You are on page 1of 6

A new colliding bodies optimization for solving

optimal power flow problem in power system


*Harish Pulluri1, R. Naresh2, Veena Sharma3, Preeti4
Electrical Engineering Department, National Institute of Technology, Hamirpur, India
Email: (*1harishpulluri, 2rnareshnith, 3veenanaresh, 4preetieednith) @gmail.com

Abstract In the present article, a new heuristic algorithm Therefore, evolutionary algorithms are developed to alleviate
colliding bodies optimization (CBO) is applied to solve optimal the drawbacks in conventional methods and to provide near
power flow (OPF) problem in an electrical power system. CBO is optimal solution such as, genetic algorithm (GA) [4],
simple in structure and does not have any control parameters. differential evolution (DE) [5], artificial bee colony (ABC)
Recently, the CBO is applied on different mathematical and [6], particle swarm optimization (PSO) [7], biogeography
structural problems owing to its appreciable efficacy other than based optimization (BBO) [8], black hole based optimization
the several existing algorithms. The performance and efficiency (BHBO) [9], and league championship algorithm [10].
of the proposed CBO method are demonstrated on IEEE 30-bus
system with different objective functions. Results obtained with Recently, CBO [11] is a newly proposed heuristic
the proposed CBO method is compared with other methods technique developed by Kaveh and Mahdavi. It is a multi-agent
present in the recent literature. From these results, it can be algorithm inspired from the natural phenomenon of one-
concluded that the CBO algorithm provides better solutions for dimensional collision between two objective bodies. In which
different OPF problems. every object is modeled, as a colliding body (CB) with
predefined velocity and mass. A collision between any two pair
Keywords colliding bodies optimization; heuristic algorithm; of objects and its new positions are updated with new velocities
optimal power flow. based on the collision laws. The key feature of CBO is that
does not have any tuning variables and simple in structure. The
I. INTRODUCTION CBO algorithm is tested on various practical world
Optimal power flow (OPF) is one of the important tools for optimization problems [12,13] and the results are proven that
power system planning and operation control. Dommel and the CBO is more effective and superior to solve the
Tinney first defined OPF in 1960. After that, this problem has optimization problems. In the present article, CBO is applied to
drawn considerable attention of many researchers. The main solve the OPF problems of IEEE 30-bus system with different
aim of OPF is to minimize the given objective function by objective functions.
adjusting the some of the independent variables to satisfy the The rest of the paper is arranged as below: Section II
set of various diverse constraints related to the power system describes the OPF problem formulation. Section III discusses a
network. The independent variables include generator buses, brief introduction of collision laws and proposed CBO method
real power outputs except slack bus, generator buses voltage to solve OPF problems. Section IV explains the simulation
magnitudes, shunt var compensators, and transformer taps results of different objective functions and finally, conclusions
connected between various buses. The equality constraints are are given in Section V.
the nonlinear power flow equations, and inequality constraints
of the power system that include load buses voltage
magnitudes, the line flows, slack bus active power output, and II. MATHEMATICAL FORMULATION
generator reactive power limits [1]. Minimization of total fuel In general, the goal of OPF problem is to minimize a
cost is the most commonly considered objective function. To chosen objective function while satisfying a set of equality and
improve the reserve capacity minimization of transmission loss inequality constraints by adjusting the some of control
is considered as a part of the OPF problem. Moreover, due to variables. Mathematically, it is described as below:
ever increase in the electricity demand with unequal generation
and transmission capacity expansion, voltage instability min f ( x, u ) (1)
evolving a new challenge for power system operation and
control. In some cases due to lack of reactive power sources, g ( x, u ) = 0
subjected to (2)
the voltages of the heavily loaded buses may be less than the h( x, u ) 0
normal operating limits which lead to voltage collapse in the
network. In such cases, voltage profile improvement and where f is an objective function to be optimized; x is
voltage stability enhancement are considered as objectives of vector of state variables containing slack bus active power
the OPF problem. output ( Pg1 ), load buses voltage magnitudes ( Vl ), MVA flows
The literature on OPF is considerably high. Various in transmission lines ( Sl ) and reactive power of generator
conventional methods [2, 3] have been successfully applied to buses ( Qg ), and mathematically formulated as follows:
finding the solution of OPF problem. But, these all these
techniques are generally based on differentiation and easily
trapped in local optimum solution or may divergent also. x = Pg1 , Vl1 Vlnl , Sl1 Slntl , Qg1 Qgng (3)

