Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2016.2628783, IEEE
Transactions on Power Electronics
IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. X, NO. X, XXXX 20XX 1

A Modeling and Analysis Method for


Fractional-order DC-DC Converters
Xi Chen, Yanfeng Chen, Member, IEEE, Bo Zhang, Senior Member, IEEE, and Dongyuan Qiu, Member, IEEE

Abstract—This paper proposes a modeling and analysis method Therefore, numerical algorithms are widely applied to the
for fractional-order DC-DC converters operating in continuous analysis of fractional-order systems, which may cause an ex-
conduction mode (CCM). As an example, a fractional-order Boost ponential increasing of computational efforts [15], [16]. Unlike
converter is studied in detail. Instead of using fractional calculus,
the method presented here uses a general state vector differential integer-order systems, fractional-order systems do not permit
equation to describe the converter. By combining the principle fractional derivatives to be defined uniformly. Thus, existing
of harmonic balance and equivalent small parameter method numerical methods for finding derivatives are ineffective [17].
(ESP), an approximate analytical steady-state solution of the state Taking the fractional-order DC-DC Boost converter as an
variables could be obtained. Subsequently, the CCM-operating example, this paper presents a time-domain modeling and
criterion of the converter is discussed based on the proposed
method, and the appropriate parameters are provided to make analysis scheme, which is different from the methods reported
sure that the fractional-order converter operates in CCM. In in [18] and [19]. And the main theoretical basis of the scheme
addition, a statistical analysis of harmonics is performed, from is derived from the formula proposed in [20], that is
which the effects of fractional orders on harmonics can be
observed. Moreover, numerical simulations are performed using dµ eωt
= ω µ eωt (1)
the Adams-Bashforth-Moulton typed predictor-corrector method dµ t
and the Oustaloup’s filter-based approximation method. Both where µ is the order of the differential operation, which can be
the DC components and ripples of the state variables obtained
from these two methods are combined with those obtained by an integer or non-integer, and the angular frequency ω can be
the proposed method, and they are in good agreement. Finally, any real number. Equation (1) shows that the differential op-
the fractional-order capacitor and inductor are designed and eration affects only the amplitude of the exponential function,
verified by simulations, and on the basis of which experiments of but has no effect on the angular frequency. Thus, the principle
the fractional-order Boost converter are carried out to further of harmonic balance can be used, and a scheme based on the
verify the proposed method.
ESP method [21] is put forward. According to this scheme,
Index Terms—Fractional-order, DC-DC converters, harmonic the conventional ESP method is further extended to model
balance, approximate analytical steady-state solution. the fractional-order DC-DC converter and acquire its steady
state solution. And the approximate analytical time-domain
I. I NTRODUCTION solution obtained can be simply expressed as the summations
of harmonic contents, on the basis of which the steady-state
ONCEPTS from fractional-order circuits and systems
C have recently attracted much attention from the elec-
trical engineering community [1]. Many novel ideas have
performance of the converter can be analyzed.
This paper is organized as follows: Section II deals with the
modeling work of the fractional-order DC-DC Boost converter
been generated by exploiting concepts of fractional circuits. in CCM. In Section III and Section IV, an equivalent model
For example, fractional-order models of capacitors, inductors, based on the ESP method and the approximate solution are
memristors, and CMOS metamaterial transmission lines are offered, respectively. Afterwards, the order-related phenom-
built [2]–[11]. Practical fractional-order elements are fabri- ena, such as effects of fractional order on the CCM-operating
cated, such as those reported in [12], [13]. In addition, new boundary and the harmonic amplitude of the converter, are
topologies of circuits based on fractional-order elements are discussed in Section V. In Section VI, simulations and exper-
constructed [14], though the underlying characteristics of these iments are performed to verify the proposed method. Finally,
circuits remain elusive. conclusions are given in Section VII.
In general, systems including energy storage components
could be described by integral or differential equations. Be-
II. M ATHEMATICAL MODEL OF A FRACTIONAL - ORDER
cause of their long memory characteristic, fractional-order
DC-DC B OOST CONVERTER IN CCM
calculus equations do not have simple analytical solutions.
The fractional-order DC-DC Boost converter is shown in
Manuscript received April 28, 2016; revised July 18 and September 18, Fig. 1, in which both the inductor L and the capacitor C are
2016; accepted October 31, 2016. fractional-order energy-storage components, the switch ST
The authors Xi Chen, Yanfeng Chen, Bo Zhang and Dongyuan Qiu
are with the School of Electric Power, South China University of Tech- and the diode SD are ideal. Typically, when the converter
nology, Guangzhou 510640, China (e-mail: xichen 1021@hotmail.com; operates in CCM, two switch states can be identified as
eeyfchen@scut.edu.cn; epbzhang@scut.edu.cn; epdyqiu@scut.edu.cn). This State 1: ST on and SD off, for nTs <t≤ (n + D)Ts .
work is supported by National Natural Science Foundation of China under
Grant 51437005, 51077057, and Guangdong Provincial Natural Science State 2: ST off and SD on, for (n + D)Ts <t≤ (n + 1)Ts .
Foundation, China, under Grant 2014A030313247. where n is an integer, Ts is the switching period and D is the

0885-8993 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2016.2628783, IEEE
Transactions on Power Electronics
IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. X, NO. X, XXXX 20XX 2

duty ratio, which is defined as the ratio of the turn-on time    


of ST to the switching period Ts . State 1 and state 2 toggle α 1 1
p 0 −
G1 (pα , pβ ) =  1 L L
periodically in the steady state. A switching function δ(t) is 1  , G2 =  1  (6)
  
β
introduced to represent the switching state, which is defined − p + 0
C RC C
as δ(t) = 1 and δ(t) = 0 for state 1 and state 2 respectively.
It can be seen from (6) that, as to the fractional-order Boost
converter, the orders of derivations can only affect the matrix
G1 , yet have no effects on the matrix G2 .

