Escolar Documentos
Profissional Documentos
Cultura Documentos
I. I NTRODUCTION
A number of methods of reconfiguration in distribution networks have been proposed by researchers and also available in
literature. Recently, Kavousi-Fard and Niknam [3] have proposed the reconfiguration problem with respect to the reliability
using a self-adaptive modified optimization algorithm.
In restructured power systems, the use of distributed generation energy resources including photovoltaic (PV), fuel cells,
small wind turbines, etc., is playing an important role because
of various advantages. The advantage of distributed generation
energy resources includes reduction in power loss, improvement in VP, and increase in the reliability of the network.
To achieve the benefits of DG units, the selection of optimal locations and capacity is becoming the major problem.
Various methods have been proposed by researchers in order
to find the optimal placement and capacity of DG units. These
methods are often based on artificial intelligence and heuristic algorithms. Kollu et al. [4] proposed a Harmony Search
(HS) algorithm-based novel method to allocate DG units optimally in distribution system for power loss reduction and VP
improvement.
Distribution flexible ac transmission system (DFACT)
devices are used in distribution systems with different applications and controlling methods for improving the power quality
indices. DSTATCOM, unified power flow controller (UPFC),
and dynamic voltage restorer (DVR) are widely used DFACT
devices. To find the optimal location and capacity of DFACT
devices has a considerable impact in distribution systems. Some
researchers have proposed various methods to find the optimal location and size of DFACT unit. Farhoodnea et al. [5]
presented a firefly algorithm-based novel method to optimally
place the DSTATCOM in distribution system.
In addition, several methods have been proposed for the
simultaneous reconfiguration and optimal allocation of DG or
DFACT unit in distribution system. However, simultaneous
reconfiguration with both DG and DFACT is rarely available
in the literature.
In this paper, the ant colony optimization (ACO) approach
has been used for simultaneous multiobjective reconfiguration
and optimal allocation of PV and DSTATCOM in a distribution
network. The main objective of the work includes loss reduction, VP improvement, and equalizing the feeder load balancing
(LB). To avoid the convergence problem, the input and output
data are normalized in the same range based on fuzzy sets.
1949-3029 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
BAGHERI TOLABI et al.: SIMULTANEOUS RECONFIGURATION, OPTIMAL PLACEMENT OF DSTATCOM, AND PV ARRAY IN A DISTRIBUTION SYSTEM
211
2 2
2 R 2 + X 2
= Vk + k 2 k Pk2 + Q2k + Yk Vk
Vk
2
(2)
2 R k P k + X k Q k + Yk V k
where
Pk
Qk
Ploss,k
Qloss,k
PLk+1
QLk+1
Rk
Xk
Yk
Vk
B. PV Array Equations
A PV array is a series-parallel combination of PV
panels.
A PV system for residential, commercial, or industrial energy
supply normally contains PV array, dcdc converter, inverter,
and a tracking system. An example of the electricity generated
by PV system is shown in Fig. 1.
The active and reactive powers of PV unit can be calculated
using the following equations [7]:
2
V
Ppv = k Ppv,loss Pk2 + Q2k
Rk
G 1/2
Q2pv 2Pk Ppv 2Qk Qpv
(3)
L
2
V
Qpv = k Ppv,loss Pk2 + Q2k
Rk
2
G 1/2
Ppv 2Pk Ppv 2Qk Qpv
(4)
L
where
Ppv
Qpv
G
L
C. DSTATCOM Equations
The DSTATCOM is one of the DFACT devices that basically
consists of a coupling transformer with a leakage reactance, a
GTO/IGBT voltage source convertor (VSC), and a dc capacitor.
Fig. 2 shows a bus in a distribution system equipped with a
DSTATCOM.
The active and reactive powers of DSTATCOM can be
calculated using the following equations [8]:
PDSTATCOM = (Vk VM /XL ) sin
QDSTATCOM = Vk 2 /XL (Vk VM /XL ) cos
(5)
(6)
212
ns
Ppvk + PDSTATCOMk
(Pk + Ploss,k )
k=1
k=1
4) det(A1 ) = 1 (This constraint checks the radial structure
of the network and feeding all available nodes. The details
are given in the Appendix.)
where
Vk
voltage at bus k after configuration;
maximum voltage at bus k;
Vkmax
minimum voltage at bus k;
Vk min
current in line section between buses k and
Ik,k+1
k + 1after reconfiguration;
Ik,k+1 max maximum current limit of line section
between buses k and k + 1;
total number of sections in the system;
ns
reduced bus incidence matrix.
