Harmonic Optimization for the Multi-stage
Saturable Magnetically Controlled Reactor Using
Particle Swarm Optimization Algorithm
Xuxuan Chen, Jianxun Chen
School of information science and engineering
Wuhan University of Science and Technology
Wuhan, China
chenxx82@gmail.com,cjxwh@wust.edu.cn
Abstract---- Magnetically controlled reactors (MCRs) as shunt
reactors have offered flexible ways regulating the reactive power
in the power system. MCRs can be implemented in ultra high
voltage power systems and are much more economical and
operational than the thyristor controlled reactors (TCRs).
However, one remarkable defect of the MCRs is that they will
inject great harmonics into the power system. In this paper, the
structure and the mathematical model of a harmonic free multi-
stage saturable MCR (MSMCR) are proposed. This paper
illustrated the use of the particle swarm optimization (PSO)
algorithm to decide the areas of the stages based on the
harmonics mathematical model of the MSMCR. Since the 3rd
harmonic will not be injected to the power systems if delta
connected symmetrically, the current harmonics injected to the
power system can be limited to as low as 0.8% of the rated output
current of the MSMCR, theoretically. A simulation is finally
produced to verify the research in PSCAD/EMTDC.
Keywords: magnetization characteristic, harmonic analysis,
magnetically controlled reactor (MCR), magnetic variable
control, saturable reactor, PSCAD/EMTDC.
I. INTRODUCTION
N owadays, there already exist several methods to control
the reactive power in the power system such as static var
compensators (SVC) and the static synchronous
compensators (STATCOM). The most common SVCs are
thyristor controlled reactors (TCR) [1]. And MCR is a new type
SVC of which the working principle is similar to TCR[2]. The
reactive power control of the TCR, MCR and STATCOM is
stepless adjustable and the response time is as fast as
20ms[3][4]. However, the advantage of the MCRs over TCRs
and STATCOM is that MCRs can be connected directly to the
ultra high voltage up to 1000 kV without numerous thyristor
valves in series[4].
MCR are used as controlled shunt reactors for electrical
energy transmission over great distance in Russia, China and
Brazil. However, MCR is a device of transformer type which
suffers from the core saturation. The saturation in the MCR
results in a distortion of the output current and will inject great
harmonics into the unbalanced three-phase power system. The
maximum 3rd and 5th harmonic component of the
conventional MCRs are about 7% and 3% of the fundamental
current component, respectively[5]. It is not permissible to
install the MCR without harmonic filters which lead to the
978-1-4799-4315-9/14/$31.00 2014
c IEEE
Baichao Chen
School of electrical engineering
Wuhan University
Wuhan, China
whgycbc@163.com
increase of the cost.
In order to reduce the harmonics of the MCRs, the core
structure and the mathematical model for harmonics analysis
of a multi-stage saturable MCR (MSMCR) are proposed in
this paper. The cores of the MSMCR are designed to more
than four stages only using the same materials as the normal
power transformers. For each stage, there is a pair of
parameters consist of the area value and the length value. A
mathematical model for harmonics analysis of the MSMCR is
presented based on these pairs of parameters. The problem is
to choose a sequence of these parameters in order to get a
minimal harmonics of the output current of the MSMCR. The
