Sei sulla pagina 1di 11

Electric Power Components and Systems

ISSN: 1532-5008 (Print) 1532-5016 (Online) Journal homepage: http://www.tandfonline.com/loi/uemp20

Group Search Optimization for Solution of


Different Optimal Power Flow Problems

Mousumi Basu

To cite this article: Mousumi Basu (2016): Group Search Optimization for Solution of
Different Optimal Power Flow Problems, Electric Power Components and Systems, DOI:
10.1080/15325008.2015.1122109

To link to this article: http://dx.doi.org/10.1080/15325008.2015.1122109

Published online: 25 Mar 2016.

Submit your article to this journal

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at


http://www.tandfonline.com/action/journalInformation?journalCode=uemp20

Download by: [University of Nebraska, Lincoln] Date: 27 March 2016, At: 01:33
Electric Power Components and Systems, 00(0):1–10, 2016
Copyright C Taylor & Francis Group, LLC

ISSN: 1532-5008 print / 1532-5016 online


DOI: 10.1080/15325008.2015.1122109

Group Search Optimization for Solution of Different


Optimal Power Flow Problems
Mousumi Basu
Department of Power Engineering, Jadavpur University, Kolkata, India

CONTENTS
Abstract—This article presents group search optimization for the
Downloaded by [University of Nebraska, Lincoln] at 01:33 27 March 2016

1. Introduction solution of different optimal power flow problems of a power sys-


2. Problem Formulation tem with generators that may have either convex or non-convex fuel
cost characteristics. Different operational constraints, such as gen-
3. GSO erator capacity limits, power balance constraints, line flow, and bus
4. Application of the Proposed Method voltages limits, have been considered. Settings of transformer tap
5. Conclusion ratio and reactive power compensating devices have also been in-
cluded as the control variables in the problem formulation. Group
References
search optimization, inspired by the animal searching behavior, is a
biologically realistic algorithm. Group search optimization has been
implemented to solve four different objectives such as fuel cost min-
imization, emission minimization, voltage profile improvement, and
voltage stability enhancement with the optimal power flow embed-
ded on IEEE 30-bus, 57-bus, and 118-bus test systems. The results
of the proposed approach are compared with those obtained by other
evolutionary methods reported in the literature. It is found that the
proposed group search optimization based approach is able to provide
a better solution.

1. INTRODUCTION
Optimal power flow (OPF) is an important tool for power sys-
tem operators both in power system planning and operation
for many years. The main purpose of an OPF is to deter-
mine the settings of control variables for economic and se-
cure operation of a power system. The OPF minimizes the
power system operating objective function while satisfying
a set of equality and inequality constraints. The equality con-
straints are power flow equations and inequality constraints are
the limits on control variables and functional operating con-
straints. OPF is a highly non-linear, non-convex, large scale
static optimization problem. Several optimization techniques
have emerged so far and have been applied to solve the OPF
problem.
Earlier, OPF algorithms were based on classical
Keywords: group search optimization, optimal power flow, fuel cost
minimization, emission minimization, voltage profile improvement, voltage mathematics-based methods. The gradient-based method [1],
stability enhancement non-linear programming (NLP) [2], quadratic programming
Received 23 May 2014; accepted 8 November 2015
(QP) [3], Newton-based method [4, 5], linear programming
Address correspondence to Mousumi Basu, Department of Power
Engineering, Jadavpur University, 2nd Campus, Salt Lake, Sector III, (LP) [6], and interior point methods (IPMs) [7–9] have been
Kolkata 700 098, India. E-mail: mousumibasu@yaoo.com successfully applied to solve OPF problems.

1
2 Electric Power Components and Systems, Vol. 00 (2016), No. 00

The problem of OPF was originally formulated in 1962 Due to its high efficiency, GSO has been applied in many
by Carpentier [1], and he solved the OPF problem by using fields. Moreover, some papers also indicate GSO as solutions
the reduced gradient method. Later, Dommel and Tinney [2] to some power system optimization problems, such as non-
worked out the problem, based on Kuhn–Tucker optimality cri- convex economic dispatch problems [24], etc.
terion using a combination of the gradient method and penalty The effectiveness of the proposed GSO algorithm is tested
functions. In the following years, much research took place to on IEEE 30-bus, 57-bus, and 118-bus test systems for four
improve the quality of the OPF solution. single and compound-objective problems. The single objec-
These classical optimization techniques have been widely tive problems are (1) fuel cost minimization, (2) emission
applied to varieties of OPF problems. However, these tech- minimization, (3) voltage deviation minimization, and (4) en-
niques fail to deal with systems having complex non-smooth, hancement of voltage stability. It has been seen that perfor-
non-convex, and non-differentiable objective functions and mance of the proposed GSO method is better compared to
constraints. The NLP-based procedure has many drawbacks, other evolutionary methods.
such as insecure convergence properties, convergence to local
optimum, and algorithmic complexity. The LP formulation re-
Downloaded by [University of Nebraska, Lincoln] at 01:33 27 March 2016