978-1-5090-0128-6/16/$31.00 2016 IEEE


where ng , nl , and ntl represents the number of ng

generators, transmission lines and load buses respectively. H km mV


Lk = 1 m =1 where m = 1, 2, , ng (9)
u is vector of independent variables comprising of Vk
generators real power output except slack bus, generator buses
voltage magnitudes ( Vg ), shunt var compensators ( bsh ), and
transformer tap settings connected between various buses ( t ). H km = [inv(Ykk )] * [Ykm ] (10)
u can be expressed as follows:
where Lk is VSI value of the k th load bus, Ykk is self-
u = Pg2 , , Pgng , Vg1 , , Vgng , t1 , , tnt , b sh1 , b shnc (4) admittance of k th load bus, Ykm is mutual admittance between

where nt , and nc represents the number of transformer k th and mth buses.


taps shunt var compensators respectively.
B. Constraints
In this study, various objective functions are adopted to
show the effectiveness of the CBO approach, which can be The equality constraints g ( x, u ) are nonlinear power flow
expressed as follows: equations, which are expressed as below:
nb

A. Objective Functions Pgk Pdk Vk Vm ( Gkm cos km + Bkm sin km ) = 0 (11)


m =1
1) Minimization of total fuel cost (TFC): The objective is
to minimize the TFC of all the thermal generators in a power nl

system, it is expressed as follows: Qgk Qdk Vk Vm ( Gkm sin km Bkm cos km ) = 0 (12)
m =1
ng
min f1 = ( ak + bk Pgk + ck pgk
2
) (5) where Pgk , Qgk are active and reactive power generations
k =1
of the k th generator respectively; Pdk , Qdk are active and
where f1 is the total fuel cost of all the generators, reactive power demands of k th bus respectively; Vk , Vm are
th
ak , bk , c k represent the cost coefficients of the k voltage magnitudes of k th and mth buses respectively;
th
generator. Pgk is the active power generation at k generator. Gkm , Bkm are conductance and susceptance between k th and

2) Voltage profile improvement: The aim decreases the mth buses respectively.
sum of voltae deviation of all the load buses from reference The inequality constraints h( x, u ) are defined within their
voltage.
prescribed limits, which are expressed as follows:
nl
min f 2 = (Vk Vref ) (6) Pgkmin Pgk Pgkmax
k =1 min
Vgk Vgk Vgk
max
k = 1, 2, , ng (13)
where f 2 is the sum of voltage deviation of all the load min max
Qgk Qgk Qgk
buses; Vk represents the voltage magnitude of the k th bus;
tkmin tk tkmax k = 1, 2, , nt (14)
and Vref =1 p.u. min max
b shk bsh b shk k = 1, 2, , nc (15)
3) Minimization of active power loss: Mathematically, V min
Vlk V max
k = 1, 2, , nl (16)
lk lk
active power loss is expressed as given below:
max
Slk S lk k = 1, 2, , ntl (17)
ntl
min f 3 = Gn (Vk2 + Vm2 2VkVm cos km ) (7)
n =1
where, Pgkmin , Pgkmax are the minimum and maximum active

