Building Serv. Eng. Res. Technol. 30,4 (2009) pp.

287–304

Fuzzy multi-criteria evaluation model of HVAC schemes in

optimal combination weighting method
JJ Wang BEng ME, YY Jing BEng and CF Zhang BEng ME
School of Energy and Power Engineering, North China Electric Power University, Baoding, Hebei Province, China

Six heating, ventilation and air conditioning (HVAC) schemes, constant air volume, variable
air volume, fan-coil, induction unit, variable refrigerant volume and distributed HVAC
system are evaluated and compared in multi-criteria. Because of the uncertainty of
qualitative criteria, this paper presents a fuzzy multi-criteria model to select the optimal
HVAC system. The methods of fuzzy set theory, linguistic value, analytic hierarchy process,
entropy weighting and optimal weighting are used to obtain the criteria weights. Fuzzy
technique for order preference by similarity to ideal solution (TOPSIS) method is employed
to get the ranking of HVAC schemes. Finally, the application of the fuzzy multi-criteria
evaluation model demonstrates detailedly the computation and evaluation procedure in the
selection of HVAC schemes.
Practical applications: The fuzzy multi-criteria evaluation model combined with the
optimal weighting method in this paper is believed by the authors to contribute to the
selection and comparison of HVAC systems or schemes in the building. It is proposed that
both qualitative and quantitative criteria existing in reality are used to evaluate HVAC
schemes and the decision-making method combines both subjectivity and objectivity. In
addition, simple and practical computation and evaluation procedure is easily applied by
HVAC designers or decision-makers. Such an approach could become a powerful tool to
make an optimal selection for HVAC schemes.

1 Introduction quality and performance. Professionals have

been well aware that the performance of
Nowadays, building energy consumption has HVAC systems in the whole building lifecycle
reached approximately 40% of the total is significant to building sustainability. Many
world annual energy consumption. Energy ways have been developed to improve the
consumption of heating, ventilation and air lifecycle performance of HVAC systems, such
conditioning (HVAC) accounts for about as setting up high expectation and design
50% of building energy consumption on an criteria from the very beginning, selecting
average.1 HVAC systems are required to be energy efficient materials and equipment and
more energy efficient while adhering to an implementing optimal control strategies.
ever-increasing demand for better indoor air Moreover, economical considerations and
environmental issues also need to be taken
Address for correspondence: Jiangjiang Wang, No. 62 mailbox,
into account.2
North China Electric Power University, North Yonghua HVAC designers usually focus only on one
Street, Baoding, Hebei Province, 071003, China. or two design aspects, such as load calcula-
Tel: þ86-312-7522443; Fax: þ86-312-7522440.
E-mail: [email protected] tions, energy consumption estimation, duct
Figure 8 appears in color online: http://bse.sagepub.com design, pipe design, etc.2–4 However, HVAC
ß The Chartered Institution of Building Services Engineers 2009 10.1177/0143624409338502

Downloaded from bse.sagepub.com at UCSF LIBRARY & CKM on March 15, 2015
288 Fuzzy multi-criteria evaluation model of HVAC schemes
design is a complex task including various management14–16 and environment.17–19
interactive factors which requires experts The applications in building industry have
from different disciplines. Although a pleth- risen gradually, for instance, the selection of
ora of advanced HVAC systems and equip- combined cooling, heating and power
ment have been adopt, developers are (CCHP) systems20–22 and cool storage sys-
confronted with the quandary of choosing tems23,24 for buildings, the optimal designs of
the apposite components or products to suit building,25,26 and the assessment of building
the needs and to accomplish the unique performance.5,27–29 Currently, the multi-
configuration of a particular HVAC project. criteria evaluation of HVAC systems is still
The problems inherent in justifying the at the starting stage. There are few literatures
options of HVAC systems and components, about the multi-criteria evaluation of HVAC,
or the HVAC schemes, can be attributed to which case studies or comparison of HVAC
two factors.5 schemes exist only in internal reports. Some
Firstly, there is a dearth of systematic and studies have attempted to establish the assess-
rigorous methods in existence for selecting ment criteria system of HVAC systems.30–34
new HVAC technologies. Currently many Jiang et al.30 proposed energy conversion
approaches focus only on cost performance coefficient (ECC) indicators to evaluate the
that was easily quantifiable,6–9 but ignore efficiency of HVAC systems based on the
other benefits such as improved thermal exergy analysis method. Zhang et al.31 set up
comfort, environmental sustainability, and the comprehensive assessment index system of
building flexibility. Consequently, the classi- HVAC facility management in analytic hierar-
cal cost evaluation approach probably led to chy process (AHP) method by means of a
a biased decision. Accordingly, methodology questionnaire from energy consumption,
that can combine qualitative factors (i.e. maintenance, environmental protection, ther-
human judgments) with the quantitative mal comfort, indoor air quality and economy.
approach is helpful in selecting the HVAC Zhang and Zhang32 advised to use ‘air-
scheme. conditioning quality’ as judge standard for
Secondly, many performance criteria that design level and set up six evaluation grades
are not easily expressed or quantified fail to and eight evaluation rules. Long33 syntheti-
be captured in many evaluation approaches. cally summarised the evaluation criteria of
Sometimes both qualitative and quantitative building including HVAC system, especially
criteria exist in reality, for example thermal the building energy saving and efficiency man-
comfort. Moreover, due to the availability agement. He and Feng applied a comprehen-
and uncertainty of information in our deci- sive evaluation method to select the optimal
sion process as well as the vagueness of air-conditioning project from four scenarios.34
human feeling and recognition, it is difficult Based on the current research deficiencies,
to make an exact evaluation and express the this paper proposes a fuzzy multi-criteria
feeling and recognition of decision makers or evaluation model to select the optimal
designers. HVAC scheme. In the evaluation process,
Multi-criteria analysis is a technique the relative importance of criteria that is
devoted to simplifying a complex selection defined to weight is critical to the final
or decision problem and to solving it. Multi- result.10,11 Usually, the priority is given to
criteria evaluations have been employed to one of criteria, and the weights of other
many fields such as energy planning,10–13 criteria are all equal, which is subjective and

Downloaded from bse.sagepub.com at UCSF LIBRARY & CKM on March 15, 2015
 JJ Wang et al. 289
cannot reflect the external data. To solve this The selection problem involves m HVAC
point, an improved combination weighting schemes evaluated on n criteria. Thus the
method is applied to assist the fuzzy evalua- judgment matrix can be modelled:
tion model in this paper. This paper is
organised as follows. The fuzzy multi-criteria G ¼ ½Gij mn
model including weighting methods is C1 C2    Cn
described in Section 2. The presented model 2 3
S1 G11 G12    G1n
is applied to select the optimal scheme from
6G 7 ð1Þ
six HVAC alternatives in Section 3. The ¼ S2 6 21 G22    G2n 7
6 .
sensitivity analysis and discussion are pre- .. 6 . .. .. .. 7
7
sented in Section 4. Finally, some concluding . 4 . . . . 5
comments are presented in the last section. Sm Gm1 Gm2    Gmn
and the weight vectors are

