Mathematical Modelling of Engineering Problems: Received: 30 November 2021 Accepted: 9 June 2022

Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Mathematical Modelling of Engineering Problems

Vol. 9, No. 3, June, 2022, pp. 606-614


Journal homepage: http://iieta.org/journals/mmep

Dynamic Economic Load Dispatch Using Linear Programming and Mathematical-Based


Models
Ahmad Al-Subhi

Electrical and Electronics Engineering Technology Department, Yanbu Industrial College, Yanbu 46452, Saudi Arabia

Corresponding Author Email: subhi@rcyci.edu.sa

https://doi.org/10.18280/mmep.090307 ABSTRACT

Received: 30 November 2021 Economic dispatch (ED) is one of the most important topics in power system operation
Accepted: 9 June 2022 and planning. The main purpose of this paper is to develop simple and effective
mathematical models for the ED problem. Two stages were considered to solve this
Keywords: problem. First, the ED problem is formulated using linear piecewise functions and then
economic dispatch, TuringBot, linear optimally solved using the LP technique at various load values. The effectiveness of the
programming, mathematical models LP in optimally solving the ED problem is verified by applying it to two different test
systems. The results are compared with those obtained using other ED optimization
techniques. The LP optimization performance of the proposed method is found to be
similar to those of the reported techniques. In the second stage, the data collected from
the optimization process in the first stage are transferred to TuringBot software. This
software is adopted to build efficient mathematical models for the optimal power
generation (output parameters) as functions of the load values (input parameters). The
main objective of these models is to easily evaluate the optimal power sharing of the
generators in an online fashion under rapid variable loading conditions without the need
to solve the ED-LP based problem. Optimization techniques, including the LP,
generally require considerable simulation times for linearization and optimization code
execution, particularly under fast load variations. Thus, the main features of the
developed models in this paper are simplicity, accessibility, as well as the ability in
obtaining an efficient and optimal solution with a faster execution time.

1. INTRODUCTION A literature survey has recently explored several studies


reported on solving the ED problem using both established and
The economic dispatch (ED) problem is a crucial topic in new techniques. Examples of such techniques include the
the field of power system planning and operation. The main gradient method [1], fast lambda iteration (FLA) [2], and
objective of the ED solution is to determine the power enhanced lambda iteration (ELI) algorithm [3].
generated by each generator in order to minimize the total cost The PSO is frequently used for ED applications. A thorough
while meeting several constraints including load demand and literature survey [4] has been conducted on several studies that
minimum and maximum generation capacity for each employed the PSO along with its modified versions to solve
generator. An efficient and accurate evaluation of the power the ED problem. Other recent heuristic techniques include the
generated by each generator can improve the system symbiotic organism search (SOS) [5], ant lion optimization
significantly and lead to a reliable operation of the power (ALO) [6], differential evolution (DE) [7], grey wolf
system. In recent years, various optimization methods have optimization (GWO) [8], flower pollination algorithm (FPA)
been proposed to solve the ED problem. The methods differ in [9], and gravitational search algorithm (GSA) [10].
different aspects including accuracy, execution time, Hybrid techniques have been recently employed to optimize
complexity and simulation workspace. In general, the the ED problem. A gradient search method with an improved
techniques can be categorized into classical or traditional Jaya algorithm has been proposed to solve the ED problem
techniques and heuristic techniques. considering environmental issues [11]. Moreover, the bees
Classical techniques have been frequently employed in algorithm (BA) along with the TS has been studied [12],
solving the ED problem. Examples of such techniques include wherein the results were better than those obtained using the
lambda iteration algorithm (LIA), Lagrange multiplier and general BA. In addition, the hybrid QP and compact
gradient method. In addition, heuristic and artificial intelligent formulation method (CFM) have been used for the ED with
approaches have been widely adopted for the ED optimization line losses and prohibited operating zones [13]. Firefly and
purposes. These techniques include genetic algorithm (GA), self-regulating particle swarm optimization (FSRPSO)
tabu search (TS), particle swarm optimization (PSO) and algorithms are hybridized to solve and optimize the ED
simulated annealing (SA). Other methods include the cost problem [14].
composite function, dynamic programming (DP) and The ED problem has been also adopted in the field of
quadratic programming (QP). Moreover, hybrid methods have Distributed Generation (DG) and smart grids. The objective of
been also used to solve the ED problem. ED in the DG field is to optimize the optimal power sharing

