Académique Documents
Professionnel Documents
Culture Documents
Research Article
ISSN 1751-8687
Received on 5th August 2014
Revised on 6th January 2015
Accepted on 1st February 2015
doi: 10.1049/iet-gtd.2014.0757
www.ietdl.org
Abstract: This study proposes a method for designing advanced power distribution system (PDS) including distributed
generations, using a combination of fundamental loop generator and multi-objective seeker-optimisation algorithm
(MOSOA). The proposed approach reduces the searching space using fundamental loop generator technique to obtain
initial feasible solutions which is further improved by SOA to generate new set of solutions with improved aptitude.
The proposed methodology uses a contingency-load-loss-index for reliability evaluation, which is independent of the
estimation of failure rate and fault repair duration of feeder branches. This planning strategy includes distribution
automation devices such as automatic reclosers (RAs) to enhance the reliability of PDS. The proposed algorithm
generates a set of non-dominated solution by simultaneous optimisation of two conflicting objectives (economic cost
and system reliability) using Pareto-optimality-based trade-off analysis including a fuzzy-operation to automatically
select the most suitable solution over the Pareto-front. The performance of the proposed approach is assessed and
illustrated on 54-bus and 100-bus PDS, considering realtime design practices. Extensive comparisons are made against
some well-known and efficient MO algorithms such as fast non-dominated sorting genetic algorithm-II, MO particleswarm-optimisation and MO immunised-particleswarm-optimisation. Simulation results show that the proposed
approach is accurate and efficient, and a potential candidate for large-scale PDS planning.
List of acronyms:
SP
PQ
SAIFI
SAIDI
CAIDI
ASAI
ASUI
MAIFI
CAIFI
NDL
LP
NOL
Spacing
Active and reactive power.
System average interruption frequency index.
System average interruption duration index.
Customer average interruption duration index.
Average Service Availability Index.
Average Service Unavailability Index.
Momentary Average Interruption Frequency Index.
Customer Average Interruption Frequency Index.
Non-delivered load.
Load point.
Number of lines.
Introduction
Problem formulation
Objective function 1
The overall annual cost (u1) representing the yearly investment costs
plus the yearly operational (variable) costs of the network, plus the
cost of maintenance per year including the cost of the energy
losses in the circuit branches, subject to equality as well as
inequality constraints, represents one of the main concerns in
planning the APDS. The cost minimisation problem can be
formulated as a mixed integer non-linear programming model, as
described in (1) [9]. The main innovation is integrating the
planning of DGs, RAs and cost associated with energy losses in
the planning model. Thus, the objective function u1 is dened as
u1 = CSystem = CFixed + CVariable + CMaintenance
(1)
where
CFixed =
[(CkF )b (Yk )b ] +
[(CDF )g (YD )g ]
D[ND g[Ng
k[NS b[Nb
[(Ci,F j )a (Yi, j )a ]
[(Ci,F j )m (Yi, j )m ]
[[(Ci,F j )E ]a (Yi, j )a ]
[[(CkF )E ]b (Yk )b ]
k[NSE b[Nb
CVariable =
D[ND g[Ng
k[NS b[Nb
k[NSE b[Nb
CMaintenance =
constant
residential
commercial
industrial
0
0.92
1.51
0.18
0
4.04
3.4
6
D[ND g[N g
k[NS b[Nb
k[NSE b[Nb
(2)
Qi (V ) = Qoi Vig
(3)
where Pi and Qi are the load active and reactive power consumptions
at bus i; Vi is the terminal voltage magnitude (per unit (pu)) at bus i;
Poi and Qoi are the active and reactive power operating points at bus i.
and are the real and reactive power exponents. Table 1 shows the
real and reactive power exponents used in the present work for
constant power, industrial, commercial and residential type loads
[25]. In practical cases, loads are not explicitly constant power,
industrial, commercial and residential; rather load class mix may
be seen by PDS depending on the nature of area being supplied.
Therefore a load class mix of constant power, industrial,
commercial and residential loads are also adopted in this paper.
2.2
Objective function 2
NDLAvg.
=
LTotal
NF
NDLi /NF
LTotal
i=1
(4)
and CAIFI etc.) [19] are calculated as the weighted averages of the
basic load point indices (failure rate and fault repair duration).
CLLI is calculated considering all single contingency events (all
branch failures, taken one at a time). However, robust operation of
a distribution system requires anticipation of unplanned multiple
contingencies that could lead to dramatic and costly blackouts.
