Optim 2014 6850887

Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

Application of the Firefly Algorithm to

Optimal Over-Current Relay Coordination


S. S. Gokhale1, Dr. V. S. Kale2
1
Associate Prof., Y.C.C.E, Nagpur, India, 2Associate Prof., V.N.I.T, Nagpur, India,
[email protected], [email protected]

plied to optimally coordinate the relays. Two sample cases


Abstract- Optimization of over current relay settings is a ma- have been shown.
jor problem in protection of electrical power systems. This paper
describes the Firefly algorithm for optimal time coordination of
these relays. The algorithm has been implemented in MATLAB II. PRINCIPLE OF OVER-CURRENT RELAY COORDINATION
and tested on a radial system, as well as a single end fed system
with a parallel feeder. The results obtained by the Firefly algo- In the two section radial system shown in Fig.1, for a fault
rithm (FA) are compared with those obtained by the Genetic beyond bus B (zone k), the relay RB at bus B, operates first.
algorithm (GA) and the Linear Programming method. The novel The relay RA at bus A, serves the purpose of back-up protec-
feature of this paper is the application of the Firefly algorithm to tion. If the operating time of relay RB, is set at 0.1s, the relay
the problem of over-current relay coordination. RA is coordinated such that it should operate at 0.1s. Plus the
operating time of circuit breaker at bus B, plus the overshoot
I. INTRODUCTION
time of relay RA..
The primary protection in distribution systems mainly In general the coordination of directional over-current re-
comprises of over current relays. Many a times they are also lays in power systems can be stated as an optimization prob-
used as a back-up protection in transmission systems. These lem ,where the sum of operating times of the relays 'z' of the
relays have to operate fast, without mal-operating, so that the system ,for near end fault is to be minimized.
circuit breaker isolates the faulty section. This makes coordi- Minimize:
nation of these relays important.
m
The relay coordination problem can be concisely stated as,
"Given the magnitude of the load and fault current at every
z= ∑W p .t p,k (1)
p =1
bus, how to set the relays at the different buses, so that the
entire system gets primary as well as backup over-current
Here m denotes the total number of relays,
protection?”[1]. Correct coordination is necessary to ensure
tp,k is the operating time of the primary relay, for fault in zone
that the relays do not operate inadvertently. Optimal coordi-
k, for near end fault, and Wp is the weight assigned for operat-
nation is necessary so that the system has the lowest possible
ing time of the relay Rp.
fault clearing time. Hence the optimal coordination of relays
As in distribution systems the lines are short and are nearly
is formulated as a relay operating time minimization problem
equal in length, equal weights are assumed and (W=1) is as-
under selectivity constraints. With the advent of digital relays
signed to all the relays [12].
the Time Multiplier Setting (TMS) can be set at any non inte-
gral value, and therefore has made the exact optimal time
coordination more possible.
This problem has been dealt with by several optimization
methods in the past [2, 3]. Very recently [4] has applied a
Seeker algorithm to solve the relay coordination problem. The
Artificial Bee colony algorithm by [5], the Genetic algorithm
by [6, 7, 8], a hybrid evolutionary algorithm based on Tabu
search by [9], and a Particle Swarm Optimization algorithm
(PSO) by [10] are some of the other notable approaches.
In this paper the relay coordination problem has been for- The constraints in the problem are:
mulated as a Linear programming problem by considering the
Plug Setting (PS) of the relays fixed, and the operating time
of the relays, a linear function of the Time Multiplier A. Coordination constraints
Setting (However, this need not be the case always and the The relays in the system have to be coordinated by the criteri-
problem can be formulated as a non-linear problem [11]). The on,
paper then describes how the Firefly algorithm has been ap- t q , k − t p , k ≥ STI (2)