606
from different energy sources rather than only one developed models in this paper can be executed instantly
conventional source. Examples of DG sources include wind, without a need to any iterative algorithms or trial and error
solar, battery storage and fuel cell. Thus, adopting such operations, hence leading to considerable time saving.
sources add more complexity to the ED optimization model as In this paper, all quadratic cost functions are first linearized
each DG source has its own limitations and constraints that using piecewise linearized cost functions. After the linearized
have to be considered in the optimization process. Various model is formulated, two ED test systems are solved using the
approaches have been proposed to optimize the power LP optimization model. Accurate mathematical models are
generated by different DG systems in smart grid environment then developed using TuringBot software for each generating
[15-17]. unit based on the ED simulations and data collected from the
In the field of LP, a quick and effective LP algorithm has LP model. The main objective of such mathematical models is
been applied to solve the ED problem where there are several to accurately and rapidly determine the optimal solution under
interconnected micro-grids [18]. Computational experiments fast load variations without having to solve the ED-LP based
demonstrated the accuracy and speed of the LP model [18]. An algorithm for each load value. The computational analysis in
algorithm to solve the ED problem while integrating this study was carried out using MATLAB environment on a
renewable energy resources has been described [19]. The PC with a 64-bit operating system and a 2.6 GHz processor.
algorithm adopts LP owing to its flexibility and reliability. An The rest of this paper is organized as follows. Section 2
LP-driven multi-criteria decision making approach has been discusses the ED problem and the linearization approach.
presented for a multi-objective ED in smart grids [20]. In Section 3 explains how the general optimization mathematical
another study, LP based hybrid method for the ED of both models are formulated. TuringBot software is briefly
energy and reserve is discussed. Assuming the reserve presented in section 4. The simulation results and comparative
requirement is evaluated in the optimization process, the analysis are discussed in section 5 and 6 for the ED-LP based
technique is used to estimate the optimal power flow and the solution and the mathematical models performance,
reserve assigned to each generating unit. This methodology respectively. Finally, the conclusions of this study are given in
combines LP with two meta-heuristic algorithms: mean- section 7.
variance mapping optimization (MVMO) and evolutionary
particle swarm optimization (EPSO) [21].
In general, the complexity of the ED optimization problem 2. OPTIMAL ECONOMIC DISPATCH USING LINEAR
depends mainly on the objective function and the operational PROGRAMMING
system constraints. For example, this would include ignoring
or considering transmission line losses. In addition, various This section discusses the ED problem concept and explains
other constraints may be added in the problem including the in detail the linearization methodology of the quadratic cost
prohibited zone, ramp rate limits, valve point effect, and functions of the generators in order to be appropriately used in
consideration of environmental issues and carbon emissions. the optimization process using the LP concept.
This would add more complexity to the ED problem and more
advanced techniques are thus required to find the optimal 2.1 Economic load dispatch problem
solution. For example, several studies have analyzed the ED
problem in consideration of carbon emissions [22-24]. In The ED problem is an important optimization problem in
addition, the ED problem considering the valve point effect the area of power generation, operation and planning. In the
has been also discussed [25, 26]. ED problem, several generating units with their respective cost
This high number of research work in this field indicates functions are given. The objective of the ED is to evaluate the
that this area of study is still under consideration and produced power by each generating unit in order to minimize
investigation, and there is still an opportunity to improve the the total fuel cost while satisfying the generation-load balance
ED optimization process in terms of simplification and and other technical and operational constraints. In power
accuracy. Although the previously published algorithms have plants, the ED is considered as a major activity for power
shown efficient performance in solving the ED problem, they system operation and planning for engineers. Therefore, it is
are still in need for some requirements. The majority of such highly recommended to accurately estimate the optimal power
algorithms are not easy for anyone to understand. It is not a produced by each generating unit. This can lead to a significant
simple task to understand the algorithm, write the code and enhancement to the power system reliability.
tune the parameters and carry out the simulations. Special A typical power plant has several generating units. At any
software programs are also required to perform such instant of time, the available generating units should be able to
simulations. Such programs are usually costly and not easily meet the total load requirement. The ED optimization process
available in any PC. The main contribution of this work is to ultimately estimates the optimal power sharing by each
develop simple mathematical models that can easily evaluate generating unit to supply a certain load in the minimum
the optimal power sharing among all generators in very simple possible cost [27]. Therefore, the objective of the ED
and quick way. The models can be easily evaluated without optimization process is to find the optimal values of the
requiring to solve the ED problem. Moreover, the models can powers outputted by the generators which will minimize the
be easily implemented in any software such as Excel, total cost and satisfy the power demand, units capacity and
MATLAB or any other computational platform. other operational constraints.
Another issue in the optimization techniques is the need of Figure 1 shows the structure of a system composed of N
solving the ED problem and performing and executing the generating units connected to a common bus, supplying a
optimization process at each load value. The actual power certain load Pload. The input to each unit Ci(Pi) represents the
systems are usually subjected to rapid load variations. Thus, cost of power generation. The output Pi represents the power
carrying out the optimization process at each and every load produced by that unit. The total cost of the system is the
value may increase the simulation time. On other hand, the summation of the respective generation costs of all units. The