W ¼ ½w1 , w2 , . . . , wn T ð2Þ
2 Fuzzy multi-criteria evaluation model
where S1 , S2 , . . . , Sm are the HVAC schemes,
2.1 Problem formulation C1 , C2 , . . . , Cn are the criteria, Gij
The purpose of this paper is to select the (i ¼ 1, 2, . . . , m, j ¼ 1, 2, . . . , n) is the perfor-
optimal HVAC scheme from alternatives and mance of alternative Si on criteria Cj and Wj
to establish a hierarchical structure for is the related weight of Cj .
tacking the evaluation problem of HVAC
alternatives. Multi-criteria evaluation is a 2.2 Hierarchical structure of multi-criteria
well-known branch of decision making. It is evaluation
a branch of a general class of operations AHP36 is widely used for tackling multi-
research models which deal with decision criteria decision-making problems in real
problems under the presence of a number of situations. Herein, AHP is used to solve the
decision criteria. A vast number of multi- selection problem from HVAC schemes. By
criteria models and approaches are available means of a systematic hierarchy structure,
in the literature. However, the general multi- complex estimation criteria can be clearly and
criteria evaluation process can be shown in distinctly presented. The hierarchical struc-
Figure 1. Firstly, the applicable alternatives ture adopted in this study to deal with the
of HVAC systems are developed. Secondly, problem of HVAC systems evaluation for an
the criteria of efficiency to be used analysing office building is shown in Figure 2. The key
the alternatives are established. Based on the dimensions of the criteria for evaluation and
judgments of experts or measure data, the cri- selection of HVAC alternatives were derived
teria weights are determined. Then, the pref- through reviewed literatures,30–34,37–42 com-
erence ranking of schemes are determined by prehensive investigation and consultation
multi-criteria decision making (MCDM) with several experts from Building
methods,35 such as technique for order pref- Environment and Service Engineering in
erence by similarity to ideal solution North China Electric Power University.
(TOPSIS) and AHP. Finally, the optimal These individuals were asked to rate the
HVAC scheme is selected after sensitivity accuracy, adequacy and relevance of the
analysis. criteria and dimensions and to verify their

Downloaded from bse.sagepub.com at UCSF LIBRARY & CKM on March 15, 2015
290 Fuzzy multi-criteria evaluation model of HVAC schemes

Establish decision goals

Identify the alternatives

Identify the criteria

Assign the criteria scores

Normalisation

Weighting
methods Determine criteria weights

MCDM Combine weights & priority scores by

methods aggregation

Preference ranking of the alternatives

Sensitivity analysis

Final recommendation

Figure 1 Multi-criteria evaluation process

‘content validity’ in terms of HVAC systems’ 16 evaluation criteria for the hierarchical
assessment. The AHP structure in Figure 2 structure are employed to evaluate the inte-
involves decomposing a complex decision into grated performance of HVAC schemes.
a level with goal (objective) at the top of the
hierarchy, rules and criteria at levels 2 and 3
2.3 Normalisation
of the hierarchy, and HVAC schemes at the
In Figure 2, there are qualitative criteria
bottom of the hierarchy. The schemes include
such as control property, visual impact,
six styles of HVAC systems:
and convenience, and quantitative criteria
such as investment cost and service life.
S1: Constant air volume HVAC system These qualitative criteria are not directly
S2: Variable air volume (VAV) HVAC used to evaluate the system, which need to
system be assigned in fuzzy number. Furthermore,
S3: Fan-coil HVAC system the evaluation criteria are of their own
S4: Induction unit HVAC system characteristics or extensity, and each data of
S5: Distributed HVAC system: air- criteria has its own dimension and distribu-
conditioner tion, it is difficult to directly compare or
S6: Variable refrigerant volume (VRV) operate. As a result, the original data of
HVAC system evaluation criteria should be dimensionless
and unit-free by normalisation method to
There are five rules and dimensions directly compare. For the sake of clarity,
including technology, economy, environ- some notions of fuzzy set theory are intro-
ment, health and management. From these, duced to assign the scores of criteria, Gij .

Downloaded from bse.sagepub.com at UCSF LIBRARY & CKM on March 15, 2015
 JJ Wang et al. 291

Level 1: Ultimate goal Level 2: Rules Level 3: Criteria Level 4: Schemes

Constant air
C11: Maturity volume
HVAC system
R1: Technology C12: Temperature control
C13: Humidity control
VAV
C21: Investment cost HVAC system
R2: Technology C22: Operation maintenance cost
C23: Service life
C31: Noise pollution Fan-coil
Selecting a best HVAC system
HVAC system R3: Environment C32: Visual impact
from alternatives C33: Space use
C41: Cleanliness Induction unit
HVAC system
R4: Health C42: Fresh air
C43: Thermal comfort
Distributed HVAC
C41: Convenience system: air-
C42: Simplicity conditioner
R5: Management
C43: Concentration
VRV
C44: Flexibility HVAC system

Figure 2 Hierarchy structure of multi-criteria evaluation of HVAC schemes

2.3.1 Fuzzy number where l and r stand for the lower and upper
Fuzzy set theory proposed by Zadeh in bounds of the fuzzy number A, respectively,
196543 is suitable for dealing with the uncer- and m for the middle value (see Figure 3). The
tainty and imprecision associated with infor- TFN can be denoted by A ¼ ðl, m, rÞ and the
mation concerning various parameters. following is the operational laws of two
Human judgment is generally characterised TFNs, A ¼ ðla , ma , ra Þ and B ¼ ðlb , mb , rb Þ, as
by vague language, like ‘equally’, ‘moder- shown:
ately’, ‘strongly’, ‘very strongly’ and (1) Addition
‘extremely’. It provides numerous methods
to represent the qualitative judgment of the
decision making as quantitative data. Fuzzy A þ B ¼ ðla þ lb , ma þ mb , ra þ rb Þ ð4Þ
numbers are a fuzzy subset of real numbers,
representing the expansion of the idea of the
(2) Multiplication
confidence interval. A fuzzy number A on R
is a triangular fuzzy number (TFN) if its
membership function fA ðxÞ: R [0,1] is equal to A  B  ðla lb , ma mb , ra rb Þ ð5Þ
8
>
> 1, x¼m
< (3) Subtraction
ðx  lÞ=ðm  lÞ, l  x5m
fA ðxÞ ¼ ð3Þ
>
> ðr  xÞ=ðr  mÞ, m5x  r
:
0, otherwise A  B ¼ ðla  rb , ma  mb , ra  lb Þ ð6Þ

Downloaded from bse.sagepub.com at UCSF LIBRARY & CKM on March 15, 2015
292 Fuzzy multi-criteria evaluation model of HVAC schemes
fA(x ) conversion scales is generally intuitive: while
too few conversion scales reduce analytical
1 discrimination capability, too many conver-
sion scales make the system overly complex
and impractical. In this study, a scale of 1–5 is
used to present five states of attributes shown
in Table 1.
0
l m r x 2.3.4 Fuzzy normalisation
After assigning the fuzzy numbers of crite-
Figure 3 The membership function of triangular fuzzy number A ria, the judgment matrix in Equation (1) is
normalised. If the criteria has a characteristic
(4) Division of ‘higher is better’ such as service life, the
criteria is benefit item and Gij is normalised as
A=B  ðla =rb , ma =mb , ra =lb Þ ð7Þ follows:
!
Gij glij gm
ij grij
2.3.2 Ranking of TFN Rij ¼ þ ¼ rþ , mþ , lþ ^ 1 ð9Þ
Gj gj gj gj
In a fuzzy decision-making environment,
ranking the alternatives under consideration is
essential. For matching the fuzzy algorithm Otherwise, Gij is a cost item that has a
developed in this paper, and solving the characteristic of ‘lower is better’ such as
problem powerfully, the graded mean integra- investment cost, and it is normalised:
tion representation method proposed by Chen !
and Hsieh44 is used to rank the final ratings of Gj gl
j gm
j gr
j
Rij ¼ ¼ , , l ^1 ð10Þ
alternatives. By the graded mean interaction Gij grij gm ij gij
representation method, the graded mean inte-
gration representation RðAÞ of TFN A is: where Gþj ¼ maxfGij
lþ mþ rþ
ji ¼ 1, 2, . . . , mg ¼ ðgj , gj , gj Þ
la þ 4ma þ ra
RðAÞ ¼ ð8Þ l m r
6 G
j ¼ minfGij ji ¼ 1,2, .. . ,mg ¼ ðgj , gj , gj Þ

Suppose RðAÞ and RðBÞ be the graded The normalised judgment matrix is rewrit-
mean integration representations of the TFLs ten to:
A and B, respectively and define that
G0 ¼ ½Rij mn ð11Þ
A4B , RðAÞ4RðBÞ,
All criteria values locate the intervals (0, 1)
A ¼ B , RðAÞ ¼ RðBÞ, after normalisation. The normalised values
can be compared and operated directly. The
A5B , RðAÞ5RðBÞ: greater the term in G0 is, the better the
alternative is.

