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/TIE.2020.3014562, IEEE
Transactions on Industrial Electronics

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS

Optimized EPS Control to Achieve Full Load

Range ZVS with Seamless Transition for
Dual Active Bridge Converters
Guo Xu, Member, IEEE, Liting Li, Student Member, Xiaoying Chen, Student Member,
Yonglu Liu, Member, IEEE, Yao Sun Member, IEEE, and Mei Su, Member, IEEE

 an output high frequency rectifier. The leakage inductance of

Abstract— Dual active bridge (DAB) converters are the transformer is essential, because the transferred power is
widely studied for applications requiring isolation, controlled by controlling the phase shift angle or voltage level
bidirectional power flow and high efficiency. To solve the of the two voltages at the two ports of leakage inductance.
problem of limited zero voltage switching (ZVS) range and
achieve seamless transition for extended-phase-shift (EPS)
Single phase shift (SPS) control [6] is a classical method for
control, the magnetizing current is utilized and an DAB converter, which is easy to implement. It only controls the
optimized control law is proposed in this paper. With the phase shift angle and can achieve wide zero voltage switching
injection of magnetizing current, the two ZVS regions (ZVS) under unity gain condition. However, when the gain is
under EPS control will be creatively intersected with a large far away from the unity gain, the converter would suffer from
region, instead of disconnected or just intersected at a high turn-off current loss, high circulating current loss and
particular point as in previous studies. The ZVS model
considering switch junction capacitors, ZVS region
losing of zero voltage switching [7].
analysis with magnetizing current injection and power To improve the performance of DAB converter, modulation
transfer characteristics are analyzed. Based on these strategies are widely investigated [8]-[15], which could be
analyses, the optimized EPS control is proposed to achieve regarded as improvements of SPS control. Unlike SPS control
full load range ZVS, low root-mean-square (RMS) current only controlling the phase shift angle, the “zero voltage level”,
and maximum transmission power capability. A 250 kHz, which is obtained by controlling the inner phase shift angle of
500W GaN-based DAB prototype is built to verify the
effectiveness of the proposed solution.
individual bridges, can be applied at the two ports of leakage
inductor. Consequently, extended phase shift (EPS) control [8],
Index Terms—dual-active-bridge, extended-phase-shift dual phase shift (DPS) control [9] and triple phase shift (TPS)
(EPS), zero-voltage switching (ZVS). control [10] are proposed. In addition, the operating frequency
of the converter can also be utilized [11]. With the increase of
I. INTRODUCTION control degree of freedoms, the challenge for these methods is

D ue to the advantages of galvanic isolation, high power how to determine the optimal combinations of multiple control
density, bidirectional energy flow, wide voltage range and variables, which could include phase shift ratio, two duty cycles
easy of realizing soft-switching control, dual active bridge and the operating frequency. As a result, various control
(DAB) dc-dc converter is gaining more and more popularity in methods are investigated aiming at different optimization goals,
industry applications [1], such as energy conversions [2], such as low reactive power [9], minimum root-mean-square
distributed generating systems [3], energy storage systems and (RMS) current [12], minimum reactive power [13], light-load
so on [4]. enhancement [14], full load range ZVS achievement [15], etc.
DAB converter is firstly proposed in [5], and it consists of an Among these optimization goals, ZVS for power switches
input high frequency inverter, a high frequency transformer and could reduce the switching loss, lighten the burden of EMI
filters design and benefit the thermal design. Besides, with the
development of GaN based devices, the operating frequency
Manuscript received January 02, 2020; revised April 29, 2020;
accepted July 25, 2020. This work was supported in part by the Nature can be pushed further high. This will increase the weight of full
Science Foundation of China under Grant 51907206, 61933011 and load range ZVS achievement compared with other optimization
61873289, the Hunan Provincial Key Laboratory of Power Electronics goals.
Equipment and Grid under Grant 2018TP1001, the CRRC Zhuzhou
Electric Locomotive Institute Co., Ltd.(YJYJS2018-017), and the To achieve full load range ZVS under wide voltage range for
Fundamental Research Funds in the Central South University under DAB converters, a three-degree-of-freedom control is proposed
Grant 2020zzts580. (Corresponding author: Yonglu Liu). in [12], and a modulation strategy with seamless transition is
G. Xu, L. Li, X. Chen, Y. Liu, Y. Sun and M. Su are with the School of also proposed in [16]. However, in practice, because of many
Automation, Central South University and with Hunan Provincial Key
Laboratory of Power Electronics Equipment and Grid, Changsha, coupled control variables and multi operation control laws, the
410083, China, (e-mail: xuguocsu@[Link]; lyleetine@[Link]; control complexity is high, and real-time control is somehow
chenxiaoying01@[Link]; liuyonglu@[Link]; complex to achieve. To decrease the control complexity,
yaosuncsu@[Link]; sumeicsu@[Link]).

0278-0046 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
Authorized licensed use limited to: Carleton University. Downloaded on August 26,2020 at [Link] UTC from IEEE Xplore. Restrictions apply.
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/TIE.2020.3014562, IEEE
Transactions on Industrial Electronics