optimization problem can be regarded as a nonlinear min-max
problem. Although classical method can be used to solve the
problem, evolution algorithms still have an advantage on
getting the solver in a much shorter time and reducing the
difficulties in the mathematical analysis. The PSO algorithm
and its variants have been successfully applied to solve many
technical areas on the power systems. But there is still limited
information about harmonics suppression for parameters
design of the electrical equipment. For this reason, the PSO
algorithm is considered in this work as a supplement to the
application on power systems.
The work is showing that the PSO algorithm is able to
choose the parameters of the MSMCR successfully. The peak
values of the current harmonics in single-phase MSMCR are
greatly reduced. The peak values of the 3rd, 5th and 7th
current harmonics are 2.55%, 0.9% and 0.7%, and RMS
values are 1.80%, 0.64%, 0.49%, respectively. The harmonics
are much less than the traditional MCRs. The results are tested
and verified in PSCAD/EMTDC.
II. MODEL AND PARAMETER DESIGN OF MSMCR
A. Structure and Model of the Proposed MSMCR
The MCR is based on the working principle of the
magnetic magnifier. The working principle of the
conventional MCR is detailed in literature [6]. By controlling
the dc current through the control winding, the iron cores of
the MCR are deeply saturated, and the reactance of the MCR
will be changed depends on the saturation degree of the iron
cores.
There is only one stage (magnetic valve) in each iron core
1804
of the conventional MCR. The conventional model is shown
in Fig. 1. Control windings are coupled by winding taps on
each half of the iron cores and thyristors are connected
between the taps. The turn for each control winding is Nk /2.
High rated thyristors is not required because the value of Nk /
N (į) is usually 5%. The wind taps results that the voltages
upon the thyristors are relatively lower. For a 500 kV power
system, there is only 25kV voltage on the thyristors. So the
quantities of the thyristors in series are much fewer than the
Fig. 3. Design of the multi-stage saturable iron cores
STATCOM.
+ H = f ( B) =
­0 B < Bt1
− ° n −1
Ask
°B−
° ¦
k =1 Ab
Btk
° Bt ( n −1) < B < Btn
° μ0
°° n
A (1)
® B − ¦ sk Btk
° k =1 Ab
B > Btn
° μ0
° n −1
° Ask
° B + ¦ A Btk
k =1
° b
− Btn < B < − Bt ( n −1)
Fig. 1. Model of the conventional MCR
°̄ μ0
The reactance of the MCR is regulated by controlling the dc
current magnitude through the control winding which changes where Bt1 and Btn (n=2,3,4…) represent the magnetic flux
the magnetic field strength in the magnetic valves. The density when the first and the nth stage begin to saturate,
equivalent working circuit of the MCR is shown in Fig.2a. respectively.
The commutation circuits when thyristors T1 or T2 is In (1), Bts represent the magnetic flux density when the
conducted are shown in Fig.2b and Fig.2c, respectively. The iron cores begin to saturate. The relationship between Bts and
dc control current ik1 and ik2 can be regulated by changing the Btn can be described in (2),
switching angles of the thyristors. As the switching angle of A
Btn = sn Bts (2)
the thyristors increases, the dc control current in the control Ab
winding decreases and the reactance of the reactor will The analyses of B1 and B2 are the same because the structure
increase. of the MSMCR is symmetrical. Taking B1 in the left iron core
as example, as shown in Fig. 4, if the peak value of B1 is
(N-Nk)/2