2.3.3 Linguistic variables 2.4 Determining the criteria weights

In fuzzy set theory, conversion scales are Weight is assigned to criteria to indicate its
applied to transform linguistic terms into relative importance in multi-criteria evalua-
fuzzy numbers. Determining the number of tion. The bigger the weight is, more it

Downloaded from bse.sagepub.com at UCSF LIBRARY & CKM on March 15, 2015
 JJ Wang et al. 293
Table 1 The Linguistic scales for attributes of pairwise comparison when there are n
Linguistic variables Fuzzy number criteria at a given level can be formed as:
Very low (Very poor) (0, 0, 0.3)
2 3
C1 =C1 C1 =C2    C1 =Cn
Low (Poor) (0, 0.3, 0.5)
6 C2 =C1 C2 =C2    C2 =Cn 7
Moderate (Middle) (0.3, 0.5, 0.7) 6 7
High (Good) (0.5, 0.7, 1.0) V ¼ 6 .. .. .. .. 7 ð12Þ
Very high (Excellent) (0.7, 1.0, 1.0) 4 . . . . 5
Cn =C1 Cn =C2  Cn =Cn
contributes to the evaluation result. Different
weights influence directly the ranking order of The geometric mean of each row is
HVAC alternatives. Accordingly, the ratio- computed as:
nality and veracity of criteria weights deter-
mines the reliability of the evaluation results. !1=n
Y
n
Often the weights depend on the decision- Yi ¼ Vij ¼ ðyli , ym r
i , yi Þ ð13Þ
maker’s subjectivity and it may leads to a j¼1
biased result. For the determination of the
rational weights of criteria, the combination and then the fuzzy weights are calculated as:
weighting method is employed in this study. X X X
There are rule hierarchy and criteria hier- Ws ¼ ½Y1 = Yi , Y2 = Yi , . . . , Yn = Yi  T
archy on levels 2 and 3 in Figure 2, respec- ð14Þ
tively. The rule hierarchy is given higher
priority than the corresponding criteria hier- Finally, the integrated weights can be
archy. Therefore, the rule hierarchy and the obtained from the fuzzy weights.
criteria hierarchy should be weighted inde- Pairwise comparison method is subjective.
pendently. After obtaining the criteria weights To different decision-makers from different
at given level, each performance is then viewpoint such as producers, users or gov-
multiplied with its weight and the weighted ernment, the ranking and weights are not
performances are summed to get the score at always same. Commonly, more experts or
a higher level. The procedure is repeated decision-makers from different disciplines are
upward for each hierarchy, until the top of employed to give out the relative important
the hierarchy is reached. degree of criteria.

2.4.1 Subjective weights 2.4.2 Objective weights

Subjective weighting method was the most Entropy weighting method45 can effectively
commonly used to evaluate many kinds of measure the average essence of information
complex systems. AHP is a descriptive deci- quantity, and the larger the entropy value is,
sion analysis methodology that calculates the lower the information express quantity.
ratio-scaled importance of criteria through This paper tries to solve the objective weight
pairwise comparison. of criteria above the alternative level. Thus, it
Experts in HVAC projects are presented a can represent actual conditions of decision-
worksheet and are asked to compare the making, and express the explanation ability
importance of two criteria at a time: ‘Which and reliability of criteria. The first step is to
one of these two criteria is more important, obtain the integrated data from the fuzzy
and how much more important?’ Then the numbers in the normalised judgment matrix,
relative importance is scored. The scales are Equation (11), in the graded mean integration
shown in Table 2 and in Figure 4. The matrix representation method, Equation (8), and

Downloaded from bse.sagepub.com at UCSF LIBRARY & CKM on March 15, 2015
294 Fuzzy multi-criteria evaluation model of HVAC schemes
Table 2 The linguistic scales for relative importance of criteria

Linguistic scales for importance Triangular Triangular Definition

fuzzy scale fuzzy reciprocal
scale

Just equal (1,1,1) (1,1,1)

Criteria i is equally important when
compared to criteria j
Weakly more important (WMI) (1,3/2,2) (1/2,2/3,1) Criteria i seems weakly more important
when compared to criteria j
Strongly more important (SMI) (3/2,2,5/2) (2/5,1/2,2/3) Criteria i is strongly more important when
compared to criteria j
Very strongly more important (VSMI) (2,5/2,3) (1/3,2/5,1/2) Criteria i is very strongly more important
when compared to criteria j
Absolutely more important (AMI) (5/2,3,7/2) (2/7,1/3,2/5) Criteria i is absolutely more important
    when compared to criteria j
nþ1 nþ2 nþ3 2 2 2
Higher n classes , , , , Criteria i is higher n classes than criteria j
2 2 2 nþ3 nþ2 nþ1

fA(x) 2.4.3 Combination weights

WMI SMI VSMI AMI
The above two weighting methods have
1 their own advantages and disadvantages. In
order to obtain rational weights, some com-
bination weighting methods have been devel-
oped and applied to many complex systems.
0 13,21,23,46,47
1 2 3 n n +1 x The two typical combination
weighting methods include multiplication
Figure 4 The Linguistic scales for relative importance of criteria
method and additive method. Due to the
‘multiplier effect’ of multiplication method, it
then the objective weights can be calculated as could lead to a biased result. In this paper, we
follows: adopted the optimal additive method pro-
Define posed by Chen HY47 to get the weights.
Xm The combination weights are expressed to:
Dj ¼ rij ð15Þ
i¼1 Wc ¼ k1 Ws þ k2 Wo ¼ WK ð18Þ
and the entropy measurement of the j-th where K ¼ ½k1 , k2 T is the vector of the com-
criteria is bination coefficients and 0  ki  1,
k1 þ k2 ¼ 1.
X
m
rij rij To obtain the combination coefficients, the
ej ¼ k ln ð16Þ below optimal model can be constructed as:
i¼1
Dj Dj
( )
Xm X m X n
where k ¼ 1=ln m40 and ej  0. maxJðWc Þ ¼ ½ðrij  ri1 j Þwcj 2
The element of the objective weight eigen- i¼1 i1 ¼1 j¼1
vector, Wo , is calculated as: ( " # )
Xn X
n X m X
m
1  ej ¼ ðrij1 ri1 j1 Þðrij2 ri1 j2 Þ wcj1 wcj2
woj ¼ , j ¼ 1, 2, . . . , n ð17Þ j1 ¼1 j2 ¼1 i¼1 i1 ¼1
P
n
( )
ð1  ej Þ X
n X
n
j¼1
¼ Bj1, j2 wcj1 wcj2 ð19Þ
P
n
where 0  woj  1 and woj ¼ 1. j1 ¼1 j2 ¼1
j¼1

