237

BALKAN JOURNAL OF ELECTRICAL & COMPUTER ENGINEERING, Vol. 10, No. 3, July 2022

Dynamic Economic Dispatch with Valve-Point

Effect Using Crow Search Algorithm
Cenk Andic, Ali Ozturk and Salih Tosun 

production have production costs so the generators are

Abstract—This paper presents a method based on meta- spending fuel. These fuel they spend are modeled with a
heuristic to solve Dynamic Economic Dispatch (DED) problem in quadratic equation called fuel cost function. The purpose of
a power system. In this paper, Crow Search Algorithm (CSA), the Economic Dispatch (ED) problem is to distribute the
which is one of the heuristic methods is proposed to solve the
DED problem in a power system. In this study, line losses, power required by the loads on the power systems taking into
generation limit values of generators, generation-consumption account the fuel costs of the generators. Thus, more electric
balance, valve-point effect and ramp rate limits of generator are power generation is expected from the generator, which has a
included as constraints. The CSA is implemented on two low fuel cost. However, when performing these loads to the
different test cases. Finally, the CSA results are compared with generators, the lower and upper bound of parameters will be
the results of well-known heuristics in the literature such as taken to produce. The ED problem can be solved by traditional
Particle Swarm Optimization (PSO), Genetic Algorithm (GA),
Symbiotic Organism Search (SOS) algorithm, Artificial Bee methods [2-19]. A more realistic analysis is made by including
Colony (ABC) algorithm, Simulated Annealing (SA), Imperial the valve-point effect of the generators. In this case, the fuel
Competitive Algorithm (ICA), Modified Ant Colony cost function transforms from a quadratic equation to a non-
Optimization (MACO) algorithm. The results show that the convex equation [3-6]. In this case, it becomes difficult to find
proposed algorithm has a better operating cost. With the results the global best solution to the problem. Therefore, heuristic
of the algorithm proposed in the test system 1, a profit of methods are used to solve this problem, which is a non-convex
$2,056,5931 per day and $751,751,4815 per year is obtained. It is
seen that with the results of the algorithm proposed in the test structure [7-31]. [13] used the method of Artificial Neural
system 2, a daily profit of $12,279,7328 and a yearly profit of Networks (ANN) to solve the ED problem.
$4,482,102,472 are obtained. Test systems are operated by using In the literature, when we examine the dissolution of this
less fuel with the results of the proposed algorithm and thus the problem by heuristic methods, we see that the ED problem is
harmful gas emissions released by thermal production units to solved by Genetic Algorithm (GA) in the study of [14]. [15]
the environment are also reduced. Solve the ED problem with Particle Swarm Optimization
(PSO) algorithm. [16] solves with Artificial Bee Colony
Index Terms—Crow search algorithm, dynamic economic (ABC) algorithm. [17] solves the ED problem with Crow
dispatch, thermal power units, valve-point effect Search Algorithm (CSA). [18] has solved the ED problem
using Grey Wolf Optimization (GWO) algorithm.
The ED problem aims to distribute the load value given for
I. INTRODUCTION a only one hour to the generators. However, in operating
conditions, the load value changes continuously [11-33]. In
T HE DEMAND for electrical energy is increasing day by day.
On the contrary, fossil energy resources are decreasing.
Thus, it is an important issue to operate the electrical energy
this case, the Dynamic Economic Dispatch (DED) problem is
solved which takes into account the hourly load change. The
with the least cost [1-3]. load sharing is provided to the generators according to the 24-
Thermal power plants that provide electrical energy hour load change [35]. However, the part to be considered
here should be paid to the production increase or decrease
CENK ANDIC, is with Department of Electrical Engineering University of limits (ramp limit) of generators for hourly load changes [19].
Istanbul Technical University, Istanbul, Turkey, (e-mail: [email protected]). [20] has solved the DED problem using the Symbiotic
https://orcid.org/0000-0003-1123-899X Organisms Search (SOS) algorithm. [21] has solved the DED
ALI OZTURK, is with Department of Electrical-Electronics Engineering
problem using Harmony Search Algorithm (HSA). [22] has
University of Duzce, Duzce, Turkey, (e-mail: [email protected]). solved the DED problem using Simulated Annealing (SA)
https://orcid.org/0000-0002-3609-3603 technique.
In this study, the CSA, which is an heuristic methods is
SALIH TOSUN, is with Department of Electrical-Electronics Engineering
University of Duzce, Duzce, Turkey, (e-mail: [email protected]). proposed to solve the DED problem. The CSA was run on two
different test systems. The results obtained were compared
https://orcid.org/0000-0002-5698-6628
with other results in the literature such as the SA [22], the GA
Manuscript received February 18, 2022; accepted May 08, 2022. [24], the PSO [25], Imperial Competitive Algorithm (ICA)
DOI: 10.17694/bajece.1075860 [26], Sequential Quadratic Programming (SQP) [27], the PSO

