Академический Документы
Профессиональный Документы
Культура Документы
Abstract — In this paper a fuzzy adaptive comfort temperature preferences through a Graphical User Interface (GUI) [4]. For
(FACT) model has been proposed for the intelligent control of indoor temperature control, users define their own comfort
smart buildings. A multi-agent control system is applied for the zone by setting the parameters of minimum temperature and
energy management and building operation. Particle Swarm maximum temperature [Tmin, Tmax]. Through the use of Particle
Optimization (PSO) is applied to optimize the set points based on Swarm Optimization (PSO) the control system is able to
the comfort zone. Integrating a grey predictor to predict outdoor autonomously find the optimal set point that lies in the user-
temperature with the FACT model shows great promise in defined comfort zone. According to previous works regarding
systematically determining the customer temperature comfort the variation of indoor temperature [5]-[8], the indoor comfort
zone for smart buildings. With the application of the FACT
temperature is determined by a combination of the control
model and other intelligent technologies, the multi-agent control
system has successfully provided a high-level of temperature
system and the variations in outdoor temperature. Other factors
comfort with low power consumption to customers in smart including the business culture and metabolic rates of
building environments. Case studies and corresponding individuals also have an impact when determining the most
simulation results are presented and discussed in this paper. comfortable indoor temperature.
In this paper we introduce a fuzzy adaptive comfort
Keywords- Fuzzy logic; grey prediction; optimization; smart temperature (FACT) model integrated with a grey predictor
building; energy efficiency
that enables the prediction of outdoor temperatures. When
applying the FACT model one important issue is the prediction
I. INTRODUCTION of meteorological parameters. This is an interesting and
At the cutting-edge of modern building management challenging issue since the weather data is typically
technologies, smart buildings are utilizing computers and complicated and disordered. Short-term prediction is used to
intelligent technologies to manage energy consumption and the make use of the most recent data to forecast the future. This is
indoor environment autonomously. Smart buildings have called local prediction. A Grey predictor is employed in this
shown several advantages such as providing high-level comfort model due to its good performance in local predictions despite
to customers, minimizing power consumption, and reducing requiring a limited amount of past data to make these
impacts on the environment [1, 2]. In order to make the smart predictions. As compared to methods that use only customer-
building environmentally friendly, renewable energy sources defined comfort zones, the FACT model improves the
have been employed as the primary power supply. Through the determination of the temperature limits systematically without
use of an intelligent control system a high-level of comfort and the intervention of users, achieves energy savings, and
power-efficiency should be easily achieved in a smart building enhances the intelligence of the building management system.
environment. Managers are also provided an upper level of overreaching
controls in order to simplify the control process.
The indoor temperature, the indoor air quality, and the
indoor illumination are the most important comfort-impact The remainder of the paper is organized as follows. Section
factors in smart buildings [3]. Our previous work [4] has II briefly reviews the multi-agent control system for smart
proposed a multi-agent control system with heuristic buildings. Section III represents the fuzzy adaptive comfort
optimization for the comfort and energy management of smart temperature model including a detailed description of the grey
buildings. Occupants are provided with a high level of predictor. In section IV case studies and simulation results are
flexibility in order to participate in the control of system presented. The conclusion and future work are given in Section
behaviors. This means that they can set up their own VI.
729
4) Adjust the velocity and location of each particle based Considering the different functionalities of buildings and the
on the update rules (2)(3)(4); variety of equipment they contain, customers should be given
5) Repeat 2) to 4) until the desired number of iteration is the authority to manage the controllable loads based on their
achieved or the optimal solution is attained. own preferences. The load agent, when it is activated, is used to
decide the priority and the amount of load shed based on the
The Set-Point Particle Swarm Optimizer (SP-PSO) is users preferences. More energy may be dispatched to the
embedded in the central coordinator-agent to tune the set points comfort-related elements to maintain high-level comfort after
according to the outdoor information and the customer comfort the load agent is working.
zone. The optimization goal is to maximize the objective
function defined in (1). In our previous work people are III. FUZZY ADAPTIVE COMFORT TEMPERATURE (FACT)
provided with the flexibility to set their own comfort zones. In MODEL
this paper we provide a new comfort temperature model which
can automatically find the temperature comfort zone while A. Adaptive Comfort Temperature Model
customers still have flexibility to fine tune the control system Beyond a simple logical algorithm, adaptive control is an
by adjusting their degree of acceptability. artificial intelligence technology which can be used for
empirical and judgmental information [6, 7]. The adaptive
D. Local controller-agents models have recently been applied to define the indoor comfort
Local controller-agents are implemented in three local temperature as a linear regression which is related to outdoor
subsystems to control temperature comfort, visual comfort, and information. This can be expressed by an equation of the form:
air quality. Fig. 2 shows the structure of the local subsystems.