607
first constraint considered here is that the summation of the where, pi,min and pi,max are the minimum and maximum power
power generated by each unit must be equal to the supplied outputs of generator i, respectively.
load. The second constraint is the generating capacity limits of
each unit. These limits define the maximum and minimum 2.2 Linearization methodology and linear programming
powers that can be produced by each unit.
As discussed previously, the cost function presented in (2)
is in a quadratic form. Given this nonlinear cost equation for
generator i, it is possible to approximate the nonlinear curve
by a series of straight-line segments [28]. Figure 2 shows an
example of a linearized cost function represented by three
linear segments. The linearization methodology has been
explained in details in Ref. [28].
For simplicity, in the piecewise linearization approach, the
segment widths W are generally considered equal. Therefore,
Figure 1. ED problem representation with N generating units the segment width W can be obtained as follows:

In the optimization process, the objective function Z W = ( pi ,max − p )/K (5)


i ,min
represents the total cost of the power generated for supplying
a given load, while subjected to the defined constraints. Thus,
the objective function of the optimization process of the ED where, K is the number of segments. Notably, employing more
problem is represented as follows: segments will decrease the segment width and improve the
linearization accuracy. However, this increases the
N
computation time.
Minimize Z =  C i ( Pi ) (1)
i =1

This function describes the relationship between the fuel


cost and the power produced by each generating unit. There is
one cost function for each unit representing the actual behavior
of the generator. In the optimization process, it is required that
the power sharing by each generator for supplying a certain
load to be accurately optimized such that the fuel cost is
minimized.
The cost function usually has the form of a quadratic
equation. This function may be also represented by a cubic
function for some generating units. Therefore, the quadratic
representation of the total cost in $/h for a given generator, i
Figure 2. Quadratic cost function linearized by three
takes the following form:
piecewise linear functions

Ci ( Pi ) = ai + bi Pi + ci Pi , i=1, …, N
2
(2) For generator i, the last point of segment k is evaluated by
the following equation:
where, Ci(Pi) is the total generation cost of generator i; ai, bi,
and ci are the cost constants of generator i; N is the number of p = p + k  W , i = 1,...., N , k = 0,...., K (6)
ik i , min
generators; and Pi is the power output of generator i.
The following equality constraint is defined as the power
balance equation and is used to impose a balance between the where, pi0=pi,min and piK=pi,max.
total generation and demand. By ignoring the transmission line The linear segment slope, sik is evaluated by estimating the
losses, this balance is expressed in the form of the following cost values at the starting and end points of a segment (pi,k–1
constraint: and pi,k, respectively). Then the difference is divided by the
segment width. This is represented by the following equation:
N
 Pi = Pload (3)
i =1
s = [C ( p ) − C ( p )] / W (7)
ik i ik i i , k −1

The following inequality constraints are the capacity Hence, the cost function for generator i can be re-written as
bounds of the power generator. Each unit i has a lower limit follows.
(pi,min) and an upper limit (pi,max) on the power generation.
These minimum and maximum limits represent the generation
C (P ) = C + s P + s P + .... + s P (8)
capacity of the generating unit which are related to the i i i , min i1 i1 i2 i2 iK iK

technical specifications and design of the generator. These


limits are represented by the following constraints: where,

pi ,min  P  pi ,max , i = 1,...., N (4) 0  P  W , k = 1, 2,...., K (9)


i ik

608
and A mathematical model for each generator can be formulated
using the data collected and recorded in the table. This model
P = P + P + P + .... + P (10) considers the load value as the input parameter and the optimal
i i , min i1 i2 iK
powers obtained using the LP as the output parameters. The
mathematical formulation for each generator takes the
Therefore, as shown from Eq. (8), this equation is a linear following forms:
function of the Pik values, which can be optimally evaluated
using the LP optimization technique. It is worth to note that
the fixed cost constants Ci,min are not considered in the LP Popt 1 = f ( L ) (12)
1
optimization process. However, these constants will be added
later after solving the optimization problem and evaluating the
optimal total cost. Popt 2 = f ( L ) (13)
2
The decision variables in the linearized problem include Pik, .
where Pik is calculated from the beginning of segment k. For .
each segment k, the value of the corresponding parameter Pik
is obtained as follows. PoptN = f ( L ) (14)
N

