(5)
2) Building air ventilation
Time-varying ventilation systems, such as variable air
volume (VAV) and demand-controlled ventilation (DCV)
systems, have shown better performance in terms of energy
efficiency and control effects than constant air volume (CAV)
systems [9]. For a well-ventilated space, the mass balance of
CO
2
concentration is as follow [10]:
L N A A G
dt
dA
V
supply supply
* ) ( * + = (6)
where V is the volume of the ventilated space; A is the CO
2
concentration in the ventilated space; L is the average CO
2
concentration per person.
supply
A is the CO
2
concentration in
the supply air in the ventilated space, and it indicates the CO
2
concentration of the mixed air of the returned air and the
intake air which can be calculated as follows [11]:
supply return out return supply supply
G A G A G G A / ) * * ) (( + = (7)
In a short period of time, we can assume that N,
supply
A and
A are constants, Equation (6) can be solved as follows [10]:
It
supply
supply
supply
supply
e
G
L N
A A
G
L N
A A
+ + = * )
*
) 0 ( (
*
(8)
where, A(0) is the indoor CO
2
concentration at time 0 = t ; and
I = V G
supply
/ , is the space air change rate.
B. HVAC Equipment Models
The energy models of the primary equipment or
components will be established in advance for the sake of the
whole system's energy simulation. They mainly include:
1) Energy model of the fan
For a variable speed fan, the power is represented as
follows [9]:
) ( * ) ( *
* *
3
G vfd G motor trans
o o G air
fan
X X
H G X
E
= (9)
and,
o
G
G
G
X = (10)
where,
air
is the specific weight of the air;
G
X is the ratio of
air mass flow rate G to the rated air mass flow rate
o
G ;
o
H is
the rated head of fan (m);
trans
is the efficiency of
transmittance. The efficiency of motor ) (
G motor
X and the
efficiency of variable-frequency driver (VFD) ) (
G vfd
X vary
with the variation of rotation speed. They can be modeled as
follows:
3
3
2
2 1 0
) (
G G G G motor
X b X b X b b X + + + = (11)
3
3
2
2 1 0
) (
G G G G vfd
X c X c X c c X + + + = (12)
where b
0
, b
1
, b
2
, b
3
, c
0
, c
1
, c
2
, c
3
, are experimental factors.
2) Energy model of the heating unit and the cooling unit
The energy consumption of the heating coil [11]:
heat
mix supply p supply
heat
T T c G
E
) ( * *
= (13)
where,
supply
G is the mass flow rate of the supply air;
heat
is
the efficiency of the heating coil;
mix
T is the temperature of the
mixed air from the mix chamber, and it can be represented as
follows:
supply return out return supply mix
G T G T G G T / ) * * ) (( + = (14)
The energy consumption of the cooling coil:
COP
T T c G
E
mix supply p supply
heat
) ( * *
= (15)
where COP is the coefficient of performance, and it is
determined by the chiller part load ratio (PLR) [12].
The total consumed energy in the HVAC system is the sum
of the energy consumed by each component. The total energy
consumption is given as:
cool heat return fan supply fan
E E E E E + + + =
_ _
(16)
III. PROBLEM FORMULATION AND OPTIMAL CONTROL
APPROACH
Heating, ventilating and air-conditioning equipment is
primarily designed to provide a relatively constant and
comfortable temperature in the home. Secondary goals are to
provide fresh and filtered air, and a comfortable humidity
level. Effective control will be implemented to achieve these
goals with high energy efficiency in the building operations.
A. Problem formulation
When implementing control on HVAC system, the control
strategy is adjusted according to the real-time environment
condition. The HVAC controller uses information from
previous time steps to predict system response in the next time
step and makes decisions accordingly. There are two
objectives for the HVAC controller: minimizing the energy
consumption and maintaining a comfortable indoor
environment.
The control variables are the mass flow rate of the supply
air
supply
G , the mass flow rate of the return air
return
G and the
temperature of the supply air
supply
T . Since
supply return
G G 0 , we will use the ratio
supply return
G G / = instead of
return
G as the control variable.