greater than the magnetic flux density Btn, the nth stage begins
(N-Nk)/2
(N-Nk)/2

to saturate then. ȕ1 and ȕn represent the saturation degrees of

the first and the nth stage in a power frequency.
(N-Nk)/2

(N-Nk)/2
(N-Nk)/2

Fig. 2. Equivalent circuit of the MCR

In this paper, the iron cores of the MCR are specially
designed. The magnetic valves consist of more than four
stages. The stages are also called the multi-stage saturable
magnetic valves. The purpose of the design is to reduce the
output harmonics of the MCR. A multi-stage saturable iron Fig. 4. Schemetic of the magnetic flux density versus particular saturated
core design is shown in Fig.3. Ab is the area of the iron core, stages
As1 is the area of the first stage, As2 is the area of the second The mathematical model for current harmonics analysis of
stage and Asn is the area of the nth stage. If consider l1= the MSMCR is shown in (3),
l2=...=ln, the B-H characteristic of the iron cores can be
described in (1),

2014 IEEE 9th Conference on Industrial Electronics and Applications (ICIEA) 1805
­βn = 0 β1 ∈ [0, 2 cos−1 (2 − kn )] x11 x12 " x1 j
°
° β = 2 cos−1 (k − 1 + cos β1 ) β ∈ (2 cos −1 (2 − k ), β ] X =
x21 x22 " x2 j (4)
° n n
2
1 n max
%
° n
®i* = 1 [( β n − sin β n ) Ln ] (3) xi1 xi 2 " xij
° 1 2π ¦
° 1 v11 v12 " v1 j
°* 1 n
­ sin(mβ n ) sin[(m + 1) β n ] ½
°i2 m +1 = (2m + 1)π ¦ Ln ® 2m − 2(m + 1) ¾ V=
v21 v22 " v2 j (5)
¯ 1 ¯ ¿
%
(n = 2,3,...7; m = 1, 2,3...)
vi1 vi 2 " vij
where ȕn (n=1,2,3…) is the saturation degree of each stage.
The lager the saturation degree is, the bigger current the Step3: Evaluate the fitness from (4) to get personal best
MSMCR will output. The relationship between ȕn when n=1 P and global best Pg value;
and n>1 is given in the first two expression in (1). This will p11 p12 " p1 j
result in only one variable when the design parameters of the p21 p22 " p2 j (6)
MSMCR are determined. kn (n=2,3,…) is the area ratios of pij (t ) =
%
the nth stages over the first stage which is the smallest stage
in the valves. It means that k1 =1 and kn >1. ȕmax is decided by pi1 pi 2 " pij
the area ratio of the iron core over the first stage. The highest pgj (t ) = pg1 pg 2 " pgj (7)
ratio of the iron core over the smallest stage is ranged from
2.5 to 3 in the MSMCR design. That is, the maximal value of Step4: Update velocity using (8). The velocity of each
ȕmax is 2ʌ when the maximal area ratio is 3. particle depends on the distance of the current position to the
positions that result in good fitness values.
B. PSO Algorism for the Design of the MSMCR vij (t + 1) = ωvij (t ) + c1r1 j (t )[ pij (t ) − xij (t )]
The PSO algorithm is chosen to calculate the areas ratios (8)
+ c2 r2 j (t )[ pgj (t ) − xij (t )]
among these stages and to get the optimization results of the
harmonics performance of the MSMCR. PSO is very effective where t and t+1 refer to the current and the next index of
solving this nonlinear optimization problem [8]. A modified iterations; r1j(t) and r2j(t) are uniform distributed random
PSO algorithm is shown in Fig.5 as a flowchart diagram. numbers between 0 and 1;c1 and c2 are weight of the personal
best position and the global best position, respectively. ω can
be changed linearly from 0.8 to 0.4 during the iteration so that
precise optimal position can be obtained.
Step 5: Update current position using (9) and a new
particle is generated.
X (t + 1) = X (t ) + V (t ) (9)
Step6: Output optimal position Pg or return to step 3 if the
termination condition is not achieve such as iteration limit is
reached.
For current problem, we are interesting in designing the
MSMCR with minimal current harmonics. To achieve this,
from (3), the following function iopt of the MSMCR in relation
to rated current is used to evaluate the fitness value.
­ m
°i = (i2*m +1 ) 2 / (irated
*
) ( m ≥ 3)
° thd ¦
1
°
®imax = β ⊂max (ithd ) (10)
1 [0, β max ]
°
°iopt = min n (imax )
° k1 , k2 ⋅⋅⋅k7 ⊂ (1,3),¦ Ln =1
Fig. 5 A modified PSO algorithm for the MSMCR design ¯ 1

It must take steps as follows using the PSO Algorithm: ‫כ‬ *

Step1: define the fitness function and the solution space;
Step2: For an N-dimensional problem, N=j, the initial
position and velocity can be specified by i × j matrices as
follows where i is the population in the swarm. Each row of
the position X is a possible solution to the optimal problem;
where ‹୰ୟ୲ୣୢ = max(i 1).Total current distortion ithd is defined
in the first expression in (10). For each pairs of sequences kn
and Ln in (3), there is a maximal value imax of the total
harmonics ithd during the regulation of ȕ1 from 0 to ȕmax. iopt is
the minimal value of imax. iopt will also have different values if
ȕmax is taken into consideration. Here, we only consider that
m=3 and ȕmax=2ʌ because the harmonics are much smaller
when the harmonic order is 9 or higher. The parameters Ln is a

