IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 49, NO. 1, JANUARY/FEBRUARY 2013
1
An Energy Management System for Building
Structures Using a Multi-Agent Decision-Making
Control Methodology
Peng Zhao, Student Member, IEEE, Siddharth Suryanarayanan, Senior Member, IEEE, and
Marcelo Godoy Simões, Senior Member, IEEE
Abstract—Building energy management systems (BEMS) must
consider energy utilization efficiency improvement, energy cost
reduction, and renewable energy technology utilization in order
to serve local energy loads in building structures with dispersed
resources. The distributed management of building energy system
proposed in this paper utilizes a semi-centralized decision-making
methodology using multi-agent systems for BEMS for electrical,
heating, and cooling energy zones with combined heat and power
system optimizations aimed at improving energy efficiency and
reducing energy costs. A case study is presented to demonstrate
the validity of the proposed energy management scheme.
Index Terms—Building energy management systems (BEMS),
cyber-physical systems (CPS), distributed generation, multi-agent
systems (MAS), net-zero energy buildings.
I. I NTRODUCTION
T
HE PATH forward for sustainable energy for society is in
the forefront of public interest, and it is a high priority
for policy makers. In the US, two energy policy acts have
been passed in the last decade, in 2005, and 2007, which
include programs for conservation and energy development,
with Grants and tax incentives for both renewable energy and
non-renewable energy [1], [2]. There have been multiple goals
and provisions for the improvement of energy efficiency by
incentives for investments in modernization of the energy infrastructure based on the fact that the cost of generating energy
is greater than the costs for savings. The US National Science Foundation has been supporting the application of cyberphysical systems (CPS), i.e., a contemporary breed of systems
Manuscript received July 31, 2011; revised February 6, 2012; accepted
May 1, 2012. Paper 2010-IACC-545.R1, presented at the 2010 Industry Applications Society Annual Meeting, Houston, TX, October 3–7, and approved
for publication in the IEEE T RANSACTIONS ON I NDUSTRY A PPLICATIONS
by the Industrial Automation and Control Committee of the IEEE Industry
Applications Society. This work was supported by National Science Foundation Award 0931748 CPS Cyber-Enabled Efficient Energy Management of
Structures.
P. Zhao is with the Department of Civil, Environmental, and Architectural Engineering, University of Colorado, Boulder, CO 80309 USA (e-mail:
Peng.Zhao@Colorado.edu).
S. Suryanarayanan is with the Department of Electrical and Computer
Engineering, Colorado State University, Fort Collins, CO 80513-1373 USA
(email: sid@colostate.edu).
M. Godoy Simões is with the Department of Electrical Engineering and
Computer Science, Colorado School of Mines, Golden, CO 80401 USA
(e-mail: mgs@mines.edu).
Digital Object Identifier 10.1109/TIA.2012.2229682
with integrated computational and physical capabilities that can
interact with humans, expanding system capabilities through
computation, communication, and control. The work described
in this paper deals with an application of CPS for building
energy management via data fusion and analysis for real-time
monitoring, prediction, and control of energy management systems in modern buildings and cyber integration along the lines
of contemporary mandates. The primary objective of this paper
is the framework and algorithmic aspects of a CPS for future
generation building energy management systems (BEMS) and
does not include aspects of hardware and controller design in
this study.
The US electricity grid is steadily increasing in supply and
demand growth for the next two decades, while the trend in
transmission systems indicates lack of investments and generally restrictive regulatory barriers for new transmission lines
[3], [4], [18]. Therefore, management and control of distributed
energy systems at the consumer end present a significant avenue
for investigations and improvements [5]. Due to the limited
capabilities of centralized computing on large-scale distributed
systems, decentralized or semi-centralized decision-making
process is viewed as a suitable option for employment in
distributed energy systems. A possible approach to the solution
of managing distributed energy systems is by the utilization of
multi-agent systems (MAS) [19].
MAS can be considered an aggregation of networked agents
or controllers, for achieving some global objectives by coordination and communication among the agents [6]. In this
paper, the applicability of a MAS-based control methodology
for BEMS, as an example of a distributed energy system, is
presented.
