International Journal of Power Electronics and Drive Systems (IJPEDS)

Vol. 14, No. 3, September 2023, pp. 1885~1893

ISSN: 2088-8694, DOI: 10.11591/ijpeds.v14.i3.pp1885-1893  1885

Artificial intelligence of optimal real power dispatch with

constraints of lines overloading

Abdesselam Abderrahmani, Nasri Abdelfatah, Gasbaoui Brahim

Department of Electrical Engineering, Faculty of Technologies, University Tahri Mohammed, Bechar, Algeria
Smart Grids and Renewable Energies Laboratory, University Tahri Mohammed, Bechar, Algeria

Article Info ABSTRACT

Article history: In the literature on optimal power flow (OPF), it has been shown that the
suggested ways offer a higher degree of satisfaction in optimizing overall
Received Mar 8, 2023 production costs while fulfilling power flow equations, system security, and
Revised Apr 25, 2023 equipment operational constraints. Despite this, the overloaded of the
Accepted May 5, 2023 transmission lines are taken as a performance index but not a primary
constraint. This article presents an improved approach to artificial intelligence
algorithms of optimal real power dispatch with the security of lines; the main
Keywords: difference concerning our point seen relies on the additional penalization of
the choices, which does not respect this constraint. The problem is
Algorithm hybrid implemented in the IEEE 14-bus system with "5" generator units. The results
Economic dispatch of the simulations of the metaheuristic algorithms without/with constraint
Meta-heuristics algorithms (overloaded lines) were compared. Furthermore, this article suggests
Optimal power flow hybridizing ant colony optimization (ACO) and genetic algorithm (GA) as a
Security constrained means to enhance the optimization performance of these algorithms. This
hybridization involves using ACO to generate a set of initial solutions, which
are then refined using GA. The compound results obtained by the ant system-
genetic algorithm hybrid (H-ASGA) for the problem of overloaded lines
validated its potential.
This is an open access article under the CC BY-SA license.

Corresponding Author:
Abdesselam Abderrahmani
Department of Electrical Engineering, Faculty of Technologies, University Tahri Mohammed
The Independence Street, PB 417 Bechar, Algeria
Email: [email protected], [email protected]

1. INTRODUCTION
The economic dispatch problem plays an important role in power system operation. Their
development during the last two decades has tracked on the one hand progress in advanced computer
technology and the other in artificial intelligence optimization technology. The capacity of OPF to quickly
identify the best solution while taking system security into account is the main area of attention at the moment.
In this context, we discuss a new formulation of transmission line safety constraints that are assessed as
penalties. Five optimization approaches are proposed to optimize the model, particle swarms (PSO), firefly
algorithm (FA), artificial bee colony (ABC), genetic algorithm (GA) and ant colony optimization (ACO) [1].
Numerous comparisons are made on a reference test system to show the effectiveness of the proposed model
and the five approaches metaheuristic.
The goal of the OPF is to identify the best possible mix of power generation that satisfies the overall
demand and power system restrictions while minimising total fuel costs. The operating cost function of power
generation units, it is expressed as (1) [2]–[4].
2
𝐹𝑇 = ∑𝑁𝐺 𝑁𝐺
𝑖=1 𝑓(𝑃𝐺𝑖 ) = ∑𝑖=1(𝑎𝑖 𝑃𝐺𝑖 + 𝑏𝑖 𝑃𝐺𝑖 + 𝑐𝑖 ) (1)

Journal homepage: http://ijpeds.iaescore.com

1886  ISSN: 2088-8694

Where 𝑓(𝑃𝐺𝑖 ) : Generating unit operation cost of a generator connected to bus i ($/h); 𝑃𝐺𝑖 : Real power
generation of a generator connected to bus i (MW); and 𝑎𝑖 , 𝑏𝑖 , 𝑐𝑖 : are the generator's cost-of-fuel coefficients

2. SECURITY CONSTRAINED ECONOMIC DISPATCH BY ARTIFICIAL INTELLIGENCE

The goal of security constrained economic power distribution (SCED) is to reduce the fuel
consumption of the generators or the overall operating costs of the system by determining the power output of
each generating unit under the constraints of the system load demands [5], By observing the input-output
characteristics of the generating unit, it can be seen that the power output is constrained by the minimal and
maximal capacitie of the generating unit.
The outputs of the "N-1" generators may initialize freely within bounds when using metaheuristic
methods, while the power balancing uses the Gauss-Seidel algorithm to adjust the output of the "reference
generator" or extended bus generator. New points in the search space are generated by techniques known in
the artificial intelligence (AI) literature. The AI type algorithms are in fact optimization without constraint; all
information and constraint must be expressed in a fitness functions [6].

2.1. Constraint 1: Real power flow equation

The essential constraint in the operation of an electrical system is that the sum of the output powers
must equal the load demand. The class of models that can be estimated using pool estimation, can be written
as (2) and (3) [7]–[9], where NB: Total number of buses; NG: The number of generating units; NL: The number
of transmission lines; and ℎ1 : Penalty factors of loss.