III. T HE EQUIVALENT MODEL BASED ON THE ESP


METHOD
According to the principle of the ESP method, the state
vector x in (5) can be expanded into the following series
n
X
x = x0 + εi x i (7)
i=1

where the zero order approximation x0 and the ith order


approximation xi are called as the main wave and correction
Fig. 1. Fractional-order DC-DC Boost converter.
respectively [21]. Similarly, δ(t) is also expanded as follows
n
According to the model in [2], the relationships between X
δ(t) = δ0 + εi δi (8)
the voltage and current of the fractional-order components
i=1
are defined as
Note that the parameter εi in (7) and (8) is only a
dα iL symbolic mark introduced temporarily to indicate that the
vL (t) = L α (2a)
dt correction xi is much smaller than the main wave x0 [22],
dβ vC that is εi+1 xi+1  εi xi  x0 . Substituting (7) and (8) into
iC (t) = C (2b) f(x) = δ(t)x, and combining terms with the same order of εi ,
dtβ
one can get the expression for f(x)
Here, α and β are orders of derivations. When α = 1 and
β = 1, the inductor and capacitor are conventional integer-
f = f0 + εf1 + ε2 f2 + . . . (9)
order components. Taking iL and vC as state variables, the
fractional-order Boost converter in state 1 can be described where
by the following equations f0 = δ 0 x 0 (10a)
dα i L 1 f1 = δ0 x1 + δ1 x0 (10b)
= Uin (3a)
dtα L
f2 = δ0 x2 + δ1 x1 + δ2 x0 (10c)
dβ vC vC
+ =0 (3b)
dtβ RC
and so on.
And when operating in state 2, it can be described by
Based on (1), the order of the fractional differential
dα iL 1 1 operation has no effect on the angular frequency of the
+ vC = Uin (4a)
dtα L L exponential function. Then similar to the method described
in [21], each correction xi in (7) can also be chosen as
dβ vC vC 1
β
+ − iL = 0 (4b)
dt RC C X X
xi = xik = ai0 + (aik ejkτ + c.c) (11)
When we use p to represent the differential operator d/dt, k∈Eir
the differential operator dα /dtα and dβ /dtβ can be replaced
in which c.c denotes the conjugate item, k is an integer,
by pα and pβ respectively. By combining (3) and (4) with
and τ = ωt (where ω = 2π/Ts ) is the normalized time.
δ(t), the fractional-order DC-DC converter in CCM can
The spectral content set {E0 } of the state vector x0 , which
be described by the following general vector state variable
consists of a set of numbers representing relative frequencies
differential equation
of harmonics, is determined by the physical sense of the
study object. For example, as DC-DC converters have
G1 (pα , pβ )x + G2 f(x) = U (5)
low-pass filtering properties, the main wave x0 includes the
where the nonlinear function f(x) = δ(t)x, the state vector DC component only, i.e. x0 = a00 , thus, the set {E0 } is
x = [iL vC ]T and the constant vector U = [Uin /L 0]T , here {0}. Similarly, the set {Eir } contains the spectral contents
the subscript T means the transpose of a matrix. The square for the correction xi , yet it is not known in advance and is
matrices G1 (pα , pβ ) and G2 are given by determined in the process of iterated calculations, i.e. it is

0885-8993 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2016.2628783, IEEE
Transactions on Power Electronics
IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. X, NO. X, XXXX 20XX 3

determined by x0 , x1 , . . . , and xi−1 during the calculating x1 , x2 . . . , etc. Thus with sufficient numbers of corrections,
process. The terms ai0 and aik represent the DC component the steady-state solution of (5) can be acquired, which is
and the kth harmonic magnitude of xi respectively. approximated by x0 + x1 + x2 + . . . .
The switching function δ(t) is expanded into Fourier series In theory, under the criterion of minimum mean-square
as error, the Fourier series is the best approximation of the

periodic signal [23]. From the discussion above, we can
δ(t) = b0 +
X
(bk ejkτ + c.c) (12) see that the nonlinear switching function δ(t) is expanded
k=1
into Fourier series, and in fact the expansion (7) of the
RT  state vector x is based on the principle that the periodic
here, b0 = 0 s δ(t)dt /Ts = D, bk = (αk − jβk )/2 and bk function can be represented as Fourier series. Actually, as
is the conjugate complex of bk , in which DC-DC converters have the low-pass filtering property, the
Z Ts magnitudes of harmonics with order higher than 3 are small.
2 It means that, the steady-state solution can be approximated
αk = δ(t) cos(kωt)dt
Ts 0 by finite numbers of corrections xi . Thus based on the
(13)
principle in [23], a tolerance error index is introduced here
sin(2Dkπ) as a termination criterion of the iterative calculation process
=
kπ of xi , which is defined as the ratio between 2-norm of aiK
and and a00 , i.e.
Z Ts
2 kaiK k2
βk = δ(t) sin(kωt)dt T olerance = × 100% (18)
Ts 0 ka00 k2
(14)
in which aiK represents the magnitude of the highest Kth
1 − cos(2Dkπ) harmonic in the correction xi . In this paper, the tolerance error
=
kπ index is set to be 1%, which is sufficient for most engineering
The terms δ0 and δi have the same definitions as those applications. And the correcting process is terminated once
described in [21], i.e. this index is smaller than 1%, which means the harmonic
magnitude in xi is much smaller than the DC component,
δ0 = b0 + b1 ejτ + c.c (15a) and the approximate solution obtained under this condition
is accurate enough for most DC-DC converters.
δi = b2i ej2iτ + b2i+1 ej(2i+1)τ + c.c (15b)
When xi and δi are substituted into (10), the spectral
content of the term fi usually becomes wider than the spectral IV. P ERIODIC STEADY- STATE SOLUTIONS OF THE
content of the term xi , thus, each fi can be rewritten as FRACTIONAL - ORDER B OOST CONVERTER IN CCM

fi = fim + εRi+1 (i = 0, 1, 2, . . . ) (16) According to the low-pass filtering property of