A1
3)
ns
nf ns
R k I k 2
(7)
j=1 k=1
nf
Vk Vref,k
(8)
j=1 k=1
1/
nf
Ik
nf
k=1
Ik
(9)
When there are different objectives to be satisfied simultaneously, a comparison is required to get the best solution. Since
different terms of the multiobjective function are in various
ranges, all the values are normalized in the same range to avoid
the convergence problem [9]. Each variable has a membership
function () that determines the rank and effectiveness of its
variable as follows:
wi .f ij (X)
(11)
i=1
j=1 klb
nf ns
LB =
ij
, that is Tij is the obtained value for the ith
where fij = Ti0
term of the objective function (i = 1, 2, 3) for the jth solution,
Ti0 is the base value for the ith term of the objective function
before optimization, and fimin and fimax are the best and worst
where wi is the weighting factor for the ith term of the objective function. The weighting factors are considered w1 = w2 =
w3 = 0.33, in which the three objectives are assumed to have
equal importance.
III. ACO
The ACO is a swarm intelligence-based technique, proposed
by Dorigo et al. for the solution of combinatorial problems
[10]. The ACO algorithm is originally inspired by the biological
behavior of the ants and specifically their way of communication. This inspiration comes from the ability of real ants to
find the short paths in their movement from and to their nests
when searching for food source [11]. However, ants do not communicate with each other in a direct way but they exchange
information through what is known as pheromones. A simple
diagram of ant foraging behavior is shown in Fig. 3.
In this figure, two ants leave their nest in different directions
at the same time for searching the food. As they move about,
they deposit a pheromone trail that evaporates slowly and is
detectable by other ants. If no pheromone exists initially outside
the nest, the paths of the two ants are generally random [12],
[13]. The evaporation of pheromone is also an important part of
the foraging process that helps deemphasize older trails leading
to exhausted food sources.
The ACO algorithms attempt to exploit the efficiency of ant
foraging behavior by creating an abstract environment of possible paths, and simulating ants traveling along these paths.
Each individual chooses a different path and comes up with
different solutions in the search space region. When an ant
passes through a path, it deposits the pheromone. This approach
BAGHERI TOLABI et al.: SIMULTANEOUS RECONFIGURATION, OPTIMAL PLACEMENT OF DSTATCOM, AND PV ARRAY IN A DISTRIBUTION SYSTEM
[i,j ] [i,j ]
, if i, j T
[i,j ] [i,j ]
(12)
P k i,j (t) =
iT
0,
otherwise
where k is the kth ant, > 0 and > 0 are two parameters
(the choice of and is determined experimentally), i,j is
the pheromone value for the ijth path of the ACO algorithm,
i,j is heuristic desirability for the ijth path of the ACO algorithm, and T is the total number of paths currently not visited
by an ant. As suggested in (12), the transition probability is
determined by the pheromone and heuristics of the trail, which
corresponds to the individual cost of the ijth path. This path is
said to be accepted when higher probability value is gained as
a result of high quantity of pheromone or heuristic desirability
[10]. After all ants have completed their tours, the pheromone
level is updated by
i,j (t + 1) = (1 ) i,j (t) +
k
i,j
(t)
(13)
k=1
213
4) updating the list of feasible operation and probability values so that ants schedule their next operation until they
reach food node; and
5) analysis of the best solution, updating pheromones, and
checking the stopping criterion [11], [14], [15].
IV. P ROPOSED F UZZY-ACO M ETHOD
In this section, the proposed fuzzy-ACO approach has been
used for reconfiguration and optimal placement of DSTATCOM
and PV array in distribution system.
An optimal feeder topology can be represented as
S 1 = tie switches1 , PV1 , DSTATVOM1 .
In the S 1 vector, PV1 is a matrix with two columns as
follows:
PV1 = sizepv 1 locationpv 1
and DSTATCOM is a 1 2 matrix as follows:
1
DSTATCOM1 = sizeDSTATCOM 1 locationDSTATCOM
where size1pv and location1pv are the initial proposed capacity and bus location for the PV unit, respectively, and
size1DSTATCOM and location1DSTATCOM are the initial capacity
and bus candidate to install the DSTATCOM unit, respectively.
By updating the ACO algorithm, the second, third, and the
ith solution vectors are generated with the new proposed tie
switches, PV, and DSTATCOM size and location as follows:
S i = tie switchesi , PVi , DSTATVOMi .