Copyright © BAJECE ISSN: 2147-284X http://dergipark.gov.tr/bajece

 238
BALKAN JOURNAL OF ELECTRICAL & COMPUTER ENGINEERING, Vol. 10, No. 3, July 2022

[28], Evolutionary Programming–Sequential Quadratic A. Demand Constraint

Programming (EP-SQP) [29], Particle Swarm Optimization-
Bacterial Foraging (PSO-BF) [30], Chaotic Sequence-
Differential Evolution (CS-DE) algorithm [31], Artificial
Immune System (AIS) [24], the ABC [24], CDBCO [32], the
SOS [20], Modified Ant Colony Optimization (MACO) [33], The sum of the power demanded by the loads (PD) and the
combining Biogeography-Based Optimization with Brain lost power (PL) must be equal to the sum of the power
Storm Optimization (BBOSB) [34] and it was seen that the produced by N thermal power plants (Pİ). There must be a
CSA results found best operating conditions in terms of fuel balance of power. Power loss can be calculated using B-
cost. coefficients, given as follows:

II. PROBLEM FORMULATION OF DED PROBLEM

A large part of electrical energy production is met by
thermal generation units. In order for thermal power plants to
produce electrical energy, they need to burn fuel. The power B. Real Power Operating Limits
demanded by the loads given for any given hour is shared
between the thermal power plants in the power system with The active power produced by thermal power plants has a
certain minimum-maximum limit range. Thermal power plants
the ED problem. It is based on the idea that the thermal
generation unit with less fuel cost should produce more should be operated within this limit range.
electrical energy. In addition, The DED problem allocates the
thermal power plants according to the 24-hour (one-day)
changing load values. The fuel cost function of m thermal C. Generating Unit Ramp Rate Limits
generation units is shown in Eq. (1) below.

where PGi,t and PGi,t-1 are power outputs of the ith

generating units at time t and (t−1), URi and DRi are up and
where, PGi: it is the active power value produced by the ith down ramp-rate limits of ith generating units.
thermal power plant. ai, bi and ci are the fuel cost function co-
efficients of ith thermal power plant. The fuel cost function of III. CROW SEARCH ALGORITHM
the thermal generation units is expressed as a second order The CSA is one of the metaheuristic optimization
function as shown in Eq. (1). However, in reality, valves are algorithms introduced by Askarzadeh in 2016 [23]. The CSA
used to adjust the output power of steam turbines. If there is an is inspired by the intelligent behavior of crows and is a
increase in the output power, the valves are activated in order population-based optimization algorithm. Crows live in flocks
to increase the input power of the plant. However, while the and each crow has a home in its memory. Crows store their
control mechanism trying to achieve this balance tries to excess food in their nests. While one crow takes its excess
maximize efficiency, a ripple occurs in the output power and food to its nest, it is followed by another crow. Two different
this effect is called the valve-point effect [16-17]. In this situations can occur. i) The crow flies to its nest without
study, valve-point effects of steam turbines are included in the noticing that it is being followed and steal it once the owner
fuel cost function. With the valve-point effect, the DED leaves. ii) The crow realizes that it is being followed and flies
problem turns into a non-convex problem and becomes a to a random place to protect the food in its nest. One of these
complex problem without a global solution point. The fuel two situations occurs. We can think of each crow in the flock
cost function of a thermal power plant with the valve-point as a solution to an optimization problem. The steps of the
effect included is shown in Eq. (2) below. crow search algorithm are given below.
(a) Initialize problem and adjustable parameters
(b) Initialize position and memory of crows
(c) Evaluate fitness function
where, di and ei are the fuel cost function co-efficients (d) Generate new position
showing the valve-point effect. PGimin is the minimum limit (e) Check the feasibility of new positions
value that the ith thermal power plant can produce. In this (f) Evaluate fitness function of new positions
study, the DED problem with valve-point effects is subjected (g) Update memory
to constraints given as follows: (h) Check termination criterion

The flowchart of the crow search algorithm is shown in Fig.

1.

Copyright © BAJECE ISSN: 2147-284X http://dergipark.gov.tr/bajece

 239
BALKAN JOURNAL OF ELECTRICAL & COMPUTER ENGINEERING, Vol. 10, No. 3, July 2022

situations. If crow i notices that crow j is following him, crow

i flies to a random place in search space. If crow i does not
notice that crow j is following him, crow i flies to his nest. It is
the Awareness Probability (AP) that determines which of
these two situations will occur. A random number is generated
between 0 and 1, and if this number is greater than the AP, the
condition occurs that the crow does not notice that it is being
followed. The condition that the crow does not notice that it is
following is realized. If the randomly generated number is less
than the AP, the crow realizes that it is being followed and
flies to a random place in the search space.

where rj is a random number with uniform distribution

between 0 and 1. In the CSA, the parameter of awareness
probability is primarily responsible for intensification and
diversification (AP). By lowering the awareness probability
value, the CSA is more likely to focus its search on a local
area where a current good answer can be located.
E. Update Position and Memory
The positions of the crows in the flock change according to
the above conditions and are updated in each iteration. Thus,
Fig. 1. Flowchart of the proposed method the fitness values of crows also change.