Building structures in the US consume significant levels of
electrical energy and are responsible for substantial greenhouse
gas (GHG) emissions [7]. A framework for addressing energy
management has been proposed in [8] with assumed objectives
of increased energy efficiency, decreased operation cost of energy utilization, decreased dependence on use of fossil fuel for
energy needs, and consequently decreased GHG emissions. The
definition of “net-zero energy costs” for zero energy buildings
(ZEB) given in [7] is used to set an optimization goal for
the MAS-based BEMS described in this paper [8]. The main
purpose of this paper is to describe the framework and the algorithmic aspects of a CPS for increasing the energy utilization
efficiency by decision-making control methodology. A detailed
0093-9994/$31.00 © 2012 IEEE
2
IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 49, NO. 1, JANUARY/FEBRUARY 2013
Fig. 1. Physical layer of a building energy management system [8].
description of the physical aspect of CPS-enabled BEMS, the
CEBEMS, is available in [8]. The methodology presented in
the paper assumes an integrated data analysis program as in
[9]. The rest of the paper is organized as follows: Section II
describes the system organization of BEMS; Section III
presents the decision-making agents in three energy zone,
Section IV presents a case study for achieving minimum energy
cost, and Section V concludes.
II. S YSTEM S TRUCTURES OF BEMS
According to [10], for typical commercial buildings, 50%
of the energy is consumed in heating, ventilating, and airconditioning (HVAC) system; when combined with hot water
needs, the energy consumption in the heating and cooling energy zones may reach up to 60%. Lighting and office equipment
take most of the rest of the energy consumption, around 20% in
a commercial building [10].
During the last few years, there have been several advances
in the architecture and civil engineering fields related to energysaving techniques such as passive solar heating, passive cooling, natural ventilation, natural day lighting, LED light bulbs,
and certified appliances approaching net-ZEB goals [7].
The cyber-enabled BEMS (CEBEMS) based on the conceptual framework [8] is discussed in the following sections.
The CEBEMS proposed in [8] has particular focus on the
physical aspects, including novel hardware, and cyber aspects,
including energy management schemes for heating, cooling,
and electrical energy zones of a commercial building.
electrical zone may possess some renewable energy sources
(RES). In the case study presented in this paper, a photovoltaic
(PV) array with grid connection and combined heat and power
(CHP) units are used as generation units; the generation sources
are supported by an electric energy storage unit (e.g., battery
bank). Such a configuration can supply the electrical loads in
the building. The heating zone in the building has a solarthermal heater, recovered heat from CHP units, with a natural
gas furnace and thermal storage as heat generation and storage units. The heating loads may be split according to space
heating and hot water needs. The cooling zone possesses an
air-conditioning unit and an absorption chiller that uses the
recovered heat from CHP to provide space-cooling needs.
Probably not all the blocks shown in Fig. 1 are eventually
employed in a practical CEBEMS. However, the selection and
combination of the appropriate structure will depend on the
building size and energy needs. In commercial buildings, onsite electric generator combined with waste heat recovery and
absorption chillers are used for building cooling, heating, and
power (BCHP) system [11], where the waste heat from the
generator is utilized for both building heating and cooling.
Typically, the fuel utilization efficiency BCHP system is around
80%, some systems may exceed 90%, [12], which is much
higher than the maximum efficiency for delivered power of
a central power plant (i.e., 55%–60%), when the carbon savings of BCHP system is compared with a traditional boiler
and chiller system, the result is significant [12]. Commercial
buildings can also employ solar thermal panels for domestic hot
water (DHW), and solid oxide fuel cell in the CCHP system as
clean and RES for the purpose of GHG reduction and fossil fuel
independence.
B. Cyber Aspect of CEBEMS
The CPS of the proposed CEBEMS is achieved by a MAS
approach as shown in Fig. 2. Reference [13] lists some applications of MAS in building energy management. In each
zone identified in the previous subsection, the energy conversion, storage, and consumption are precisely measured and
dispatched by the intelligent agent embedded in that zone.
The respective agents are: 1) the E-agent for electricity; 2) the
H-agent for heating; and 3) the C-agent for cooling zones,
respectively. The three agents communicate with each other
through local area network when the energy management task
is beyond the capability of a single agent, or the agents are
required to work together for a series of tasks. The energy
management and control methods of three agents and their
communication are discussed next section.
III. M ULTI -AGENT D ECISION -M AKING S YSTEM
A. Physical Aspect of CEBEMS
Fig. 1 shows the building energy generation and storage and
consumption units and the energy flow paths. The objective of
such a system is to achieve overall high energy efficiency, low
emissions, and economic feasibility, without compromising the
preferences and comfort of consumers.
The proposed local BEMS has three zones of interest: the
electrical zone, the heating zone, and the cooling zone. The
In a commercial BCHP system, the energy system sizing
for electrical, heating, and cooling zones is very important,
because when the on-site generation is working, the recovered
waste heat should also be utilized simultaneously, otherwise,
the overall energy efficiency will go down, and its advantage of
energy efficiency and cost effectiveness will be lost. The system
sizing has three comparative references: 1) tracking electrical load, 2) providing electrical base load, and 3) following
ZHAO et al.: ENERGY MANAGEMENT SYSTEM FOR BUILDING STRUCTURES
Fig. 2.
