You are on page 1of 7

Journal of Information & Computational Science 10: 1 (2013) 17

Available at http://www.joics.com
Application of Combination Weighing and Outranking
Relation Method in Evaluation of Power Quality
Liguo Fan

, Yanxia Zhang
Key Laboratory of Smart Grid of Ministry of Education, Tianjin University, Tianjin 300072, China
Abstract
A New Decision-making (MADM) method based on combination weighing and outranking relation is
presented by introducing combination weighing to traditional outranking relation in the paper. Firstly,
the subjective weights and objective weights of power quality indices are decided by FAHP and entropy
methods. Secondly, uncertainty of the data is expressed by thresholds according to Decision-maker
(DM)s risk attitudes by outranking relation. Thirdly, the power quality of all observation points are
compared in pairs, the score of all observation points are calculated under single attribute and the results
are expressed with index of concordance and discordance. Finally, the degree of outranking is calculated
by integrating the index of concordance and discordance. The nal ranking of the power quality of all
observation points are obtained by net credibility degree. The eectiveness of the method is validated
by simulation results.
Keywords: Power Quality; Outranking Relation; Fuzzy Analytic Hierarchy Process; Entropy
1 Introduction
Since the technology of digital substation, smart grid, microgrid and so on are developed in fast,
power system is converting quantity into quality. In a liberalized electricity market in the future,
it is no wonder that end-users have dierent level demands of power quality (voltage quality
and reliability) associated with the power energy pricing variations. To support the multi-level
transactions of power quality in the future market, it is necessary to establish a set of Power
Quality Comprehensive Evaluation (PQCE) system [1], which is reasonable, acceptable and easily
understood by the general public. In recent years many experts and scholars have tended to focus
on the methods of Power Quality Comprehensive Evaluation (PQCE). Some advance fuzzy model
and comprehensive indices based on fuzzy principle are used to evaluate power quality [2]. Some
advance evaluation methods are based on combination enpowerment and the deciency of single
enpowerment methods is overcome [3]. Another fuzzy mathematics and fuzzy AHP in PQCE are
applied and fuzzy uncertainty of subjective judgment is resolved [4]. The concept of combination
weighing and outranking relation is proposed to evaluate power quality in the paper. A actual
case of power quality of ve observation points is applied in the paper.

Corresponding author.
Email address: fanliguo@sohu.com (Liguo Fan).
15487741 / Copyright 2013 Binary Information Press
January 1, 2013
2 L. Fan et al. / Journal of Information & Computational Science 10: 1 (2013) 17
2 Combination Weighing Method
The weights of indices of power quality comprehensive evaluation are decided by FAHP and
entropy methods. The deciency of single enpowerment methods is overcome and the weight is
decided rationally.
2.1 Decision of Subjective Weights
Fuzzy judgement matrix is formed by introducing triangular fuzzy numbers [5] in consideration
of uncertainty of subjective judgement formation. Indices seqencing and weight are decided by
using fuzzy numbers compare principle.
Judgement matrix is indicated by 1 9 scale [6] and reciprocal according to relative important
level of each level element.
The quantity of power quality indices that is related with upper index is set and index assem-
blage is set as X = {x
1
, x
2
, . . . , x
m
}. Triangular fuzzy numbers b
ij
=
_
l
ij
e
ij
p
ij
_
of fuzzy
judgement is decided to important level of index i comparing with index j. Left and right exten-
sion l
ij
and p
ij
is indicated fuzzy level of judgement. The value of p
ij
l
ij
is bigger, fuzzy level of
comparing judgement is higher. Fuzzy comparing judgement matrix B is obtained by comparing
nally:
B = (b
ij
)
mn
=
_

_
_
l
11
e
11
p
11
_

_
l
1m
e
1m
p
1m
_
_
l
21
e
21
p
21
_

_
l
2m
e
2m
p
2m
_
.
.
.
.
.
.
.
.
.
.
.
.
_
l
m1
e
m1
p
m1
_

_
l
mm
e
mm
p
mm
_
_

_
(1)
Fuzzy relative weight vector of index i comparing with other indices is decided in fuzzy comparing
judgement matrix:
Q
i
=
_