They can provide planners and operators more condence in the
security of the system beyond N1 requirements [26], where N is
the total number of branches in the system. Furthermore, there is a
possibility of purposeful and malicious N k scenarios (where k is
the simultaneous fault of k branches); those have severe
consequences in the planning issue [26]. The occurrences of
simultaneous contingencies, such as simultaneous faults in two or
more branches are not considered in this planning work as the
frequency of occurrence of such faults are very small for any
practical distribution system.
Two cases have been considered for the computation of CLLI.
Similarly, for two feeder radial conguration with two RAs and
one DG at bus 3 as shown in Fig. 2 (TS being open), the CLLI
will be 0.33, with an improvement of 12% (0.375: without DG
and 0.33: with DG) in comparison with, without DG. Thus, this
network is more reliable than the single feeder radial network.
Now, if TS is closed and DG is present at bus 3, it forms a loop
conguration and it causes the most reliable network with CLLI
index of 0.23. Flowchart for calculation of CLLI index for the
aforementioned two cases is shown in Fig. 3.
Thus, the loop network shows higher reliability as compared with
other congurations. Power utilities such as Hong Kong Electric
Company, Singapore power etc. have adopted closed-loop
congurations to serve their consumers with high reliability.
However, the loop conguration possesses high cost as compared
with the radial congurations because of more number of circuit
branches, and more CBs or switchgears for higher short-circuit
level. Thus, from several past decades the PDS conguration has
been typically designed in radial form for easy access and control,
lower protection cost and operational convenience [15]. The MO
problem can be formulated as
Min f = [CSystem , CLLI]
(6)
max
Qmin
SS, i QSS, i QSS, i
(8)
max
min
max
min
In these inequalities, PSS,
i , PSS, i , QSS, i and QSS, i are the maximum
and minimum amount of active and reactive powers that can be
generated by the ith substation.
(c) DG capacity limits
min
max
PDG,
i PDG, i PDG, i
(9)
max
Qmin
DG, i QDG, i QDG, i
(10)
max
min
max
min
In these inequalities, PDG,
i , PDG, i , QDG, i and QDG, i are the maximum
and minimum amount of active and reactive powers that can be
generated by the ith DG unit.
(d) Bus voltage magnitude limits: The voltage magnitude limits in a
distribution system must be within the specied range
n [ N
(11)
N
PD (i) +
i=2
QSS =
N
N
OL
(12)
(13)
n=1
QD (i) +
i=2
NOL
n=1
where PSS and QSS are the total active and reactive power
generations from the substation of the distribution system, PD(i)
and QD(i) are the total active and reactive power demands at bus i,
NOL is the number of lines, PLoss(n) and QLoss(n) are the net
active and reactive power losses of line n connected between two
buses in a distribution system, PDG,i and QDG,i are the net active
and reactive powers ow from the ith DG.
(f) Network connectivity and radiality constraint of the distribution
system
NB = NN NSS
(14)
Basics of SOA
min
max
PSS,
i PSS, i PSS, i
f (X ) =
PSS =
(5)
V min Vn V max ;
NEq
1. Dene the input data: Initialise the good acceptable initial set of
seekers of size NIS using fundamental loop generator that
preserves the radiality operation, where each seeker represents one
radial conguration of dimension P, each dimension represents
one load bus. For each seeker, we have optimised RA positions
using binary coded GA (BCGA) (BCGA parameters are cross-over
rate of 0.8, mutation rate of 0.02 with a population size of 100 and
the evaluation was performed for 50 independent runs where each
run consists of 100 generations). Instead of placing the RAs
randomly in the network, we have optimised its position which
enables acceptable set of solutions, keeping all the system
constraints into consideration. At each step of the algorithm, it is
required that the set of system constraints (6)(14) should be
satised. To handle these constraints [27], convert the constraints
of MO problem to an unconstrained one using (15). In this paper,
in order to handle the constraints penalty factor method is used
(see (15))
where u1(X ) and u2(X ) are the objective functions related to cost and
reliability function as shown in (1) and (4), respectively. sk(X ) and
vk(X ) are the equality and inequality constraints, respectively. NEq
and NIEq are the number of equality and inequality constraints,
respectively. l1 and l2 are the penalty factors, respectively. The
values of the penalty factors are considered to be 5000. At rst,
backwardforward load ow method is used to evaluate the
objective function values (1) and (4) for each seeker, where each
NIEq
(sk (X )) + l2
(Max[0, vk (X )])
u1 (X ) + l1
k=1
k=1
f1 (X )
=
N
N
f2 (X )
Eq
IEq
(sk (X ))2 + l2
(Max[0, vk (X )])2
u2 (X ) + l1
k=1
(15)
k=1
i = 1, 2, . . . , N ND
(16)
Actual search direction of each individual, i, Si (t) is based on a
compromise among self-centric or egotistic behaviour,
group-centric behaviour and pro-activeness behaviour, whereas the
step length for an individual i at any dimension j is dened as
aij = sj log (rand(mi , 1)
(17)
sj = w (Xbest Xrand )
(18)
sj = w (Xbest Xworst )
where Xworst, the worst value of the objective function in the same
subpopulation to which the jth seeker belongs. Addition of global
search component to the overall search direction reduces the
chances of the seekers being fall at local optimum, and the
addition of inter-sub-population learning mechanism is more
benecial to remove the worst seeker positions. According to
inter-sub-population learning scheme, if there are M
subpopulations, the M1 worst seekers in each subpopulation are
replaced by the best positions of the remaining M1 subpopulation.