3
Building energy management system via a multi-agent system approach [8].
thermal load. When sizing for tracking electrical load, the
building should be considered islanded from the grid, so it can
be energy independent from the service provider. Due to the
fuel cost, on-site generator maintenance cost, and time of use
electricity rate, this option 1) is not cost effective for buildings
connected with the grid. When providing electrical base load as
in 2), the system is designed for a grid connected building, so
that the surplus electricity generated during the hours of offpeak grid conditions may be sold back to the grid to offset
the electricity purchased during on-peak hours. However, the
thermal demand is higher during the day and drops to a lower
setpoint at night; this may lead to the thermal energy output
being over generated during the night and under-utilized during
the day. This condition results in surplus thermal energy wasted
during the night, and requirement of supplemental boilers for
the daytime. This results in extra installation costs, operational
budget, higher energy cost, and eventual loss of BCHP system
benefits. The approach taken in this current study is to size
the BCHP system based on 3), i.e., to follow thermal load and
electricity becomes a byproduct of the cycle, thus eliminating
the need for thermal storage. The generator is viewed as an
overall system, composed of a boiler to provide hot water for
building heating and also a heat source of absorption chiller for
building cooling demand. The electricity generation is viewed
as an additional service, which can at least offset some level of
demand, and any surplus is to be sold back to grid for profit.
According to [10], the thermal load is higher than electric load
in a typical commercial building, so sizing and scheduling the
BCHP system based on comparative reference 3) may provide
high energy utilization efficiency of BEMS, which will be
examined in the case study shown in Section IV whereupon to
the methodology to fully utilize the BCHP system in a BEMS
will be presented.
A. Energy Management in the Heating
Zone—Decision-Making Control of the H-Agent
The goal of the CEBEMS is to minimize the energy cost,
so the real energy cost in each zone must be examined before
any optimization technique is implemented for “minimized
cost.” The proposed BCHP system in the notional commercial
building is sized based on the heating load. The H-Agent is
responsible for fully utilizing the recovered heat for heating.
However, the space heating demand in each room of the building may not be accurately predicted, so a boiler is needed. In
the heating zone, the possible energy cost comes from the generation and distribution side of hot water. Hot water recovered
from generator is viewed as byproduct of electricity generation,
so the fuel cost is already considered in the electrical zone, and
this part of hot water is viewed as free of charge in heating
4
IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 49, NO. 1, JANUARY/FEBRUARY 2013
zone. Hot water produced from the supplemental boiler needs
natural gas, so this cost should be counted in the heating
zone. On the distribution side, regardless of the sources of hot
water—be it powered by natural gas boiler or recovered waste
heat of a generator, pumps need to be working all the time.
The electricity consumption is also counted in the electrical
zone, so the hot water distribution is viewed as free of charge
in the heating zone. Therefore, the minimum energy cost of a
day in the heating zone can be shown as (1), where the energy
consumption is computed every 15 min and the sum of the costs
in these time intervals will represent the cost for one day
min
24
NG
$N G × Et
max
Ei(t−t
D)
ij
min
Ei(t−t
D)
ij
ηi
provided by the ith supplier during t − tD
ij (kg),
minimum amount of hot water that must be provided
by the ith supplier during t − tD
ij (kg),
efficiency of the hot water distribution pipe connected with the ith heating supplier,
subject to the constraints:
Ei(t−tD ) ≥ Djt
ij
(1)
where $N G is the natural gas price, and EtN G is the hourly
natural gas consumption, and t is the time.
However, (1) does not show the energy-saving process (i.e.
the recovered hot water generation, distribution, and natural
gas burning reduction), and has no controllability on the energy
savings and natural gas burning minimization.
Therefore, the objective function in the optimization should
be changed from minimizing energy cost to maximizing energy
utilization efficiency. Due to the intrinsic losses in the hot
water pipeline and other thermal losses and from the hot water
generation side to the consumption end (i.e., heating coils or
radiators), and the water flow speed limit in distribution pipes,
the hot water distribution lagging time cannot be neglected
and feedforward control must be used. In [14], a Just-in-Time
(JIT) supply chain management method was first introduced for
a district heating system; JIT requires the desired amount of
hot water to be dispatched to the desired consumption end at
desired time [14]. This subsection will introduce the hot water
distribution optimization based on the idea of JIT aiming at
increasing hot water utilization efficiency for space heating.
