Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

2016 IEEE 8th International Power Electronics and Motion Control Conference (IPEMC-ECCE Asia)

Economic Evaluation of Harmonic Cost

Zhemin Lin, Xuchang Zhang, Xun Mao, Yong Zhan Xiaotao Peng, Bin Zhang
Dept. of Power System Analysis School of Electrical Engineering
State Grid Anhui Electric Power Company Electric Power Wuhan University
Research Institute Wuhan, China
Hefei, China

Abstract—With the development of electronic technology, an the economic evaluations are mainly affected by two aspects:
increasing number of electronic devices are widely used in probability of occurrence and time of occurrence. Normally,
modern power grid. Accordingly, harmonic distortion and harmonic distortions are treated as a power quality variation,
associated economic losses have become of increasing concern to which is a continuous phenomenon [13]. However, as most
both power suppliers and end users. For the purpose of harmonic devices are designed to operate under a certain level of
regulations and investment analysis, the harmonic cost, mainly harmonic distortions without affecting performance, it is
the potential economic losses due to equipment fail or important to consider this trigger level while evaluating the
malfunction that caused by high levels of harmonic distortions probability of mis-operation. In this case, unlike standard ways,
are discussed in this paper. A Wei-bull distribution is used to
harmonic distortions are considered as a discrete phenomenon
represent the interruption probability of harmonic distortions.
Then, the effect of time of occurrence is quantified by a time
with a trigger level (power quality event) in probability of
varying coefficient, which is derived from daily load curve. occurrence evaluation.
Together, a time dependent harmonic related cost model is In power quality cost calculations, the effects of time of
proposed. As a whole, this cost model provides a flexible and easy occurrence have already been discussed in many literatures
way to estimate the expected value of potential interruption cost [14]-[16]. In [14]-[16], customers are specifically asked to
due to harmonic distortions. estimate the degrees of financial damages in terms of variations
associated with hourly, daily, or monthly values based on their
Keywords—power quality, power system ecomomic, harmoincs.
stated power quality disturbance costs. According to these
variations, the costs for specific time of occurrence are then
I. INTRODUCTION estimated. Whereas, [17]-[20] evaluate the time of occurrence
Nowadays, a growing amount of harmonic distortions are according to a time varying fault rate. This time varying fault
introduced by widely used electronic devices in modern power rate is then used to estimate the long term power quality
systems [1], it becomes of increasing concern to both power disturbance cost. However, the evaluation method in [17]-[20]
suppliers and users. As the economic consequences of is only suitable for cost estimation in terms of years, and the
harmonic distortions might be significant [2], it is necessary to accuracy of evaluation in [14]-[16] mainly depends on
evaluate harmonic cost in detail before any harmonic reduction customer estimations. As the economic consequences of
investments and regulations are made. Generally, the effects of interruptions caused by harmonics are usually results of short
harmonic distortions on system and associated equipment will term distortions in terms of hours rather than years, and the
fall into three aspects: additional energy consumption, values of costs are varying from hours to hours, a short term
premature aging and mis-operation [3]. In most papers [4]-[8], economic calculation model that could truly reflects the
the economic consequences of additional energy consumption variation of time of occurrence is required for the evaluation of
and premature aging have been well studied, and mis-operation interruption cost caused by harmonics.
cost has been normally identified as short interruption cost. In this paper, based on the features of occurrence when
However, when the mis-operation cost due to harmonic harmonic distortions are above the trigger level, a Wei-bull
distortions is evaluated in terms of short interruption cost, it is distribution probability model is proposed and used in
hard to distinguish the losses caused by harmonic distortions probability of occurrence calculation. Then a predefined time
from the ones caused by other power quality issues. This will varying coefficient is used to describe the time dependent
lead to a bias evaluation on harmonic cost, especially for the interruption cost caused by harmonic distortions.
profit/investment balancing analysis of harmonic reduction
measures. Therefore, the economic consequences due to
interruptions caused by harmonic distortions should be II. PROBABLITY OF OCCURRENCE
particularly evaluated.
A. The Trigger Level
Generally, there are many factors should be considered in Normally, the Total Harmonic Distortion (THD) is a
economic evaluations of power quality disturbances, such as common tool to measure the contents and severity of
type of customer, customer consumptions, disturbed duration, harmonics. The IEEE Standard 519 [13] recommends that the
etc. [9]-[12]. For interruptions caused by harmonic distortions, THD design value (the worst case lasting more than one hour)
Sponsored by The Major Program of the National Natural Science
Foundation of China (51190104)