3.2
3.3
(19)
any other closed loop. For Fig. 5, the closed loops selected are
Loop 1 = [L1 L3 L5 ]; Loop 2 = [L3 L4 L6 L7 ];
Loop 3 = [L2 L4 L7 L8 ]
To create a radial topology, one should select from the group of
fundamental loop vector elements to be disconnected (one for
loop). It is important that the vectors elements are not repeated in
a selected topology. The combination of elements can be
generated with these vectors, creating all possible radial topologies
of the system, which satisfy the radiality constraints, for Fig. 5
(condition 1: the solution must have ve (total number of load
buses) number of branches. Condition 2: the solution must be
connected; that means the solution must satisfy the conservation of
power ow constraints (it must supply the power demand at every
load bus, so that a path between the substation and each other bus
exists)). Therefore every bus is linked with the substation bus,
forming a connected graph.
For the system shown in Fig. 5, the total number of feasible radial
topologies, using the proposed method, is 30, whereas 56 topologies
were observed using a random methodology. This is one of the great
advantage of the proposed method. These considerations allow the
proposed SOA to limit the generation of non-feasible individuals.
This also reduces the combinatorial searching space.
One topological presentation is shown in Fig. 6, which satises
the radiality constraint. One radial topology represents one seeker.
For a 6-bus system with one source, the dimension of each seeker
is 5. The basic SOA structure for formation of new set of seekers
(new feasible radial topologies) is shown in Fig. 7, which forms a
radial conguration consisting of ve branches. From, Fig. 7, start
and end buses represents the line start and end bus numbers. Now,
the search direction and step length for each dimension of a seeker
is evaluated, as dened in Section 3.1 (step 2). Now, the position
of each dimension of a seeker is updated according to (16), which
forms one new radial topology, satises the radiality constraint.
Finally, the backward/forward load ow programme is carried out
to evaluate the tness function (5).
3.4
i=1
(20)
mkfi
1,
max
fi fi
m fi =
,
fimax fimin
0,
fi fimin
fimin , fi , fimax
fi fimax
4.1
The NSGA was developed by Srinivas and Deb [6]. It has certain
drawbacks such as computational complexity, lack of elitism and
choosing the optimal parameter value for sharing parameter. These
drawbacks have been rectied in a modied version, NSGA-II,
which has a better sorting algorithm, incorporates elitism and no
sharing parameter needs to be chosen such as NSGA. Once the
populations are initialised, each population represents one radial
topology, then based on BCGA, position of RAs is placed
optimally to each radial topology. BCGA parameters are same as
used for MOSOA. After the evaluation of objective functions, the
initialised populations are sorted based on non-domination into
each front. After this the usual selection, cross-over and mutation
operations are performed.
In the ranking procedure, the non-dominated individuals are
identied rst from the current set of population which is called
the rst-front population. Thus, the rst-front solution is a
complete non-dominated set of solution and the second-front
individuals are only dominated by the rst-front individuals and
the front goes so on. Accordingly all the fronts are dened. Each
individual in each front are assigned rank values (e.g. individuals
in the rst (second) front are assigned rank 1(2)). After assignment
of rank, crowding distance is calculated for each individual. Now,
only either all the rst front solutions are selected or individuals
are selected using tournament selection based on rank and
crowding distance (An individual is selected if the rank is lesser
than the other or if the crowding distance is greater than the other.)