∑𝑁𝐺 𝑁𝐵 𝑁𝐿
𝑖=1 𝑃𝐺𝑖 = ∑𝑖=1 𝑃𝐿𝑜𝑎𝑑𝑖 + ∑𝑖=1 𝑃𝐿𝑜𝑠𝑠𝑖 (2)

𝐹𝑙𝑜𝑠𝑠 = ℎ1 (∑𝑁𝐺 𝑁𝐵
𝑖=1 𝑃𝐺𝑖 − ∑𝑖=1 𝑃𝐿𝑜𝑎𝑑𝑖 )
2
(3)

2.2. Constraint 2: Upper and lower thresholds for the generators' active power output
The active power generated must satisfy the maximum and minimum operating limits, that is, each
unit's power output must be higher than or equal to the minimum power allowed and lower than or equal to
the maximum power allowed for that specific unit [10]–[12].
𝐼𝐹 𝑃𝐺𝑖 < 𝑃𝐺𝑖𝑚𝑖𝑛 ⟹ 𝑃𝐺𝑖 = 𝑃𝐺𝑖𝑚𝑖𝑛
{ (4)
𝐼𝐹 𝑃𝐺𝑖 > 𝑃𝐺𝑖𝑚𝑎𝑥 ⟹ 𝑃𝐺𝑖 = 𝑃𝐺𝑖𝑚𝑎𝑥

The stresses of the node voltage limits can be expressed by:

0 𝑃𝐺𝑖𝑚𝑖𝑛 < 𝑃𝐺𝑖 < 𝑃𝐺𝑖𝑚𝑎𝑥 𝑛𝑜 𝑣𝑖𝑜𝑙𝑎𝑡𝑖𝑜𝑛
𝑁𝐺𝑚𝑖𝑛 2
𝐹𝑃𝐺 = { ℎ2 ∑𝑘=1 (𝑃𝐺𝑖 (𝑘) − 𝑃𝐺𝑖𝑚𝑖𝑛 (𝑘)) 𝑃𝐺𝑖 < 𝑃𝐺𝑖𝑚𝑖𝑛 (5)
2
ℎ3 ∑𝑁𝐺𝑚𝑎𝑥
𝐿=1 (𝑃𝐺𝑖 (𝐿) − 𝑃𝐺𝑖𝑚𝑎𝑥 (𝐿)) 𝑃𝐺𝑖 > 𝑃𝐺𝑖𝑚𝑎𝑥

Where 𝑁𝐺𝑚𝑖𝑛: The set of generators violating minimum production limits; 𝑁𝐺𝑚𝑎𝑥 : The set of generators
violating maximum production limits; ℎ2 : Penalty factors to the violation of minimum power output; and ℎ3 :
Penalty factors to the violation of maximum power output.

2.3. Constraint 3: Upper and lower limits of node voltages

In general, functional constraints are more often soft limits than hard limitations in a strictly
mathematical sense [5]. The following are possible expressions for the inequality restrictions on control
parameters, such as the generator bus voltage limits [13].
2
ℎ4 ∑𝑁𝑉𝑚𝑖𝑛
𝑘 (𝑉𝑖 (𝑘) − 𝑉𝑖𝑚𝑖𝑛 (𝑘)) 𝑉𝑖 < 𝑉𝑖𝑚𝑖𝑛
𝐹𝑉 = { 2 (6)
ℎ5 ∑𝑁𝑉𝑚𝑎𝑥
𝐿 (𝑉𝑖 (𝐿) − 𝑉𝑖𝑚𝑎𝑥 (𝐿)) 𝑉𝑖 > 𝑉𝑖𝑚𝑎𝑥

Where 𝑁𝑉𝑚𝑖𝑛: The set of generators violating minimum voltage limits; 𝑁𝑉𝑚𝑎𝑥 : The set of generators
violating maximum voltage limits; ℎ4 : Penalty factors to the violation of minimum voltage; ℎ5 : Penalty factors
to the violation of maximum voltage

2.4. Constraint 4: Available transfer capacity of the transmission lines

If a single contingency (or numerous contingencies) takes place, there is no assurance that regular
operation won't violate the line limits. The generations must be redistributed if such a circumstance occurs in

Int J Pow Elec & Dri Syst, Vol. 14, No. 3, September 2023: 1885-1893
Int J Pow Elec & Dri Syst ISSN: 2088-8694  1887

order to satisfy the line limitations. There is consequently a need for an effective method of integrating the
security limitations into ED [14].
In practical electrical power supply systems, it is very important to solve the OPF problem with lines
overload constraints. The power balance constraint is included in the formulation and the incremental losses
are accurately represented at the classic economic dispatch if power flow is solved concurrently with generation
and cost minimization. As a result, the security of the lines can be included to be checked periodically during
the optimum seeking process to ensure that the dispatch solution is within the operating limits [15].

