Turk J Elec Eng & Comp Sci

() : –

c TÜBİTAK
doi:10.3906/elk-

1 Heuristic based binary Grasshopper optimization algorithm to solve unit

2 commitment problem

3 Muhammad SHAHID1∗, Tahir Nadeem MALIK1 , Ahsan SAID2

1
Department of Electrical Engineering, University of Engineering and Technology taxila, Pakistan
ORCID iD: https://orcid.org/XXXX-XXXX-XXXX-XXXX
2
Maynooth University, Co. Kildare, Maynooth, Ireland,

Received: .201 • Accepted/Published Online: .201 • Final Version: ..201

4 Abstract: The unit commitment problem in power system is a highly nonlinear, non-convex, multi-constrained, complex,
5 highly dimensional, mixed integer and combinatorial generation selection problem. The phenomenon of committing and
6 de-committing represents a discrete problem that requires binary/discrete optimization techniques to tackle with unit
7 commitment optimization problem. The key functions of the unit commitment optimization problem involve deciding
8 which units to commit and then to decide their optimum power (economic dispatch). This paper confers a binary
9 grasshopper optimization algorithm to solve the unit commitment optimization problem under multiple constraints.
10 The grasshopper optimization algorithm is a metaheuristic, multiple solutions-based algorithm inspired by the natural
11 swarming behavior of grasshopper towards food. For solving the binary unit commitment optimization problem, the
12 real/continues value grasshopper optimization algorithm is mapped into binary/discrete search-space by using an S-
13 shaped sigmoid function. The proposed algorithm is tested on IEEE benchmark systems of 4, 5, 6, 10, 20, 26, 40, 60,
14 80 and 100 generating units including the IEEE 118-bus system and the results are compared with different classical,
15 heuristics, metaheuristics, quantum, and hybrid approaches. The results confer better performance of binary grasshopper
16 optimization algorithm to solve the unit commitment optimization problem.

17 Key words: Heuristic, unit commitment, optimal scheduling, constraints handling, power operation, grasshopper opti-
18 mization algorithm.
19

20 1. Introduction
21 In a smart power system network, the increasing demand for energy is fulfilled by different conventional (hydro,
22 thermal and nuclear) and non-conventional (solar, tidal and wind) energy resources. Also, the load demand curve
23 is not constant but changes with time along different peaks. Thus, it is an essential requirement of the power
24 system operation to optimally decide which unit (generator) to turn on (committed) and which unit to turn off
25 (de-committed). This whole optimal decision-making process of on/off, committed/de-committed and selection/
26 not selection of units under system, unit, network, security, environmental and cost constraints is termed as
27 unit commitment optimization problem. Unit commitment (UC) is important to thermal power plants only, not
28 for hydro and nuclear (base load) power plants. So due to boiler operational constraints, thermal units cannot
29 be turned on immediately to fulfill the power demand. The proper generator selection is an essential feature to
30 fulfill the load demand with ample reverse generating capacity in order to avoid malfunctions and failures under
∗ Correspondence: [email protected]

1
This work is licensed under a Creative Commons Attribution 4.0 International License.
 AUTHOR and AUTHOR and AUTHOR/Turk J Elec Eng & Comp Sci

1 severe conditions. The earliest techniques to solve the UC optimization problem include classical, conventional,
2 traditional and gradient-based optimizer such as dynamic programming (DP) [1], Lagrangian relaxation (LR),
3 extended Lagrangian relaxation (ELR), dynamic programming Lagrangian relaxation [2], mixed-integer linear
4 programming (MILP) [3], Priority list approach (PL) [4], branch and bound (BB).These techniques suffer from
5 numerous iterations, local optima stagnation, premature convergence, larger execution time and parameter
6 sensitivity. The heuristic and evolutionary techniques are based on mimicking natural phenomena. These
7 natures inspired algorithms are also adopted to optimize the UC problem under different constraints. These
8 heuristic techniques include genetic algorithm (GA) which is based on the principle of biological evolution and
9 natural selection of genes for the survival of the fittest [5].
10 Ant colony optimization (ACO) [6] and particle swarm optimization (PSO) [7] mimics the social inter-
11 action, behavior, and coordination among the swarms. There is an extended list of nature inspired algorithms
12 such as binary grey wolf optimization algorithm (BGWO) [8], ring crossover genetic algorithm (GA) [9], firefly
13 algorithm (FF) [10], evolutionary algorithms (EA) [11] and improve binary cuckoo search (IBCS) [12] etc. to
14 solve the UC optimization problem. Some hybrid approaches compromising of the benefits of both heuristic and
15 classical techniques such as Lagrangian relaxation PSO [13] and Lagrangian relaxation GA [14] are developed
16 to improve the solution quality of UC optimization problem. With low population size evolutionary quantum
17 approaches improve the exploitation and exploration of the evolutionary techniques as compared to the oth-
18 ers evolutionary techniques. Hybridization of two natured inspired algorithms such as PSO-GWO [15] is also
19 investigated to UC optimization problem for better solution quality.
20 Recently Seyedali Mirjalili proposed multi solutions based metaheuristic approach named as grasshopper
21 optimization algorithm (GOA) which mimics swarming behavior of grasshopper in nature to solve the optimiza-
22 tion problems. This is used to solve the multiple real value and binary optimization problems such as optimal
23 reconfiguration for partial shaded photo voltaic array [16], frequency control of the load for interconnected
24 multi area micro grid power system by tuning the gains of Fuzzy proportional integral derivative (Fuzzy PID)
25 controller through GOA [17], economic load dispatch (ED) with renewable energy (wind mill) integration [18],
26 optimal selection of conductor for radial distribution system [19], optimal control of voltage and frequency for
27 an islanded micro grid [20], feature selection problem and short term load forecasting specific to the region [21]
28 etc. Inspired by the successful applications of GOA and BGOA (binary grasshopper optimization algorithm) to
29 research and industrial problems, this paper proposes a binary grasshopper optimization algorithm to solve the
30 combinatorial generation selection problem with a better solution quality as compared to the traditional GOA.
31 The paper is organized in such a way that section 2 and 3 explain UC formulation including constraints.
32 The principles of grasshopper optimizer and binary grasshopper optimizer are described in section 4. Section
33 5 describes the BGOA-UC approach. Section 6 presents the IEEE test systems, computational results, per-
34 formance, statistical significance and parametric analysis. Conclusion and contributions are demonstrated in
35 section 7.