978-1-5090-1210-7/16/$31.00 ©2016 IEEE

978-1-5090-1210-7/16/$31.00 ©2016 IEEE


2016 IEEE 8th International Power Electronics and Motion Control Conference (IPEMC-ECCE Asia)

of bus voltage at PCC for utilities below 69kV is 5%. Wei-bull distribution appears to be the best model to describe
Accordingly, it is believed that any harmonic variations with such a probability problem.
THD of bus voltage at PCC larger than 5% and lasting more
than one hour will probably result in the system and associated In terms of Weibull distribution, the measuring independent
equipment fail or malfunction. In contrast, if THD of bus variable is the duration of harmonic distortions, expressed as t.
voltage at PCC is less than 5% or lasting less than one hour, the Therefore, the cumulative occurrence probability of
system and associated equipment will not have the probability interruptions due to harmonic distortions could be represented
of failure or malfunction. Hence, the trigger level of mis- as [21],
operation due to harmonic could be set as 5% THD of bus
voltage at PCC for utilities below 69kV. β
t
F (t ; α , β ) = p (t ) = 1 − exp[ −   ] (1)
B. The Duration of Harmoinc Distortion a
Due to the fact that some equipment could even survive
from THD of voltage up to 50% within very short period of where F(t; α, β) indicates the probability of interruption
time, such as start-up or unusual conditions [13], rather than within a duration lasting less than or equal to t. And α, as the
the THD of voltage, the duration of harmonic distortion scale parameter, represents the time when the occurrence
becomes the main concerned factor in this probability probability is 1-1/e = 0.632. β, as the shape parameter, suggests
calculation. the range of number of interruptions, the larger β is, the smaller
the range of numbers is [22].
C. The Probablity of Occurrence Accordingly, the expected economic losses of interruptions
The probability of interruption due to harmonic distortion is due to harmonic distortions with duration t could be
evaluated according to historical data. For the purpose of represented as,
calculation and prediction, the raw historical data is further
manipulated in terms of number of harmonic distortions with
THD>5%, number of interruptions and associated durations.    t  β 
And the durations are divided into different time frames with a C ( t ) = Caverage × N average × 1 − exp  −     (2)
time interval. For example, if the historical durations of    α  
harmonic variations vary within 6 hours, and the time interval
is one hour. Then it is possible to have the following time where, Caverage is the average economic losses each
frames, 0-1 hours, 1-2 hours, 2-3 hours, 3-4 hours, 4-5 hours interruption caused by harmonic distortion, Naverage is the
and 5-6 hours. Any duration allocated within a given frame is average number of harmonic distortions with THDV>5%.
represented as the associated maximum duration, as shown in
Table I. Based on (2), the expected economic losses vary with the
probability of interruptions. The more likely the occurrence of
interruptions, the larger economics losses are.
TABLE I. EXAMPLE OF HISTORICAL DATA

No. of Harmonic No. of III. TIME OF OCCURRENCE


Time Distortions with Interruptions due to
Approximate
Frame THD>5% Harmonic A. The relationship between Load Demand and Econmic
Duration (Hour)
(Hour) (NTHD ) Distortions
(Ntrip )
Losses
0-1 1 n1 i1 Taking an industrial customer as example, when load
1-2 2 n2 i2 demands of this customer stay at a relatively high level, it
2-3 3 n3 i3 means that most industrial processes are in operation.
3-4 4 n4 i4 Compared with the normal load condition, an interruption at
4-5 5 n5 i5
this time will impact all the operating processes. In contrast,
5-6 6 n6 i6
during off-peak hours, if anything, only parts of the industrial
With these approximate durations, it is possible to gain the
processes are in operation, accordingly load demand of
information on the historical data as shown in Table I. For
industrial customer drops to a relatively low level. The severity
example, n2 represents the total number of harmonic distortions
of an interruption at this time will certainly be much smaller
with duration lasting between 1 and 2 hours. Accordingly, the
than normal load condition.
number of interruptions due to associated harmonic distortions
is i2. It is easy to conclude that, during peak periods, electricity
consumers, especially large and sensitive consumers, suffer
Unlike most of the regular probability distribution
higher economic losses due to interruptions caused by
calculations, the occurrence of interruptions is quite random
harmonic distortions. However, during off-peak periods, they
among harmonic distortions with THD>5%. It is thus not
will experience much lower economic losses.
always possible to describe this kind of probability distribution
with common distributions. Furthermore, the sample size of
historical data is usually relative small, the highly adaptable
2016 IEEE 8th International Power Electronics and Motion Control Conference (IPEMC-ECCE Asia)