Pik =  min( Pi , pi , k ) − pi , k −1 , if Pik  pi , k −1


0, otherwise
(11)
Therefore, each unit has its own optimization model as a
function of the input demand. TuringBot software is adopted
to build and formulate the mathematical models based on the
collected data. Figure 3 shows a flow chart, summarizing the
3. ACCURATE REPRESENTATION OF GENERAL steps followed to build the general optimization model for a
OPTIMIZATION PREDICTIVE MODELS FOR selected generator. Figure 4 shows a descriptive block diagram
ONLINE ECONOMIC DISPATCH of the general optimization model.

Several previously reported studies have solved the ED


problem in an offline manner. Thus, the ED optimization
algorithm was required to be simulated whenever the load
changes to determine the optimal solution. Consequently, a
longer time is required to solve the ED problem for each load
value. Hence, an accurate and simple model is needed to deal
with such fast load variations. This section proposes an online
dynamic ED model that can be used to determine the optimal
solution for any load value without simulating the ED-LP
based optimization algorithm.
The proposed model is first formulated by considering the
generating units shown in Figure 1. This figure shows N
generating units with known cost functions required to supply
the connected load Pload. Thus, given the range of the load
values, the ED problem is solved for each specific load value,
wherein the output parameters are the optimal power
generated by each generation unit. The proposed LP
optimization process is employed to solve the dynamic ED
problem wherein the input parameters (load values) and output
parameters (optimal power values) are stored and recorded.
After finalizing the simulations of the load values, the results
are recorded in a table, as listed in Table 1. Here, L1, L2, L3,….,
Ln represent the load values selected to develop the model
(training process). The values must be selected to cover a
representative and uniform range of possible connected
demands and must be within the range of the sum of the
minimum and maximum generator limits.
Figure 3. Flowchart of the proposed methodology for general
Table 1. Collected data used to build the general optimization model
optimization model for each generator

Input
Output parameters
parameters
Load (kW) Popt1 (kW) Popt2 (kW) . . PoptN (kW)
L1 Popt11 Popt21 . . PoptN1
L2 Popt12 Popt22 . . PoptN2
L3 Popt13 Popt23 . . PoptN3
. . . . . .
. . . . . . Figure 4. Descriptive block diagram of the general
Ln Popt1n Popt2n . . PoptNn optimization model

609
4. TURINGBOT SOFTWARE: OVERVIEW model. The MATLAB function linprog is employed to
determine the optimal solution to this problem. Table 2 lists
TuringBot is a desktop software that uses symbolic the optimal solutions found using the LP for a load value of
regression to find mathematical models from data values. This 200 kW. In addition, the table compares the ED-LP based
software is employed to determine mathematical relationships solutions with some other techniques including the QP using
that describe sets of measured inputs and outputs data in their the CPLEX solver and iterated-based algorithm (IBA) [30],
simplest form [29]. Figure 5 shows the block diagram of the LIA [27], and ALO [6].
basic function of TuringBot.
Table 2. Optimal solution of the three DG system microgrid

Load (kW) 200


Method CPLEX IBA LIA ALO LP
[30] [30] [27] [6] (proposed)
P1 (kW) 67.949 67.96 67.949 67.949 67.95
P2 (kW) 73.996 74 73.996 73.996 73.997
P3 (kW) 58.055 58.03 58.055 58.055 58.053
Figure 5. Basic function of TuringBot Total cost
($/h) 11.076 11.076 11.076 11.076 11.0763
In this figure, the input parameters, u1….un and the
corresponding output parameter y are measured 5.2.2 Ten thermal units test system
experimentally or by simulations, at which sufficient samples This system comprises ten thermal units having quadratic
are necessary to represent all operating conditions of the cost functions. Constants values and generator limits can be
system. The collected input and output data are transferred to found in Ref. [31]. Table 3 lists the optimal solutions to the
TuringBot software to start the training process and formulate ED problem obtained using the LP, as well as those obtained
the mathematical models. The mathematical relationship using the LIA [27] and ALO [6] for a 2000-kW demand.
representing the output parameter as a function of the input
parameters is given as follows. Table 3. Optimal solutions of the ten-unit test system