In (2), it is shown to maximize the recovered heat utilization
efficiency, which will result in less natural gas burned in the
supplemental boiler, so the energy cost in heating zone is
minimized
Ejt
i=1 j=1 t=0
Ei(t−tD )
Eimax
(t−tD ) × ηi
(2)
where the indices are:
i ith heating supplier (from 1 to n),
j jth heating consumer (from 1 to m),
t time interval t,
and the variables are:
Ejt
amount of hot water delivered to the jth consumer
during t (kg);
amount
of hot water sent by the ith supplier during
Ei(t−tD
)
ij
t − tD
ij (kg),
and the parameters are:
Djt
hot water demand of the jth consumer during t (kg),
hot water distribution time from the ith heating
tD
ij
supplier to the jth heating consumer,
(4)
ij
ij
(5)
ij
B. Energy Management in the Cooling
Zone—Decision-Making Control of the C-Agent
In typical commercial central cooling systems, the cold water
from a central chiller system is distributed to cooling coils
located at handlers, where fans keep blowing when cycled on;
then, the cooled air flows into air-conditioned rooms with ventilation air. As the BCHP system is sized based on heating load,
the absorption chiller might not provide enough chilled water
for space cooling because of limited recovered heat. Therefore,
an electric chiller (i.e., water-compression chiller) must be
included in the cooling energy zone. Similarly, the chilled water
in BCHP system should be optimized for maximum usage. The
electric chiller is cycled on when the absorption chiller is not
able to provide enough cooling for the demand. In a commercial
central cooling system, (with only one electric chiller device)
the cooling demand is satisfied with only one absorption chiller
and one electric chiller. Therefore, the objective function in
(6) shows an optimization function embedded in the C-Agent,
where the chilled water provided from electric chiller is also included, because the cooling demand is supposed to be satisfied
by both the absorption chiller and the electric chiller
max
24
k q=1 t=0
ij
max
Eimin
(t−tD ) ≤ Ei(t−tD ) ≤ Ei(t−tD ) .
ij
n m 24
(3)
Djt ≤ Ejt ≤
t=0
max
maximum amount of hot water can possibly be
Eqt
E at−taD + E et−teD
( q )
( q )
(6)
where the indices are:
q qth cooling coil (from 1 to k),
t time interval t,
and the variables are:
amount of chilled water delivered to the qth cooling
Eqt
coil during t (kg),
a
amount of chilled water sent by the absorption
E(t−t
aD )
q
e
E(t−t
eD )
q
chiller during t − taD
q (kg),
amount of chilled water sent by the electric chiller
during t − teD
q (kg),
and the parameters are:
chilled water demand of the qth
Dqt
cooling coil during t (kg),
cold water distribution time from
taD
q
the absorption chiller to the qth
cooling coil,
ZHAO et al.: ENERGY MANAGEMENT SYSTEM FOR BUILDING STRUCTURES
TABLE I
D EMAND M ANAGEMENT OF E-AGENT [15]
teD
q
max
a
e
(E(t−t
aD ) + E(t−teD ) )
q
q
min
a
e
(E(t−t
aD ) + E(t−teD ) )
q
q
ηq
cold water distribution time from
the electric chiller to the qth cooling coil,
maximum amount of chilled water can possibly be provided by
the two chillers during the interval
that chilled water is sent for Eqt
(kg),
minimum amount of chilled water
that must be provided by the two
chillers during the interval that
chilled water is sent for Eqt (kg),
efficiency of the chilled water distribution pipe connected with the
cooling coil q,
subject to the constraints:
max
× ηq
Dqt ≤ Eqt ≤ E(at−taD ) + E(et−teD )
q
q
E(at−taD ) + E(et−teD ) ≥ Dqt
q
(7)
(8)
5
demand response to make a profit. On the other hand, the
building users’ specific levels of comfort vis-à-vis thermostat
setpoints and lighting levels are also respected. The local energy
consumption and generation data will be collected by metering
devices installed in the building, and then dispatched through
an apt communication protocol to a database server (e.g., host
server). The database server stores and archives the data, and
building users can access and download appropriate real-time
reports using dedicated or generic tools.
D. Interactions Between E-, H-, and C- Agents
The interaction between the three agents is shown in Fig. 3
where global information is shared in the computer simulation environment. However, the interaction between agents in
real deployments will require specific candidate networking
platforms.