B. The Time Varying Coefficient distortions in the past 5 years is sorted out as Table III, where
Based on the load demand at different time, i.e. the daily only concerned data are needed.
load curve, the time varying coefficient is represented in terms
of load demand, which can be expressed as: TABLE II. TYPICAL DAILY LOAD VARIATION OF INDUSTRIAL
CUSTOMER

Time of Load Demand Time of Load Demand


1 + L − Lα  × 100
  (3) Observation (kW) Observation (kW)
ω
λω =
 Lα  1:00 150 13:00 285
2:00 170 14:00 289
where L is the load demand at observed time point ; La is 3:00 180 15:00 290
the total average load demand, (L - La)/La is the load demand 4:00 185 16:00 280
deviation percentage at observed time point from the average 5:00 190 17:00 275
load demand. And the average load demand can be calculated 6:00 200 18:00 270
thus 7:00 220 19:00 260
8:00 240 20:00 230
j 9:00 250 21:00 220
L ω
10:00 270 22:00 210
ω =1 11:00 275 23:00 200
La = (4)
j 12:00 280 0:00 180

j is the number of observation. Accordingly, the average TABLE III. HISTORICAL DATA OF HARMONIC DISTORTIONS IN THE PAST
5 YEARS
load demand is derived from historical data. It should be noted
that load demands are normally recorded discretely rather than The Number of The Number of
continuously. Hence, the observed load demand varies at an Time Approximate Harmonic Interruptions due to
interval. To make the time varying coefficient more precise, the Frame Duration t Distortions with Harmonic
interval of observation is defined as every one hour. (Hour) (Hour) THD>5% Distortions
(NTHD ) (Ntrip )
Therefore, the time dependent interruption cost due to 0-1 1 3 2
harmonic distortions could be evaluated as 1-2 2 5 0
2-3 3 4 2
Ci = λωti × C ( t ) (5) 3-4 4 5 2
4-5 5 7 4
5-6 6 9 6
In Equation (5), the duration t is varying within one load
6-7 7 9 7
demand observation interval. When the duration t covers more
than one interval, the cost could be calculated as, 7-8 8 19 11
8-9 9 20 14
9-10 10 18 13
γ

= λωt1 × C ( t1 ) + λωt x [C (t x ) − C ( t x −1 )] (6)


10-11 11 17 11
Ct total 11-12 12 15 10
x=2
12-13 13 12 9
13-14 14 13 8
where, γ indicates the number of intervals that the duration
may cover, t1 is the duration in the first load demand 14-15 15 14 7
observation interval; tx is the duration lasting from the 15-16 16 6 5
beginning to the xth observation interval, which is larger than t1; 16-17 17 4 3
λωt1 indicates the first time varying coefficients; λωtx 17-18 18 3 2
18-19 19 4 3
represents the following time varying coefficient in xth
19-20 20 2 2
observation interval, respectively. According to (5) and (6), the
time dependent cost model is capable of evaluating the According to above discussions, the historical durations of
economic losses with various durations. harmonic distortions are represented by an approximate
duration t (the maximum duration) in further calculations. As
IV. NUMBERIC EXAMPLE shown in Table III, the durations of harmonic distortions of this
industrial customer are lasting less than 20 hours in the past 5
A. Historial Data years. And the occurrences distribution of interruptions via
durations could be represented by Fig.1. Due to larger amount
The typical daily load variation of an industrial customer is of harmonic distortions with THD>5%, there are more
shown in Table II, and the historical data of harmonic interruptions when harmonic distortions lasting between 7 to
2016 IEEE 8th International Power Electronics and Motion Control Conference (IPEMC-ECCE Asia)

12 hours, as shown in Fig.1. It should be noted that this doesn’t historical data in Table III, the average number of harmonic
mean harmonic distortions with durations from 7 to 12 hours distortions with THD>5% is 37.8. Together, according to
are more likely to cause interruptions. Equation (2), the expected economic losses due to industrial
process interruptions caused by harmonic distortions are

  t  2.93 
C ( t ) = 1614 × 37.8 × {1 − exp  −    (8)
  12.05  
Accordingly, for any harmonic distortions covering one
load demand observation interval, the expected value of
economic losses due to harmonics could be calculated
according to (8). And the relationship between duration and
harmonic distortion cost could be described as Fig.2.