_
m

j=1
l
ij
m

i=1

j = 1
m
p
ij
m

j=1
e
ij
m

i=1

j = 1
m
e
ij
m

j=1
p
ij
m

i=1

j = 1
m
l
ij
_

_
1
(2)
Triangular fuzzy numbers of fuzzy relative weight vector are dened clearly in order to sequence.
Subjective weight of Q
i
is decided:
w
i
=
l
i
+ 2e
i
+ p
i
4
(3)
2.2 Decision of Objective Weights
Objective weight based on entropy is applied in the paper. Information entropy is indicated
uncertainty magnitude for a random event or occurrence probability of some specied information.
L. Fan et al. / Journal of Information & Computational Science 10: 1 (2013) 17 3
Entropy is bigger, degree of out-of-order is higher. Otherwise, degree of in-order is higher. Weights
are decided by comparing entropy value of indices according to entropy value in PQCE. Entropy
value of indices is smaller, degree of variation of indices data series is bigger, weight of indices is
bigger.
Entropy value of index i:
S
i
= k
n

j=1
P
ij
lnP
ij
(i = 1, , m; j = 1, , n) (4)
where k =
1
lnn
when P
ij
= 0, P
ij
lnP
ij
= 0.
Objective weight
i
is decided:

i
=
1 S
i
m

i=1
(1 S
i
)
=
1 S
i
m
m

i=1
S
i
(i = 1, , m; j = 1, , n) (5)
2.3 Decision of Synthetic Weights
a
i
=
w
i

i
n

j=1
w
j

j
, i = 1, , n (6)
3 Outranking Relation
ELECTRE method in [7, 8] is presented by Benoyown in 1969. The key of ELECTRE method is
using the concept of outranking relation [9] which demands decision-maker to regard alternative 1
is superior than alternative 2 according to some risk degree. Decision-maker can determine order
of quality by a series of evaluating for outranking relation.
Outranking relation [10] is regarded the third method as treating uncertain factor except proba-
bility and fuzzy. Among ELECTRE III is the most representation method of applying outranking
relation. It begins from decision-makers point of view, establishes outranking relation according
to decision-makers risk attitude and considers the concept of imperfect compensation among at-
tribute. The theory is easily accepted because it is embody of mens daily thought. ELECTRE III
is multi-attribute decision-making method which is recognized and applied extensively in Europe.
A MADM problem is usually formulated by a set of alternatives A = {a
1
, a
2
, , a
n
}, a set of
estimate attribute F = {f
1
, f
2
, , f
n
} and evaluation of estimate X
ik
of alternatives under esti-
mate attribute. Where X
ik
presents evaluation of estimate of alternative under estimate attribute
f
k
[11, 12] without loss of generality, given that all the criteria values are to be maximized.
3.1 Denition of Thresholds
For any ordered pair (a
i
, a
k
) of alternatives, the three thresholds are as follows:
4 L. Fan et al. / Journal of Information & Computational Science 10: 1 (2013) 17
The indierence threshold q
i
: for the attribute f
j
, a
i
and a
k
are indierent if a
ij
a
kj
q
j
and
a
kj
a
ij
q
j
.
The strict preference threshold p
j
: for the attribute, f
j
is strictly preferred to a
k
if a
ij
a
kj
> p
j
.
The veto threshold
j
: for the attribute f
j
, reject the hypothesis of outranking of a
i
over a
k
if
a
ij
a
kj
>
j
.
It is implied that:
j
> p
j
> q
j
> 0.
3.2 Calculating Concordance
A concordance index C(a
i
, a
k
) is computed for each ordered pair (a
i
, a
k
) of alternatives and dened
by:
C(a
i
, a
k
) =
m

j=1

j
c
j
(a
i
, a
k
) (7)
where
j
is the weight determining the relative importance of the attribute f
j
and c
j
(a
i
, a
k
) is
dened by:
c
j
(a
i
, a
k
) =
_

