See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/308853614

Particle Swarm Optimization Based Optimal Reactive Power Dispatch

Conference Paper · March 2015

DOI: 10.1109/ICECCT.2015.7225981

CITATIONS READS

89 197

2 authors, including:

Pandya Sundaram
Shri K .J. Polytechnic
16 PUBLICATIONS 184 CITATIONS

SEE PROFILE

All content following this page was uploaded by Pandya Sundaram on 27 April 2021.

The user has requested enhancement of the downloaded file.

 Particle Swarm Optimization Based

Optimal Reactive Power Dispatch

Sundaram Pandya Ranjit Roy

Electrical Engineering Department Electrical Engineering Department
Sardar Vallabhbhai National Institute of Technology Sardar Vallabhbhai National Institute of Technology
Surat, India Surat, India
[email protected] [email protected]

Abstract- Optimal reactive power dispatch is one of the new functions which is derived based on a local voltage
major and important optimization problem in electrical power stability index,called ,DSY related with voltage stability
system operation and control. This is nothing but multi margin.The new objective function is tested on the New
objectives, nonlinear, minimization problem of power system England 39- bus test system. [4]
optimization. This paper focuses on this sub problem of optimal
power flow. The particle swarm optimization (PSO) is one of the Some researchers are dealing with a multi objective
best population based intelligent technique of optimization. The optimization problem with newly developed teaching learning
basic PSO algorithm is used for optimal reactive power dispatch. based optimization (TLBO) algorithm to solve ORPD problem
This approach applies to the IEEE-14 bus, IEEE-30 bus, IEEE- by minimizing the objective functions.[5-S].M.Varadarajan
57 bus and IEEE-39 New England bus test systems for et.al. presents the ORPD problem as a mixed integer nonlinear
minimization of active power loss. Simulation results are optimization problem with penalty coefficients for constraint
compared with the other optimization algorithm. handling. The algorithm is tested on standard IEEE-14, IEEE-
30 and IEEE-lIS bus test system [9].
Keywords- reactive power dispatch, optimal power flow,
particle swarm optimization,metaheuristic methods
II. OPTIMAL REACTIVE POWER DISPATCH PROBLEM

FORMULATION
I. INTRODUCTION

For the last several decades, so many researchers focused A. Optimal reactive power dispatch
on the optimal operation of the power system. This is because Optimal reactive power dispatch [10-21] is a non-linear,
of demand of power system is increasing drastically. Also, non-convex optimization problem. The main goals of ORPD
there is increasing renewable energy sources required proper problem is to minimize the active power loss of the system
planning of the system. For reliability and efficiently working and to keep the voltage profile within the certain limits
of system power losses and voltage deviation should be followed by the equality and inequality constraints. The multi­
rmmmum. objective optimal reactive power planning problem seeks to
The optimal power flow problem was first introduced by optimize one or more objective function while satisfying their
Carpentier in 1960s [l] and after then many researchers have defined constraints. In optimal reactive power planning all
worked on the OPF problem with the different methodology. control variables are set in such a way that, the system
Optimal reactive power dispatch problem is actually sub parameters are to be optimized. The optimal location and their
problem of optimal power flow (OPF).This ORPD is also values are also coming into the picture of the reactive power
multi objectives, nonlinear in nature. planning problems.

A.H.Khazali.et.al.presents the harmony search algorithm B. Problem Formulation

for ORPD problem. In this paper objective functions are
optimized separately and the inequality constraints are handled The mathematically optimal reactive power dispatch
by penalty coefficients. The study is implemented on IEEE 30 problem is defined as follows
and IEEE 57 bus system and the results are compared with a Min f (x, u) (1)
simple genetic algorithm (GA) and particle swarm S.t g (x, u) 0
=

optimization (PSO).[2-3]. h (x,u)::; 0

Abbas Rabiee.et.al deals with the management of on load Where f is the objective function to be optimized, g is the
tap changes (OLTC) and dynamic VAR sources to improve equality constraints which are nothing but nonlinear power
voltage stability margin (VSM) of power system and proposed flow equations, and h is the power system operating