IV. C ASE S TUDY: ACHIEVING M INIMUM E NERGY
C OST IN E FFICIENT B UILDINGS
In this section, a case study is presented to show the energy
saving capabilities of CEBEMS and a typical summer day and
winter day based on local weather data of Golden, CO. A building simulation prototype [16] is used to examine the overall
performance, where the utility energy price is based on raw
data provided by the EnergyPlus software [15]. The following
subsections describe some of the software applications used in
this case study.
q
min ≤ E(at−taD ) + E(et−teD )
E(at−taD ) + E(et−teD )
q
q
q
q
max
.
(9)
≤ E(at−taD ) + E(et−teD )
q
q
C. Energy Management in the Electrical
Zone—Decision-Making Control of the E-Agent
The energy management in electrical zone has two main
objectives: 1) demand management, i.e., the reduction of the
peak electric load; and, 2) communication with the utility
(service provider/grid) for demand response protocols and realtime prices of electricity and natural gas, trying to engage
the building to participate in demand response, and uploading
the building energy usage information on to an appropriate
database for certain authorized entities to download and use in
control actions.
CEBEMS allows demand management by reducing lighting
set-point levels and also setting back the cooling setpoint during
peak hours [15]. The demand management has three levels as
shown in Table I. The demand management level changes from
moderate to more aggressive. The purpose of load management
is to engage the building in demand response participation with
the utility, so there is a goal of making profits toward net-zero
costs for the building.
The E-Agent continuously communicates with the utility
for receiving energy prices and demand response information,
and attempts to reduce local load demand to participate in
A. Building Energy Simulation Environment
EnergyPlus: EnergyPlus is a building energy analysis and
simulation software, developed based on BLAST and DOE-2
[14]. EnergyPlus can calculate the building heating and cooling
loads, simulate building HVAC system, and energy generation
and consumption, based on user’s description of the building
and the associated energy system [14]. Newer versions of
EnergyPlus also include DG and RES (e.g., PV, fuel cell, wind
turbine and micro-CHP system) for energy efficient building
applications.
AMPL: AMPL is a powerful modeling language for both
linear and nonlinear optimization problems [17].
AMPL does not solve optimization problems directly, but
calls outside solvers, such as CPLEX, SNOPT, MINOS, IPOPT,
KNITRO, LANCELOT. Appropriate nonlinear programming
methods are SQP, based on Lagrange multipliers and IPM.
AMPL is a global solver and there is no concern about reaching
local extrema.
Control Optimization From AMPL to the Building Model
in EnergyPlus: Space heating and cooling water flows are
optimized (as introduced in Section III) by AMPL. The results
are the amount of water that should be sent from hot or chilled
water suppliers during each time interval (i.e., 15 min in this
case). Actuators in the water pumps change the water flow rate
(i.e., Qijt and Qcqt in heating and cooling zones, respectively)
according to the optimized value calculated by H- and C-agents.
EnergyPlus supports user-defined schedules of variable flow
pumps. Thus, water demanded at the supply side in each time
6
IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 49, NO. 1, JANUARY/FEBRUARY 2013
Fig. 3. Complete CEBEMS.
interval is input into the schedule of appropriate pump as
control signal from actuator, then the building heating and
cooling loads are satisfied by the optimized water flow schedule
calculated by AMPL.
TABLE II
B UILDING E NERGY C ONSUMPTIONS C OMPARED ON A T YPICAL
S UMMER DAY IN G OLDEN , CO
B. Building Test System
The CEBEMS is simulated for a single-floor, rectangular,
five-room setting of area 102.19 m2 , with details available in
EnergyPlus [16]. The building energy and cost saving will be
compared among: 1) base case where the building is connected
to the grid and air-conditioned by electric chiller and heated
by heating coils with hot water from a natural gas boiler;
2) BCHP model where the building has a micro-turbine combined with heat recovery for space heating and absorption
chiller for space cooling. The natural gas burner and electric
chiller are also installed as supplemental devices in BCHP, and
3) a CEBEMS model, based on the BCHP model but advanced
with the proposed E-, H-, and C-Agents.
C. Results
Tables II and III show the energy consumption at the end
use compared with the base case, BCHP model, and CEBEMS
model. The energy consumers in the base case are HVAC
system (i.e., cooling, boiler, and fans as shown in Table II),
interior and exterior lighting, and interior equipment. The
BCHP and CEBEMS model also have on-site generation with
of the hot water loop, the cold water loop, the condensing loop,
and the heat recovery loop and their associated water pumps
installed as the BCHP system. The interior lighting, exterior
lighting, and interior equipment (appliances and miscellaneous
loads) are base loads that are kept in the base load unless load
management from E-Agent is performed. The space heating
and DHW needs are satisfied by boilers in the base case and by
recovered heat in the other two models. Analysis can be made
from the three energy management system models compared in
Tables II and III, and the following observations are made:
1) The comparison of base case and BCHP model shows the
electricity usage in BCHP model has been significantly
reduced (particularly in the summer), but the natural gas
consumption has been greatly increased due to the BCHP
TABLE III
E ND U SES OF B UILDING E NERGY C ONSUMPTIONS C OMPARED
ON A T YPICAL W INTER DAY IN G OLDEN , CO
system installation that has changed the major energy
consumption from electricity to natural gas.