quires linearization of the objective function and constraints 2. PROBLEM FORMULATION


in non-negative variables. Though LP method is fast and reli-
The OPF problem optimizes the steady state performance of
able, it has some disadvantages in dealing with the non-smooth
power system with respect to specified objective function sub-
cost functions of modern generators. QP is a special case of
ject to various equality and inequality constraints. Here, four
NLP, where the objective function must be quadratic and con-
different single objectives, i.e., (1) fuel cost minimization, (2)
straints are linear. QP-based approaches have drawbacks re-
emission minimization, (3) reduction of voltage deviation, and
lated to approximation of the exact cost function with piece-
(4) improvement of voltage stability and compound-objective
wise quadratic approximation. Newton-based techniques have
problems are considered. Four single objective functions and
the common drawback that the convergence characteristics are
two compound-objective functions and constraints are formu-
sensitive to the choice of starting values and they may even fail
lated as follows.
to converge due to inappropriate starting values. Though the
IPM is considered to be one of the most efficient techniques for 2.1. Objective Functions
solving OPF problems, it suffers from bad initialization, termi-
2.1.1. Minimization of Fuel Cost
nation, and optimality criteria and in most cases, is unable to
solve non-linear and quadratic objective functions. Improper The fuel cost function of each thermal generating unit, con-
selection of step size may also give an infeasible solution in sidering the valve-point effects, is expressed as the sum of a
IPM. quadratic and a sinusoidal function given below.
Due to tremendous improvement in capability of computers 
NG

in recent years, evolutionary algorithms, such as genetic algo- F1 = ai + bi PGi + ci PGi
2

rithm (GA) [10], evolutionary programming (EP) [11], particle 


i=1
  min 
swarm optimization (PSO) [12, 13], simulated annealing (SA) + di × sin ei × PGi − PGi  , (1)
[14], enhanced GA [15], improved evolutionary programming where ai , bi , and ci are the fuel cost coefficients of the ith
[16], differential evolution (DE) [17, 18], biogeography-based generator; di and ei are the coefficients of the ith generator
optimization (BBO) [19], improved PSO (IPSO) [20], and the reflecting valve-point effect; PGi is the active power genera-
faster evolutionary algorithm (FEA) [21] have been applied tion of the ith generator; PGimin
is the minimum active power
for solving various complex OPF problems to overcome the generation limit of the ith generator; and NG is the number of
drawbacks of classical techniques. committed generators.
Group search optimization (GSO) is a biologically realistic The vector of dependent variables x may be represented as
algorithm. It takes inspiration from animal (such as lions and 
wolves) searching behavior. He et al. proposed GSO in 2009 x T = PGslack , VL1 , . . . , VL N P Q , Q G1 , .., Q G N P V , Sl1 , . . . , Sl NT L ,(2)
[22]. where PGslack denotes the slack bus power, VL is the PQ bus
GSO utilizes a special framework where individuals are voltage, Q G is the reactive power output of the generator, Sl
divided into three classes and grow separately. Shen et al. [23] is the transmission line flow, NPV is the number of generator
showed that GSO is an important alternative for constrained buses, NPQ is the number of PQ buses, and NTL is the number
optimization. of transmission lines.
Basu: Group Search Optimization for Solution of Different Optimal Power Flow Problems 3

The vector of control variables u may be represented as as described in Eq. (8):






u T = VG1 , ...., VG N P V , PG2 , ...., PG N P V , Q c1 , . . . , Q cNC , IP Q
=
Y1 Y2 VPQ
.
IPV Y3 Y4 VPV
(8)
× T1 , . . . , TNT , (3)
The L-index is calculated as follows:
where NC and NT are the number of shunt VAR compensators
 
and the number of tap changing transformers, VG is the ter- L = max L j (where j= 1, 2, . . . ,N P Q ). (9)
minal voltage at the generator bus, Q c is the output of shunt
The lower value of L represents a more stable system. To
VAR compensator, and T is the tap setting of the tap changing
improve voltage stability and to move the system far from the
transformer.
voltage collapse point, the objective function can be defined
2.1.2. Minimization of Emission as follows

