 Open Access
 Total Downloads : 406
 Authors : Moses Peter Musau, Nicodemus Abungu Odero, Cyrus Wabuge Wekesa
 Paper ID : IJERTV5IS050532
 Volume & Issue : Volume 05, Issue 05 (May 2016)
 Published (First Online): 16052016
 ISSN (Online) : 22780181
 Publisher Name : IJERT
 License: This work is licensed under a Creative Commons Attribution 4.0 International License
Single Objective Dynamic Economic Dispatch with Cubic Cost Functions using a Hybrid of Modified Firefly Algorithm with Levy Flights and Derived Mutations
Moses Peter Musau
Department of Electrical and Information Engineering, The University of Nairobi,
Nairobi, Kenya.
Abstract: The single objective dynamic economic dispatch (SODED) problem been formulated in quadratic form and solved extensively using pure and hybrid methods. SODED solution can be improved by introducing higher order generator cost functions since the fuel cost functions become more nonlinear when the actual generator response is considered. Cubic cost functions models the actual response of thermal generators more accurately, thus it is an industry practice to adopt cubic polynomials for modelling fuel costs of generating units. Previous works have considered cubic static ED (SED).Therefore, there is need to consider the formulation of the dynamic SODED with all the possible constraints. Further the hybrid methods used in the solution of this vital problem need to be revisited and better ones developed. The modern trend of hybrids are the twomethod and three method hybrids. In this paper, cubic SODED is formulated and validated on IEEE 3unit,5unit and 26unit systems using Modified Firefly Algorithm with Levy Flights and Derived Mutation(MFALF DM).The proposed method proved better than Genetic Algorithm (GA), Particle Swarm Optimization (PSO) in determining optimal dispatch in the industry using the fully constrained SODED.
Key words: Cubic cost functions, Modified Firefly Algorithm with Levy Flights and Derived Mutation (MFALFDM), Single objective dynamic economic dispatch (SODED)
I: INTRODUCTION
Economic Dispatch (ED) with cubic cost functions has been extensively studied in the past researches. According Z.X Liang and J.D Glover ,1991[1],a very crucial issue in SODED studies is to determine the order and approximate the coefficients of the polynomial used to model the cost function. This helps in reducing the error between the approximated polynomial along with its coefficients and the actual operating cost. According to Z.X Liang and J.D Glover,1992 [2] and A.Jiang and S.Ertem,1995 [3] to obtain accurate SODED results, a third order polynomial is realistic in modelling the operating cost for a non monotonically increasing cost curve. SODED works using cubic cost functions include Bharathkumar.S et al, 2013[4], Hari M.D et al, 2014[5], Deepak Mishra et al, 2006[6], and N.A.Amoli et al, 2012[7]. Krishnamurthy, 2012 [8] used the static cubic function of the emissions dispatch in the Multi Objective Static ED(MOSED) using the Lagrange method(LM).This provided better results as compared to the quadratic functions. In all these studies, however, the cubic cost function provided more accurate and practical results as compared to lower order cost functions. A
N. O Abungu 2, C. W Wekesa3
Department of Electrical and Information Engineering, The University of Nairobi,
Nairobi, Kenya.
functions have been considered in a great extent. Only B.S et al, 2013 [4] has considered the SODED, thus there is need to consider the SODED with all the possible constraints in place. The thermal cost functions has been considered with only the work in [4], [5] and [8] incorporating emission cost functions. Further, the pure heuristic deterministic methods which are strong and weak at the same time have been applied, only the works in [5] have considered a two method hybrid. Thus there is need to use more advanced hybrid methods for better results in these vital and complex cubic cost functions.
Contribution: In this paper a fully constrained dynamic SODED (with ramp rates, valve points and prohibiting zones) with cubic cost function is formulated. A new method, Modified Firefly Algorithm (MFA) and its hybrids is proposed for its solution. These hybrids include MFA with Levy Flights (MFALF) and MFALF with Derived Mutation (MFALFDM). The results are compared with those for pure methods, for example, Genetic Algorithm (GA) and Particle Swarm Optimization (PSO). Cubic and Quadratic cost functions results are also compared and presented.
II: PROBLEM FORMULATION
Economic dispatch (ED) may sometimes be classified as a static optimization (SOSED) problem in which costs associated with the act of changing the outputs of generators are not considered.
According to Jizhong Zhu, pp. 8788, (2009) [13], the single objective function corresponding to the production cost can be approximated to be a quadratic function of the active power outputs from the generating units. This static ED (SED) is formulated as
min = 2 + + (1)
A general formulation for the order SOSED was proposed by Yusuf Sonmez, 2013[9]. It can be given by the equation
F( ) = 0, + = + (2)
summary of ED works using cubic cost functions is provided in Table 1.0, from which, it is clear that static cubic cost
=1
,
Table 1.0. ED with Cubic Cost Functions.
Reference 
Ob 
Nature Ob function 
Con 
Method 
Z.X Liang et al,1991[1] 
T 
Static 
– 
GramSchmidt(GS), Least Squares(LS) 
Z.X Liang et al,1992 [2] 
T 
Static 
3 
Dynamic programming(DP) 
A.Jiang and S.Ertem,1995 [3] 
T 
Static 
2 
Newton Method(NM) 
B.S et al ,2013[4] 
T,E 
DED with ramp rates 
4 
Fuzzy Logic (FL), Bacteria Foraging(BF) and Nelder 
and valve points 
Mead(NM) (FLBFNM) 

