4440 IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 8, NO.

4, DECEMBER 2020

General Mathematical Solution for Selective

Harmonic Elimination
Mahrous Ahmed , Senior Member, IEEE, Mohamed Orabi , Senior Member, IEEE,
Sherif S. M. Ghoneim , Senior Member, IEEE, Mosleh M. Al-Harthi, Farhan A. Salem,
Basem Alamri, and Saad Mekhilef , Senior Member, IEEE

Abstract— This article proposes a general mathematical

solution for a selective harmonic elimination (SHE) technique. NS Number of the switching angles per a quar-
The proposed procedure is valid for both symmetrical and ter time period including optional notches.
unsymmetrical multilevel inverter (MLI) inverter configurations α1 , α2 , . . . , αk Switching angles.
with a wide range of modulation indexes (mi). The proposed l Number of eliminated harmonics.
procedure eliminates the targeted number of harmonics “l” for
1+l Number of equations groups.
switching angles NS = 2l regardless to the MLI output waveform
levels. The proposed procedure receives the values and ratios k = 1, 2, 3, . . . , NSmin Switching angles
of the dc sources, and then, it calculates the NS values to order.
generate a suitable set of harmonic equations. In this proposed S Ratio among the input dc voltage sources.
algorithm, the nonlinear transcendental harmonic equations were k1 Number of dc sources.
transformed into a linear system of equations ( AX = B); the N Maximum available number of output volt-
resulting equations can be solved online easily by formulating
matrix A and matrix B and then multiplying the inversion of age levels.
matrix A by matrix B. It has been found that only the last M Total normalized input dc voltage of the
element in matrix B needs to be updated for each change in the MLI.
modulation index or in the ac output fundamental component, THD Total harmonic distortion.
resulting in reduced computation time. The proposed technique mi Modulation index.
generates the switching angles α1 , α2 , . . . , α k based on a preset
objective functions. The simulation and experimental results
in some selected cases are provided to validate the proposed I. I NTRODUCTION
algorithms.
Index Terms— Multilevel inverters (MLIs), selective harmonic
elimination (SHE), total harmonic distortion (THD).
M ULTILEVEL inverters (MLIs) [1]–[4] have garnered
a lot of attention due to their inherent advantages
over conventional two-level H-bridge inverters. MLIs generate
a staircase of output voltage waveforms with low-harmonic
N OMENCLATURE
contents. Moreover, MLI switches are turned on/off at a low
hn Normalized harmonic-order component.
switching frequency; thus, they have low switching losses, low
n = 1, 3, 5, . . . Harmonic orders.
switching stresses, low electromagnetic interface (EMI), and
NSmin Minimum number of switching angles per
high system efficiency. Both the input side boosting stage and
a quarter time period.
the output side power transformer can be eliminated in MLIs,
Manuscript received May 1, 2019; revised June 24, 2019 and July 21, resulting in reducing the system’s size and efficiency.
2019; accepted July 25, 2019. Date of publication August 2, 2019; date Recently, the concept of selective harmonic elimina-
of current version November 5, 2020. This work was supported by
the Scientific Research Deanship, Taif University, Saudi Arabia, under tion (SHE) has received much attention [4]–[6]. SHE has
Grant 1-439–6072. Recommended for publication by Associate Editor the ability to remove the harmful lower harmonics from the
Joseph O. Ojo. (Corresponding author: Mahrous Ahmed.) output voltage waveforms. Thus, filtering higher harmonics
M. Ahmed is with the Electrical Engineering Department, College of
Engineering, Taif University, 21974 Ta’if, Saudi Arabia, and also with can be easily done using a small filter. The main challenge in
the Aswan Power Electronics Applications Research Center (APEARC), employing the SHE control algorithm is its nonlinear transcen-
Faculty of Engineering, Aswan University, 81542 Aswan, Egypt (e-mail: dental equations, which have complex solutions. In addition
[email protected]).
M. Orabi is with the Aswan Power Electronics Applications Research Center to their complexity, they drastically increase the number of
(APEARC), Faculty of Engineering, Aswan University, 81542 Aswan, Egypt. selected harmonics; therefore, the number of equations has
S. S. M. Ghoneim is with the Electrical Engineering Department, College also increased. Recently, several trials and techniques have
of Engineering, Taif University, 21974 Ta’if, Saudi Arabia, and also with
the Electrical Department, Faculty of Industrial Education, Suez University, been introduced to simplify the transcendental equations’
43527 Suez, Egypt (e-mail: [email protected]). solution with high accuracy for high levels of MLIs. These
M. M. Al-Harthi, F. Salem, and B. Alamri are with the Electrical Engi- can be categorized into three main groups: 1) numerical
neering Department, College of Engineering, Taif University, 21974 Ta’if,
Saudi Arabia. solutions [7]–[12]; 2) evolutionary solutions (EAs) [13]–[21];
S. Mekhilef is with the Department of Electrical Engineering, University of and 3) the mathematical solutions [22]–[31].
Malaya, 50603 Kuala Lumpur, Malaysia. The first group, numerical solutions, includes the well-
Color versions of one or more of the figures in this article are available
online at http://ieeexplore.ieee.org. known Newton–Raphson (NR), gradient optimization, and
Digital Object Identifier 10.1109/JESTPE.2019.2932933 sequential quadratic programming, in which the solution is fast
2168-6777 © 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.

Authorized licensed use limited to: National Institute of Technology Patna. Downloaded on December 14,2022 at 18:11:30 UTC from IEEE Xplore. Restrictions apply.
AHMED et al.: GENERAL MATHEMATICAL SOLUTION FOR SELECTIVE HARMONIC ELIMINATION 4441

iterative [6]. The solutions converge of the above-mentioned demonstrated with examples. Section IV presents the results
techniques mainly depends on the initial guesses for the values and discussion of the suggested simulations and experiments
of the unknowns. Therefore, choosing the proper initial values used to verify and validate the performance of the proposed
of the unknowns is the main challenge of these algorithms. technique. Section V gives numerical calculated examples for
The second group, EAs, includes particle swarm optimization the selected cases of studied methods. Section VI gives a
(PSO), genetic algorithms (GAs), and bee algorithms (BAs). detailed comparison between the proposed algorithm and its
EAs transform the transcendental equations of the fundamental existence and recent counterpart algorithms to highlight the
and lower order harmonics into a fitness function in order performance of the proposed algorithm. Finally, Section VII
to fetch the optimum firing angles by optimizing the fitness presents a summary of the main remarks on the proposed
function. The main advantage of EAs over numerical methods technique and presents the study’s conclusion.
is that EAs are less influenced by the initial guesses of the
unknowns [5]. II. M ATHEMATICAL P ROCEDURE
The third group, mathematical solutions [22]–[31], includes
Fig. 1(a) shows the general N-levels single-phase cascaded
many trials to solve the nonlinear transcendental equations
H-bridge inverter. Each unit has a single dc source and four
using algebraic algorithms. In [22]–[25], the transcendental
switches. The general ratio among MLI input dc sources
equations are transformed into polynomial equations, and the
is S (S = 1 for symmetrical MLI). Any dc sources ratio
resultant new system is solved mathematically without the
yields inverter output voltage waveforms as shown in Fig. 1(b)
need of the initial guesses for the values of the unknowns.
with α1 , α2 , α3 , . . . αn as MLI switching angles. The total
The method suffers from the computational complexity of
normalized input dc voltage of the MLI (M) can be calculated
the solutions; thus, it is not applicable in real time, and it
from the following equation:
is limited for low-level inverters. In [26] and [27], simple
closed-form analytical solutions were proposed where the non- M = 1 + S 1 + S 2 + · · · + S k1−1 (1)
linear transcendental equations are converted into a nonlinear
system that are solved analytically. The solution is very fast where k1 refers to the number of dc sources.
and very accurate. Moreover, while it guarantees harmonic Therefore, the maximum number of available levels is
cancellation, as stipulated, the solution is limited for two N = 2M + 1. (2)
or three switching angles; thus, they are not applicable to
higher levels. In [28], a novel analytical method, called selec- Using the Fourier series of the staircase function as shown
tive harmonic elimination pulse amplitude modulation (SHE in Fig. 1(b), the general normalized harmonic component (h n )
PAM), was presented to calculate the switching angles for all is normalized to Vdc1 as follows:
levels. However, the method mainly depends on equating the 4 
voltage second area under the reference sine signal and the hn = cos(nαk ), k = 1, 2, 3, . . . , NSmin (3)
nπ
n=1,3,5,....
synthesized multilevel waveform; thus, both the magnitude
of the voltage and the switching angles must be selected where α1 < α2 . . . αNS < (π/2), n = 1, 3, 5, . . . harmonic
at different levels. Other analytical techniques have been number. The minimum required number of switching angles,
presented in [29] and [30]; those techniques are simple, real- NS, is defined as the ceiling of the natural logarithm (ln) of
time analytical, and cost-effective, and they are succeeding in M as follows:
 