978-1-4799-5183-3/14/$31.00 ' 2014 IEEE 150


where tp,k is the operating time of the primary relay at k, for Create Initial population of fireflies ;.
near end fault and tq,k is the operating time of the back-up Formulate light intensity Im , associated with f(x)
relay, for the same near end fault. STI is the Selective Time Define absorption coefficient γ
Interval [13]. In this paper it has been taken as 0.57. While (t < Maximum _Iterations)
for i=1:n (all n fireflies)
B .Bounds on the relay operating time for j=1:n (n fireflies)
if (Ij>Ii ),
Relays take a certain minimum time to operate. Also they move firefly i towards j;
should not take too long a time to operate. This puts the con- end if
straint, Vary attractiveness with distance r via ; exp( - γr2)
Evaluate new solutions and update light intensity;
t p.k,min ≥t p,k ≥t p,k,max (3) end for j
end for i
where tp,k,min is the minimum operating time and tp,k,max is the Rank fireflies and find the current best;
maximum operating time of relay at 'k' for near end fault. end while
end
C. Relay characteristics
----------------------------------------------------------------------
The relays are assumed to be identical, having normal Inverse FA initially produces a swarm of fireflies located random-
Definite Minimum Time (IDMT) characteristics, which can ly in the search space. The location of each firefly in the
be described by the equation, search space represents a potential solution of the optimiza-
tion problem. The dimensions of the search space are equal to
0.14*(TMS ) the number of optimizing parameters in the given problem.
t op = (4) The fitness function depends upon the position of the firefly.
PSM (0.02) −1
The brightness of each firefly depends on the fitness value of
that firefly.
where top is the relay operating time ,TMS is the time multi-
Each fly is attracted by the brightness of the other flies and
plier setting and PSM is the plug setting multiplier.
moves to the fly that relatively appears brighter. This 'leads' to
I relay the concept of attractiveness. Attractiveness depends on
PSM = (5)
PS brightness as well as distance. A brighter firefly far off may
not be as attractive as a less bright firefly that is near. The
velocity with which a firefly moves towards another firefly
In the linear problem as PSM is fixed, (4) reduces to depends on the attractiveness. Each iteration the algorithm
computes the brightness and attractiveness of each firefly.
t op = a p (TMS ) (6) Depending on these values, the positions of the fireflies are
where updated. After sufficient iterations all fireflies converge to the
best possible position in the search space.
0.14 (7) A firefly is specified by a vector xm :
ap =
PSM (0.02) −1
⎡ 1 ⎤
xm = ⎢ X m , X m2 ,... X mnd ⎥ (9)
⎣ ⎦
Hence the objective function can be written as ,
where 'nd' denotes the dimensions of the search space.
Minimize:
m The search space is bounded by the constraints:
z= ∑ a p (TMS ) p (8)
x v (min) ≤ x v ≤ x v (max) ; v = 1,2,...nd (10)
p =1
The initial population of fireflies is generated from a uniform
III. FIREFLY ALGORITHM distribution by the equation:
v
The Firefly algorithm (FA) is a meta-heuristic formulated xm = x v (min) + ( x v (max) − x v (min)) *rand (11)
by Yang in 2007. It is a swarm intelligence type of algorithm
inspired by the flashing behavior of fireflies [14]. The FA where 'rand' is a random number between 0 and 1, taken from
algorithm is as follows: a uniform distribution.

------------------------------------------------------------------------- The light intensity of the mth firefly Im is related to the fit-
Define Objective function f(x) ; ness function. The relationship maybe proportional or inverse

151
TABLE 1: CALCULATION OF ap CONSTANT FOR RELAY.
⎧ ⎫
I m = f ⎨ Fitness ( Χ m )⎬ (12)
⎩ ⎭ Sr.
Fault
Relay RB- ap con-
Loca- Relay RA - ap constants
In this paper FA searches for an optimal solution by maximiz- No.
tion
stants
ing the function ‘Light Intensity’ which is obtained from the
relay operating time minimization function by: 0.14
Just = 2.63
Maximize: 1 beyond (13.33) ( 0.02 ) −1 -------
Bus A
1 (13)
I m=
1+ z Just 0.14 0.14
2 beyond = 2.97 =2
The attractiveness between the mth and nth firefly, βmn, is giv- Bus B (10)(0.02) −1 (30) (0.02) −1
en by,
'---' indicates that the fault is not seen by that relay

β m ,n = ( β max, m ,n − β min, m ,n ) exp( −γrm2,n ) + β min, m ,n (14)


These lead to the following optimization equations:
where rm,n is the Cartesian distance between the mth and nth Minimize:
fly. z =2.63x1+2 x2 (16)
th
If the light intensity of n firefly is larger than the intensity
of the mth firefly, then the mth firefly moves towards the nth Subject to the coordination constraint,
firefly and its position at the kth iteration is denoted by the
following equation: 2.97x1-2 x2 ≥ 0.5 (17)