where f 3 is the active power loss of all transmission lines, power limits of the k th generating unit respectively; Vgkmin ,
Gn is conductance of nth transmission line, Vk , Vm are voltage Vgkmax are minimum and maximum voltage limits of the k th
magnitudes of k th and mth buses respectively, km is phase generator bus respectively; Vgk is voltage magnitude of k th
angle between k th and mth buses. generator bus; tkmin , tkmax are minimum and maximum tap
4) Voltage stability enhancement: Mathematically voltage settings of k th transformer respectively; tk is tap setting of
stability index (VSI) is expressed as below: the k th transformer tap; bshk min max
, bshk are minimum and
th
min f 3 = min(max( Lk )) (8) maximum susceptance limits of k shunt var compensator
respectively; bshk is susceptance value of k th shunt capacitor;
Vlkmin , Vlkmax are minimum and maximum voltage limits of k th
load bus respectively; Vlk is voltage magnitude of k th load v12 v11
= (22)
bus; and Slkmax , Slk are maximum MVA flow and MVA flow v2 v1
in k th transmission line respectively.
Two different cases are occurred depending upon the value
of COR ( ) for each impact as described below:
III. COLLIDING BODIES OPTIMIZATION
1) Perfect Elastic Collision: Here, there is no loss of
A. Laws of Momentum and Energy energy when impact occurs between two objects. Therefore,
Laws of energy and momentum govern the impact the coefficient of restitution is considered as equal to one
between the positions of the two objects. When an impact ( =1) and final velocities are very high after impact
happens in an isolated power system, the positions of the happened.
objects before and after impact are shown in Fig. 1. During 2) Inelastic Collision: Here, there is loss of energy (some
this process, the momentum of the objects is preserved. The amount of energy is converted into other forms of energy)
momentum of all the objects before and after impact are equal when impact occurs between two objects. Therefore, the
if there is no external force acting upon these objects and can coefficient of restitution is considered as less than one ( <1)
be expressed as follows: and the final velocities are low after collision occurred.
m1v1 + m2 v2 = m1v '1 + m2 v '2 (18)
B. Detailed Description of CBO Algorithm for solving OPF
Similarly, the total kinetic energy before and after the Step 1: Each colliding body, which represents the complete
collision is expressed as follows: solution set comprising of generator buses active real power
1 1 1 1 outputs, generator buses voltage magnitudes, tap settings of
m1v12 + m2 v22 = m1v '12 + m2 v '22 + P (19) the transformer connected between various buses and reactive
2 2 2 2 power of the shunt var compensators are generated randomly
where m1 and m2 are masses of the two objectives within their limits, which represent the CB positions. Thus, the
respectively, v1 , v2 , v '1 & v '2 are velocities of the two objects k th CB that is expressed as follows:
before and after impact respectively. P is the loss of kinetic X k = Pgk ,2 , , Pgk , ng , Vgk ,1 , , Vgk , ng ,
energy. The final velocities v '1 , v '2 of the two objectives after (23)
the impact are defined as follows; tk ,1 , , tk , nt , bshk ,1 , b shk , nc

( m1 m2 ) v1 + ( m2 + m2 ) v2 The complete search space for CBO algorithm having


v '1 = (20) population NP is expressed as follows:
m1 + m2
X = [ X 1 , , X k , , X NP ]
T
(24)

Step 2: Calculate the fitness value of each CB as follows


( m2 m1 ) v2 + ( m1 + m1 ) v1 [7].
v '2 = (21)
m1 + m2
F = fi + wP (|PG1 -PG1lim |) + wV (| VL VLlim |) +
2 2

where is the coefficient of restitution (COR) of the (25)


wQ (| QG QGlim |) + wS (| Sl Sllim |)
2 2
colliding objects, it is defined as the ratio of relative velocities
of two objects after impact to before impact.
where, f i is the i th objective function, i = 1, 2, 3, 4 ,
wP , wQ , wV and wS are penalty coefficients of the respective
constraints. These values are to be selected high in order to
reject the infeasible strings during the iteration process.
Step 3: The body mass of each CB is calculated as.
1
F (k )
mk = (26)

N
1
m =1 F ( m)

where, F ( k ) is the fitness value of the k th CB.


Step 4: Sort all the CBs in ascending order according to
their mass values. These sorted values are divided equally into
two parts (Fig. 2). (i) Stationary group (ii) moving group. The
Fig. 1. The principle of collision between to objective bodies (a) before velocity of the stationary and moving objects before the
impact (b) after impact.
collision is expressed as follows:
Step 7: If any control variable is violating its minimum
limit made equal to its minimum limit and any variable
violates it maximum limit made equal to its maximum limit.
When the maximum number of iterations is reached, then stop
the procedure and print the optimum solution from the last
iteration. Otherwise, go to step 2.