Downloaded from bse.sagepub.com at UCSF LIBRARY & CKM on March 15, 2015
 JJ Wang et al. 295
P Pm
where Bj1, j2 ¼ mi¼1 i1 ¼1 ðrij1 ri1 j1 Þðrij2 ri1 j2 Þ, Then the weighted distances of alternative
and JðWc Þ is rewritten to: Ai versus Aþ and A are calculated respec-
tively as:
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
JðWc Þ ¼ WcT BWc ¼ KT WT BWK ð20Þ u
u wcj 2  lþ 2  2
u r  r l
þ2 r mþ
 r m
Xn u 4 j ij j ij
Then, the optimal coefficients vector is u
eþ ¼ u !
solved as: i u  2
j¼1 t
þ rrþ j  rij
r

max J ¼ K T WT BWK ¼ max ð21Þ

ð24Þ
where max is the maximum characteristic
root of WT BW, K is the corresponding and
characteristic vector of max . vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u
Finally, K is normalised and k1 and k2 are u wc 2  2  2
u
n u j r l
 r l
þ2 r m
 r m
obtained. Thus, the combination weights can X j ij j ij
e u 4
be determined. i ¼ u  
j¼1 t 2
þ rr j  rij Þ
r

2.5 Ranking of HVAC schemes

TOPSIS method is employed to determine ð25Þ
the ranking of schemes in this study. TOPSIS
is based on the concept that the ideal alter- The relative closeness coefficient Ai of
native has the best level for all criteria, alternative Ai , which shows the similarities
whereas the negative ideal is the one with all to ideal solution, is computed as:
the worst criteria values.20,48 The principle is e
simple: the selected best alternative should i
Ai ¼ , i ¼ 1, 2, . . . , m ð26Þ
have the shortest distance from the positive e
i þ eþ
i
ideal solution in geometrical sense while it has
the longest distance from the negative solu- Obviously from (26), 0  Ai  1. As Ai ¼ 1,
tion. The positive ideal solution Aþ, and alternative Ai is ideal solution. Oppositely, Ai
negative ideal solution (anti-ideal solution) will be anti-ideal solution as Ai ¼ 0. Thus
A are defined as: alternative Ai is closer to ideal solution and
farther from anti-ideal solution as Ai
Aþ ¼ ½Rþ þ þ approaches to 1. So the ranking order of
1 , R2 , . . . , Rn  ð22Þ
schemes by their closeness coefficients is deter-
and mined and the alternative with the maximum
closeness coefficient is the best alternative.
A ¼ ½R  
1 , R2 , . . . , Rn  ð23Þ
where Rþ 3 Application
j ¼ maxfRij ji ¼ 1, 2, . . . , mg
lþ mþ rþ
¼ ðrj , rj , rj Þ
To present the evaluation of HVAC
systems more clearly, the fuzzy multi-criteria
R
j ¼ minfRij ji ¼ 1, 2, . . . ,mg evaluation model is employed to evaluate the
 
¼ rl m r six HVAC schemes of an office building
j ,rj , rj
in Figure 2. The features and performances

Downloaded from bse.sagepub.com at UCSF LIBRARY & CKM on March 15, 2015
296 Fuzzy multi-criteria evaluation model of HVAC schemes
of HVAC schemes are summarised in Table 3 3.2 Determining the weights
according to the experts’ subjective judg- Firstly, the subjective weights are respec-
ments. The last row of Table 3 shows the tively calculated in AHP pairwise comparison
attributes of criteria. method from the users’ viewpoint to ensure
the reasonable evaluation. The three criteria
3.1 Normalising the performance matrix of technology are seen as an example and the
The scales of 0–1 in Table 1 are employed detailed deduction is given out here. Based on
to transform the linguistic terms into fuzzy the knowledge of experts, the relative impor-
numbers and they are expressed in Table 4. tance order is C1 4 C2 4 C3 and the pairwise
Then, the fuzzy criteria are normalised and comparison is summarised in Table 6.
the normalised performances of HVAC The fuzzy local weights can be obtained in
schemes are shown in Table 5. Equation (13) and the local integrated weights

Table 3 The features of six HVAC alternatives

Criteria S1 S2 S3 S4 S5 S6 Attributes

C11 High Moderate High Moderate Very high Low 4

C12 Excellent Moderate Excellent Low Excellent Moderate 4
C13 Excellent Very low Low Low Low Low 4
C21 (¥/m2) 320 368 187.76 250 162 512 5
C22 (¥/m2 year) 30.05 16.70 13.00 20.00 43.03 22.84 5
C23 (year) 20.000 20.000 15.000 20.000 8.000 15.000 4
C31 Moderate Moderate High Very high Very high High 5
C32 Moderate Moderate High High Very high High 5
C33 Very high Very high High High Low Moderate 5
C41 Excellent Excellent Low Low Low Low 4
C42 Good Excellent Low Low Low Low 4
C43 Good Good Moderate Moderate Low Moderate 4
C51 High Moderate High Moderate Very low Very low 4
C52 Moderate Very low Excellent Moderate Excellent Moderate 4
C53 Good Good Moderate Moderate Low Moderate 4
C54 Low Low Moderate Moderate Excellent Moderate 4

Note: 4 represents the benefit item and 5 is the cost item.

Table 4 The fuzzy numbers of the features of six HVAC alternatives

Criteria S1 S2 S3 S4 S5 S6

C11 (0.5, 0.7, 1.0) (0.3, 0.5, 0.7) (0.5, 0.7, 1.0) (0.3, 0.5, 0.7) (0.7, 1.0, 1.0) (0, 0.3, 0.5)
C12 (0.7, 1.0, 1.0) (0.3, 0.5, 0.7) (0.7, 1.0, 1.0) (0, 0.3, 0.5) (0.7, 1.0, 1.0) (0.3, 0.5, 0.7)
C13 (0.7, 1.0, 1.0) (0, 0, 0.3) (0, 0.3, 0.5) (0, 0.3, 0.5) (0, 0.3, 0.5) (0, 0.3, 0.5)
C21 (¥/m2) 320 368 187.76 250 162 512
C22 (¥/m2 year) 30.05 16.70 13.00 20.00 43.03 22.84
C23 (year) 20.000 20.000 15.000 20.000 8.000 15.000
C31 (0.3, 0.5, 0.7) (0.3, 0.5, 0.7) (0.5, 0.7, 1.0) (0.7, 1.0, 1.0) (0.7, 1.0, 1.0) (0.5, 0.7, 1.0)
C32 (0.3, 0.5, 0.7) (0.3, 0.5, 0.7) (0.5, 0.7, 1.0) (0.5, 0.7, 1.0) (0.7, 1.0, 1.0) (0.5, 0.7, 1.0)
C33 (0.7, 1.0, 1.0) (0.7, 1.0, 1.0) (0.5, 0.7, 1.0) (0.5, 0.7, 1.0) (0, 0.3, 0.5) (0.3, 0.5, 0.7)
C41 (0.7, 1.0, 1.0) (0.7, 1.0, 1.0) (0, 0.3, 0.5) (0, 0.3, 0.5) (0, 0.3, 0.5) (0, 0.3, 0.5)
C42 (0.5, 0.7, 1.0) (0.7, 1.0, 1.0) (0, 0.3, 0.5) (0, 0.3, 0.5) (0, 0.3, 0.5) (0, 0.3, 0.5)
C43 (0.5, 0.7, 1.0) (0.5, 0.7, 1.0) (0.3, 0.5, 0.7) (0.3, 0.5, 0.7) (0, 0.3, 0.5) (0.3, 0.5, 0.7)
C51 (0.5, 0.7, 1.0) (0.3, 0.5, 0.7) (0.5, 0.7, 1.0) (0.3, 0.5, 0.7) (0, 0, 0.3) (0, 0, 0.3)
C52 (0.3, 0.5, 0.7) (0, 0, 0.3) (0.7, 1.0, 1.0) (0.3, 0.5, 0.7) (0.7, 1.0, 1.0) (0.3, 0.5, 0.7)
C53 (0.5, 0.7, 1.0) (0.5, 0.7, 1.0) (0.3, 0.5, 0.7) (0.3, 0.5, 0.7) (0, 0.3, 0.5) (0.3, 0.5, 0.7)
C54 (0, 0.3, 0.5) (0, 0.3, 0.5) (0.3, 0.5, 0.7) (0.3, 0.5, 0.7) (0.7, 1.0, 1.0) (0.3, 0.5, 0.7)