The atmospheric pollutants such as sulfur oxides (SOx ) and Minimize F4 = L . (10)
nitrogen oxides (NOx ) caused by thermal generating units can
Downloaded by [University of Nebraska, Lincoln] at 01:33 27 March 2016

be modeled as a sum of a quadratic and an exponential function 2.2. Constraints


given below:
The objective functions are subjected to the equality con-

NG
 straints imposed by the physical laws governing the trans-
F2 = αi + βi PGi + γi PGi
2
+ ηi exp (λi PGi )] , (4) mission system as well as the inequality constraints imposed
i=1
by the equipment ratings given in what follows.
where αi , βi , γi , ηi , λi are the emission coefficients of the ith
generator. 2.2.1. Equality Constraints.
These constraints are load flow equations described as
2.1.3. Minimization of Voltage Deviation

NB
  
The objective is to minimize the voltage deviation of all load PGi − PDi − Vi V j G i j cos δi − δ j
(PQ) buses from 1 p.u. for the operating power system more j=1
 
securely. The objective function can be formulated as follows: + Bi j sin δi − δ j = 0, i = 1, 2, .., N B , (11)

NP Q 
NB
  
Minimize F3 = |Vi − 1.0|, (5) Q Gi − Q Di − Vi V j G i j sin δi − δ j
j=1
i=1  
where NPQ is the number of load buses in the power system. − Bi j cos δi − δ j = 0, i = 1, 2, .., N B , (12)
where NB is the number of buses, PGi and QGi are active and
2.1.4. Voltage Stability Enhancement reactive power generation at the ith bus, PDi and QDi are active
The voltage stability problem is the ability of a power system and reactive power demand at the ith bus, and G i j and Bij are
to maintain acceptable voltages at all busbars in the system un- the transfer conductance and susceptance between ith bus and
der normal operating condition and even after being subjected j th bus, respectively.
to disturbances. A weak system, a system with long transmis-
sion lines, and a heavily loaded system are much prone to the 2.2.2. Inequality Constraints
voltage instability problem. Voltage stability enhancement can Generator constraints: The generator voltage magnitudes
be done by minimizing the voltage stability indicator, i.e., the and reactive power outputs are constrained by design spec-
L-index value at each bus of a power system. The L-index of ifications. The lower and upper limits of generator voltage
a bus indicates the proximity of voltage collapse condition of magnitude and reactive power output are given below:
that bus. L-index L j of the jth bus is defined as follows [25]: min
VGi ≤ VGi ≤ VGi
max
, i = 1, 2, . . . , N P V , (13)
 
 
N PV
Vi  PGi ≤ PGi ≤ PGi , i = 1, 2, . . . , N P V ,
min max
 (14)
L j = 1 − F ji  (where j= 1, 2, . . . , N P Q) (6)
 Vj  Gi ≤ Q Gi ≤ Q Gi , i = 1, 2, . . . , N P V .
Q min max
(15)
i=1
−1
F ji = − [ϒ1 ] [ϒ2 ] , (7) Shunt VAR compensator constraints: Reactive power output
of shunt VAR compensators must be restricted within their
where NPV is the number of PV buses, and NPQ is the number of
lower and upper limits as follows:
the PQ bus. Y 1 and Y 2 are the submatrices of the system YBUS
ci ≤ Q ci ≤ Q ci , i = 1, 2, ...., NC .
Q min max
obtained after segregating the PQ and PV busbar parameters (16)
4 Electric Power Components and Systems, Vol. 00 (2016), No. 00

Transformer constraints: The upper and lower values for resource and is bestowed with the best fitness value called the
the discrete transformer tap settings are limited by physical producer. Some numbers of the group except the producer are
considerations and these are given below: randomly selected as scroungers and the rest of the members
are rangers. Scroungers perform area copying to join the re-
Timin ≤ Ti ≤ Timax , i = 1, 2, . . . , N T . (17)
source found by the producer and search around it. Rangers
Security constraints: These include the constraints on volt- employ ranging behavior by random walk in the searching
age magnitudes at PQ buses and transmission line loadings. space to raise the chance of escaping local optima.
Voltage of each PQ bus must be within its lower and operat- By using vision ability, producer scans three points around
ing limits. Line flow through each transmission line must be its position in certain distances and head angles in the search
within its capacity limits. These are described as follows: space for the better states. At the kth iteration, the producer
behavior is as follows:
VLimin ≤ VLi ≤ VLimax , i = 1, 2, .., N P Q , (18)
Sli ≤ Slimax , i = 1, 2, . . . , N T L . (19) (1) The producer scans at 0◦ and tests three points toward
its position using Eqs. (21)–(23):
 
Downloaded by [University of Nebraska, Lincoln] at 01:33 27 March 2016

X z = X kP + r1lmax D kp φ k , (21)