IV. SIMULATION RESULTS


To demonstrate the feasibility and validity of CBO
algorithm, IEEE 30-bus system with various objective
functions are considered. The complete work was done on a
2.2 GHz CPU, i3 core processor and program code is written
in MATLAB 2009a. The number of colliding bodies and
maximum number of iterations are considered as 50 and 100
respectively for all objective functions.
Fig. 2. The pairs of objects.
A. IEEE 30-Bus System
NP It comprises of six generators, four transformer taps
vk = 0 k = 1, 2, (27) connected between various buses, and nine shunt
2
compensators. The voltage magnitudes of the buses and the tap
NP settings of a transformer lie between [0.9 1.1] p.u. Shunt
vk = X k NP X k k= + 1,, NP (28) compensators are varied between [0, 0.05] p.u. Moreover, the
2
2
complete information about IEEE 30-bus system can be
where X k is position vector of k th CB. referred from [6].
Step 5: The velocity of the stationary and moving objects Here the CBO method is applied to solve all the four
after collision is expressed as follows: objective functions considered in this work. The best set of the
control variables along with the total fuel cost (TFC), active
(m k + NP
2
+ mk + NP v
2
) k + NP NP
power loss (APL), sum of voltage deviation (SVD) and
voltage stability index (Lmax) for all the cases are given in
vk ' = 2
k = 1, 2, (29)
mk + mk + NP 2 Table I. The optimal TFC obtained in Case 1 is compared with
2
LCA [10], DE [5], simulated annealing (SA) [10], enhanced