2) The comparison of BCHP model and CEBEMS model
shows that both the electricity and natural gas consumption have been reduced in the CEBEMS model, because
ZHAO et al.: ENERGY MANAGEMENT SYSTEM FOR BUILDING STRUCTURES
7
Fig. 4.
Comparison of BCHP and CEBEMS models on a typical summer day in Golden, Colorado.
Fig. 5.
Comparison of BCHP and CEBEMS models on a typical winter day in Golden, Colorado.
the H- and C- Agents have optimized the water generation
and dispatch from on-site generator, so excessive hot
water production is avoided.
3) The end energy uses indicated in Tables II and III did
not specify the energy sources, so on-site generation from
the BCHP model and CEBEMS model are further investigated in Figs. 4 and 5. The electric and thermal demand of
BCHP model are plotted in Figs. 4 and 5 for comparative
reference, because if no load management is performed
by the E-Agent, the energy demand is the same in these
two models. On the other hand, if load management is
performed, the energy demand in CEBEMS model will
be less, so only the energy demand in the BCHP model
will need to be plotted as a comparative reference.
The following results are observed from Figs. 4 and 5:
1) In the BCHP model, both the thermal and electric generations are greater than the demand; the surplus electricity
generation is sold back to the grid. Tables II and III also
shows that natural gas usage in boilers is kept zero in
the BCHP and CEBEMS model, showing that heating
demand is totally satisfied by recovered heat. Spacecooling electricity usage in the BCHP model is greatly
reduced when compared with the base case. However,
the absorption chiller did not totally replace the electric
chiller, because the BCHP system is sized based on the
space heating demand, so the absorption chiller’s chilled
water output is limited by the hot water produced by the
on-site generator.
8
IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 49, NO. 1, JANUARY/FEBRUARY 2013
2) In the CEBEMS model, the thermal energy output of
BCHP system is optimized by H- and C-Agents, the
thermal generation spikes are corrected with the calculated data, which is enough for the end-use requirement;
thus, excessive hot water generation is avoided. Notice
that the thermal generation in CEBEMS model can be
higher than the thermal generation of BCHP model. This
can be used as a preparation for the upcoming higher
energy demand period. As stated in Section III, the BCHP
model is following the thermal load, but that is high in
the early morning hours (7 AM) and evening (7 PM),
so the associated electric generation from the on-site
sources should also ramp up and down concomitantly,
which will potentially decrease the energy efficiency of
the generator. In the CEBEMS model, the thermal output
from the generator is relatively smoother than that in the
BCHP model, so the associated electric output is also
smoother, which helps the generator to avoid some fast
response actions to the sudden rising demand.
The electrical and thermal demand shown in Tables II and
III are the energy usage at the end-user level. The electrical
demand includes interior lighting, exterior lighting, interior
equipment, and fans, pumps, and air-conditioning system of the
HVAC system; and the thermal demand is the amount of energy
calculated to meet the specific indoor air heating or cooling
setpoints during a certain time. However, when employing the
EnergyPlus software for calculating the electrical and thermal
demand values at the end-user level, the following are not taken
into account: any energy consumed in the generation and the
distribution processes, any losses in the processes, and any of
the efficiencies of the system and equipment.
Tables II and III indicate the respective base loads of the
base case, the BCHP model, and the CEBEMS model for
comparisons between the three energy models, as well as show
the changes in energy usage with the progressive installations
of CHP system in the BCHP model and the cyber aspect in
the CEBEMS model. The electrical and thermal energy generation shown in Figs. 4 and 5, however, considers the energy
consumed in the process, losses, and efficiency of the system
and components.
Since CHP system plays a key role for energy generation in
the BCHP and CEBMES model, these values can be significant.