IV. CSA IMPLEMENTATION FOR DED PROBLEM

This section shows the steps in implementation the CSA to
the DED problem with valve-point effect. In each iteration, a comparison is made between the crows'
new fitness value and the best fitness value in memory. If the
A. Generate The Initial Population of The Flock new fitness value is less than the fitness value in the memory,
In the crow flock population, there are as many crows as the the fitness value in the memory is updated with the new fitness
problem size in each flock (flock size is N). And each crow value.
represents the solution to the DED problem. For example,
since there are five variables in test case 1, the d-dimensional
should be five. Initially, the positions of the crows in the flock
are randomly generated to fit the lower and upper bounds of Else, the fitness value in the memory continues to be
the variable in the problem. iterated without updating.
F. Stopping Criterion Testing
Iteration continues until it reaches the maximum number of
iterations (itermax).
B. Calculate The Fitness Value of Flock
The fitness values of each crow in the flock are calculated.
Fitness function is shown in Eq. (2).
G. The Sample Problem Solution with CSA
C. Memory Initialization
In the memory of each crow in the flock there is the The function given in Eq. (2) is taken as a sample problem.
location of their nest. Crow’s positions represent the solutions The number of problem dimension (variables) is assigned to 3,
to the problem. At first, the locations of the randomly the flock (population) size is assigned 2, the awareness
generated crow flock are stored in memory as the best value in probability is assigned 0.03, the flight length is assigned 2,
the first iteration. while the iteration number of the algorithm is assigned to 10.
At the first step of the algorithm, the crows positions is
created randomly. This initial population is given in Table I.
D. Generate A New Positions
As a mentioned, crows follow each other and there are two

Copyright © BAJECE ISSN: 2147-284X http://dergipark.gov.tr/bajece

 240
BALKAN JOURNAL OF ELECTRICAL & COMPUTER ENGINEERING, Vol. 10, No. 3, July 2022

TABLE I
THE INITIAL POPULATION OF CROWS CREATED RANDOMLY AND CALCULATED FITNESS VALUE FOR EACH FLOCK

Flock1 Flock2
Hour Demand Crow1 Crow2 Crow3 Cost Crow1 Crow2 Crow3 Cost
1 210 161,323 112,687 159,448 2240495 153,838 134,393 115,797 1945802,8
2 200 190,93 140,522 157,873 2899900 151,367 116,055 178,423 2464525
3 190 178,170 84,947 169,862 2435740 178,140 84,182 170,096 2430118
4 185 176,716 149,925 141,714 2839928 164,319 122,589 165,849 2683737
5 190 85,281 102,629 172,330 1707372 184,784 92,273 114,779 2023760,5
6 215 165,579 140,375 170,903 2625049 157,520 76,486 113,896 1333852
7 220 188,945 113,189 122,894 2056104 195,549 96,863 134,994 2079942
8 230 127,142 132,716 164,797 1952345 111,826 133,939 172,949 1892855
9 235 127,624 100,191 137,913 1312305 160,952 85,957 146,416 1588656,5
10 240 178,407 54,306 134,387 1276128 96,930 134,120 110,868 1023892
11 245 77,258 130,910 157,964 1216344 150,175 123,819 148,483 1780530
12 250 190,439 102,133 125,414 1685595 166,093 109,386 136,184 1622263
13 245 194,946 102,271 113,852 1666344 187,306 84,504 159,335 1867377,7
14 255 164,186 93,547 144,415 1476999 165,604 129,499 157,542 1982606
15 260 175,062 123,981 172,951 2126367 167,132 50,729 177,316 1357274
16 255 156,005 144,196 111,942 1577065 182,675 140,968 131,336 2005997
17 250 175,846 101,357 111,813 1395523 152,122 78,995 129,501 1111173,4
18 245 136,916 118,067 173,495 1840623 177,520 149,674 147,016 2298544
19 240 156,562 111,207 160,417 1887705 142,872 78,609 130,964 1129338
20 235 157,574 126,917 120,716 1707607 178,321 108,320 95,977 1481451
21 230 199,419 92,452 88,407 1508064 194,891 117,766 169,552 2528667,2
22 220 125,233 144,511 137,632 1879316 193,117 122,271 90,801 1867484
23 215 178,391 79,761 132,861 1765529 181,223 91,416 132,517 1907126
24 210 181,547 77,373 130,274 1797323 157,890 103,619 131,102 1831493
Total Cost ($/24h) 44,875,770 44,238,465

As seen in Table 1, it represents the positions of the crows in the flocks. The positions of the crows also represent the solution
to the problem. The fitness value of each flock is calculated according to Eq. (2) for each hour. As a result of the 24-hour
dynamic economic dispatch, the least total cost value that gives the best fitness value in the flock is thrown into memory. Thus,
the best initial value is stored in memory. The value in memory actually shows the best result up to the current iteration. This is
the value for the initial population 44,238,465 $/24h as has been shown in Table 1. Crows store their food in their nests. The
positions of the crows indicate the position of their nests. Crows try to reach more food by following each other. By paying
attention to the flight length of the crows in the search space, they update the positions of the crows in each iteration. The values
of the positions of the crows in the flock in the 9th and 10th iterations are shown in Table 2.
TABLE II
THE CHANGE OF POSITIONS OF THE CROWS ACCORDING TO ITERATIONS