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS

frequency domain analysis for DAB converter is studied in [17], two ZVS regions under EPS control will be creatively
while because only the fundamental frequency signal is intersected with a large region, instead of disconnected or just
considered, the accuracy of this method is decreased. On the intersected at a particular point. This makes the seamless
other hand, to decrease the control complexity and achieve full transition to be possible, and also makes the control to be more
load range ZVS, modifications of converter topology have also flexible. With the obtaining of possibility to achieve seamless
been investigated [18]-[20]. In [18], a series resonate tank is transition, a seamless EPS control law is also proposed to
adopted, and one capacitor is parallel with a bidirectional achieve full load range ZVS, low RMS current and maximum
switch; an LCC branch is also added as in [19] to achieve ZVS power transfer capability, based on the accurate ZVS modeling,
under wide voltage range. Auxiliary LC networks [20] are analysis of ZVS range with consideration of magnetizing
injected between two dual active bridges to effectively provide
inductance and power transfer characteristics analysis.
reactive current to achieve ZVS. However, for these methods,
This paper is organized as follows. In Section II, the
additional components will make the converter complex and
topology of DAB dc-dc converter and its operating principle
also cause additional losses. Based on detail analysis of ZVS
with EPS control are introduced. Then ZVS regions
considering junction capacitors, an improved SPS scheme with
considering the junction capacitors of power switches are
two DC blocking capacitors is provided in [21] to obtain a wide
discussed in Section III. Section IV explains how to design the
soft-switching range in wide voltage range applications. But a
magnetizing inductance to connect the separated ZVS regions.
pair of switches on secondary side is always turn-on or turn-off
After that, the proposed power control law is analyzed in
with the proposed modulation, which reduces the utilization of
Section V. In Section VI, design considerations and
the switches and the transmission power capability of the
implementation of the proposed control strategy are provided.
converter.
Section VII presents the comparison of the proposed control
As a result, in general, to achieve full load range ZVS of
law and the existing methods in the literature. In Section VIII,
DAB converter, the tradeoff must be made between the control
the experimental results are given, which verifies the validation
complexity and the complexity of the converter topology based
of the utilization of magnetizing inductance and the proposed
on different application requirements. In view of the control
control law. At last, conclusions are made in Section IX.
complexity, compared with SPS control, EPS control can adjust
one inner phase shift angle and is able to reduce the inductor S1 S3 Q1 Q3
peak and RMS current by 37.8% and 26.8% [22], and the im
A T C
overall efficiency can be increased by 20% [15]. On the other ik Lk is
hand, compared with TPS control, EPS control has less control C1 uAB Lm uCD C2
V1 V2
variables and is easier to be implemented. Meanwhile, twelve B D
operation modes are needed under the whole operation range, n:1
S2 S4 Q2 Q4
which makes the real time control to be an issue under TPS
control [10]. But under EPS control, very few operation modes Fig. 1. Topology of the DAB dc-dc converter.
are needed, which reduces the control complexity and makes it
possible to achieve real-time control without look-up tables. II. TOPOLOGY AND CONVENTIONAL EPS CONTROL
However, the existed EPS control methods proposed in
The topology of DAB dc-dc converter is depicted in Fig. 1.
[8],[23][24] seldom investigated the full load range ZVS
Both of the input and output sides are active full bridges, and
performance. Even though, full load range ZVS control method
they are linked by a high frequency transformer. The active
is studied in [15], the method needs step-to-step calculation to
bridge in V1 side generates a high frequency voltage uAB, and
determine the working region, which is not a seamless
the other one generates voltage uCD. Bidirectional power
transition and is sensitive to the system parameters. In addition,
transfer can be controlled by manipulating the phase shift ratio
just as most of the previous studies [7]-[9], [12], [15], [19],
between the two high frequency voltages and their duty cycles.
[23]-[26], the polarity of the instant current is used to determine
Fig. 2 shows the operation principle of EPS control. Dφ is the
ZVS region without considering the charging/discharging of
duty cycle corresponding to the phase-shift angle between uAB
junction capacitors of switches.
and uCD, and Dy represents the duty cycle of uAB. V1 and V2 stand
As in [25], ZVS regions for controls with different control
degrees are analyzed. The results show that the two ZVS S1 S2 S1 S2
regions only intersect at one particular point for EPS control. S4 S3 S4 S3
This means that the control law must across that particular point Q1,Q4 Q2,Q3 Q1,Q4 Q2,Q3
to achieve the seamless transition, which is extremely critical in DyTs/2 DyTs/2
real application. Meanwhile, if the charging/discharging of V1 uAB(t) V1 uAB(t)
junction capacitors of switches are considered, there will be no
intersection for the two ZVS regions, which makes the V2 uCD(t) V2 uCD(t)
seamless transition to be impossible. This phenomenon will be
Dφ Ts/2 Dφ Ts/2
further explained in Section III.C of this paper. nik(t),is(t) nik(t),is(t)
In this paper, unlike the previous studies, the charging/
discharging of junction capacitors and the magnetizing t0 t1t2t3t4 t5t6 t7t8t9t10 t11t12 t0 t1t2t3t4 t5t6 t7 t8 t9 t10 t11t12
(a) (b)
inductance of the transformer are added into consideration of Fig. 2. Working waveforms under EPS control. (a) Case 1:
ZVS analysis. With the injection of magnetizing current, the 0≤Dφ≤(1-Dy)/2. (b)Case 2: (1-Dy)/2≤Dφ≤0.5.

0278-0046 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
Authorized licensed use limited to: Carleton University. Downloaded on August 26,2020 at [Link] UTC from IEEE Xplore. Restrictions apply.
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/TIE.2020.3014562, IEEE
Transactions on Industrial Electronics