DC-DC converters, the main wave x0 is chosen as
In (16), f0m contains all the items that have the same spec- x0 = a00 = [I00 V00 ]T , where I00 and V00 are DC
tral contents with the main wave x0 , while the remaining ones values. Based on the discussion on the term fi in Section III,
belong to the term R1 . Generally, the harmonic magnitudes in the term f0m should include the same spectral content with x0 ,
R1 are smaller than those in f0m , so the indicator parameter and the rest harmonics in f0 belong to R1 , so it can be deduced
ε is used here again. These newly generated harmonics in
f0 determine the spectral content in the following spectral f0m = a00 b0 = a00 D (19a)
content set {E1r }, that is, f1m has the same spectral content set

{E1r } with the correction x1 . Similarly, the spectral contents R1 = b1 a00 e + c.c (19b)
in R2 determine the spectral contents of the correction x2 , etc.
Then substituting (7) and (9) into (5), and collecting the terms
Then introducing x0 and f0m into (17a) gives
with the same εi , one can get the following equivalent linear
fractional-order differential equations as (G10 + G2 D)a00 = U (20)
α β α β
G1 (p , p )x0 + G2 f0m = U (17a) The coefficient matrix G10 = G1 (0 , 0 ), thus (20) can be
rewritten in the matrix form as
G1 (pα , pβ )x1 + G2 (f1m + R1 ) = 0 (17b)
   
α β
G1 (p , p )x2 + G2 (f2m + R2 ) = 0 (17c) 1 " # 1 " # "U #
 0 L 
I00  0 − L  I00 in
L
 1 1 +D 1 = (21)
V00 V00
  
and so on. − 0 0
Note that equations in (17) are all linear and can be solved C RC C
step by step by using the principle of harmonic balance, Thereby a00 can be obtained. Subsequently, according
where the first equation is used to obtain the main wave x0 , to the spectral contents of R1 , we deduce that the spectral
and the following equations are used to solve the corrections content set of the correcting term x1 is {E1r } = {1}. Thus

0885-8993 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2016.2628783, IEEE
Transactions on Power Electronics
IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. X, NO. X, XXXX 20XX 4

the first order correction x1 can be assumed to be


(G10 + G2 b0 )a20 = −G2 (b1 a11 + b1 a11 ) (29a)
x1 = a11 ejτ + c.c (22)
(G12 + G2 b0 )a22 = −G2 (b1 a11 + b2 a00 + b3 a11 ) (29b)
in which a11 = [I11 V11 ]T . Substitution of x0 , x1 , δ0 and δ1
(G13 + G2 b0 )a23 = −G2 (b1 a22 + b2 a11 + b3 a00 ) (29c)
into f1 can deduce the following expressions

f1m = b0 a11 ejτ + c.c



(23a) Likewise, the coefficient matrix G1k =G1 (jkω)α , (jkω)β
(k=0, 2, 3), and (29) can be rewritten as
R2 = (b1 a11 + b1 a11 ) + (b1 a11 + b2 a00 )ej2τ + b3 a00 ej3τ + c.c
(23b)    
1 D " #
 0   0 −L
  I
20
where f1m should consist of all the terms in f1 with the same L
 1 1  + D 
V
harmonics as x1 , and R2 includes the rest. Introducing x1 , − 0 20
C RC C (30a)
R1 and f1m into (17b) leads to the following equation  
1 " #
0 b
 1 11 I + b I
1 11
G1 (pα , pβ )a11 ejτ + G2 b0 a11 ejτ + b1 a00 ejτ = 0 =  1 L
 
(24)
− 0 b V
1 11 + b1 V 11
C
According to (1), it can be derived that, pα ejωt = (jω)α ejωt    
1 D
and pβ ejωt = (jω)β ejωt , then a11 can be obtained by solving  (j2ω)α
" #
0 −   I22
L L
1  + D
  
the following equation 1 V22

− (j2ω)β + 0
C RC C
(G11 + G2 b0 )a11 = −G2 b1 a00
(25)  
1 " #
 0 b1 I11 + b2 I00 + b3 I 11
in which the coefficient matrix G11 = G1 (jω)α , (jω)β . =  1 L
 
Thus, (25) can be rewritten in the matrix form as − 0 b1 V11 + b2 V00 + b3 V 11
C

1
 
D
 (30b)
 (jω)α
" #
  0 −L
 I11
L
1  + D

 1 
1
 
D

V11

− (jω)β +  (j3ω)α
" #
0 0 −   I23
 C RC C L
1  + D
L
  
(26) 1 V23
 
1 " # − (j3ω)β + 0
0 b I
=  1 L
  1 00  C  RC C (30c)
b1 V00 1 "

#
0  0 b I
1 22 + b I
2 11 + b I
3 00
C = 1 L
b V + b V + b 3 V00

− 0 1 22 2 11
Obviously, due to the effects of (jω)α and (jω)β in the C
matrix G11 , the harmonic amplitude in the correction x1 where bk = [sin 2kDπ − j(1 − cos 2kDπ)]/2kπ (k=1, 2, 3).
has a nonlinear relationship with the order α and β. Then Following a similar way, aik = [Iik , Vik ]T can be solved,
based on the spectral contents of R2 in (23b), the spectral and then the correction xi can be obtained. Without doubt,
content set of x2 can be determined, that is {E2r } = {0, 2, 3}. higher-order correction xi would bring a more accurate
Therefore the second-order correction x2 can be assumed to be solution. The detailed expression of each aik is listed in Tab.
I, where ∆(jkω) is defined as
x2 = a20 + a22 ej2τ + a23 ej3τ + c.c (27)
∆(jkω) = (1 − D)2 + (jkω)α+β LC + (jkω)α L/R (31)
T
here a20 = [I20 V20 ] is a vector with DC values, it will give
the correction of a00 . a22 = [I22 V22 ]T and a23 = [I23 V23 ]T When the components x0 , x1 and x2 are obtained, the
are the second- and third-harmonic magnitudes of the state approximate periodic solution can be expressed as
vector x respectively. Similarly, the term f2m , which should
include the same harmonics with x2 , can be derived by  T
x = iL vC
introducing xi and δi (here i = 0, 1, 2) into f2 .
≈x0 + x1 + x2 (32)
f2m =b0 a20 + (b0 a22 + b1 a23 + b3 a11 )ej2τ =(a00 + a20 ) + (a11 ejτ + a22 ej2τ + a23 ej3τ + c.c)
(28)
+ (b0 a23 + b1 a22 + b2 a11 )ej3τ + c.c The components of this vector are
Next, substituting x2 , R2 and f2m into (17c), and using iL ≈ (I00 + I20 )...
the method of harmonic balance, one can get the following
+ 2(<I11 cos ωt − =I11 sin ωt...
equations (33)
+ <I22 cos 2ωt − =I22 sin 2ωt...
+ <I23 cos 3ωt − =I23 sin 3ωt)