Load (kW) 2000


y = f (u1 , u2 ,....., un ) (15) Method LIA [27] ALO [6] LP (proposed)
P1 (kW) 600 600 600
In TuringBot, different mathematical models are developed P2 (kW) 600 599.99 600
and the user can simply select the best model that fits the P3 (kW) 300 300.01 300
P4 (kW) 100 100 100
input/output data with the lowest error.
P5 (kW) 50 50 50
P6 (kW) 100 100 100
P7 (kW) 100 100 100
5. SOLUTION OF THE ED PROBLEM USING LINEAR P8 (kW) 50 50 50
PROGRAMING P9 (kW) 50 50 50
P10 (kW) 50 50 50
5.1 Overview Total cost ($/h) 16579.75 16579.759 16579.75

As stated earlier, the proposed ED-LP based algorithm is 5.3 Discussion


applied to different benchmark test systems. Two test systems
are selected here to carry out this analysis. The first system is As can be seen from Table 2 for the DG system and Table
a microgrid system consisting of a microturbine and two diesel 3 for the ten thermal units system, the solution obtained using
generators, while the second system is comprising the LP is similar to that obtained using the other techniques,
conventional thermal generating units. For each system, the thus verifying the effectiveness of the proposed methodology.
optimal solution for a selected load value is found using the Hence, at this stage, the ED-LP based algorithm can be
LP and then compared with the values obtained using some efficiently used to build the general predictive optimization
other reported techniques. models for modeling the optimal solution under varying load
conditions without having to solve the ED-LP based algorithm
5.2 Simulation results for each load value.

5.2.1 DG test system


Three DG systems are incorporated into this system, 6. DYNAMIC SOLUTION OF THE ED PROBLEM
representing a small microgrid. The cost functions for the DG USING MATHEMATICAL MODELS
have the same form as that of the quadratic cost equation of
the conventional generators, shown in Eq. (2). Constant 6.1 Overview
parameters values and power generation limits for each DG
can be found in the study [30]. The main objective of this subsection is to formulate
The system is discussed here in terms of the ED using the accurate mathematical predictive models of the optimal power
LP for a fixed load value. The quadratic cost functions of the generated by each unit under variable loading conditions.
generators are linearized using piecewise linearization as Therefore, for training data collection, the ED-LP based
described previously. The problem can then be formulated as problem is solved for a wide and representative range of loads.
a linear equation, which is appropriately used in the LP based For the DG system, the demand is varied from 90 kW up to

610
400 kW in steps of 1 kW. For the ten thermal units system, the 20, 90  L  95.75
load is varied from 800 kW to 3200 kW. It should be noted 0.6265L − 40.06, 95.75  L  143.25
that any load value out of this range gives an infeasible 
P = 0.4284 L − 11.6868, 143.25  L  207.75 (17)
solution, as it would be beyond the sum of the minimum and opt 2
0.5753L − 42.18, 207.75  L  316.75
maximum generator limits. For each load value, the optimal 
power generated by each generation unit is determined. This 140, elsewhere
process is repeated for all the selected load values and the
optimal solutions obtained in each case are recorded and stored. L − 70, 90  L  95.85
Figure 6 shows the trend in the optimal solutions obtained for 
0.3735L − 9.9424, 95.85  L  143.25
each DG with respect to the response to the load variations. P = (18)
0.2554 L + 6.9778, 143.25  L  207.58
opt 3
Also, for the ten thermal units system, Figure 7 shows the
optimal solutions for demand values ranging from 800 kW to 
60, elsewhere
3200 kW, obtained using the same approach as that applied to
the three-DG microgrid.
Thus, Eqns. (16)-(18) are employed here to determine the
optimal solution of the DG microgrid for any load value within
the feasibility range. These equations can be simply coded in
MATLAB or any other software to directly evaluate the
optimal solution. To check the validity of the developed
models, a set of new random load values are used as input
variables for both the ED-LP based algorithm and the
developed general optimization models. Figure 8 shows 50
new load values selected randomly within the feasibility range.