36 2. Unit commitment
37 The power system includes hydro thermal coordination, load forecasting, unit commitment, and economic
38 dispatch. UC optimization problem is defined as the optimal turn off and turn on the schedule of a set of
39 generators to obtain a minimum production cost for given power demand satisfying the physical and system
40 operational constraints. Production cost consists of fuel cost, shutdown cost, and startup cost. Constraints that
41 must be handled are: (a) Total power of generated units should be equal to the power demand for the given

2
 AUTHOR and AUTHOR and AUTHOR/Turk J Elec Eng & Comp Sci

1 hour, (b) In case of any shortfall of generated power, there should be a sufficient amount of spinning reserve,
2 (c) Power for each generator should be within its minimum and maximum rating for a given hour, (d) Every
3 generating unit must satisfy the minimum uptime and minimum downtime.

4 3. Problem formulation
5 The formulation of the UC binary minimization problem consists of a single objective function and multiple
6 constraints. The main objective of the problem is to find out the optimal on/off schedule of the generating units
7 to obtain minimum operating cost including fuel cost, shut down cost and start-up cost subject to the power
8 demand, generation limit, spinning reserve and minimum up and downtime constraints within a specific time
9 period. Electricity generation cost is an essential index in the power sector. The highest contribution to the
10 operational cost for the power plant consists of fuel costs. In order to minimize the tariff, the fuel cost should be
11 minimized by allocating the optimal power to the given generators. All the thermal committed units include the
12 cost in order to satisfy the minimum power limits and these on units are economically dispatched to minimize
13 the overall cost of the fuel. The calculation of fuel cost is done on the basis of data given by the characteristics
14 of generating units. Price of fuel, heat rates, initial status, turn-on and turn-off time are the characteristics,
15 which are mathematically represented by a non-convex and non-smooth quadratic equation with power output
16 and without losses using economic dispatch (ED) of the load as as equation (1).

Fi (Pih ) = ai Pi2 + bi Pi + ci (1)

17 Where i is thermal unit index, Fi = fuel cost function, Pi = power of unit i, ai , bi , ci = fuel cost
18 coefficients.
19 Startup cost represents the cost to restart a de-committed unit depending on the committed and de-
20 committed states of the unit. It varies with the temperature of the boiler as a hot start, cold start and warm
21 start (banking). when going back to the committed state from the de-committed state, the startup cost also
22 depends on the de-committed hours of the unit. If de-committed hours of the unit are greater than or equal to
23 the cold startup hours after the minimum de-committed time, cold start cost is related to the committed event.
24 However, if the de-committed hours of the unit are less than cold start-up hours, the hot start cost is related
25 to the committed event. The mathematical equation for the startup cost is given by the following equation.
(
HSCi ; f or M DTi ≤ M DTiON ≤ (M DTi + CSHi )
SU Cih = (2)
CSCi ; f or M DTiON > (M DTi + CSHi )

26 Where h is scheduling hour index, SU Cih is start-up cost, CSHi and CSCi are hot start and cold start
27 cost, M DTi and M DTiON are minimum down time and continuously on time, respectively.
28 In the proposed methodology to solve the UC optimization problem shutdown cost is neglected, which
29 is often given as a constant value for the de-committed status of the corresponding unit. The total cost of fuel
30 (TC) for the specific interval h and on/off status of the unit i unit Uih include shutdown cost, startup cost
31 and fuel cost is given by the equation (3).

H X
X N
TC = Fi (Pih ) ∗ Uih + SU Cih (1 − Ui(h−1) ) ∗ Uih (3)
h=1 i=1

32 Where N and H are total number of units and scheduling hours respectively.

3
 AUTHOR and AUTHOR and AUTHOR/Turk J Elec Eng & Comp Sci

1 3.1. Constraints
2 The maximum power of each generator is limited due to the corresponding thermal consideration and cannot
3 produce power more than its rated value. The minimum power of each unit is limited due to the stability issue
4 of the machine. For an optimal operation of the given system, if the generated power of the unit is less than
5 its rated value Pmin then this unit cannot be put on the related bus bar. Thus actual generated power of each
6 unit should satisfy the generation limit as by equation (4).

Pimin < Pi < Pimax (4)

7 Where Pimin and Pimax are minimum and maximum generation limits, respectively. Power balance or
8 load balance constraint satisfaction requires the sum of generated power of all the committing units at specific
9 hour h should be greater than or equal to the power demand at that specific hour h.

N
X
Pih Uih ≥ Pd (5)
i=1

10 Where Pd is the system load. Spinning reserve SRt requirement is implemented to satisfy the adequate
11 online capacity of a generation which is needed in case of running unit failure or sudden increase of the load
12 demand. This generated adequate power capacity is termed as spinning reserve and mathematically represented
13 by the following equation.

N
X
Pih Uih ≥ Pd + SRt (6)
i=1

14 The switching states viz. committed and de-committed for the thermal generating units in the total
15 time range is constrained by considering the operating features of the generating units. Due to the boiler
16 operating characteristics, there must be a predefined time horizon between committed and de-committed state.
17 A generating unit must remain committed for this specific time horizon when it is ON termed as a minimum
18 uptime.

ON
T(i,h) ≥ M U Ti (7)

19 A generating unit must remain de-committed (OFF) for the specific time horizon when it is OFF termed as a
20 minimum downtime.

OF F
T(i,h) ≥ M DTi (8)

ON OF F
21 Where T(i,h) and T(i,h) are continuously on and off time for unit i, respectively. M U Ti is the minimum
22 up time.
23 At the start of organizing the time horizon UC optimization problem parameters like uptime and
24 downtime constraint, startup cost, etc. are affected by the committed and de-committed status of the generating
25 thermal units. Thus initial status for every operating unit must take the previous day’s prior scheduling time
26 horizon into account in order to satisfy the minimum uptime and downtime.