978-1-4799-6085-9/15/$31.00 ©2015 IEEE

constraints. Also, u is the vector of independent control As the system reaches its voltage collapse point,
variables including: when both active and reactive power losses increases rapidly.
There are close relationship between the voltage stability and
1. Generator active power output Pe; except at slack bus Pm.
reactive power of the system. As voltage stability indicator L­
2. Generator bus voltage Ve;. index is minimized, system voltage stability is increased. The
L-index shows that the system is how many close to the
3. Online transformer tap setting T. voltage collapse. The minimization of the L-index is one of
4. Reactive power compensation Qc. the objective functions of the optimal reactive power dispatch
problem. The L-index is briefly discussed as follows:
Hence, u can be expressed as:
Consider Nc is the number of generators in N bus
uT= [PG2 ....Pe;Ng'VGI....VGNg,TI.....TNl,QCI ...QCNcl (2)

[ ][ ][ ]
system .The currents and voltages are related as

where Ng, Nt and NC signify the number of generating IG YGG YGL VG (6)
units, online tap setting transformers and reactive shunt =
IL YLG YLL VL
compensators, respectively. The active powers(excluding slack
bus) and bus voltages of generators are treated as continuous Where IG,lL and VG,VL are currents and voltages of the

[ ]-[ ][ ]
variables.while, transformer tap setting and shunt reactors are generator buses and load buses
denoted as discrete variables. Also, x is the vector of dependent VI Zu. F1G /I
(7)
variables including:
1 K GI YGG VG
I. Active power output, generated by slack bus PGI. G

2.The voltage generated by load buses VL.

Here, F LG =
- [y LL ]-1 [Y LG ]
The L-index of the j'h node is given by the expression as,
3. Reactive power output of generator QG.

I I
Ng v.
L j I - I F - . L ((J ij + c5 i
- c5 j ) (8)
4. Line loading capacity of transmission lines SL. = '

i=1 V i
t
Hence, x can be expressed as: where Vi is the voltage magnitude of i l1 generator bus, Vj is
the voltage magnitude of jth PQ bus, 8ij is the phase angle of
xT= [PGI,VLl ....VLNpq,QGI ....Qe;Ng,SLl .....SLNll (3)
the term Fji, 8 i is the voltage phase angle of ith generator unit,
8j is the voltage phase angle of jth generator unit, and Npq is
Where Npq is the number of PQ buses, and NI is the total
the number of PQ buses. For all load buses ,calculate L-index.
number of transmission lines.
As mentioned above,the value of L-index is lies between zero
C. Objectivefunctions to one for each load bus. So, a global system L-index
describing the voltage stability of the entire system can be
The different objective functions in different literatures
defined by:
are:-
f3=L = max ( Lj ) j=1,2,3 ...........,Npq (9)
(I) The active power loss minimization
The aim of this objective function is to minimize the active The lower the value of L-index,larger the voltage stability
power loss of the system and defined as margin. Hence, in order to increase the voltage stability of the
power system and to move the system far from the voltage
. Nt! 2 2
fl= mlll (Pr) =I Gk(V.1 + V}· - 2V·V .cos ( 6· - 6 .) (4) collapse point, the voltage stability indicator L defined can be
I } 1 }
k=! employed.
where PL is the active power 10ss;Gk is the conductance of
t
the line connected between the i h and j'h lines; Vi and Vj are D. Constraints
the voltage magnitudes at bus i and bus j, respectively; 8 i and
The constraints considered in the literatures for optimal
8j are the voltage angles at bus i and bus j, respectively
reactive power dispatch problem are:-
(2) Improvement of voltage profile
(i) Equality constraints:-
The voltages of the system must be within their
permissible limits for reliable operation of the power system. The equality constraints g are typical power flow equations
So the voltage deviation of the system is within the tolerable which are defined as follows:
limits. Hence, the objective function is defined as Pe;i-PDi-Vi (Gij cos(8i-8j) + Bij sin(8i-8j)) = 0 i=1,2, .....,N(I 0)
QGi-Q Di-Vi (Gij sin(8i---bj)+ Bij cos (8i-8j)) = 0 i=I,2, .....,N(lI)
(5)
i =1 where N is the number of buses in the system. PGi and
where Npq is the number of load buses;Vi is the voltage of QGi are respectively the active and reactive power generations
t
i l1bus. at bus i; PDi and QDi are the active and reactive power loads
at bus i, respectively; Gij and Bij are the real and imaginary
(3) Improvement of voltage stability index parts of the ijth element of the bus admittance matrix,
respectively; Vi and Vj are the voltage magnitudes at bus i and
bus j, respectively; oi and oj are the voltage angle at bus i and t t
x:' : The position of i h particle at k h iteration
bus j, respectively.
X : constriction factor
(ii) Inequality constraints:­
The inequality constraints are:- Here the weight factor 'w' is changing at iteration to
iteration provides the balancing between particle best and
(A) Generator constraints: Generator active power PG, global best solution according to the following equation:
generator reactive power QG, and generator voltage magnitude
VG are restricted by their lower and upper limits: w max -
w mm
·