The problem is formulated with multiple objectives that can be
expressed as follows:
Minimize
) , , (
_ _
supply supply
cool heat return fan supply fan
T G E
E E E E E
=
+ + + =
(17)
The total energy consumption is a function of the three
variables as shown in Eqn. (16).
Minimize
set in
T T = (18)
where is the difference between indoor temperature
in
T and
the set temperature
set
T .
Subject to
max
A A (19)
where
max
A is the maximum carbon dioxide concentration
allowed by occupants.
B. Multi-objective particle swarm optimizer
Since the two objectives are in conflict with each other,
more than one optimal solutions can be located in this
problem. That is, a set of trade-off solutions that represents the
best possible compromises among the objectives can be found
for the optimization problem. For multi-objective problems, it
is useful to find the Pareto-optimal front which is created by
the derived Pareto-optimal solutions.
Particle swarm optimization has been proven to be an
efficient method in solving optimization problems. A general
PSO algorithm works by updating the position of particles
from a swarm to find the optimal solutions. Each particle
adjusts its flying trajectory by continuously updating its
position and velocity [13]. The basic concept of the updating
criteria for a particle is to follow both the swarm leader and
the best solution of the particle itself.
To implement the PSO algorithm, the swarm is first
initialized including the initialization of both position and
velocity for each particle in the population. The personal best
pbest is initialized for each particle, and the corresponding
global best gbest is chosen to be the leader of the swarm.
Then, for a maximum number of iterations, each particle flies
over the search space by updating its velocity and position.
The updating rules of the velocity and the position of a particle
i are represented as follows:
) ( ()
) ( ()
2
1
i
i i i i
l gbest rand c
l pbest rand c v w v
r
r
r r
+
+ =
(20)
i i i
v l l
r
r r
+ = (21)
where
i
v is the velocity of particle; w is the inertia weight,
which represents the inertia effect for the particle;
1
c and
2
c are the learning factors:
1
c represents the learning ability
for flying towards the best position that the particle itself has
experienced;
2
c represents the learning ability for flying
towards the group leader;
i
l is the position of particle;
pbest for each particle and the leader gbest for the whole
population is also updated in the iterations.
The algorithm for multi-objective particle swarm
optimization (MOPSO) can be extended from the single-
objective PSO algorithm [14]. In this paper, a weighted
aggregation approach is introduced for solving this MO
problem. The weighted aggregation method converts a multi-
objective problem into a single-objective problem by
multiplying each objective with a user-defined weight and
adding them together. The weights are usually selected by the
importance of the corresponding objectives in the overall
problem. Normalization process is usually needed in weighted
aggregation method in order to scale objectives into an
identical order of magnitude.
For a bi-objective problem, it can be converted into a
single-objective problem as follows [15]:
Minimize
) ( * * ) 1 ( ) ( x f W x f * W
2 1
r r
+ (22)
where the weight W varies in the range of [0,1], and is the
scaling factor to enable the two objectives to be equally
treated.
To enhance the search efficiency, during the optimization
run of the PSO the inertia weight factor
u
decreases linearly as
follows:
iter *
iter
u - u
- u u
max
min max
max
=
(23)
where
max
iter is the number of total iterations and iter is the
current iteration number;
max
u and
min
u are the maximum and
minimum values of the inertia weight u , respectively.
By applying the MOPSO algorithm, the controller
determines the amount of energy dispatched to each HVAC
component and controls the corresponding variables.
IV. CASE STUDY
In this section, a case study is carried out based on a
developed building model. The control time step for decision
making is 15 minutes, that is, the system needs to adjust its
control strategy in every 15 minutes based on the
environmental change. The results of decision making in both
a time step and a 24-hour period are given in following two
subsections. The corresponding environment parameters and
the energy consumption are also given in this section.
The building under consideration is used for office and
business purposes and consists of two floors. The floor area is
15m*20m, and the height for each floor is 3m. The whole
building is considered as one thermal zone for control
purpose. The heat convection coefficients for the walls
are C m W h
wall
=
2
1 , and the heat convection coefficient for
the ceiling is C m W h
ceiling
=
2
3 . 3 .