3. GSO r2 θmax
X r = X kP + r1lmax D kP φ k + , (22)
GSO [22, 23] is inspired by animal searching behavior 2

and group living theory. The framework is based on the r2 θmax
X l = X kP + r1lmax D kP φ k − , (23)
producer–scrounger (PS) model assuming that group mem- 2
bers search either for “finding” or for joining opportunities. where XP is the position of the producer, is a normally
The concept of animal scanning mechanisms is employed distributed random number with mean 0 and standard
metaphorically for designing an optimum searching strategy deviation 1, r2 is a uniformly distributed random num-
in order to solve optimization problems. ber in the range of (0, 1), and lmax is maximum pursuit
The population of GSO is called a group and each indi- distance and θmax is maximum pursuit angle.
vidual in the population is called a member. Each member is
(2) The producer will then find the best point. If the best
defined by its position and head angle. In n-dimensional search
point has a better value in comparison with its current
space, the ith member of GSO at the kth iteration has a current
position, the producer will fly to that point. If not, it will
position X ik ∈ R n and a head angle φik = (φi1k
, . . . , φi(n−1)
k
)∈
remain in its current position turning its head using the
R . The search direction of the ith member Dik (φik ) =
n−1
equation given below:
k
(di1 , ...., din
k
) ∈ R n can be calculated from φik via polar to
Cartesian coordinate transformation [24]: φ k+1 = φ k + r2 αmax , (24)

n−1
 k where αmax ∈ R 1 is the maximum turning angle.
k
di1 = cos φiq
q=1 (3) If the producer cannot find a better area after
aiterations, it will turn its head back to 0◦ as follows:
  n−1  k
dikj = sin φi(k j−1) cos φiq , j = 2, ...., (n − 1) .
φ k+a = φ k , (25)
q= j
 k 
k
din = sin φi(n−1) (20) where a ∈ R 1 is a constant.

The GSO algorithm comprises three kinds of members, The scroungers keep searching for opportunities to join
i.e., producers, scroungers, and rangers, assuming that there is the resources found by the producer. At the kth iteration, the
only one producer and the remaining members are scroungers area copying behavior of the ith scrounger can be modeled as
and rangers. Here, the simplest joining policy is used where follows:
all scroungers will join the resource found by the producer.  
X ik+1 = X ik + r3 ◦ X kP − X ik , (26)
Group members search for the patches by moving over the
search space. The producer and the scroungers can switch where X ik is the position of the ith scrounger at the kth iteration.
between the two roles [26] assuming that they do not differ in r3 is a uniform random number in the range of (0, 1). Operator
their relevant phenotypic characteristics. “◦“ calculates the entry wise product of the two vectors.
At each iteration, a group member, located in the most Rangers are dispersed from their positions by random walk
promising area, adopts animal scanning to seek the optimal at the search space. At the kth iteration, a ranger produces a
Basu: Group Search Optimization for Solution of Different Optimal Power Flow Problems 5

4.1. IEEE 30-bus System


The line data, bus data, generator data, and the minimum and
maximum limits for the control variables have been adapted
from [27]. The total system active power demand is 2.834 p.u.
at a 100-MVA base. In this study, 50 test runs are performed to
solve the OPF problem for four different objective functions.

4.1.1. Minimization of Fuel Cost


The proposed GSO approach is applied for minimization of
fuel cost as the objective function. Here, the population size
and maximum iteration number have been selected, 100 and
100 respectively, for this test system. The optimal values of
control variables obtained from the proposed GSO are given
Downloaded by [University of Nebraska, Lincoln] at 01:33 27 March 2016

in Table 1. The convergence characteristic obtained from pro-


posed GSO for cost minimization is shown in Figure 2. It is
seen that the minimum cost obtained from GSO is the lowest
among BBO [19], DE [18], and PSO.