9th Iteration 10th Iteration

Flock1 Flock2 Flock1 Flock2
Crow1 Crow2 Crow3 Crow1 Crow2 Crow3 Crow1 Crow2 Crow3 Crow1 Crow2 Crow3
87,647 93,312 92,680 52,864 117,75 71,889 80,797 88,698 87,359 83,244 79,607 78,845
54,232 38,852 63,340 67,064 45,249 154,24 92,930 50,504 82,684 65,052 64,209 60,864
65,378 93,789 96,936 89,324 82,442 137,10 105,582 60,429 86,689 81,497 70,006 73,643
109,60 72,975 148,89 65,606 92,020 106,48 141,102 118,158 67,268 65,243 68,05 125,077
111,99 75,358 78,835 54,752 107,38 122,90 108,314 53,725 89,442 53,051 105,113 120,574
94,387 89,541 63,819 71,060 146,89 58,613 93,239 51,808 53,369 54,072 65,794 103,341
62,609 78,046 146,14 57,173 85,021 66,761 120,953 49,398 74,921 98,117 42,069 75,329
54,478 125,79 77,353 57,374 39,525 110,66 68,335 75,485 61,819 91,004 90,403 68,074
113,21 47,008 59,094 152,10 47,534 60,495 59,497 53,228 126,374 54,929 62,947 124,018
72,229 58,884 154,78 94,087 83,024 78,361 51,442 93,644 135,180 100,56 58,173 72,826
98,341 59,343 108,28 57,386 105,60 101,09 73,722 131,224 58,081 92,015 63,966 97,555
118,85 76,829 75,123 143,49 93,263 76,928 79,056 125,003 45,030 86,437 50,288 116,249
106,97 80,280 79,289 100,20 41,521 89,394 68,079 134,084 55,447 147,812 41,803 51,268
60,971 149,06 68,675 54,333 116,52 109,11 84,899 37,822 149,266 70,649 49,154 154,483
103,18 92,228 70,052 110,13 82,323 92,328 64,736 60,754 129,643 136,239 42,735 102,041
60,383 40,001 167,36 78,708 53,995 126,25 51,915 142,650 65,446 132,819 70,165 52,515
84,088 76,398 91,046 95,703 41,896 106,24 127,86 67,967 53,429 104,513 85,774 54,839
137,16 107,10 60,286 104,61 44,220 49,508 134,457 105,659 49,816 67,041 74,112 67,253
53,534 91,128 121,96 80,327 78,720 93,448 72,620 106,076 80,313 61,134 73,085 103,440
88,644 44,536 110,69 50,367 91,798 46,280 101,658 44,452 96,242 63,734 92,63 96,988
55,556 89,923 115,25 64,914 54,756 126,58 109,763 47,270 87,860 63,202 42,965 133,289
132,40 74,362 64,425 139,09 69,415 55,619 81,113 61,942 108,126 70,144 59,891 105,077
75,531 111,96 89,399 113,21 43,843 122,36 93,289 61,866 95,770 95,136 41,095 138,648
97,956 53,186 53,222 94,969 121,48 63,415 65,137 88,706 58,348 67,446 43,645 127,242
Total Cost ($/24h) 9,449,683 9,934,455 6,725,579 5,261,095

Copyright © BAJECE ISSN: 2147-284X http://dergipark.gov.tr/bajece

 241
BALKAN JOURNAL OF ELECTRICAL & COMPUTER ENGINEERING, Vol. 10, No. 3, July 2022

It can be seen from Table 1 and 2 that the positions of the crows change in each iteration. In each iteration, the crow’s
positions are updated and the fitness value of the crow flocks is calculated according to these new values. The best fitness value
among the flocks is compared with the best value in the memory, and if it is a better value than the value in the memory, the
memory value is updated with this new fitness value. This comparison process is performed in each iteration. Table 3 shows the
variation of the memory value in each iteration.

TABLE III
THE CHANGE OF THE VALUE IN MEMORY IN EACH ITERATION

Iter_1 Iter_2 Iter_3 Iter_4 Iter_5 Iter_6 Iter_7 Iter_8 Iter_9 Iter_10
44,238,46 36,887,14 30,403,10 26,294,20 22,989,27 19,485,01 16,247,78 12,957,75 9,449,68 5,261,09