1806 2014 IEEE 9th Conference on Industrial Electronics and Applications (ICIEA)
common divisor can be divided by both ݅ଵ‫ כ‬and ‹ଶ୫ାଵ ‫כ‬
when current harmonics are 2.55%, 0.9% and 0.7% of the rated
all lengths of the stages are the same. Accordingly, the output current, and the RMS values are 1.80%, 0.64%, 0.49%,
optimization can be described as a minimax problem which is respectively.Parameter sequences of Kn for the MSMCR are
defined as shown in Table. 1.
FITNESS= min max F{(k1 , k2 ⋅⋅⋅ kn ), β1} (11)
k1 , k2 ⋅⋅⋅k7 ⊂ (1,3) β1 ⊂[0, βmax ]

The mutant strategy of (9) is simple.

Step1: Choose a random number between 0 and 1.It means
the mutant rate is 60%. The function rand( ) choose a random
real number between 0 and 1.
if rand( )>0.4 then
go forth to step 2;
else
jump to step 4;
end if
Step2: Choose the update position. n is the dimension of the
population. The number of N is decided by the stages the Fig. 6 Convergence curve of the fitness value and harmonics of the seven-
MSMCR have. The value of N is 7 in this problem. The stage single-phase MSMCR when the lengths of each stage are the same.
function ceil(A) gets the nearest integers greater than or equal C. The B-H magnetization characteristics calculation
to A. The value got from the function rand( ) here is
independent from the step 1. The B-H magnetization characteristics of the MSMCR is
n=ceil(N*rand( )); calculated using (12) and the curve is shown in Fig.7.
step3: Update current position, j is the size of the popula- TABLE I
AREAS RATIO AMONG THE STAGES IN THE IRON CORES
tion. The function X( ) describes the sequences of Kn. The
function rands(1) get a random real number between -1 and 1. Stages Ratios B-H curve slopes
The parameter s is very important in the mutant strategy. If (Kn =Asn/ Ab) (ȝn)
the value of s is too large, the error of the solution is not 1 1.0000 7.000
tolerable. If the value of s is too small, the iterations to the 2 1.6396 3.500
solution are large and the processing time is long. Here, the 3 1.9874 2.333
value of s is 0.005. 4 2.2766 1.750
X(j,n)=X(j,n)+s*rands(1); 5 2.6270 1.200
%Xmax and Xmin are the constraints of K which is 6 2.8924 1.167
limited to 0.01 and 2. 7 2.9835 1.000
if X(j,n)>Xmax then Iron core 3.0000 1.000
X(j,n)=Xmax;
end if ȝ7 equals ȝ0 which is the magnetic permeability in the air.
if X(j,n)<Xmin then Btn − Bt( n−1)
X(j,n)=Xmin; H n = H n −1 + (n ≥ 2) (12)
end if μn −1
step 4:Choose another individual. The initial value of j is 1.
In (12), H1 is equal to zero.
j:=j+1;
if j> the maximal iterations number then
stop the process;
else
back to step 1;
end if
Finally, the minimal value of iopt was obtained by using a
swarm of population, j, of 50 particles. A number of iterations
are large enough to guarantee the convergence of the PSO
algorithm to the desired solution. The example illustrates the
optimal total harmonics of the MSMCR using the fitness
function given in (11). The convergence curve for the seven-
stage MSMCR obtained using the PSO algorithm is shown in
Fig. 6a. As shown in Fig. 6b, the 3rd, 5th and 7th harmonics Fig.7. The B-H characteristic of the stages in iron cores
related to the saturation degree ȕ1 is displayed. From the
figure, we can observe that the current THD mainly consists of
the 3rd harmonic. The peak values of the 3rd, 5th and 7th