4.1.2. Minimization of Emission


The proposed GSO approach is applied for minimization of
emission as the objective function. Here, the population size
and maximum iteration number have been selected 100 and
FIGURE 1. Flowchart of the GSO algorithm.
100, respectively, for this test system. The optimal values of
control variables obtained from the proposed GSO are given
random head angle φi by using Eq. (21), and then it chooses in Table 1. The emission convergence characteristic obtained
a random distance by using Eq. (27) to move to the new point from proposed GSO for emission minimization is shown in
by using Eq. (28): Figure 3. It is seen that minimum emission obtained from
GSO is the lower than IPSO [20].
li = br1lmax , (27)
 k+1 
k+1
Xi = X ik + li Dik φ , (28) 4.1.3. Enhancement of Voltage Stability
In this case, the proposed GSO approach is applied for en-
where bis a constant and is a normally distributed random hancement of voltage stability, i.e., minimization of L max .
number with mean 0 and standard deviation 1. Here, the population size and maximum iteration number have
The most important control parameter of GSO is the per- been selected 100 and 50, respectively, for this test system. The
centage of scroungers and rangers. optimal values of control variables obtained from the proposed
Figure 1 shows the flowchart of the GSO algorithm. GSO are shown in Table 1. It is seen that the value of L max
obtained from GSO is the lower compared to BBO [19] and
IPSO [20].
4. APPLICATION OF THE PROPOSED METHOD
To verify the effectiveness and performance of the proposed 4.1.4. Improvement of Voltage Profile
GSO method for solving four different objectives OPF prob- In this case, the proposed GSO approach is applied for the
lems, IEEE 30-bus, 57-bus, and 118-bus test systems have improvement of voltage profile. Here, the population size and
been considered. Programs have been written in MATLAB-7 maximum iteration number have been selected 100 and 100,
language (The MathWorks, Natick, Massachusetts, USA) and respectively, for this test system. The optimal values of control
executed on a 3.0 GHz Intel Pentium-IV personal computer. variables obtained from the proposed GSO are given in Table 1.
The results obtained from the proposed GSO method are com- It is seen that voltage deviation obtained from GSO is the lower
pared with those obtained from other evolutionary methods. than that obtained from BBO [19] and the FEA [21].
6 Electric Power Components and Systems, Vol. 00 (2016), No. 00

Control Fuel cost Emission Voltage stability Improvement of


variable minimization minimization enhancement voltage profile
PG1 (MW) 190.40 117.51 139.05 135.17
PG2 (MW) 46.04 70.91 80.00 67.17
PG5 (MW) 20.43 44.34 15.00 49.77
PG8 (MW) 12.99 30.81 10.00 18.11
PG11 (MW) 10.000 25.84 18.58 10.29
PG13 (MW) 12.000 0 29.13 12.22
V1 (p.u.) 1.0500 1.0500 1.0500 1.0500
V2 (p.u.) 1.0338 1.0335 1.0341 1.0337
V5 (p.u.) 1.0058 1.0054 1.0057 1.0062
V8 (p.u.) 1.0230 1.0227 1.0233 1.0237
V11 (p.u.) 1.0913 1.0915 1.0914 1.0910
V13 (p.u.) 1.0400 1.0059 1.0360 1.0393
Fuel cost ($/h) 796.68 868.34 837.96 875.49
Downloaded by [University of Nebraska, Lincoln] at 01:33 27 March 2016

Emission (ton/h) 0.4076 0.1980 0.2929 0.2815


Loss (MW) 8.46 6.00 8.35 9.34
Voltage deviation 0.4369 0.3348 0.9070 0.0627
L max 0.0550 0.0755 0.0256 0.0535

TABLE 1. Optimal value of control variables obtained from GSO for IEEE 30-bus system for different cases

4.2. IEEE 57-bus System


The standard IEEE 57-bus system consists of 80 transmission
lines, seven generators at buses 1, 2, 3, 6, 8, 9, 12, and 15
branches under load tap setting transformer branches. The
reactive power sources are considered at buses 18, 25, and
53. The system line data, bus data, generator data, and the
minimum and maximum limits for the control variables have
been adapted from [28, 29]. The upper and lower limits of
reactive power sources and transformer tap settings are taken
from [30]. The total system active power demand is 12.508 p.u.
and reactive power demand is 3.364 p.u. at a 100-MVA base.
Transmission line limits is set to 100 MVA. In this study, 50
FIGURE 2. Cost convergence characteristics for IEEE 30-bus test runs are performed to solve the OPF problem for different
system. objective functions.

4.2.1. Minimization of Fuel Cost


The proposed GSO approach is applied for minimization of
fuel cost as the objective function. Here, the population size
and maximum iteration number have been selected 100 and
100, respectively, for this test system. The optimal values of
control variables obtained from the proposed GSO are given
in Table 2.

4.2.2. Minimization of Emission


The proposed GSO approach is applied for minimization of
emission as the objective function. Here, the population size
FIGURE 3. Emission convergence characteristics for IEEE and maximum iteration number have been selected 100 and
30-bus system. 100, respectively, for this test system. The optimal values of
Basu: Group Search Optimization for Solution of Different Optimal Power Flow Problems 7