Now, the selected population generates offspring population from
crossover and mutation operators. Now, the population with the
current population and the current offspring populations is sorted
based on non-domination and only the best N individuals are
selected (elitism is ensured), where N is the population size.
Continue the same process till the stopping criterion is reached
(either a maximum number of iterations is reached or the objective
function does not improve).
(21)
4.2
MO immunised particle-swarm-optimisation
Step 8: Then, the fuzzy theory approach is applied to select the most
suitable MO planning solution from the obtained Pareto-front as
dened in Section 3.5.
Table 6
Parameters
Value
X
318 000
29
7
Size,
kcmil
R,
/mi
X,
/mi
B.I.C.,
$/km
P.M.C.,
$/km/
year
C.M.C.,
$/km/
year
Current
rating, A
6
4
2
1/0
3/0
4/0
4.257
2.719
1.753
1.145
0.762
0.628
0.76
0.72
0.67
0.61
0.58
0.56
10 000
10 000
10 000
10 000
10 000
10 000
53.5
53.5
53.5
53.5
53.5
53.5
6.5
6.5
6.5
6.5
6.5
6.5
105
140
184
242
315
357
Turkey
Swan
Sparrow
Raven
Pigeon
Penguin
0(1)
1(2)
2(3)
3
4
5
6
7
1.5
11.5
7.5
8.5
12.5
11
8
11
5.5
3.5
Load,
kVA
Load,
kVA
Load,
kVA
0
2
15
4
12
11.5
10
7
5.5
13.5
17.5
15.5
10.5
17.5
7.5
6
5.5
8.5
0
25
25
25
50
63
63
50
25
16
16
25
50
63
63
25
16
16
13
14
16.5
5.5
20.5
8
5
8
10.5
10.5
9
7.5
5.5
3
13
14
12.5
11
(36)
8
13
14
17
12
9
7
5.5
8
15
19
19.5
19.5
17.5
15.5
16.5
19
20
16
63
25
25
50
100
100
100
50
50
25
63
63
25
50
50
25
25
5
2
3
6
9
14
15
15
15.5
12
14.5
13.5
13
13.5
4
9.5
9.5(53)
12(54)
15.5
10.5
3.5
4
4.5
11.5
10
14.5
12.5
12
7.5
6
4.5
18
5
6.5
17
2.5
50
50
63
25
25
50
50
25
25
63
63
25
16
16
25
16
25
50
MOSOA
NSGA-II
MOPSO
MOIPSO
100
(0.95, 0.0111)
(0.1, 0.9)
100
50
100
0.8
0.02
100
50
100
(0.4, 0.9)
(2, 2)
100
50
100
(0.4, 0.9)
0.1
(2, 2)
100
50
Table 5 Specification of CB
CB type
1
2
3
Installation cost, $
25
35
50
20 000
40 000
60 000
10
Fig. 11 Possible paths between source and load points for a 54-bus system
5.2
11
12
76
1511
4118
3246
2529
2453
1753
2531
1625
1544
2128
4840
1618
3841
1439
1925
3439
4136
1731
2136
4916
67
3228
2622
5012
3228
2848
2438
3027
5015
3546
2728
1931
1831
3145
2416
1923
3228
1524
239
2139
1531
2233
1118
305
2732
1245
2252
1821
4143
3639
2327
3942
3334
1443
1618
5021
2147
4129
3450
4546
1537
3645
2942
1137
1653
1643
1441
1741
3036
2115
2427
3247
3916
3450
1147
1837
1343
4849
2433
3549
1841
2433
3212
2843
4919
3237
3138
4239
3435
3332
4329
1946
3152
2425
3514
1219
2033
2325
2953
2244
1632
2721
2244
2043
1842
1736
1822
1132
2935
1218
3219
2329
4325
2853
2244
2045
(1)24-15
(2)1233
1145
2133
3249
2936
1142
2853
2841
2233
1338
4233
1316
3652
1632
2617
2832
5148
1451
2624
3024
818
1413
2529
2642
1013
1513
3934
3129
1844
1742
1948
2928
2844
2935
3024
1927
1213
1751
1028
2654
3032
3034
3848
4250
4639
5213
1454
2327
3452
4549
4352
1940
1215
1333
3546
3642
2215
2312
2916
1435
4146
1751
1944
4146
4049
2046
1252
2143
1433
2247
1648
5215
3039
1916
1335
4146
1122
3536
2338
2833
1140
2134
4640
3349
4254
3139
2541
1811
1311
5339
5153
1119
3234
4623
3739
3049
2735
3235
3536
1349
2946
1345
1349
2250
1546
2249
1046
2551
1525
4853
1649
2246
4235
4546
1349
3550
0.553
0.423