Fig. 1. Occurrence distribution of interruptions due to harmonic distortions

B. Expected Economic Losses due to Harmonic Distortions


Based on the above data, it is possible to estimate the
interruption probability by the cumulative density function (cdf)
of Weibull distribution. In this paper, the business analytics
software Statistical Analysis System (SAS) is used to estimate
the Weibull distribution parameters. Given the number of
interruptions and associated durations as inputs, the outputs of
estimated Weibull parameters are shown in Table IV, where
the ‘estimate’ represents the estimated value of Weibull scale
parameter α and shape parameter β respectively. The ‘standard
error’ indicates the standard deviation of the estimate, whereas Fig. 2. The relationship between duration and economic losses
the 95% confidence with lower and upper limits indicates the
general range of the estimate [23]. It can be seen that the expected economic losses derived by
(8) increases with duration. For harmonic distortions lasting
less than two hours, the economic losses are relatively small.
TABLE IV. RESULTS OF WEIBULL DISTRIBUTION PARAMETERS
ESTIMATION Thereafter, the longer the durations are, the larger the economic
losses are. And the increment of economic losses becomes
95% Confidence Limits
steady when economic losses approaches maximum value. The
Weibull Standard variation in Fig.2 descripts relative realistic changes of
Estimate
Parameter Error Lower Upper expected economic losses with duration. For short harmonic
Limits Limits distortions, it will be less likely lead to interruptions, and the
α 12.05 0.39 11.30 12.85 expected losses are smaller. With the duration increasing,
β 2.93 0.21 2.54 3.36 customer are suffering more harmonic distortions and more
As shown in Table IV, the estimated Weibull parameters α likely result in interruptions, therefore, the expected losses are
and β are acceptable and reliable. Therefore, according to the increasing. As approaching the 100% probability of
Weibull distribution calculation, the interruption probability of interruptions, the expected economic losses become relatively
harmonic distortions (THD>5%) with duration t could be stable.
expressed as
C. Time Dependent Economic Losses
Based on the typical daily load data in Table II, the average
  t  2.93  load demand is calculated according to (4) as,
p ( t ) = 1 − exp  −    (7)
  12.05  
24

As described by (7), the probability of interruption L ω

occurrence will equal to nearly 100% when harmonic distortion La = ω =1


= 233.29kW
with THD>5% is lasting long enough, which is around 28 24
hours in this case.
As the data of daily load variations are observed every one
Given the average economic losses of interruptions due to
hour, according to (5), the time dependent economic losses due
harmonic distortions are around $1,614, and as derived from
to harmonic distortions within one hour could be expressed
2016 IEEE 8th International Power Electronics and Motion Control Conference (IPEMC-ECCE Asia)

harmonic distortion cost appears larger when load demand is


  ti  2.93 
C ( ti ) = λωti × 1614 × 37.8 × {1 − exp  −    (9) higher, which are peak hours.
  12.05   2) t10 =10 hour
As for duration longer than one hour or cover more than one When the duration goes up to 10 hours, the harmonic
observation interval, the harmonic distortion cost could be distortions cover more than one observation intervals, which is
represented based on (6), 10 here. In this case, γ=10, and the harmonic distortion cost
could be represented as


( )  

2.93

t1
= λω t × 1614 × 37.8 × 1 − exp  −
( )  
Ct
   
2.93
total 1

12.05 1
Ct = λω t × 1614 × 37.8 × 1 − exp  −
 

 (

)    
( ) 
total 1

γ
2.93 2.93
12.05
tx t x −1
 1 − exp  

λ

 ( )   ( ) 