Control Fuel cost Emission Voltage stability Improvement of


variable minimization minimization enhancement voltage profile
PG1 (MW) 574.66 307.87 570.76 489.69
PG2 (MW) 0 0 0 0
PG3 (MW) 107.34 162.34 44.43 51.64
PG6 (MW) 0 0 0 0
PG8 (MW) 122.57 344.64 415.68 540.52
PG9 (MW) 0 0 0 0
PG12 (MW) 485.98 464.52 253.49 200.75
V1 (p.u.) 1.0400 1.0400 1.0400 1.0400
V2 (p.u.) 1.0100 1.0104 1.0102 1.0108
V3 (p.u.) 0.9850 0.9857 0.9855 0.9858
V6 (p.u.) 0.9800 0.9809 0.9803 0.9807
V8 (p.u.) 1.0050 1.0053 1.0046 1.0054
V9 (p.u.) 0.9800 0.9805 0.9807 0.9801
Downloaded by [University of Nebraska, Lincoln] at 01:33 27 March 2016

V12 (p.u.) 1.0150 1.0147 1.0155 1.0156


Cost ($/h) 7658.28 13, 228.76 15, 120.45 21, 889.03
Emission (ton/h) 2.7495 1.7160 2.7821 2.6898
Power loss (MW) 39.7573 28.5720 33.5526 31.7881
Voltage deviation 4.0759 4.9534 1.0722 0.6740
L max 0.2501 0.3664 0.1036 0.1371

TABLE 2. Optimal value of control variables obtained from GSO for IEEE 57-bus system for different cases

control variables obtained from the proposed GSO are given 4.3. IEEE 118-bus System
in Table 2. The standard IEEE 118-bus system consists of 186 transmis-
sion lines, 54 generator buses, 64 load buses, 9 branches under
4.2.3. Enhancement of Voltage Stability load tap setting transformer, and 14 reactive power sources.
In this case, the proposed GSO approach is applied for en- The system line data, bus data, generator data, and the mini-
hancement of voltage stability, i.e., minimization of L max . mum and maximum limits for the control variables have been
Here, the population size and maximum iteration number have adapted from [29, 31]. The upper and lower limits of reac-
been selected 100 and 100, respectively, for this test system. tive power sources and transformer tap settings are taken from
The optimal values of control variables obtained from the pro- [31]. The total system active power demand is 42.4200 p.u.
posed GSO are given in Table 2.

4.2.4. Improvement of Voltage Profile


In this case, the proposed GSO approach is applied for im-
provement of voltage profile. Here, the population size and
maximum iteration number have been selected 100 and 100,
respectively, for this test system. The optimal values of con-
trol variables obtained from the proposed GSO are given in
Table 2.

Best cost Average Worst cost CPU time


Techniques ($/h) cost ($/h) ($/h) (sec)
GSO 68,115.29 68,119.42 68,124.97 108.0753
FIGURE 4. Cost convergence characteristics for IEEE 118-
TABLE 3. Cost minimization of IEEE 118-bus system bus system.
8 Electric Power Components and Systems, Vol. 00 (2016), No. 00

Best Average Worst Average CPU time


emission emission emission CPU time Techniques Best L max L max Worst L max (sec)
Techniques (lb/h) (lb/h) (lb/h) (sec)
GSO 0.0570 0.0574 0.0581 110.8794
GSO 315.8618 317.6841 321.3608 111.0805

TABLE 5. L max minimization of IEEE 118-bus system


TABLE 4. Emission minimization of IEEE 118-bus system

4.3.3. Enhancement of Voltage Stability


and reactive power demand is 14.3800 p.u. at a 100-MVA base. In this case, the proposed GSO approach is applied for en-
Transmission line limits are set to 100 MVA. In this study, 50 hancement of voltage stability, i.e., minimization of L max .
test runs are performed to solve four different objectives by Here, the population size and maximum iteration number
using GSO. have been selected, 200 and 100, respectively, for this test
system. The best, average, and worst L max and average CPU
4.3.1. Minimization of Fuel Cost time among 50 runs of solutions obtained from proposed GSO
Downloaded by [University of Nebraska, Lincoln] at 01:33 27 March 2016

The proposed GSO approach is applied for minimization of are summarized in Table 5. The convergence characteristic ob-
fuel cost as the objective function. Here, the population size tained from proposed GSO for L max minimization is shown in
and maximum iteration number have been selected, 200 and Figure 6.
100, respectively, for this test system. The best, average, and
worst fuel cost and average CPU time among 50 runs of solu- 4.3.4. Improvement of Voltage Profile
tions obtained from proposed GSO are summarized in Table 3. In this case, the proposed GSO approach is applied for im-
The cost convergence characteristic obtained from proposed provement of the voltage profile. Here, the population size
GSO for minimum fuel cost solution is shown in Figure 4. and maximum iteration number have been selected, 200 and
100, respectively, for this test system. The best, average, and
worst voltage deviation and average CPU time among 50 runs
4.3.2. Minimization of Emission
of solutions obtained from proposed GSO are summarized in
The proposed GSO approach is applied for minimization of Table 6. The convergence characteristic obtained from pro-
emission as the objective function. Here, the population size posed GSO for voltage deviation is shown in Figure 7.
and maximum iteration number have been selected, 200 and
100, respectively, for this test system. The best, average, and
worst emission and average CPU time among 50 runs of
solutions obtained from proposed GSO are summarized in
Table 4. The emission convergence characteristic obtained
from proposed GSO for minimum emission solution is shown
in Figure 5.

