids is hinged on availability of a tariff decision making was also allowed within the hierarchy.
system to guide service within the network. This may mean
2. In all the cases, there was no interaction with the grid.
charging for services in relation to the designed performances
This could be due to the fact that existing regulations on
as highlighted above. This makes it also logical that the
electrical installations until recently did not allow for
framework for MAS based interactions between buildings and
interference with the meter. The meter still remains part of
electrical smart grids should also be constructed around value
electrical utility company assets and communication between
and specifically on value-based tariff contract.
buildings and the grid is still below par. However, in the ABB
study, allowance was made for a smart meter.
III. MAS FRAMEWORK
3. All the cases idealize incorporation of local
The previous section has outlined the concept of value and environment conditions and user preference in decision
associated performance characteristics in the interactions making.
between buildings and electrical smart grids. This section
examines details of MAS frameworks and their suitability in Having examined available options in structuring MAS, it
coordination and decision making for buildings-electrical smart is imperative that this be contextualized to the case of
grids interactions. Key components of any multi-agent system interactions of buildings and the smart grids. Generally, two
are agents, environment and interactions. options are available for MAS based problem solving.
Philosophically, MAS framework design must realize Bond and Gasser [29] reveal these as either coordination or
orderly implementation of functions. To accomplish this a tasks decomposition based. In the coordination approach,
number of parameters are considered. These are mainly several agents coordinate their knowledge, activities and
grouped as authority delineation and communication, decision reason about the processes whereas the decomposition
making and tasks allocation and access to resources [25]. In approach breaks down the problem into numerous functional
authority and communication considerations, the choice is nodes for task distribution and knowledge sharing in efforts to
between collaborative and hierarchical frameworks. find practicable solution. The decomposition approach is
Collaborative frameworks tend to be task oriented and random chosen for this case due to its ability to simplify the problem.
in nature. Hierarchical structure tends to be regimented with a The decomposition is undertaken in two-phases. First, the
clearly defined top down authority structure. Communication processes are decomposed according to the level of energy
follows the lines of authority in hierarchical case. Available
abstraction (that is, decomposition is at grid, building, floor,
options when considering tasks allocation and decision
structures are (i) specialization framework where agents are room, workplace and user levels). The second phase
developed only for dedicated tasks and decisions are mono- decomposes value concepts associated with the tasks. This is
channel are dedicated to specific agents, and (ii) shared pool or similar to the proposal by Zeiler et al. [30] except a layer on
synergistic agents whereby tasks are pooled, decisions are value perspective is added atop the functional decomposition.
made in network mode. The later scenario is more economical This would take cognizance of the decomposed processes and
and avoids domino like failure common for mono-channel associated value attached. The value attributes could be
situations. expressed in terms of cost equivalent to simplify decision
processes. This is in line with a suggestion by van Overbeeke
When considering resources, focus is on whether they are and Vaughan [21] that the ultimately, desired energy
either dedicated for use to a multiple agents (Non-specialized
characteristics for active networks are easily are realisable by
resource structure) or to one or some few similar agents
interlinking service to tariff structure. In our approach, the cost
(Specialized resource structure). It is important that these
structural formations be located within past studies for office categories are made for various services within the network.
Agent Role
WEATHER AGENT
Building Manager This is the super-agent. It undertakes overall coordination of activities of agents, it’s the ultimate
enforcer of order. Also tasked with ultimate decision making for comfort & energy management. It’s
the ultimate value enforcer.
Grid Value Communicates the effect of building operations on the local grid to the Building Manager Agent. Also
communicates quantitative energy and production to the grid SCADA.
Comfort Value Communicates user comfort preferences to the Building Manager Agent.
Device Manager Local Coordinates & communicates actual settings of locally installed services installations, such as
PREDICTION lighting, radiator panels and plug load activities to the Building Manager Agent.
AGENT
Device Manager Coordinates, communicates actual settings & load profiles of centralised installed services
ARCHIVE AGENT Central installations, such as lighting, boilers, HVAC Controls, Humidifiers to the Building Manager Agent.
Also coordinates & communicates energy generation profiles to the Building Manager Agent.
SENSOR NETWORK
AGENT Zone Communicates zonal environmental conditions to the Building Manager.
Aggregates energy requirements at workplace, room and floor level installation to the Building
Manager Agent.
Prediction Predicts energy production and load profiles in several time frames (10 minutes, 1 hour, 3 hours, 1
day, 3 days) then communicates the results to Building Manager Agent.
Sensor Network Communicates measurements on equipment set-points, operation conditions and ambient
environment to the Building Manager Agent and Prediction Agent.
Archive Communicates archive data from sensor network to the Prediction Agent to improve accuracy of
prediction.
Weather Undertakes basic analysis of prevailing weather parameters and communicates them to the Building
Manager Agent.