Downloaded from bse.sagepub.com at UCSF LIBRARY & CKM on March 15, 2015
 Table 5 The normalised features of six HVAC alternatives

Criteria S1 S2 S3 S4 S5 S6

C11 (0.500, 0.700, 1.000) (0.300, 0.500, 0.700) (0.500, 0.700, 1.000) (0.300, 0.500, 0.700) (0.700, 1.000, 1.000) (0.000, 0.300, 0.714)
C12 (0.700, 1.000, 1.000) (0.300, 0.500, 0.700) (0.700, 1.000, 1.000) (0.000, 0.300, 0.714) (0.700, 1.000, 1.000) (0.300, 0.500, 1.000)
C13 (0.700, 1.000, 1.000) (0.000, 0.000, 0.429) (0.000, 0.300, 0.714) (0.000, 0.300, 0.714) (0.000, 0.300, 0.714) (0.000, 0.300, 0.714)
C21 (0.506, 0.506, 0.506) (0.440, 0.440, 0.440) (0.863, 0.863, 0.863) (0.648, 0.648, 0.648) (1.000, 1.000, 1.000) (0.316, 0.316, 0.316)
C22 (0.432, 0.432, 0.432) (0.778, 0.778, 0.778) (1.000, 1.000, 1.000) (0.650, 0.650, 0.650) (0.302, 0.302, 0.302) (0.570, 0.570, 0.570)
C23 (1.000, 1.000, 1.000) (1.000, 1.000, 1.000) (0.750, 0.750, 0.750) (1.000, 1.000, 1.000) (0.400, 0.400, 0.400) (1.000, 1.000, 1.000)
C31 (0.429, 1.000, 1.000) (0.429, 1.000, 1.000) (0.300, 0.714, 1.000) (0.300, 0.500, 1.000) (0.300, 0.500, 1.000) (0.300, 0.714, 1.000)
C32 (0.429, 1.000, 1.000) (0.429, 1.000, 1.000) (0.300, 0.714, 1.000) (0.300, 0.714, 1.000) (0.300, 0.500, 1.000) (0.300, 0.714, 1.000)
C33 (0.300, 0.500, 1.000) (0.300, 0.500, 1.000) (0.300, 0.714, 1.000) (0.300, 0.714, 1.000) (0.429, 1.000, 1.000) (0.429, 1.000, 1.000)
C41 (0.700, 1.000, 1.000) (0.700, 1.000, 1.000) (0.300, 0.714, 1.000) (0.300, 0.714, 1.000) (0.300, 0.714, 1.000) (0.300, 0.714, 1.000)
C42 (0.500, 0.700, 1.000) (0.700, 1.000, 1.000) (0.300, 0.714, 1.000) (0.300, 0.714, 1.000) (0.300, 0.714, 1.000) (0.300, 0.714, 1.000)
C43 (0.500, 1.000, 1.000) (0.500, 1.000, 1.000) (0.300, 0.714, 1.000) (0.300, 0.714, 1.000) (0.000, 0.429, 1.000) (0.300, 0.714, 1.000)
C51 (0.500, 1.000, 1.000) (0.300, 0.714, 1.000) (0.500, 1.000, 1.000) (0.300, 0.714, 1.000) (0.000, 0.000, 0.600) (0.000, 0.000, 0.600)
C52 (0.300, 0.500, 1.000) (0.000, 0.000, 0.429) (0.500, 0.700, 1.000) (0.300, 0.500, 1.000) (0.700, 1.000, 1.000) (0.300, 0.500, 1.000)
C53 (0.500, 1.000, 1.000) (0.500, 1.000, 1.000) (0.300, 0.714, 1.000) (0.300, 0.714, 1.000) (0.000, 0.429, 1.000) (0.300, 0.714, 1.000)
C54 (0.300, 0.714, 1.000) (0.300, 0.714, 1.000) (0.300, 0.500, 1.000) (0.300, 0.500, 1.000) (0.700, 1.000, 1.000) (0.300, 0.500, 1.000)

is solved as:
computed to:
HVAC users.

cT

Downloaded from bse.sagepub.com at UCSF LIBRARY & CKM on March 15, 2015
the average weight.

c
JðW Þ ¼ W BW ¼


max ¼ 1:2607
on the pairwise comparison of experts.
JJ Wang et al.

0:9471 1:7910
0:6306 0:9471


Then, its maximum characteristic root max

13.3 times than S2 and S6 schemes.
It can be seen that the weight of cleanliness

Obviously, the objective weights are different

Cleanliness (C41), and fresh air (C42) follows

marised in Table 9 and the global normalised
computed according to the pairwise compar-

Section 2.4.3. The construction function is

difference of six HVAC schemes, which the
lated in Equation (15) and (16) are sum-

found that the weight of humidity control

fresh air (C42) and thermal comfort (C43) are
HVAC systems from experts’ subjective judg-
(C41) is the maximum and cleanliness is
Then the rule weights at level 2 are

lated in optimal weighting method in

The final combination weights are calcu-
to C13 closely. Temperature control (C12)
(C13) is the maximum because of the greatest
weights, Wo , are displayed in Figure 6.
Similarly the local objective weights calcu-

integrated performance of S1 scheme is about

to the subjective weights in Figure 5. It can be
all greater than the average weight, 0.063.
(C22), space use (C33), service life (C23),
ments. Operation and maintenance cost
thought of the important criteria evaluated
Equation (14), Ws , are shown in Figure 5.
The global subjective normalised weights in
ison (Table 8) and they are shown in Table 7.
Similarly, other criteria are calculated based
in Equation (8) and they are filled into Table 7.
297

and convenience (C51) are also greater than

These criteria are also paid more attention by
298 Fuzzy multi-criteria evaluation model of HVAC schemes
The corresponding characteristic vector is So the combination weights can be calcu-
normalised as: lated in Equation (18) to:
K ¼ ½0:4263, 0:5737T
Table 6 Pairwise comparison for the criteria in technology Wc ¼ 0:4263Ws þ 0:5737Wo
based on the experts knowledge
and the final weights are shown in Figure 7.
C1 C2 C3
As can be seen, the rank order of the
C1 (1,1,1) (1,3/2,2) (3/2,2,5/2) importance in multi-criteria evaluation is
C2 (1/2,2/3,1) (1,1,1) (3/2,2,5/2) C414C424C134C224C514C124C434
C3 (2/5,1/2,2/3) (2/5,1/2,2/3) (1,1,1)
C524C114C234C334C214C31¼C544