reducing the computational complexity. However, the studied ceil ln(M)
methods are limited to five-level inverter topologies. In [31], NSmin = 2 ln(2)
. (4)
a simple mathematical solution to solve transcendental nonlin- It is clear from (4) that the minimum required number of
ear equations was obtained by finding the direct relationships
switching angles will be equal or greater than M. The number
among the switching angles instead of their cosines. The k is the switching angles orders. Equation (3) can be applied
solution is simple and accurate, and it has a low computation
for voltage waveform that was given in Fig. 1(c) and (d) [27]
time; however, it is limited to a specific even number of
as follows:
switching angles, and it failed to control the fundamental
4
component. h n = [cos (nα1 ) + cos (nα2 ) − cos (nα3 )] (5.a)
This article presents a new online analysis and generalized π
method to calculate the switching angles for both symmetrical 4
h n = [cos (nα1 ) − cos (nα2 ) + cos (nα3 )]. (5.b)
and unsymmetrical MLI configurations regardless of the MLI π
output waveforms number of levels. The advantages of the It could be noticed from (5.a) and (5.b) that the rising edge
proposed algorithm are: it is a mathematical solution, very of the step voltage at the switching angle gives a positive sign
accurate, very simple, and valid for a wide range of mod- for the corresponding cosine term, but the falling edge of the
ulation index and provides voltage control. The remainder step voltage at the switching angle produces a negative sign
of this article is organized as follows. Section II presents for the corresponding cosine term. Thus, (3) can be written in
a detailed explanation of the proposed analytical procedure, a general form as follows [6], [8], [25]:
such as the assumptions and development of the equations. 
4
Section III presents the general solution for the switching hn = ± cos(nαk ), k = 1, 2, 3, . . . , NSmin . (6)
angles algorithm, and the construction of the main matrices is nπ
n=1,3,5,....

Authorized licensed use limited to: National Institute of Technology Patna. Downloaded on December 14,2022 at 18:11:30 UTC from IEEE Xplore. Restrictions apply.
4442 IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 8, NO. 4, DECEMBER 2020

eight switching angles. In this technique, the fundamental

voltage is controlled, and the lower orders of harmonics were
removed. Equation (6) can be written for the lower order of
harmonics as follows:
4 
h3 = 0 = cos(3αk )
3π
k=1,2,3,...,8
4 
= [cos(3αk ) + cos (3αk+1 )]. (7)
3π
k=1,3,5,7
Using algebra formulas of cosines, (7) can be rewritten as
follows:
   
4 3
h3 = 0 = 2 cos (αk + αk+1 )
3π 2
k=1,3,5,7
  
3
× cos (αk − αk+1 ) . (8)
2
From (8) and in order to cancel h 3 and its multiples, the
following condition must be satisfied:
   
3 3
2cos[ (αk + αk+1 )]] = 0 => (αk + αk+1 )
2 2
π
= , k = 1, 3, 5, 7. (9)
2
Decomposing (9) yields
π
α1 + α2 = (10.a)
3
π
α3 + α4 = (10.b)
3
π
α5 + α6 = (10.c)
3
π
α7 + α8 = . (10.d)
3
To cancel out the fifth harmonic, (8) is rewritten once more
for h 5 = 0 as follows:
   
4 5
h5 = 0 = 2 cos (αk + αk+1 )
5π 2
k=1,3,5,7
  
5
× cos (αk − αk+1 )] . (11)
2
Substituting from (9) and (10) into (11) to omit the first
term, then, (11) becomes
   
4 5 
h5 = 0 = 2 cos (cos (αk − αk+1 ) . (12)
5π 2
k=1,3,5,7
Applying the cosines formulas, (12) can be rewritten as
follows:
  
4  5
h5 = 0 = 2 cos (cos (αk −αk+1+α k+2 −αk+3)
5π 4
k=1,5
   
5
Fig. 1. (a) Single-phase N -level inverter. (b) General level waveform. × cos cos αk − αk+1 −α k+2 + αk+3 . (13)
(c) Five-level waveform with NS = 3. (d) Three-level waveform with three
4
switching angles (NS = 3). Similarly, from (13) and in order to cancel h 5 and its
multiples, the following condition must be satisfied:
  
Equation (6) is the final general harmonic equation that 4M  5
depends on the switching angles and their number (NSmin ), h5 = 0 = 2 cos (cos(αk −αk+1+α k+2 −αk+3 ) .
and therefore, it is dependent on the normalized input voltage 5π 4
k=1,5
(M). The proposed technique will be explained in detail for (14)

Authorized licensed use limited to: National Institute of Technology Patna. Downloaded on December 14,2022 at 18:11:30 UTC from IEEE Xplore. Restrictions apply.
AHMED et al.: GENERAL MATHEMATICAL SOLUTION FOR SELECTIVE HARMONIC ELIMINATION 4443

Decomposing (13) yields Substituting from (15) into (21) and after mathematical
processing, it yields
2π    
α1 − α2 +α 3 − α4 = (15.a) 1
5 cos αk − αk+1 −α k+2 + αk+3
2π 4
α5 − α6 +α 7 − α8 = (15.b) k=1,5
      
5 π π
= πh 1 /4 / 4cos cos . (22)
Equation (14) is rewritten to cancel out the seventh 6 10
harmonic (h 7 = 0) as follows: Then, (22) becomes
  
   1
4  7 2 cos (αk − αk+1 +α k+2 − αk+3 +α k+4
h7 = 0 = 2 cos (cos (αk − αk+1 +α k+2 − αk+3) . 8
7π 4 k=1

k=1,5
(16) − αk+5 +α k+6 − αk+7 )

By applying the cosine formulas, (16) can be rewritten as × cos[(1/8)(αk − αk+1 +α k+2
follows: − αk+3 +α k+4 − αk+5 +α k+6 − αk+7 )]
 π π
4  = (πh 1 /4)/ 4 cos cos . (23)
h7 = 0 = 2 6 10
7π
k=1
  Substituting from (15) into (21) and after mathematical
7 processing, it yields
× cos (cos(αk − αk+1 +α k+2   
8 1
 cos (α1 − α2 +α 3 − α4 +α 5 − α6 +α 7 − α8 )
− αk+3 +α k+4 − αk+5 +α k+6 − αk+7 ) 8
  
π π π
× cos[(7/8)(cos(αk − αk+1 +α k+2 − αk+3 +α k+4 = (πh 1 /4)/ 8 cos cos cos .
6 10 14
− αk+5 +α k+6 − αk+7 )]. (17) Thus