and the lower bounds on the relay decided by the minimum


xm(k)= xm(k −1)+βmn(xm(k −1)−xn (k −1))+α(rand−0.5) (15)
time for the relay to operate. (This time is taken as 0.2 sec.)
where α is the random movement factor. A large α explores 2.63x1 ≥. 0.2 (18)
the distant search space, whereas a smaller α facilitates local
search. β is limited between 0 and 1. If βmax is kept at 1 the 2 x2 ≥ 0.2 (19)
algorithm performs cooperative local search with the brightest
firefly strongly determining the position of the other fireflies. The upper limit is taken at 1.2.
If βmax is set to zero, there is only non-cooperative random 'x' denotes the TMS. The subscripts of x denote the relay
search. number. This is a two dimensional problem ('nd' =2). The
The parameter γ determines the variation of attractiveness optimization equations were solved using the Firefly algo-
with distance. Setting γ to 0 corresponds to no variation or rithm in the MATLAB environment. The population size was
attractiveness is constant and the FA reduces to the PSO. Put- taken as 40 and the number of iterations was taken to be 500.
ting γ = ∞ results in attractiveness equal to zero which means, The value of α was chosen as 0.5. The penalty method was
a complete random search. In this paper γ is chosen as 10. used for handling the constraints [16, 17].
Choice of these parameters is important for the fast conver- The results obtained for the TMS are as shown in Table 2.
gence of the algorithm [15].Equations (14) and (15) produces where they are also compared against those obtained by the
better off-springs each iteration. In general FA is considered genetic algorithm and the linear programming method. The
to be better than Particle Swarm Optimization, as far as time latter two tools are readily available in the MATLAB optimi-
taken for the optimum value to be generated is concerned. zation toolbox. Fifty iterations of each were taken.
TABLE 2: TMS OBTAINED USING FIREFLY, GA AND LINEAR
IV. RESULTS PROGRAMMING METHOD

Case 1: The Radial System


Firefly Genetic Linear
To test the algorithm, initially the simple radial system Relay TMS
Algorithm Algorithm programming
shown in Fig.1 was considered. The maximum fault current
RA x1 0.265 0.27 0.289
just beyond bus A and bus B are 4000A and 3000A respec-
tively, the plug settings of the relays are taken as 1, the CT RB x2 0.1 0.1 0.11
ratios at bus A and bus B are taken as 300:1 and 100:1 respec-
tively. Minimum operating time for each relay is considered
as 0.2s and STI is taken as 0.57 s. Calculation of value of ap is The total operating time 'z' for all the relays given by (1) for
done by using (7) and is as shown in Table1. the three methods is shown in Table No. 3.

152
TABLE 3: 'z' OBTAINED USING FIREFLY, GA AND LINEAR Minimize
PROGRAMMING METHOD
.
z = 2.97x1 + 5.749x2 + 5.749x3 + 2.97x4 (20)
Total Operat- Subject to coordination constraints,
Firefly Genetic Linear
ing
Algorithm Algorithm Programming
Time(sec.)
5 . 749 x 4 − 5 . 749 x 2 ≥ 0 . 57 (21)
'z' 0.365 0.37 0.399
5.749 x1 − 5.749 x3 ≥ 0.57 (22)

and the minimum time constraints,


As seen, the Firefly algorithm gives total operating time, 'z'
=0.365 s compared to that obtained by genetic algorithm and 2 . 97 x1 ≥ 0 . 2 (23)
Linear programming.
5 . 749 x 2 ≥ 0 . 2 (24)
Case 2: A PARALLEL FEEDER SYSTEM
5 .749 x 3 ≥ 0 .2 (25)

2 . 97 x 4 ≥ 0 . 2 (26)

This problem has four dimensions ('nd ' =4). The equations
were solved using the Firefly algorithm with α taken as 0.2.
The population size was taken as 40 and the number of itera-
tions was taken to be 500.
The TMS obtained are shown in Table 5 and the results ob-
tained are compared with those obtained by Genetic algorithm
and linear programming technique. Again fifty iterations of
each were taken.
The total operating time for all the relays given by (1) for
the three methods is shown in table No. 6.
As seen FA gives total operating time 'z' =0.326 s which is
Apart from the radial system of Fig.1, the case of the paral- least compared to that obtained by genetic algorithm and Lin-
lel feeders, fed from a single end is also tested by the Firefly ear programming method.
algorithm. Fig.2 shows this system. Relays 1 and 4 are non-
TABLE 5: TMS OBTAINED USING FIREFLY, GA AND LINEAR
directional relays, whereas 2 and 3 are directional relays. The PROGRAMMING METHOD.
relays are assumed to have plug setting of 1 and the corre-
sponding current transformer ratios are 300:1. Relay 4 will Firefly Genetic Linear
Relay TMS
back up relay 2 for fault at A and relay 1 will back up relay 3 Algorithm Algorithm programming
for fault at B. Total fault current in each case is assumed to 1 x1 0.131 0.134 0.159
be 4000A.
2 x2 0.032 0.035 0.039
The currents and ap constants of the relays are shown in
Table 4. 3 x3 0.032 0.035 0.039

4 x4 0.131 0.134 0.159


TABLE 4: CURRENT SEEN BY RELAY AND ap CONSTANT FOR
RELAY

TABLE 6: 'z' OBTAINED USING FIREFLY, GA AND LINEAR


Fault Relay No. PROGRAMMING METHOD
point 1 2 3 4
Relay Total Operating Firefly Genetic Linear
10 3.33 ---- 3.33
current Time(sec.) Algorithm Algorithm Programming
A
ap constants 2.97 5.749 ---- 5.749 'z' 0.326 0.338 0.396
Relay
3.33 --- 3.33 10
current
B
V. CONCLUSION
ap constants 5.749 ---- 5.749 2.97
'---' indicates that the fault is not seen by that relay The problem of optimal coordination has been solved using
the Firefly algorithm. The programs have been developed in
These lead to the equations: MATLAB, so as to solve the problem for many numbers of