0885-8993 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2016.2628783, IEEE
Transactions on Power Electronics
IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. X, NO. X, XXXX 20XX 5

TABLE I
VALUES OF aik

aik
(i, k)
Vik Iik

Uin Uin
(0, 0) 1−D R(1−D)2

(1−D)b1 V00 −(jω)α Lb1 I00 b1 V00 −(1−D)V11


(1, 1) ∆(jω) (jω)α L

b̄1 V11 +b1 V̄11 V20 /R+b̄1 I11 +b1 I¯11


(2, 0) 1−D 1−D

(1−D)(b1 V11 +b2 V00 +b3 V̄11 )−(j2ω)α L(b1 I11 +b2 I00 +b3 I¯11 ) (b1 V11 +b2 V00 +b3 V̄11 )−(1−D)V22
(2, 2) ∆(j2ω) (j2ω)α

(1−D)(b1 V22 +b2 V11 +b3 V00 )−(j3ω)α L(b1 I22 +b2 I11 +b3 I00 ) (b1 V22 +b2 V11 +b3 V̄00 )−(1−D)V23
(2, 3) ∆(j3ω) (j3ω)α

... ... ...

and is no unified definition for the unit of the fractional-order


vC ≈ (V00 + V20 )... inductor now, so the unit µH would still be used. Then, a
contour plot of the CCM-operating boundary can be obtained
+ 2(<V11 cos ωt − =V11 sin ωt...
(34) in Fig. 2.
+ <V22 cos 2ωt − =V22 sin 2ωt...
+ <V23 cos 3ωt − =V23 sin 3ωt)
where <(·)and =(·) represent the real- and imaginary- part of
the complex term in the bracket.

V. D ISCUSSIONS OF THE ORDER - RELATED PHENOMENA


According to the boundary condition of the conventional
Boost converter, if the converter operates in CCM, the
average value of the inductor current must be greater than its
peak-peak ripple [24]. Here, we use I¯L and ∆iL to represent
the average value and peak-peak ripple of inductor current
respectively, where I¯L can be approximated by the summation
of I00 and I20 , and ∆iL can be calculated from the equation
(3a) based on the principle of inductor volt-second balance. Fig. 2. The operating boundary of the fractional-order Boost converter.
Then according to Caputos definition of fractional integral
[17], the CCM-operating criterion of the fractional-order
Boost converter can be derived as
Additionally, according to the solution obtained by the
Φ =I¯L − ∆iL proposed method, a statistical analysis of harmonics under
Uin (DT )α (35) different fractional orders is made. As shown in Fig. 3, the
≈I00 + I02 − RMS values of harmonics increase with the decrease of α and
2αLΓ(α)
β, which will lead to the increasing of the ripples of state
in which Γ(α) can be solved by the ”Gamma function” in variables, and is usually undesirable for DC-DC converters.
MATLAB. If Φ > 0, the fractional-order Boost converter Besides, orders also affect the terminating criterion of the
operates in CCM, otherwise, it operates in discontinuous- proposed method. It can be seen from Fig. 4 that, when
conduction mode (DCM). (α, β) = (1, 1), one time of correction is enough as the
The parameters of the fractional-order Boost converters in tolerance error index for a11 is less than 1%. And when
Fig. 1 are as follows, i.e. the switching frequency fs = (α, β) = (1, 0.9), more than one correction should be per-
25kHz, the input voltage Uin = 12V , the load resistance formed, as the index for a11 is greater than 1%. One can also
R = 40Ω, the duty ratio D = 0.4, L = 477µH, C = see that the tolerance error index increases with the decrease
10µF/(second)1−β . Note that the unit of fractional-order of α and β, which means that more correction times should
capacitor, is µF/(second)1−β , instead of µF [2]. As there be done for smaller fractional orders.

0885-8993 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2016.2628783, IEEE
Transactions on Power Electronics
IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. X, NO. X, XXXX 20XX 6

(a)
Fig. 4. The tolerance error index.

the state variables from zero initial state to steady-state.


Also, by using Oustaloups’s approximation method [25], a
MATLAB/Simulink based simulation is offered. The results
from these two methods are listed together with those from
our method in Tab.II.
It can be seen that, the DC components from the three
methods agree well with each other, and they are obviously
dependent on the orders of the inductor and capacitor. Further,
(b)
the steady-state ripples from these three methods are compared
Fig. 3. The RMS values of harmonics in state variables under different with each other in Fig. 5 and Fig. 6, where the blue dotted
fractional orders: (a) for iL , (b) for vC . lines, black solid lines and red dash-dot lines represent the
results from the PECE-ABM method, Oustaloup’s approxi-
mation method and our method respectively.
VI. C OMPARISON AND VERIFICATIONS
As shown in Fig. 5 and Fig. 6, the harmonic magnitudes
A. Comparison of different methods are also order dependent. More specifically, the harmonic
Using the parameters in Section V, a comparison is made magnitudes of iL and vC increase with the decreases of α and
between the results from different methods in this part. An β respectively. Waveforms coming from these three methods
Adams-Bashforth-Moulton typed predictor-corrector method show good agreement with each other, which coincides with
(PECE-ABM) is applied to perform numerical simulations the analysis in Section V.
[15]. Subroutines are written in MATLAB, so (3) and (4)
can be solved in each switching state, and a cycle-by-cycle
numerical simulation is carried out. Then the results obtained B. Implementations of fractional-order devices
in each switching state would be collected in one matrix,
in which the values of iL and vC are included. The final Theoretically, the impedance of an order-α inductor is
matrix obtained would contain the full information about defined as