Tables II and III when combined with Figs. 4 and 5 provide
a qualitative description of the energy generation and demand
as follows. In the base case model, the electrical generation
exactly matches the demand, since it is assumed that the losses
are supplied by the grid/utility. When the thermal generation
from the gas boiler is multiplied by the efficiency and losses
are subtracted, it equals the thermal demand. In the BCHP
model, when the electrical generation from the CHP system
is accounted for the efficiency for converting natural gas to
electric power, any differences between the electric energy
bought from the utility and the electric energy used inside
the BCHP model and losses should equal the demand. Similar
accounting procedures are handled for balancing the thermal
generation and the thermal demand. In the CEBEMS model,
the physical aspect remains the same as the BCHP model; thus,
the same statements are also hold true here; the addition of the
TABLE IV
E NERGY C OSTS C OMPARED ON A T YPICAL W INTER DAY
AND S UMMER DAY
cyber aspect helps to increase the efficiency of the system, thus,
less energy is consumed in the processes.
Table IV shows the energy cost for the base case, BCHP
model, and CEBEMS model based on the utility dynamic
energy price provided as raw data in the example model in
EnergyPlus [16].
The total utility cost shown in Table IV is calculated by
the following methodology: the sum of energy charges, demand charges, and service charges constitutes the basis, which
is added with adjustments and surcharges that constitute the
subtotal, added with taxes (8% in this case). In this case, the
adjustments and surcharges are zero, so they are not shown in
Table IV. For energy charges, unit cost of electricity and natural
gas price are defined as 10.23 cents/kWh and $9.55/MCF,
respectively.
The service charges are determined by the electricity demand
of the building from the utility. If the building peak demand is
above 10 kW, then the service charge will be $87.3 per month;
if not, then the service charge will be $12.74 per month. The
services charges applied here are divided by 30 to reflect the
cost for one day of the month. The demand charges are employing block rates determined by building service charges type.
For detailed block rates, refer to [16]. In this case study, the
energy cost of the BCHP model has a significant cost savings
compared to the base case, because the BCHP system is energy
efficient in nature and the CEBEMS model has more energy
cost savings compared with the BCHP model due to the cyberenabled MAS decision-making control system working toward
optimizing overall energy utilization. For real implementation,
the real cost savings will be dependent of the prescribed utility
tariffs.
Fig. 6 compares the performance of BCHP and CEBEMS
energy management approaches, showing that in a typical 24 h
time span, there are significant reductions of thermal power
generation, and at the same time the electrical power generation
is met without constraints, showing the improved performance
achieved by the present control methodology.
V. C ONCLUSION
This paper investigated an application of MAS for cyberenabled energy management of building structures known as
CEBEMS. The efficient energy management system is achieved
by tapping both physical and cyber aspects of the building,
such that the BCHP building model provided an applicable
ZHAO et al.: ENERGY MANAGEMENT SYSTEM FOR BUILDING STRUCTURES
Fig. 6. Comparison of electrical and thermal energy consumption for BCHP
and CEBEMS models.
energy management computational structure for energy efficient buildings. The CEBEMS model advances a BCHP model
and optimizes the energy generation and distribution. The test
building chosen in the example shown in this paper is for a
typical food service center, whose thermal demand is relatively
constant compared with common commercial office building;
it is expected that the control methodology studied in this work
can be applied for office building as well.
ACKNOWLEDGMENT
The authors acknowledge the Center for Advanced Control
of Energy and Power Systems at Colorado School of Mines and
are grateful to the help of Dr. A. Newman at Colorado School
of Mines and Dr. S. Leyffer at Argonne National Laboratory
on AMPL.
R EFERENCES
[1] 109th Congress of the United States of America, Energy Policy Act 2005
(EPACT05), Aug. 2005.
[2] 110th Congress of the United State of America, Energy Independence
and Security Act of 2007 (EISA07), Dec. 2007.
[3] GRID 2030: A National Vision for Electricity’s Second 100 Years, United
States Department of Energy, Washington, DC, Jul. 2003.
[4] B. L. Dorgan, “The case for a 21st century electricity transmission system,” Democrat. Policy Committee, Washington, DC, Mar. 2009.
[5] H. E. Brown, S. Suryanarayanan, and G. T. Heydt, “Some characteristics
of emerging distribution systems under the smart grid initiative,” Elsevier
Elect. J., vol. 23, no. 5, pp. 64–75, Jun. 2010.
[6] M. Wooldridge, “Intelligence agents,” in Multiagent Systems—A Modern
Approach to Distributed Artificial Intelligence. Cambridge, MA: MIT
Press, 1999, pp. 27–77.
[7] P. Torcellini, S. Pless, and M. Deru, “Zero energy buildings: A critical look
at the definition,” presented at the ACEEE Summer Study, Pacific Grove,
CA, Aug. 2006. [Online]. Available: http://www.nrel.gov/docs/fy06osti/
39833.pdf
[8] R. Carnieletto, D. I. Brandão, S. Suryanarayanan, F. Farret, and M. Godoy
Simões, “Smart grid initiative,” IEEE Ind. Appl. Mag., vol. 17, no. 5,
pp. 27–35, Sep./Oct. 2011.