Similarly, from (17) and in order to cancel h 7 and its α1 − α2 +α 3

− α4 +α5 − α6 +α 7 − α8 
πh π π π
= 8cos−1
multiples, the following condition must be satisfied: 1
/ 8 cos cos cos .
4 6 10 14
4π (24)
α1 − α2 +α 3 − α4 +α 5 − α6 +α 7 − α8 = . (18)
7
Equations (10), (15), (18), and (24) represent a linear
To control the fundamental value, (7) is rewritten as follows: 8×8 equations with eight unknowns α1 , α2 , α3 , α4 , α5 , α6 , α7 ,
   and α8 , which can be solved mathematically as follows:
4  1
h1 = 2 cos (αk + αk+1 ) AX = B => X = A−1 B. (25)
π 2
k=1,3,5,7
   Matrix A and Matrix B equal
1 ⎛ ⎞
× cos (αk − αk+1 ) . (19) 1 1 0 0 0 0 0 0
2
⎜0 0 1 1 0 0 0 0 ⎟
⎜ ⎟
Substituting from (1) into (19) and after mathematical ⎜0 0 0 0 1 1 0 0 ⎟
⎜ ⎟
processing it yields ⎜0 0 0 0 0 0 1 1 ⎟
A=⎜ ⎜ ⎟
       ⎜ 1 −1 1 −1 0 0 0 0 ⎟⎟
4 1 1 ⎜0
h 1 = cos (α1 + α2) 2cos (αk −αk+1 ) . ⎜ 0 0 0 1 −1 1 −1 ⎟⎟
π 2 2 ⎝ 1 −1 −1 1 1 −1 −1 1 ⎠
k=1,3,5,7
1 −1 −1 1 −1 1 1 −1
Thus ⎛ ⎞ ⎛ ⎞ ⎛ ⎞
B1 π/3 α1
       ⎜ B1 ⎟ ⎜ π/3 ⎟ ⎜ α2 ⎟
1 π ⎜ ⎟ ⎜ ⎟ ⎜ ⎟
2 cos (αk −αk+1 ) = (πh 1 /4)/ 2cos . ⎜ B1 ⎟ ⎜ π/3 ⎟ ⎜ α3 ⎟
2 6 ⎜ ⎟ ⎜ ⎟ ⎜ ⎟
k=1,3,5,7 ⎜ B1 ⎟ ⎜ π/3 ⎟ ⎜ ⎟
B=⎜ ⎟=⎜ ⎟ , X = ⎜ α4 ⎟ (26)
(20) ⎜ B2 ⎟ ⎜ 2π/5 ⎟ ⎜ α5 ⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜ B2 ⎟ ⎜ 2π/5 ⎟ ⎜ α6 ⎟
Then, (20) becomes ⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎝ B3 ⎠ ⎝ 4π/7 ⎠ ⎝ α7 ⎠
  
1 B4 B4 α8
2 cos (αk − αk+1 +α k+2 − αk+3 )   
4 −1 πh 1 π π π
k=1,5
   B4 = 8cos / 8 cos cos cos
4 6 10 14
1         
× cos (αk − αk+1 −α k+2 + αk+3 ) πh B B B3
= 8cos−1
1 1 2

4 / 8∗cos 1 cos 2 cos 3 .
π 4 2 2 2
= (πh 1 /4)/ 2 cos . (21) (27)
6

Authorized licensed use limited to: National Institute of Technology Patna. Downloaded on December 14,2022 at 18:11:30 UTC from IEEE Xplore. Restrictions apply.
4444 IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 8, NO. 4, DECEMBER 2020

The number 8 in (27) represents the number of switching

angles (NSmin ). The modulation index (mi) is defined in terms
of the normalized fundamental component to the normalized
dc input as follows:
H1 h1
mi = = (28)
M V DC1 M
where H1 is the fundamental component of the required output
voltage. Substituting from (28) into (27), (27) can be rewritten
in a general form as

B4 = NSmin ∗ cos−1 [c ∗ NSmin ∗ Mmi ] (29) Fig. 2. Building Row1(4) in Group1.

where c is a certain constant.

Using the general equation for B4 to solve (25) and (24) 6) And so on the generated equations can cancel harmonics
gives the general solution for the values of switching angle as h 3 , h 5 , h 7 , h 11 , . . . . and their multiples based on the
follows: number of groups.
7) The last group of equation is a single equation, and it
αk = c1k + c2k cos−1 [c3k ∗ NSmin ∗ Mmi ]
is devoted to control the fundamental component h 1 .
k = 1, 2, 3, . . . ,NSmin (30) 8) α1 < α2 < α3 < · · · < αk < π/2, k = 1, 2, 3, 4, . . . .
9) If |αk | < π/2==> αk = |αk |, this means that the
where c1k , c2k , and c3k are certain constants (30).
voltage step at this angle is increased from the lower
Equation (30) gives the direct closed-form relationship
level to the next higher level.
between the switching angles and the modulation index for the
10) If |αk | > π/2 ==> αk = π − |αk | , this means that
calculated NSmin value that is suitable for the online operation.
the voltage step at this angle is decreased from a higher
Moreover, the proposed algorithm has the merit of can-
level to the next lower level.
celing more harmonic orders for the same number of inputs
11) The number (M) must be less or equal than the number
(S, M, k1 , and NSmin ) by introducing extra switching angles
of switching angles (NS).
(notches), thanks to the notches in the output voltage wave-
12) All elements of matrix A and matrix B except the last
form. Then, the total number of switching angles is calculated
element of matrix B are independent of the modulation
as
index. Therefore, they are calculated only single time.
NS = 2k NSmin . (31) Thus, the proposed technique has reduced computation
complexity because it needs only to update the last
For example, when S = 2 and k1 = 2, thus, from (1), element of matrix B and this is shown in (29).
M = 3, and from (4),NSmin = 4, and then, NS can be 4, 8,
16,...; the number of the eliminated harmonics “l” is defined III. G ENERAL S OLUTION OF S WITCHING
as A NGLES A LGORITHM
ln (NS) Based on the aforementioned analysis, the proposed tech-
l= . (32) nique can be extended to any number of levels for symmetrical
ln (2)
and unsymmetrical cascaded MLIs. Building of matrix A and
“l + 1” defines the number of equation groups. For the matrix B in a general form is the main part of the proposed
above-mentioned case (26), NS = 8, one can notice that there analysis. The following summarizing steps explain the general
are four groups as depicts in matrix B elements as B1 , B2 , B3 , algorithm.
and B4 . 1) First, M is calculated, from (1), based on the number of
Generally, there are some observations for the proposed the available dc input sources and their ratio.
technique, which are summarized in the following points. 2) Second, a suitable value for the switching angles NS is
1) Matrix A has NS × NS elements and each element is calculated based on (31).
either 0, 1, or −1, while Matrix B has NS × 1 elements. 3) The generated set of groups of equations is defined
2) The first group of the generated equations has by (32).
NS/2 equations and they cancel the third harmonics h 3 4) The first group elements have either 0 or 1. Fig. 2 shows
and its multiples. how to generate the non-zero values at any row inside
3) The second group of the generated equations has the first group: each row has two non-zero values, their
NS/4 equations and they cancel the fifth harmonics h 5 coordinates can be found by multiplying the row number
and its multiples. by the value of “2” that is then subtracted by “1,” and
4) The third group of the generated equations has the next element coordination is defined by multiplying
NS/8 equations and they cancel the seventh harmonics the row number by the value of “2,” as shown in Fig. 2.
h 7 and its multiples. 5) All matrix B elements values inside this first group are
5) The overall groups of equations are (1 +l). equal to b1 = 20 π/3, where “1” represents the group

Authorized licensed use limited to: National Institute of Technology Patna. Downloaded on December 14,2022 at 18:11:30 UTC from IEEE Xplore. Restrictions apply.
AHMED et al.: GENERAL MATHEMATICAL SOLUTION FOR SELECTIVE HARMONIC ELIMINATION 4445

Fig. 3. Building Row2(1) in Group2.