4
 AUTHOR and AUTHOR and AUTHOR/Turk J Elec Eng & Comp Sci

1 4. Grasshopper Optimizer (GO)

2 4.1. Overview of GO
3 Grasshoppers are basically insects. Due to their severe damaging effect on agriculture and crop production,
4 they are often considered a pest. Their life-cycle consists of three stages as egg, Nymph, and adult. There is
5 no caterpillar stage in their two-month whole life cycle due to incomplete metamorphosis. The nymph stage is
6 different from adults due to smaller size, no wings, and no reproductive organs. They join to make the largest
7 swarm in nature as compared to other creatures. The swarming size of grasshoppers may be as large as the
8 continental scale. Nymph and adult both show the swarming behavior is the interesting aspect of their swarm
9 in nature. Billions of Nymph move just like spinning cylinders and jump with a short distance. In their way,
10 they eat all agriculture and crop production.
11 Grasshoppers in their adult form make a huge swarm to travel a long distance. In the nymph phase,
12 small steps and slow movement is the main feature of the swarming behavior (exploitation). In contrast, in
13 the adult phase abrupt and long-range movement is the main property of the swarming behavior (exploration).
14 For nature-inspired optimization algorithms; the searching process consists of two steps: exploitation and
15 exploration. Grasshoppers performed these two steps and target seeking naturally. Their natural behavior
16 to tackle optimization problems is mathematically modeled by Seyedali Mirjalili [22]. He proposed a model
17 in order to simulate the attraction (exploitation) and repulsion (exploration) forces among the grasshoppers.
18 Exploitation takes place during the attraction forces towards a local solution while repulsion forces cause to
19 explore the search space to a global optimum. In order to make a balance between exploitation and exploration,
20 Grasshopper optimizer (GO) is provided with an adaptive coefficient to change the comfort zone for the
21 grasshoppers.

22 4.2. Continues values grasshopper optimization (GOA)

23 The continuous-valued mathematical model of the swarming behavior of the grasshoppers in nature to simulate
24 the social interaction forces is given by equation 9.

Xk = r1 Sk + r2 Gk + r3 Ak (9)

25 Where k is the grasshopper index Xk is the position, Sk is the force of social interaction (attraction or
26 repulsion), Gk is the force of gravity and Ak is the wind propagation of grasshopper k , r1 , r2 and r3 are
27 random numbers between [0, 1].Where the strength of social forces is given by the function Sk .

Ngs
X
Sk = Sf (dkl )(dˆkl ) (10)
l

28 dkl is the distance between grasshopper l and k which is calculated as dkl = |Xl − Xk | , Ngs is the total
29 number of the grasshoppers (search agents) and dˆkl is the unit vector between grasshopper l and k

Xl − Xk
dˆkl = (11)
dkl

30 Social forces are defined by Sf function, calculated by the equation (12).

r
Sf = f e l − e−r (12)

5
 AUTHOR and AUTHOR and AUTHOR/Turk J Elec Eng & Comp Sci

1 f is the intensity of attraction=0.5, l is the length scale of attraction=1.5, f and l change the social behavior
2 (force of repulsion and attraction) of the grasshoppers to a large extent. Comfort zone, repulsion region and
3 attraction region of the grasshoppers are significantly changed by f and l . Gk is calculated by the following
4 equation.

Gk = −gêg (13)

5 Where g represents the gravitational constant and êg indicates a unity vector towards the center of the
6 earth. The Ak component of the equation (7) is calculated by the following equation

Ak = µêk (14)

7 Where µ represents drift/flow constant and êk is the unit vector towards the wind flow direction.
8 Substituting the value of Sk , Gk and Ak in the equation in equation (9), equation (15) is obtained.

Ngs
X Xl − Xk
Xk = Sf (Xl − Xk ) ∗ − gêg + µêk (15)
dkl
l

9 The above equation is used to simulate the interaction between grasshoppers in the swarm. However,
10 the above model cannot be used directly to simulation purposes due to improper convergence. So the modified
11 equation used for optimization is (16).

Ngs
X 
ub − lb Xl − Xk
Xkd = α1 α2 d d
∗ Sf (|Xl − Xk |) ∗ ∗ T̂d (16)
2 dkl
j=l

12 Where ub represents upper bound, lb is the lower bound, T̂d represents the best solution obtained so
13 far α1 is just like an inertial weight in particle swarm optimization (PSO) in order to make a balance between
14 exploitation and exploration around the optimum target and α2 causes to decrease the comfort zone, repulsion
15 zone, and attraction zone. It reduces the comfort zone with proportional to iterations as below equation.

αmax − αmin
α = αmax − 1 (17)
L

16 Where L is maximum number of iterations, αmax and αmin are set as 1 and 0.00001, respectively.

17 4.3. Binary values GO (BGOA)

18 UC is a binary problem consisting of 1 and 0 states at each interval of time and iteration. So, in order to optimize
19 this problem real value to binary value mapping has to be done. In this BGOA the position vector of search
20 agents/grasshopper, search space and food location is mapped into the binary value by an s-shaped sigmoid
21 function. Due to easy computation and the ability to differentiate across the entire domain, the sigmoid function
22 is quite useful for binary conversion. Equation (18) represents a simple s-shaped sigmoid function proposed by
23 Kennedy and Eberhart (1997).
1
T (∆Xt ) = (18)
1 + e−Xt

6
 AUTHOR and AUTHOR and AUTHOR/Turk J Elec Eng & Comp Sci

1 ∆Xt is velocity of a grasshopper for a specific iteration. The position of each search agent will be updated
2 according to the equation (19) depending on the probability value of ∆Xt .
(
d
d 1 if r1 < T (Xk+1 )
Xk+1 = d
(19)
0 if r1 ≥ T (Xk+1 )

d
3 Where Xk+1 is the dkh dimension of the search agent in the next iteration and r1 is a random number
4 [0, 1].

5 5. BGOA implementation to UC optimization problem

6 In this paper, the BGOA is used to obtain optimal feasible commitment scheduling of thermal units. The power
7 allocation among the committed thermal units through economic dispatch (ED) is observed using the Quadratic
8 Programming technique. The generalized form for the UC optimization problem is shown by a flow chart in
9 Fig.1.
10 A search agent represents a commitment schedule of a unit within a given time horizon. Mathematically
11 it is described in the form of integer matrix U with N ∗ H matrix. Where uti represents on/off status of ith
12 unit at a time interval of t . A random set of search agents is formed in this initial process. The element uti is
13 generated either 1 or 0 by a random function.
 1
u21 u31 uH

u1 ... 1
 u12 u22 u32 ... uH 
2 
 1
 u3 u23 u33 ... uH 
3 
U = . . . (20)
 ... .... .... ...