Figure 6. Behavior of the optimal power of each DG under


variable loading conditions

Figure 8. Randomly generated load profile within the


feasibility range of the microgrid

6.2.2 Ten thermal units system


For this system, the general predictive optimization models
have been also formulated using the data collected from the
Figure 7. Behavior of optimal power of each thermal unit ED-LP based simulations. The formulation procedure is the
under variable loading conditions same as that performed for the three-DG microgrid system.
Based on the load variations and by following the solution
6.2 Simulation results trend shown in Figure 7, the optimal powers of each generation
unit are formulated as a function of the load values as follows:
6.2.1 DG system
From Figure 6, it is clear that some mathematical
relationships can be obtained for each parameter as a function L − 700, 800  L  1300
P =  (19)
of the load demand. A single mathematical model for each
600, elsewhere
opt 1
parameter can be formulated. However, for more accurate
representation, optimal power mathematical relationship can
be divided into approximately 4–5 functions based on the load 100, 800  L  1300
values. Thus, following this arrangement, the final general 
P = L − 1200, 1300  L  1800 (20)
predictive optimization models for each DG are as shown in opt 2
Eqns. (16)-(18): 600, elsewhere

50, 90  L  143.25
0.3162L + 4.7084, 143.25  L  207.5 100, 800  L  1800

P = (16) P = L − 1700, 1800  L  2100 (21)
0.4247 L − 17.81, 207.5  L  316.75
opt 1 opt 3
400, elsewhere
L − 200, elsewhere

611
100, 800  L  2175 6.3 Discussion

0.4444 L − 866.7, 2175  L  2490
P = (22) Thus, for both test systems, the performance of the
L − 2250, 2490  L  2650
opt 4
mathematical models in estimating the optimal power

400, elsewhere generation is compared with the LP solution. Figure 10 shows
the comparison of the optimal solutions obtained using both
approaches for each DG, including the difference absolute
50, 800  L  2100 percentage error. Also, Figures 11 and 12 show the

L − 2050, 2100  L  2175 comparison between the optimal solutions found using the
P = (23) ED-LP based algorithm and the developed mathematical
0.5556L − 1080, 2175 < L  2490
opt 5

models for the ten thermal units system. The figures clearly

300, elsewhere show that the percentage error is very low for all cases, thus
verifying the high accuracy of the developed mathematical
100, 800  L  2650 models and the ability of such models in accurately
 determining the optimal solutions for any load value within the
P = L − 2550, 2650  L  2850 (24) feasibility range without having to simulate the ED-LP based
opt 6

300, elsewhere algorithm.

100, 800  L  2850



P = L − 2750, 2850  L  2950 (25)
opt 7

200, elsewhere

50, 800  L  3000



P = L − 2950, 3000  L  3150 (26)
opt 8

200, elsewhere

50, 800  L  2950



P = L − 2900, 2950  L  3000 (27)
opt 9

100, elsewhere

50, 800  L  3150


P =  (28)
L − 3100, elsewhere
opt 10

Therefore, Eqns. (19)-(28) are used now as general Figure 10. Optimal solution comparisons between proposed
predictive models that can evaluate the optimal solution for ED-LP based algorithm and the developed general
any load value without the need to solve the ED-LP optimization models of the microgrid
optimization problem.
The validity of the proposed models is tested now using
some random and new input load values. Figure 9 shows the
randomly generated load profile selected from the feasibility
range.

Figure 9. Randomly generated load profile within the Figure 11. Optimal solution comparisons between proposed
feasibility range of the 10 thermal units ED-LP based algorithm and proposed general optimization
models for units 1-5