0 |𝑃𝑖𝑗 | ≤ 𝑃𝑖𝑗−𝑚𝑎𝑥 𝑛𝑜 𝑣𝑖𝑜𝑙𝑎𝑡𝑖𝑜𝑛

𝐹𝐿𝑂 = { 𝑃𝑖𝑗 (𝑘)−𝑃𝑖𝑗−𝑚𝑎𝑥 (𝑘) 𝑘 ∈ 𝑁𝑂𝐿 (7)
∑𝑁𝑂𝐿
𝑘=1 |𝑃𝑖𝑗 | > 𝑃𝑖𝑗−𝑚𝑎𝑥
𝑃 (𝑘)
𝑖𝑗−𝑚𝑎𝑥

Where NOL: The set of overloaded lines; 𝑃𝑖𝑗 : The power flow at the line from bus "i" to bus "j"; 𝑃𝑖𝑗−𝑚𝑎𝑥 :
The power limit flow at the line from bus "i" to bus "j".
Given everything, the stochastic model's corresponding transformation is formally expressed by a new
objective function as [16]:
2
𝐹𝑇 = ∑𝑁𝐺
𝑖=1(𝑎𝑖 𝑃𝐺𝑖 + 𝑏𝑖 𝑃𝐺𝑖 + 𝑐𝑖 ) + 𝐹𝑙𝑜𝑠𝑠 + 𝐹𝑃𝐺 + 𝐹𝑉 + 𝐹𝑂𝐿 (8)

Since the issue in (8) is a high-dimension, non-convex optimisation problem, finding an analytical solution is
very challenging. Numerous artificial intelligence (AI) techniques have developed recently that make it easier
to solve optimisation issues that were previously challenging or impossible. The main benefit of AI approaches
is that they are not constrained by restricted search space hypotheses like continuity, the presence of an
objective function derivative, etc. These techniques are comparable in certain ways [17].

3. IMPLEMENTATION
The optimization method with "IA" can be summarized in the following steps: The optimization
method with "IA" can be summarized in the following steps:
Step 1: Pick the initial control variable,
Step 2: To find a workable solution that complies with the equality condition for the power balance, solve the
power flow issue. {PG(i)} i∈NG,
Step 3: load flow calculation to determine: production bus reference, load, loss, line overload,
Step 4: Evaluate fitness function,
Step 5: Obtain the power flow solution with updated control variables,
Step 6: Examine the convergence. In this case, the solution converges if ΔPGi are lower than the user-defined
tolerance. Otherwise
A well-known benchmark system for power system analysis and optimisation algorithm testing is the
IEEE 14-bus network. The literature presents the cost coefficients, maximum and lowest generation limitations,
and load demand for each interval of the IEEE 14-bus [5], [18]. The optimization strategies of GA, FA,ABC,
ACO and PSO with this new formulation for solution of the economic dispatch load problem were introduced
in Table 1. It is evident from the comparison that the suggested technique offers an improvement in the overall
cost reduction.

Table 1. Results of economic dispatch without lines overload constraint

IEEE 14 Bus Particle warm Sunflower Gravitational
GA FA ABC PSO ACO
System [19] [20] search [21]
PG1 (MW) 172.2998 194.6693 210.3368 167.3213 * 172.3926 * 171.6565 * 171.6277 * 171.6196 *
PG2 (MW) 47.3037 36.7904 20.2300 49.8893 47.2066 47.5756 47.5768 47.7001
PG3 (MW) 20.8993 27.9933 17.3500 20.3736 20.818 20.9549 20.9603 20.9795
PG4 (MW) 15.0895 0.0000 10.0400 17.2207 15.9887 15.8 15.8289 16.0389
PG5 (MW) 11.6331 8.8547 11.5000 12.1845 11.0563 11.4277 11.4159 11.0825
Total generation 267.2254 268.3077 269.2221 266.9894 267.4621 267.4147 267.4096 267.4207
Real load (MW) 259.0000 259.0000 259.0000 259.0000 259.0000 259.0000 259.0000 259.0000
Real loss (MW) 8.2254 9.3077 10.2221 1 7.9894 8.4621 8.4147 8.4096 8.4207
Fuel cost ($/hr) 715.2875 725.0132 ** 735.7454 ** 715.5140 2 716.0666 ** 716.0568 ** 716.0454 ** 716.0604 **
Constraint 1,2,3 Satisfied Satisfied Satisfied Satisfied Satisfied Satisfied Satisfied Satisfied
L2-3{58.57} L1-2{130.63}L1-2{142.63} L2-3{58.60} L2-3{57.87} L2-3{57.76} L2-3{57.76} L2-3{57.77}
Constraint 4 L4-5{52.48} L2-3{56.51} L1-5{65.69} L4-5{52.33} L4-5{49.61} L4-5{49.39} L4-5{49.39} L4-5{49.51}
overloaded lines L4-5{51.48} L2-3{60.54}
L4-5{55.79}
*The power at the reference bus is calculated by the Gauss-Seidel algorithm
**The costs are calculated by the coefficients in the appendix

Artificial intelligence of optimal real power dispatch with constraints … (Abderrahmani Abdesselam)
1888  ISSN: 2088-8694

Figure 1 shows the topology of the IEEE 14-bus test system's network. The results shown are
from [20]. The lines that are overloaded are in red. In the context of the practical application of the OPF, the
optimal load flow solution presents an unavoidable limitation, of capacity to transit the powers of the lines;
owever, these solutions result in a violation of lines {1-2, 2-3 and 4-5} limits. The distribution of powers is of
course only a constituent, but very necessary, for the operation and planning of networks. Therefore, the result
obtained does not meet the security constraints of the system.