For each ith solution, a power flow program is carried out, the
fuzzy membership values, and the fitness of objective function
are evaluated and compared with the previous solution and the
better solution has been selected and replaced. The proposed
method is summarized in the following steps.
Step 1) Read data of distribution system (bus, load, branch,
sectionalizing, number of tie switches, etc.), initialize the ACO parameters and run the power flow
program [6].
Step 2) Generate a solution vector as S by offering the tie
switches, PV, and DSTATCOM sizes and locations
in the network without violating the constraints. Run
the power flow program, evaluate the fuzzy membership value () based on (10) for each objective
function. Calculate the fitness of fuzzy objective
function according to (11). Save the solution as the
best solution.
Step 3) Update the ACO pheromone value using (13) and
(14). Generate a new solution by updating ACO, and
compare the fitness value of the new solution with
the best solution. If the new generated solution has a
better fitness value than best solution, save it as the
best solution.
Step 4) Check the number of iterations, and if the number of
iterations does not exceed from the specified value,
repeat step 3).
214
TABLE I
S ELECTED PARAMETERS FOR THE ACO
BAGHERI TOLABI et al.: SIMULTANEOUS RECONFIGURATION, OPTIMAL PLACEMENT OF DSTATCOM, AND PV ARRAY IN A DISTRIBUTION SYSTEM
TABLE II
R ESULTS OF 33-B US T EST S YSTEM
215
TABLE III
R ESULTS OF TPC T EST S YSTEM
B. Obtained Results
The results of the proposed approach for 33-bus test system are presented in Table II for the five tested cases. It is
observed from Table II that base case power loss in the system is 202.5 kW, which is reduced to 143.69, 93.30, 110.22,
and 48.73 kW for cases II to V respectively. VP index is calculated as 1.7, 1.47, 1.23, 1.08, and 0.82 and LB index is obtained
67.71, 58.42, 49.64, 49.51, and 43.05 for cases I to V, respectively. Also, Table II shows the optimal location and size for PV
and DSTATCOM units for the cases III to V.
By comparing the improvements obtained for cases II to V
with the base case (case I), it is found that simultaneous reconfiguration with PV and DSTATCOM allocation (case V) has
75.93%, 36.42%, and 51.76% improvements in Ploss , LB, and
VP indices, respectively.
Also, the results of ACO method (without fuzzy system) for
33-bus test system shows that the power loss is 184.12, 136.93,
1145.27, and 89.36 kW, VP index is 1.56, 1.48, 1.30, and 1.07,
and LB index is 63.37, 56.79, 52.18, and 49.43 for cases II to
V, respectively. Obtained results show that the performance of
fuzzy-ACO approach is better as compared to the ACO (without
fuzzy system) for 33-bus test system.
Similarly, the proposed approach is also applied on second
test system, i.e., the TPC, and it is observed that all three indices
of Ploss , LB, and VP have been improved for all cases II to V.
The results of the proposed method on TPC test system are
presented in Table III for all five tested cases. It is observed
from Table III that the base case power loss in the system
is 531 kW, which is reduced to 383.15, 371.42, 368.09, and
293.55 kW for cases II to V, respectively. VP index is calculated as 2.5, 2.34, 2.03, 1.92, and 1.67 and LB index is obtained
as 140.4, 111.38, 115.70, 102.65, and 98.47 for cases I to V,
respectively. Also, Table III includes the optimal location and
size for PV and DSTATCOM units for the cases III to V on
TPC test system. The results of ACO approach (without fuzzy
system) on TPC test system shows that the power loss is 427.55,
376.21, 392.80, and 352.04 kW, VP index is 2.31, 2.17, 2.10,
and 1.93, and LB index is 132.04, 128.19, 121.53, and 117.99
for cases II to V, respectively. Obtained results show that the
performance of fuzzy-ACO approach is better performance as
compared to ACO (without fuzzy system) on TPC test system.
Obtained results are compared on both the systems under
study and the following conclusions have been obtained.
1) The performance of fuzzy-ACO method is better as compared to ACO (without fuzzy system).
2) As can be seen from Tables II and III, by applying the
fuzzy-ACO method, the active power loss is reduced, VP
is improved, and the LB of the feeders has been increased
for cases II to V.
3) From Tables II and III, the results prove the superiority of
case V (fuzzy-ACO approach) as compared to other cases.