. . . ... .... .... ...
u1N u2N u3N ... uH
N

14 The priority list for the UC optimization problem is based on fuel cost acquired from full-load average
15 production cost. The unit having the lowest cost value will be at the highest priority. This unit scheduling may
16 not be satisfied with the minimum up and minimum downtime constraints. In order to determine the violation
17 of these constraints, on and off times of the given units should be computed by the equation (21). In order to
18 repair these types of constraints, heuristic adjustment is used.
(
t−1
t Ti,on + 1 if uti = 1
Ti,on = (21)
0 if uti = 0
(
t−1 t
t Ti,of f + 1 if ui = 1
Ti,of f =
0 if uti = 0

19 The reliability of the system depends upon the spinning reserve and load demand satisfaction. For all
20 the tested system spinning reserve is 10% and 5% . Modification of the search agents in order to repair the
21 minimum up and downtime constraints may extend the spinning reserves of corresponding units. Power is
22 allocated to each committed unit through economic dispatch using Quadratic programming. The fitness of each
23 search agent is calculated by equation (3). After the fitness calculation, each search agent is sorted according to
24 fitness values. BGOA is used to update the position of each search agent. Termination criteria are the number
25 of iterations. Upon reaching the maximum number of iterations optimal scheduling of the units with optimal
26 power allocation is obtained.

7
 AUTHOR and AUTHOR and AUTHOR/Turk J Elec Eng & Comp Sci

Start Take all the grasshopper position

from main algorithm

Initialize all the parameters of Initialize the parameters of all

BGOA thermal units

Set k =1 Set t=1

Convert Real to binary

value using sigmoid
No

function
Initialize the grasshopper position

t=t+1
within search space Return
If t<=H

Heuristic Adjustment
Fitness calculation of each yes
grasshopper using equation (3)
Apply the minimum
No up/down time
algorithm satisfying equation (7) and
Find out the best grasshopper (8)
according to fitness value

No Apply the spinning reserve algorithm

satisfying equation (6)
k <=max itr stop
t=t+1

yes
Apply the de commitment algorithm
Update the position of grasshopper
using equation (18) and (19)

Fitness calculation of each

grasshopper using equation (3)

Best grasshopper based on

fitness value

Figure 1 Flow chart of UC optimization problem by BGOA

Figure 1. Flow chart of UC optimization problem by BGOA.

1 6. Numerical results, comparison, and discussion

2 The proposed BGOA is modeled to find the solution of UC optimization problem with IEEE benchmark systems
3 of 4 [23],5 [24],6 [24],10 [24],20 [23],26 [24],40 [23], 60 [23], 80 [23] and 100 [23] generating units including IEEE
4 118-bus system. 24 hour time horizon is utilized for the 5 to 100 units while for 4-unit system 8 h time horizon is
5 used. MATLAB R2016a software is used to implement unit commitment problem by grasshopper optimization
6 with Intel(R) Core(TM) i3-4030 CPU @ 1.90GHz processor. By observing the effect of changing population size
7 on the total cost, 30 is chosen as the best population size. By observing optimality, the controlling parameters

8
AUTHOR and AUTHOR and AUTHOR/Turk J Elec Eng & Comp Sci

Figure 2. Convergence curves of IEEE unit systems.

9
 AUTHOR and AUTHOR and AUTHOR/Turk J Elec Eng & Comp Sci

Cost vs Number of Trials

6000000

5000000

4000000
Total Cost ($)

3000000

2000000

1000000

0
1 6 11 16 21 26
Number of Trials
IEEE 4-unit system IEEE 5-unit system IEEE 6-unit system
IEEE 10-unit system IEEE 20-unit system IEEE 26-unit system
IEEE 40-unit system IEEE 60-unit system IEEE 80-unit system
IEEE 100-unit system IEEE 118-bus system

Figure 3.Figure 1 Variations

Variations of of
of total cost total cost of
different different
units units with
with different trials. different trials

1 l , f and interval are set to 1.5, 0.5 and [0 2.079], respectively.

2 6.1. Performance of BGOA for small test systems

3 The small test system includes 4, 5, 6 and 10-unit systems. The best optimal solution for IEEE 4, 5 and 6-unit
4 system for 8 h and 12 h is obtained using BGOA and its comparison with different approaches is described in
5 Table 5. Table 1 shows the simulation results for the IEEE 10-unit system and its comparison is given in the
6 Table 5. Their convergence characteristics and cost variation w.r.t different trials are shown in Fig. 2 and 3,

10
 AUTHOR and AUTHOR and AUTHOR/Turk J Elec Eng & Comp Sci

Table 1. Unit commitment of IEEE 10 unit system by BGOA.

Table 1 unit commitment of IEEE 10 unit systems by BGOA

Hour Unit1-10
Commitment Power allocation Fuel cost ($) Startup cost ($)
1 1100000000 455 245 0 0 0 0 0 0 0 0 13683 0
2 1100000000 455 295 0 0 0 0 0 0 0 0 14554 0
3 1101000000 455 265 0 130 0 0 0 0 0 0 16892 560
4 1101000000 455 365 0 130 0 0 0 0 0 0 18638 0
5 1111000000 455 285 130 130 0 0 0 0 0 0 20133 550
6 1111100000 455 360 130 130 25 0 0 0 0 0 22387 900
7 1111100000 455 410 130 130 25 0 0 0 0 0 23262 0
8 1111100000 455 455 130 130 30 0 0 0 0 0 24150 0
9 1111110000 455 455 130 130 110 20 0 0 0 0 26589 340
10 1111111100 455 455 130 130 162 33 25 10 0 0 30058 580
11 1111111110 455 455 130 130 162 73 25 10 10 0 31916 60
12 1111111111 455 455 130 130 162 80 25 43 10 10 33890 60
13 1111111100 455 455 130 130 162 33 25 10 0 0 30058 0
14 1111110000 455 455 130 130 110 20 0 0 0 0 26589 0
15 1111100000 455 455 130 130 30 0 0 0 0 0 24150 0
16 1111100000 455 310 130 130 25 0 0 0 0 0 21514 0
17 1111100000 455 260 130 130 25 0 0 0 0 0 20642 0
18 1111100000 455 360 130 130 25 0 0 0 0 0 22387 0
19 1111100000 455 455 130 130 30 0 0 0 0 0 24150 0
20 1111111100 455 455 130 130 162 33 25 10 0 0 30058 750
21 1111111000 455 455 130 130 85 20 25 0 0 0 27251 0
22 1111011000 455 340 130 130 0 20 25 0 0 0 23085 0
23 1101000000 455 315 0 130 0 0 0 0 0 0 17764 0
24 1100000000 455 345 0 0 0 0 0 0 0 0 15427 0
Total cost ($) = 559226.576292+ 3800.00= 563026.576292
Elapsed time 30.73 seconds

Table 2. Unit commitment of IEEE 40 unit system by BGOA.