Figure 1: Agent Framework for the case of Breda Building


The specific costs categories conceived are comfort, building (see Figure 1). Ten agents that are hierarchically
administration, connectivity, network access, reliability, structured are suggested for the case building. The agents
reactive and time of use (see Table 1). Specific weightings are identification and roles are also outlined in Figure 1.The
then allocated to each category to assist respective agents with Building Manager Agent makes decision on the basis of a
decision making. In this case it is suggested that the weighting scoring rule. This is set such that any decision that promotes
for comfort be more than that for other grid characteristics comfort is rewarded at for all agents. This borrows greatly from
which are mostly power based. However, a maximal limit rule Ramchurn et al. 2011 which uses a similar system for decision
is imposed to ensure that comfort is within allowable budget of making in a demand side management scenario for households
the building manager. This approach is a bottom up in decision in a smart grid. Also, the maintenance of operations in a
making. Consequently individual comfort requirements and manner that does not go beyond power quantity budgets and
interactions with the building system is given priority in conformance to power quality requirements is rewarded. This
decision making. The grid/network requirements are encourages value based on both comfort and energy
considered peripheral and are only adhered to after comfort performance. The suggested multi-agent framework is largely
provision. This implies that in attempts to balance budgetary hierarchical but allows for minimal localised decisions to
provision against resource requirements, the building manager improve performance.
prioritizes comfort value against provision of additional
services to the grid. V. CONCLUSIONS
We have presented the concept of value as a basis of
IV. THE CASE SPECIFICS negotiating the conflict between comfort demand and optimal
The above MAS framework is idealized for implementation energy management in buildings and the local grid. We have
in an ongoing study of an office building in Breda, also introduced the issue of maintenance of power quality
Netherlands. This paper has been written as part of requirements at local grid as important when coordinating
supplementary studies aimed at realizing design of multi-agent comfort and energy management in buildings. The paper has
based building energy management system. This system will outlined the multi-agent system framework for use in
be based on a proprietary insiteview building management negotiating value in the interaction between building and the
system that is currently on use at the test building. The agent smart grid. Preparation is currently underway to practically
platform used in the study is EVE . The test building is a 3 operationalize the agents. This will ultimately contribute to
floor one with an approximately total floor area of the test sustainable comfort and energy management in building-grid
building is 1000 m2. Building services at this office is interactions. Though not contributing directly to energy
centralized with distinct ventilation, cooling and heating zones. efficiency and renewability, the use of multi-agent system
Heating is provided by a gas fired boiler with hot water enables additional flexibility and dynamism that is potentially
radiators grouped according into North East or South West crucial for further maximisation operational efficiencies,
orientation. Ventilation is effected by a large supply fan comfort and energy efficiency. These can be indirectly tapped
according to the 3 supply groups (North East, South West, and in the integration of renewable energy supply.
Electrical Engineering). Cooling system is also organised
based on orientations similar to the ventilation system. REFERENCES
Application of the findings discussed in sections above [1] H. Akkermans, and Fredrik Ygge. "Decentralized markets versus central
results to the following agents interactions structure for the test control: A comparative study." arXiv preprint arXiv:1106.0223 (2011).
[2] Z. Wang, A. I. Dounis, L. Wang, and R. Yang, "An information fusion [17] B. Dupont, L. Meeus, and R. Belmans, "Measuring the “smartness” of
based multi-agent control system for indoor energy and comfort the electricity grid." Energy Market (EEM), 2010 7th International
management in smart and green buildings." In Power and Energy Conference on the European. IEEE, 2010.
Society General Meeting, 2011 IEEE, pp. 1-8. IEEE, 2011. [18] K. Voss, and E. Musall, “Net zero energy buildings: International
[3] Z. Wang, R. Yang, and L. Wang, "Multi-agent control system with projects of carbon neutrality in buildings”, Walter de Gruyter, 2013.
intelligent optimization for smart and energy-efficient buildings." In [19] J. F. G. Cobben, “Power Quality: About the problems and solutions”,
IECON 2010-36th Annual Conference on IEEE Industrial Electronics COED, Arnhem, 2012.
Society, pp. 1144-1149, IEEE, 2010.
[20] S. Bhattacharyya, Power quality requirements and responsibilities at the
[4] M. Pipattanasomporn, H. Feroze, and S. Rahman. "Multi-agent systems point of connection, PhD dissertation, Eindhoven: Technische
in a distributed smart grid: Design and implementation." In Power Universiteit Eindhoven, 2011.