number, while “3” represents the harmonic order that

will be removed.
6) Matrix A elements values at any group (from Group2 to
the last Group) can be generated from its previous group
elements. For example, the elements of group 2 are
generated from group 1 elements. And the elements of
group 3 are generated from the elements of group 2 and
so on.
7) Fig. 3 shows how to generate the elements values of
matrix A; for example, to build Row2(1), “2” refers to
the group number and “1” refers to the row number
inside the corresponding group. While all matrix B
elements values inside this group are equal to b2 = Fig. 4. Flowchart for the general solution of the proposed technique.
21 π/5, where “2” represents the group number, while
“5” represents the harmonic order that will be removed.
Therefore
8) All elements of matrix B,except the last element, can  
be calculated using bl = 2(l−1) π/ h n , where “l” is the π −1 π Mmi
α1 = + cos ∘ (34.a)
6
group number and “h n ” refers to the harmonic order that  4 3
will be canceled by this group of equations. π π Mmi
α2 = −cos−1 ∘ . (34.b)
9) The last group has always single row, which is responsi- 6 4 3
ble for controlling the fundamental component h 1 . It is It should be noticed that M is less or equal NS.
generated in a similar way as shown in Fig. 3. Substituting the calculated elements of matrix A and
10) By following the above-mentioned steps, Table I demon- matrix B in (23) and after manipulations, it yields the general
strates the above-mentioned general summarizing steps closed-form solutions in terms of mi for this MLI, which has
of algorithm. Tables II–IV demonstrate the examples for NS = 2, l = 1, and M = 4 as follows:
building elements of both matrices A and B for = 2, 4, ⎡π ⎤
and 16, respectively. ⎡ ⎤ ⎡ ⎤
α1 0.5 0 0.25 0.25 ⎢3 ⎥
Fig. 4 shows a block diagram to build the elements of both ⎢π ⎥
⎢α2 ⎥ ⎢ ⎥ ⎢ ⎥
matrix A and matrix B for a general number of switching X=⎢ ⎥ = A−1 B = ⎢ 0.5 0 − 0.25 − 0.25 ⎥ ⎢ 3 ⎥.
angles. The proposed algorithm receives the data of the input
⎣α3 ⎦ ⎣ 0 0.5 0.25 − 0.25 ⎢ 2π ⎥
⎦ ⎢
⎥
α4 0 0.5 − 0.25 0.25 ⎣ ⎦
dc sources (S and k1 ) and the data of the outputs (mi or h 1 ). 5
It could be noticed that only the last element in matrix B needs b3
to be updated when the modulation index or the fundamental Therefore
 
component (mi or h 1 ) changes. While all other elements of π π πM
both matrix A and matrix B are constants. Also there are no α1 = + + cos−1  mi  π  (35.a)
6 10 16 cos π6 cos 10
loops or iteration with initial variables guess, resulting in a  
reduced computation complexity π π πM
Substituting the calculated elements of matrix A and matrix α2 = − − cos−1  π mi  π  (35.b)
6 10 16 cos 6 cos 10
B in (22) and after manipulations, it yields the general closed-  
form solutions in terms of mi for this MLI, which has π π πM
α3 = + − cos−1  π mi  π  (35.c)
NS = 2 and l = 1 as follows: 6 10 16 cos 6 cos 10
    π   
α1 −1 0.5 0.5 π π π Mmi
X= =A B= 3 . α4 = − + cos−1   π  . (35.d)
α2 0.5 −0.5 b2 6 10 16 cos π6 cos 10

Authorized licensed use limited to: National Institute of Technology Patna. Downloaded on December 14,2022 at 18:11:30 UTC from IEEE Xplore. Restrictions apply.
4446 IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 8, NO. 4, DECEMBER 2020

TABLE I
G ENERAL A LGORITHM (ROW 2 (1): “2” R EFERS TO G ROUP N UMBER AND “1” R EFERS TO ROW N UMBER I NSIDE THE G ROUP )

It should be noticed that M is less or equal NS = 4. by using MATLAB/SIMULINK software. The experimental
Moreover, (34) and (35) are matched with the general closed- results are done using MOSFET IRF640 as a switch, optoiso-
form solution given in (30). later 6N137 in the gate drive, and the code is built and run
using microcontroller PIC16F917. Two main studied cases
IV. R ESULTS AND D ISCUSSION have been adopted to validate the proposed concept as follows.
The proposed technique is validated through both simulation Case Study I: Two cells (S = 2 and k1 = 2) of the
and experimental results. The simulation results are carried out asymmetrical cascaded MLI shown in Fig. 1 are used with

Authorized licensed use limited to: National Institute of Technology Patna. Downloaded on December 14,2022 at 18:11:30 UTC from IEEE Xplore. Restrictions apply.
AHMED et al.: GENERAL MATHEMATICAL SOLUTION FOR SELECTIVE HARMONIC ELIMINATION 4447

Fig. 5. Harmonic analysis simulation results using a two-cell asymmetrical cascaded inverter with a ratio V DC1 : V DC2 = 1 : 2 for (a) NS = 4 and
(b) NS = 16.

TABLE II Fig. 6 shows both the simulation and experimental results of

B UILDING E LEMENTS OF B OTH M ATRICES A AND B the output voltage waveforms and the corresponding harmonic
FOR NS = 2 AND l = 1
spectrum using a number of switching angles NS = 4 for two
cases of modulation indexes 0.25 and 1.03, respectively. These
two cases of modulation indexes, 0.25 and 1.03, generate
three-level and five-level output voltage waveforms, respec-
tively. From (2), N = 7, which means that the output voltage
has up to seven levels (Vdc1 , Vdc2 , Vdc1 + Vdc2 , 0, −V dc1 ,
−Vdc2 , and −V dc1 − Vdc2 ) based on the modulation index
values. For a very low value of mi ≤ 0.25, only Vdc1 is used to
generate the required output voltage. While in the case of high
values of mi and over modulation, both Vdc1 and Vdc2 are used
to generate the output voltage. When NS = 4, thus, from (31),
there are three groups of equations (l+1= 3). Group 1 will
cancel the third harmonics and its multiples; group 2 cancels
the fifth harmonics and its multiples in addition to group 3
to control the fundamental components. It could be noticed
that in all the two cases, the third and fifth harmonics are
canceled, and the harmonics starts from the seventh order and
this is matched with the aforementioned analysis provided in
Sections II and III.
Fig. 7 shows both the simulation and experimental results of
Vdc1 = 9 V. Therefore, this inverter can generate up to seven the output voltage waveforms and the corresponding harmonic
levels in the output voltage. Based on the aforementioned spectrum using a number of switching angles NS = 16 for the
analysis provided in Sections II and III, N = 7 and M = 3; two cases of modulation indexes 0.25 and 0.83, respectively.
therefore, NS = 4, 8, 16, . . . , and so on. From (31), NS = 4 These two cases of modulation indexes generate five-level
and NS = 8 are the two values that have been selected for and seven-level output voltage waveforms, respectively. When
both the simulation and experimental results, and these results NS = 16, thus, from (31), there are five groups of equations
are shown in Figs. 5–7. (l+1= 5). Group 1 cancels the 3rd harmonics and its mul-
Fig. 5(a) and (b) shows the simulation results for the per- tiples, group 2 cancels the 5th harmonic and its multiples,
centage of harmonic analysis for a wide range of modulation group 3 cancels the 7th harmonics and its multiples, and
index for this case study I. In Fig. 5(a), NS = 4, and thus, group 4 cancels the 11th harmonics and its multiples in
there are three groups of equations (l + 1 = 3), as explained addition to group 5 to control the fundamental components.
before in Sections II and III, which can cancel the third and It could be noticed that in all the two cases, the 3rd, 5th,
fifth harmonics and their multiples besides another equation to 7th, and 11th harmonics are canceled. And the harmonics
control the fundamental component. It is clear in Fig. 5(a) that starts from the 13th order and this is also matched with the
the third and fifth harmonics have been canceled for all values aforementioned analysis provided in Sections II and III.
of mi. In Fig. 5(b) = 16, there are five groups of equations Case Study II: Three cells (S = 3 and k1 = 3) of the
(l +1 = 5) that can cancel the 3rd, 5th, 7th, and 11th harmonic asymmetrical cascaded MLI shown in Fig. 1 are used with
orders besides another equation to control the fundamental Vdc1 = 5 V. Therefore, this inverter can generate up to
components. Fig. 5(b) shows that the 3rd, 5th, 7th, and 11th 27 levels in the output voltage. Based on the aforementioned
harmonics have been vanished for all values of mi. analysis provided in Sections II and III, N = 27 and M = 13.