Table 5 unit commitment of IEEE 40 unit systems by BGOA

Hour Unit commitment of 1-40 units by BGOA Fuel Cost ($) Startup cost ($)
1 1100000000110000000011000000001000000000 53787 0
2 1100000000110000000011000000001100000000 58218 5000
3 1101000000110000000011000000001100000000 65794 560
4 1101000000110100000011010000001101000000 74551 1680
5 1111000000111100000011010000001101000000 79285 1100
6 1111100000111100000011110000001111000000 88025 2000
7 1111100000111110000011110000001111000000 92034 900
8 1111100000111110000011111000001111100000 96601 1800
9 1111110000111111000011111100001111110000 1.0636e+05 1360
10 1111111100111111110011111110001111111000 1.1885e+05 2200
11 1111111110111111111011111111001111111100 1.2626e+05 240
12 1111111111111111111111111111101111111110 1.3419e+05 240
13 1111111100111111110011111110001111111000 1.1885e+05 0
14 1111110000111111000011111100001111110000 1.0636e+05 0
15 1111100000111110000011111000001111100000 96601 0
16 1111100000111110000011111000001111100000 86055 0
17 1111100000111110000011111000001111100000 82567 0
18 1111100000111110000011111000001111100000 89548 0
19 1111100000111110000011111000001111100000 96601 0
20 1111111100111111110011111110001111111000 1.1885e+05 2880
21 1111011000111111100011111110001111111000 1.0859e+05 0
22 1101011000110111100011010110001100011000 90342 0
23 1101000000110010000011000000001100000000 69801 0
24 1100000000110010000011000000001100000000 62218 0
Total cost ($) = 2220317.766613+19960.00 = 2240277.7666
Elapsed time 116.925396 seconds

1 respectively.

11
 AUTHOR and AUTHOR and AUTHOR/Turk J Elec Eng & Comp Sci

Table 3. Unit commitment of IEEE 60 unit system by BGOA.

Table 6 unit commitment of IEEE 60 unit systems by BGOA

Hour Unit commitment of 1-60 units by BGOA Fuel Cost ($) Startup
cost ($)
1 110000000011000000001100000000110000000011000000001000000000 81151 0
2 110000000011000000001100000000110000000011000000001000000000 86389 0
3 110100000011000000001100000000110000000011000000001100000000 98397 5560
4 110100000011010000001101000000110100000011010000001101000000 1.1183e+05 2800
5 111100000011110000001111000000110100000011010000001101000000 1.1893e+05 1650
6 111110000011110000001111000000111100000011110000001111000000 1.3178e+05 2550
7 111110000011111000001111100000111100000011110000001111000000 1.3805e+05 1800
8 111110000011111000001111100000111100000011110100001111110000 1.4485e+05 1580
9 111111000011111100001111110000111111000011111100101111110000 1.6027e+05 3220
10 111111110011111111001111111100111111100011111110001111111000 1.7827e+05 3300
11 111111111011111111101111111110111111110011111111001111111100 1.8938e+05 360
12 111111111111111111111111111111111111111011111111101111111110 2.0129e+05 360
13 111111110011111111001111111100111111100011111110001111111000 1.7827e+05 0
14 111111000011111100001111110000111111000011111100001111110000 1.5953e+05 0
15 111110000011111000001111100000111110000011111000001111100000 1.449e+05 0
16 111110000011111000001111100000111110000011111000001111100000 1.2908e+05 0
17 111110000011111000001111100000111110000011111000001111100000 1.2385e+05 0
18 111110000011111000001111100000111110000011111000001111100000 1.3432e+05 0
19 111110000011111000001111100000111110000011111000001111100000 1.449e+05 0
20 111111110011111111001111111100111111100011111110001111111000 1.7827e+05 4320
21 111111100011111110001111111000111101100011110110001101111000 1.6251e+05 0
22 110111100011011110001100011000110001100011010110001100011000 1.3569e+05 0
23 110010000011001000001100000000110000000011000100001100000000 1.0455e+05 0
24 110010000011000000001100000000110000000011000100001000000000 92614 0
Total cost ($) = 3329074.112433+27500.00= 3356574.112433
Elapsed time 201.706896 seconds

Table 4. Unit commitment of IEEE 80 unit system by BGOA.

Table 7 unit commitment of IEEE 80 unit systems by BGOA

Hour Unit commitment of 1-80 units by BGOA Fuel Cost Startup

($) cost ($)
1 11000000001100000000110000000011000000001100000000110000000010000000001000000000 1.0757e+05 0
2 11000000001100000000110000000011000000001100000000110000000011000000001000000000 1.155e+05 5000
3 11000000001100000000110000000011000000001100000000110000000011000000001101000000 1.31e+05 5560
4 11010000001101000000110100000011010000001101000000110100000011010000001101000000 1.491e+05 3920
5 11110000001111000000111100000011010000001101000000110100000011010000001101000000 1.5795e+05 1650
6 11111000001111100000111100000011110000001111000000111100000011110000001101000000 1.7543e+05 4000
7 11111000001111100000111110000011110000001111000000111100000011110000001111100000 1.8407e+05 2350
8 11111000001111100000111110000011111000001111100000111110000011110000001111100000 1.9276e+05 2700
9 11111100001111110000111111000011111100001111110000111111000011111100001111110000 2.1271e+05 3620
10 11111111001111111100111111110011111111001111111000111111100011111110001111111000 2.3769e+05 4400
11 11111111101111111110111111111011111111101111111100111111110011111111001111111100 2.5251e+05 480
12 11111111111111111111111111111111111111101111111110111111111011111111101111111110 2.677e+05 420
13 11111111001111111100111111110011111111001111111000111111100011111110001111111000 2.3769e+05 0
14 11111100001111110000111111000011111100001111110000111111000011111100001111110000 2.1271e+05 0
15 11111000001111100000111110000011111000001101100000111110000011111100001111100000 1.9333e+05 0
16 11111000001111100000111110000011111000001101100000111110000011111000001111100000 1.7149e+05 0
17 11111000001111100000111110000011111000001101100000111110000011111000001111100000 1.6451e+05 0
18 11111000001111100000111110000011111000001101100000111110000011111000001111100000 1.7848e+05 0
19 11111000001111100000111110000011111000001101100000111110000011111000001111100000 1.9291e+05 0
20 11111111001111111100111111110011111111001111111000111111100011111110001111111000 2.3769e+05 6310
21 11111110001111111000111111100011110110001111011000111111100011110110001111011000 2.1649e+05 0
22 11010110001101011000110101100011010110001111011000110111100011000110001100011000 1.8056e+05 0
23 11010000001100000000110000000011000000001110000000110010100011000000001100010000 1.4033e+05 0
24 11000000001100000000110000000011000000001110000000110010100011000000001000010000 1.2482e+05 0
Total cost ($) = 4434997.754713+40410.00= 4475407.754713
Elapsed time 246.984534 seconds