Systems Conference and Exposition, 2009. PSCE'09. IEEE/PES, pp. 1-
[21] F. van Overbeeke, and V. Roberts, "Active networks as facilitators for
8. IEEE, 2009.
embedded generation." Cogeneration and On-Site Power Production 3,
[5] A. I. Dounis, and C. Caraiscos, "Advanced control systems engineering no. 2 ,2002, pp. 37-42.
for energy and comfort management in a building environment—A
[22] M. F. McGranaghan, and H. W. Beaty, “Electric power systems
review." Renewable and Sustainable Energy Reviews 13, no. 6, 2009,
quality”, McGraw-Hill, 2002.
pp. 1246-1261.
[23] I. Lampropoulos, W. Kling, P.F. Ribeiro and Berg, J. van Den, History
[6] S. D. Ramchurn, P. Vytelingum, A. Rogers, and N. Jennings, "Agent-
of Demand Side Management and Classification of Demand Response
based control for decentralised demand side management in the smart
control Schemes, Proceedings of the 2013 IEEE PES General Meeting,
grid." In The 10th International Conference on Autonomous Agents and
Vancouver, Canada, (2013)
Multi-agent Systems-Volume 1, pp. 5-12. International Foundation for
Autonomous Agents and Multi-agent Systems, 2011. [24] J. Agüero, F. Rodríguez, and A. Giménez, "Energy management based
[7] M. Wooldridge, An introduction to multi-agent systems. Wiley. com, on productiveness concept," Renewable and Sustainable Energy
2008. Reviews vol. 22, 2013, pp. 92-100.
[8] D. H. W. Li, L.Yang, and J. C. Lam, "Impact of climate change on [25] K.M. Carley and L. Gasser, “Computational Organisation Theory”. In
energy use in the built environment in different climate zones–a review." A.H. Bond and L. Gasser, editors, Readings in Distributed Artificial
Energy 42, no. 1, 2012, pp. 103-112. Intelligence, pages 3-35. 1Morgan Kaufmann, San Mateo, CA, 1988.
[9] R. De Dear, "Thermal comfort in practice." Indoor Air 14, no. s7, 2004, [26] S. H. Clearwater and B. A. Huberman, “Thermal markets for
pp. 32-39. controlling building environments. Energy Engineering; (United States)
91.3, 1994.
[10] ANSI/ASHRAE Standard 55, Thermal Environmental Conditions for
Human Occupancy, ASHRAE Inc., Atlanta, 2004. [27] J. Kwak, P. Varakantham, R. Maheswaran, Y.Chang, M. Tambe, B.
Becerik-Gerber, and Wendy Wood, "TESLA: An energy-saving agent
[11] E. Halawa, and J. van Hoof, "The adaptive approach to thermal comfort: that leverages schedule flexibility." In Proceedings of the 2013
A critical overview." Energy and Buildings vol. 51, 2012, pp. 101-110. international conference on Autonomous agents and multi-agent
[12] J. F. Nicol, and M. A. Humphreys, "Adaptive thermal comfort and systems, pp. 965-972. International Foundation for Autonomous Agents
sustainable thermal standards for buildings." Energy and buildings vol. and Multi-agent Systems, 2013.
34, no. 6, 2002, pp. 563-572. [28] L. Klein, J. Kwak, G. Kavulya, F. Jazizadeh, B. Becerik-Gerber, P.
[13] W. Saman, "Study of the effect of temperature settings on AccuRate Varakantham, and M. Tambe. "Coordinating occupant behavior for
cooling energy requirements and comparison with monitored data, PhD building energy and comfort management using multi-agent systems."
diss., University of South Australia, 2008. Automation in Construction vol. 22, 2012, pp. 525-536.
[14] Y. Yanshan, J. Yang, and B. Chen, "The smart grids in China—a [29] D-Y. Yu, E. Ferranti, and H. Hadeli, "An intelligent building that listens
review."Energies 5, no. 5, 2012, pp. 1321-1338. to your needs." In Proceedings of the 28th Annual ACM Symposium on
[15] Y. Cheng, “Architecture and principles of smart grids for distributed Applied Computing, pp. 58-63. ACM, 2013.
power generation and demand side management,” in SMARTGREENS, [30] A.H. Bond and L. Gasser. “An analysis of problems and research in
2012. DAI”, In A.H. Bond and L. Gasser, editors, Readings in Distributed
[16] J. G. Slootweg, E. Veldman, and J. Morren, "Sensing and control Artificial Intelligence, pages 3-35. Morgan Kaufmann, San Mateo, CA,
challenges for Smart Grids." In Networking, Sensing and Control 1988.
(ICNSC), 2011 IEEE International Conference on, pp. 1-7. IEEE, 2011. [31] W. Zeiler and P. Savanovic, "General Systems Theory based Integral
Design Method." In Proceedings of the 17th International Conference on
Engineering Design (ICED'09), Vol. 5, pp. 217-228. 2009.

S-ar putea să vă placă și