0.489
0.479
0.641
0.329
0.444
0.489
0.223
0.593
0.389
0.488
0.432
0.477
0.418
0.489
0.679
0.249
0.327
0.344
0.589
0.535
0.498
0.398
0.428
0.492
0.456
0.481
0.493
0.497
0.649
0.399
0.379
0.423
0.529
0.476
0.328
0.398
0.512
0.627
0.317
0.717
0.577
0.419
0.624
0.399
0.369
0.427
0.397
0.423
0.439
0.576
0.419
0.389
0.456
0.443
0.512
0.623
0.386
0.429
0.438
0.554
0.512
0.476
0.399
0.445
0.398
0.541
0.511
0.499
0.523
0.543
0.556
0.396
0.412
0.643
0.366
0.192
DG sizes, MW
0.364
0.541
0.262
0.239
0.289
0.318
0.413
0.517
0.621
0.289
0.419
0.399
0.512
0.412
0.419
0.643
0.329
0.781
0.431
0.541
0.398
0.456
0.651
0.341
0.239
0.416
0.395
0.431
0.445
0.476
0.319
0.422
0.362
0.369
0.328
0.378
0.319
0.571
0.311
1.544
1.281
1.468
1.295
1.349
1.271
1.256
1.375
1.271
1.279
1.231
1.326
1.520
1.308
1.226
1.588
1.451
1.542
1.381
1.271
1.416
1.429
1.703
1.251
1.143
1.307
1.296
1.310
1.479
1.484
1.467
1.344
1.284
1.348
1.253
1.266
1.290
1.335
1.015
Sum (DG
sizes, MW)
20
24
4
8
18
6
12
11
5
13
17
32
12
16
23
20
12
11
7
8
12
2
18
29
30
6
15
42
21
13
27
18
32
41
18
21
31
20
23
23
44
32
42
31
20
19
18
34
48
51
42
34
23
52
28
28
39
21
3
51
14
32
5
42
18
31
12
52
43
29
28
39
49
32
42
6
16
24
DG locations
41
31
42
21
52
52
41
51
52
21
20
50
50
52
31
50
52
42
43
50
29
52
52
51
52
42
48
8
16
26
52
49
49
13
49
51
52
51
32
80.03
88.54
92.14
94.96
96.89
108.96
102.78
112.34
145.49
103.45
101.32
100.98
104.54
108.49
109.21
111.21
102.56
111.34
123.32
119.54
110.31
100.80
102.39
112.32
119.12
101.23
100.34
132.23
103.45
101.36
121.36
122.35
121.38
101.49
100.23
100.89
112.34
114.45
102.80
Losses, kW
2 851 128
2 883 412
2 899 834
2 912 769
2 920 342
2 928 763
2 930 245
2 936 523
2 939 876
2 942 784
2 949 385
2 957 893
2 965 487
2 969 879
2 971 398
2 979 934
2 989 986
3 042 664
3 086 594
3 095 487
3 099 789
3 109 432
3 132 889
3 149 989
3 152 134
3 152 786
3 161 245
3 170 234
3 178 965
3 189 567
3 198 345
3 209 456
3 234 217
3 249 789
3 269 838
3 277 985
3 281 425
3 293 452
3 301 945
Ck
0.3311
0.3241
0.3118
0.2912
0.2768
0.2712
0.2601
0.2513
0.2281
0.2116
0.2054
0.1914
0.1723
0.1534
0.1402
0.1304
0.1264
0.1186
0.1154
0.1142
0.1135
0.1054
0.1052
0.0989
0.0969
0.0945
0.0912
0.0889
0.0832
0.0739
0.0699
0.0645
0.0601
0.0541
0.0533
0.0532
0.0523
0.0512
0.0501
CLLIk
single
single
single
single
single
two
two
two
two
two
two
two
two
two
two
two
two
two
two
two
two
two
two
two
three
three
three
three
three
three
three
three
three
three
three
three
three
three
three
Feeder
configuration
Fig. 12 Pareto-fronts of MOSOA, NSGA-II, MOPSO and MOIPSO for 54-bus system
13
14
suitable
solutions
with
different
Parameters
MOSOA
NSGA-II
MOPSO
MOIPSO
cost, $
CLLI
maximum voltage
deviation, pu
worst voltage, pu
active power loss, kW
reactive power loss, kVAr
3 109 432
0.1054
0.0015
3 150 314
0.1202
0.0029
3 146 874
0.144
0.003
3 115 768
0.112
0.0018
0.9985
0.5102
0.4025
0.9971
0.9163
0.5862
0.997
1.1109
0.7214
0.9982
0.8963
0.5218
The obtained Pareto-front for 54- and 100-bus systems (Figs. 12 and
20, respectively) clearly shows that MOSOA dominates the other
techniques, for both the systems which show the superiority of the
proposed algorithm over others.