Hari Mohan D.et al,2014[5] 
T,E 
Static 
5 
PSOGeneral Search Algorithm (PSOGSA) 
Yusuf Somez,2013[9] 
T 
General static 
2 
Artificial Bee Colony (ABC) 
Deepak Mishra et al,2006 [6] 
T 
General static 
2 
ORHopfield Neural Network(ORHNN) 
N.A Amoli et al,2012 [7] 
T 
Static 
2 
Firefly Algorithm(FA) 
Krishnamurthy .S et al ,2012[8] 
T,E 
Static 
2 
Langrange Method(LM 
T. Adhinarayanan M.Sydulu,2006[10] 
T 
Static 
2 
Lambdalogic based(LLB) 
T.Adhinarayanan M.Sydulu,2010 [11] 
T 
Static 
2 
Lambdalogic based(LLB) 
E.B Elanchezhian et al,2014[12] 
T 
Static 
8 
Teaching learning based optimization (TLBO) 
Key: ObObjective function, TThermal cost objective functionEmissions cost objective function, ConNumber of constraints
On the other hand, dynamic SODED is one that considers changerelated cost and takes the ramp rate limits, valve points and prohibited operating zone of the generating units into consideration. The general form of the SODED is given by
=
,
,
F( ) = {0, + + }
=1
+  sin ( ) (3)
Where 0,,,, and are the cost coefficients of the ith unit, is the lower generation bound for it unit and is the error associated with the ith equation.
When L=1 the linear form of the SODED results.
F(,1) = 1,, + , + +  sin ( ) (4)
This is also called the first order model this is of no practical significance ED studies.
When L=2, the most popular quadratic SODED results. This is given by
,
,
F(,2) = 2,2 + 1,, + , +
+  sin ( ) (5)
When L=3, the cubic form of the SODED results. This can be expressed as
, ,
, ,
F(,3) = 3, 3 + 2,2 + 1,, + , +
+  sin ( ) (6)
The problem in equation (6) is solved subject to the following constraints:
= + (7)
=1
(8)
1 (9)
1 (10 )
= 1,2,3 . (11)
, (12)
, (13)
III: PROPOSED METHODOLOGY
Introduction to Fireflies
The fireflies are the most charismatic species among the insects and their spectacular display have inspired the poets, writers and scientists. Today more than 2000 species exists and the flashings of the fireflies can be seen in the summer sky in the tropical and temperate regions with warm weather and most active in the nights [15]. These fireflies produce the short rhythmic patterns of flashing lights and these patterns of flashes are unique in species to species, and the flashing light is produced by a bioluminescence process. Moreover, flashing is produced to attract their mating partners; the first signalers are flying males who tries to attract the females on ground. In response females also emit flashing lights and move towards the brightest firefly. However the flashing lights obey certain physical rules, the light intensity, I , decrease with the increase of distance r according to the term 1/2 [16] .Also the flashing is produced for communication purpose among each other and also to attract prey, but still the flashing behavior is a topic of discussion among scientists and engineers. Thus the flashing behavior of fireflies plays a key role in reproduction, protection, communication and feeding.
Firefly Algorithm (FA)
Firefly Algorithm (FA) [14] is a new nature inspired algorithm developed by XinShe Yang in the year 2007, based on the flashing behavior of the fireflies. The flashing signifies the signal to attract other fireflies, where the objective function is associated with the flashing light or the light intensity which helps the fireflies to move to brighter and more attractive locations to achieve optimal solution.
The FA has three idealized rules or assumptions which are been developed to define the characteristics of fireflies: i) All fireflies are unisex and they move towards the more attractive and brighter one irrespective of their sex. ii) The level of attraction of firefly is proportional to brightness which reduces with the increase in the distance between two
fireflies 1/2 since air absorbs the light. If there is no
brighter or more attractive firefly than a particular one, it will then move randomly. iii) The brightness or light intensity is determined by the value of the objective function of a given problem and it is proportional to the light intensity for a maximization or optimization problem.
Need for Improved FA (IFA)
The reasons behind making FA [16] so popular and successful include: i) The method automatically divides its population into subgroups, because of the fact that local attraction is stronger than long distance (global) attraction.