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS

for the voltages of high side and low side respectively. Ts is the ZVS are analyzed only within one half cycle from t0 to t6, as
switching period. ik and is are the current of leakage inductance denoted in Fig. 2. The average value of ik over one switching
Lk and transformer current in V2 side respectively. n:1 period must be equal to zero at steady state, which satisfies ik(t0)
represents the transformer turns ratio. =-ik(t6). According to this, through integrating the voltage
According to the difference in the position of the rising edges crossing the inductance, the current values at the instants when
of uAB and uCD, the forward working condition can be divided switches are turned OFF can be summarized as in Table I and
into three cases: 0<Dφ<(1-Dy)/2, (1-Dy)/2<Dφ<(1+Dy)/2, Table II for case 1 and case 2 respectively, where k=nV2/V1 is
(1+Dy)/2<Dφ<1. Since serious circulating energy will be the voltage conversion ratio.
generated when Dφ is larger than 0.5 [9], only two cases are
B. ZVS Restrictions Considering Switch Junction
used in real application. The operation waveforms of these two
Capacitors
cases are shown in Fig. 2(a) and Fig. 2(b) respectively.
Generally, to achieve ZVS, the energy stored in the
III. ZVS ANALYSIS CONSIDERING JUNCTION CAPACITORS inductance should be enough to charge/discharge the switch
junction capacitors. As a result, the minimum values of ik when
A. Instant Current Expressions for Leakage Inductance the power switches are turned OFF, defined to be Imin, need to
ZVS conditions are related to the current values when power be guaranteed to provide enough energy for ZVS. Meanwhile,
switches are turned OFF; hence it is necessary to calculate the the instants when ik reaches zero are defined to be tm.
expressions of ik based on Fig. 2. Due to symmetrical The equivalent ZVS circuits for case 1 and case 2 are
characteristic of the PWM, the current values and conditions of illustrated in Fig. 3 and Fig. 4 respectively. Since the
TABLE I impedance for the junction capacitor is much larger compared
CURRENT VALUES IN CASE 1 with other parasitic impedances during the charging process,
Stage Value Time and the minimum charging current highly depends on the
ik (t )  ik (t0 ) （ t  t0 )nV2 /Lk , (1  D y  2 D )Ts / 4 junction capacitors rather than other parasitic parameters, only
t0~t1
ik (t1 )= ( k  D y ) nV2Ts / (4 kLk ) junction capacitors are considered in ZVS analysis.
ik (t )  ik (t2 )  (t  t2 )nV2 / Lk (1  D y  2 D )Ts / 4 Take case 1 for instance. At t1, Q2 and Q3 are turned OFF.
t2~t3
ik (t3 )=(kDy  Dy  2kD )nV2Ts / (4kLk ) During time t1 to t2, is discharges the junction capacitors of Q1
ik (t )  ik (t4 )  (t  t4 )(1  k )nV2 / kLk and Q4 while charges the junction capacitors of Q2 and Q3, as
t4~t5
ik (t5 )=( Dy  kDy  2kD )nV2Ts / (4kLk )
D yTs / 2 shown in Fig. 3(a). The junction capacitors on V2 side resonate
with the leakage inductance equivalent to V2 side. In this
TABLE II process, there is no energy loss, and the ZVS condition for Q1
CURRENT VALUES IN CASE 2 and Q4 can be obtained as
Stage Value Time
is (t1)  nik (t1 )  0 (1)
ik (t )  ik (t0 ) （t  t0 )nV2 /Lk
t0~t1 (1  Dy )Ts / 2 At t3, S3 is turned OFF. ik starts to discharge the junction
ik (t1 )=[2k  ( k  1) D y  2kD ]nV2Ts / (4kLk )
capacitor of S4 and charge the junction capacitor of S3. In this
ik (t)  ik (t2 )  (t  t2 )(1 k)nV2 / (kLk ) process, the energy of Lk is transmitted to the voltage sources V1
t2~t3 ( D y  2 D  1)Ts / 4
ik (t3 )  (k 1 2D )nV2Ts / (4kLk ) and nV2 as shown in Fig. 3(b). Hence, at t=t3, ik(t3) min=Imin1,
ik (t )  ik (t4 )  nV2 (1  k（
) t  t4）/(kLk ) uCs3=0, uCs4=V1, and at t=tm1, ik(tm1) min=0, uCs3=V1, uCs4=0.
t4~t5 (1  Dy  2 D )Ts / 4 During the stage from t3 to tm1, the leakage current can be
ik (t5 )=(Dy  kDy  2kD )nV2Ts / (4kLk )
expressed as
S1 S3 S1 S3
Q1 Q3
iCq1 nik iCs3
2
Lk nV2 iCs1 Lk nV2
nim Lk/n
iCq2 Lm/n2
C2 V2 V1 C1 ik iCs4 V1 C1 iCs2 ik

S2 S4 S2 S4
Q2 Q4
(a) (b) (c)
Fig. 3. Equivalent ZVS circuits of case 1. (a) t1-t2. (b) t3-t4. (c) t5-t6.
S1 S3 S1 S3
Q1 Q3
nik V1/n
iCs3 iCq1
Lk nV2 2
iCs1 Lk nV2
nim Lk/n
V1 C1 ik iCs4 iCq2 C2 V2 V1 C1 iCs2 ik
Lm/n2

S2 S4 S2 S4
Q2 Q4
(a) (b) (c)
Fig. 4. Equivalent ZVS circuits of case 2. (a) t1-t2. (b) t3-t4. (c) t5-t6.

0278-0046 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
Authorized licensed use limited to: Carleton University. Downloaded on August 26,2020 at [Link] UTC from IEEE Xplore. Restrictions apply.
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/TIE.2020.3014562, IEEE
Transactions on Industrial Electronics

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS

Dy (1-k)/2 Dy (1-k)/2
duCs 4 du du
ik  t   Cs
1 1
 Cs Cs 3  2Cs Cs 3 (2) 0.9
Dy  1  2 D ZVS
0.9 ZVS
dt dt dt Non ZVS case 2
Non ZVS
0.8 k=0.4 0.8 k=0.5
where Cs represents the junction capacitor on the V1 side, Cs= 0.7 case 2 0.7 Dy  2k (1 
D ) 
Dy  2k (1  D a (1
Cs1= Cs2= Cs3= Cs4. 0.6 case 1 )a
(1  2k
)
0.6 case 1 C 1 k
 2k )
0.5 1 k 0.5
According to the energy balancing principle, the following 0.4
C
0.4 case 1 case 2
relationship holds 0.3 case 1 case 2 0.3 case 1
E t  E t  E loss  E delivered 1 (3) 0.2 case 1
2kD  a (1  2k )
0.2 2kD
3 m1 Dy 
0.1 Dy  0.1 1 k Dy  1  2 D
1 k
Assuming that the energy of circuit is lossless, it can be 0
0 0.1 0.2 0.3 0.4 0.5
0
0 0.1 0.2 D 0.3 0.4 0.5
D
deduced as 

(a) (b)
L k I m2 in 1 / 2  E d e live re d 1 (4) Dy (1-k)/2 Dy (1-k)/2
1 1
tm1 C case 2 ZVS
Edeliver1   [nV2ik (t )  V1iCs 4 (t )]dt 0.9 case 2 ZVS 0.9
Dy  2k (1 Non ZVS
C Dy  2k (1 Non ZVS case 1 D
t3 0.8  0.8 case 1 )a k=0.8
(5) D ) k=0.7 
case 1 a
tm1 du du nV 0.7 (1  0.7 1 k
(1 
2k )
  [nV2 (2Cs Cs 3 )  V1Cs Cs 4 ]dt  Cs ( 2 ) 2 (1  2k ) case 1 1 k 2k )