6.1.1 Convergence metric: Using C-metric, the quality of the
two given sets of non-dominated solutions is compared. For two
given Pareto-sets A and B, the function Ic maps the ordered pair
(A, B) to the interval [0, 1] as follows
Ic(A, B) =
size({b [ B,
a [ A:ab})
|B|
15
Most suitable
Cost($)
MOSOA
NSGA-II
MOPSO
MOIPSO
2.780 106
2.863 106
3.213 106
2.821 106
Most reliable
CLLI
Cost($)
CLLI
0.1321
0.1426
0.1600
0.1389
2.898 106
2.996 106
3.386 106
2.931 106
0.0821
0.1311
0.1376
0.1231
Fig. 18 Voltage prole for 54-bus system with and without deploying DGs using different techniques
16
Fig. 19 Possible paths between source and load points for a 100-bus system
Fig. 20 Pareto-fronts of MOSOA, NSGA-II, MOPSO and MOIPSO for 100-bus system
17
i=1
where Xi Yi is the Euclidean distance between the
non-dominated solutions Xi and Yi . The obtained results for
D-metric are depicted in Tables 17 and 18. Results clearly show
that NSGA-II, MOPSO and MOIPSO have more diversity than
MOSOA in the obtained Pareto-fronts for both the test systems.
The results are obvious as the MOSOA dominates the NSGA-II,
MOPSO and MOIPSO (shown in Figs. 12 and 20) and all
18
suitable
solution
with
different
Parameters
MOSOA
NSGA-II
MOPSO
MOIPSO
cost, $
CLLI
maximum voltage
deviation, pu
worst voltage, pu
active power loss, kW
reactive power loss, kVAr
6 134 523
0.121
0.0012
6 146 342
0.1525
0.002
6 149 285
0.1725
0.0024
6 139 828
0.1401
0.0019
0.9988
1.1985
0.8965
0.998
1.9404
1.1762
0.9976
2.2021
1.6825
0.9981
1.2242
0.9826
MOSOA
NSGA-II
MOPSO
MOIPSO
Most suitable
Most reliable
Cost, $
CLLI
Cost, $
CLLI
5.81 106
5.94 106
6.02 106
5.92 106
0.21
0.23
0.26
0.22
6.01 106
6.10 106
6.386 106
6.096 106
0.12
0.16
0.1676
0.1411
Fig. 24 Voltage prole for 54-bus system with and without deploying DGs using different techniques
19
54-bus
100-bus
MOSOA
NSGA-II
MOPSO
MOIPSO
Shortest
time, s
Longest
time, s
Average
time, s
Shortest
time, s
Longest
time, s
Average
time, s
Shortest
time, s
Longest
time, s
Average
time, s
Shortest
time, s
Longest
time, s
Average
time, s
4900.65
4996.43
5980.56
5998.89
5400.5
5434.6
7649.8
7946.9
8200.8
8289.2
7800.3
7864.8
5682.4
5996.2
7272.8
8184.6
6141.3
6894.5
5848.8
6148.1
7301.2
8489.2
6468.3
6964.2
Fig. 25 Box-plot for average C-metric values for 54-bus (test system 1) and 100-bus (test system 2)
MOSOA
mean
std
t-test difference
0.59
1.8 102
NSGA-II
MOIPSO
0.29
0.25
0.51
3.6 102
2.6 102
2.1 102
3
2.84 10 extremely significant
MOSOA
mean
std
t-test difference
0.69
3.4 102
NSGA-II
MOSOA
NSGA-II
MOPSO
mean
std
t-test difference
0.029
2 102
0.056
0.066
3 102
3.2 102
0.0452 non-significant
MOIPSO
0.045
2.8 102
MOIPSO
0.28
0.31
0.49
9.9 102
5.9 102
4.8 102
3
3.18 10 extremely significant
SP-metric
Conclusion
SP-metric
MOSOA
mean
std
t-test difference
0.043
6.718 102
NSGA-II
MOPSO
0.076
0.081
8.834 102
8.52 102
0.0086 non-significant
MOIPSO
0.059
7.4 102
Fig. 26 Box-plot for SP-metric values for 54-bus (test system 1) and 100-bus (test system 2)
20
MOSOA
NSGA-II
MOPSO
mean
std
t-test difference
0.43
3 102
0.59
0.61
5 102
4.3 102
0.63 significant
MOIPSO
0.49
3.9 102
MOSOA
NSGA-II
MOPSO
mean
std
t-test difference
0.513
5.7 101
0.725
0.764
8.2 101
7.8 102
0.498 significant
MOIPSO
0.631
6.2 102
method takes into account the total annual cost including xed and
variable costs of DGs and substations, energy loss cost and the
reliability function (CLLI). Extensive testing is carried out in order
to design a real PDS including distribution automation devices and
DNs are generated based on trade-off between cost and reliability.