2014 IEEE 9th Conference on Industrial Electronics and Applications (ICIEA) 1807
 III. SIMULATIONS OF THE SINGLE PHASE MSMCR IN TABLE II
SATURATION CURVE CONFIGURATION IN UMEC TRANSFORMER
PSCAD/EMTDC
Point Current (%) Voltage (pu)
A. The Circuit of the Single Phase MSMCR 1 0 0
In order to verify the results of the theoretical harmonics 2 0.1 1
analyses in section II, a transformer module which can 3 9.14 1.6396
simulate the saturation characteristics of the transformers can 4 19.08 1.9871
fulfill the simulation requirement. In this paper, the single- 5 31.47 2.2766
phase MSMCR is modeled using two UMEC transformers
6 51.49 2.6270
model in PSCAD/EMTDC. The simulation can also be done
7 70.45 2.8924
in MATLAB/Simulink or ATPdraw/EMTP, etc.
The primary windings of the UMEC transformers are 8 78.26 2.9835
connected with different polarity and secondary windings are 9 178.26 3.98935
connected with the same polarity. The dc current controller is 10 278.26 4.9835
composed of a single-phase full-bridge controlled rectifier.
The source of the rectifier derives from the ideal transformer
of which the tap ratio is 20/1. As the switching angle of the
thyristors increasing, the dc control current in the control
winding decreases and the reactance of the reactor will
increase. Complex controller or control algorithm is not
required for this device. The circuit design is shown in Fig.8.

Fig. 9. The flux waveforms of the UMEC transformers

The output current and the RMS value of the MSMCR are
shown in Fig.10. The peak and RMS value are 5.05 A and
3.512 A, respectively.

Fig. 8. The Circuit of the single phase MSMCR

B. UMEC Transformer Configurations

The configurations of two UMEC transformers are the
same. The primary voltage (RMS) and secondary voltage
(RMS) are both 220 V which is a half of the source voltage of
the ac side. The apparent power rating of the UMEC
transformers are all 2 kVA. The ratio of core yoke length to
the core winding-limb length and the ratio of core yoke area to
the core winding-limb area are 1.6 and 1, respectively.
The current harmonics of the ac side can be reduced by Fig. 10. The output current and the RMS value of the MSMCR
configuring the saturation characteristics of the UMEC The 3rd, 5th, 7th and 11th harmonics of the output current
transformers. From Table I, the saturation curve in UMEC are shown in Fig.11. The waveform is similar to Fig. 6.The
transformers can be calculated which is shown in Table II. In 3rd harmonic is the largest in all harmonics and the peak RMS
this table, the gradients of the lines after point 8 are the same. value is 0.0956 A which is 1.89% of the peak value of the
Actually, the peak value of the magnetic flux in the primary rated output current. The peak values of the 5th, 7th and 11th
winding can only reach 3 pu. harmonics are 0.032 A, 0.0195 A and 0.013 A which are
0.63%, 0.39% and 0.26% of the peak value of the rated output
C. Simulation Results
current, respectively. The THD of the rated output current is
The stages are saturated one after another when MSMCR 2.45%. The simulation results are very close to the theoretical
begins to output its rated current. Note that, the iron cores of results which are 1.80%, 0.64%, 0.49% for the 3rd, 5th and
the largest areas in the MSMCR will not saturate which results 7th harmonics, respectively.
that the flux in the cores is limited to 3 pu. The limitation can
be configured by setting the value of the resistor Rdc in the
control winding. The flux waveform is shown in Fig.9 when
MSMCR output the rated current.