1   2kD

 2 kD
0.6 0.6
t3 dt dt k

k  1)
0.5 0.5

1 k
)
1
case 2 case 2

k
Thus, the ZVS condition for S4 can be derived as

Dy  a (2
2k
0.4 case 1 0.4 case 1

 a (
0.3
ik (t3 )  I min1  nV2 2Cs (1  2k ) / Lk / k ,0  k  0.5 0.3
2kD

y
0.2 Dy 

D
0.2
 (6) 2kD Dy  1  2 D 1 k Dy  1  2 D
ik (t3 )  I min1  0, 0.5  k  1 0.1 Dy  1  k 0.1
0 0
0 0.1 0.2 D 0.3 0.4 0.5 0 0.1 0.2 D 0.3 0.4 0.5
S1 is turned OFF at t5. The junction capacitor of S2 starts to be  

discharged by ik, and its drain-source voltage begins to decrease (c) (d)
Fig. 5. ZVS regions when the switch junction capacitors are considered.
as illustrated in Fig. 3(c). In the same way, the leakage current (a) k=0.4. (b) k=0.5. (c) k=0.7. (d) k=0.8.
expression and the relationship with energy can be listed as According to Fig. 2 and Table I as well as Table II, the
duCs1 du du
ik  t   Cs  Cs Cs 2  2Cs Cs1 (7) average power expressions of case 1 in (10) and case 2 in (11)
dt dt dt can be obtained.
tm 2
Edeliver 2   [nV2ik (t )  V1iCs1 (t )]dt 2 Ts / 2 2 t5 ( nV2 ) 2 D y D
t5
tm 2 du du nV (8) Pcase1  V1 
Ts  0
i1 (t ) dt V1 
Ts  t3
ik (t ) dt 
2 kLk f s
(10)
  [nV2  2Cs Cs1  V1Cs Cs1 ]dt  Cs ( 2 )2 (2k  1)
t5 dt dt k 2 Ts / 2 2 t5
Ts 0
Therefore, to achieve ZVS for S2, it should satisfy the Pcase 2  V1  i1 (t ) dt V1   ik (t ) dt
Ts t1
condition that (11)
 ( nV2 ) 2 [1  (1  2 D ) 2  (1  D y ) 2 ] / (8 kLk f s )
ik (t5 )  I min 2  0, 0  k  0.5
 (9) Then, the plots for relationships of power, duty cycle, phase
ik (t5 )  I min 2  nV2 2Cs (2k  1) / Lk / k ,0.5  k  1 shift angle and ZVS regions are shown in Fig. 6. The output
Substituting the parameters in Table I into (1) (6) and (9), the power is normalized by Pbase=(nV2)2/(8Lkfs). As seen in Fig. 6(a),
constraints of ZVS can be summed up in Table III. Likewise, when switch junction capacitors are ignored, the two ZVS
the restrictions of ZVS for the case 2 can also be obtained, as regions only interacted at one particular point P. While the two
listed in Table III, where a=(32LkCs)1/2/Ts. ZVS regions are separated when the junction capacitors are
TABLE III considered, as demonstrated in Fig. 6(b), which means that it is
CONSTRAINTS OF ZVS FOR ALL POWER SWITCHES impossible to achieve full load range ZVS no matter how to
Case 1: 0≤Dφ≤(1-Dy)/2 2: (1-Dy)/2≤Dφ≤0.5 control. To avoid this, the operation of DAB can only work at
Q1-Q4 Dy ≤ k Dφ ≥ (1-k)/2
the lower ZVS region, however that will sacrifice the capability
2kD  a (1  2k ) of the maximum power transfer.
Dy  , k  0.5 2 k (1  D )  a (1  2 k )
S3&S4 1 k Dy 
2kD 1 k
Dy  , k  0.5 1.5 1.5
1 k
1 1
a (2k  1)  2kD a (2k  1)  2kD P_pu P_pu
Dy  , k  0.5 Dy  , k  0.5
S1&S2 1 k 1 k 0.5 P 0.5
2kD 2kD
Dy  , k  0.5 Dy  , k  0.5 0 0
1 k 1 k 1 Non ZVS
1 Non ZVS
Region 0.5 Region
0.4 0.5
0.5 0.3 0.4 0.5 0.3
Dy 0.2 Dy 0.2
0 0 0.1 Dφ 0 0 0.1 Dφ
C. Discontinuous Power Transfer in ZVS Region (a) (b)
without Magnetizing Current Fig. 6. Power transfer curves under different junction capacitors when
The ZVS regions and Non ZVS regions, taking the junction Lm =∞. (a) Ignoring switch junction capacitors. (b) With 65pF junction
capacitors.
capacitors into consideration, are illustrated in Fig. 5 based on
Table III. As seen in Fig. 5, ZVS region is divided into two
parts, and these two parts are not connected under different IV. MAGNETIZING CURRENT INJECTION TO ACHIEVE
voltage ratios. Therefore, the control can’t be seamlessly ENTIRE LOAD RANGE ZVS
changed if the converter needs to work in both two ZVS regions. In order to solve the problem of disconnected ZVS regions,
To better illustrate this phenomenon, the transferred power can the magnetizing inductance is utilized to extend ZVS range,
also be calculated as follows. which is usually ignored in previous studies. The waveforms

0278-0046 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
Authorized licensed use limited to: Carleton University. Downloaded on August 26,2020 at [Link] UTC from IEEE Xplore. Restrictions apply.
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/TIE.2020.3014562, IEEE
Transactions on Industrial Electronics