vk ' =
(m m ) v
k k NP
2
k
k=
NP
+ 1,, NP (30)
GA (EGA) [4], BHBO [9], genetic evolving ant direction
hybrid DE (GEADHDE) [10], adaptive GA with adjusting the
mk + mk NP 2 population size (AGAPOP) [10], and EADDE [10] that are
2 furnished in Table II. From these results, it is proved that the
where is the coefficient of restitution and it is defined CBO method is capable to minimize the TFC at least value of
as follows. 799.1124 ($/hr) as compared to all the other methods reported
in the literature. The convergence characteristic obtained in
J Case 1 is shown in Fig. 3. In Case 2, CBO method is capable
= 1 (31) to decrease the APL at least value of 2.9369 (MW), while the
J max
other approaches ABC [6], and enhanced GA (EGA) [4] are
where, J is the current iteration number and Jmax is resulting in 3.1078 (MW), and 3.2008 (MW) respectively. The
percentage of APL with respect to the base case for all the
maximum number of iterations. methods is shown in Fig. 4. From Fig. 4, it is confirmed that
Step 6: Based on the new velocities, each colliding body the CBO leads to the best APL savings compared to the other
has updated its position by the following equations. algorithms. The SVD obtained in Case 3 using the proposed
CBO is 0.0932 p.u., which is the optimum SVD as compared
NP to the teaching learning based optimization (TLBO) [10],
X kJ +1 = X kJ + rand (1, NCV ) * VkJ k = 1, 2,, (32)
2 BBO [8], PSO gravitational search algorithm (PSOGSA) [8],
NP AGAPOP [10], and BHBO [9]. The voltage profile in Case 3
X kJ +1 = X J NP + rand (1, NCV ) *VkJ k= + 1,, NP (33) compared to Case 1 is illustrated in Fig. 5. From this, it is
k
2 2 observed that the voltage profile in this case is greatly
improved with respect to the Case 1, which results in an
where NCV is total number of control variables. X kJ , improvement of 94.9165 % in the SVD, but there is a slight
X kJ +1 are position vectors of k th CB in J th , ( J + 1) iterations
th
increment 1.77% in the TFC as compared to Case1. In Case 4
CBO enhanced the voltage stability by minimizing the net
respectively. voltage stability index (VSI) of all the load buses. The net VSI
values for all the load buses are calculated with CBO method
for Case 1 and Case 4 are shown in Fig. 6. From which is ide-
TABLE I. OPTIMAL CONTROL VARIABLES FOR ALL CASES
Control
Case 1 Case 2 Case 3 Case 4
variables
Pg1 (MW) 176.2460 52.8437 183.9404 78.8687
Pg2 (MW) 49.1931 78.7410 33.3919 59.3520
Pg5 (MW) 21.5458 50.0000 23.4786 50.0000
Pg8 (MW) 21.4153 35.0000 16.0317 35.0000
Pg11 (MW) 11.5906 30.0000 24.5152 26.5470
Pg13 (MW) 12.0000 39.7521 12.1504 37.1473
Vg1 (p.u.) 1.1000 1.1000 1.0321 1.0979
Vg2 (p.u.) 1.0852 1.0944 1.0189 1.0943
Vg5 (p.u.) 1.0566 1.0799 1.0185 1.0923
Vg8 (p.u.) 1.0689 1.0853 1.0094 1.0903
Vg11 (p.u.) 1.1000 1.0792 0.9849 1.1000
Vg13 (p.u.) 1.0990 1.0939 1.0011 1.1000
t6-9 (p.u.) 1.0230 0.9474 0.9967 0.9873
t6-10 (p.u.) 0.9135 1.0603 0.9000 0.9000
Fig. 4. Percentage of APL savings for Case 2.
t4-12 (p.u.) 0.9854 0.9963 0.9645 0.9897
t28-27 (p.u.) 0.9689 0.9640 0.9645 0.9526
bsh10 (p.u.) 0.0105 0.0493 0.0428 0.0000 TABLE II. COMPARISON OF SIMULATION RESULTS OF DIFFERENT CASES
bsh12 (p.u.) 0.0500 0.0330 0.0214 0.0419
Different
bsh15 (p.u.) 0.0340 0.0408 0.0500 0.0344 Method TFC ($/hr)
cases
bsh17 (p.u.) 0.0387 0.0500 0.0000 0.0150 CBO 799.1294
bsh20 (p.u.) 0.0500 0.0443 0.0500 0.0000 LCA[10] 799.1974
bsh21 (p.u.) 0.0485 0.0295 0.0428 0.0391 SA [10] 799.45
bsh23 (p.u.) 0.0436 0.0239 0.0500 0.0100 EGA [4] 799.56
bsh24 (p.u.) 0.0485 0.0500 0.0500 0.0000 Case 1
AGAPOP [10] 799.8441
bsh29 (p.u.) 0.0239 0.0188 0.0214 0.0079 BHBO [9] 799.9217
TFC ($/hr) 799.1294 963.9374 813.5987 919.9993 GEADHDE [10] 800.1579
APL (MW) 8.5910 2.9369 10.1085 3.5152 EADDE [10] 800.2014
SVD (p.u.) 1.8334 1.84971 0.0932 1.9867 DE [5] 800.5609
Lmax 0.12809 0.12683 0.14889 0.1245 Method APL (MW)
CBO 2.9369
ntified that bus 30 is identified as most week bus as its L- Case 2 ABC [6] 3.1078
index value with respect to Case1 is highest. In Case 3, the L- EGA[4] 3.2008
index value is reduced by 2.8 % as compared with Case 1. Base Case [6] 5.812
Hence, CBO method enhanced the voltage stability by Method SVD (p.u.)
CBO 0.0932
minimizing the net L-index value for all the load buses. From
TLBO [10] 0.0945
Table II it is confirmed that the max L- index value obtained Case 3
BBO [8] 0.0951
in DE [5], and ABC [6]. From all the above results, it is PSOGSA [7] 0.0968
confirmed that the CBO is an efficient approach for solving BHBO [9] 0.1262
the OPF problems. Method Lmax
Case 4 CBO 0.1245
DE [5] 0.1246
860 ABC[6] 0.1379