TABLE II
DC COMPONENTS OF STATE VARIABLES (iL (A), vC (V )) OBTAINED BY DIFFERENT METHODS

Order (α, β) Our Method (a00 + a20 ) PECE-ABM Method Oustaloup’s Method
(1, 1) (0.8306, 19.9662) (0.8236, 19.8810) (0.8720, 19.8628)
(1, 0.9) (0.8204, 19.7870) (0.8377, 19.7376) (0.8379, 19.6360)
(1, 0.8) (0.7771, 18.9121) (0.8270, 19.0599) (0.8001, 18.5520)
(0.95, 1) (0.8403, 19.9406) (0.8339, 20.0031) (0.8335, 19.9315)
(0.95, 0.95) (0.8355, 19.8657) (0.8439, 19.9326) (0.8331, 19.8752)
(0.95, 0.9) (0.8261, 19.7043) (0.8502, 19.7877) (0.8266, 19.6705)
(0.95, 0.85) (0.8077, 19.3609) (0.8493, 19.4905) (0.8162, 19.1852)
(0.95, 0.8) (0.7729, 18.6494) (0.8346, 18.8933) (0.7656, 18.3543)
(0.9, 1) (0.8678, 19.8946) (0.8531, 20.1115) (0.9947, 19.8894)
(0.9, 0.95) (0.8609, 19.7831) (0.8662, 20.0015) (0.8693, 19.7838)
(0.9, 0.9) (0.8483, 19.5561) (0.8743, 19.7852) (0.8553, 19.4890)
(0.9, 0.85) (0.8252, 19.0958) (0.8729, 19.3608) (0.8063, 19.0011)
(0.9, 0.8) (0.7861, 18.1884) (0.8559, 18.5507) (0.7901, 17.8460)

0885-8993 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2016.2628783, IEEE
Transactions on Power Electronics
IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. X, NO. X, XXXX 20XX 7

(a) (b) (a) (b)

(c) (d) (c) (d)

(e) (f) (e) (f)

Fig. 5. Comparison of steady-state ripple of iL with different (α, β): (a) Fig. 6. Comparison of steady-state ripples of vC with different (α, β): (a)
(α, β) = (1, 1), (b) (α, β) = (1, 0.9), (c) (α, β) = (1, 0.8), (d) (α, β) = (α, β) = (1, 1), (b) (α, β) = (1, 0.9), (c) (α, β) = (1, 0.8), (d) (α, β) =
(0.9, 1), (e) (α, β) = (0.9, 0.9), (f) (α, β) = (0.9, 0.8). (0.9, 1), (e) (α, β) = (0.9, 0.9), (f) (α, β) = (0.9, 0.8).

Zα = (jω)α L
(36)
απ απ
= (ω)α L[cos( ) + j sin( )]
2 2
And Zα has the following frequency-domain characteristics.
• Magnitude: with a constant slope of 20α dB/dec.
(a)
• Phase: a horizontal line with the value of απ/2.

Similarly, the impedance of an order-β capacitor is defined as


1
Zβ =
(jω)β C
(37)
1 βπ βπ
= β
[cos( ) − j sin( )] (b)
(ω) C 2 2
And Zβ has the following frequency-domain characteristics Fig. 7. The chain structure of fractional-order components: (a) the fractional-
order capacitor, (b) the fractional-order inductor.
[26].
• Magnitude: with a constant slope of −20β dB/dec. Based on the method described in [32], the implementation
• Phase: a horizontal line which value is −βπ/2. circuits of the fractional-order capacitor and inductor are
As there are currently very few commercially available constructed in typical chain structures, as shown in Fig. 7,
fractional-order elements, many methods are proposed to where the parameters such as Ri , Ci and Li can be obtained by
construct the equivalent implementation circuit of a fractional the synthesis method of impedance networks. For the order-0.8
element [27]–[29], although it may lead to flaws when dealing and order-0.9 fractional capacitor, the maximum error of the
with problems such as the stability analysis of nonlinear magnitude response curve between the theoretical fractional
dynamic systems [30], [31]. impedance and the built circuit impedance is set to be less

0885-8993 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2016.2628783, IEEE
Transactions on Power Electronics
IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. X, NO. X, XXXX 20XX 8

TABLE III
Ri AND Ci VALUES OF FRACTIONAL CAPACITORS UNDER DIFFERENT ORDER β

β 0.8 0.9
i Ri (Ω) Ci (µF ) Ri (Ω) Ci (µF )
Type By synthesis/actual By synthesis/actual By synthesis/actual By synthesis/actual
0 0 0 0 0
1 18.6m/20m 0.40389/0.4 0.23487m/0.25m 2.5524/2.5
2 177.1m/200m 0.75315/0.78 23.4m/20m 4.2749/4.27
3 1.7662/1.8 1.3426/1.3 2.3392/2.55 7.1311/7.12
4 17.6598/20 2.3879/2.34 233.9241/225 11.895/11.89
5 176.5976/170 4.2463/4.2 23.476k/23.47k 19.772/19.7
6 1.7661k/1.74k 7.5505/7.5 5.8263M/5.75M 13.289/13.26
7 17.69k/17.7k 13.405/13.38 \ \
8 182.10k/182k 23.158/23 \ \
9 3.9267M/3.91M 19.097/19.09 \ \

than 2dB, and based on this tolerance error, the values of and α = 0.8 are listed in Tab. IV.
Ri and Ci in Fig.7(a) can be determined, as listed in Tab. Note that some values of Ri in Tab. V are extremely small,
III. Note that some resistors have milli-Ω values, so we will especially when α = 0.8, which will bring difficulties in
choose constantan resistors or cement resistors in experiments implementations. However, according to Fig. 2, the converter
with considering the power stress. would be in DCM for α = 0.8, which is not included in the
Moreover, according to the duality principle, the fractional- proposed model, so only the circuit-oriented simulation for
order inductors can be constructed. More specifically, as this situation is provided, and the result of which is used to
the impedance of the fractional-order inductor has a similar verify of the CCM-operating criterion.
style with the admittance of a fractional-order capacitor, the Using the values of Ri , Ci and Li listed in Tab. III and Tab.
reciprocal of a fractances transfer function can be regarded as IV, The impedance of the 10µF/(second)1−β capacitor and
the transfer function of a fractional-order inductor. So we can the impedance of 477µH inductor with different order α and
follow the same way described in [32] to construct the transfer β are listed in Tab. V.
functions of fractional-order inductors. When the maximum Additionally, the Bode diagrams of the constructed
error between the approximate curves and the theoretical lines fractional-order devices are depicted in Fig. 8, where the black
of frequency-domain characteristics is set to be 1dB, the solid lines represent the theoretical value calculated according
values of Ri and Li for the inductors with the order α = 0.9 to (36) and (37), and the red dash-dot and blue dash curves