612
[3] Singhal, P.K., Naresh, R., Sharma, V., Kumar, G. (2014).
Enhanced lambda iteration algorithm for the solution of
large scale economic dispatch problem. In International
Conference on Recent Advances and Innovations in
Engineering (ICRAIE-2014), pp. 1-6.
http://dx.doi.org/10.1109/ICRAIE.2014.6909294
[4] Abbas, G., Gu, J., Farooq, U., Asad, M. U., El-Hawary,
M. (2017). Solution of an economic dispatch problem
through particle swarm optimization: A detailed survey-
part I. IEEE Access, 5: 15105-15141.
http://dx.doi.org/10.1109/ACCESS.2017.2723862
[5] Vedik, B., Naveen, P., Shiva, C.K. (2020). A novel
disruption based symbiotic organisms search to solve
economic dispatch. Evolutionary Intelligence, 1-36.
http://dx.doi.org/10.1007/s12065-020-00506-5
[6] Nischal, M.M., Mehta, S. (2015). Optimal load dispatch
using ant lion optimization. Int. J Eng Res Appl, 5(8): 10-
19.
[7] Zou, D., Gong, D. (2022). Differential evolution based
on migrating variables for the combined heat and power
Figure 12. Optimal solution comparisons between proposed dynamic economic dispatch. Energy, 238: 121664.
ED-LP based algorithm and proposed general optimization http://dx.doi.org/10.1016/j.energy.2021.121664
models for units 6-10 [8] Tung, N.S., Chakravorty, S. (2015). Grey wolf
optimization for active power dispatch planning problem
considering generator constraints and valve point effect.
7. CONCLUSIONS International Journal of Hybrid Information Technology,
8(12): 117-134.
In this paper, an ED solution methodology is proposed http://dx.doi.org/10.14257/ijhit.2015.8.12.07
based on the LP concept by linearizing the quadratic cost [9] Vijayaraj, S., Santhi, R.K. (2016). Multi-area economic
functions of the generation units. A piecewise linearization is dispatch using flower pollination algorithm. In 2016
employed, wherein thousands of segments are used to International Conference on Electrical, Electronics, and
precisely convert the quadratic functions to linear functions. Optimization Techniques (ICEEOT), pp. 4355-4360.
The effectiveness of the methodology is verified by applying http://dx.doi.org/10.1109/ICEEOT.2016.7755541
it to two different benchmark test systems, and the results are [10] Hota, P.K., Sahu, N.C. (2015). Non-convex economic
compared with those of other recently reported ED dispatch with prohibited operating zones through
optimization techniques. A general optimization predictive gravitational search algorithm. International Journal of
model is then proposed for each generation unit using simple Electrical and Computer Engineering, 5(6).
mathematical models, in which the input variable is the load http://dx.doi.org/10.11591/ijece.v5i6.pp1234-1244
value and the output is the optimally generated power. [11] Azizipanah-Abarghooee, R., Dehghanian, P., Terzija, V.
The simulations have shown the ability of the developed (2016). Practical multi-area bi-objective environmental
mathematical models to evaluate the optimal power generation economic dispatch equipped with a hybrid gradient
from all generating units. Moreover, simplicity is another search method and improved Jaya algorithm. IET
feature of the proposed method. Only a few and simple Generation, Transmission & Distribution, 10(14): 3580-
mathematical equations could represent the feasible operating 3596. http://dx.doi.org/10.1049/iet-gtd.2016.0333
conditions for the selected systems. The models are accessible [12] Chansareewittaya, S. (2017). Hybrid BA/TS for
and easy to be implemented and coded in any software. The economic dispatch considering the generator constraint.
models can also be evaluated instantly without performing any In 2017 International Conference on Digital Arts, Media
complicated algorithms and iterative operations. Thus, the and Technology (ICDAMT), pp. 115-119.
developed models have the features of simplicity, accessibility, http://dx.doi.org/10.1109/ICDAMT.2017.7904946
accuracy and fast execution time. [13] Wu, Z.L., Wu, Q.H., Zhou, X.X., Li, M.S. (2015).
Hybrid quadratic programming and compact formulation
method for economic dispatch with prohibited operating
REFERENCES zones and network losses. In 2015 IEEE Innovative
Smart Grid Technologies-Asia (ISGT ASIA), pp. 1-6.
[1] Imen, L., Mouhamed, B., Djamel, L. (2013). Economic http://dx.doi.org/10.1109/ISGT-Asia.2015.7386963
dispatch using classical methods and neural networks. In [14] Nasir, M., Sadollah, A., Aydilek, İ.B., Ara, A.L., Nabavi-
2013 8th International Conference on Electrical and Niaki, S.A. (2021). A combination of FA and SRPSO
Electronics Engineering (ELECO), pp. 172-176. algorithm for combined heat and power economic
http://dx.doi.org/10.1109/ELECO.2013.6713826 dispatch. Applied Soft Computing, 102: 107088.
[2] Zhan, J.P., Wu, Q.H., Guo, C.X., Zhou, X.X. (2013). Fast http://dx.doi.org/10.1016/j.asoc.2021.107088
λ-iteration method for economic dispatch with prohibited [15] Cherukuri, A., Cortés, J. (2017). Distributed coordination
operating zones. IEEE Transactions on Power Systems, of DERs with storage for dynamic economic dispatch.
29(2): 990-991. IEEE Transactions on Automatic Control, 63(3): 835-
http://dx.doi.org/10.1109/TPWRS.2013.2287995 842. http://dx.doi.org/10.1109/TAC.2017.2731809