w =
w max -
. X iter (21)
PGi.min::; PGi ::; P Gi.max i=l, .......Ng (12) I termax
QGLmin::; QCJi ::; QCJLmax i=l, .......Ng (13)
where W max :Inertia weight at beginning of iterations
VGi.min::; VGi ::; VGi.max i=l, .......Ng (14)

Wmin :Value of iteration weight at the end of iterations

(B) Transformer constraints: Transformer taps have minimum
and maximum setting limits:
iter : Current iteration number
i=l, .......Nt (15)
itermax : Maximum number of iterations
(C) Switchable VAR sources: The switchable VAR sources
have restrictions as follows: IV. IMPLEMENTATION OF PSO FOR OPTIMAL REACTIVE POWER

i=I, .......Nc (16) DISPATCH

Qci,min::; Qci ::; QCi,max
The proposed algorithm can quickly and fast converge in
(D) Security constraints: These include the limits on the load the near about the best solutions for optimal reactive power
bus voltage magnitudes and transmission line tlows limits: dispatch problem. The steps of PSO algorithm are as follows.
VLi.min::; VLi ::; VLi.max i=I, .......Npq (17) Step 1) Between upper and lower limits randomly generate
SLi ::; SLi,max i=I, ........,NI (18) particles.

III. THE PROPOSED PSO ALGORITHM Step 2) Assign the initial agents (particles) as particle best Phes1
values.
Particle swarm optimization is simple, fast and population
based algorithm. This was proposed by Eberhart and Kennedy Step 3) Evaluate the objective function with particle best Phest
[22-23] in year 1995.lt has been motivated by the behavior of and assigned ghest among all the Pbcst.
groups such as fish schooling and bird tlocking. Because of
Step 4) Change the velocity and position using equations (19)
inherent simplicity, it is fast and requires less memory. The
and (20) for each particle.
swarm initially has population of random velocity and is flown
through the problem space. All particles have a memory and Step 5) Compare the objective function for each particle with
track of best position (pbest) and the corresponding fitness that of its Phes1 . If it is better than pbcst. set current value as Phes1
value. The swarm has another best value called (gbest) which and locate it as a current location in the search area.
is the best value of all particle' pbest.All particles update its
position based upon its position, global best position and Step 6) According to fitness value, find the best agent and
previous velocity according to the following equations: reset as global best ghest.
Step 7) Repeat steps (4) to (6) till the specified iterations are
(19) met.
k k+1
X ;+l = Xi +X X Vi (20)
th th V. NUMERICAL RESULTS
where v {'+1 :The velocity of i particle at (k+1) iteration
To verify the effectiveness and efficiency of the proposed
w:Inertia weight of the particle algorithm, different IEEE bus test systems are used as test
systems. The PSO algorithm has been implemented in
k 1 t
Vi :The velocity of i h particle at k h iteration MATLAB programming language incorporated with
matpower 3.2 [24] package and numerical tests are carried out
CbC2 :Positive constants between 0-2.5 on a core i5, 3.20GHz, 4 GB RAM computer.
r l h :Random numbers between 0-1
A. IEEE-14 Bus power system
1
PbcsLi :Best position of i h particle The IEEE-14Bus system and system data and other
ghest :Global best position operating conditions are given in the [25]. The network
consists 20 branches, 5 generator-buses, 3 numbers of
k+l . . f· ·t . l . .
Xi :
The posltIOn 0 I h patlcIe at (k +l) h IteratIOn adjustable transformers and 2 reactive power compensators.
The IEEE-14 Bus contains 10 control variables and their
settings limits are shown in Table I.
 TABLE. 1 SETTINGS OF CONTROL VARIABLES. 8 -10 40
30 Bus 11 -6 24
IEEE BUS TEST SYSTEM Variables Min Max
13 -6 24
Transformer tap settings l.05 0.95
Generator node 1.1 0.9
14 Bus TABLE.5 COMPARATIVE RESULTS OF IEEE-30 BUS SYSTEM.
Qc.9(p.u) 0.20 0
Qc-14 0.20 0
CONTROL VARIABLES PSO EPI261 SARGAI261