FA does not use historical individual best and explicit global best. This reduces the potential drawbacks of premature convergence. iii) Also FA does not use the velocities hence problems associated with velocities in PSO is automatically eliminated. iv) FA has an inbuilt ability to modify and therefore to control the parameters such as , leading to improved results. Hence it can be clearly seen that the FA is more efficient in respects of controlling parameters, local search ability, robustness and elimination
The minimum distance between any two fireflies and at
and is thus given by
= min = min (16)

Attractiveness
The attractiveness between two fireflies and at a separation distance is given by
0
0
= 2 (17)
Where 0 is the attractiveness at = 0.
In actual implementation, the actual implementation () is a monotonically decreasing function generalized as
0
0
() = 1 (18)

Movement
The movement of a firefly is attracted to another more attractive (brighter) firefly by the relation
= + 2 ( )
of premature convergence.
+1
0
N.A Amoli et al, 2012 [7] used the basic FA in solving static ED with cubic cost functions. However, the method is poor in global searching and optimization, long convergence time, requires more iterations, and low computational speed
.These problems can be addressed by using modified (improved) FA [17] and using heuristic and deterministic methods to form hybrid FA [18]. In a hybrid method the weaknesses of the base method are suppressed while its strengths are exalted leading to better realistic results and improved performance of the method. In this paper therefore, a hybrid of Modified FA (MFA) with Levy Flights (LF) [MFALF] coupled with Derived Mutation (DM); [MFALFDM] is proposed.
Modified Firefly Algorithm (MFA) with Levy Flights (LF) and Derived Mutation (DM) [MFALFDM]
The MFALFDM proposed in this paper has six operators. These include brightness, distance, attractiveness, movement, randomness reduction and mutation. These are formulated as follows:
i) Brightness
The brightness of a firefly at a particular location can be chosen as
() () (14)
+ sign [ 1] (19)
2
Where is the current position of a firefly,the second term defines the fireflies attractiveness to light intensity as seen by the adjacent firefly and the third term is for the random movement of a firefly is no brighter firefly is left, is a randomization parameter, and is a random number generator uniformly distributed over the space [0,1],that is, [0,1].
In general the solutions can be improved by reducing the randomness by
+ (0 ) (20)
Where [0, ]the pseudo is time for simulation and
is the maximum number of generations, and 0 are the final and initial values of the randomness parameter
v) Randomness Reduction
Levy flight is a random walk of step lengths having direction of the steps as isotropic and random. The concept propounded by Paul Pierre Levy (18861971) is very useful in stochastic measurements and simulations of random and pseudorandom phenomena.
The movement of a firefly with Levy Flights is defined by the relation
ii) Distance
= + 2 ( )
+1 0
+1 0
+ sign [ 1] Levy (21)
The distance between any two fireflies and at and
respectively is the Cartesian distance
Where the second term is due to
2
attraction, while the third
= = (, ,)2
=1
(15)
term is randomization via the Levy Flights with being the randomization parameter.The product means entry wise multiplication
The sign [ 1] where [0,1] essentially
2
provides a random sign or direction while the random step
Where , is the component of the spatial component
of the firefly
length is drawn from a Levy distribution given by
~ = , (1 < 3) (22)
Which has an infinite variance with an infinite mean
vi) Derived Mutation (DM)
To further improve the exploration of or diversity of the candidate solution, the simple mutation corresponding to from the ant colony optimization (ACO) genetic algorithm (GA), evolutionary programming ( EP) and differential evolution (DE)algorithms is adopted in the MFALF process. This enhances the accuracy of the optimum results in solving the fullyconstrained SODED problem.
3.2 Algorithm for MFALFDM
The proposed MFALFDM algorithm is implemented using the following procedure:
Step 1: Define objective function ().
Step 2: Read the system data, cubic cost coefficients, loss coefficients, minimum and maximum power limits of all the generating units and power demand Step 3: Input the algorithm parameters randomness (), attractiveness(), light absorption coefficient(), randomness reduction parameter () , number of fireflies (), maximum iterations, and stopping criteria. Step 4: Generate initial population of fireflies ( =
1,2,3 . ) in a random manner
Step 5: Set the iteration counter to 1
Step 6: Evaluate the light intensity or function value at
by value of ().
Step 7: while ( < )
for = 1: n all n fireflies
for = 1 all n fireflies
if >
Move firefly towards in ddimension via Levy flights
end if
Find the minimum variation distance of all fireflies
= (( ))
Attractiveness varies with distance r via exp[]
Evaluate new solutions and update light intensity
end for
end for
Random
Mutation if <
Rank the fireflies and find the current best
end of while loop
Step 8: Post process results and visualize the same Step 9: Find the firefly with the highest light Intensity among all fireflies, Gbest
Step 10: Plot the increase of light intensity with time per iteration
Step 11: Plot the objective with respect to time, % best solution with time
Step 12: End of MFALFDM
IV: RESULTS AND ANALYSIS
In this method the initial solution is generated randomly within the feasible range, The FA parameters used in the problem are as shown in table 2.0.The mapping of the parameters to the SODED problem is also given .
The cubic cost coefficient, maximum and minimum power limits, ramp rates and valve points have been taken from [4] and [10].A lossless system is assumed. The results are divided into four parts: SOCED and SODED comparison, ED with cubic cost function under various demands, ED with Cubic and quadratic cost functions and finally a comparison of MFALFDM with other methods in solving the SODED with cubic cost functions.
TABLE 2.0 Parameters for MFALFDM
Parameter 
Value 
Brightness 
() 
Alpha () 
0.9 
Beta () 
0.5 
Gamma () 
1.0 
Number of fireflies () 
50 
Maximum no. of iterations 
100 
Attraction at = 0, (0) 
2.5 
Lambda () 
1.5 
SOCED and SODED
The optimal generation of the six generating units and the optimal costs are displayed for each of the intervals. The algorithm is first run without any constraints and the optimization does not include the ramp rate constraints, that is, the algorithm is run to optimize a classic economic dispatch problem. The algorithm is then run to solve the classic economic dispatch with minimum generation constraints.
Modifications are then done to include maximum generation constraints. Finally, the algorithm is run to include the inequality, equality and ramp rate constraints. The algorithm optimizes a dynamic economic dispatch problem. The power demand for each interval is taken as [150MW, 300MW, 400MW 500MW]. The key used in interpreting the results in this section include A: SOCED without constraints, B: SOCED with min generation constraints, C: SOCED with min and MAX constraints, D: SODED with valve points and ramp rate limits. Further, in the tables,t represents computation time(seconds),n ,the number of iterations,L,losses(MW) and C,the optimal cost($). From the results tabulated in Table 3.05.0, it is clear that the optimal cost increases with the power demand. The cost of operation is directly proportional to the power demand. The cost is highest for the SOCED, then slightly less for the SOCED with minimum generation constraints, lesser when the algorithm is used for SOCED with max and min generation constraints and the cost is least when SODED is used with valve point effects and ramp rate generation constraints. The difference in the optimal cost is more pronounced at higher power demand. This is so because at lower power demand only minimum generating constraints are violated hence the costs tend to be similar. At higher power demands, line constraints and max generation constraints are violated hence the need to keep them in check by not overloading the
generators. This is done by defining the highest power demand that a generator can supply.
Table 3.0: Results for a load demand of 150 MW
Unit 
A 
B 
C 
D 