Figure 1. The 14-bus system with lines overload of ref [20]

4. OPTIMIZATION OF ECONOMIC DISTRIBUTION UNDER CONSTRAINT OF LINE

OVERLOAD
The five meta-heuristics algorithms used are PSO, FA, GA, ACO and ABC for solving the security
ED problem is tested on the IEEE 14-bus system with constraints of lines overload. The total system load is
259.00 MW. The corresponding load-scaling factor (LSF) is 1.0. Multiple iterations of the tests have been run
to ensure the stability of the meta-heuristics optimisation strategies. Table 2 shows the best solutions to the
objective function of the 14-bus problem taking into account lines overload. The elimination of line overload
constraints has an impact on the primary objective function, this costs an increase in fuel cost from 9.83 to
9.9% depending on the method.

Table 2. Results of economic dispatch under constraint of line overload (LSF=1).

IEEE 14 Bus System GA FA ABC PSO ACO
PG1 (MW) 123.0996 122.0295 123.6483 123.3977 122.8166
PG2 (MW) 29.3979 27.9233 26.3671 29.1437 31.5821
PG3 (MW) 50.2217 49.2821 49.1659 50.0694 50.9488
PG4 (MW) 40.0848 40.5972 38.8869 37.5642 38.7391
PG5 (MW) 20.7895 23.6637 25.4379 23.3765 19.5339
Total generation 263.5936 263.4957 263.5062 263.5515 263.6206
Real load (MW) 259.0000 259.0000 259.0000 259.0000 259.0000
Real loss (MW) 4.5936 4.4957 4.5062 4.5515 4.6206
Fuel cost ($/hr) 794.3059 794.1652 794.6702 794.1583 794.6661
Constraint 1,2,3 Satisfied Satisfied Satisfied Satisfied Satisfied
Constraint 4 Overloaded lines Satisfied Satisfied Satisfied Satisfied Satisfied

Int J Pow Elec & Dri Syst, Vol. 14, No. 3, September 2023: 1885-1893
Int J Pow Elec & Dri Syst ISSN: 2088-8694  1889

Figure 2 show a comparison between the convergences of the objective function using the five meta-
heuristics methods. After 100 iterations, the algorithm that yielded the lowest cost function value was the 'PSO'
algorithm, with a cost function value of 794.1583$/h. The FA algorithm had the second-lowest cost function value
of 794.1652 $/h, followed by GA with a value of 794.3059 $/h, and ACO with a value of 794.6661 $/h, The ACO
algorithm also seems to converge to a certain value after the first few iterations. However, the values are much
smaller than those of the ABC algorithm. The ACO algorithm seems to have a slower convergence rate than the
ABC algorithm, but it eventually reaches a stable value after a larger number of iterations. The first value of the
ABC algorithm is quite large, which indicates that the algorithm starts with a high value of the objective function.
However, the values decrease significantly after the first few iterations and then remain stable, fluctuating only
slightly around a certain value. We also see that the objective function converged fast, smoothly and stably,
providing a better optimization performance to achieve the best solution with the PSO algorithm.

Figure 2. Convergence characteristics of 14-bus (5) generator system.

When comparing the various meta-heuristic methods used to solve ED problems, the algorithms are
able to achieve similar results. However, there may be slight differences in precision between the different
methods. From an engineering perspective, these small variations can be deemed acceptable and can be
overlooked. Finally, in Table 3 the IA techniques, we show the lines overload are annulled in full.
It is important to note that the differences between the cost function values of the algorithms are
relatively small. Overall, it is difficult to draw definitive conclusions about the performance of each
algorithm [22], especially with the change in the parameters used for each algorithm. It may be helpful to run
each algorithm multiple times with different parameters and a load scale factor of 1.2 to obtain a more robust
comparison of their performance. The calculation results of ED with lines overload for the IEEE 14-bus system
are shown in Table 4.

Table 3. Results of classic economic dispatch with lines overload

Line Line
Lines GA FA ABC PSO ACO Lines GA FA ABC PSO ACO
Limit Limit
1⟶2* 0.8339 0.8307 0.8438 0.8357 0.8269 1.200 6⟶11 0.1286 0.1261 0.1207 0.1206 0.1274 0.1800
1⟶5 0.3971 0.3896 0.3927 0.3983 0.4013 0.6500 6⟶12 0.0863 0.086 0.0853 0.0853 0.0862 0.3200
2⟶3* 0.36 0.3601 0.3597 0.36 0.36 0.3600 6⟶13 0.2068 0.2055 0.2027 0.2027 0.2062 0.3200
2⟶4 0.3165 0.3069 0.3049 0.3149 0.324 0.6500 8⟶7 0.2079 0.2366 0.2544 0.2338 0.1953 0.3200
2⟶5 0.2224 0.2141 0.2136 0.2232 0.2299 0.5000 7⟶9 0.2723 0.2817 0.2914 0.2865 0.2704 0.3200
4⟶3 0.0863 0.0958 0.0974 0.0879 0.079 0.6500 9⟶10 0.0008 0.0016 0.0068 0.0069 0.0003 0.3200
5⟶4* 0.3998 0.3937 0.3879 0.3896 0.3996 0.4500 9⟶14 0.058 0.0595 0.0628 0.0629 0.0587 0.3200
4⟶7 0.063 0.0441 0.0362 0.0516 0.0735 0.5500 11⟶10 0.0917 0.0893 0.084 0.0839 0.0906 0.1200
4⟶9 0.0774 0.0721 0.071 0.0759 0.0809 0.3200 12⟶13 0.0244 0.0241 0.0234 0.0234 0.0243 0.1200
5⟶6 0.1238 0.1152 0.1229 0.1351 0.1346 0.4500 13⟶14 0.093 0.0915 0.0881 0.0881 0.0924 0.1200
*Solved problem of lines overload

Overall, FA outperforms the other algorithms in terms of minimizing the objective function for the
14-bus IEEE system after 100 iterations. However, it is important to note that the performance of these
algorithms can be affected by the initial parameter values and stopping criteria used, so the results may vary

Artificial intelligence of optimal real power dispatch with constraints … (Abderrahmani Abdesselam)
1890  ISSN: 2088-8694

depending on these factors. In Table 5 of the simulation results, all of the line overloads for the five
metaheuristic algorithms have been eliminated.