VGl l.0995 - -
TABLE. 2 CONSTRAINTS OF STATE VARIABLES.[18]
VG2 1.0778 l.097 l.094
IEEE BUS TEST Generation QMin QMax l.049 l.053
VG5 1.0459
SYSTEM Variables
VG8 1.0576 1.033 1.059
1 0 10
2 -40 50 VGll l.0526 l.092 l.099
14 Bus 3 0 40 VG13 l.0330 l.091 l.099
6 -6 24 TI 1.0050 l.01 0.99
8 -6 24
T2 l.0239 l.03 l.03

T3 l.0183 1.07 0.98

Table.2 shows the limits of reactive power in MW of
T4 0.9927 0.99 0.96
generation variables. The simulation results are shown in
QC3 0.1964 0.19 0.19
Table.3 with the comparison of EP and SARGA algorithm.
QCIO 0.1840 0.04 0.04

QC24 0.1315 - -
TABLE.3 COMPARATIVE RESULTS OF lEEE-14 BUS SYSTEM.
TOTAL ACTIVE 16.1810 16.38962 16.09292
POWER LOSS MW MW MW
CONTROL PSO EP[26] SARGA[26]
VARIABLES DE[18] DE-ABC[18] ABC[18]

VGl l.0917 - - 16.3176MW 16.2163MW 16.5908MW

VG2 1.0862 1.029 1.06
VG3 1.0876 1.016 1.036
Total Active power losses (MVV)
VG6 1.0577 1.097 1.099 19 ,-
-.--�--,---.--,
VG8 1.0278 1.053 1.078
T4-7 1.0039 1.04 0.95 185
T4-9 1.0324 0.94 0.95
T5-6 l.0083 l.03 0.96
Qc9 0.0513 0.18 0.18
Qc14 0.1920 0.06 0.06
TOTAL ACTIVE 12.9691 MW 13.34620 MW 13.21643 MW
POWER LOSS