Authorized licensed use limited to: National Institute of Technology Patna. Downloaded on December 14,2022 at 18:11:30 UTC from IEEE Xplore. Restrictions apply.
4448 IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 8, NO. 4, DECEMBER 2020

Fig. 6. MLI output voltage performance using a two-cell asymmetrical cascaded inverter with a ratio V DC1 :V DC2 = 1 : 2 for NS = 4. (a) Simulation
result of output voltage for mi = 0.25. (b) Experimental result of output voltage for mi = 0.25. (c) Simulation result of harmonic spectrum for mi = 0.25.
(d) Simulation result of output voltage for mi = 1.03. (e) Experimental result of output voltage for mi = 1.03. (f) Simulation result of harmonic spectrum for
mi = 1.03.

From the constraints relating to NS, a reasonable values for groups of equations (l+1 = 6) that can cancel 3rd, 5th, 7th,
NS = 16, 32, 64, . . . and so on. The number of equation 11th, and 13th harmonics in addition to the last equation that
groups “l” can be calculated from (31). There are two values can control the fundamental component. Fig. 8(b) shows that
for NS that are selected, and they are 16 and 32 for both 3rd, 5th, 7th, 11th, and 13th harmonics have been vanished
the simulation and experimental results and these results are for all values of mi.
shown in Figs. 8–10. Fig. 9 shows both the simulation and experimental results of
Fig. 8(a) and (b) shows the simulation results for the per- the output voltage waveforms and the corresponding harmonic
centage of harmonic analysis for a wide range of modulation spectrum using a number of switching angles NS = 16 for the
index. In Fig. 8(a), NS = 16, thus, there are five groups of two cases of modulation indexes 0.25 and 0.95, respectively.
equations (l+1= 5), as explained before in Sections II and III, When NS = 16, thus, from (31), there are five groups of
which can cancel the 3rd, 5th, 7th, and 11th harmonics and equations (l+1= 5). Group 1 cancels the 3rd harmonics
their multiples in addition to the fifth group that can control and its multiples, group 2 cancels the 5th harmonics and its
the fundamental component. It is clear in Fig. 8(a) that the multiples, group 3 cancels the 7th harmonics and its multiples,
3rd, 5th, 7th, and 11th harmonics have been disappeared for and group 4 cancels the 11th harmonics and its multiples in
all values of mi. In Fig. 8(b), NS = 32, thus, there are six addition to group 5 to control the fundamental components.

Authorized licensed use limited to: National Institute of Technology Patna. Downloaded on December 14,2022 at 18:11:30 UTC from IEEE Xplore. Restrictions apply.
AHMED et al.: GENERAL MATHEMATICAL SOLUTION FOR SELECTIVE HARMONIC ELIMINATION 4449

Fig. 7. MLI output voltage performance using a two-cell asymmetrical cascaded inverter with a ratio V DC1 :V DC2 = 1 : 2 for NS = 16. (a) Simulation
result of output voltage for mi = 0.25. (b) Experimental result of output voltage for mi = 0.25. (c) Simulation result of harmonic spectrum for mi = 0.25.
(d) Simulation result of output voltage for mi = 0.83. (e) Experimental result of output voltage for mi = 0.83. (f) Simulation result of harmonic spectrum for
mi = 0.83.
It could be noticed that in all the two cases, the 3rd, 5th, It could be noticed that in all the two cases, the 3rd, 5th, 7th,
7th, and 11th harmonics are canceled, and the harmonics 11th, and 13th harmonics are canceled. And the harmonics
starts from the 13th order and this is also matched with the starts from the 17th order and this is also matched with the
aforementioned analysis provided in Sections II and III. aforementioned analysis provided in Sections II and III.
Fig. 10 shows both the simulation and experimental results The percentage of the THD is defined by the following
of the output voltage waveforms and the corresponding har- equation:
monic spectrum using a number of switching angles NS = 32 
for the two cases of modulation indexes 0.25 and 0.95, 2
n=3,5,7,... [h n ]
respectively. Since = 32, thus, there are five groups of THD (%) = . (36)
equations (l+1= 6). Group 1 cancels the 3rd harmonics h1
and its multiples, group 2 cancels the 5th harmonics and its Using the definition of h 1 as given in (25), therefore
multiples, group 3 cancels the 7th harmonics and its multiples,

group 4 cancels the 11th harmonics and its multiples, and 2
group 5 cancels the 13th harmonics and its multiples in n=3,5,7,... [h n ]
THD (%) = . (37)
addition to group 6 to control the fundamental components. Mmi

Authorized licensed use limited to: National Institute of Technology Patna. Downloaded on December 14,2022 at 18:11:30 UTC from IEEE Xplore. Restrictions apply.
4450 IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 8, NO. 4, DECEMBER 2020

Fig. 8. Harmonic analysis simulation results using a three-cell asymmetrical cascaded inverter with a ratio V DC1 : V DC2 : V D3 = 1 : 3 : 9 for (a) NS = 16
and (b) NS = 32.

Fig. 9. MLI output voltage performance using a three-cell asymmetrical cascaded inverter with a ratio V DC1 : V DC2 : V D3 = 1 : 3 : 9 for NS = 16.
(a) Simulation result of output voltage for mi = 0.25. (b) Experimental result of output voltage for mi = 0.25. (c) Simulation result of harmonic spectrum for
mi = 0.25. (d) Simulation result of output voltage for mi = 0.95. (e) Experimental result of output voltage for mi = 0.95. (f) Simulation result of harmonic
spectrum for mi = 0.95.

From (5), (27), and (29), the harmonic components h n have a The dynamic response of the proposed procedure has been
certain function in mi. Therefore, all terms of (36) have certain verified through the simulation results. Fig. 11 shows the
functions in mi. dynamic response of the proposed procedure for two cases

Authorized licensed use limited to: National Institute of Technology Patna. Downloaded on December 14,2022 at 18:11:30 UTC from IEEE Xplore. Restrictions apply.
AHMED et al.: GENERAL MATHEMATICAL SOLUTION FOR SELECTIVE HARMONIC ELIMINATION 4451

Fig. 10. MLI output voltage performance using a three-cell asymmetrical cascaded inverter with a ratio V DC1 : V DC2 : V D3 = 1 : 3 : 9 for NS = 32.
(a) Simulation result of output voltage for mi = 0.25. (b) Experimental result of output voltage for mi = 0.25. (c) Simulation result of harmonic spectrum for
mi = 0.25. (d) Simulation result of output voltage for mi = 0.95. (e) Experimental result of output voltage for mi = 0.95. (f) Simulation result of harmonic
spectrum for mi = 0.95.

of MLI. Fig. 11(a)–(c) is carried out for an MLI with k1 = 2, as shown in Fig. 11(d). The dynamic performance of the MLI
S = 2, and NS = 4, and therefore, l = 2. The dynamic output voltage is shown in Fig. 11(e), which demonstrates
response is done in the modulation index from 0.25 to 1.03, the computation complexity discussion in Section VI. The
as shown in Fig. 11(a). The dynamic performance of the MLI harmonic spectrums of the output voltage for the two cases of
output voltage is shown in Fig. 11(b), which demonstrates mi are shown in Fig. 11(f). Because l = 4, it can be noticed
the computation complexity discussion in Section VI. The that the three lower order harmonics, the third, fifth, and
harmonic spectrums of the output voltage for the two cases seventh harmonics and their multiplies, are canceled regardless
of mi are shown in Fig. 11(c). Because l = 2, it can be of the value of mi.
noticed that the two lower order harmonics, the third and fifth
harmonics and their multiplies, are canceled regardless of the V. N UMERICAL E XAMPLES
value of mi. In these two numerical examples, the output voltage wave-
Fig. 11(d)–(f) is carried out for an MLI with k1 = 3, form in Fig. 6(a) and (d) will be explained in detail, and
S = 3, and NS = 8, and therefore, l = 3. The dynamic moreover, the parameters related to these waveforms will also
response is done in the modulation index from 0.25 to 0.95, be explained in detail.

Authorized licensed use limited to: National Institute of Technology Patna. Downloaded on December 14,2022 at 18:11:30 UTC from IEEE Xplore. Restrictions apply.
4452 IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 8, NO. 4, DECEMBER 2020

Fig. 11. Simulation results of the dynamic response due to step changes in mi at the step times of 25 and 55 ms for two studied cases. Case I: MLI
with k1 = 2, S = 2, NS = 4, and l = 2. (a) modulation index. (b) Output voltage waveform. (c) Harmonics spectrum. Case II: MLI with k1 = 3, S = 3,
NS = 16, and l = 4. (d) Modulation index. (e) Output voltage waveform. (f) Harmonics spectrum.