G1 
45.9510 
45.9510 
45.9510 
45.9510 

G2 
34.4019 
34.4019 
34.4019 
34.4019 

G3 
19.8484 
19.8484 
19.8484 
19.8484 

G4 
9.8010 
9.8010 
9.8010 
9.8010 

G5 
11.0204 
11.0204 
11.0204 
11.0204 

G6 
15.9781 
15.9781 
15.9781 
15.9781 

L 
0.3010 
0.3025 
0.3055 
0.3100 

t 
2.5 
2.6 
2.8 
3.0 

n 
10 
11 
12 
15 

C 
489.0303 
452.9328 
452.9328 
454.3845 
Table 4.0: Results for a load demand of 300 MW
Unit 
A 
B 
C 
D 
G1 
95.0530 
95.0530 
95.0530 
95.0530 
G2 
71.1643 
71.1643 
71.1643 
71.1643 
G3 
41.0587 
41.0587 
41.0587 
41.0587 
G4 
20.2745 
20.2745 
20.2745 
20.2745 
G5 
22.7970 
22.7970 
22.7970 
22.7970 
G6 
33.0525 
33.0525 
33.0525 
33.0525 
L 
0.6050 
0.6085 
0.6090 
0.7100 
t 
2.6 
2.7 
3.0 
3.2 
n 
12 
15 
17 
18 
C 
927.6174 
927.6174 
891.7733 
886.5009 
Table 5.0: Results for a power demand of 400MW
Unit 
A 
B 
C 
D 
G1 
125.5410 
125.5410 
125.5410 
125.5410 
G2 
93.9901 
93.9901 
93.9901 
93.9901 
G3 
54.2282 
54.2282 
54.2282 
54.2282 
G4 
26.7775 
26.7775 
26.7775 
26.7775 
G5 
30.1091 
30.1091 
30.1091 
30.1091 
G6 
43.6540 
43.6540 
43.6540 
43.6540 
L 
1.0500 
1.0580 
1.1000 
1.2000 
t 
2.6 
2.9 
3.9 
4.1 
n 
13 
16 
17 
19 
C 
1266 
1266 
1139 
1,084.8 
The computation time and the number of iterations increase with system demand. It should be noted that the parameter gamma () which is set to 1.0 in this case characterizes the variation of the attractiveness, beta ,and it is very crucial in determining the speed of convergence and how the MFA LFDM behaves. Theoretically, [0, but in practice (1) and is determined by the characteristic length of the system to be optimized. By varying the computation speed can be improved
Unit 
A 
B 
C 
D 
G1 
160.7583 
160.7583 
160.7583 
160.7583 
G2 
120.35 
120.35 
120.35 
120.35 
G3 
69.44 
69.44 
69.44 
69.44 
G4 
34.2829 
34.2829 
34.2829 
34.2829 
G5 
38.5554 
38.5554 
38.5554 
38.5554 
G6 
55.900 
55.900 
55.900 
55.900 
L 
1.8010 
1.8080 
2.1015 
2.3050 
t 
3.0 
4.0 
4.5 
5.0 
n 
15 
17 
18 
20 
C 
1719.8 
1719.8 
1221.6 
1,209.9 
Unit 
A 
B 
C 
D 
G1 
160.7583 
160.7583 
160.7583 
160.7583 
G2 
120.35 
120.35 
120.35 
120.35 
G3 
69.44 
69.44 
69.44 
69.44 
G4 
34.2829 
34.2829 
34.2829 
34.2829 
G5 
38.5554 
38.5554 
38.5554 
38.5554 
G6 
55.900 
55.900 
55.900 
55.900 
L 
1.8010 
1.8080 
2.1015 
2.3050 
t 
3.0 
4.0 
4.5 
5.0 
n 
15 
17 
18 
20 
C 
1719.8 
1719.8 
1221.6 
1,209.9 
Table 6.0: Results for a load demand of 500MW
4.2 ED with Cubic Cost Function under Various Demands With the demand of [500MW, 600MW, 700MW, 800MW], the results for the cubic cost function under various demands are tabulated in Table 7.0 .In this case the IEEE 6unit system is used. The optimal cost and the losses increase with power demand. However the Computation time and the number of iterations are not affected by demand in a great extend
Table 7.0: SODED with cubic cost function under various demands
[MW] 
500 
600 
700 
800 
G1 
48.7954 
56.9279 
65.0605 
73.1931 
G2 
37.8673 
44.1786 
50.4898 
56.8010 
G3 
21.4006 
24.9673 
28.5341 
32.1009 
G4 
11.3017 
13.1856 
15.0689 
16.9525 
G5 
12.7317 
14.8537 