At the end of the 10th iteration, the fitness function value is System data for five thermal units is given in Table 6.
seen as $5,261,09 per day. In this way, the dynamic economic TABLE VI
SYSTEM PARAMETERS FOR FIVE-UNIT TEST SYSTEM [20]
dispatch problem is solved by using the crow search
algorithm. Parameters P1 P2 P3 P4 P5
ai ($/h) 25 60 100 120 40
V. SIMULATION RESULTS bi ($/MWh) 2.0 1.8 2.1 2.0 1.8
ci ($/(MW)2h) 0.0080 0.0030 0.0012 0.0010 0.0015
In this study, the CSA was used to solve the DED problem. di ($/h) 100 140 160 180 200
The DED problem includes some constraints such as line ei (1/MW) 0.042 0.040 0.038 0.037 0.035
Pi min (MW) 10 20 30 40 50
losses, ramp-rate limits, valve-point effects and power Pi max (MW) 75 125 175 250 300
balance. Thus, the DED problem become a non-convex UR (MW/h) 30 30 40 50 50
problem. The proposed algorithm has been tested on two DR (MW/h) 30 30 40 50 50
different test systems which are five units with loss and ten The 24-hour changing load data for solving the DED
units without loss. Simulations were done using a 2.50 GHz problem on test case 1 are given in Table 7.
Windows 10 personal computer with 16 GB-RAM and
TABLE VII
MATLAB program package and the CSA was run at least 30 24-HOUR LOAD CHANGE DATA FOR FIVE-UNIT TEST SYSTEM [20]
times for two test cases. The results obtained from the CSA
were compared with the results of other previously reported Hour Load Hour Load Hour Load Hour Load
methods in the literature. Test cases and results have been (h) (MW) (h) (MW) (h) (MW) (h) (MW)
given subsections. The parameters of the CSA used in this 1 410 7 626 13 704 19 654
2 435 8 654 14 690 20 704
study are given in Table 4. 3 475 9 690 15 654 21 680
4 530 10 704 16 580 22 605
TABLE IV 5 558 11 720 17 558 23 527
SETTING PARAMETERS OF THE CSA FOR THE DED PROBLEM
6 608 12 740 18 608 24 463
pd AP fl
Test
(Problem (Awareness (Flight Flock_size Iter_max The results show a total operating cost of 41,030.9994
Cases
Dimension) Probability) Lenght)
Test $/24h. In Table 8, the results (minimum, average and
5 0.3 2 30 3000
Case 1 maximum value) of the CSA are compared with other results
Test
10 0.3 2 40 3000 in the previously reported literature and it is seen that the
Case 2
proposed algorithm gives the best result.
A. Test Case 1
There are five thermal generation units in the test case 1. In TABLE VIII
this case, losses are taken into account and the B co-efficient FUEL COST VALUES FOR TEST CASE 1
is given in Table 5.
TABLE V Method Min. ($/24h) Average ($/24h) Max. ($/24h)
TRANSMISSION LOSS CO-EFFICIENTS FOR FIVE-UNIT TEST SA [22] 47,356.0000 - -
SYSTEM [22] GA [24] 44,862.4200 44,921.7600 45,893.9500
PSO [25] 44,253.2400 45,657.0600 46,402.5200
ABC [25] 44,045.8300 44,064.7300 44,218.6400
ICA [26] 43,117.0500 43,144.4700 43,302.2300
SOS [20] 43,090.5925 43,103.0828 43,162.2146
CSA 41,030.9994 41,079.5397 41,103.6947

Copyright © BAJECE ISSN: 2147-284X http://dergipark.gov.tr/bajece

 242
BALKAN JOURNAL OF ELECTRICAL & COMPUTER ENGINEERING, Vol. 10, No. 3, July 2022

TABLE IX
RESULTS FROM THE CROW SEARCH ALGORITHM FOR TEST CASE 2

Hour P1 (MW) P2 (MW) P3 (MW) P4 (MW) P5 (MW) P6 (MW) P7 (MW) P8 (MW) P9 (MW) P10 (MW) Cost ($/h)
1 153.5076 235.6846 81.6252 60.1069 88.9610 159.9993 129.7902 51.2789 20.0463 55 28.1963
2 174.1104 220.5821 158.5912 60.0262 84.7006 159.9844 129.9999 47.0064 20.0004 55 29.7468
3 207.0047 272.3180 227.2118 60.0006 78.8044 159.9958 129.6627 47.0054 21.0429 55 32.9361
4 224.8680 338.2997 295.7740 60.0042 73.0179 159.9943 129.6720 48.0839 21.2884 55 36.1262
5 194.4103 368.0764 339.9999 60.0000 102.9533 159.9999 129.9999 49.3601 20.2221 55 37.7134
6 273.1290 416.5603 339.9959 70.9246 114.2374 157.6254 129.9905 50.5381 20.0004 55 40.9957
7 348.4823 421.9858 339.9998 62.5551 119.1232 158.0220 129.9992 47.0290 20.0021 55 42.6297
8 413.1924 459.9158 339.2464 62.0455 85.2900 159.9999 129.9999 51.2597 20.0501 55 44.2668
9 469.6148 459.9999 339.9999 75.3451 131.9493 159.9999 129.9999 79.5731 22.5176 55 47.6461
10 469.7190 459.4386 339.3061 124.9983 180.6288 159.9999 129.9758 106.6304 46.3029 55 51.2269
11 469.4233 459.9995 339.9998 122.7567 227.7601 154.6573 129.9998 92.4028 20.0005 55 51.0756
12 468.3962 459.8852 339.9926 169.0843 242.7376 159.9998 129.9989 119.9861 34.9188 55 53.6800
13 449.4786 459.7576 339.9988 141.0363 223.0026 159.9983 129.9988 89.0004 24.7282 55 51.0958
14 408.6655 446.9922 339.2466 91.0022 211.2103 159.8815 129.9997 59.0053 22.9962 55 47.6328
15 355.1128 442.8886 329.4883 73.0002 163.7124 159.9999 129.3275 47.4679 20.0021 55 44.2889
16 275.0022 362.0988 289.7085 66.7447 146.8688 159.9909 129.9944 48.1494 20.4419 55 39.4097
17 195.0020 333.8780 321.4467 60.0063 159.5385 159.9194 128.2062 47.0001 20.0027 55 37.7423
18 203.1714 399.4876 339.9952 63.9597 208.0304 159.9653 129.9993 47.0211 21.3697 55 40.9917
19 282.9999 454.4229 339.7089 76.5863 209.1755 159.9995 129.5644 48.5425 20.0002 55 44.2861
20 361.9901 459.5435 339.9986 117.1854 242.9992 159.5332 129.9970 77.9877 27.7651 55 48.8086
21 399.9707 459.9998 338.1435 68.2714 242.9999 159.9249 129.9999 48.3217 21.3680 55 47.5604
22 323.2210 391.9479 265.8824 61.5723 192.0673 150.8171 120.3835 47.1076 20.0006 55 41.1339
23 248.1792 311.0990 185.8063 60.0043 122.1070 159.9983 112.7484 54.7652 22.2921 55 34.6970
24 168.0059 257.7098 167.4774 60.0001 16.6554 159.9993 129.9676 49.1781 20.0063 55 31.3638
Total Cost ($/24h) 1,005,250.6