12
 AUTHOR and AUTHOR and AUTHOR/Turk J Elec Eng & Comp Sci

Table 5. Comparison of unit commitment best cost ($) and elapsed time (Sec) of IEEE test systems
Table 1 Comparison of unit commitment best cost ($) of IEEE test systems
Method Best Cost ($) of IEEE unit systems
4 unit 5 unit 6 unit 10 unit 20 unit 26 unit 40 unit 60 unit 80 unit 100 unit
A.SMP [25] 74812 nr nr 563937.26 1124490 nr nr nr nr nr
LRPSO [26] 74808 nr nr 566297 1128281 nr 2252330 3377718 4499347 5623607
GA [27] 74675 nr nr 565825 1126243 nr 2251911 3376625 4504933 5627437
BWOA [23] 74644.07 nr nr 565771 1126625 nr 2235113 nr 4490172 5614038
ILR [28] 75231 nr nr 563977 1123297 nr 2244237 3363491 4485633 5605678
BDE [29] nr nr nr 563997 112399 nr 2245700 3367066 4489022 5609341
PSO-GWO [30] nr 12281 13600 565210 nr nr nr nr nr nr
IDA-PSO [24] nr 11830 13292.28 565807 nr 741587.7 nr nr 4498479 5623885
EP [32] nr Nr nr 564551 1126494 nr 2249093 3371611 4526022 5657277
BPSO [32] nr nr nr 563977 1128192 nr 2243210 nr 4498943 5630838
SDP [33] nr nr nr 563777 1122622 nr 2242178 3363491 4485633 5605189
IBPSO [34] nr nr nr 563977 1125216 nr 2248581 3367865 4491083 5610293
ESA [35] nr nr nr 565828 1126254 nr 2250063 nr 4489022 5,609,341
BFWA [36] nr nr nr 563977 1124858 nr 2248228 3367445 4491284 5610954
BGWO1 [37] 73933.1 nr nr 563976.64 1125546 nr 2252475 3368934 4483381 5604146
BGWO2 [37] 73933.1 nr nr 563937.31 1123297 nr 2244701 3362515 4498479 5623885
BGOA 74370.32 11648.49 12945 563026 1120470 709354 2240277 3356574 4475407 5596414
(proposed)
Parameters Number of iterations= 30 Number of populations= 30 Intensity of attraction f= 1.5
Attractive length scale l= 0.5 Repulsion interval= [0 2.079]

Method Elapsed time (sec) of IEEE unit systems

4 unit 5 unit 6 unit 10 unit 20 unit 26 unit 40 unit 60 unit 80 unit 100 unit
A.SMP [25] nr nr nr nr nr nr nr nr nr nr
LRPSO [26] nr nr nr 45 96 nr 218 384 595 856
GA [27] nr nr nr 221 733 nr 2697 5840 10036 15733
BWOA [23] nr nr nr nr nr nr nr nr nr nr
ILR [28] nr nr nr 4.0 16 nr 52 113 209 345
BDE [29] nr nr nr nr nr nr nr nr nr nr
PSO-GWO [30] nr nr nr nr nr nr nr nr nr nr
IDA-PSO [24] nr nr nr nr nr nr nr nr nr nr
EP [31] nr nr nr 100 340 nr 1176 2267 3584 6120
BPSO [32] nr nr nr nr nr nr nr nr nr nr
SDP [33] nr nr nr 25.41 63.94 nr 157.73 260.76 353.84 392.56
IBPSO [34] nr nr nr 27 55 nr 110 172 235 295
ESA [35] nr nr nr nr nr nr nr nr nr nr
BFWA [36] nr nr nr 65.42 106.03 nr 238.02 422.29 676.53 1043.47
BGWO1 [37] nr nr nr 64.19 80.47 nr 169.24 281.2 473.4 836.5
BGWO2 [37] nr nr nr 66.15 87.533 nr 153.5 268.2 469.6 822.23
BGOA 9.98 17.22 20.32 30.73 57.55 86.43 116.93 201.70 246.98 337.04
(proposed)
*
nr= not reported by the corresponding author

1 6.2. Performance of BGOA for medium test systems

2 20, 26, 40 and 60 unit systems are simulated and the results of 40 and 60 unit systems are shown in the Table
3 2 and 3, respectively. The simulation results show feasible commitment and optimal power allocation as well.
4 The comparison for different approaches is given in Table 5 for IEEE 20, 26, 40 and 60-unit system. The results
5 obtained and their comparison shows the best quality of the purposed algorithm to tackle with the binary
6 optimization problem.

7 6.3. Performance of BGOA for large test systems

8 For the 80 and 100-unit system, the optimal cost and convergence quality results show as the number of units
9 increasing the ability of BGOA to obtain the best optimal solution enhances as shown in the Tables 4 and 6,
10 respectively. In order to observe the proposed algorithm IEEE 118-bus system is also observed with better
11 results as compared to other techniques. Total cost obtained and execution time by the proposed methodology

13
 AUTHOR and AUTHOR and AUTHOR/Turk J Elec Eng & Comp Sci

Table 6. Unit commitment of IEEE 100 unit system by BGOA.