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS

DyTs/2 DyTs/2  D _ D  [1  k (1  Lk / Lm )] / 2
V1 uAB(t) V1 uAB(t) D: (19)
 D y _ D  k (1  Lk / Lm )
V2 V2 The position of point D can be changed by changing the
uCD(t) uCD(t) value of magnetizing inductance Lm. As seen from Fig. 8, to
Dφ Ts/2 Dφ Ts/2
connect the ZVS regions, point D shall be at the upper left of
point C. Therefore, the following relationship for Lm to connect
im(t) ik(t) im(t)
ik(t) two ZVS regions are established.
is(t) 1  k (1  Lk / Lm ) 1  k  a 1  2 k L
is(t) (  )  ( k  a 1  2k  （
k 1+ k ))
t0 t1 t2 t3 t4t5t6 t7 t8 t9 t10 t11 t12 t0 t1t2 t3 t4 t5 t6 t7 t8 t9 t10 t11 t12 2 2 Lm
(a) (b)
Fig. 7. Current waveforms considering Lm. (a) Case 1. (b) Case 2. (20)
with magnetizing current are shown in Fig. 7. As seen, the From (20), the range of the magnetizing inductance can be
magnetizing inductance has an effect on the waveforms of is, calculated as
whereas the waveforms of ik remain the same as shown in Fig.2. L m  kL k / ( a 1  2 k ) (21)
The intervals utilizing the magnetizing current, the one from t1 The corresponding curve for the normalized power P_pu is
to t2 in case 1 as shown in Fig. 7(a) and another one from t3 to t4 illustrated in Fig. 9. As shown, compared with Fig. 6, the two
in case 2 as shown in Fig. 7(b), are reanalyzed below. The value ZVS regions are intersected with a large region, instead of
of is is determined by disconnected or just intersected at a particular point. This
is (t )  n[ik (t )  im (t )] (12) suggests the possibility to manipulate variables Dy and Dφ in
Considering the symmetry of the waveform of magnetizing order to operate the DAB dc-dc converter under soft switching
current over one switching cycle, the value of im at t1 based on for the whole operation range.
Fig. 7(a) can be denoted as
im (t1 )  nV2Ts / (4 Lm ) (13) 1.5
Meanwhile, the expression of im at t3 based on Fig. 7(b) is
obtained in the same way, and it is the same as (13). 1
P_pu
According to Fig. 3(a), the ZVS condition for switches of V2
0.5
side in case 1 can be drawn as
is (t1 )  n[ik (t1 )  im (t1 )]  0 (14) 0
1 Non ZVS
Based on (13), (14) and Table I, the constraint of ZVS can be Region 0.5
deduced as 0.5 0.3 0.4
Dy
0 0 0.1 0.2 Dφ
D y  k (1  Lk /Lm ) (15)
Fig. 9. Power transfer curve with magnetizing inductance.
As illustrated in Fig. 4(b), there is a voltage source V1/n to
provide energy. ZVS for switches Q1-Q4 in case 2 only needs to V. SEAMLESS POWER CONTROL LAW TO ACHIEVE ZVS
satisfy that WITH LOW RMS CURRENT AND MAXIMUM POWER
is (t3 )  n[ik (t3 )  im (t3 )]  0 (16) TRANSFER CAPABILITY
Based on (13), (16) and Table II, the restriction of ZVS can
As analyzed above, reasonable selection of Dy and Dφ could
be expressed as
enable ZVS of all switches in the full load range at different
D  [1  k (1  Lk / Lm )] / 2 (17)
voltage ratios k. Since different combinations of Dy and Dφ will
In terms of (15) (17) and Table III, the regions when ZVS lead to different RMS currents Irms of leakage inductance in a
regions are connected utilizing the magnetizing inductance are given power, an optimization of Dy and Dφ to achieve low Irms at
illustrated in Fig. 8. After calculation, the coordinates of the precondition of ZVS achievement is further discussed.
intersection point C and point D are Based on the currents in Table I, the normalized value of
 D _ C  (1  k  a 1  2 k ) / 2 leakage RMS current Irms_pu in a switching period can be
C : (18) calculated as
 D y _ C  k  a 1  2 k t6
(1-k)/2 I rms _ pu  2  ik2 (t ) dt / Ts / I base
1 t0
case 2 ZVS (22)
Lk D Non ZVS  2 3( kD y3  2 D y3  3 D y2  12 kD y D2  3kD y  k 2 ) / 3
case 1

k (1  ) Dy  2k (1
Lm D
C  )
a (1 where Ibase=nV2/(8Lkfs).
k 1 k  2k
) For a given power, taking the derivation of Irms_pu versus Dy
leads to
Dy

2kD case 2
Dy  9 k 3 P_ pu 2
1 k  (9 k  18 D y  9 kD y2  18 D y2  )
I rms _ pu 4 D y2

case 1 D y kP_ pu
Dy  1  2 D 3 3( kD y3  2 D y3  3 D y2  12 kD y ( ) 2  3kD y  k 2 )
4Dy
0
0 [1  k (1  Lk / Lm )] / 2 Dφ 0.5
Fig. 8. ZVS regions when the magnetizing current is considered.
(23)

0278-0046 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
Authorized licensed use limited to: Carleton University. Downloaded on August 26,2020 at [Link] UTC from IEEE Xplore. Restrictions apply.
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/TIE.2020.3014562, IEEE
Transactions on Industrial Electronics

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS

By setting (23) equal to zero and replacing P_pu with 4DyDφ/k,

an equation can be got as follow 1.5
Control
k  2 D y  kD y2  2 D y2  4 kD2  0 (24) Law B
1
Solving (24) leads to the solutions of Dy P_pu
Control
1  k 
2
D y _ I min 1a  (1   4 k (2  k ) D2 ) / (2  k ) (25) 0.5 Law A

1  k 
2
D y _ I min 1b  (1   4 k (2  k ) D2 ) / (2  k ) (26) 0
1 Non ZVS
Region 0.5
When Dy is larger than Dy_Imin1a, әIrms_pu/әDy is larger than 0.5 0.3 0.4
zero, and when Dy is less than Dy_Imin1a, әIrms_pu/әDy is less than Dy
0 0 0.1 0.2 Dφ
zero, so (25) is the solution to obtain minimal RMS current. In a Fig. 11. The simplified control laws on the surface of the power transfer.
similar way, the minimum solution of Irms_pu for case 2 can be could also be a control candidate according to the actual
obtained as situation. The expression of the solid line with circles is
[2D  k  1  4(k 2  1) D2  4(k 2  k  1) D  2k 2  2k  1] / k  2[ Lk k  Lm k  (k  2) Lm / k ] k
   
 D  , case1
  Lm  Lk k  Lm k 2k
Dy _ Imin 2  if ((1  k ) / 2  D  (k  1  1  k 2 ) / (2k )) Dy _ B =  (29)

if ((k  1  1  k 2 ) / (2k )  D  1/ 2)  2( Lm  Lk k  Lm k ) D  2 Lk k  Lm  2 Lm k ,case2
1,  ( Lk  Lm )k

( Lk  Lm ) k
(27) Based on (10), (11), (28) and (29), the two control laws are
[1-k(1+Lk/Lm)]/2 (1-k)/2
E F plotted on the power transfer surface, as shown in Fig. 11.
1 Control law A is the curve for control in (28), and also, control
[Link] law B represents (29). As seen, the two power transfer curves
Lk D
k (1  ) Full ZVS are monotonously increasing with the increasing of Dφ and Dy.
Lm
C Curve A No step change occurs in the curves, and the curves change
B Curve B seamlessly. Compared with (28), the calculation of Dy in (29) is
Dy

simple and can be easily implemented by digital controller.

k/(2-k) A
Therefore, (29) is selected to be the control law to achieve
simple calculation, and full load range ZVS in this paper.