A. Decision making in a time step
In this session we will discuss how to assign energy to fans
and heating/cooling unit at a certain time point to ensure best
performance. Similar decision making process is executed in
every 15 minutes based on the real time performance of the
building system. The targeted temperature which is set by
building occupants is C T
set
= 22 ; and the maximum CO
2
concentration allowed is ppm A 1000
max
= .This sample
decision making process is operated at the time point 0 = t . At
the time of 0 = t , the indoor temperature is C t T
in
= = 10 ) 0 ( ;
the indoor CO
2
concentration is ppm t A
in
1000 ) 0 ( = = ; the
outside temperature is C t T
out
= = 10 ) 0 ( ; the total number of
occupants in the building is 5 = N .
Using MOPSO, the Pareto front of the total energy
E
to the
difference between the target temperature and the indoor
temperature can be created. Each point in this Pareto front
represents for a possible solution.
0 2 4 6 8 10 12
5
10
15
20
25
30
35
40
45
50
55
|T
in
-T
set
| (degree Celsius)
E
n
e
r
g
y
C
o
n
s
u
m
p
t
i
o
n
(
k
W
)
Fig.2 Pareto front of the bi-objective problem
Table I gives three sample solutions of the mass flow rate
of the supply air
supply
G , the ratio of the mass flow rate of the
return air to the mass flow rate of the supply air
,
and the
temperature of the supply air
supply
T . The result indicates that
more consumed energy leads to higher indoor temperate.
TABLE I. SAMPLE SOLUTIONS
supply
G ) / s (m
3
supply
T C
E (kW)
in
T C
2521.65 0.9361 64.72 48.96 22
2633.82 0.6909 56.30 43.79 20.49
1278.38 0.2062 35.26 12.21 12.45
In this formulated problem, the indoor CO
2
concentration is
served as a constraint. In order to observe the impact of
different CO
2
concentration allowances on the results, the
Pareto fronts with different constraints are shown in Fig. 3.
Red points represent for the possible solutions with the
limitation of indoor CO
2
concentration ppm A 1000
max
= ;
Blue points represent the possible solutions with the limitation
of indoor CO
2
concentration ppm A 800
max
= . The results
show that with a looser restriction, the optimizer is capable of
finding more and better solutions.
0 2 4 6 8 10 12
5
10
15
20
25
30
35
40
45
50
55
|T
in
-T
set
| (degree Celsius)
E
n
e
r
g
y
C
o
n
s
u
m
p
t
i
o
n
(
k
W
)
A
max
=800ppm
A
max
=1000ppm
Fig.3 Pareto front of the bi-objective problem with different constraint
From the derived solutions, users can select the preferred
solution to realize the specific control goals in the building
environment. In every 15 minutes, possible solutions would be
generated and provided to users, from which users need to
choose a preferred solution for the control of next time period.
The controller can also select solutions automatically for users.
A possible procedure for selecting solution is to allow user to
set up a tolerance temperature and select the solution from the
range of tolerance randomly. This proposed procedure is
applied in the following study to select solutions automatically
for users.
B. Simulation in a 24-hour period
In this section, the proposed building is continuously
controlled by implementing the proposed control strategy in
HVAC system for a 24-hour period. Simulations are
conducted in both cold and hot weather to reveal the control
abilities of both the heating unit and the cooling unit. The
simulation results of indoor temperature and CO
2
concentration variation as well as the corresponding energy
usage are given in the 24-hour period.
As specified by the occupants, the target temperature
set
T
is set as C 22 , and the limitation of the indoor CO
2
concentration
max
A is set as 1000ppm. The outdoor CO
2
concentration
out
A remains in 350ppm for the whole day. The
variation of occupancy in a day is shown in Fig. 4. The CO
2
generation per person is person s m L / / 10 * 0052 . 0
3 3
= in an
office building.