TABLE IV
Ri AND Li VALUES OF OF FRACTIONAL INDUCTORS UNDER DIFFERENT ORDER α

α 0.9 0.8
i Ri (Ω) Li (µF ) Ri (Ω) Li (µF )
Type By synthesis/actual By synthesis/actual By synthesis/actual By synthesis/actual
0 0 65.642/65 0 13.857
1 1.6397k/1.635k 18.634/18 82.8873 4.0363
2 167.0723/165 24.523/24 28.2491 5.8010
3 16.73/16.5 31.716/32 9.0663 7.8511
4 1.6732/1.5 40.967/40 2.8768 10.505
5 0.1673/0.165 52.911/51 0.9104 14.020
6 16.7m/20m 68.335/68 0.2880 18.699
7 1.7m/2m 88.221/90 0.0911 24.937
8 0.1663m/1m 113.33/113.4 0.0288 33.254
9 0.0155m/1m 136.09/136 0.0091 44.343
10 \ \ 0.0029 59.126
11 \ \ 0.9100m 78.805
12 \ \ 0.2872m 104.86
13 \ \ 0.0900m 138.55
14 \ \ 0.0273m 177.53
15 \ \ 0.0071m 195.24

0885-8993 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2016.2628783, IEEE
Transactions on Power Electronics
IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. X, NO. X, XXXX 20XX 9

TABLE V
I MPEDANCE OF FRACTIONAL - ORDER DEVICES

Zα Zβ
0.9 0.8 0.8 0.9
Theoretical values 3.5429 + 22.3690j 2.1154 + 6.5106j 2.1532 − 6.6269j 0.3295 − 2.0802j
By Synthesis 3.2088 + 22.0600j 2.1068 + 6.5222j 2.1566 − 6.6240j 0.3195 − 2.0976j
Actual values 3.1247 + 21.7475j \ 2.1829 − 6.7095j 0.3007 − 2.1352j

belong to the devices obtained by synthesis and approximated


by actual devices respectively. It can be seen, the magnitude-
frequency characteristic curves meet well with each other in a
certain frequency band, and the phase plot of the implemented
fractional impedance fluctuates in a certain range around the
theoretical ones. Of course, more precise results can be got by
reducing the maximum error, but it would lead to an increase
in the numbers of Ri Ci or Ri Li blocks in the chain structure.
(a) (b)

(c) (d)
(a) (b)
Fig. 9. Steady-state waveforms of state variables with different values of
(α, β): (a) iL in CCM, (b) vC in CCM, (c) iL in CCM and DCM, (d) vC
in CCM and DCM.

D. Experimental results
To further evaluate the performance of the proposed method,
experimental works are considered. The types of power semi-
conductor devices ST and SD are IRF640B and MUR1620CT,
(c) (d)
respectively. Additionally, to reduce the influences from par-
asitic parameters of the devices, the metalized polypropylene
Fig. 8. Bode diagrams of fractional-order devices with different orders: (a) film capacitors are used to build the fractance. Meanwhile, to
C β with β = 0.9, (b) C β with β = 0.8, (c) Lα with α = 0.9, (d) Lα with
α = 0.8.
reduce the interference between power inductors in the chain
structure, the Coilcraft MSS1278 and MSS1583 series shielded
surface mount power inductors are used to build the fractional-
order inductor. Experimental results are shown in Fig. 10 and
Fig. 11.
C. Circuit-oriented simulations in PSIM Furthermore, the RMS values and peak-to-peak values of
Circuit-oriented simulations are performed in PSIM based state variables are listed in Tab. VI. Considering parasitic
on the implementation circuits of fractional elements. Data parameters, such as the maximum instantaneous forward volt-
obtained during simulations are reloaded to the workspace of age of the MOSTFET and the diode, the errors between
MATLAB, and the waveforms of state variables are plotted experimental results and simulations are in an acceptable
in Fig. 9. It can be seen that the ripple amplitudes of iL range.
and vC increase significantly with the decrease of α and
β, respectively. These results meet well with those from VII. C ONCLUSION
theoretical analysis and numerical simulations. The concepts on fractional-order circuits and systems have
Note that from Fig.9(c), one can see that when α = 0.8, received widespread attention in engineering applications re-
the converter operates in DCM, which is in accord with the cently. This paper offers a modeling and analysis method for
theoretical analysis of CCM-operating boundary in Section V. the fractional-order DC-DC converters, by which the approx-
imate analytical solution can be obtained without considering

0885-8993 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2016.2628783, IEEE
Transactions on Power Electronics
IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. X, NO. X, XXXX 20XX 10

(a) (b) (a) (b)

(c) (d) (c) (d)

(e) (f) (e) (f)

Fig. 10. Steady-state waveforms of iL got in experiments under differ- Fig. 11. Steady-state waveforms of vC got in experiments under differ-
ent fractional orders: (a) (α, β) = (1, 1), (b) (α, β) = (1, 0.9), (c) ent fractional orders: (a) (α, β) = (1, 1), (b) (α, β) = (1, 0.9), (c)
(α, β) = (1, 0.8), (d) (α, β) = (0.9, 1), (e) (α, β) = (0.9, 0.9), (f) (α, β) = (1, 0.8), (d) (α, β) = (0.9, 1), (e) (α, β) = (0.9, 0.9), (f)
(α, β) = (0.9, 0.8). (α, β) = (0.9, 0.8).