FIGURE 6. L max convergence characteristics for IEEE 118-


bus system.

Best Average Worst


voltage voltage voltage CPU time
Techniques deviation deviation deviation (sec)
GSO 1.5916 1.5924 1.5936 115.0158
FIGURE 5. Emission convergence characteristics for IEEE
118-bus system. TABLE 6. . Voltage deviation of IEEE 118-bus system
Basu: Group Search Optimization for Solution of Different Optimal Power Flow Problems 9

[7] Yan, X., and Quintana, V. H., “Improving an interior point


based OPF by dynamic adjustments of step sizes and toler-
ances,” IEEE Trans. Power Syst., Vol. 14, No. 2, pp. 709–717,
1999.
[8] Yan, W., Yu, J., Yu, D. C., and Bhattarai, K., “A new optimal
reactive power flow model in rectangular form and its solution
by predictor corrector primal dual interior point method,” IEEE
Trans. Power Syst., Vol. 21, No. 1, pp. 61–67, 2006.
[9] Wang, H., and Murillo-Sánchez, C. E., Zimmerman, R. D.,
Thomas, R. J., “On computational issues of market-based opti-
mal power flow,” IEEE Trans. Power Syst., Vol. 22, No. 3, pp.
1185–1193, 2007.
[10] Lai, L. L., and Ma, J. T., “Improved genetic algorithms for
optimal power flow under both normal and contingent operation
states,” Electr. Power Energy Syst., Vol. 19, No. 5, pp. 287–292,
FIGURE 7. Voltage deviation convergence characteristics for 1997.
Downloaded by [University of Nebraska, Lincoln] at 01:33 27 March 2016

IEEE 118-bus system. [11] Yuryevich, J., and Wong, K. P., “Evolutionary programming
based optimal power flow algorithm,” IEEE Trans Power Syst.,
5. CONCLUSION Vol. 14, No. 4, pp. 1245–1250, 1999.
[12] Abido, M. A., “Optimal power flow using particle swarm op-
In this article, GSO, one of the most important computational timization,” Electr. Power Energy Syst., Vol. 24, No. 7, pp.
intelligence-based techniques, has been successfully applied 563–571, 2002.
[13] Yumbla Onate, P. E., Ramirez, J. M., and Coello, C. A., “Optimal
to solve four different OPF problems. The OPF problem is
power flow subject to security constraints solved with a particle
formulated as a non-linear optimization problem with equality swarm optimizer,” IEEE Trans. Power Syst., Vol. 23, No. 1, pp.
and inequality constraints of the power system. In this study, 33–40, 2008.
different objective functions such as fuel cost minimization, [14] Roa-Sepulveda, C. A., and Pavez-Lazo, B. J., “A solution to the
emission minimization, and enhancement of the voltage pro- optimal power flow using simulated annealing,” Electr. Power
file and voltage stability are considered. The proposed GSO Energy Syst., Vol. 25, No. 1, pp. 47–57, 2003.
[15] Bakirtzis, A. G., Biskas, P. N., Zoumas, C. E., and Petridis,
approach is tested on IEEE 30-bus, 57-bus, and 118-bus test V., “Optimal power flow by enhanced genetic algorithm,” IEEE
systems to demonstrate its effectiveness. The results obtained Trans. Power Syst., Vol. 17, No. 2, pp. 229–236, 2002.
from the proposed GSO approach are better than the results [16] Ongsakul, W., and Tantimaporn, T., “Optimal powers flow by
obtained from other evolutionary methods reported in the lit- improved evolutionary programming,” Electr. Power Compon.
erature. Syst., Vol. 34, No. 1, pp. 79–95, 2006.
[17] Sayah, S., and Zehar, K., “Modified differential evolution algo-
rithm for optimal power flow with non-smooth cost functions,”
REFERENCES Energy Convers. Manage., Vol. 49, No. 11, pp. 3036–3042,
2008.
[1] Carpentier, J., “Contribution a l’etude du dispatching [18] Abou El Ela, A. A., Abido, M. A., and Spea, S. R., “Opti-
economique,” Bull. Soc. Franc. Electr., Vol. 3, pp. 431–474, mal power flow using differential evolution algorithm,” Electr.
1962. Power Syst. Res., Vol. 80, No. 7, pp. 878–885, 2010.
[2] Dommel, H., and Tinney, W., “Optimal power flow solution,” [19] Bhattacharya, A., and Chattopadhyay, P. K., “Application of
IEEE Trans. Power Appar. Syst., Vol. PAS-87, No. 10, pp. biogeography-based optimization to solve different optimal
1866–1876, 1968. power flow problems,” IET Gener. Transm. Distrib., Vol. 5, No.
[3] Burchett, R. C., Happ, H. H., and Vierath, D. R., “Quadratically 1, pp. 70–80, 2011.
convergent optimal power flow,” IEEE Trans. Power Appar. [20] Niknam, T., Narimani, M. R., Aghaei, J., and Azizipanah-
Syst., Vol. PAS-103, No. 11, pp. 3267–3276, 1984. Abarghooee, R., “Improved particle swarm optimization for
[4] Sun, D. I., Ashley, B., Brewer, B., Hughes, A., and Tinney, W. multi-objective optimal power flow considering cost, loss, emis-
F., “Optimal power flow by Newton approach,” IEEE Trans. sion and voltage stability index,” IET Gener. Transm. Distrib.,
Power Appar. Syst., Vol. PAS-103, No. 10, pp. 2864–2875, Vol. 6, No. 6, pp. 515–527, 2012.
1984. [21] Surender Reddy, S., Bijwe, P. R., and Abhyankar, A. R., “Faster
[5] Tinney, W. F., and Hart, C. E., “Power flow solution by Newton’s evolutionary algorithm based optimal power flow using incre-
method,” IEEE Trans. Power Appar. Syst., Vol. PAS-86, No. 11, mental variables,” Electr. Power Energy Syst., 54, No. 1, pp.
pp. 1449–1460, 1967. 198–210, 2014.
[6] Mota-Palomino, R., and Quintana, V. H., “Sparse reactive power [22] He, S., Wu, Q. H., and Saunders, J. R., “Group search optimizer:
scheduling by a penalty-function linear programming tech- An optimization algorithm inspired by animal searching behav-
nique,” IEEE Trans. Power Syst., Vol. 1, No. 3, pp. 31–39, ior,” IEEE Trans. Evol. Computat., Vol. 13, No. 5, pp. 973–990,
1986. 2009.
10 Electric Power Components and Systems, Vol. 00 (2016), No. 00