0 4 8 12 16 20 24
0
10
20
30
40
50
60
70
80
90
100
Time (h)
N
u
m
b
e
r
o
f
O
c
c
u
p
a
n
c
t
s
Fig.4 Occupancy variation in a day
Firstly, the building performance in cold weather is
simulated to examine the control ability of the heating unit. At
the time of 0 = t , the indoor temperature C t T
in
= = 8 . 7 ) 0 ( ,
and the indoor CO
2
concentration ppm t A
in
600 ) 0 ( = = . The
temperature below C 20 is seen as uncomfortable. Thus, the
comfort range for temperature is [20 C , 22 C ]. The outdoor
temperature variation in a cold day is shown in Fig.5.
0 4 8 12 16 20 24
6
8
10
12
14
16
18
Time(h)
O
u
t
d
o
o
r
T
e
m
p
e
r
a
t
u
r
e
(
d
e
g
r
e
e
C
e
l
s
i
u
s
)
Fig.5 Outdoor Temperature variation in a cold day
Using the particle swarm optimizer, the controller derives a
set of possible solutions and a specific solution is selected
randomly from the solution set based on the user-defined
tolerance range. After running the optimizer for 96 times, the
simulation results of the variation of indoor temperature and
indoor CO
2
concentration are shown in Fig. 6 and Fig. 7,
respectively. As shown in Fig.6, the indoor temperature is well
maintained within the range of [20 C , 22 C ] from 00:30 in
this cold day. As shown in Fig.7, the indoor CO
2
concentration
is also successfully controlled below the limitation 1000ppm.
0 4 8 12 16 20 24
10
12
14
16
18
20
22
24
Time(h)
I
n
d
o
o
r
T
e
m
p
e
r
a
t
u
r
e
(
d
e
g
r
e
e
C
e
l
s
i
u
s
)
Fig.6 Indoor temperature variation in a cold day
0 4 8 12 16 20 24
550
600
650
700
750
800
850
900
950
1000
Time(h)
C
O
2
C
o
n
c
e
n
t
r
a
t
i
o
n
(
p
p
m
)
Fig.7 Indoor CO
2
concentration variation in a cold day
The energy consumption in a day is shown in Fig. 8. Table
II shows 24 solutions selected from all the 96 solution derived.
0 4 8 12 16 20 24
0
5
10
15
20
25
30
35
40
45
Time(h)
E
n
e
r
g
y
c
o
n
s
u
m
p
t
i
o
n
(
k
W
)
Fig.8 Energy consumption in a cold day
TABLE II. SELECTED SOLUTIONS
Time
supply
G ) / s (m
3
supply
T C
E (kW)
0:00 2250.8 0.548 61.19 41.54
1:00 1971.8 0.931 53.51 22.72
2:00 1713.8 0.956 56.16 21.12
3:00 4912.7 0.996 33.49 20.61
4:00 2635.7 0.999 44.88 21.45
5:00 2737.2 0.830 49.64 29.76
6:00 1641.7 0.617 55.86 23.63
7:00 1416.0 0.462 60.25 23.20
8:00 4837.5 0.507 25.81 18.91
9:00 7730.9 0.762 22.93 12.77
10:00 3302.4 0.276 25.70 15.07
11:00 2933.7 0.166 27.55 15.30
12:00 2366.7 0.124 26.61 10.30
13:00 2898.8 0.136 21.41 5.68
14:00 4418.1 0.425 20.56 3.30
15:00 2430.1 0.194 21.11 2.95
16:00 7628.9 0.635 20.55 1.92
17:00 4988.8 0.436 20.43 2.34
18:00 1530.9 0.492 30.18 5.90
19:00 1878.3 0.252 26.97 6.50
20:00 2069.3 0.612 34.69 11.94
21:00 2021.2 0.818 45.98 19.58
22:00 1558.3 0.922 55.33 19.66
23:00 6412.0 0.990 28.42 15.83
Secondly, the building performance in hot weather is
simulated to observe the control ability of the cooling unit. At
the time of 0 = t , the indoor temperature is C t T
in
= = 9 . 27 ) 0 ( ,
and the indoor CO
2
concentration is ppm t A
in
600 ) 0 ( = = . The
temperature above C 24 is deemed uncomfortable. Thus, the
comfort range for temperature is [22 C , 24 C ]. The outdoor
temperature variation in a hot day is shown in Fig. 9.