1808 2014 IEEE 9th Conference on Industrial Electronics and Applications (ICIEA)

the 3rd harmonic
Current Harmonics
the 5th harmonic
the 11th harmonic
the 7th harmonic
Time (Sec)
Fig. 11. The 3rd, 5th, 7th and 11th harmonics of the output current
IV. SIMULATIONS OF THE THREE-PHASE MSMCR IN
PSCAD/EMTDC
The three-phase MSMCR can be either delta or star-
connected to the power system. Because the harmonics results
are the same to the single phase MSMCR if delta connected,
so we focus on the simulations when MSMCR is the delta
connected to the power system in this chapter. The schematic
of the delta connected three-phase MSMCR and its control are
shown in Fig.12
Fig. 12. Schematic of the delta connected three-phase MSMCR
The voltages of the sources are set to 254 V in order to get
the same voltage with the star connect method, while the other
parameters remain the same with the single phase MSMCR.
The output currents (ac operation source currents) and the
RMS value of the MSMCR are shown in Fig.13. The peak and
RMS value are 8.69 A and 6.08 A, respectively.
Fig. 13. The output current and the RMS value of the MSMCR
2014 IEEE 9th Conference on Industrial Electronics and Applications (ICIEA)
The 3rd, 5th, 7th and 11th harmonics of the output current
are shown in Fig.14. The 5th harmonic is the largest in all
harmonics and the peak value is 0.0548 A which is 0.63% of
the peak value of the rated output current. The peak values of
the 3rd, 7th and 11th harmonics are 0.0476 A, 0.0311 A and
0.0206 A which are 0.55%, 0.36% and 0.24% of the peak
value of the rated output current, respectively. The THD of the
rated output current is 0.8%.
Fig. 14. The 3rd, 5th, 7th and 11th harmonics of the output current
V. CONCLUSIONS
The model and simulations of a harmonic free multi-stage
saturable magnetically controlled reactor are proposed in this
paper. MSMCR is designed to solve the harmonics problems
and provide a more flexible way to control the reactive power
in the power system. The peak values of the 3rd, 5th and 7th
current harmonics in single-phase MSMCR are 2.55%, 0.9%
and 0.7%, and RMS values are 1.80%, 0.64%, 0.49%,
respectively. Simulations for single-phase and three-phase
MSMCR systems show that the harmonics are much less than
the TCR and traditional MCR.
The research shows that the MCRs have great potential and
advantages for power quality control and reactive power
compensation and offer much more flexible and economical
ways regulating the reactive power in the power systems.
VI. REFERENCES
[1] R. R. Karymov, M.Ebadian, “Comparison of magnetically controlled
reactor (MCR) and thyristor controlled reactor (TCR) from harmonics
point of view,” Electrical Power and Energy Systems. Vol. 29, pp. 191-
198, Jun. 2006.
[2] M. X. TIAN, Q. LI, and Q. F. LI Eason, “A controllable reactor of
transformer type,” IEEE. Trans. Power Delivery. vol. 19, pp. 1718-
1726, Oct. 2004.
[3] Sing, B., Saha, R. Chandra A. and Al-Haddad, K., “Static synchronous
compensators (STATCOM): a review”, IET. Power Electronics. Vol. 2,
no.4, pp. 297-324, Jul. 2009.
[4] Bryantsev, A., Bryantsev, M., Bazylev, B., et al. Power compensators
based on magnetically controlled shunt reactors in electric networks with
a voltage between 110 kV and 500 kV[C]. Transmission and distribution
conference and exposition, Sao Paulo, Brazil, 2010, pp:239-244.
[5] X. X. Chen, B. C. Chen, C. H. Tian, J. X. Yuan and Y. Z. Liu,
“Modeling and Harmonic Optimization of a Two-Stage Saturable
Magnetically Controlled Reactor for an Arc Suppression Coil”, IEEE.
Trans. Industrial Electronics. vol. 59, pp. 2824-2831, Jul. 2004.
[6] B. C. Chen, J. M. Kokernak, “Thyristor controlled two-stage magnetic-
valve reactor for dynamic var-compensation in electric railway power
supply systems,” APEC2000, Fifteenth annual IEEE. Conf., Louisiana,
USA, 2000.
[7] Eberhart, R.C.; Shi, Y. Guest editorial special issue on particle swarm
optimization [J]. IEEE Transactions on Evolutionary Computation, Vol.
8 , No. 3, pp: 201 – 203. Mar. 2004.
1809