TABLE III Example 1: This case is related to input parameters; S = 2

E LEMENTS OF B OTH M ATRICES A AND B FOR NS = 4 AND l = 2 and k1 = 2, mi = 0.25 (calculated based on h 1 ).
Then, using (1), (2), and (4) yields M = 3, N = 2M+
1 = 7, and NSmin = 4 (k = 1, 2, 3, and 4), respectively.
From (32), l = 2, and the number of group of equations
is l +1 = 3. Therefore, the third and fifth harmon-
ics are canceled in addition to controlling h 1 . Matrix A
and Matrix B can be calculated by following the general
procedure explained in Section III or from Table III as
follows:
A = [1 1 0 0; 0 0 1 1; 1 −1 1 −1; 1 −1 −1 1]; B = [ π3 ;
(π/3); (2π/5); b3].
From (27) and (28) (NS = 4), b3 can be calculated and it
is found to be b3 = 5.5641.
Thus, the values of matrix “X” from (25) can be found
by multiplying the inversion of matrix A by matrix B, and
thus, the theoretical solutions of X = [2.2288 − 1.1816 −
0.55331.6005].
The same results of the theoretical solutions of “X” can
also be found by directly substituting into (35), mi = 0.25
and M = 3, which results in X = [2.2288 − 1.1816 −
0.55331.6005].
Apply the last decision on the flowchart of Fig. 4 (taking
the absolute values of X, and checking whether the value
is within the correct bound or not (|αk | < π/2) ) yields
X = [π − 2.22881.18160.5533π − 1.6005] = [0.9128(Ck =
−1)1.1816(Ck = 1)0.5533(Ck = 1)1.5411(Ck = −1)].

Authorized licensed use limited to: National Institute of Technology Patna. Downloaded on December 14,2022 at 18:11:30 UTC from IEEE Xplore. Restrictions apply.
AHMED et al.: GENERAL MATHEMATICAL SOLUTION FOR SELECTIVE HARMONIC ELIMINATION 4453

TABLE IV
B UILDING E LEMENTS OF B OTH M ATRICES A AND B FOR NS = 16 AND l = 4

Taking the absolute values of “X” and rearranging them By applying the same procedure as mentioned earlier,
yield X = [α1 = 0.5533(Ck = 1) α2 = 0.9128(Ck = −1) X = [−1.5805 − 0.53330.09500.9522] = [1.5611(Ck = −1)
α3 = 1.1816(Ck = 1) α4 = 1.5411(Ck = −1) ]. −0.5333(Ck = 1) 0.0950(Ck = 1) 0.9522(Ck = 1)].
It could be noticed that Ck = −1 means that the output Rearranging the absolute values of “X” yields X = [α1 =
voltage has a falling edge at this switching angle, while Ck = 1 0.0950(Ck = 1) α2 = 0.5333(Ck = 1) α3 = 0.9522(Ck = 1)
refers to a rising edge output voltage. From the numerical α4 =1.5611 (Ck = −1)].
results, it is clear that the output voltage in this case has rising From the numerical results, it is clear that the output voltage
edge at α1 , falling edge at α2 , rising edge at α3 , and falling in this case has rising edges at α1 , α2 , and α3 and falling edge
edge at α4 , as shown in Fig. 6(a). at α4 , as shown in Fig. 6(d).
Example 2: This case is related to input parameters; S = 2
and k1 = 2, mi = 1.03 (calculated based on h 1 ). VI. C OMPARISON A MONG THE H ARMONIC
Then, using (1), (2), and (4) yields M = 3, N = 2 M+ E LIMINATION P ROCEDURES
1 = 7, and NSmin = 4 (k = 1, 2, 3, and 4), respectively. To highlight the performance of the proposed procedure, it is
From (32), l = 2, and the number of group of equations compared with its main counterpart existence procedures based
is l +1 = 3. Therefore, the third and fifth harmonics are on the generalized methods, MLI configurations, computation
canceled in addition to controlling h 1 . Matrix A and Matrix B complexity, continuity of modulation index, and initial guess
can be calculated by following the general procedure explained of variables, as shown in Table V. In [23], it used the theory of
in Section III or from Table III as follows: resultants based on the trigonometric identities to develop an
A = [1 1 0 0; 0 0 1 1; 1 −1 1 −1; 1 −1 −1 1]; B = equivalent set of polynomial equations from the transcendental
[(π/3); (π/3); (2π/5); b3]. equations. As the number of harmonics to be eliminated is
From (27) and (28) (NS = 4), b4 = 4 acos ((π/4) Mmi / 4 increased, the method suffers from the computation complex-
cos(π/6)cos(π/10)) = 2.9709. ity due to the increasing number of polynomial degree. In [27],
Thus, the theoretical solutions of the switching angles are: it solved the harmonic equations using the classical PI control
X = [1.5805 − 0.53330.09500.9522]. technique. The PI is used to tune the values of switching
The same results of the theoretical solution of “X” for this angles based on an initial guess to a single angle value, and
case study can also be found by directly substituting into (35), finally, the stipulated harmonics are diminished. The initial
mi = 1.03 and M = 3, which results in: X = [1.5805 − guess for the angle leads to generate a single iteration loop
0.53330.09500.9522]. that increases the computation required. Moreover, the method

Authorized licensed use limited to: National Institute of Technology Patna. Downloaded on December 14,2022 at 18:11:30 UTC from IEEE Xplore. Restrictions apply.
4454 IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 8, NO. 4, DECEMBER 2020

TABLE V
C OMPARISON OF THE P ROPOSED A LGORITHM W ITH C OUNTERPART E XISTENCE A LGORITHMS

is limited only to three switching angles. In [28], it proposed changed, which has a reduced computational time. Since the
a solution of the harmonic equations using univariate linear system was linear, the resulted solution was very accurate.
equations. The proposed method generates switching angles The algorithm was validated through both the simulation and
that has a ratio of 0.5 between any switching angle and the experimental results.
next one. The outputs of this procedure are the dc sources The proposed control algorithm has the following
values. The method is simple with reduced computational advantages.
complexity. However, the method is very sensitive to the dc 1) It is generalized for symmetrical or unsymmetrical cas-
sources variations, and also it is not valid for the symmetrical caded MLI.
MLI. The algorithm in [30] proposed a simple, cost-effective, 2) It has the ability to operate at any required levels’
and real-time implementation method to solve the harmonic number (even or odd) for the MLI.
equations. This method is limited to a five-level inverter, and 3) It is simple, cost-effective, reduced computational com-
moreover, the procedure has four calculations loops resulting plexity, as it uses a single loop of calculation and does
in large computation complexity. The proposed method in [31] not require angles initial guesses.
introduced a simple, cost-effective, real-time implementation, 4) It is a real-time implementation.
and reduced computation complexity method to solve the har- 5) It has the ability to control the fundamental voltage by
monic equations. The method received the number of MLIs to having the modulation index in the developed equations/
generate the switching angles and the dc sources values. This blocks.
method is valid only for a particular number of symmetrical
MLI levels, and also it is not applicable for unsymmetrical ACKNOWLEDGMENT
MLI systems. The proposed method in this article proposed The authors would like to acknowledge the financial support
an extension procedure to the method proposed in [31] that received from the Scientific Research Deanship, Taif Univer-
overcomes all the drawbacks explained in this comparison sity, KSA, through Grant No. 1-439–6072.
discussion. The proposed method is simple, cost-effective,
real-time implementation, reduced computational complexity, R EFERENCES
and generalized for symmetrical or unsymmetrical cascaded [1] I. Colak, E. Kabalci, and R. Bayindir, “Review of multilevel voltage
MLI at any levels without the need for using loops of calcu- source inverter topologies and control schemes,” Energy Convers. Man-
lation or angles initial guess. age., vol. 52, no. 2, pp. 1114–1128, Feb. 2011.
[2] S. Debnath, J. Qin, B. Bahrani, M. Saeedifard, and P. Barbosa, “Oper-
ation, control, and applications of the modular multilevel converter:
VII. C ONCLUSION A review,” IEEE Trans. Power Electron., vol. 30, no. 1, pp. 37–53,
In this article, a general mathematical solution using SHE Jan. 2015.
[3] M. Malinowski, K. Gopakumar, J. Rodríguez, and M. A. Pérez,
for symmetrical and unsymmetrical MLIs with switching “A survey on cascaded multilevel inverters,” IEEE Trans. Ind. Electron.,
angles NS, N = maximum available number of output voltage vol. 57, no. 7, pp. 2197–2206, Jul. 2010.
levels and k = number of eliminated harmonics, has been [4] M. S. A. Dahidah, G. Konstantinou, and V. G. Agelidis, “A review of
multilevel selective harmonic elimination PWM: Formulations, solving
introduced to generate MLI switching angles θ1 , θ2 , . . . , θk . algorithms, implementation and applications,” IEEE Trans. Power Elec-
General linear harmonic equations’ system has been formu- tron., vol. 30, no. 8, pp. 4091–4106, Aug. 2015.
lated in the form (AX = B), which has an easy and simple [5] A. M. Amjad and Z. Salam, “A review of soft computing methods
for harmonics elimination PWM for inverters in renewable energy
mathematical solution. A general algorithm was introduced conversion systems,” Renew. Sustain. Energy Rev., vol. 33, pp. 141–153,
and solved, which enables the elimination of k (NS = 2k ) May 2014.
harmonics and their respective odd multiples in addition to [6] M. A. Memon, S. Mekhilef, M. Mubin, and M. Aamir, “Selective har-
monic elimination in inverters using bio-inspired intelligent algorithms
control the fundamental component. Only a single element in for renewable energy conversion applications: A review,” Renew. Sustain.
the matrix B is needed to be updated as the modulation index Energy Rev., vol. 82, pp. 2235–2253, Feb. 2018.