0 4 8 12 16 20 24
25
26
27
28
29
30
31
32
33
34
35
Time (h)
O
u
t
d
o
o
r
T
e
m
p
e
r
a
t
u
r
e
(
d
e
g
r
e
e
C
e
l
s
i
u
s
)
Fig.9 Outdoor temperature variation in a hot day
The simulation results of the variation of indoor
temperature and indoor CO
2
concentration in hot weather are
shown in Fig. 10 and Fig. 11, respectively. As shown in Fig.
10, the indoor temperature is well maintained within the range
of [22 C , 24 C ] from 00:45 in this cold day. As shown in
Fig.11, the indoor CO
2
concentration is also successfully
controlled below the limitation 1000ppm. The corresponding
energy consumption in this 24-hour period is shown in Fig.12.
0 4 8 12 16 20 24
10
12
14
16
18
20
22
24
Time (h)
I
n
d
o
o
r
T
e
m
p
e
r
a
t
u
r
e
(
d
e
g
r
e
e
C
e
l
s
i
u
s
)
Fig.10 Indoor temperature variation in a hot day
0 4 8 12 16 20 24
600
650
700
750
800
850
900
950
1000
Time (h)
C
O
2
c
o
n
c
e
n
t
r
a
t
i
o
n
(
p
p
m
)
Fig.11 Indoor CO
2
concentration variation in a hot day
0 4 8 12 16 20 24
0
5
10
15
20
25
Time (h)
E
n
e
r
g
y
c
o
n
s
u
m
p
t
i
o
n
(
k
W
)
Fig.12 Energy consumption in a hot day
The results in this section show that the proposed control
strategy renders the HVAC system to control the building
environment with high energy efficiency in both cold and hot
weather conditions.
C. Comparison with a CAV system and a non-optimized VAV
system
In this section, the control performance of the proposed
HVAC system is compared with two other systems: a CAV
system and a non-optimized VAV system. The CAV system
changes its supply air temperature to control the room
temperature and air quality with constant air volume flow rate
[16]. The non-optimized VAV system changes its supply air
volume flow rate to control the building environment with a
constant supply air temperature. The CAV system and the
non-optimized VAV system are tested individually on the
same building model with the same outdoor and indoor
conditions as depicted in Figs. 4-5. The simulation results
demonstrate that both the CAV system and the non-optimized
VAV system are capable of maintaining the indoor
temperature and the CO
2
concentration within the proper
range, which are illustrated in Fig. 13 and Fig. 14,
respectively.
0 4 8 12 16 20 24
10
12
14
16
18
20
22
24
26
Time (h)
I
n
d
o
o
r
T
e
m
p
e
r
a
t
u
r
e
(
d
e
g
r
e
e
C
e
l
s
i
u
s
)
CAV system
Non-optimized VAV system
Optimized HVAC System
Fig.13 Indoor temperature variation in three different systems
0 4 8 12 16 20 24
200
300
400
500
600
700
800
900
1000
Time (h)
C
O
2
C
o
n
c
e
n
t
r
a
t
i
o
n
(
p
p
m
)
CAV system
Non-optimized VAV system
Optimized HVAC System
Fig.14 Indoor CO
2
concentration variation in three control designs
Fig. 15 shows the comparison of the energy consumption of
three different control designs in the same building
environment. It can be seen from Fig. 13 that the non-
optimized VAV system achieves the higher energy efficiency
than the CAV system. And the proposed MOPSO optimizer
for HVAC system features the highest energy efficiency
among all three systems, especially in high-load time. The
intelligent optimization based approach has shown advantages
in saving energy during building operations.