In other words, by investigating various cases involving
reconfiguration, PV allocation, DSTATCOM allocation,
and combination of them, it is found that reconfiguration
simultaneous placement of PV and DSTATCOM units
can be more beneficial than separate single-objective optimization. By comparing the improvements obtained by
cases II to V as compared to the base case (case I) on the
33-bus test system, it is found that simultaneous reconfiguration with PV and DSTATCOM allocation (case V) has
the most improvements in Ploss (75.93%), LB (36.42%),
and VP (51.76%) indices. These results are repeated for
the TPC test system with 44.71% improvement in Ploss ,
29.86% in LB, and 33.2% in VP indices.
4) Since the obtained Ploss index for case III is better than
the corresponding value for case IV in both test systems, it can be concluded that for power loss reduction,
simultaneous reconfiguration and PV allocation is more
effective than simultaneous reconfiguration along with
DSTATCOM allocation. On the other hand, the calculated LB and VP indexes for case IV are better than
the corresponding values for case III in both test systems. Therefore, in order to improve the LB or VP
indexes of distribution system, simultaneous reconfiguration and DSTATCOM allocation can be more effective
than reconfiguration along with PV allocation.
5) By comparing the improvements obtained by cases II to
V from Tables II and III, it is confirmed that reconfiguration and simultaneous allocation of PV or DSTATCOM
(cases III and IV) lead to more improvement in Ploss ,
LB, and VP indices as compared to single reconfiguration
(case II).
216
TABLE IV
O BTAINED R ESULTS BASED ON VARIOUS L OAD L EVELS
(TPC T EST S YSTEM )
To assess the performance of the proposed approach at different load levels, the TPC test system is simulated at three load
levels: 0.5 (light), 1.0 (nominal), and 1.6 (heavy) for cases I
and V. The obtained results are presented in Table IV. It can
be seen from Table IV that at light load (for case V), the base
case power loss is 122.76 kW, which is reduced to 103.61 kW
and the base case VP index is 1.7 that has been improved to
1.61. Also, at this load level, the LB index for the base case is
108, indicating that it is improved to 100.43. Similarly, it can be
seen from Table IV that all three indices of Ploss , VP, and LB
have been improved as compared to the base case at heavy load.
The obtained results show that the performance of the proposed
approach at different load levels is satisfactory.
Case V (at nominal load) is simulated using fuzzy-genetic
algorithm (GA) and fuzzy-particle swarm optimization (PSO)
and compared with the results obtained by fuzzy-ACO method.
All parameters of each algorithm are tuned for achieving the
best solution. The results of different methods on both test
systems are presented in Table V for the case V).
From Table V, for the 33-bus test system Ploss is calculated as
73.26, 69.15, and 48.73 kW, VP is obtained 0.87, 0.84, and 0.82
and LB is 48.59, 45.67, and 43.05 using fuzzy-GA, fuzzy-PSO,
and fuzzy-ACO approaches, respectively. Due to similar results
obtained for the TPC test system (second part in Table V), it
is confirmed that the performance of the fuzzy-ACO is better than that of the fuzzy-GA and fuzzy-PSO in all terms of
loss reduction, improving VP, and equal LB for the both test
systems. Also, it is observed that the performance of fuzzyACO approach is better than that of other tested approaches
because of the lowest fitness equal to 0.147 in comparison with
the fuzzy-PSO (0.164) and fuzzy-GA (0.181). Fig. 6 shows the
percentage improvements in Ploss , LB, and VP indices using the
fuzzy-GA, fuzzy-PSO, and fuzzy-ACO approaches for case V
as compared to the base case (case I) at nominal load. It can
be seen from Fig. 6 that among three terms Ploss , LB, and VP,
the improvement in Ploss is more significant as compared to
LB and VP for all three fuzzy-GA, fuzzy-PSO, and fuzzy-ACO
approaches. This means that reconfiguration and simultaneous
placement of PV and DSTATCOM units has a greater impact
on loss reduction as compared to VP or LB improvement.
TABLE V
C OMPARISON OF THE S IMULATION R ESULTS FOR D IFFERENT M ETHODS
BAGHERI TOLABI et al.: SIMULTANEOUS RECONFIGURATION, OPTIMAL PLACEMENT OF DSTATCOM, AND PV ARRAY IN A DISTRIBUTION SYSTEM
217
Fig. 6. Percentage improvements in Ploss , LB, and VP indexes using fuzzyGA, fuzzy-PSO, and fuzzy-ACO methods.
1,
if branch b is directed toward node n.