:t 17
B. iEEE -30 Bus system and iEEE- 57 Bus test systems ro
15
>--

The bus data and line data of IEEE-30 and IEEE-57 Bus 16.5
test systems are taken from [25].TableA shows the state
variables constraints. There are 6 generator nodes,4
16 L--=
O--�
6�O--= --
O�1�6 0�� 18=O--=
transformer tapsettings and 3 shunt reactive devices 0 20�- 4� 1 4�
80��1�OO�� 12=O � 200
Iteration number
connected in 30 bus system.So 13 control variables in IEEE
Fig.1 Convergence characteristics ofTEEE-30 Bus system
30- Bus test system. Similarly IEEE-57 Bus test systems
having 27 control variables as tabulated in Table.6.and
TABLE.6 RESULTS OF IEEE-57 BUS TEST SYSTEM.
compared [18] with other cited results in references.
Convergence characteristic and comparison results of IEEE-30 CONTROL PSO DE DE-ABC ABC
VARIABLES
Bus system are shown as below. This result also compares
VGI 1.0430 1.0555 1.0576 1.0708
with EP,SARGA and other optimization algorithms[26].Here
VG2 1.0510 1.0409 1.0426 1.0660
penalty function method is used for controlling the variables 1.0214 1.0191
VG3 1.0370 1.0606
within their maximum and minimum limits. The maximum or VG6 l.0218 0.9993 l.0000 l.0493
minimum limits of variables are rounded off for converting VG8 l.0383 l.01l0 l.0147 l.0714
discrete variables into continuous variables. There are 17 VG9 1.0159 0.9872 0.9892 1.0302
transformer tap setting and 3 shunts, so 20 discrete control VGI2 1.0438 1.0036 1.0013 1.0339
variables are used in IEEE-57 bus test system. Tl9 l.0434 l.2 1.1 0.98
T20 0.9598 l.06 l.04 l.08
T31 1.0330 1.08 1.02 1.02
TABLEA LIMITS OF STATE VARIABLES. T35 1.0381 1.06 0.94 I
T36 0.9598 0.92 0.98 l.04
IEEE BUS TEST Variables QMin QMax
T37 l.0068 0.92 l.08 1
SYSTEM
T41 0.9642 1.08 1.0 0.92
1 0 10
T46 1.0093 0.94 0.9 0.98
30 Bus 2 -40 50
T54 1.0361 1.08 0.92 1.04
5 -40 40
T58 0.9580 1.0 1.0 0.98
 T59 0.9965 1.1 1.04 1.06 [6] AAAbou El Ela,M AAbidos,S. R .Spea " Differential evolution
T65 1.0072 1.08 0.92 1.04 algorithm for optimal reactive power dispatch"Electrical power system
T66 0.9578 1.02 0.92 0.94 research vo1.81(2011) 458-464.
T71 1.0191 1.0 0.98 1.1 [7] Barun Mandal,Provas kumar roy "Optimal reactive power dispatch
T73 0.9583 1.0 1.1 1.1 using quasi oppositional teaching learning based algorithm" Electrical
T76 1.0491 1.02 1.02 1.04 power and energy systems vo1.53 (2012) 123-134..
T80 1.0432 0.92 1.02 0.98 [8] RMallipeddi,S.Jeyadevi,S.Baskar "Efficient constraint handling for
Qc18 0.1845 0.05 0 0.10 optimal power reactive power dispatch problems"Swarm and
Qc25 0.0390 0.18 0.18 0.18 evolutionary computation 5 (2012) 28-38.
Qc53 0.0532 0.18 0.18 0.18 [9] M.Varadarajan,KS.Swarup "Differential evolution approach for
TOTAL ACTIVE 26.5140 25.6228 25.5634 24.1846 reactive power dispatch"Applied soft compiting vo1.8(2008)1549-
POWER LOSS MW MW MW MW 1561.

[10] Xiong Hugang,Cheng haojang.Li Haiyu "Optimal reactive power flow

C. lEEE-39New England Bus test system. incorporating static voltage stability based on multi objective adaptive
imune algorithm"Energy vol.49 (2008)1175-1181.
The IEEE-39 bus test contains 46 lines, 10 generators, 6 Ali Ghasemi,Akbar Tohidi "Multi objective optimal reactive power
[11]
reactive compensator devices, 5 online tap setting dispatch using a new multi objective strategy" Electrical power and
transformers [4] .Table.7 shows the result of simulations as energy systems vo1.57 (2014) 318-334.
below. [12] Chaohua Dai,Weirang Chen,Xuexia zhang "Reactive power dispatch
considering volatge stability with seeker optimization
algorithhm"Electrical power system reasearch vol.79 (2009)1462-
TABLE.7 SIMULATION RESULTS OF 39-BUS SYSTEM. 1471.

CONTROL VARIABLES PSO CONTROL PSO [13] M.Granada,Mcros J Rider) R S Mantovani "A decentralized approach
VARIABLES for optimal reactivepower dispatch using a lagrangian decomposition
VG30 1.0496 T l 2-11 0.9917 method" Electrical power system reasearch vol.89 (2012)148-156.