Table 9. The results show a total operating cost of 1,005.250.6

B. Test Case 2
$/24h. In Table 12, the results (min., average and max.) of the
CSA are compared with other results in the previously
There are ten thermal generation units in the test case 2. In
reported literature and it is seen that proposed algorithm gives
this case, the DED problem was considered without a
the best result.
transmission losses. System data for ten thermal units is given TABLE XII
in Table 10. FUEL COST VALUES FOR TEST CASE 2

TABLE X Method Min ($/24h) Average ($/24h) Max ($/24h)

SYSTEMS PARAMETERS FOR TEN-UNIT TEST SYSTEM [32] SQP [27] 1,051,163.0000 - -
PSO [28] 1,036,506.0000 1,040,496.0000 -
Parameters P1 P2 P3 P4 P5 P6 P7 P8 P9 P10

ai ($/h) 958.2 1313.6 604.97 471.6 480.29 601.75 502.7 639.4 455.6 692.4 EP-SQP 1,034,100.0000 - -
bi ($/MWh) 21.6 21.05 20.81 23.9 21.62 17.87 16.51 23.23 19.58 22.54 [29]
ci .00043 .00063 .00039 .0007 .00079 .00056 .00211 .0048 .10908 .00951
($/(MW)2h) PSO-BF 1,026,537.2600 1,028,826.7400 1,033,565.2700
di ($/h) 450 600 320 260 280 310 300 340 270 380 [30]
ei (1/MW) .041 .036 .028 .052 .063 .048 .086 .082 .098 .094 CS-DE 1,023,432.0000 1,026,475.0000 1,027,634.0000
Pi min (MW) 150 135 73 60 73 57 20 47 20 55
Pi max (MW) 470 460 340 300 243 160 130 120 80 55
[31]
UR 80 80 80 50 50 50 30 30 30 30 AIS [24] 1,021,980.0000 1,023,156.0000 1,024,973.0000
(MW/h)
ABC [24] 1,021,576.0000 1,022,686.0000 1,024,316.0000
DR 80 80 80 50 50 50 30 30 30 30
(MW/h) CDBCO 1,021,500.0000 1,024,300.0000 -
[32]

The 24-hour changing load data for solving the DED SOS [20] 1,020,894.0757 1,021,072.6846 1,021,194.9972
problem on test case 2 are given in Table 11. MACO 1,019,093.1700 1,019,254.2100 1,024,310.8000
[33]
TABLE XI BBOSB 1,017,530.3328 1,018,487.8504 1,031,843.5673
24-HOUR LOAD CHANGE DATA FOR TEN-UNIT TEST SYSTEM [20] [34]
CSA 1,005,250.6000 1,007,546.0000 1,009,304.0000
Hour Load Hour Load Hour Load Hour Load
(h) (MW) (h) (MW) (h) (MW) (h) (MW)
1 1036 7 1702 13 2072 19 1776
2 1110 8 1776 14 1924 20 2072
VI. CONCLUSION
3 1258 9 1924 15 1776 21 1924 The minimum fuel cost and operating conditions of a power
4 1406 10 2072 16 1554 22 1628 system are found by the economic dispatch problem.
5 1480 11 2146 17 1480 23 1332 However, for a more realistic analysis, this problem should be
6 1628 12 2220 18 1628 24 1184
solved by considering the 24-hour load change. While making
hourly load sharing of generators, ramp rate limits were taken
into consideration. With the proposed algorithm, the DED
The operating value results of the generators are shown in