16.9757 
19.0976 
G6 
17.9033 
20.88 
23.8711 
26.8549 
L 
2.3065 
2.3090 
2.5000 
2.9950 
t 
5.5 
5.8 
6.2 
6.0 
n 
15 
18 
18 
20 
C 
1,977.1 
3,523.3 
4,211.0 
4,951.29 
SODED with Quadratic and Cubic Cost Functions
.The Algorithms were tested with 3 unit, 5 unit and 26 unit test systems and the results compared with the basic methods; FA, MFA, and MFALF. The system demands considered are 850MW, 1800MW, 2000MW and 2500MW. The results presented are for the 2500MW demand.
From the results in Table 8.0 it, it is clear that the cubic cost functions provide better and more realistic costs (higher costs) than the quadratic cost functions. The MFALFDM method gave the best optimal results as compared to the FA, MFA and MFALF.
SODED wit Cubic Cost Functions
Further comparison was done using the 5unit and the 26 unit systems .The results are as tabulated in table 9.0 10.0. The results are compared with those in [5] since this is the only work that has considered cubic cost functions in DED. From these tables, it can be observed that optimal cost in industrial power systems increases with the complexity of the system. Further, the system losses also are directly proportional to the system size. The execution time and the number of iterations dont vary to a great extend with the system size and the nature of cost function. It is worth noting that the MFALFDM provide better optimal costs, losses and total output power than all the lower versions of FA, GA and PPSO.
V: CONCLUSION
The objective of this paper was to propose a method for solving SODED with cubic cost functions. Cubic cost functions provided more realistic higher costs which are applicable in an industrial setting in a fully constrained environment. MFALFDM proved effective than FA, MFA, MFALF and the basic heuristic methods in the solutions of the industrial cubic SODED, which is a good example of NP hard problems. The pure MFA is also found to be more effective than GA and PSO in cost optimization. This effectiveness is measured in terms of efficiency and success rate. MFALFDM has been found to be very efficient, however a further improvement on the convergence can be achieved by carrying out sensitivity
studies by varying of parameters such as0,, and more interestingly. Other than mutation, other operators of the biologically inspired heuristic methods can also be considered. For more realistic results, a multi objective dynamic economic dispatch (MODED) problem with thermal cubic cost functions need to be considered. That is, the SODED problem need to be considered simultaneously with Renewable energy, transmission losses and emissions. The security and power wheeling aspects under SODED and MODED with higher order cost functions may form an exciting area for further research
Acknowledgement
The authors gratefully acknowledge The Deans Committee Research Grant (DCRG) The University of Nairobi, for funding this research and the Department of Electrical and Information Engineering for providing facilities to carry out this research Work.
Table 8.0: Cubic Cost Function on 3Unit System
Quadratic 
Cubic 