Table 4. Results of economic dispatch with lines overload (LSF = 1.2).

IEEE 14 Bus System GA FA ABC PSO ACO
PG1 (MW) 136.37 128.46 155.286 130.92 117.3557
PG2 (MW) 44.50 35.62 42.3297 35.87 47.6234
PG3 (MW) 77.49 72.04 83.2422 72.82 74.0549
PG4 (MW) 41.08 54.94 26.9137 52.49 53.618
PG5 (MW) 17.27 25.08 10 24.14 23.2172
Total generation 316.712 316.1397 317.7717 316.2463 315.8692
Real load (MW) 310.8 310.8 310.8 310.8 310.8
Real loss (MW) 5.912 5.3397 6.9717 5.4463 5.0692
Fuel cost ($/hr) 1114.66341094.44051148.76441096.26051107.5641
Constraint 1,2,3 Satisfied Satisfied Satisfied Satisfied Satisfied
Constraint 4 Overloaded linesSatisfied Satisfied Satisfied Satisfied Satisfied

Table 5. Results of classic economic dispatch with lines overload

Line Line
Lines GA FA ABC PSO ACO Lines GA FA ABC PSO ACO
Limit Limit
1⟶2 0.8904 0.8647 1.0089 0.8788 0.7672 1.200 6⟶11 0.1469 0.1642 0.1288 0.1606 0.1634 0.1800
1⟶5 0.4732 0.4200 0.5440 0.4304 0.4064 0.6500 6⟶12 0.1024 0.1045 0.1003 0.1041 0.1044 0.3200
2⟶3 0.3601 0.3600 0.3552 0.3600 0.3563 0.3600 6⟶13 0.2449 0.2538 0.2357 0.2519 0.2534 0.3200
2⟶4 0.4030 0.3468 0.4519 0.3548 0.3600 0.6500 8⟶7 0.1727 0.2508 0.1 0.2414 0.2322 0.3200
2⟶5 0.2983 0.2407 0.347 0.2489 0.2565 0.5000 7⟶9 0.3165 0.3180 0.3172 0.319 0.3143 0.3200
4⟶3 0.0015 0.0562 0.0513 0.0484 0.0396 0.6500 9⟶10 0.0065 0.0102 0.0242 0.0068 0.0095 0.3200
5⟶4 0.4463 0.4498 0.4485 0.449 0.4401 0.4500 9⟶14 0.0749 0.0644 0.0859 0.0665 0.0649 0.3200
4⟶7 0.1406 0.0658 0.2124 0.0762 0.0804 0.5500 11⟶10 0.1025 0.1195 0.0847 0.1160 0.1187 0.1200
4⟶9 0.1152 0.0874 0.1423 0.0915 0.0921 0.3200 12⟶13 0.0280 0.0300 0.0259 0.0296 0.0299 0.1200
5⟶6 0.2030 0.1001 0.3076 0.1175 0.1113 0.4500 13⟶14 0.1067 0.1173 0.0956 0.1152 0.1169 0.1200

5. HYBRIDIZATION OF ANT SYSTEM-GENETIC ALGORITHM [23], [24]

The ant system-genetic algorithm hybrid (H-ASGA) combines the advantages of two commonly used
optimization algorithms to improve the efficiency of searching for optimal solutions. This method uses ant
colonies to construct candidate solutions using meta-heuristics rules and pheromone trails left by other ants.
Then, the genetic algorithm is applied to the candidate solutions to perform selection, crossover, and mutation
operations to generate even better solutions. These new solutions are then provided to the ants to further explore
the search space. Thus, H-ASGA is a powerful and effective method for solving optimization problems by
combining the advantages of both algorithms. The steps involved in this algorithm are as follows [25], [26]:
- Step 1: Initialize the population: generate a set of random solutions for the genetic algorithm (GA). Set up
the initial pheromone values for the ant colony optimization (ACO) algorithm.
- Step 2: Evaluate fitness: utilise a fitness function to determine each population member's level of fitness.
f(PG), where PG(Real power generation) is the solution.
- Step 3:Update pheromones: update the pheromone values based on the fitness of the solutions found by
the ants using the following formula:

𝜏(𝑖, 𝑗) = (1 − 𝜌) 𝜏(𝑖, 𝑗) + 𝛥𝜏(𝑖, 𝑗) (9)

were
τ(i,j): is the pheromone value on edge (i,j)
ρ ∶ is the pheromone evaporation rate
Δτ(i,j) ∶ is the amount of pheromone deposited by an ant on edge (i,j)
- Step 4:Select parents: use selection techniques to choose the fittest individuals from the population to be
parents for the next generation. This can be done using the fitness proportional selection method, where the
chance of picking a person is related to how fit they are.
- Step 5: Apply genetic operators: apply genetic operators, such as crossover and mutation, to the selected
parents to generate new offspring for the next generation. This can be done using the following formulas:

𝑂𝑓𝑓𝑠𝑝𝑟𝑖𝑛𝑔1 = 𝑃𝑎𝑟𝑒𝑛𝑡1 + 𝛼 (𝑃𝑎𝑟𝑒𝑛𝑡2 − 𝑃𝑎𝑟𝑒𝑛𝑡1)

{ (10)
𝑂𝑓𝑓𝑠𝑝𝑟𝑖𝑛𝑔2 = 𝑃𝑎𝑟𝑒𝑛𝑡2 + 𝛼 (𝑃𝑎𝑟𝑒𝑛𝑡1 − 𝑃𝑎𝑟𝑒𝑛𝑡2)

Int J Pow Elec & Dri Syst, Vol. 14, No. 3, September 2023: 1885-1893
Int J Pow Elec & Dri Syst ISSN: 2088-8694  1891

where α is the crossover parameter.

Step 6: Apply local search: Apply a local search algorithm, such as hill climbing or simulated annealing,
to refine the solutions found by the GA. This can be done using the following formula:

𝑥′ = 𝑚𝑖𝑛 𝑓(𝑥 + 𝛿) (11)

were,
x' is the current solution, δ is a small perturbation,
“min” finds the solution with the smallest fitness.
- Step 7: Combine solutions: combine the solutions found by both the ACO and GA algorithms to create a
hybrid solution. This can be done using a weighting factor, where the hybrid solution is a linear combination
of the two solutions:

𝐻𝑦𝑏𝑟𝑖𝑑 𝑠𝑜𝑙𝑢𝑡𝑖𝑜𝑛 = 𝜔 ∗ 𝐴𝐶𝑂 𝑠𝑜𝑙𝑢𝑡𝑖𝑜𝑛 + (1 − 𝜔) ∗ 𝐺𝐴 𝑠𝑜𝑙𝑢𝑡𝑖𝑜𝑛 (12)

where ω is the weighting factor.

- Step 8: Evaluate fitness: evaluate the fitness of the hybrid solution using the fitness function.
- Step 9: Update pheromones: update the pheromone values based on the fitness of the hybrid solution using
the formula from step 3.
- Step 10: Repeat: repeat the process from step 4 until a satisfactory solution is found, or until a predetermined
stopping criterion is met.
The proposed idea and method have also been tested on IEEE 14-bus systems. The system was
simulated using the following scenarios [27]:
- Case 1: The original data, but with the load-scaling factor applied LSF=1.2
- Case 2: The original data, but with the power limit of the lines:
L1-2 reduced from 1.2 to 0.8 pu
L4-5 reduced from 0.45 to 0.3 pu
L6-13 reduced from 0.32 to 0.2 pu
- Case 3: The initial data with a line outage L1-5;
- Case 4: The original data including outage of line L1-2;
In power systems, the economic dispatch (ED) model is used to optimize the allocation of power
generation resources, such as generators, to meet the electricity demand while minimizing the cost of
production. Violations can occur when the power generation exceeds the transmission capacity or violates other
operational constraints. Therefore, the ED model with ant system-genetic algorithm hybrid can be used to
adjust the generators' output to prevent such violations. The results of the adjustments made using the ED
model are presented in Table 6.
In the context of the IEEE 14-bus network, the hybridization of ACO and GA has shown promising
results. These results demonstrate how the ED model can successfully optimize the allocation of power
generation resources and prevent any operational violations, leading to a more efficient and cost-effective
power system. One study found that the hybrid approach outperformed both ACO and GA individually in terms
of convergence rate and solution quality, the hybrid approach was able to find multiple optimal solutions,
which is important in power system analysis where there may be multiple optimal solutions.

Table 6. Results H-ASGA of economic dispatch with lines overload

IEEE 14 Bus System Case 1 Case 2 Case 3 Case 4
PG1 (MW) 129.05 78.3739 93.289 52.3651
PG2 (MW) 33.61 73.4122 31.2652 93.5825
PG3 (MW) 71.46 64.9058 53.727 44.0837
PG4 (MW) 55.87 22.1545 55.097 40.6134
PG5 (MW) 26.13 23.6405 29.9899 32.0005
Total generation 316.125 262.4869 263.3681 262.6452
Real load (MW) 310.8 259.0000 259.0000 259.0000
Real loss (MW) 5.3254 3.4869 4.3681 3.6452
Fuel cost ($/hr) 1092.8796891.7594 842.0107 864.9385
Constraint 1,2,3 Satisfied Satisfied Satisfied Satisfied
Constraint 4 Overloaded linesSatisfied Satisfied Satisfied Satisfied

6. CONCLUSION
This article demonstrated a new optimization approach, without taking into account the lines overload.
The performance of the proposed methods (GA, FA, ABC, ACO, and PSO) was compared to that of recently
Artificial intelligence of optimal real power dispatch with constraints … (Abderrahmani Abdesselam)
1892  ISSN: 2088-8694

published optimization techniques, and it was found that the proposed methods achieved better results. The
results showed that the H-ASGA hybrid approach provided more diverse, efficient, and faster solutions than
pure ACO or GA approaches; his advantage also lies not only within the best production cost but also in
reliability and speed of execution.
This Paper also discussed and gave a solution for violation of an unavoidable constraint of overloading
transmission lines. When line overload limits are taken into account, optimal load-flow calculations become a
potent and practically useful tool for the operation and design of the system. The OPF technique is flexible
enough to accommodate a number of different limitations. The application of this algorithm can lead to more
efficient and cost-effective operation of electrical networks. In order for the OPF to accurately depict ideal
performance in real-world situations, there are still several obstacles to overcome in its development.