Authorized licensed use limited to: National Institute of Technology Patna. Downloaded on December 14,2022 at 18:11:30 UTC from IEEE Xplore. Restrictions apply.
AHMED et al.: GENERAL MATHEMATICAL SOLUTION FOR SELECTIVE HARMONIC ELIMINATION 4455

[7] P. N. Enjeti, P. D. Ziogas, and J. F. Lindsay, “Programmed PWM [28] P. L. Kamani and M. A. Mulla, “Middle-level SHE pulse-amplitude
techniques to eliminate harmonics: A critical evaluation,” IEEE Trans. modulation for cascaded multilevel inverters,” IEEE Trans. Ind. Elec-
Ind. Appl., vol. 26, no. 2, pp. 302–316, Mar. 1990. tron., vol. 65, no. 3, pp. 2828–2833, Mar. 2018.
[8] M. T. Hagh, H. Taghizadeh, and K. Razi, “Harmonic minimization in [29] C. Buccella, C. Cecati, M. G. Cimoroni, and K. Razi, “Analytical method
multilevel inverters using modified species-based particle swarm opti- for pattern generation in five-level cascaded h-bridge inverter using
mization,” IEEE Trans. Power Electron., vol. 24, no. 10, pp. 2259–2267, selective harmonic elimination,” IEEE Trans. Ind. Electron., vol. 61,
Oct. 2009. no. 11, pp. 5811–5819, Nov. 2014.
[9] Y. Liu, H. Hong, and A. Q. Huang, “Real-time calculation of switching [30] C. Buccella, C. Cecati, M. G. Cimoroni, G. Kulothungan, A. Edpuganti,
angles minimizing THD for multilevel inverters with step modulation,” and A. K. Rathore, “A selective harmonic elimination method for five-
IEEE Trans. Ind. Electron., vol. 56, no. 2, pp. 285–293, Feb. 2009. level converters for distributed generation,” IEEE J. Emerg. Sel. Topics
[10] H. Taghizadeh and M. T. Hagh, “Harmonic elimination of cas- Power Electron., vol. 5, no. 2, pp. 775–783, Jun. 2017.
cade multilevel inverters with nonequal DC sources using particle [31] C. Buccella et al., “Harmonic elimination procedure for cascaded
swarm optimization,” IEEE Trans. Ind. Electron., vol. 57, no. 11, multilevel inverters having a particular even number of dc sources,” in
pp. 3678–3684, Nov. 2010. Proc. 44th Annu. Conf. IEEE Ind. Electron. Soc. (IECON), Washington,
[11] K. Shen et al., “Elimination of harmonics in a modular multilevel DC, USA, Oct. 2018, pp. 1249–1254.
converter using particle swarm optimization-based staircase modulation
strategy,” IEEE Trans. Ind. Electron., vol. 61, no. 10, pp. 5311–5322,
Oct. 2014. Mahrous Ahmed received the B.S. and M.Sc.
[12] A. N. Kumle, S. H. Fathi, F. Jabbarvaziri, M. Jamshidi, and degrees from Assiut University, Assiut, Egypt,
S. S. H. Yazdi, “Application of memetic algorithm for selective har- in 1996 and 2000, respectively, and the Ph.D. degree
monic elimination in multi-level inverters,” IET Power Electron., vol. 8, from the University of Malaya, Kuala Lumpur,
no. 9, pp. 1733–1739, 2015. Malaysia, in 2007, all in electrical engineering.
[13] M. Balasubramonian and V. Rajamani, “Design and real-time imple- Since 2007, he has been an Assistant Professor
mentation of SHEPWM in single-phase inverter using generalized with the Aswan Faculty of Engineering, Aswan
hopfield neural network,” IEEE Trans. Ind. Electron., vol. 61, no. 11, University, Aswan, Egypt. In 2014, he became an
pp. 6327–6336, Nov. 2014. Associate Professor. He is currently an Associate
[14] F. Filho, L. Tolbert, C. Yue, and B. Ozpineci, “Real-time selective Professor with the Faculty of Engineering, Taif Uni-
harmonic minimization for multilevel inverters connected to solar panels versity, Ta’if, Saudi Arabia. He has been awarded
using artificial neural network angle generation,” IEEE Trans. Ind. Appl., more than ten research-funded projects in the field of power electronics
vol. 47, no. 5, pp. 2117–2124, Sep./Oct. 2011. applications. He has published more than 70 papers in international journals
[15] K. S. Neralwar, P. M. Meshram, and V. Borghate, “GA based hybrid and conferences. His current research interests include power conversion
selective harmonic elimination (SHE) technique applied to five-level techniques and power electronics applications.
nested neutral point clamped (NNPC) converter,” in Proc. IEEE 1st
Int. Conf. Power Electron., Intell. Control Energy Syst. (ICPEICES), Mohamed Orabi (SM’08) received the Ph.D.
New Delhi, India, Jul. 2016, pp. 1–6. degree from Kyushu University, Fukuoka, Japan, in
[16] S. S. Lee, B. Chu, N. R. N. Idris, H. H. Goh, and Y. E. Heng, “Switched- 2004.
battery boost-multilevel inverter with GA optimized SHEPWM for He was with Enpirion Inc., Hampton, NJ, USA,
standalone application,” IEEE Trans. Ind. Electron., vol. 63, no. 4, and Altera Corporation, San Jose, CA, USA, from
pp. 2133–2142, Apr. 2016. June 2011 to July 2014. He is currently a Professor
[17] M. A. Memon, S. Mekhilef, and M. Mubin, “Selective harmonic with Aswan University, Aswan, Egypt, where he is
elimination in multilevel inverter using hybrid APSO algorithm,” IET also the Founder and the Director of the Aswan
Power Electron., vol. 11, no. 10, pp. 1673–1680, 2018. Power Electronics Application Research Center. He
[18] M. H. Etesami, N. Farokhnia, and S. H. Fathi, “Colonial competitive has published about 200 papers in international
algorithm development toward harmonic minimization in multilevel conference proceedings and journals. His current
inverters,” IEEE Trans. Ind. Informat., vol. 11, no. 2, pp. 459–466, research interests include high frequency dc–dc converters and power factor
Apr. 2015. correction converters for LEDs and Electric Vehicles applications, integrated
[19] A. Kavousi, B. Vahidi, R. Salehi, M. K. Bakhshizadeh, N. Farokhnia, power management, analysis and modeling of nonlinear circuits, and inverter
and S. H. Fathi, “Application of the bee algorithm for selective har- control and design for renewable energy applications.
monic elimination strategy in multilevel inverters,” IEEE Trans. Power Prof. Orabi is a member of the National University Ranking Committee,
Electron., vol. 27, no. 4, pp. 1689–1696, Apr. 2012. the Secretary of the Energy and Electricity National Specialized Scientific
[20] K. Sundareswaran, K. Jayant, and T. N. Shanavas, “Inverter harmonic Council, a member of the National Committee for Space Research, a member
elimination through a colony of continuously exploring ants,” IEEE of the Upper Egypt Scientific Research Center, Suhag, Egypt, and the
Trans. Ind. Electron., vol. 54, no. 5, pp. 2558–2565, Oct. 2007. Consultant Engineer and Operating Committee Member of several 50MW
[21] S. Barkati, L. Baghli, E. M. Berkouk, and M.-S. Boucherit, “Harmonic PV farm projects in Banban, Egypt. He received the 2002 Excellent Student
elimination in diode-clamped multilevel inverter using evolutionary Award of the IEEE Fukuoka Section and the Best Young Research Award from
algorithms,” Electr. Power Syst. Res., vol. 78, no. 10, pp. 1736–1746, the IEICE Society, Japan, in 2004. He also received the SVU Encouragement
2008. Award in 2009, the State Encouragement Award for Engineering Science in
[22] J. N. Chiasson, L. M. Tolbert, K. J. McKenzie, and Z. Du, “Control of 2010, and the State Excellence Award for Engineering Science, 2018. He is the
a multilevel converter using resultant theory,” IEEE Trans. Control Syst. Chair of the IEEE “Power Electronics” Egypt Chapter. He is a Guest Editor
Technol., vol. 11, no. 3, pp. 345–354, May 2003. of the IEEE J OURNAL OF E MERGING AND S ELECTED T OPICS IN P OWER
[23] J. N. Chiasson, L. M. Tolbert, K. J. McKenzie, and Z. Du, “Elimination E LECTRONICS .
of harmonics in a multilevel converter using the theory of symmetric
polynomials and resultants,” IEEE Trans. Control Syst. Technol., vol. 13, Sherif S. M. Ghoneim received the B.Sc. and M.Sc.
no. 2, pp. 216–223, Mar. 2005. degrees from the Faculty of Engineering at Shoubra,
[24] K. Yang, Q. Zhang, R. Yuan, W. Yu, J. Yuan, and J. Wang, “Selective Zagazig University, Zagazig, Egypt, in 1994 and
harmonic elimination with groebner bases and symmetric polynomi- 2000, respectively, and the Ph.D. degree in electrical
als,” IEEE Trans. Power Electron., vol. 31, no. 4, pp. 2742–2752, power and machines from the Faculty of Engineer-
Apr. 2016. ing, Cairo University, Giza, Egypt, in 2008.
[25] Z. Yuan, R. Yuan, W. Yu, J. Yuan, and J. Wang, “A groebner bases Since 1996, he has been a Teaching Staff with
theory-based method for selective harmonic elimination,” IEEE Trans. the Faculty of Industrial Education, Suez University,
Power Electron., vol. 30, no. 12, pp. 6581–6592, Dec. 2015. Suez, Egypt. From 2005 to 2007, he was a Guest
[26] M. Ahmed, E. Hendawi, and I. B. M. Taha, “Simple analytical solu- Researcher with the Institute of Energy Transport
tion for selective harmonic elimination technique,” IET Electron. Lett., and Storage (ETS), University of Duisburg-Essen,
vol. 52, no. 9, pp. 749–751, 2016. Duisburg, Germany. He joined the Electrical Engineering Department, Faculty
[27] M. Ahmed, A. Sheir, and M. Orabi, “Real-time solution and imple- of Engineering, Taif University, Ta’if, Saudi Arabia, as an Assistant Pro-
mentation of selective harmonic elimination of seven-level multilevel fessor. His current research interests include grounding systems, dissolved
inverter,” IEEE J. Emerg. Sel. Topics Power Electron., vol. 5, no. 4, gases analysis, breakdown in SF6 gas, and artificial intelligent technique
pp. 1700–1709, Dec. 2017. applications.