Tc = C + KTo (5)
730
~ ~
In [7], Peeters uses a new parameter named To,ref to define where α~ , β and T are fuzzy numbers and
the comfort temperature. The modified set of equations is
shown as: α~1 = (lα 1 , mα 1 , rα 1 ) (18)
~
β1 = (lβ 1 , mβ 1 , rβ 1 ) (19)
T + 0.8Tk −1 + 0.4Tk −2 + 0.2Tk −3
To,ref = k (10)
2 .4 α~2 = (lα 2 , mα 2 , rα 2 ) (20)
~
Tc = 16.63 + 0.36To,ref , To, ref ≥ 12.5 °C (free-running) (11) β 2 = (lβ 2 , mβ 2 , rβ 2 ) (21)
~
Tc = 20.4 + 0.06To, ref , To, ref <12.5 °C (heating on) (12) To, ref = (lo , mo , ro ) (22)
~
Tc = (lc , mc , rc ) (23)
where To,ref is the reference outdoor temperature (°C), Tk is the
arithmetic average of today’s maximum and minimum outdoor Equations (16) and (17) above are combined with Equation
temperature (°C), Tk-1 is the arithmetic average of yesterday’s (10) to form the FACT model to the multi-agent control system
maximum and minimum outdoor temperature (°C), Tk-2 is the of the smart building.
arithmetic average of the day before yesterday’s maximum and
minimum outdoor temperature (°C), and Tk-3 is the arithmetic C. Fuzzy Arithmetic Operations
average of maximum and minimum outdoor temperature of ~
three days ago (°C). Assume two generalized triangular fuzzy numbers X1 and
~ ~ ~
For air-conditioned buildings the correlation between X 2 where X 1 = (a1 , b1 , c1 ) and X 2 = (a2 , b2 , c2 ) . Using the
comfort temperature and outdoor temperature is given by [13]: fuzzy arithmetic operators from [14] and [15] the following
equations can be given:
Tc = 18.6 + 0.16To (13) 1) Fuzzy Numbers Addition ⊕
~ ~
As the primary goal of the smart building is to provide a X1 ⊕ X 2 = (a1 , b1 , c1 ) ⊕ (a2 , b2 , c2 )
high comfort level to customers, a naturally ventilated building = (a1 + a2 , b1 + b2 , c1 + c2 ) (24)
cannot be the best choice for a smart building. We propose a
mixed building model which provides heat to the building in
the winter and cooling in the summer. The comfort temperature 2) Fuzzy Numbers Multiplication ⊗
correlated to the outdoor temperature for smart buildings has ~ ~
been proposed using the following equations: X 1 ⊗ X 2 = (a1 , b1 , c1 ) ⊗ (a2 , b2 , c2 )
= (a1 × a2 , b1 × b2 , c1 × c2 ) (25)
Tc = 18.6 + 0.16To, ref , To, ref ≥ 18 °C (air-conditioner on) (14)
We consider three fuzzy
~ ~
Tc = 20.4 + 0.06To, ref , To, ref <18 °C (heating on) (15)
numbers α = (lα , mα , rα ) , β = (lβ , mβ , rβ )
~
and To, ref = (lo , mo , ro ) , where l, m, and r are real numbers.
B. Fuzzy Adaptive Comfort Temperature Model Utilizing the fuzzy arithmetic operators ⊕ and ⊗ , the
Changing outdoor weather conditions, internal heat gains, following fuzzy adaptive comfort temperature model is
ventilation, and the preferences of users all influence the indoor developed:
environment of a smart building leading to different comfort
profiles for different persons. The acceptable comfort ~
temperature regions are formulated by an upper and a lower Tc = (lc , mc , rc )
limit which defines the comfort temperature band. In this paper ~ ~
= α~ ⊕ β ⊗ T o, ref (26)
a new method to determine the upper (Tmax) and the lower
(Tmin) bounds of the temperature band based on Equations (14) = (lα + l β lo , mα + mβ mo , rα + rβ ro )
and (15) is proposed. This method can be implemented using
the following fuzzy equations: The relationship of these fuzzy numbers is lc=lα+lβlo,
mc=mα+mβmo and rc=rα+rβro.