Copyright © BAJECE ISSN: 2147-284X http://dergipark.gov.tr/bajece

 243
BALKAN JOURNAL OF ELECTRICAL & COMPUTER ENGINEERING, Vol. 10, No. 3, July 2022

problem was solved on two different test systems specified in [11] Yang, H. T., Yang, P. C., & Huang, C. L. (1996). Evolutionary
programming based economic dispatch for units with non-smooth fuel
the literature. The CSA results show that it is operated with cost functions. IEEE transactions on Power Systems, 11(1), 112-118.
the lowest fuel cost in the conditions determined in the test [12] Ross, D. W., & Kim, S. (1980). Dynamic economic dispatch of
case 1. With the results of the CSA, the test system 1 is generation. IEEE transactions on power apparatus and systems, 6(1),
2060-2068.
operated at a cost of $41,030,9994 per day. With the results of
[13] Yalcinöz, T., & Altun, H. (2000). Comparison of simulation algorithms
the SOS algorithm, which is the best result in the literature, the for the Hopfield neural network: an application of economic dispatch.
test system 1 is operated at a cost of $43,090,5925 per day. Turkish Journal of Electrical Engineering & Computer Sciences, 8(1),
With the proposed algorithm, a daily profit of $2,059.5931 is 67-80.
[14] Zou, D., Li, S., Kong, X., Ouyang, H., & Li, Z. (2019). Solving the
obtained. The CSA results show the operation of the system combined heat and power economic dispatch problems by an improved
with the lowest fuel cost under the conditions specified in the genetic algorithm and a new constraint handling strategy. Applied
initial test system (including losses). In the test case 2, the energy, 237, 646-670.
[15] Mahor, A., Prasad, V., & Rangnekar, S. (2009). Economic dispatch
number of thermal generation units increases, but the losses using particle swarm optimization: A review. Renewable and sustainable
are neglected. In this test system, the CSA found the lowest energy reviews, 13(8), 2134-2141.
operating conditions (min., average and max.). The results of [16] Sonmez, Y. (2011). Multi-objective environmental/economic dispatch
solution with penalty factor using Artificial Bee Colony algorithm.
the CSA show that test system 2 is operated at a cost of Scientific Research and Essays, 6(13), 2824-2831.
$1,005,250,6000 per day. With the results of the BBOSB [17] Andic, C., Ozturk, A., & Tosun, S. (2020). Türkiye’deki güc sisteminde
algorithm, which is the best result in the literature, this system karga arama algoritması kullanilarak ekonomik yük dagitimi. Düzce
Üniversitesi Bilim ve Teknoloji Dergisi, 8(1), 428-436.
is operated with a daily cost of $1,017,530,3328. With the [18] Pradhan, M., Roy, P. K., & Pal, T. (2018). Oppositional based grey wolf
proposed algorithm, a daily profit of $12,279.7328 is obtained. optimization algorithm for economic dispatch problem of power system.
The results show that the CSA gives better results than other Ain Shams Engineering Journal, 9(4), 2015-2025.
[19] Al-Bahrani, L., Seyedmahmoudian, M., Horan, B., & Stojcevski, A.
algorithms mentioned in the literature. By using the CSA, a
(2021). Solving the real power limitations in the dynamic economic
yearly profit of $751,751.4815 and $4,482,102.472 are dispatch of large-scale thermal power units under the effects of valve-
provided in test system 1 and 2, respectively. point loading and ramp-rate limitations. Sustainability, 13(3), 1274.
With the CSA algorithm, multi-objective DED problems [20] Sonmez, Y., Kahraman, H. T., Dosoglu, M. K., Guvenc, U., & Duman,
S. (2017). Symbiotic organisms search algorithm for dynamic economic
can also be successfully solved, for example environment dispatch with valve-point effects. Journal of Experimental & Theoretical
emission dispatch and including photovoltaic (PV) and energy Artificial Intelligence, 29(3), 495-515.
storage systems and electric vehicles (EV). [21] Sahoo, A. K., Panigrahi, T. K., Paramguru, J., & Hota, A. P. (2021).
Dynamic economic dispatch using harmony search algorithm. In
advances in machine learning and computational intelligence, 425-435.
REFERENCES [22] Panigrahi, C. K., Chattopadhyay, P. K., Chakrabarti, R. N., & Basu, M.
[1] Vlachogiannis, J. G., & Lee, K. Y. (2009). Economic load dispatch—A (2006). Simulated annealing technique for dynamic economic dispatch.
comparative study on heuristic optimization techniques with an Electric Power Components and Systems, 34, 577–586.
improved coordinated aggregation-based PSO. IEEE Transactions on [23] Askarzadeh, A. (2016). A novel metaheuristic method for solving
Power Systems, 24(2), 991-1001. constrained engineering optimization problems: crow search algorithm.
[2] Jayabarathi, T., Jayaprakash, K., Jeyakumar, D. N., & Raghunathan, T. Computers & Structures, 169, 1-12.
(2005). Evolutionary programming techniques for different kinds of [24] Hemamalini, S., & Simon, S. P. (2011a). Dynamic economic dispatch
economic dispatch problems. Electric power systems research, 73(2), using artificial immune system for units with valve-point effect.
169-176. International Journal of Electrical Power & Energy Systems, 33, 868–
[3] Mandal, B., & Roy, P. K. (2021). Dynamic economic dispatch problem 874.
in hybrid wind based power systems using oppositional based chaotic [25] Hemamalini, S., & Simon, S. P. (2011b). Dynamic economic dispatch
grasshopper optimization algorithm. Journal of Renewable and using artificial bee colony algorithm for units with valve-point effect.
Sustainable Energy, 13(1), 013306. European Transactions on Electrical Power, 21, 70–81.
[4] Xiong, G., & Shi, D. (2018). Hybrid biogeography-based optimization [26] Mohammadi-Ivatloo, B., Rabiee, A., Soroudi, A., & Ehsan, M. (2012).
with brain storm optimization for non-convex dynamic economic Imperialist competitive algorithm for solving non-convex dynamic
dispatch with valve-point effects. Energy, 157, 424-435. economic power dispatch. Energy, 44, 228–240.
[5] Zheng, Z., Li, J., & Han, Y. (2020). An improved invasive weed [27] Attaviriyanupap, P., Kita, H., Tanaka, E., & Hasegawa, J. (2002). A
optimization algorithm for solving dynamic economic dispatch problems hybrid EP and SQP for dynamic economic dispatch with nonsmooth fuel
with valve-point effects. Journal of Experimental & Theoretical cost function. IEEE Transactions on Power Systems, 17, 411–416.
Artificial Intelligence, 32(5), 805-829. [28] Yuan, X., Su, A., Yuan, Y., Nie, H., & Wang, L. (2009). An improved
[6] Dai, C., Hu, Z., & Su, Q. (2021). An adaptive hybrid backtracking PSO for dynamic load dispatch of generators with valve-point effects.
search optimization algorithm for dynamic economic dispatch with Energy, 34, 67–74.
valve-point effects. Energy, 122461. [29] Basu, M. (2013). Hybridization of bee colony optimization and
[7] Younes, Z., Alhamrouni, I., Mekhilef, S., & Reyasudin, M. (2021). A sequential quadratic programming for dynamic economic dispatch.
memory-based gravitational search algorithm for solving economic International Journal of Electrical Power & Energy Systems, 44, 591–
dispatch problem in micro-grid. Ain Shams Engineering Journal, 12(2), 596.
1985-1994. [30] Saber, A. Y. (2012). Economic dispatch using particle swarm
[8] Sinha, N., Chakrabarti, R., & Chattopadhyay, P. K., (2003). optimization with bacterial foraging effect. International Journal of
Evolutionary programming techniques for economic load dispatch. IEEE Electrical Power & Energy Systems, 34, 38–46.
Transactions on Evolutionary Computation, 7(1), 83-94. [31] He, D., Dong, G., Wang, F., & Mao, Z. (2011). Optimization of dynamic
[9] Irving, M.R., & Sterling, M.J.H., (1983). Economic dispatch of active economic dispatch with valve-point effect using chaotic sequence based
power with constraint relaxation. IEEE Proceedings C - Generation, differential evolution algorithms. Energy Conversion and Management,
Transmission and Distribution, 130(4), 172-177. 52, 1026–1032.
[10] Chowdhury, B. H., & Rahman, S., (1990). A Review of recent advances [32] Lu, P., Zhou, J., Zhang, H., Zhang, R., & Wang, C. (2014). Chaotic
in economic dispatch. IEEE Transactions on Power Systems, 5(4), 1248- differential bee colony optimization algorithm for dynamic economic
1259. dispatch problem with valve-point effects. International Journal of
Electrical Power & Energy Systems, 62, 130–143.