0 VI. DESIGN CONSIDERATION

0 D 0.5
(1  k  a 1  2k ) / 2 (k  1  1  k 2 ) / (2k ) A. Turns Ratio n
Fig. 10. Power control laws for different goals.
As seen in (25) and (27), the expressions for Dy_Imin are The change of k will change the position of point C in Fig.10.
nonlinear functions that lead to complex calculation. In To ensure that point C is below line Dy=1, k should be designed
addition, as seen from the ZVS region shown in Fig. 10, the in a suitable range. Then, the relationship between the turns
curve for minimum RMS current which is from point A to B, ratio and voltage gain can be expressed as
then goes to E and ends with F (A-B-E-F) will go across the non n  k m ax (V1 / V 2 ) m in (30)
ZVS region. To simplify the complex control, starting points
B. Leakage Inductance Lk
and ending points of the curve can be taken to linearize the
control law in (25) and (27). In addition, a new intersection Since the leakage inductance is the key component for power
point C, given as (18), can be chosen as the middle point to transmission in DAB converter, the choice of Lk can be
satisfy ZVS condition. As a result, three segments as the dotted determined by the maximum power. It can be derived from (11)
lines (A-C-E-F) can be used. To further simplify the control that the maximum power is Pbase/k. Besides, the minimum value
under the precondition of realizing maximum power transfer of Pbase/k should be larger than rated power Pr value, which is
capability, the two segments in the upper ZVS region (C-E and expressed as
E-F) can be linearized into one based on the starting point C Pbase / k max  ( nV2 ) 2 / (8 k max Lk f s )  Pr (31)
and ending point F. Finally, curve A shown in Fig. 10 (the solid Then the range of Lk can be deduced from (31).
line with triangles, A-C-F) is obtained, and the corresponding
expression is C. Magnetizing Inductance Lm
 k After choosing the suitable devices, the magnetizing
 2( k  k  2  a 2 k  1) k inductance can be designed by (21). Because k is a variable, to
 D  , case1 (28)
Dy _ A  k  a 2k  1  1 2k achieve ZVS in whole voltage range, Lm should be satisfied as
 2( k  a 2 k  1  1) L m  k m in L k / ( a 1  2 k m in ) (32)
2 k  2a 2 k  1  1
 D  , case2
 k  a 2k  1 k  a 2k  1 where a=(32LkCs)1/2/Ts as mentioned before.
The coefficients of (28) have square-root functions, it is still D. Implementation of The Control Strategy
complicated if digital controller such as DSP is used in To achieve ZVS in full load range containing forward and
engineering implementation. The middle point C can be reverse transmission, an absolute value of Dφ can be taken to the
modified to point D to make the calculation simple. As results, proposed control law. Based on (29), the control law can be
curve B shown in Fig. 10 (the solid line with circles, A-D-F) expressed as

0278-0046 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
Authorized licensed use limited to: Carleton University. Downloaded on August 26,2020 at [Link] UTC from IEEE Xplore. Restrictions apply.
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/TIE.2020.3014562, IEEE
Transactions on Industrial Electronics

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS

2[ Lk k  Lm k  ( k  2) Lm / k ] k Irms_pu Proposed in [13] Proposed in [26] [Link] Control Law A Control Law B
D y _ B1 = | D |  , | D | | D _ C | (33)
Lm  Lk k  Lm k 2k 1.6 k=0.4 1.6 k=0.5

2( Lm  Lk k  Lmk ) 2L k  Lm  2Lmk
Dy _ B 2 = | D |  k ,| D | | D _ C | (34) 1.2 1.2
( Lk  Lm )k ( Lk  Lm )k
Fig. 12 is the control block diagram to show how the unified 0.8 0.8

control is implemented. Voltage V2 is compared with the

0.4 0.4
reference voltage Vref to obtain the error which generates Dφ by
a PI compensator. Since Dφ=0.5 and Dφ=-0.5 correspond to the 0
0 0.2 0.4 0.6 0.8 1 1.2
0
0 0.2 0.4 0.6 0.8 1 1.2
maximum power in forward and reverse power transfer Irms_pu (a) P_pu (b)
respectively, Dφ is limited to the range [-0.5,0.5]. When 1.6 k=0.7 1.6 k=0.8
|Dφ|≤|Dφ_C|, Dy is obtained based on (33). In contrast, Dy is
obtained by (34) when |Dφ|>|Dφ_C|. And Dy is limited to the 1.2 1.2
range [0,1]. In addition, V1 and V2 are used to calculate the
0.8 0.8
voltage conversion ratio k.
0.5
S1 0.4 0.4
Vref D D S2
PI S3
1 0 0
|Dφ|≤|Dφ_C | S4 0 0.2 0.4 0.6 0.8 1 1.2 0 0.2 0.4 0.6 0.8 1 1.2
-0.5 PWM (c) P pu (d)
Dy=Dy_B1 Q1
Modulation Fig. 14. Comparisons of RMS current with different EPS control
V2 nV2 k
Q2
methods. (a) k=0.4. (b) k=0.5. (c) k=0.7. (d) k=0.8.
k 0 1 Q3
V1 V1 Dy Q4 addition, the differences of RMS values among control law A,
Dy=Dy_B2
|Dφ|>|Dφ_C |
control B and minimum Irms control law are very small, and the
0 differences become smaller as the P_pu increases, which implies
Fig. 12. Control block diagram with the proposed control law. that control law A and control law B can still obtain very low
RMS current. Besides, the curves between control law A and
VII. COMPARISONS control law B in Fig. 14 are almost coincident, which shows
Fig. 13 shows leakage RMS currents comparisons among that the RMS currents are almost the same under the two
minimum Irms control law and the two control laws after control laws. Hence, control law B can also achieve small
tradeoff between ZVS region and complexity. As seen, for leakage RMS current.
different power transfer, the operating RMS currents under the A further comparison is made between proposed control law
three control methods are very close. and minimum Irms control law. The specifications used in
Irms_pu [Link] Control Law A Control Law B comparison are: 1) V1=200-400V, V2=48V; 2) turns ratio n:
1 1.5 20/6; 3) rated power Pr=500W; 4) switching frequency
case 1 case 2
0.8 fs=250kHz; 5) GaN device GS66508T is chosen as the power
P_pu=0.5
P_pu=0.4 1 switches of V1 side and its junction capacitor is 65pF, and
0.6
P_pu=0.3 GS61008T are the power switches of V2 side with a 250pF
0.4 junction capacitor; 6) voltage conversion ratio k=0.4-0.8. Lk is
0.5
P_pu=0.2
designed to be 32μH by (31). Also, Lm is designed to be 142μH
0.2
P_pu=0.1 to satisfy (32). Take case when V1=400V as an example, the
00 0.2 0.4 0.6 0.8 1
00 0.2 0.4 0.6 0.8 1
comparisons of the calculated losses are shown in Fig. 15,
(a) Dy (b) including switches losses, inductor losses, transformer losses
Fig. 13. Curves of leakage RMS current versus Dy for three different and total losses. As seen, since switches conduction loss,
control laws in given power (k=0.6). (a) Case 1. (b) Case 2. inductor and transformer copper losses are closely relative to
Fig. 14 shows comparisons of RMS currents with different RMS current, their loss curves versus power are similar, and the
EPS control methods. The results show that RMS currents with losses with proposed control law are a bit larger than that with
the control laws proposed in [13] and [26] are significantly minimum Irms control law. Inductor iron loss with proposed
higher than the RMS current with minimum Irms control law. In control law is larger than that with minimum Irms control law,
10 7 30
[Link] [Link] [Link] [Link] [Link]
[Link] [Link] 6 [Link] [Link] 25 [Link]
8 [Link] [Link] [Link] [Link]
5 [Link] [Link] 20
[Link](W)
Ploss(W)