Moreover, a fuzzy theory approach is employed to automatically
select the most suitable solution over the Pareto-front. Extensive
performance comparisons are made with efcient MO algorithms
such as NSGA-II, MOPSO and MOIPSO for 54-bus and 100-bus
PDSs. The comparison is made qualitatively by plotting the
Pareto-front, and considering comparison metrices such as
C-metric, the SP-metric and the D-metric. The Pareto-front and the
three performance metrices results clearly indicate that MOSOA
has superior performance in both convergence and uniform
diversity compared with other MOEAs. Moreover, MOSOA is
computationally less intensive than the others. Nevertheless, the
method presented in this paper is highly effective and helpful to
system planners for obtaining typical designs of an APDS.
Acknowledgment
References
1 Samui, A., Singh, S., Ghose, T., Samantaray, S.R.: A direct approach to optimal
feeder routing for radial distribution system, IEEE Trans. Power Deliv., 2012,
27, (1), pp. 253260
2 El-Kady, M.A.: Computer-aided planning of distribution substation and primary
feeders, IEEE Trans. Power Appl. Syst., 1984, PAS-103, (6), pp. 11831189
3 Ranjan, R., Venkatesh, B., Das, D.: A new algorithm for power distribution system
planning, Electr. Power Syst. Res., 2002, 62, (1), pp. 5565
4 Kumar, D., Samantaray, S.R., Kamwa, I.: A radial path building algorithm for
optimal feeder planning of primary distribution networks considering reliability
assessment, Electr. Power Compon. Syst., 2014, 42, (8), pp. 861877
5 Ramrez-Rosado, I.J., Bernal-Agustn, J.L.: GA applied to the design of large
power distribution systems, IEEE Trans. Power Syst., 1998, 13, (2), pp. 696703
6 Ramrez-Rosado, I.J., Bernal-Agustn, J.L.: New multi-objective Tabu search
algorithm for fuzzy optimal planning of power distribution systems, IEEE
Trans. Power Syst., 2006, 21, (1), pp. 224233
7 Mendoza, F., Agustin, J.B., Dominguez-Navarro, J.A.: NSGA and SPEA applied
to multiobjective design of power distribution systems, IEEE Trans. Power Syst.,
2006, 21, (4), pp. 19381945
8 Kumar, D., Samantaray, S.R.: A multi-objective design of advanced power
distribution network using an evolutionary approach. SCES 2014, May 2014,
pp. 16
9 Ramrez-Rosado, I.J., Bernal-Agustn, J.L.: Reliability and costs optimization for
distribution networks expansion using an evolutionary algorithm, IEEE Trans.