_
1 if X
kj
X
ij
q
j
0 if X
kj
X
ij
p
j
p
j
+ X
ij
X
kj
p
j
q
j
otherwise
(8)
where c
j
(a
i
, a
k
) shows the degree of concordance with the judgmental statement that a
i
outranks
a
k
under the attribute f
j
, the index of global concordance C(a
i
, a
k
) represents the amount of evi-
dence to support the concordance among all the attribute, under the hypothesis that a
i
outranks
a
k
.
3.3 Calculating Discordance
d
j
(a
i
, a
k
) =
_

_
1 if X
ij
X
kj
p
j
0 if X
ij
X
kj

j
X
ij
X
kj
p
j

j
p
j
p
j
< X
ij
X
kj
<
j
(9)
where d
j
(a
i
, a
k
) shows the degree of discordance with the judgmental statement that a
i
outranks
a
k
. Discordance includes non-compensable ingredient of attribute. It is unacceptable of using
good data to substitute defective data.
3.4 Calculating the Degree of Outranking
The degree of outranking S(a
i
, a
k
) of order alternatives couple (a
i
, a
k
) is dened by:
S(a
i
, a
k
) =
_
_
_
C(a
i
, a
k
) if J(i, k)
C(a
i
, a
k
)
jJ(i,k)
1 d
j
(a
i
, a
k
)
1 C(a
i
, a
k
)
otherwise
(10)
L. Fan et al. / Journal of Information & Computational Science 10: 1 (2013) 17 5
where J(i, k) is dened as the set of attribute for which d
j
(a
i
, a
k
) > C(a
i
, a
k
). If J(i, k) = ,
we have d
j
(a
i
, a
k
) C(a
i
, a
k
) for all attribute, then, S(a
i
, a
k
) is the same as C(a
i
, a
k
) . S(a
i
, a
k
)
shows the degree of credibility of outranking with the judgmental statement that a
i
outranks a
k
.
4 Case Study
The paper simulates and analyzes eld data of power quality from ve observation point of some
area according to the method. Field data of power quality from ve observation point of the area
are shown as Table 1.
Table 1: Field data of power quality from observation point
Observ point Observ point 1 Observ point 2 Observ point 3 Observ point 4 Observ point 5
Voltage deviation (%) 3.21 6.68 4.35 5.33 4.22
Voltage uctuation (%) 1.33 1.53 1.95 1.37 1.58
Voltage sag 1.745 0.826 1.396 2.048 2.935
Unbalance (%) 0.30 0.72 1.55 1.10 1.95
Harmonics (%) 1.72 4.28 2.67 3.36 4.57
Frequency deviation (Hz) 0.092 0.095 0.100 0.098 0.096
Interruption (h/a) 170.4 204 156 208.8 182.4
4.1 Determination of Weights of Indices
Indices weight is establishes by using subjective enpowerment based on fuzzy AHP method and
by using objective enpowerment based on entropy-right method in the paper. The weight is
indicated in Table 2.
Table 2: Weight about single index of power quality indices
Indices Subjective weight w
j
Objective weight
j
Synthetic weight a
i
Voltage deviation 0.186 0.119 0.155
Voltage uctuation 0.110 0.107 0.082
Voltage sag 0.152 0.118 0.126
Unbalance 0.074 0.143 0.074
Harmonics 0.151 0.186 0.197
Frequency deviation 0.286 0.142 0.284
Interruption 0.064 0.184 0.082
4.2 Calculation of the Degree of Outranking
Concordance index and disconcordance index of all indices are calculated by using the above
method. Finally the degree of outranking is calculated by formula (10) according to concordance
and discordance. The result is in Table 3.
6 L. Fan et al. / Journal of Information & Computational Science 10: 1 (2013) 17
Table 3: Outranking matrix
a a
1
a
2
a
3
a
4
a
5
a
1
1 0.874 0.859 0.874 0.874
a
2
0.407 1 0.566 0.636 0.802
a
3
0.208 0.452 1 0.777 0.580
a
4
0.229 0.795 0.627 1 0.748
a
5
0.314 0.914 0.734 0.716 1
4.3 Sequence and Decision of Optimal Points
The paper adopts net credibility method to sequence alternatives. The sequence result is calcu-
lated by formula (11) in [13] and is showed in Table 4.
(a
i
) =
+
(a
i
)