Copyright © BAJECE ISSN: 2147-284X http://dergipark.gov.tr/bajece

 244
BALKAN JOURNAL OF ELECTRICAL & COMPUTER ENGINEERING, Vol. 10, No. 3, July 2022

[33] Secui, D. C. (2015). A method based on the ant colony optimization

algorithm for dynamic economic dispatch with valve‐point effects.
International Transactions on Electrical Energy Systems, 25(2), 262-
287.
[34] Xiong, G., & Shi, D. (2018). Hybrid biogeography-based optimization
with brain storm optimization for non-convex dynamic economic
dispatch with valve-point effects. Energy, 157, 424-435.
[35] Gharehchopogh, F. S., Shayanfar, H., & Gholizadeh, H. (2020). A
comprehensive survey on symbiotic organisms search algorithms.
Artificial Intelligence Review, 53(3), 2265-2312.
[36] Zou, D., & Gong, D. (2022). Differential evolution based on migrating
variables for the combined heat and power dynamic economic
dispatch. Energy, 238, 121664.
[37] Ali Shaabani, Y., Seifi, A. R., & Kouhanjani, M. J. (2017). Stochastic
multi-objective optimization of combined heat and power
economic/emission dispatch. Energy, 141, 1892-1904.
[38] Mohammadi-Ivatloo, B., Moradi-Dalvand, M., & Rabiee, A. (2013).
Combined heat and power economic dispatch problem solution using
particle swarm optimization with time varying acceleration coefficients.
Electric Power Systems Research, 95, 9-18.

BIOGRAPHIES
CENK ANDIC received the B.S. degrees in
electrical-electronics engineering from the
University of Duzce, Turkey in 2018 where
he is currently pursuing the Ph.D. degree in
electrical-electronics engineering. He has
been an Research Assistant with the
Electrical Engineering Department,
Istanbul Technical University (ITU). His
research interests include economic dispatch and power
system, state estimation and renewable energy.

ALI OZTURK received his B.S. in

Electrical Engineering Yildiz Technical
University, Turkey, in 1996, his M.Sc. in
Electrical Engineering from Sakarya
University, Sakarya, Turkey, in 2001 and
his Ph.D. in 2007 from the same
university. His research interests include
voltage stability and optimization in
electric power systems.

SALIH TOSUN received his B.S. in Electrical

Education Marmara University, Turkey, in
1986, his M.Sc. in Electrical Engineering
from Marmara University, Istanbul,
Turkey, in 1991. His research interests
include voltage stability and optimization
in electric power systems.

Copyright © BAJECE ISSN: 2147-284X http://dergipark.gov.tr/bajece