Power Syst., 2001, 16, (1), pp. 111118
10 Ramrez-Rosado, I.J., Domnguez-Navarro, J.A.: Possibilistic model based on
fuzzy sets for the multiobjective optimal planning of electric power distribution
networks, IEEE Trans. Power Syst., 2004, 19, (4), pp. 18011810
11 Smith, D.K., Walters, G.A.: An evolutionary approach for nding optimal trees in
undirected networks, Eur. J. Oper. Res., 2000, 120, (3), pp. 593602
12 Duan, G., Yu, Y.: Power distribution system optimization by an algorithm for
capacitated Steiner tree problems with complex-ows and arbitrary cost
functions, Electr. Power Energy Syst., 2003, 25, (7), pp. 515523
13 Carrano, E.G., Soares, L.A.E., Takahashi, R.H.C., Saldanha, R.R., Neto, O.M.:
Electric distribution network multiobjective design using a problem-specic
genetic algorithm, IEEE Trans. Power Deliv., 2006, 21, (2), pp. 9951005
14 Duan, G., Yu, Y.: Problem-specic genetic algorithm for power transmission
system planning, Electr. Power Syst. Res., 2002, 61, (1), pp. 4150
15 Ganguly, S., Sahoo, N.C., Das, D.: A novel MOPSO for electrical distribution
system planning incorporating distributed generation, Energy Syst., 2010, 1, (3),
pp. 291337
16 Ganguly, S., Sahoo, N.C., Das, D.: Multi-objective planning of electrical
distribution systems incorporating sectionalizing switches and tie-lines using
particle swarm optimization, Swarm Evol. Comput., 2012, 3, pp. 1532
17 Kumar, D., Samantaray, S.R.: Design of an advanced electric power distribution
systems using seeker optimization algorithm, Electr. Power Energy Syst., 2014,
63, pp. 196217
18 Nekooei, K., Farsangi, M.M., Pour, H.N., Lee, K.Y.: A multiobjective particle
swarm optimization for sizing and placement of DGs from DG owner and
distribution companys viewpoints, IEEE Trans. Power Deliv., 2014, 29, (4),
pp. 18311840
19 Kumar, D., Samantaray, S.R., Joos, G.: A reliability assessment based graph
theoretical approach for feeder routing in power distribution networks including
distributed generations, Electr. Power Energy Syst., 2014, 57, pp. 1130
20 Kumar, D., Samantaray, S.R., Kamwa, I., Sahoo, N.C.: Reliability-constrained
optimal placement of multiple distributed generators in power distribution
network using cat-swarm-optimization, Electr. Power Compon. Syst., 2014, 42,
(2), pp. 149164
21 Deb, K., Pratap, A., Agarwal, S., Meyarivan, T.: A fast and elitist multiobjective
genetic algorithm: NSGA-II, IEEE Trans. Evol. Comput., 2002, 6, (2),
pp. 182197
22 Nanda, S.J., Panda, G.: Automatic clustering algorithm based on multi-objective
Immunized PSO to classify actions of 3D human models, Eng. Appl. Artif.
Intell., 2013, 26, pp. 14291441
23 Azizipanah-Abarghooee, R., Niknam, T., Zare, M., Gharibzadeh, M.:
Multi-objective short-term scheduling of thermoelectric power systems using a
novel multi-objective -improved cuckoo optimisation algorithm, IET Gener.
Transm. Distrib., 2014, 8, (5), pp. 873894
24 Azizipanah-Abarghooee, R., Niknam, T., Gharibzadeh, M., Golestaneh, F.:
Robust, fast and optimal solution of practical economic dispatch by a new
enhanced gradient-based simplied swarm optimisation algorithm, IET Gener.
Transm. Distrib., 2013, 7, (6), pp. 620635
25 El-Zonkoly, A.M.: Optimal placement of multiple distributed generation units
including different load models using particle swarm optimization, IET Gener.
Transm. Distrib., 2011, 5, (7), pp. 760771
26 Zhang, H., Vittal, V., Heydt, G.T., Quintero, J.: A mixed-integer linear
programming approach for multi-stage security-constrained transmission
expansion planning, IEEE Trans. Power Syst., 2012, 27, (2), pp. 11251133
27 Narimani, M.R., Vahed, A.A., Azizipanah-Abarghooee, R., Javidshari, M.:
Enhanced gravitational search algorithm for multi-objective distribution feeder
reconguration considering reliability, loss and operational cost, IET Gener.
Transm. Distrib., 2014, 8, (1), pp. 5569
28 Niknam, T., Narimani, M.R., Azizipanah-Abarghooee, R., Bahmani-Firouzi, B.:
Multiobjective optimal reactive power dispatch and voltage control: a new
opposition-based self-adaptive modied gravitational search algorithm, IEEE
Syst. J., 2013, 7, (4), pp. 742753
29 Derrac, J., Garca, S., Molina, D., Herrera, F.: A practical tutorial on the use of
nonparametric statistical tests as a methodology for comparing evolutionary and
swarm intelligence algorithms, Swarm Evol. Comput., 2011, 1, pp. 318
21