~ ~ ~
Tc = α~1 + β1 × To,ref , To, ref ≥ 18 °C (air-conditioner on) (16) The user participates in the control process by defining the
“α” that is the acceptability degree of comfort temperature. The
~
~ ~ ~ α level determines the crisp set, that is, the lower Tcα, min and the
Tc = α~2 + β 2 × To ,ref , To, ref <18 °C (heating on) (17) ~
upper Tcα, max comfort temperatures in Equations (27) and (28).
731
~
Tcα, min = mc − α (mc − lc ) (27) dx (1)
= x (1) (i ) + x (1) (i − 1) = x ( 0) (i ) (32)
dt
~
Tcα, max = mc + α ( rc − mc ) (28)
And by approximating x(1) with ½[x(1)(i)+x(1)(i-1)], Equation
(30) can be rewritten as,
The acceptability degree, α, will substantially change the
acceptance comfort zone. It is obvious that the bandwidth of
the comfort zone will extend when the acceptable degree, α , is x ( 0) (i ) + a ⋅ y (1) (i ) = b (33)
changed.
The parameters a and b can be obtained by using the Least
D. Grey Predictor Square Error Method in Equation (34)
Generally, a grey system means that the information
regarding the system is incomplete or uncertain. By using a
grey model that requires little previous data to perform a real- ⎡a ⎤ T −1 T
time forecast, the grey predictor has been successfully ⎢ ⎥ = ( B B) B Z n (34)
b
⎣ ⎦
employed in many areas. A first order linear dynamic grey
model, GM(1,1), is applied to make short-term predictions of
where
the future average outdoor temperature on a daily basis [16].
Local prediction forecasts of the future based on the most ⎡− y (1) (2) 1⎤
recent data set are a type of time series. Suppose we have the ⎢ (1) ⎥
− y (3) 1⎥
previous values of x from the time k-m to k-l, that is x(k-1), x(k- B=⎢ (35)
2), …, x(k-m). The next time interval values, x(k), can be
⎢ # #⎥
⎢ (1) ⎥
predicted by a grey model. The algorithm of this first order ⎣⎢ − y ( n) 1⎦⎥
grey model GM(1,1) is as follows [17].
and
Assume the original raw data series x(0) with n samples is
defined as x(0)=[ x(0)(1), x(0)(1), x(0)(2), .., x(0)(n)]. All values in ⎡ x ( 0) (2) ⎤
this data sequence are required to be positive. When negative ⎢ ( 0) ⎥
x (3) ⎥
values appear in this sequence the absolute value of the Zn = ⎢ (36)
maximum negative data is added to all the data in the sequence ⎢ # ⎥
to make the data sequence positive. ⎢ ( 0) ⎥
⎢⎣ x (n)⎥⎦
To weaken the randomness of the original raw data, the
original raw data x(0) is pre-processed and transformed into a The solution of equation (30) is an exponential function
new sequence x(1) using the accumulated generating operations with the initial condition x(1)(0)= x(0)(1). The predicted value
(AGO). can be obtained as
i b b
xˆ (1) (n + 1) = ( x ( 0) (1) − ) ⋅ e − an + (37)
x (1) (i ) = AGO( x ( 0) (i )) = ∑
k =1
x ( 0) (k ), i = 1,..., n (29) a a
732
prediction and the temperature for Day 4 through Day 15 is comfort temperature zones. The temperature is taken as the
forecasted. Table I and Table II show the comparison between most important comfort impact factor in the control of the
the predicted temperature and the realistic temperature. smart building. The user-defined weighting factors are set as
μ1=1 and μ2=μ3=0. According to Equations (16) and (17) all of
TABLE I. THE PREDICTED TEMPERATURE AND THE REALISTIC
the fuzzy members should be defined. Based on Equations (14)
TEMPERATURE IN WINTER and (15), we set the fuzzy member as:
Winter data set α~ = (l , m , r ) = (16.6,18.6,20.6)
1 α1 α1 α 1 (38)
(Jan.4-Jan.15) /°C ~
Date Real Predicted
β1 = (l β 1 , mβ 1 , rβ 1 ) = (0.13,0.16,0.19) (39)
temperature temperature
α~2 = (lα 2 , mα 2 , rα 2 ) = (18.4,20.4,22.4) (40)
Day 1 5.4
~
Day 2 -6.3 β 2 = (l β 2 , mβ 2 , rβ 2 ) = (0.03,0.06,0.09) (41)
Day 3 -2.9
~
Day 4 -0.9 1.4 To, ref = (lo , mo , ro ) = (To, ref − 2, To, ref , To, ref + 2)(°C ) (42)
Day 5 -5.6 -8.1
For T0,ref, we use Equation (10) to calculate the reference
Day 6 -3.4 -0.5 temperature using the information from the grey predictor.