complex definitions of fractional calculus, and a reference for As for the potential applications of fractional elements in
the optimal design of parameters can be supplied. power electronic systems, and the effects of fractional orders
In the fractional-order DC-DC Boost converter, both the on the properties of systems, further works should still be
DC components and harmonics of state variables are heavily done. Moreover, during our research work, we found that
dependent on orders. As a result, the change of orders would there is an association between the fractional-order capacitor
affect the steady-state characteristics of the converter. For and the equivalent circuit model of the electrolytic capacitor,
example, the orders have a noticeable effect on the CCM- thus detailed works can be done and more accurate models of
operating boundary, especially the order α of the inductor; the electrolytic capacitors may be developed in the future.
ripples of state variables are also order-dependent. As reflected
in the simulation and experimental results, when the orders of R EFERENCES
fractional devices decrease, the ripples of state variables rise,
and vice versa. These phenomena have been explained by the [1] I. Petras, Fractional-Order Nonlinear Systems. Springer-Verlag Berlin
Heidelberg, 2011.
theoretical analysis of the proposed method, and confirmed by [2] S. Westerlund and L. Ekstam, “Capacitor theory,” IEEE Transactions
simulations and experimental works. on Dielectrics and Electrical Insulation, vol. 1, pp. 826–39, Oct. 1994.

TABLE VI
R ESULTS OF CIRCUIT- ORIENTED SIMULATIONS AND EXPERIMENTS (C IRCUIT- ORIENTED S IMULATIONS / E XPERIMENTS )

xRM S ∆xpp
(α, β)
vRM S (V) iRM S (A) ∆vpp (V) ∆ipp (A)
(1,1) 19.99/17.93 0.83/0.78 0.80/1.00 0.4025/0.3959
(1,0.9) 19.89/17.75 0.82/0.75 2.55/2.20 0.4025/0.4079
(1,0.8) 19.39/17.09 0.78/0.71 7.35/6.60 0.4025/0.3899
(0.9,1) 19.99/17.57 0.96/0.84 0.83/0.84 1.3591/1.297
(0.9,0.9) 19.54/17.43 0.94/0.83 2.91/2.80 1.3627/1.287
(0.9,0.8) 18.38/16.40 0.89/0.81 7.65/7.87 1.3645/1.258

0885-8993 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2016.2628783, IEEE
Transactions on Power Electronics
IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. X, NO. X, XXXX 20XX 11