(a
i
) =
n

a
k
=1
S(a
i
, a
k
)
n

a
k
=1
S(a
k
, a
i
)
a
i
= 1, 2, , n (11)
where (a
i
) is net reliability, optimal points is decided by (a
i
).
Table 4: Net credibility degree of alternatives
Net reliability a
1
a
2
a
3
a
4
a
5
(a
i
) 2.323 -0.624 -0.769 -0.604 -0.326
The order of quality of points is made as a
1
, a
5
, a
4
, a
2
, a
3
by comparison of net reliability.
5 Conclusion
A PQCE method based on combination weighing and outranking relation is presented in order
to strengthen reliability of power quality comprehensive evaluation in the paper. The weights of
power quality indices are decided by FAHP and entropy methods. It benets to consider subjective
and objective elements overall and avoids information loss for single weight. Meanwhile, complete
judgement course is composed by using ELECTRE III method and sequence is performed by using
net reliability method. The sequence diculty of traditional ELECTRE III is simplied eciently
and calculation eciency and accuracy is improved.
References
[1] H. Tang, J. Peng, Research on synthetic and quanticated appraisal index of power quality based
on fuzzy theory, Power System Technology, 27(12), 2003, 85-88
[2] G. Lei, W. Gu, X. Yuan, Application of gray theory in power quality comprehensive evaluation,
Electric Power Automation Equipment, 29(11), 2009, 62-70
L. Fan et al. / Journal of Information & Computational Science 10: 1 (2013) 17 7
[3] L. Li, J. Yao, L. Long et al. Application of combination weighing method in fuzzy synthetic
evaluation of powerquality, Automation of Electric Power Systems, 31(4), 2007, 56-60
[4] S. Kang, J. Peng, Y. He, Comprehensive evaluation of power quality bas ed on the in tegration of
fuzzy analytic hierarchy process with multi objective decision making, Power System Technology,
33(19), 2009, 113-118
[5] D. Wu, H. Cheng, X. Xi et al, Annual peak power loadforecasting based on fuzzy AHP, Proceedings
of the CSU-EPSA, 19(1), 2007, 55-58
[6] Y. Wu, R. Gao, Interval approach to analysis of hierarchy process, Journal of Tianjin University,
28(5), 1995, 700-705
[7] R. Benayoun, B. Roy, N. Sussman, Manual de reference du programme Electre, Notede Synthese
et Formation, No. 25, Direction Scientique SEMA, Paris, 1966
[8] C. L. Hwang, K. Yoon, Multiple Attribute Decision Making, Methods and Applications, Springer-
Verlag, New York
[9] Z. Zeng, C. Huang, J. Wang, Research of multinational investment of enterprise based on outrank-
ing methods, Science and Management Association of 2004, 2004
[10] P. Nijkamp, A multicriteria analysis for project evaluation, economic-ecological evaluation of res-
idential environment, Papers of the Regional Science Association, 35, 1974, 87-111
[11] B. Roy, The outranking approach and the foundations of the ELECTRE method, Theoryand
Decision, 31, 1991, 49-73
[12] M. Rogers, M. Bruen, A new system for weighting environmental criteria for use within ELECTRE
III, European Journal of Operational Research, 107, 552-563
[13] H. Li, J. Wang, An improved ranking method for ELECTRE III, Wireless Communications, Net-
working and Mobile Computing, 2007, 6653-6656