Authorized licensed use limited to: National Institute of Technology Patna. Downloaded on December 14,2022 at 18:11:30 UTC from IEEE Xplore. Restrictions apply.
4456 IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 8, NO. 4, DECEMBER 2020

Mosleh M. Al-Harthi was born in Ta’if, Saudi Basem Alamri received the B.Sc. degree (Hons.) in
Arabia, in 1966. He received the B.Sc. and M.S. electrical engineering from the King Fahd University
degrees in electronics technology and engineering of Petroleum and Minerals (KFUPM), Dhahran,
from Indiana State University, Terre Haute, IN, Saudi Arabia, the two M.Sc. degrees (Hons.) in
USA, in 1996 and 1997, respectively, and the Ph.D. electrical power systems and sustainable electrical
degree in electrical engineering from the University power from King Abdulaziz University, Jeddah,
of Arkansas, Fayetteville, AR, USA, in 2001. Saudi Arabia, and Brunel University, London, U.K.,
He was an Assistant Professor with the Jeddah in 2007 and 2008, respectively, and the Ph.D. degree
College of Technology, Jeddah, Saudi Arabia, from in electrical power engineering from Brunel Univer-
2001 to 2009. He is currently a Professor with the sity London, Uxbridge, U.K., 2017.
Electrical Engineering Department, Taif University, He is currently an Assistant Professor of electrical
Ta’if, Saudi Arabia, where he is also the Dean of the College of Engineering. engineering with the College of Engineering, Taif University, Ta’if, Saudi
His current research interests include control engineering, electronics, and Arabia. His current research interests include power systems, power quality,
signal processing. power filter design, and smart grids with a particular emphasis on the
integration of renewable energy sources with power grids.
Dr. Alamri is a member of many international and local professional
organizations. He is also a Certified Energy Auditor (CEA) and a Certified
Energy Manager (CEM) by the Association of Energy Engineers (AEE),
USA. He has received many awards and prizes, including a certificate from
Advance Electronics Company (AEC) in recognition of outstanding academic
achievement during the B.Sc. Program at KFUPM. He also received the
National Grid (NG) Prize, the power grid operator in U.K., for being the top
distinction student for the M.Sc. of SEP Program, Bunel University London.

Saad Mekhilef (SM’11) has been a Professor with

the Department of Electrical Engineering, Univer-
sity of Malaya, Kuala Lumpur, Malaysia, since
Farhan A. Salem was born in Amman, Jordan. June 1999, where he is currently the Director of the
He received the M.Sc. degree in mechatronics of Power Electronics and Renewable Energy Research
production systems from the Moscow State Acad- Laboratory (PEARL). He has authored or coauthored
emy of Instrument Making and Information Tech- more than 300 publications in international journals
nology, Moscow, Russia, in 1996, and the Ph.D. and proceedings. He is actively involved in industrial
degree in production and manufacturing/mechanical consultancy for major corporations in the power
and physico-technical processing machines and tools electronics projects. His current research interests
from the Russian State University, Moscow, in 2000. include power conversion techniques, control of
He is currently an Associate Professor with power converters, renewable energy, and energy efficiency.
the Industrial Engineering Program, Department Dr. Mekhilef is an IET Fellow. He is an Associate Editor of the IEEE
of Mechanical Engineering, Taif University, Ta’if, T RANSACTIONS ON P OWER E LECTRONICS and the Journal of Power
Saudi Arabia. Electronics.

Authorized licensed use limited to: National Institute of Technology Patna. Downloaded on December 14,2022 at 18:11:30 UTC from IEEE Xplore. Restrictions apply.