Table 3 unit commitment of IEEE 100 unit systems by BGOA

Hour Unit commitment of 1-100 units by BGOA Fuel Cost

($)
1 1100000000110000000011000000001100000000110000000011000000001100000000100000000010000000001000000000 1.34e+05
2 1100000000110000000011000000001100000000110000000011000000001100000000110000000010000000001000000000 1.4367e+05
3 1101000000110000000011000000001100000000110000000011000000001100000000110000000011000000001100000000 1.636e+05
4 1101000000110000000011110000001101000000110100000011010000001101000000110100000011010000001101000000 1.8641e+05
5 1111000000111100000011110000001111000000110100000011010000001101000000110100000011010000001101000000 1.9759e+05
6 1111100000111110000011110000001111000000111100000011110000001111000000111100000011110000001111000000 2.1981e+05
7 1111100000111110000011111000001111100000111110000011110000001111000000111100000011110000001111000000 2.3008e+05
8 1111100000111111000011111000001111100000111110000011111000001111100000111110000011110000001111000000 2.4103e+05
9 1111110000111111000011111100001111110000111111000011111100001111110000111111000011111100001111110000 2.6589e+05
10 1111111100111111110011111111001111111100111111100011111110001111111000111111100011111110001111111000 2.9642e+05
11 1111111110111111111011111111101111111110111111110011111111001111111100111111110011111111001111111100 3.1493e+05
12 1111111111111111111111111111111111111111111111111011111111101111111110111111111011111111101111111110 3.3479e+05
13 1111111100111111110011111111001111111100111111100011111110001111111000111111100011111110001111111000 2.9642e+05
14 1111110000111111000011111100001111110000111111000011111100001111110000111111000011111100001111110000 2.6589e+05
15 1111100000111110000011111000001111100000111110000011111000001111100000111110000011111100001111100000 2.4192e+05
16 1111100000111110000011111000001111100000111110000011111000001111100000111110000011111100001111100000 2.1561e+05
17 1111100000111110000011111000001111100000111110000011111000001111100000111110000011111100001111100000 2.0689e+05
18 1111100000111110000011111000001111100000111110000011111000001111100000111110000011111100001111100000 2.2434e+05
19 1111100000111110000011111000001111100000111110000011111000001111100000111110000011111100001111100000 2.4192e+05
20 1111111100111111110011111111001111111100111111100011111110001111111000111111100011111110001111111000 2.9642e+05
21 1111111000111111100011111110001111011000111101100011111110001111011000111101100011111110001111011000 2.7061e+05
22 1101011000110101100011011110001101011000110101100011011110001101011000110101100011000010001100011000 2.2541e+05
23 1101001000110001000011001100001100010000110000000011001000001100000000110000000011000000001100000000 1.7551e+05
24 1100001000110001000011001100001100000000110000000011001000001100000000110000000010000000001000000000 1.5511e+05
Total cost ($) = 5544304.094150+52110.00= 5596414.094150
Elapsed time 337.043674 seconds

1 is drastically less than many other algorithms found in the literature. The ability of BGOA approach to reduce
2 total cost with less execution time goes on increasing from small to large IEEE unit system, which makes this
3 approach to have robustness in finding results, most efficient searching ability and real-time practical based UC
4 problem solving ability.

5 7. Conclusion
6 This study presents binary grasshopper optimization algorithm (BGOA) models to obtain an optimal solution
7 of UC optimization problem. The objective function of the UC optimization problem is formulated as the
8 cost function under the load balance, generation limit, spinning reserve ( 10% and 5% ), minimum up time
9 & minimum down time and de commitment constraints. Sigmoid function is used for binary mapping. The
10 simulation results of the small, medium and large unit systems show the superiority and searching efficiency
11 of the proposed algorithm as compared to the other modern techniques. This approach can also be used to
12 tackle profit based unit commitment and multi objective such as reliability maximization, emission reduction
13 and many other constraints optimization problem. The proposed algorithm also has ability to better results
14 of UC optimization problem with the integration of renewable energy resources, security, ramp rate and many
15 other constraints.

16 References

17 [1] C.-C. Su and Y.-Y. Hsu, ”Fuzzy dynamic programming: an application to unit commitment,” IEEE transactions
18 on power systems, vol. 6, no. 3, pp. 1231-1237, 1991.
19 [2] W. Ongsakul and N. Petcharaks, ”Unit commitment by enhanced adaptive Lagrangian relaxation,” IEEE Transac-
20 tions on Power Systems, vol. 19, no. 1, pp. 620-628, 2004.