Unit 
FA 
MFA 
MFALF 
MFALFDM 
FA 
MFA 
MFALF 
MFALFDM 
1 
393.170 
393.170 
393.170 
393.169 
725.02 
724.99 
724.99 
724.99 
2 
334.604 
334.604 
334.603 
334.603 
910.19 
910.19 
910.19 
910.19 
3 
122.226 
122.226 
122.226 
122.226 
864.88 
864.88 
864.88 
864.88 
L 
850.00 
850.00 
850.00 
850.00 
2,500.00 
2,500.00 
2,500.00 
2,500.00 
t 
5.0 
6.0 
6.5 
7.0 
5.2 
6.7 
7.0 
8.0 
n 
60 
58 
52 
50 
68 
62 
55 
53 
C 
8,194.35 
8,194.35 
8,193.30 
8,193.20 
12,730.14 
12,729.35 
12,728.15 
12,728.05 
Table 9.0: Five Unit System with Static Cubic Cost Functions
Unit 
GA[5] 
PSO[5] 
MFFA 
MFFALF 
MFFALFDM 
1 
320.00 
319.90 
320.00 
320.05 
320.10 
2 
343.74 
343.70 
343.73 
343.70 
343.74 
3 
472.60 
472.50 
472.40 
472.45 
472.68 
4 
320.00 
320.08 
319.95 
320.00 
320.00 
5 
343.74 
343.77 
343.65 
343.74 
343.74 
L 
1800.00 
1800.00 
1800.00 
1800.00 
1800.00 
t 
8.5 
9.5 
8.5 
9.0 
10.0 
n 
72 
68 
60 
55 
53 
C 
18,611.07 
18,610.40 
18,609.35 
18,609.05 
18,608.65 
Table 10.0: 26Unit System with Static Cubic Cost Functions
Unit 
GA[5] 
PSO[5] 
MFFA 
MFFALF 
MFFALFDM 
19 
2.40 
2.40 
2.40 
2.40 
2.40 
1012 
15.20 
15.20 
15.20 
15.20 
15.20 
1316 
25.00 
25.00 
25.00 
25.00 
25.00 
17 
129.71 
124.69 
124.69 
124.69 
124.69 
18 
124.71 
124.69 
124.69 
124.69 
124.69 
19 
120.42 
120.40 
120.40 
120.40 
120.40 
20 
116.72 
116.70 
116.70 
116.70 
116.70 
2123 
68.95 
68.95 
68.95 
68.95 
68.95 
24 
337.76 
337.85 
337.85 
337.85 
337.85 
2526 
400.00 
400.00 
400.00 
400.00 
400.00 
L 
2000.00 
2000.00 
2000.00 
2000.00 
2000.00 
t 
24 
26 
20 
22 
25 
n 
95 
92 
90 
88 
85 
C 
27,671.24441 
27,671.2276 
27,671.3926 
27,672.1113 
27,672.3345 
REFERENCES

Z.X Liang and J.D Glover Improved cost functions for Economic Dispatch compensations Power Systems IEEE Transactions on Vol 6.pp 821829,1991.