Table 7 The subjective local weights of rules and criteria

Rules Fuzzy weights Integrated Criteria Fuzzy Integrated

weighs weights weighs

Technology (0.067, 0.097, 0.152) 0.098 C11 (0.299, 0.455, 0.659) 0.451
C12 (0.237, 0.347, 0.523) 0.349
C13 (0.142, 0.199, 0.294) 0.200
Economoy (0.160, 0.251, 0.391) 0.251 C21 (0.132, 0.181, 0.259) 0.182
C22 (0.326, 0.480, 0.678) 0.476
C23 (0.235, 0.340, 0.507) 0.342
Environment (0.116, 0.182, 0.289) 0.183 C31 (0.220, 0.304, 0.424) 0.304
C32 (0.133, 0.177, 0.248) 0.179
C33 (0.376, 0.519, 0.700) 0.517
Health (0.220, 0.339, 0.502) 0.335 C41 (0.487, 0.600, 0.730) 0.600
C42 (0.174, 0.200, 0.233) 0.200
C43 (0.174, 0.200, 0.233) 0.200
Management (0.086, 0.132, 0.210) 0.133 C51 (0.169, 0.200, 0.242) 0.200
C52 (0.288, 0.400, 0.531) 0.400
C53 (0.169, 0.200, 0.242) 0.200
C54 (0.169, 0.200, 0.242) 0.200

Table 8 Pairwise comparison for the rules based on the experts knowledge

C1 C2 C3 C4 C5

C1 (1,1,1) (1/3,2/5,1/2) (2/5,1/2,2/3) (2/7,1/3,2/5) (1/2,2/3,1)

C2 (2,5/2,3) (1,1,1) (1,3/2,4/2) (1/2,2/3,1) (3/2,2,5/2)
C3 (3/2,2,5/2) (1/2,2/3,1) (1,1,1) (2/5,1/2,2/3) (1,3/2,4/2)
C4 (5/2,3,7/2) (1,3/2,4/2) (3/2,2,5/2) (1,1,1) (2,5/2,3)
C5 (1,3/2,4/2) (2/5,1/2,2/3) (1/2,2/3,1) (1/3,2/5,1/2) (1,1,1)

0.252
0.201
0.189

0.119
0.126
0.086 0.094
0.044 0.045 0.056 0.067 0.067 0.053
0.063
0.034 0.033 0.027 0.027 0.027
0.020
0.000
C11 C12 C13 C21 C22 C23 C31 C32 C33 C41 C42 C43 C51 C52 C53 C54

Figure 5 The global subjective normalised weights of criteria

Downloaded from bse.sagepub.com at UCSF LIBRARY & CKM on March 15, 2015
 JJ Wang et al. 299
C32 ¼ C53. The weight of the cleanliness 3.3 Ranking the schemes
(C41) is the maximum, fresh air (C42) and It is the turn to determine the ranking of
humidity control (C13) follows, and they are HVAC schemes after determining the criteria
greater than the average weight. weighs. Based on the normalised judgment
matrix G0 , the ideal and anti-ideal solutions
Table 9 The objective local weights of rules and criteria are easily obtained in Equations (22) and (23)
Rules Integrated Criteria Integrated in Table 10.
weighs weighs Then the weighted distance between
schemes and the ideal and anti-ideal solution
Technology 0.350 C11 0.145
C12 0.198 in Equations (24) and (25) are calculated and
C13 0.658 summarised in Table 11.
Economoy 0.049 C21 0.402 Finally, the closeness coefficients Ai of six
C22 0.377
C23 0.222 HVAC schemes in Equation (26) are calcu-
Environment 0.010 C31 0.368 lated and filled into Table 11. The ranking
C32 0.264
C33 0.368
order determined by their closeness coeffi-
Health 0.428 C41 0.504 cients is S1 4 S3 4 S2 4 S5 4 S4 4 S6.
C42 0.411 Therefore, constant air volume HVAC
C43 0.085
Management 0.163 C51 0.503 system (S1) is the best HVAC option, fan-
C52 0.287 coil HVAC system and VAV HVAC system
C53 0.054 (S2) are close and locate the second place in
C54 0.156
this study.

0.252
0.230
0.216
0.189 0.176

0.126
0.082
0.069
0.063 0.051 0.047
0.036
0.020 0.018 0.011 0.025
0.004 0.003 0.004 0.009
0.000
C11 C12 C13 C21 C22 C23 C31 C32 C33 C41 C42 C43 C51 C52 C53 C54

Figure 6 The global objective normalised weights of criteria

0.252
0.209
0.189
0.141
0.129
0.126

0.061 0.043
0.063 0.048 0.054 0.042
0.050 0.058 0.049
0.031 0.026 0.016
0.016 0.026
0.000
C11 C12 C13 C21 C22 C23 C31 C32 C33 C41 C42 C43 C51 C52 C53 C54

Figure 7 The global combination weights of criteria

Downloaded from bse.sagepub.com at UCSF LIBRARY & CKM on March 15, 2015
300 Fuzzy multi-criteria evaluation model of HVAC schemes
4 Analysis and discussion weights of the first criteria and the last criteria
are exchanged. Thus, 15 evaluation results are
4.1 Sensitivity analysis obtained.
In this research, in order to improve the Then, the second criteria is selected to
confidence in the evaluation results, a sensi- exchange with other criteria and 14 evalua-
tivity analysis is necessary. In this part, we tion results are obtained again. Like the
repeat the multi-criteria analysis for different exchange process, the calculation is finished
weights. after the fifteenth and the sixteenth criteria
The idea of sensitivity analysis is to are exchanged.
exchange each criteria weight with another Finally, the results of exchanged criteria
criteria weight in the final combination plus the result in Section 3.3 are totally 121
weights. Firstly, the weight of the first criteria kinds of ranking orders. The 121 different
and the weight of the second criteria are calculation results are displayed in Figure 8.
exchanged and other weighs are kept invar- As can be seen, the ranking orders are not
iant, thus, the closeness coefficients can be always as same as the order in Section 3.3,
recalculated and the six HVAC schemes are which shows the result is influenced by the
ranked again. Then, the weight of the second criteria weights and proves the fact that the
criteria is back to initial value, the first criteria criteria weights obtained in optimal weighting
and the third criteria are exchanged and the methods are necessary. However, it is also
ranking order of HVAC schemes is obtained noted that the weights does not influence the
again. The procedure is continued, until the choice of the best HVAC (S1 scheme). To find
out the robust of evaluation result, the rank
Table 10 The ideal and anti-ideal solutions of six HVAC distribution of six HVAC schemes in 121
systems on 16 criteria kinds of orders are calculated and sum-
Criteria Aþ A
marised in Table 12. It is seen that S1, S3,
S2, S5, S4 and S6 in the first place to the last
C11 (0.700, 1.000, 1.000) (0.000, 0.300, 0.714) place are 100, 88.34, 76.03, 85.95, 97.52 and
C12 (0.700, 1.000, 1.000) (0.000, 0.300, 0.714)
C13 (0.700, 1.000, 1.000) (0.000, 0.000, 0.429) 100%, respectively. Based on the place of the
C21 (1.000, 1.000, 1.000) (0.316, 0.316, 0.316) maximum value, it can be concluded that the
C22 (1.000, 1.000, 1.000) (0.302, 0.302, 0.302) ranking order of six HVAC schemes is
C23 (1.000, 1.000, 1.000) (0.400, 0.400, 0.400)
C31 (0.429, 1.000, 1.000) (0.300, 0.500, 1.000) S1 4 S3 4 S2 4 S5 4 S4 4 S6. The result is
C32 (0.429, 1.000, 1.000) (0.300, 0.500, 1.000) as same as the result in Section 3.3, which
C33 (0.429, 1.000, 1.000) (0.300, 0.500, 1.000) indicates the strong robust of fuzzy TOPSIS
C41 (0.700, 1.000, 1.000) (0.000, 0.300, 0.714)
C42 (0.700, 1.000, 1.000) (0.000, 0.300, 0.714) method.
C43 (0.500, 1.000, 1.000) (0.000, 0.429, 1.000)
C51 (0.500, 1.000, 1.000) (0.000, 0.000, 0.600)
C52 (0.700, 1.000, 1.000) (0.000, 0.000, 0.429) 4.2 Discussion
C53 (0.500, 1.000, 1.000) (0.000, 0.429, 1.000)
C54 (0.700, 1.000, 1.000) (0.000, 0.300, 0.714)
Similarly, the schemes of HVAC system can
be solely evaluated and ranked in one rule.