850 CBO
Total fuel cost ($/hr)

840

830

820

810

800

790
0 20 40 60 80 100
Fig. 5. Voltage profile in Case 1 and Case 3.
Number of iterations
Fig. 3. Convergence characteristics for Case 1.
[2] R. M. Palomino and V. H. Quintana, Sparse reactive power
scheduling by a penalty-function linear programming technique, IEEE
Trans. Power Syst., vol. 1(3), pp. 3139, 1986.
[3] A. Santos and G. R. M. Costa, Optimal powerflow solution by
Newtons method applied to an augmented lagrangian function, IET
Gen. Transm. Distrib., vol. 142(1), pp. 33-36, 1995. DOI: 10.1049/ip-
gtd:19951586
[4] M. Sailaja Kumari and S. Maheswarapu, Enhanced genetic algorithm
based computation technique for multi-objective optimal power flow,
Int. J. Electr. Power Energy Syst., vol 32(6), pp. 73642, 2010.
[5] M. A. Abido and N. A. Al-Ali, Multi-objective differential evolution
for optimal power flow, Int. conf. Power Eng. Energy Electr. Dirves,
Fig. 6. L-index value of all the load buses in Case 1 and Case 4. 2009. POWERENG 09. on 18-20 March. 2009, pp. 101-106.
[6] M. Rezaei Adaryani and M. Karami, Artificial bee colony algorithm
V. CONCLUSION for solving multi-objective optimal power flow problem, Int. J. Electr.
Power Energy Syst., vol. 53, pp. 219-30, 2013.
In this paper, a newly developed algorithm called colliding [7] R. Jordan, K. Dardan, J. Miroljub, and A. A. Nebojsa, Optimal power
bodies optimization (CBO), which has been applied to solve flow using a hybrid optimization algorithm of particle swarm
optimal power flow (OPF) problem. CBO is inspired from optimization and gravitational search algorithm, Electr. Power
Compon. Syst., vol. 43(17), pp. 1958-70, 2015.
collision happens between the two objective bodies, which
[8] A. Bhattacharya and K. P. Chattopadhyay, Application of
obey the laws of momentum and energy. The superiority of biogeography-based optimization to solve different optimal power flow
the CBO method is verified on IEEE 30-bus system with problems, IET Proc Gener. Transm. Distrib., vol 5(1), pp. 70-80,
various objective functions. The optimal results attained with 2011.
the CBO approach have been compared with the LCA, DE, [9] H.R.E.H. Bouchekara, Optimal power flow using black-hole-based
AGAPOP, BHBO, GEADHDE, EADDE, ABC, EGA, DSA, optimization approach, Appl. Soft Comput., vol. 24, pp. 879-888,
2014.
GSA, TLBO, and PSOGSA. These results confirmed that the
[10] H.R.E.H. Bouchekara, M.A. Abido, A.E. Chaib and R. Mehasni
CBO method outperforms all the other above-mentioned Optimal power flow using the league championship algorithm: A case
methods. Therefore, the results proved that the CBO is study of the Algerian power system, Energy convers. manag. vol. 87,
effective for solving small-scale test systems and pp. 5870, 2014.
implementing to solve multi-objective CBO for large and [11] A. Kaveh and V.R. Mahdvi, Colliding bodies optimization: A novel
meta-heuristic method, Comput. Struct., vol. 139, pp. 18-27, 2014.
practical power systems would be a continuation of the present
[12] A. Kaveh and V.R. Mahdvi, Two-dimensional colliding bodies
work. algorithm for optimal design of truss, Adv. Eng. Softw., vol. 83, pp.
70-79, 2015.
REFERENCES
[13] A. Kaveh and V.R. Mahdvi, Colliding Bodies Optimization method
[1] H. W. Dommel and W. F. Tinney, Optimal power flow solutions, for optimum design of truss structures with continuous variables,
IEEE Trans. Power Apparat. Syst., vol. PAS-87, pp. 18661876, Oct. Adv. Eng. Softw., vol. 70, pp. 1-12, 2014.
1968.

You might also like