Day 7 -6.7 -8.9 Considering that the reference temperature is related to the
Day 8 -8.5 -9.9 temperature of the past three days, only 12 days’ (Day 4 – Day
Day 9 -8.9 -9.2
15) reference temperatures can be obtained out of the 15 days
of outdoor temperature information. The customer acceptability
Day 10 -4.4 1.8 degree is defined as α=0.8. For comparison, the user defined
Day 11 -4.6 -4.7 temperature comfort zones in both winter and summer periods
Day 12 -5.1 -5.5 are set to [Tmin, Tmax] =[20,24 ](°C) .
Day 13 -7.1 -8.6 In the first case the predicted daily average temperature in
Day 14 -6.3 -5.4 winter is used. The FACT model is applied to calculate the
Day 15 -2.8 1.9
comfort zone [Tmin, Tmax] and the set points are tuned by the P-
PSO algorithm. Fig. 3 and Fig. 4 illustrate the comfort zone and
the set point both with and without the FACT model. The
TABLE II. THE PREDICTED TEMPERATURE AND THE REALISTIC comfort zone after applying the FACT model is lower than
TEMPERATURE IN SUMMER
before. This demonstrates that the PSO has successfully found
Summer data set the set points.
(Aug.4-Aug.15) /°C Fig. 5 shows the energy saving advantage of the FACT
Date Real Predicted model. The power consumption for the local temperature
temperature temperature
controller-agent is obviously reduced under the effect of FACT
Day 1 24.2 model. The FACT model not only provides a more rational
Day 2 23.5 comfort zone systematically to enhance customers comfort
level, but also reduces the power consumption of the building.
Day 3 25.6
These advantages achieve the requirements of the control
Day 4 25.7 25.8 system for smart building.
Day 5 25.8 25.9
The second case uses the summer set data to control the
Day 6 23.4 21.2 building and to observe the effect of the FACT model during
Day 7 23.1 22.8 hotter weather. Fig. 6 and Fig. 7 are the comfort zones and set
Day 8 24.5 25.9 points with and without FACT model in this case. Fig. 8 is the
comparison of the power consumption under these situations.
Day 9 25.7 26.9
Day 10 26.8 27.9
Day 11 25.2 23.6
Day 12 24.6 24.0
Day 13 28.1 32.0
Day 14 26.3 24.6
Day 15 26.7 27.1
Based on the two data sets, two case studies are conducted
in this section by utilizing the FACT model to determine the
733
30 35
25
30
20
Temperature(C)
Temperature(C)
15 Upper Comfort Temperature 25
Low er Comfort Temperature
10 Set Points
Outdoor Temperature 20
5
Upper Comfort Temperature
0 Low er Comfort Temperature
15
Set Points
-5 Outdoor Temperature
-10 10
4 5 6 7 8 9 10 11 12 13 14 15
4 5 6 7 8 9 10 11 12 13 14 15
Time (Day)
Time (Day)
Figure 3. Comfort zone and set points with FACT model in winter Figure 6. Comfort zone and set points with FACT model in summer
35
30
25 30
20
Temperature(C)
25
Temperature(C)
30
With FACT model
8
Without FACT model
25
Power Consumption (KW)
Power Consumption (KW)
6
20
15 4
5
0
4 5 6 7 8 9 10 11 12 13 14 15
0 Time (Day)
4 5 6 7 8 9 10 11 12 13 14 15
Time (Day) Figure 8. Power consumption for local temperature controller-agent with and
without FACT model in summer
Figure 5. Power consumption for local temperature controller-agent with and
without FACT model in winter According to our current results, the FACT model can
provide high-level of comfort temperature in any outdoor
situation and significantly save energy in order to maximize the
customer benefit. It can be concluded that the FACT model is a
suitable method for smart building as it enhances the
intelligence and optimizes the entire building system.