6
[Link] ss(W)

4
15
4 3
10
2
2 5
1
0 0 0
0 100
200 231 300 400 500 0 100 200 300 400 500 0 100 200 300 400 500
P(W) P(W) P(W)
(a) (b) (c)
Fig. 15. Comparisons of losses. (a) Switches losses. (b) Inductor and transformer losses. (c) Total losses.

0278-0046 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
Authorized licensed use limited to: Carleton University. Downloaded on August 26,2020 at [Link] UTC from IEEE Xplore. Restrictions apply.
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/TIE.2020.3014562, IEEE
Transactions on Industrial Electronics

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS

VIII. EXPERIMENTAL RESULTS

To validate the feasibility of the proposed solutions, a GaN
device-based laboratory prototype with the same specifications
used in theoretical analysis was built, as shown in Fig 16.
The steady-state waveforms under full load (500W), half
load (250W) and light load (50W) for different V1 voltages
including 200V, 300V, 400V are shown in Fig 17, Fig 18 and
Fig. 19, respectively. As seen from these figures, under a same
voltage conversion ratio k, the duty cycle of uAB increases as the
Fig. 16. The GaN device based experimental prototype. load increases, which agrees with the analysis. Furthermore,
too. For switches turn-off losses, due to the injection of transformer secondary current is is different from leakage
magnetizing current, the turn-off currents in secondary side are current ik after considering with a magnetizing inductance.
increased at light load, [Link] is higher than [Link]. As the Meanwhile, to directly judge whether ZVS is achieved or not,
input power increases, input current becomes larger, and the the current values at switch turn OFF instants are measured, and
effect of the magnetizing current decreases. When input power the corresponding calculated ZVS current limits as analyzed in
increases to 231W, hard switching of S3&S4 causes switch Section III.B are listed in Table IV. In this table, Izvs_s represents
turn-on loss. The transformer iron loss is not changed because the current values of is at the rising edge of uCD. If the condition
the voltage-second during half switch period at the V2 side of for Izvs_s is satisfied, ZVS for Q1 to Q4 is achieved. Izvs_1 and Izvs_2
the transformer is constant. The turn-on process of power stand for the current values of ik at rising edge and falling edge
device has three possible modes, including full ZVS, partial of uAB respectively. They respectively determine the
ZVS and non ZVS. It is worth noting that when analyzing achievement of ZVS for S3&S4 and S1&S2. By the way, the two
switches turn-on loss, the partial ZVS mode is not taken into operations for case 1 and case 2 shown in Fig. 2 can be
account to simplify the calculation, which causes a sharp identified by the waveforms of ik.
change as shown in Fig. 15(c). As seen in Fig.15 (c), the total TABLE IV
loss of minimum Irms control law is less than that of proposed CALCULATED ZVS CURRENT LIMITS FOR ZVS ACHIEVEMENT
control law in light load. With the addition of switch turn-on V1 200V 300V 400V
loss, the total loss of minimum Irms control law becomes larger Case 1 2 1 2 1 2
in heavy load. Therefore, due to the complex formula of IZVS_S (A) ≥0 ≥0 ≥0 ≥0 ≥0 ≥0
minimum Irms control law of (25) and (27) and the related issues IZVS_1 (A) ≤0 ≤ -0.65 ≤0 ≤ -0.87 ≤ -0.36 ≤ -1.08
caused by hard switching, the proposed control law is more IZVS_2 (A) ≥ 0.31 ≥ 0.31 ≥ 0.16 ≥ 0.16 ≥0 ≥0
preferred.

uAB(500V/div) uCD(100V/div) uAB(500V/div) uCD(100V/div) uAB(500V/div) uCD(100V/div)

6.2A
ik(10A/div)
-6.1A -3.2A 6.4A ik(5A/div) ik(5A/div)
6.2A -1.8A
is(20A/div) is(25A/div)
is(50A/div)
0.4A
13.3A 4.4A
(a) (b) (c)
Fig. 17. Experimental waveforms under full load. (a)200V. (b)300V. (c)400V.

uAB(500V/div) uCD(100V/div) uAB(500V/div) uCD(100V/div) uAB(500V/div) uCD(100V/div)

2.5A 4.2A 4.6A

ik(5A/div) ik(5A/div) ik(5A/div)
-2.6A -1.0A
is(10A/div) is(20A/div) -1.2A
is(20A/div)
4.6A
0.4A 2.0A
(a) (b) (c)
Fig. 18. Experimental waveforms under half load. (a)200V. (b)300V. (c)400V.

uAB(500V/div) uCD(100V/div) uAB(500V/div) uCD(100V/div) uAB(500V/div) uCD(100V/div)

2.5A
ik(2A/div) ik(2A/div) 3.0A ik(5A/div)
-0.9A 1.8A
is(2.5A/div) is(10A/div) -1.4A is(10A/div) -1.6A

0.3A 4.6A 5.4A

(a) (b) (c)
Fig. 19. Experimental waveforms under light load. (a)200V. (b)300V. (c)400V.