REFERENCES
[1] S. E. De Leon-Aldaco, H. Calleja, and J. Aguayo Alquicira, “Metaheuristic Optimization Methods Applied to Power Converters: A
Review,” IEEE Transactions on Power Electronics, vol. 30, no. 12, pp. 6791–6803, 2015, doi: 10.1109/TPEL.2015.2397311.
[2] J. Miao, N. Zhang, C. Kang, J. Wang, Y. Wang, and Q. Xia, “Steady-state power flow model of energy router embedded AC
network and its application in optimizing power system operation,” IEEE Transactions on Smart Grid, vol. 9, no. 5, pp. 4828–4837,
2018, doi: 10.1109/TSG.2017.2672821.
[3] M. H. Ali, A. M. A. Soliman, and S. K. Elsayed, “Optimal power flow using archimedes optimizer algorithm,” International Journal
of Power Electronics and Drive Systems, vol. 13, no. 3, pp. 1390–1405, 2022, doi: 10.11591/ijpeds.v13.i3.pp1390-1405.
[4] R. A. Abttan, A. H. Tawafan, and S. J. Ismael, “Economic dispatch by optimization techniques,” International Journal of Electrical
and Computer Engineering, vol. 12, no. 3, pp. 2228–2241, 2022, doi: 10.11591/ijece.v12i3.pp2228-2241.
[5] J. Zhu, OPTIMIZATION OF POWER SYSTEM OPERATION. Wiley, 2015.
[6] P. K. Rao, P. V. N. M. Krishna, B. S. S. Santosh, and P. Sumanth, “A study on economic load dispatch in power system in the
environment perspective by PSO,” IEEE International Conference on Power, Control, Signals and Instrumentation Engineering,
ICPCSI 2017, pp. 940–943, 2018, doi: 10.1109/ICPCSI.2017.8391850.
[7] W. Ongsakul and D. N. Vo, “Artificial intelligence in power system optimization,” Artificial Intelligence in Power System
Optimization, pp. 1–510, 2016, doi: 10.1201/b14906.
[8] A. Rizki, R. Habachi, K. Tahiry, and A. Echchatbi, “Economic dispatch problem in smart grid system with considerations for
pumped storage,” Bulletin of Electrical Engineering and Informatics, vol. 11, no. 6, pp. 3004–3012, 2022, doi:
10.11591/eei.v11i6.4207.
[9] R. Habachi, A. Touil, A. Boulal, A. Charkaoui, and A. Echchatbi, “Comparative study of metaheuristics methods applied to smart
grid network in Morocco,” International Journal of Power Electronics and Drive Systems, vol. 11, no. 1, pp. 487–495, 2020, doi:
10.11591/ijpeds.v11.i1.pp487-495.
[10] S. C. Kim and S. R. Salkut, “Optimal power flow based congestion management using enhanced genetic algorithms,” International
Journal of Electrical and Computer Engineering, vol. 9, no. 2, pp. 875–883, 2019, doi: 10.11591/ijece.v9i2.pp875-883.
[11] A. Latif and P. Palensky, “Economic Dispatch Using Modified Bat Algorithm,” Algorithms, vol. 7, no. 3, pp. 328–338, Jul. 2014,
doi: 10.3390/a7030328.
[12] M. Balasubbareddy, D. Dwivedi, G. V. K. Murthy, and K. S. Kumar, “Optimal power flow solution with current injection model of
generalized interline power flow controller using ameliorated ant lion optimization,” International Journal of Electrical and
Computer Engineering, vol. 13, no. 1, pp. 1060–1077, 2023, doi: 10.11591/ijece.v13i1.pp1060-1077.
[13] F. Benhamida and R. Belhachem, “Dynamic Constrained Economic/Emission Dispatch Scheduling Using Neural Network,”
Advances in Electrical and Electronic Engineering, vol. 11, no. 1, Mar. 2013, doi: 10.15598/aeee.v11i1.745.
[14] A. Salhi, D. Naimi, and T. Bouktir, “Fuzzy multi-objective optimal power flow using genetic algorithms applied to algerian
electrical network,” Advances in Electrical and Electronic Engineering, vol. 11, no. 6, pp. 443–454, 2013, doi:
10.15598/aeee.v11i6.832.
[15] D. Devaraj and B. Yegnanarayana, “Genetic-algorithm-based optimal power flow for security enhancement,” IEE Proceedings:
Generation, Transmission and Distribution, vol. 152, no. 6, pp. 899–905, 2005, doi: 10.1049/ip-gtd:20045234.
[16] W. Cheng and H. Zhang, “A dynamic economic dispatch model incorporating wind power based on chance constrained
programming,” Energies, vol. 8, no. 1, pp. 233–256, 2015, doi: 10.3390/en8010233.
[17] J. K. Pattanaik, M. Basu, and D. P. Dash, “Dynamic economic dispatch: a comparative study for differential evolution, particle
swarm optimization, evolutionary programming, genetic algorithm, and simulated annealing,” Journal of Electrical Systems and
Information Technology, vol. 6, no. 1, 2019, doi: 10.1186/s43067-019-0001-4.
[18] L. Benasla, A. Belmadani, and M. Rahli, “Hooke-Jeeves’ method applied to a new economic dispatch problem formulation,”
Journal of Information Science and Engineering, vol. 24, no. 3, pp. 907–917, 2008, doi: 10.6688/jise.2008.24.3.15.
[19] T. M. Mohan and T. Nireekshana, “A Genetic Algorithm for Solving Optimal Power Flow Problem,” Proceedings of the 3rd
International Conference on Electronics and Communication and Aerospace Technology, ICECA 2019, pp. 1438–1440, 2019, doi:
10.1109/ICECA.2019.8822090.
[20] M. A. M. Shaheen, H. M. Hasanien, S. F. Mekhamer, and H. E. A. Talaat, “Optimal power flow of power systems including
distributed generation units using sunflower optimization algorithm,” IEEE Access, vol. 7, pp. 109289–109300, 2019, doi:
10.1109/ACCESS.2019.2933489.
[21] K. Sarker, B. Roy, J. Sarker, and D. Santra, “A solution procedure to the economic load dispatch problem through the gravitational
search technique,” International Journal of Engineering, Science and Technology, vol. 11, no. 1, pp. 10–21, 2018, doi:
10.4314/ijest.v11i1.2.
[22] V. Suresh, P. Janik, and M. Jasinski, “Metaheuristic approach to optimal power flow using mixed integer distributed ant colony
optimization,” Archives of Electrical Engineering, vol. 69, no. 2, pp. 335–348, 2020, doi: 10.24425/aee.2020.133029.
[23] M. M. Bhaskar and S. Maheswarapu, “A hybrid genetic algorithm approach for optimal power flow,” Telkomnika, vol. 9, no. 2, pp.
211–216, 2011, doi: 10.12928/telkomnika.v9i2.689.