Table 11 The weighted distance and the closeness coefficients between six HVAC schemes and the
ideal and anti-ideal solutions

S1 S2 S3 S4 S5 S6
þ
e 0.0147 0.1049 0.0737 0.0859 0.1175 0.1131
e 0.1475 0.0700 0.0571 0.165 0.518 0.0104
A* 0.9094 0.4004 0.4365 0.1609 0.3060 0.0840

Downloaded from bse.sagepub.com at UCSF LIBRARY & CKM on March 15, 2015
 JJ Wang et al. 301
The each evaluation results from technology, system (S1), VAV system (S2) and distributed
economy, environment, heath and manage- air-conditioners (S5) are almost same.
ment are shown in Figure 9, receptively. On the health aspect, only constant air
From the technical rule, constant air volume system (S1) and VAV system (S2) are
volume HVAC system (S1) is the opti- thought of health-friendly scheme due to
mal scheme and it can be called technical- better cleanliness, fresh air and thermal
friendly scheme in this study. Fan-coil comfort.
HVAC system (S3) and distributed HVAC From the management rule, fan-coil HVAC
system (S5) follow and other schemes are the system (S3) is the ideal scheme and constant air
worst. volume system (S1), induction unit HVAC
Similarly, it is found that fan-coil HVAC system (S4) and distributed air-conditioners
system (S3) is economic-friendly scheme, (S5) are kept the same level.
induction unit HVAC system (S4) and VAV It is found that constant air volume HVAC
HVAC system (S2) follows closely and the system (S1) is only the first place in technol-
worst scheme is distributed HVAC system. ogy and health; however, the integrated
From the three criteria in environment, performance is the best scheme. Fan-coil
VRV HVAC system is ideal environmental- HVAC system (S3) locates the second place
friendly scheme. Constant air volume HVAC in better performance.

1.0000

0.8000
Closeness coefficient

0.6000

0.4000

0.2000

0.0000
1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 73 77 81 85 89 93 97 101 105 109 113 117 121

S1 S2 S3 S4 S5 S6

Figure 8 The new closeness coefficients values of HVAC schemes

Table 12 The rank distribution of six HVAC schemes

Schemes

Rank S1 (%) S2 (%) S3 (%) S4 (%) S5 (%) S6 (%)

1st 100 0 0 0 0 0
2nd 0 11.57 88.34 0 0 0
3rd 0 76.03 11.57 0 12.40 0
4th 0 11.57 0 2.48 85.95 0
5th 0 0.83 0 97.52 1.65 0
6th 0 0 0 0 0 100

Downloaded from bse.sagepub.com at UCSF LIBRARY & CKM on March 15, 2015
302 Fuzzy multi-criteria evaluation model of HVAC schemes
1.2000

0.8000

0.4000

0.0000
Technology Economy Environment Health Management

S1 S2 S3 S4 S5 S6

Figure 9 The closeness coefficients values of HVAC schemes in single rule

5 Conclusion better scheme as long as the better improve-

ment in humidity control and health aspect.
Fuzzy sets theory, when combined with the Although the specific results for this par-
multi-criteria analytic hierarchy process, can ticular analysis cannot be applied to every
be used to evaluate the HVAC systems. The HVAC project, it is believed that if this
procedure has the advantage of allowing for a procedure is applied correctly, fuzzy multi-
more systematic handling of the uncertainty criteria model can become a powerful tool to
of information and the vagueness of decision- make an optimal selection for HVAC schemes
maker’s feeling and recognition. This simple and other building service contexts. For
and practical methodology could be very further research, the opinions of different
interesting to engineers involved in HVAC authorities and professionals should be taken
systems selection or other complex systems. into account, and developing a group deci-
The fuzzy multi-criteria evaluation model for sion-making system and finding the represen-
HVAC systems proposed in this paper leads tative evaluation criteria of HVAC systems
to the following conclusions. can be very useful.
The evaluation and selection result is
influenced by the criteria weights directly. It
is necessary to obtain the rational weights. Acknowledgements
The optimal weighting method combines the
subjectivity of decision-makers and profes- This research has been supported by the Key
sional experts with the objectivity of measure Laboratory of Condition Monitoring and Control
data. After calculating weights of the criteria for Power Plant Equipment of Ministry of
in this study, it is found that the cleanliness Education, China.
(C41) is the most important criteria in HVAC The authors also wish to express their gratitude
system, and fresh air (C42) and humidity to the anonym referees for remarks and sugges-
control (C13) follows. They are paid more tions that improved this paper significantly.
attention in the design of HVAC project.
The evaluation results in fuzzy multi-crite- References
ria model combined the optimal weighting
method indicate that the constant air volume 1 Pérez-Lombard L, Ortiz J, Pout C.
HVAC system is the optimal scheme in this A review on buildings energy consumption
paper because of its better integrated perfor- information. Energy and Buildings 2008; 40(3):
mance. Fan-coil HVAC system is also the 394–98.