0 4 8 12 16 20 24
0
10
20
30
40
50
60
70
Time (h)
E
n
e
r
g
y
c
o
n
s
u
m
p
t
i
o
n
(
k
W
)
CAV system
Non-optimized VAV system
Optimized HVAC System
Fig.15 Energy consumption of three control designs
V. CONCLUSION
In this paper, the function of HVAC system in building
energy and comfort management has been discussed. The
HVAC system is applied to provide a relatively constant and
comfortable temperature in buildings and provide fresh and
filtered air with a comfortable humidity level. In this paper,
the building model and the HVAC system model are
developed. A MOPSO optimizer is embedded in the controller
to control the buildings indoor environment with energy
efficiency. The control strategy utilized swarm intelligence
technique to determine the amount of energy dispatched to
each HVAC equipment. The building performance in both hot
and cold weather is simulated. The simulation results show the
effectiveness of the MOPSO optimizer in maintaining indoor
comfort and saving energy. Compared to the CAV system and
the non-optimized VAV system, the proposed optimizer has
proven to be capable of saving more energy in building
operations under the same environmental condition.
VI. REFERENCES
[1] W.C. Turner, Energy Management Handbook, 5th Ed., Fairmont Press,
2004.
[2] K. F. Fong, V. I. Hanby, and T. T. Chow, "HVAC system optimization
for energy management by evolutionary programming," Energy and
Buildings, Vol. 38, pp. 220-231, 2008.
[3] E.H. Mathews, C. P. Botha, D. C. Arndt, and A. Malan, HVAC control
strategies to enhance comfort and minimise energy usage, Energy and
Buildings, vol. 33, pp. 853-863, 2001.
[4] A.I. Dounis, and C. Caraiscos, Advanced control systems engineering
for energy and comfort management in a building environmentA
review, Renewable and Sustainable Energy Reviews, vol.13, pp.1246-
1261, 2009.
[5] M. Stipe, Demand-Controlled Ventilation: A Design Guide,
Northwest Energy Efficiency Alliance, June 2003.
[6] L. Glicksmana, and S. Tauba, Thermal and behavioral modeling of
occupant-controlled heating, ventilating and air conditioning systems,
Energy and Buildings, vol. 25, pp. 243-249, 1997.
[7] R.D. Taylor , C. O. Pedersen, and L. K. Lawrie, "Simultaneous
simulation of buildings and mechanical systems in heat balance based
energy analysis programs," in Proceedings of the 3rd International
Conference on System Simulation in Buildings, Liege, Belgium, Dec.
1990.
[8] F. Merritt, J. Ricketts, Building Design and Construction Handbook, 6th
Ed., McGraw-Hill, 2000, pp.13.1-13.95.
[9] Y. Yao, Z. Lian, Z. Hou, and X. Zhou, Optimal operation of a large
cooling system based on an empirical model, Applied Thermal
Engineering, vol. 24, pp. 2303-2321, Nov. 2004.
[10] X. Lu, T. Lu and M. Viljanen, Estimation of space air change rates and
CO
2
generation rates for mechanically-ventilated buildings, Advances
in Computer Science and Engineering, pp.237-260, Mar. 2011.
[11] M. Mossolly, K. Ghali, and N. Ghaddar, Optimal control strategy for a
multi-zone air conditioning system using a genetic algorithm, Energy,
vol. 34, pp.58-66, 2009.
[12] F.W. Yu, and K.T. Chan, Part load performance of air-cooled
centrifugal chillers with variable speed condenser fan control, Building
and Environment, vol. 42, pp. 3816-3829, Nov. 2007.
[13] R. Poli, J. Kennedy, and T. Blackwell, Particle swarm optimisation: an
overview. Swarm Intelligence Journal, vol.1, pp.33-57, 2007.
[14] M. Reyes-Sierra, and C.A.C. Coello, Multi-objective particle swarm
optimizers: A survey of the state-of-the-art. International Journal of
Computational Intelligence Research, vol. 2, pp.287-308, 2006.
[15] L. Wang, and C. Singh, Reserve-constrained multi area
environmental/economic dispatch based on particle swarm optimization
with local search, Engineering Applications of Artificial Intelligence,
vol. 22, pp. 298-307, 2009.
[16] Y. Yao, Z. Lian, W. Liu, Z. Hou, and M. Wu, Evaluation program for
the energy-saving of variable-air-volume systems, Energy and
Buildings, vol. 39, pp. 558-568, May 2007.