[23] Shen, H., Zhu, Y., Zou, W., and Zhu, Z., “Group search [28] “The IEEE 57-bus test system,” available at: http://www.ee.
optimizer algorithm for constrained optimization,” Comput. washington.edu/research/pstca/pf57/pg tca57bus.htm
Sci. Environ. Eng. Eco. Inf. (2011) (CSEEE 2011), Part II, [29] “Index of data,” available at: http://motor.ece.iit.edu/data
Berlin, Heidelberg: Springer-Verlag, CCIS 159, pp. 48–53, [30] Duman, S., Sonmez, Y., Guvenc, U., and Yorukeren, N., “Opti-
2011. mal reactive power dispatch using a gravitational search algo-
[24] Moradi-Dalvand, M., Mohammadi-Ivatloo, B., Najafi, A., rithm,” IET Gener. Transm. Distrib., Vol. 6, No. 6, pp. 563–576,
and Rabiee, A., “Continuous quick group search opti- 2012.
mizer for solving non-convex economic dispatch problems,” [31] “The IEEE 118-bus test system,” available at: http://www.ee.
Electr. Power Syst. Res., Vol. 93, pp. 93–105, December, washington.edu/research/pstca/pf118/pg tca118bus.htm
2012.
[25] Kessel, P., and Glavitsch, H., “Estimating the voltage stability
BIOGRAPHY
of a power system,” IEEE Trans. Power Del., Vol. 1, No. 3, pp.
346–354, 1986. Mousumi Basu received her Bachelor’s degree, Master’s de-
[26] Mustard, D., “Numerical integration over the n-dimensional gree, and Ph.D. from Jadavpur Universty, Kolkata, India, in
spherical shell,” Math. Computat., Vol. 18, No. 88, pp. 578–589,
1964.
1991, 1993, and 2003, respectively. She is currently with
Downloaded by [University of Nebraska, Lincoln] at 01:33 27 March 2016

[27] Mahadevan, K., and Kannan, P. S., “Comprehensive learning the Department of Power Engineering, Jadavpur University,
particle swarm optimization for reactive power dispatch,” Appl. Kolkata, India. Her research is focused on power system opti-
Soft Comput., Vol. 10, No. 2, pp. 641–652, 2010. mization and soft computing techniques.

Potrebbero piacerti anche