Int J Pow Elec & Dri Syst, Vol. 14, No. 3, September 2023: 1885-1893
Int J Pow Elec & Dri Syst ISSN: 2088-8694  1893

[24] A. A. A. Galal, A. A. A. Mousa, and B. N. Al-Matrafi, “Ant Colony Optimization Approach Based Genetic Algorithms for
Multiobjective Optimal Power Flow Problem under Fuzziness,” Applied Mathematics, vol. 04, no. 04, pp. 595–603, 2013, doi:
10.4236/am.2013.44084.
[25] S. B. Pandya, S. Ravichandran, P. Manoharan, P. Jangir, and H. H. Alhelou, “Multi-Objective Optimization Framework for Optimal
Power Flow Problem of Hybrid Power Systems Considering Security Constraints,” IEEE Access, vol. 10, pp. 103509–103528,
2022, doi: 10.1109/ACCESS.2022.3209996.
[26] M. M. Alobaedy and K. R. Ku-Mahamud, “Scheduling jobs in computational grid using hybrid ACS and GA approach,”
Proceedings - 2014 IEEE Computers, Communications and IT Applications Conference, ComComAp 2014, pp. 223–228, 2014,
doi: 10.1109/ComComAp.2014.7017200.
[27] S. Haddi, O. Bouketir, and T. Bouktir, “Improved optimal power flow for a power system incorporating wind power generation by
using Grey Wolf Optimizer algorithm,” Advances in Electrical and Electronic Engineering, vol. 16, no. 4, pp. 471–488, 2018, doi:
10.15598/aeee.v16i4.2883.

BIOGRAPHIES OF AUTHORS

Abdesselam Abderrahmani holds a PhD from the University of Bechar. His

thesis examined the control and command of "FACTS" systems. He worked eleven years at
the electrical distribution company, including six years as a SCADA project manager. He
was an Associate Professor at Adrar University from 2011 to 2016. In 2016, he joined the
University of Bechar. His scientific research is focused on the control and command of power
systems, the study of the dynamic stability of the networks electrical supply, as well as the
operation and safety of industrial systems. Currently, he works as a teacher at the university,
and he is preparing a book on the industrial network. He can be contacted at email:
[email protected], [email protected].

Nasri Abdelfatah was born in 1978 at Bechar-Algeria, he's received the

electrical engineering diploma from Bechar Center University-Algeria in 2002, and the
Master degree from the University of Sciences and Technology of Oran (USTO), Algeria in
2006, from 2006 to 2011 he was preparing his PHD diplomat, from 2018 until this moment
he is full professor of electrical engineering department. He can be contacted at email:
[email protected].

Gasbaoui Brahim doctorate in science from the University of Bechar (Algeria).

Currently, he teaches electrical engineering at the University of Bechar. His research focuses
on power electronics, robust and optimal control, electric propulsion system and
mechatronics. He can be contacted at email: [email protected].

Artificial intelligence of optimal real power dispatch with constraints … (Abderrahmani Abdesselam)