14
 AUTHOR and AUTHOR and AUTHOR/Turk J Elec Eng & Comp Sci

1 [3] B. Venkatesh, T. Jamtsho, and H. Gooi, ”Unit commitment-a fuzzy mixed integer linear programming solution,”
2 IET Generation, Transmission & Distribution, vol. 1, no. 5, pp. 836-846, 2007.
3 [4] T. Senjyu, K. Shimabukuro, K. Uezato, and T. Funabashi, ”A fast technique for unit commitment problem by
4 extended priority list,” IEEE Transactions on Power Systems, vol. 18, no. 2, pp. 882-888, 2003.
5 [5] S. A. Kazarlis, A. Bakirtzis, and V. Petridis, ”A genetic algorithm solution to the unit commitment problem,” IEEE
6 transactions on power systems, vol. 11, no. 1, pp. 83-92, 1996.
7 [6] K. Vaisakh and L. Srinivas, ”Evolving ant colony optimization based unit commitment,” Applied Soft Computing,
8 vol. 11, no. 2, pp. 2863-2870, 2011.
9 [7] A. Saber, T. Senjyu, A. Yona, and T. Funabashi, ”Unit commitment computation by fuzzy adaptive particle swarm
10 optimisation,” IET Generation, Transmission & Distribution, vol. 1, no. 3, pp. 456-465, 2007.
11 [8] L. K. Panwar, S. Reddy, A. Verma, B. K. Panigrahi, and R. Kumar, ”Binary grey wolf optimizer for large scale
12 unit commitment problem,” Swarm and Evolutionary Computation, vol. 38, pp. 251-266, 2018.
13 [9] J. Zhao, S. Liu, M. Zhou, X. Guo, and L. Qi, ”An Improved Binary Cuckoo Search Algorithm for Solving Unit
14 Commitment Problems: Methodological Description,” IEEE Access, vol. 6, pp. 43535-43545, 2018.
15 [10] B. Koodalsamy, M. B. Veerayan, C. Koodalsamy, S. P. J. T. J. o. E. E. Simon, and C. Sciences, ”Firefly algorithm
16 with multiple workers for the power system unit commitment problem,” vol. 24, no. 6, pp. 4773-4789, 2016.
17 [11] A. Ş. UYAR, B. J. T. J. o. E. E. Tuerkay, and C. Sciences, ”Evolutionary algorithms for the unit commitment
18 problem,” vol. 16, no. 3, pp. 239-255, 2008.
19 [12] J. Zhao, S. Liu, M. Zhou, X. Guo, and L. Qi, ”An Improved Binary Cuckoo Search Algorithm for Solving Unit
20 Commitment Problems: Methodological Description,” IEEE Access, vol. 6, pp. 43535-43545, 2018.
21 [13] H. H. Balci and J. F. Valenzuela, ”Scheduling electric power generators using particle swarm optimization combined
22 with the Lagrangian relaxation method,” International Journal of Applied Mathematics and Computer Science, vol.
23 14, pp. 411-421, 2004.
24 [14] C.-P. Cheng, C.-W. Liu, and C.-C. Liu, ”Unit commitment by Lagrangian relaxation and genetic algorithms,” IEEE
25 transactions on power systems, vol. 15, no. 2, pp. 707-714, 2000.
26 [15] V. K. Kamboj, ”A novel hybrid PSO–GWO approach for unit commitment problem,” Neural Computing and
27 Applications, vol. 27, no. 6, pp. 1643-1655, 2016.
28 [16] A. Fathy, ”Recent meta-heuristic grasshopper optimization algorithm for optimal reconfiguration of partially shaded
29 PV array,” Solar Energy, vol. 171, pp. 638-651, 2018.
30 [17] D. K. Lal, A. K. Barisal, and M. Tripathy, ”Load frequency control of multi area interconnected microgrid power
31 system using grasshopper optimization algorithm optimized fuzzy PID controller,” in 2018 Recent Advances on
32 Engineering, Technology and Computational Sciences (RAETCS), 2018, pp. 1-6: IEEE.
33 [18] S. Hazra, T. Pal, and P. K. Roy, ”Renewable Energy Based Economic Emission Load Dispatch Using Grasshopper
34 Optimization Algorithm,” International Journal of Swarm Intelligence Research (IJSIR), vol. 10, no. 1, pp. 38-57,
35 2019.
36 [19] S. M. Ismael, S. H. A. Aleem, A. Y. Abdelaziz, and A. F. Zobaa, ”Optimal Conductor Selection of Radial Distribution
37 Feeders: An Overview and New Application Using Grasshopper Optimization Algorithm,” in Classical and Recent
38 Aspects of Power System Optimization: Elsevier, 2018, pp. 185-217.
39 [20] T. A. Jumani, M. W. Mustafa, M. M. Rasid, N. H. Mirjat, Z. H. Leghari, and M. S. Saeed, ”Optimal Voltage and
40 Frequency Control of an Islanded Microgrid using Grasshopper Optimization Algorithm,” Energies, vol. 11, no. 11,
41 p. 3191, 2018.
42 [21] M. Barman, N. D. Choudhury, and S. Sutradhar, ”A regional hybrid GOA-SVM model based on similar day
43 approach for short-term load forecasting in Assam, India,” Energy, vol. 145, pp. 710-720, 2018.
44 [22] S. Saremi, S. Mirjalili, and A. Lewis, ”Grasshopper optimisation algorithm: theory and application,” Advances in
45 Engineering Software, vol. 105, pp. 30-47, 2017.

15
 AUTHOR and AUTHOR and AUTHOR/Turk J Elec Eng & Comp Sci

1 [23] V. Kumar, D. J. N. C. Kumar, and Applications, ”Binary whale optimization algorithm and its application to unit
2 commitment problem,” pp. 1-29, 2018.
3 [24] S. Khunkitti, N. R Watson, R. Chatthaworn, S. Premrudeepreechacharn, and A. J. E. Siritaratiwat, ”An Improved
4 DA-PSO Optimization Approach for Unit Commitment Problem,” vol. 12, no. 12, p. 2335, 2019.
5 [25] S. Khanmohammadi, M. Amiri, and M. T. J. E. Haque, ”A new three-stage method for solving unit commitment
6 problem,” vol. 35, no. 7, pp. 3072-3080, 2010.
7 [26] H. H. Balci, J. F. J. I. J. o. A. M. Valenzuela, and C. Science, ”Scheduling electric power generators using particle
8 swarm optimization combined with the Lagrangian relaxation method,” vol. 14, pp. 411-421, 2004.
9 [27] S. A. Kazarlis, A. Bakirtzis, and V. J. I. t. o. p. s. Petridis, ”A genetic algorithm solution to the unit commitment
10 problem,” vol. 11, no. 1, pp. 83-92, 1996.
11 [28] W. Ongsakul and N. J. I. T. o. P. S. Petcharaks, ”Unit commitment by enhanced adaptive Lagrangian relaxation,”
12 vol. 19, no. 1, pp. 620-628, 2004.
13 [29] Y.-W. Jeong, W.-N. Lee, H.-H. Kim, J.-B. Park, J.-R. J. J. o. E. E. Shin, and Technology, ”Thermal unit
14 commitment using binary differential evolution,” vol. 4, no. 3, pp. 323-329, 2009.
15 [30] V. K. J. N. C. Kamboj and Applications, ”A novel hybrid PSO–GWO approach for unit commitment problem,”
16 vol. 27, no. 6, pp. 1643-1655, 2016.
17 [31] K. Juste, H. Kita, E. Tanaka, and J. J. I. T. o. P. S. Hasegawa, ”An evolutionary programming solution to the unit
18 commitment problem,” vol. 14, no. 4, pp. 1452-1459, 1999.
19 [32] Y.-W. Jeong, J.-B. Park, S.-H. Jang, and K. Y. Lee, ”A new quantum-inspired binary PSO for thermal unit
20 commitment problems,” in 2009 15th International Conference on Intelligent System Applications to Power Systems,
21 2009, pp. 1-6: IEEE.
22 [33] R. J. I. J. o. E. P. Jabr and E. Systems, ”Rank-constrained semidefinite program for unit commitment,” vol. 47,
23 pp. 13-20, 2013.
24 [34] X. Yuan, H. Nie, A. Su, L. Wang, and Y. J. E. S. w. a. Yuan, ”An improved binary particle swarm optimization
25 for unit commitment problem,” vol. 36, no. 4, pp. 8049-8055, 2009.
26 [35] D. N. Simopoulos, S. D. Kavatza, and C. D. J. I. T. o. P. S. Vournas, ”Unit commitment by an enhanced simulated
27 annealing algorithm,” vol. 21, no. 1, pp. 68-76, 2006.
28 [36] L. K. Panwar, S. Reddy, and R. J. I. J. o. S. I. R. Kumar, ”Binary fireworks algorithm based thermal unit
29 commitment,” vol. 6, no. 2, pp. 87-101, 2015.
30 [37] L. K. Panwar, S. Reddy, A. Verma, B. K. Panigrahi, R. J. S. Kumar, and E. Computation, ”Binary grey wolf
31 optimizer for large scale unit commitment problem,” vol. 38, pp. 251-266, 2018.

16