153
relays and many numbers of primary back-up relationships. [6] A. Koochaki, M.R. Asadi, R.R. Naghizadeh, "Optimal over- current
relay coordination using Genetic Algorithm",11th Inter- national Con-
Two cases have been presented in this paper. The first case ference OPTIM 2008, pp 197-202.
though trivial, illustrates the basics of the problem. It is also a [7] P.P.Bedekar, S.R.Bhide, "Optimum coordination of over-current relay
very practical example. As the number of relays goes on in- using the Continuous Genetic Algorithm", Expert System with applica-
tions(Elsevier) 38(2011), 11286-11292.
creasing the highly constrained nature of the problem be- [8] C.W. So, K.K. Li, K.T. Lai, K.Y. Fung, "Application of Genetic Algo-
comes more visible. The results obtained show that the firefly rithm to over-current relay grading coordination, Proceedings of the 4th
algorithm is more efficient than the other approaches. This is International Conference on Advances in Power Sys tems, Control, Op-
eration and Management, Hong Kong1997.
crucial because, as the relay operating time decreases, the [9] Xu. Chunlin, Zou Xiufen and Wu Chuansheng, "Optimal Coordination
damage caused by the fault can get limited. of Protection relay using New Hybrid Evolutionary Algorithm", IEEE
Still, different network topologies as in [18] should be test- Transactions on CEC,pp823-828,2008.
[10] Liu, An. Yang, Ming-Ta, "Optimal coordination of directional over-
ed. Also it should be tried on the non-linear coordination current relay using NM-PSO technique", IEEE Transactions on Interna-
problem as well, where the plug setting of the different relays tional Symposium on Computer and Control, 2012. pp 678-681.
can be considered as optimization variables. [11] P.P. Bedekar, S.R. Bhide, "Optimal Coordination of directional Over-
current relay using GA_NLP Hybrid Approach", IEEE Transactions on
Power Delivery, Vol.26,No.1,Jan.2011.
[12] H.Zeienldin, E.F.El-Saadany "A Novel Problem formulation for direc-
REFERENCES tional over-current relay coordination", Large Engineering systems con-
[1] Y.G. Paithankar and S.R. Bhide, Fundamentals of Power System Pro- ference on Power Engineering(LESCOPE) 28-30 July 2004,pp 48-52.
tection, Prentice Hall of India Pvt. Ltd, New Delhi 2013. [13] Warrington, The Protective Relays, Theory and Practice", John Wiley
[2] A.J Urdaneta ,H Restrepo, S.Marquez and J. Sanchez, Optimal Coordi- and Sons, New York, 1969.
nation of Directional relays in Interconnected Power Sys tems, IEEE [14] X.S. Yang, Nature Inspired Meta-heuristic Algorithms, Luniver Press,
Transactions on Power Delivery, July 1988, Vol.3 No.3, pp 903-911. Beckington, 2010.
[3] C.W. So, K. K. Li. Time co-ordination Method for power system pro- [15] T. Apostolopoulos, A Vlachos, "Application of the firefly algorithm for
tection by evolutionary algorithm, IEEE Transactions on Industry Ap- solving the Economic emissions load dispatch problem", International
plications, September/October 2000. Vol. 36, No.5, pp 1235-1240. Journal of Combinatorics, 2011.
[4] A. Turaj, "Coordination of directional Over-current relay using Seeker [16] S.S. Rao, Engineering Optimization-Theory and Practice, 3rd Edition,
Algorithm." ,IEEE Transactions on Power Delivery, pp. 1415-1422, New Age International Pvt. Ltd, India, 1998.
2012. [17] K. Deb, Optimization for Engineering Design, Algorithms and Exam-
[5] D Uthitsunthorn, P Pao-L-Or and T. Kulworawanichpong, "Optimal ples, Prentice Hall of India, Pvt. Ltd., New Delhi, India 2006.
over-current relay coordination using Artificial Bees Colony Algo- [18] Abbas Saberi Noghabi, Javed Sadeh, Habib Rajabi Mashhadi," Consid-
rithm", The Eighth Electrical Engineering, Electronics, Computer, Tel- ering Different Network Topologies in Optimal Over- current relay co-
ecommunications and Information Technology Conference (ECTI- ordination using a Hybrid GA", IEEE Transactions on Power Deliv ery,
2011),pp 901-904. Vol. 24,No.4,October 2009.

154

You might also like