734
V. CONCLUSION
This study applied the FACT model with grey predictor to
the control system of a smart building. From the simulation
results, the appropriateness of this model is demonstrated and
discussed. The customers systematically attain a more
reasonable comfort temperature with less power consumption
when using the FACT model. In future studies similar adaptive
comfort models for visual comfort and air quality will be
developed in order to enhance the overall comfort level of the
smart building while consuming minimum energy.
REFERENCES
[1] http://www.smart-buildings.com
[2] J. Hang, J. Tian and H. Lin, “Application of artificial neural network in
intelligent building,” Proc. IEEE Int. Joint Conf. on Machine Learning
and Cybernetics, vol.7, pp. 4215 – 4220, Aug 2007
[3] A.I. Dounis and C. Caraiscos, “Advanced control systems engineering
for energy and comfort management in a building environment—A
review,” Renewable and Sustainable Energy Reviews, vol.13, issue 6-7,
pp. 1246–1261, August – September 2009.
[4] Z. Wang, R. Yang, L. Wang and A.I. Dounis, “Customer-centered
Control System for Intelligent and Green Building with Heuristic
Optimization,” IEEE Power System Conference and Exposition,
Phoenix, AZ, March 2011
[5] K. W. H. Mui and W. T. D. Chan, “Adaptive comfort temperature model
of air-conditioned building in Hong Kong”, Building and Environment,
vol.38, pp 837 – 852, June 2003
[6] K. J. McCatney and J. F. Nicol, “Developing an adaptive algorithm for
Europe,” Energy and Buildings, vol.34, pp623-635, 2002
[7] L. Peeters, R. de Dear, J. Hensen, and W. D’haeseleer, “Thermal
comfort in residential buildings: Comfort values and scales for building
energy simulation,” Applied Energy, vol.86, pp. 772-780,May 2009
[8] J. A. Orosa, “A new modeling methodology to control HVAC systems,”
Expert Systems with Applications,vol.38, pp.4505-4513, April 2011
[9] J. Kennedy and R. Eberhart, “Particle swarm optimization,” Proc. IEEE
Int. Joint Conf. on Neural Network, vol.4, pp1942-1948, 1995.
[10] F. Lin, L. Teng, J. Lin and S. Chen, “Recurrent functional-link-based
fuzzy-neural-network-controlled induction generator system using
improved particle swarm optimization,” IEEE Transactions on Industrial
Electronics, vol.56, pp1557-1577, May 2009
[11] M. R. AlRashidi and M. E. El-Hawary, “A survey of particle swarm
optimization applications in electric power system,” IEEE Transaction
on Evolutionary Computation, vol.13, pp913-918, June 2008
[12] B. Zhao, C. X. Guo and Y. J. Cao, “A multiagent-based particle swarm
optimization approach for optimal reactive power dispatch,” IEEE
Transaction on Power Systems, vol.20, pp 1070-1078, May 2005
[13] M. Humphreys, F. Nicol, S. Roaf and O. Sykes, Standards for Thermal
Comfort Indoor air temperature standards for the 21st century,
Routledge, UK, 1995
[14] S. H. Chen, “Operations on fuzzy numbers with function principle,”
Tamkang J. Manag. Sci., vol.6, pp13-25, 1985
[15] SJ. Chen and SM. Chen, “Fuzzy risk analysis based on similarity
measures of generalized fuzzy numbers”, IEEE Transactions on Fuzzy
Systems, vol. 11, pp. 45-56, February 2003.
[16] A. I. Dounis, P. Tiropanis, D. Tseles, G. Nikolaou, and G. P. Syrcos, A
Comparison of Grey Model and Fuzzy Predictive Model for Time Series,
International Journal of Engineering and Mathematical Sciences 2:3 pp.
176-181, 2006.
[17] S.-F. Su, C.-B. Lin, Y.-T. Hsu, A high precision global prediction
approach based on local prediction approaches, IEEE SMC-C, Vol.23,
No.4, Nov. 2002, pp. 416-425.
[18] http://www.wunderground.com
735