Z.X Liang and J.D Glover A zoom feature for a Dynamic Programming Solution to Economic Dispatch including transmission Loses Power Systems IEEE Transactions on Vol 7.pp 544550, 1992.

A.jiang and S.Ertem Economic Dispatch with nonmonotonically increasin incremental cost units and transmission system losses Power Systems IEEE Transactions on Vol 10.pp 891897, 1995.

Bharathkumar.S et al Multi Objective Economic Load Dispatch using Hybrid Fuzzy, Bacterial Foraging NelderMead Algorithm
International Journal of Electrical Engineering and Technology,Vol 4,Issue 3 pp. 4352 ,May June 2013.

Hari Mohan D.et al A Fuzzy field improved hybrid PSOGSA for Environmental /Economic power dispatch International Journal of Engineering Science and Technology,Vol.6,No.4,pp.1123,2014

Deepak Mishra et al ORNeuron Based Hopfield Neural Network for Solving Economic Load Dispatch Problem Letter and Reviews for Neural Information Processing ,Vol.10,No.11 pp249259,November 2006

N.A Amoli et al Solving Economic Dispatch Problem with Cubic Fuel Cost Function by Firefly Algorithm Proceedings of the 8th International Conference on Technical and Physical Problems of Power Engineering,ostfold University College Fredrikstad,Norway.pp 15,57th September 2012

Krishnamurthy, S.; Tzoneva, R., "Impact of price penalty factors on the solution of the combined economic emission dispatch problem using cubic criterion functions," Power and Energy Society General Meeting, 2012 IEEE , vol., no., pp.1,9, 2226 July 2012

Yusuf Sonmez Estimation of Fuel cost curve parameters for thermal power plants using the ABC Algorithm ,Turkish Journal Of Electrical Engineering and Computer Science ,Vol .21 pp. 1827 1841,2013

T.Adhinarayanan and M.Sydulu Fast and effective Algorithm for Economic Dispatch of Cubic Fuel Cost based thermal units First international conference on industrial and information systems
,ICIIS ,2006 ,Sirlanka ,8th 11th August 2006.

T.Adhinarayanan and M.Sydulu An effective noniterative
algorithm for Economic Dispatch of generators with cubic fuel cost function Electrical power and energy systems ,vol 32,pp 539542,2010

E.B Elanchezhian et al Economic Dispatch with cubic cost models using Teaching learning Algorithm IET Generation,Transmission and Distribution ,vol 8 issue 7 ,pp 11871202,2014

Jizhong Zhu (2009) Optimization of Power System Operation, New Jersey &Canada: John Wiley and Sons.

Yang X.S., Firefly algorithm, stochastic test functions and design optimization,International Journal Of BioInspired Computation, vol. 2, issue 2, pp. 7884, 2010.

Yang X.S., Firefly Algorithm: Recent Advancements and Applications, International Journal Of Swarm Intelligence, vol. pp. 3650, 2013.

Fister I., Iztok Fister I. Jr., Yang X. S., Brest J., A comprehensive Review of Firefly Algorithms, Swarm And Evolutionary Computation, vol. 13, pp. 34 46, 2013.

X.S. Yang,Firefly algorithm, LÂ´evy flights and global optimization, in: Research and Development in Intelligent Systems XXVI (Eds M. Bramer, R. Ellis, M. Petridis), Springer London, pp. 209218, 2010.

Younes M., A Novel Hybrid FFAACO Algorithm for Economic Powe,r Dispatch, Journal of Control Engineering And Applied Informatics, Vol. 15, Issue 2, Pp. 6777, 2013.