[3] H. Malek, S. Dadras, and Y. Chen, “Fractional order equivalent series [27] J. Valsa and J. Vlach, “RC models of a constant phase element,”
resistance modelling of electrolytic capacitor and fractional order fail- International Journal of Circuit Theory and Applications, vol. 41, no. 1,
ure prediction with application to predictive maintenance,” IET Power pp. 59–67, JAN 2013.
Electronics, vol. 9, no. 8, pp. 1608–1613, JUN 29 2016. [28] M. C. Tripathy, D. Mondal, K. Biswas, and S. Sen, “Experimental
[4] M. C. Tripathy, D. Mondal, K. Biswas, and S. Sen, “Experimental studies on realization of fractional inductors and fractional-order band-
studies on realization of fractional inductors and fractional-order band- pass filters,” International Journal of Circuit Theory and Applications,
pass filters,” International Journal of Circuit Theory and Applications, vol. 43, no. 9, pp. 1183–1196, SEP 2015.
vol. 43, no. 9, pp. 1183–1196, SEP 2015. [29] M. S. Sarafraz and M. S. Tavazoei, “Realizability of Fractional-Order
[5] Yi-Fei Pu and Xiao Yuan, “Fracmemristor: Fractional-Order Memristor,” Impedances by Passive Electrical Networks Composed of a Fractional
IEEE Access, vol. 4, pp. 1872–88, 2016. Capacitor and RLC Components,” IEEE Transactions on Circuits and
[6] C. Yang, H. Yu, Y. Shang, and W. Fei, “Characterization of CMOS Systems I-Regular Papers, vol. 62, no. 12, pp. 2829–2835, DEC 2015.
Metamaterial Transmission Line by Compact Fractional-Order Equiva- [30] M. Tavakoli-Kakhki, M. Haeri, and M. S. Tavazoei, “Notes on the
lent Circuit Model,” IEEE Transactions on Electron Devices, vol. 62, State Space Realizations of Rational Order Transfer Functions,” IEEE
no. 9, pp. 3012–3018, SEP 2015. Transactions on Circuits and Systems I-Regular Papers, vol. 58, no. 5,
[7] J. Sabatier, M. Cugnet, S. Laruelle, S. Grugeon, B. Sahut, A. Oustaloup, pp. 1099–1108, MAY 2011.
and J. M. Tarascon, “A fractional order model for lead-acid battery [31] M. S. Tavazoei, “Comments on “Chaotic Characteristics Analysis and
crankability estimation,” Communications in Nonlinear Science and Circuit Implementation for a Fractional-Order System”,” IEEE Trans-
Numerical Simulation, vol. 15, no. 5, pp. 1308–1317, MAY 2010. actions on Circuits and Systems I-Regular Papers, vol. 62, no. 1, pp.
[8] T. J. Freeborn, B. Maundy, and A. S. Elwakil, “Measurement of Superca- 329–332, JAN 2015.
pacitor Fractional-Order Model Parameters From Voltage-Excited Step [32] A. Charef, H. Sun, Y. Tsao, and B. Onaral, “Fractal system as rep-
Response,” IEEE Journal on Emerging and Selected Topics in Circuits resented by singularity function,” IEEE Transactions on Automatic
and Systems, vol. 3, no. 3, pp. 367–376, SEP 2013. Control, vol. 37, pp. 1465–70, Sept. 1992.
[9] G. Tsirimokou, C. Psychalinos, A. Allagui, and A. S. Elwakil, “Simple
non-impedance-based measuring technique for supercapacitors,” Elec-
tronics Letters, vol. 51, no. 21, pp. 1699–1700, OCT 8 2015.
[10] K. Biswas, S. Sen, and P. K. Dutta, “Realization of a constant phase
element and its performance study in a differentiator circuit,” IEEE
Transactions on Circuits and Systems II-Express Briefs, vol. 53, no. 9,
pp. 802–806, SEP 2006. Xi Chen was born in Henan, China, in 1988. He
received the M.S. degree in Pattern Recognition and
[11] D. Mondal and K. Biswas, “Packaging of Single-Component Fractional
Intelligent Systems from the College of Electronic
Order Element,” IEEE Transactions on Device and Materials Reliability,
and Information Engineering, Henan University of
vol. 13, no. 1, pp. 73–80, MAR 2013.
Science & Technology, Luoyang, China, in 2013.
[12] A. S. Elwakil, “Fractional-Order Circuits and Systems: An Emerging
He is currently working towards the Ph.D. degree in
Interdisciplinary Research Area,” IEEE Circuits and Systems Magzines,
power electronics at the School of Electric Power,
vol. 10, no. 4, pp. 40–50, 2010.
South China University of Technology. His research
[13] A. S. Elwakil, A. Allagui, B. J. Maundy, and C. Psychalinos, “A low interests include the modeling and control of non-
frequency oscillator using a super-capacitor,” AEU-International Journal linear systems and power electronics.
of Electronics and Communications, vol. 70, no. 7, pp. 970–973, 2016.
[14] T. T. Hartley, R. J. Veillette, J. L. Adams, and C. F. Lorenzo, “Energy
storage and loss in fractional-order circuit elements,” IET Circuits
Devices & Systems, vol. 9, no. 3, pp. 227–235, MAY 2015.
[15] K. Diethelm, N. Ford, and A. Freed, “A predictor-corrector approach for
the numerical solution of fractional differential equations,” Nonlinear
Dynamics, vol. 29, no. 1-4, pp. 3–22, JUL-SEP 2002.
[16] K. Diethelm, The Analysis of Fractional Differential Equations. Yanfeng Chen (M’06) received the M.E. degree
Springer-Verlag Berlin Heidelberg, 2004. in power electronics technology from Wuhan Uni-
[17] I. Podlubny, Fractional Differential Equations. Academic Press, 1998. versity, China (1995), and the Ph.D degree in cir-
[18] W. Fa-Qiang and M. Xi-Kui, “Transfer function modeling and analysis cuits and systems from South China University of
of the open-loop Buck converter using the fractional calculus,” Chinese Technology, China (2000). From November 2005 to
Physics B, vol. 22, no. 3, MAR 2013. December 2006, she was a Research Associate with
[19] C. Wu, G. Si, Y. Zhang, and N. Yang, “The fractional-order state- the Department of Electronic and Information En-
space averaging modeling of the Buck Boost DC/DC converter in dis- gineering, Hong Kong Polytechnic University, Hong
continuous conduction mode and the performance analysis,” Nonlinear Kong. She is currently a Professor with the School of
Dynamics, vol. 79, no. 1, pp. 689–703, JAN 2015. Electric Power, South China University of Technol-
[20] C. Tseng, S. Pei, and S. Hsia, “Computation of fractional derivatives us- ogy. Her main research interests are modeling and
ing Fourier transform and digital FIR differentiator,” Signal Processing, analysis of nonlinear systems and power electronics.
vol. 80, no. 1, pp. 151–159, JAN 2000.
[21] S. Qiu and I. Filanovsky, “Harmonic analysis of PWM converters,” IEEE
Transactions on Circuits and Systems I-Regular Papers, vol. 47, no. 9,
pp. 1340–1349, SEP 2000.
[22] Y. Wu, A. P. Hu, D. Budgett, S. C. Malpas, and T. Dissanayake, “Effi-
cient Power-Transfer Capability Analysis of the TET System Using the Bo Zhang (M’03-SM’15) was born in Shang-
Equivalent Small Parameter Method,” IEEE Transactions on Biomedical hai, China, in 1962. He received the B.S. de-
Circuits and Systems, vol. 5, no. 3, pp. 272–282, JUN 2011. gree in electrical engineering from Zhejiang Uni-
[23] H. Zhu, M. Ding, and Y. Li, “Gibbs phenomenon for fractional Fourier versity, Hanzhou, China, in 1982, the M.S. degree
series,” IET Signal Processing, vol. 5, no. 8, pp. 728–738, DEC 2011. in power electronics from Southwest Jiaotong Uni-
[24] R. W. Erickson and D. Maksimovic, Fundamentals of Power Electronics. versity, Chengdu, China, in 1988, and the Ph.D.
Springer US, 2001. degree in power electronics from Nanjing University
[25] A. Oustaloup, F. Levron, B. Mathieu, and F. Nanot, “Frequency- of Aeronautics and Astronautics, Nanjing, China, in
band complex noninteger differentiator: Characterization and synthesis,” 1994. He is currently a Professor with the School of
IEEE Transactions on Circuits and Systems I-Fundamental Theory and Electric Power, South China University of Technol-
Applications, vol. 47, no. 1, pp. 25–39, JAN 2000. ogy, Guangzhou, China. He has authored or coau-
[26] D. Sierociuk, I. Podlubny, and I. Petras, “Experimental Evidence of thored more than 380 papers and 80 patents. His current research interests
Variable-Order Behavior of Ladders and Nested Ladders,” IEEE Trans- include nonlinear analysis and control of power supplies and ac drives.
actions on Control Systems Technology, vol. 21, no. 2, pp. 459–466,
MAR 2013.

0885-8993 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2016.2628783, IEEE
Transactions on Power Electronics
IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. X, NO. X, XXXX 20XX 12

Dongyuan Qiu (M’03) was born in China in 1972.


She received the B.Sc. and M.Sc. degrees from
South China University of Technology, Guangzhou,
China, in 1994 and 1997, respectively, and the Ph.D.
degree from the City University of Hong Kong,
Kowloon, Hong Kong, in 2002. She is currently a
Professor with the School of Electric Power, South
China University of Technology. Her main research
interests include design and control of power con-
verters, fault diagnosis, and sneak circuit analysis of
power electronic systems.

0885-8993 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

You might also like