A radial distribution feeder that feeds all of loads can be simulated by a tree graph, i.e., a connected graph (i.e., there is not
any isolated load) without any cycle (i.e., the network has a
radial structure). For each tree graph, satisfying the following
[1] H. B. Tolabi, M. Gandomkar, and M. Bahreyni Borujeni, Reconfiguration and load balancing by software simulation in a real distribution
network for loss reduction, Can. J. Elect. Electron. Eng., vol. 2, no. 8,
pp. 386391, Aug. 2011.
[2] A. Merlin and H. Back, Search for a minimal-loss operating spanning
tree configuration in an urban power distribution system, in Proc. 5th
Power Syst. Comput. Conf. (PSCC), Cambridge, U.K., 1975, pp. 118.
[3] A. Kavousi-Fard and T. Niknam, Multi-objective stochastic distribution
feeder reconfiguration from the reliability point of view, Energy, vol. 64,
pp. 342354, Jan. 2014.
[4] R. Kollu, S. R. Rayapudi, and V. L. N. Sadhu, A novel method for
optimal placement of distributed generation in distribution systems using
HSDO, Int. Trans. Elect. Energy Syst., vol. 24, pp. 547561, Apr. 2014.
[5] M. Farhoodnea, A. Mohamed, H. Shareef, and H. Zayandehroodi,
Optimum D-STATCOM placement using firefly algorithm for power
quality enhancement, in Proc. 7th Power Eng. Optim. Conf. (PEOCO),
Langkawi, Malaysia, 2013, pp. 98102.
[6] R. S. Rao, S. V. L. Narasimham, M. R. Raju, and A. S. Rao, Optimal
network reconfiguration of large-scale distribution system using harmony
search algorithm, IEEE Trans. Power Syst., vol. 26, no. 3, pp. 1080
1088, Aug. 2011.
[7] R. Srinivasa Rao, K. Ravindra, K. Satish, and S. V. L. Narasimham,
Power loss minimization in distribution system using network reconfiguration in the presence of distributed generation, IEEE Trans. Power
Syst., vol. 28, no. 1, pp. 317325, Feb. 2013.
[8] T. Xiangqian, X. Keqing, S. Ming, and M. Xianhong, Reactive power
and unbalance compensation with DSTATCOM, in Proc. Int. Conf. Elect.
Mach. Syst. (ICEMS), Nanjing, China, 2005, vol. 2, pp. 11811184.
[9] L. A. Zadeh, Fuzzy sets, Inf. Control, vol. 8, pp. 338353, 1965.
[10] M. Dorigo, V. Maniezzo, and A. Colorni, The ant system: An autocatalytic optimizing process, Politecnico di Milano, Italy, Tech. Rep.
91-016 Revised, 1991.
[11] M. Dorigo, Optimization, learning and natural algorithms, Ph.D. dissertation, Dipartimento di Elettronica, Politecnicodi Milano, Milano, Italy,
1992.
[12] V. Maniezzo, L. M. Gambrdella, and F. DeLuigi, Ant Colony
Optimization: New Optimization Techniques in Engineering, G. C.
Onwubolu and B. V. Babu, Eds. New York, NY, USA: Springer, 2004,
pp. 101117.
[13] M. Gandomkar and H. B. Tolabi, Investigation of simulated annealing, ant-colony and genetic algorithms for distribution network expansion
planning with distributed generation, in Proc. 9th WSEAS Int. Conf.
Instrum. Meas. Circuits Syst. (IMCAS10), China, 2010, pp. 4852.
[14] M. Dorigo, V. Maniezzo, and A. Colorni, The ant system: Optimization
by a colony of cooperating agents, IEEE Trans. Syst. Man Cybern. B,
vol. 26, no. 1, pp. 2941, Feb. 1996.
[15] M. Dorigo and L. M. Gambardella, Ant colony system: A cooperative
learning approach to the traveling salesman problem, IEEE Trans. Evol.
Comput., vol. 1, no. 1, pp. 5366, Apr. 1997.
[16] M. E. Baran and F. Wu, Network reconfiguration in distribution system
for loss reduction and load balancing, IEEE Trans. Power Del., vol. 4,
no. 2, pp. 14011407, Apr. 1989.
[17] L. W. Oliveira et al., Optimal reconfiguration and capacitor allocation in
radial distribution systems for energy losses minimization, Elect. Power
Energy Syst., vol. 32, pp. 840848, Oct. 2010.
[18] J. Grainger and William D. Stevenson, Jr., Power flow solutions, in
Power System Analysis, 1st ed. New York, NY, USA: McGraw-Hill, 1994.
218