+1614 × 37.8 × − − 1 − exp −
     
2.93 2.93
kt 10
x

12.05 12.05 tx t x −1
+1614 × 37.8 × λ 1 − exp
  
x=2
− − 1 − exp −

kt
x

x=2
12.05 12.05
(10)
Based on (9) and (10), the impacts of time of occurrence And the results are descripted in Fig.4.
are considered respectively. For harmonic distortions within
one hour, the time dependent economic losses could be
calculated according to (9); as for harmonic distortions lasting
longer than one hour, the time dependent economic losses
could be calculated according to (10).
The calculations of time dependent economic losses due to
harmonic distortions with different durations are demonstrated
below. Two scenarios are considered, harmonic distortions
(THD>5%) lasting 1 hour and lasting 10 hours.
1) t1 =1 hour
Assuming the duration of harmonic distortions just fits in
one observation interval, which means that harmonic Fig. 4. Variations of harmonic distortion cost (t=10 hour)
distortions start on each sharp o’clock, and end on the
following ones. Then, harmonic distortion cost could be The Fig.4 shows a different type of harmonic cost variation.
estimated according to (9) as With durations increased, the curve of harmonic distortion cost
variation become smooth and the expected economic losses are
much larger, which increased from around $200 to $25,000.
  1  2.93  The harmonic variation with the highest cost appears around
C ( ti ) = λωti × 1614 × 37.8 × {1 − exp  −    7am in the morning and lasts the whole peak hours, while the
  12.05   harmonic variation with the lowest cost appears around 8pm in
the evening and lasts the whole off-peak hours. For durations
And the variation of cost is descripted as Fig.3. covering more than one observation intervals, the harmonic
distortions during peak hours will result in more economic
losses. In contrast, the harmonic distortions during off-peak
hours will cause less economic losses.
Based on the results of numeric examples, this proposed
time dependent harmonic distortion cost model not only reveals
the effect of duration, but also indicates the effect of time of
occurrence.

V. CONCLUSIONS
In this paper, a time dependent cost model is proposed
based on Wei-bull probability function and the time varying
coefficient. This cost model is used to describe the characters
Fig. 3. Variations of harmonic distortion cost (t=1 hour) of interruption cost due to harmonic distortions. According to
the calculation of interruption probability due to harmonic
As shown in Fig.3, though not exactly the same, the variations with THD>5%, the effects of duration could be
harmonic distortion cost describes a similar variation as load considered in Wei-bull distribution dependent cost model.
demands. The harmonic distortion cost appears smaller when Then based on the relationship between load demand and
load demand is lower, which are off-peak hours; and the
2016 IEEE 8th International Power Electronics and Motion Control Conference (IPEMC-ECCE Asia)