VG31 1.0888 TIO-32 1.0394 [14] T.Malakar,S.KGoswami "Active and reactive power dispatch with
VG32 1.0931 T22.35 1.0304 minimum control movement" Electrical power system reasearch vol.44
VG33 1.0431 T2-30 1.0039 (2013)78-87.
VG34 1.0944 T19-20 0.9857 [15] Houssem Ben Aribia ,Nizar Derbel"The active reactive complete
VG35 1.0268 Qc l 0.1710 dispatch on electrical network" Electrical power system reasearch
VG36 1.0856 Qc5 0.1428 vol.44 (2013) 236-248.
VG37 1.0925 Qc l l 0.1248 [16] D.Devraj,.T.Pretha Roselyn "Genetic algorithm based reactive power
VG38 0.9767 Qc l 4 0.1094 dispatch for voltage stability improvement" Electrical power system
VG39 1.0435 Qc22 0.0841 reasearch vol.32 (2010) 1151-1156.
Qc27 0.1449 Kurshat Ayan,Ulas,kilik "Artificial Bee colony algoritm solution for
[17]
TOTAL ACTIVE POWER 41.8892 MW optimal reactive power tlow" Applied soft computing Vol.12 (2012)
LOSS 1477-1482

[18] Yuancheng Li,Yiliang Wang,Bin Li "A hybrid artificial Bee colony

VI. CONCLUSIONS
assisted diiferential evolotion algorithm for optimal reactive power
In this paper, Particle Swarm Optimization algorithm is flow". Electrical power system reasearch vol.52 (2013) 25-33

implemented for optimal reactive power dispatch problem. [19] Amit Saraswat, Ashish Saini "Multi objective optimal power dispatch

The ability of this algorithm has been assessed by testing on considering voltage stability in power system using
HFMOEA."Engineering application of intelligence application Vol.26
IEEE-14 Bus, IEEE-30 Bus, IEEE-57 Bus and IEEE-39 New
(2013) 390-404
England Bus systems. The obtained results are compared with
[20] RHe,G.A.Taylor,YH.Song "Multi objective optimal reactive power
other methods reported in given references. It is concluded fow including voltage security and demand profile classitlcation".
that, the obtained results in this paper are better than the Electrical power system reasearch vol.30 (2008) 327-336
results obtained in other reported papers. [21] A.lashkarAra,A.Kazemi,S.Gahramani,M.Behshad"Optimal reactive
power flow using multi objective mathematical programming".Scientia
Iranica D volume-19 (2012) 1829-1836
REFERENCES
[22] Kennedy .T. and Eberhart R, "Particle swarm optimization", In Proc. of
[1] J.Carpentier,Cotribution a Etude du Dispatching Economique.Bulletin
the IEEE international conference on Neural networks, 1995, pp. 1942-
Societe Francaise des Electriciens 8 (1962) 431-447. 1948.
[2] S.Patra,S.KGoswami "Handling inequalities and discrete variables in B. Zhao et aI., "A Multiagent-Based Particle Swarm Optimization
[23]
newton optimal power flow using optimal multieplier and fuzzy based Approach for Optimal Reactive Power Dispatch", IEEE Trans. On
limit enforcement and relaxation technique" Electrical power and Power Systems, vol. 20, no. 2, May 2005, pp. 1070-1078.
energy systems vo1.42 (2012) 359-366.
[24] RD. Zimmerman et al. ,"MATPOWER: Steady-State Operations,
[3] AH.Khazali,M.Kalantar"Optimal Reactive power dispatch based on
Planning and Analysis Tools for Power Systems Research and
harmony search algorithm" Electrical power and energy systems vo!.33 Education," IEEE Trans. On Power Systems, vol. 26, no. I, Feb. 20II,
(2011) 684-692.
pp. 12-19.
[4] Abbas Rabiee,MaziarVanouni,MostafaPamiani "Optimal power case
[25] Power system test archive;
dispatch for improving voltage stability margin using a local voltage 2012.hUp:/Iwww.ee.washington.edulresearch/pstca!.
stability index"Energy conversion and management voI.59(2012) 66-
[26] P.Subbaraj,P.N .Rajnarayan "Optimal reactive power dispatch using
73.
self adaptive real coded Genetic algorithm"Electrical power system
[5] S.Jeyadevi,S.Baskar,C.K.Babulal,Willjuice iruthayarajan "Solving
reaserch vol.79 (2009) 374-381.
multi objective optimal power dispatch using modified NSGA- II"
Electrical power and energy systems vol.33 (2011) 219-228.

View publication stats