[9] P. C. Peng, L. Jie, Y. Chongchong, and T. Li, “Design and implementation
of renewable energy and building integrated data analysis platform,” in
Proc. APPEEC, Mar. 2010, pp. 1–5.
[10] B. L. Capehart, W. C. Turner, and W. J. Kennedy, Guide to Energy
Management, 6th ed. Atlanta, GA: Fairmont, 2008.
9
[11] Building integrated cooling, heat and power for cost-effective carbon mitigation, The Climate Group. World Alliance Decentralized Energy, 2005. [Online]. Available: http://files.harc.edu/Sites/GulfCoastCHP/
Publications/BuildingIntegratedCooling.pdf
[12] D. W. Wu and R. Z. Wang, “Combined cooling, heating and power:
A review,” Prog. Energy Combust. Sci., vol. 32, no. 5–6, pp. 459–495,
Sep.–Nov. 2006. [Online]. Available: http://www.sciencedirect.com/
science/article/pii/S0360128506000244
[13] L. Klein, J.-Y. Kwak, G. Kavulya, F. Jazizadeh, B. Becerik-Gerber,
P. Varakantham, and M. Tambe, “Coordinating occupant behavior for
building energy comfort management using multi-agent systems,” Autom.
Construct., vol. 22, pp. 525–536, Mar. 2012.
[14] F. Wernstedt and P. Davidsson, “A multi-agent system architecture for
coordination of just-in-time production and distribution,” in Proc. SAC,
2002, pp. 294–299.
[15] EnergyPlus Documentation. EnergyPlus, Version 5.0.0, Apr. 2010.
[16] MicroCogeneration.idf, EnergyPlus, Version 5.0.0, US Department of
Energy, Apr. 2010.
[17] R. Fourer, D. M. Gay, and B. W. Kernighan, AMPL—A Modeling Language for Mathematical Programming, 2nd ed. Pacific Grove, CA:
Brooks/Cole, 2003.
[18] M. G. Simões, R. Roche, E. Kyriakides, S. Suryanarayanan,
B. Blunier, K. D. McBee, P. H. Nguyen, P. F. Ribeiro, and A. Miraoui,
“A comparison of smart grid technologies and progresses in Europe
and the U.S.,” IEEE Trans. Ind. Appl., vol. 48, no. 4, pp. 1154–1162,
Jul./Aug. 2012.
[19] X. Y. Yu, C. Cecati, T. Dillon, and M. G. Simões, “The new frontier
of smart grids,” IEEE Ind. Electron. Mag., vol. 5, no. 3, pp. 49–63,
Sep. 2011.
Peng Zhao (S’09) received the M.S. degree in engineering from the Colorado School of Mines, Golden,
in 2010. He is currently working toward the Ph.D.
degree in the Department of Civil, Environmental,
and Architectural Engineering, University of Colorado, Boulder.
His research interests are commercial building and
electric grid systems integration, intelligent control
systems for buildings, and building energy efficiency.
Siddharth Suryanarayanan (S’00–M’04–SM’10)
received the Ph.D. degree in electrical engineering
from Arizona State University, Tempe, in 2004.
He is an Assistant Professor in the Department
of Electrical and Computer Engineering at Colorado
State University, Fort Collins, where he serves as
the Site Director of the Center for Research and
Education in Wind, and as a 2011–2012 Resident
Faculty Fellow of the School of Global Environmental Sustainability. His research interests are in
the design, operation, and economics of finite-inertia
systems and integration of renewable energy systems to electric grids.
Marcelo Godoy Simões (S’89–M’95–SM’98) received the B.S. and M.S. degrees from the University
of São Paulo, Brazil, in 1985 and 1990, the Ph.D.
degree from The University of Tennessee, Knoxville
in 1995, and the D.Sc. degree (Livre-Docência) from
the University of São Paulo in 1998.
He is an Associate Professor with the Colorado
School of Mines (CSM), Golden, where he has been
establishing research and education activities in the
development of intelligent control for high-powerelectronics applications in renewable and distributed
energy systems, and where he currently serves as the Director of the Center for
the Advanced Control of Energy and Power Systems. He has been involved in
activities related to the control and management of smartgrid applications since
joining CSM. In 2002, he received an NSF CAREER Award “Intelligent-Based
Performance Enhancement Control of Micropower Energy Systems.” Currently,
he is the Chair for the IEEE IES Smart Grid Committee.