Downloaded from bse.sagepub.com at UCSF LIBRARY & CKM on March 15, 2015
 JJ Wang et al. 303
2 Ellis MW, Mathews EH. Needs and trends in 14 Zavadskas EK, Vilutiene T. A multiple criteria
building and HVAC system design tools. evaluation of multi-family apartment block’s
Building and Environment 2002; 37(5): 461–70. maintenance contractors: I–Model for main-
3 Hanby VI, Angelov PP. Application of uni- tenance contractor evaluation and the deter-
variate search methods to the determination of mination of its selection criteria. Building and
HVAC plant capacity. Building Service Environment 2006; 41(5): 621–32.
Engineering Research and Technology 2000; 15 Sheu J-B. A hybrid neuro-fuzzy analytical
21(3): 161–66. approach to mode choice of global logistics
4 Elkhuizen PA, Peitsman HC, Wienk WJ. A management. European Journal of Operational
new design guideline for the heating and Research 2008; 189(3): 971–86.
cooling curve in AHU units of HVAC systems. 16 Chen C-T, Lin C-T, Huang S-F. A fuzzy
Building Service Engineering Research and approach for supplier evaluation and selec-
Technology 2003; 24(3): 191–202. tion in supply chain management.
5 Wong JKW, Li H. Application of the analytic International Journal of Production Economics
hierarchy process (AHP) in multi-criteria 2006; 102(2): 289–301.
analysis of the selection of intelligent building 17 Chang N-B, Parvathinathan G, Breeden JB.
systems. Building and Environment 2008; 43(1): Combining GIS with fuzzy multicriteria
108–25. decision-making for landfill siting in a
6 Buys JH, Mathews EH. Investigation into fast-growing urban region. Journal of
capital costs of HVAC systems. Building and Environmental Management 2008; 87(1):
Environment 2005; 40(9): 1153–63. 139–53.
7 Kilkis BI. Cost optimization of a hybrid 18 Sadiq R, Khan FI, Veitch B. Evaluating
HVAC system with composite radiant wall offshore technologies for produced water
panels. Applied Thermal Engineering 2006; management using GreenPro-I–a risk-based
26(1): 10–17. life cycle analysis for green and clean process
8 Mazzei P, Minichiello F, Palma D. Desiccant selection and design. Computers & Chemical
HVAC systems for commercial buildings. Engineering 2005; 29(5): 1023–39.
Applied Thermal Engineering 2002; 22(5): 19 Geldermann J, Spengler T, Rentz O. Fuzzy
545–60. outranking for environmental assessment.
9 Arkin H, Navon R, Burg I. HVAC with Case study: iron and steel making industry.
thermal energy storage: Optimal design and Fuzzy Sets and Systems 2000; 115(1): 45–65.
optimal scheduling. Building Service 20 Wang JJ, Jing YY, Zhang CF, Shi GH,
Engineering Research and Technology 1997; Zhang XT. A fuzzy multi-criteria decision-
18(1): 31–38. making model for trigeneration system.
10 Afgan NH, Pilavachi PA, Carvalho MG. Energy Policy 2008; 36(10): 3823–32.
Multi-criteria evaluation of natural gas 21 Wang JJ, Jing YY, Zhang CF, Zhang XT,
resources. Energy Policy 2007; 35(1): 704–13. Shi GH. Integrated evaluation of distributed
11 Pilavachi PA, Roumpeas CP, Minett S, Afgan triple-generation systems using improved grey
NH. Multi-criteria evaluation for CHP system incidence approach. Energy 2008; 33(9):
options. Energy Conversion and Management 1427–37.
2006; 47(20): 3519–29. 22 Alanne K, Saari A. Sustainable small-scale
12 Pohekar SD, Ramachandran M. Application CHP technologies for buildings: the basis for
of multi-criteria decision making to sustain- multi-perspective decision-making. Renewable
able energy planning–A review. Renewable and and Sustainable Energy Reviews 2004; 8(5):
Sustainable Energy Reviews 2004; 8(4): 365–81. 401–31.
13 Wang JJ, Jing YY, Zhan CF. Weighting 23 Wang JJ, Zhang CF, Jing YY, Zheng GZ.
methodologies in multi-criteria evaluations Using the fuzzy multi-criteria model to select
of combined heat and power systems. the optimal cool storage system for air
International Journal of Energy Research 2009; conditioning. Energy and Buildings 2008;
Available at: http://dx.doi.org/10.1002/er.1527. 40(11): 2059–66.

Downloaded from bse.sagepub.com at UCSF LIBRARY & CKM on March 15, 2015
304 Fuzzy multi-criteria evaluation model of HVAC schemes
24 Zheng GZ, Jing YY. Air conditioning: 36 Saaty TL. The analytic hierarchy process.
selecting the optimal cool storage system. New York, McGraw-Hill, 1980.
Energy and Environment 2007; 18(2): 251–57. 37 Zmeureanu R, Yu Wu X. Energy and exergy
25 Park T, Kim K-J. Determination of an opti- performance of residential heating systems
mal set of design requirements using house of with separate mechanical ventilation. Energy
quality. Journal of Operations Management 2007; 32(3): 187–95.
1998; 16(5): 569–81. 38 Pan Y, Zhou H, Huang Z, Zeng Y, Long W.
26 Hsieh T-Y, Lu S-T, Tzeng G-H. Fuzzy Measurement and simulation of indoor air
MCDM approach for planning and design quality and energy consumption in two
tenders selection in public office buildings. Shanghai office buildings with variable air
International Journal of Project Management volume systems. Energy and Buildings 2003;
2004; 22(7): 573–84. 35(9): 877–91.
27 Soebarto VI, Williamson TJ. Multi-criteria 39 Ye G, Yang C, Chen Y, Li Y. A new approach
assessment of building performance: theory for measuring predicted mean vote (PMV) and
and implementation. Building and Environment standard effective temperature (SET*).
2001; 36(6): 681–90. Building and Environment 2003; 38(1): 33–44.
28 Wong J, Li H, Lai J. Evaluating the system 40 Mui KW, Chan WT. Building calibration for
intelligence of the intelligent building systems: IAQ management. Building and Environment
Part 2: Construction and validation of ana- 2006; 41(7): 877–86.
lytical models. Automation in Construction 41 Namiesnik J, Górecki T, Kozdron-Zabiega la
2008; 17(3): 303–21. B, Lukasiak J. Indoor air quality (IAQ), pollu-
29 Balcomb JD, Curtner A. Multi-criteria tants, their sources and concentration levels.
Decision-making Process for Buildings: Building and Environment 1992; 27(3): 339–56.
Proceedings of the 35th Intersociety on Energy 42 Wolkoff P, Kjærgaard SK. The dichotomy of
Conversion Engineering Conference and relative humidity on indoor air quality.
Exhibit. Las Vegas, NV, USA, 2000. Environment International 2007; 33(6): 850–57.
30 Jiang Y, Liu XH, Xue ZF, Fu L. Study of 43 Zadeh LA. Fuzzy sets. Information and
evaluation criteria for energy conversion Control 1965; 8(3): 338–53.
system. Energy of China 2004; 26(3): 27–31 (in 44 Chen SH, Hsieh CH. Representation, ranking,
Chinese). distance, and similarity of L–R type fuzzy
31 Zhang PH, Wan HH, Huang XY, Zhang XM, number and application. Australian Journal of
Liu M. The comprehensive evaluation of Intelligent Processing Systems 2000; 6(4):
HVAC facility management based on AHP. 217–29.
Journal of Shenyang Jianzhu University 45 Zeleny M. Multiple criteria decision making.
(Natural Science) 2005; 21(3): 242–45 Singapore, McGraw-Hill Book Company,
(in Chinese). 1982.
32 Zhang FZ, Zhang XF. The design grade of air- 46 Wang ZY, Gu HF, Yi XX, Zhang SR. A
conditioning project and its evaluation princi- method of determining the linear combination
ple. Building Energy and Environment 2000; weights based on entropy. Journal of Systems
19(4): 35–37 (in Chinese). Engineering-theory and Practice 2003; 23(3):
33 Long WD. Building energy saving and effi- 112–16 (in Chinese).
ciency management. Beijing, China Building 47 Chen HY. Research on optimal combination
Industry Press (in Chinese), 2005. determining weights method for multiple
34 He H, Feng SH. Comprehensive assessment attribute decision making. Operations
and optimization of air-conditioning project Research and Management Science 2003; 12(2):
design schemes. Energy Conservation 6–8 (in Chinese).
Technology 2002; 20(4): 9–11 (in Chinese). 48 Hwang CL, Yoon K. Multiple attribute
35 Kahraman C. Fuzzy multi-criteria decision decision making-methods and application:
making: theory and applications with recent a state-of-the-art survey. New York, USA,
developments. US, Springer, 2008. Springer-Verlag, 1981.

Downloaded from bse.sagepub.com at UCSF LIBRARY & CKM on March 15, 2015