613
[16] Huang, B., Liu, L., Zhang, H., Li, Y., Sun, Q. (2019). using artificial bee colony algorithm. In 2016 Biennial
Distributed optimal economic dispatch for microgrids International Conference on Power and Energy Systems:
considering communication delays. IEEE Transactions Towards Sustainable Energy (PESTSE), pp. 1-7.
on Systems, Man, and Cybernetics: Systems, 49(8): http://dx.doi.org/10.1109/PESTSE.2016.7516478
1634-1642. [24] Goudarzi, A., Swanson, A.G., Tooryan, F., Ahmadi, A.
http://dx.doi.org/10.1109/TSMC.2019.2900722 (2017). Non-convex optimization of combined
[17] Gil-González, W., Montoya, O.D., Holguí n, E., Garces, environmental economic dispatch through the third
A., Grisales-Noreña, L.F. (2019). Economic dispatch of version of the cultural algorithm (CA3). In 2017 IEEE
energy storage systems in dc microgrids employing a Texas Power and Energy Conference (TPEC), pp. 1-6.
semidefinite programming model. Journal of Energy http://dx.doi.org/10.1109/TPEC.2017.7868281
Storage, 21: 1-8. [25] Wu, Z., Ding, J., Wu, Q. H., Jing, Z., Zheng, J. (2017).
http://dx.doi.org/10.1016/j.est.2018.10.025 Reserve constrained dynamic economic dispatch with
[18] Hoke, A., Brissette, A., Chandler, S., Pratt, A., valve-point effect: A two-stage mixed integer linear
Maksimović, D. (2013). Look-ahead economic dispatch programming approach. CSEE Journal of Power and
of microgrids with energy storage, using linear Energy Systems, 3(2): 203-211.
programming. In 2013 1st IEEE Conference on http://dx.doi.org/10.17775/CSEEJPES.2017.0025
Technologies for Sustainability (SusTech), pp. 154-161. [26] Bhui, P., Senroy, N. (2016). A unified method for
http://dx.doi.org/10.1109/SusTech.2013.6617313 economic dispatch with valve point effects. In 2016
[19] Elsaiah, S., Benidris, M., Mitra, J., Cai, N. (2014). IEEE 6th International Conference on Power Systems
Optimal economic power dispatch in the presence of (ICPS), pp. 1-5.
intermittent renewable energy sources. In 2014 IEEE http://dx.doi.org/10.1109/ICPES.2016.7584108
PES General Meeting| Conference & Exposition, pp. 1- [27] Wood, A.J., Wollenberg, B.F., Sheblé, G.B. (2013).
5. http://dx.doi.org/10.1109/PESGM.2014.6939903 Power Generation, Operation, and Control. John Wiley
[20] Lazzerini, B., Pistolesi, F. (2015). A linear & Sons.
programming-driven MCDM approach for multi- [28] Al-Subhi, A.A., Alfares, H.K. (2016). Economic load
objective economic dispatch in smart grids. In 2015 SAI dispatch using linear programming: A comparative study.
Intelligent Systems Conference (IntelliSys), pp. 475- International Journal of Applied Industrial Engineering
484. http://dx.doi.org/10.1109/IntelliSys.2015.7361183 (IJAIE), 3(1): 16-36.
[21] Chamba, M., Ano, O. (2013). Economic dispatch of http://dx.doi.org/10.4018/IJAIE.2016010102
energy and reserve in competitive markets using meta- [29] TuringBot, S. (2021). Symbolic regression software.
heuristic algorithms. IEEE Latin America Transactions, https://turingbotsoftware.com.
11(1): 473-478. [30] Modiri-Delshad, M., Koohi-Kamali, S., Taslimi, E.,
http://dx.doi.org/10.1109/TLA.2013.6502848 Kaboli, S.H.A., Rahim, N.A. (2013). Economic dispatch
[22] Kuo, M.T., Lu, S.D., Tsou, M.C. (2017). Considering in a microgrid through an iterated-based algorithm. In
carbon emissions in economic dispatch planning for 2013 IEEE Conference on Clean Energy and Technology
isolated power systems. In 2017 IEEE/IAS 53rd (CEAT), pp. 82-87.
Industrial and Commercial Power Systems Technical http://dx.doi.org/10.1109/CEAT.2013.6775604
Conference (I&CPS), pp. 1-10. [31] Logenthiran, T., Srinivasan, D. (2009). Short term
http://dx.doi.org/10.1109/ICPS.2017.7945099 generation scheduling of a microgrid. In TENCON 2009-
[23] Bhongade, S., Agarwal, S. (2016). An optimal solution 2009 IEEE Region 10 Conference, pp. 1-6.
for combined economic and emission dispatch problem http://dx.doi.org/10.1109/TENCON.2009.5396184

614

You might also like