economic losses, a time varying coefficient is proposed to [8] G. Carpinelli; P. Caramia; E. Di Vito; A. Losi and P. Verde,
indicate the effects of time of occurrence in harmonic cost "Probabilistic evaluation of the economical damage due to harmonic
losses in industrial energy system," Power Delivery, IEEE Transactions
calculation. on , vol.11, no.2, pp.1021-1031, Apr 1996.
According to a numeric example, the accuracy and [9] M. Delfanti; E. Fumagalli; P. Garrone; L. Grilli and Luca Lo Schiavo,
practicability of this model is proved. The results indicate "Toward Voltage-Quality Regulation in Italy," Power Delivery, IEEE
Transactions on , vol.25, no.2, pp.1124,1132, April 2010.
relatively realistic and continuous variations of expected
[10] R. Targosz, and J. Manson, "Pan-European power quality survey,"
harmonic distortion cost. The expected interruption costs are Electrical Power Quality and Utilisation, 2007. EPQU 2007. 9th
increasing with the duration, and harmonic distortions during International Conference on , vol., no., pp.1,6, 9-11 Oct. 2007.
peak hours will result in more economic losses. In additional, [11] L.Lawton; M.sullivan, K.V.Liere and A.Katz, "A Framework and
the time dependent cost possesses the similar variation as the Review of Customer Outage Costs: Integration and Analysis of Electric
real time load demand, which reflects the differences in time of Utility Outage Cost Surveys, " ERNEST ORLANDO LAWRENCE
occurrence. Without considering the effect of time of BERKELEY NATIONAL LABORATORY, November 2003.
occurrence, the harmonic distortion cost would be a constant [12] "The Cost of Power Disturbances to Industrial & Digital Economy
Companies", Report by Primen, EPRI. 2001.
value for a given duration, which is impossible in real world.
[13] "IEEE Recommended Practices and Requirements for Harmonic Control
This cost model provides a time dependent harmonic distortion in Electrical Power Systems," IEEE Std 519-1992, vol., no., pp.1,112,
cost with different time of occurrence and duration. It reflects April 9 1993.
both impacts of the probability of occurrence and time of [14] R. Billinton and W. Wangdee, "Estimating customer outage costs due to
occurrence, and provides a useful evaluation tool for harmonic a specific failure event," Generation, Transmission and Distribution, IEE
regulations and investments. Proceedings- , vol.150, no.6, pp. 668- 672, 12 Nov. 2003.
[15] R. Billinton and W. Wangdee, "Approximate methods for event-based
customer interruption cost evaluation," Power Systems, IEEE
REFERENCES Transactions on , vol.20, no.2, pp. 1103- 1110, May 2005.
[1] I.M. Nejdawi; A.E. Emanuel; D.J. Pileggi; M. J. Corridori and R. D. [16] G.H. Kjolle; Knut Samdal; B. Singh and O.A. Kvitastein, "Customer
Archambeault, "Harmonics trend in NE USA: a preliminary survey," Costs Related to Interruptions and Voltage Problems: Methodology and
Power Delivery, IEEE Transactions on , vol.14, no.4, pp.1488,1494, Oct Results," Power Systems, IEEE Transactions on , vol.23, no.3,
1999. pp.1030,1038, Aug. 2008.
[2] Gutierrez Iglesias J. et al. "Power quality in European electricity supply [17] G. H. Kjolle and A. T. Holen. "Reliability and interruption cost
network", NE Stan-dardisation Specialist Group, N-E EMC & prediction using time-dependent failure rates and interruption costs,"
Harmonics, 2002. Quality and Reliability Engineering International, 14:159–165, 1998.
[3] Angelo Baggini, Handbook of Power Quality, [Xiao xiangning, [18] Peng Wang and R. Billinton, "Time sequential distribution system
Taoshun, Xu yonghai Translation], Beijing, China Electric Power reliability worth analysis considering time varying load and cost
Press.2010:379-380. models," Power Delivery, IEEE Transactions on , vol.14, no.3, pp.1046-
[4] E.J. Davis; A.E. Emanuel and David J. Pileggi, "Evaluation of single- 1051, Jul 1999.
point measurements method for harmonic pollution cost allocation," [19] M. Al-Muhaini and G.T. Heydt, "Evaluating Future Power Distribution
Power Delivery, IEEE Transactions on , vol.15, no.1, pp.14,18, Jan System Reliability Including Distributed Generation," Power Delivery,
2000. IEEE Transactions on , vol.28, no.4, pp.2264,2272, Oct. 2013.
[5] A.E. Emanuel; Yang Minghao and David J. Pileggi, "The engineering [20] Peng Wang and R. Billinton, , "Reliability cost/worth assessment of
economics of power systems harmonics in sub-distribution feeders: a distribution systems incorporating time-varying weather conditions and
preliminary study," Power Systems, IEEE Transactions on , vol.6, no.3, restoration resources," Power Delivery, IEEE Transactions on , vol.17,
pp.1092,1098, Aug 1991. no.1, pp.260,265, Jan 2002.
[6] F.C. Pereira; O.C.N. Souto; J.C. De Oliveira; A.L.A. Vilaca and P.F. [21] H. Rinne, The Weibull Distribution: A Handbook, Taylor & Francis
Ribeiro, "An analysis of costs related to the loss of power quality," Group, LLC, 2009,pp. 27-91.
Harmonics and Quality of Power Proceedings, 1998. Proceedings. 8th
[22] "IEEE Guide for the Statistical Analysis of Electrical Insulation
International Conference On , vol.2, no., pp.777-782 vol.2, 14-18 Oct
Breakdown Data," IEEE Std 930-2004 (Revision of IEEE Std 930-
1998.
1987), vol., no., pp.0_1-41, 2005.
[7] Yong Peng; Shun Tao; Qun Xu and Xiangning Xiao; , "Harmonic
[23] William H. Press, Numerical Recipes in Fortran 77: The Art of Scientific
pricing model based on harmonic costs and harmonic current excessive
Computing, 2nd Edition, Cambridge University Press, 25 Sep 1992. pp.
penalty," Artificial Intelligence, Management Science and Electronic
684-694.
Commerce (AIMSEC), 2011 2nd International Conference on , vol., no.,
pp.4011-4014, 8-10 Aug. 2011.

You might also like