0278-0046 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
Authorized licensed use limited to: Carleton University. Downloaded on August 26,2020 at [Link] UTC from IEEE Xplore. Restrictions apply.
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/TIE.2020.3014562, IEEE
Transactions on Industrial Electronics

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS

Through making point-to-point comparisons of the denoted Load Switching u2(50V/div)

current values in the experiment waveforms and the values
shown in Table IV, the conclusion can be made that the current 200W 400W uCD(100V/div) 200W
values denoted from Fig. 17 to Fig. 19 under different input
voltages and loads all meet the conditions in Table IV, which ik(5A/div)
demonstrates the realization of ZVS for all power switches.
One case when V1 =300V is taken to show the specific soft
is(20A/div)
switching waveforms under different loads, as displayed from
Fig. 20 to Fig. 22. ugs is the driver signal, and uds is the
drain-source voltage. Fig. 20 shows driver signal and Fig. 23. Load switching waveforms.
drain-source voltage of switch S1. S1 is turned on at peak current, 250kHz operating frequency.
and the current is significantly less than zero, so ZVS is In addition, a series of comparisons among the proposed
achieved from light load to full load. Fig. 21 illustrates ZVS solution and recently reported solutions in the literature are
occurrence of switch S3. In all the loads, the current at the time made, and the results are listed in Table V. Compared with the
of S3 turn-on is large, which demonstrates the ZVS existing solutions, the proposed solution operates at 250kHz
achievement of S3. Fig.22 shows driver signal and drain-source switching frequency and only needs four operation modes,
voltage of switch Q2. As seen, before ugs changes from zero to which is feature of high switching frequency and a medium
positive, uds already goes to zero, so ZVS of Q2 can be obtained. control complexity without auxiliary components. Besides, the
With the intention to show the dynamic performance of the highest efficiency is 96.3% and the power density is 29W/inch3.
experiment prototype, the dynamic waveforms, when output 1
power is switched from 200W to 400W and then switched back η
to 200W, are shown in Fig. 23. As seen, the output voltage V2 0.95
can keep constant when the load changes, and the leakage 0.9
current is smoothly changed. Fig. 23 indicates that the control is
a seamless closed loop control. The experimental efficiency 0.85
curves of the converter with the proposed control law for 0.8
different V1 side voltages are illustrated in Fig. 24. As shown, at k=0.8 (V1=200V)
light load, high efficiency will be obtained if the input voltage 0.75 k=0.53(V1=300V)
k=0.4 (V1=400V)
is low, mainly because of reduced reactive current. At heavy
0.7
load, it has higher efficiency when the input voltage is 300V. In 0 0.2 0.4 0.6 0.8 1 P pu
addition, the maximum efficiency point is larger than 96% with Fig. 24. Measured efficiency curves with different input voltages.

ZVS ugs(5V/div) ZVS ugs(5V/div) ZVS ugs(5V/div)

uds(300V/div) uds(300V/div) uds(300V/div)

ik(2A/div) ik(5A/div) ik(5A/div)

uCD(100V/div) uCD(100V/div) uCD(100V/div)

(a) (b) (c)

Fig. 20. Soft-switching waveforms for S1&S2. (a)light load. (b)half load. (c)full load.

ZVS ugs(5V/div) ZVS ugs(5V/div) ZVS ugs(5V/div)

uds(300V/div) uds(300V/div) uds(300V/div)

ik(2A/div) ik(5A/div) ik(5A/div)

uCD(100V/div) uCD(100V/div) uCD(100V/div)

(a) (b) (c)

Fig. 21. Soft-switching waveforms for S3&S4. (a)light load. (b)half load. (c)full load.

ZVS ugs(5V/div) ZVS ugs(5V/div) ZVS ugs(5V/div)

uds(50V/div) uds(50V/div) uds(50V/div)

ik(10A/div) ik(20A/div)
ik(10A/div)

uAB(500V/div) uAB(500V/div) uAB(500V/div)

(a) (b) (c)

Fig. 22. Soft-switching waveforms for Q1&Q2. (a)light load. (b)half load. (c)full load.

0278-0046 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
Authorized licensed use limited to: Carleton University. Downloaded on August 26,2020 at [Link] UTC from IEEE Xplore. Restrictions apply.
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/TIE.2020.3014562, IEEE
Transactions on Industrial Electronics

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS

TABLE V
COMPARISONS OF THE PROPOSED METHOD AND THE EXISTING METHOD FOR DAB
Frequency Power Density Dynamic Behavior
Article Addition Component Efficiency Control Complexity Modulation
(kHz) (W/inch3) On/Off-line Operation modes
[27] No 200 39 95.26% online 2 modes low SPS
[18] Two capacitors and a switch 100 5-15 97.5% online 2 modes low SPS
[13] No 20 —— 91% online 4 modes medium EPS
[28] No 2.5 —— 93% offline 4 modes high TPS
[12] No 50 4-6 95% online 9 modes high TPS
This paper No 250 29 96.3% online 4 modes medium EPS
Bidirectional DC–DC Converter for Power Distribution in Microgrid,” in
IEEE Transactions on Power Electronics, vol. 27, no. 11, pp. 4667-4680,
IX. CONCLUSION AND FUTURE WORK Nov. 2012.
[9] H. Bai and C. Mi, “Eliminate Reactive Power and Increase System
In this paper, an optimized EPS control is proposed for DAB Efficiency of Isolated Bidirectional Dual-Active-Bridge DC–DC
dc-dc converters. The ZVS regions based on the ZVS working Converters Using Novel Dual-Phase-Shift Control,” in IEEE
modes considering switch junction capacitors are analyzed, and Transactions on Power Electronics, vol. 23, no. 6, pp. 2905-2914, Nov.
it shows that the ZVS region is separated into two parts with 2008.
[10] F. Krismer and J. W. Kolar, “Closed Form Solution for Minimum
conventional control methods. To achieve ZVS for all power Conduction Loss Modulation of DAB Converters,” in IEEE Transactions
switches in full load range and obtain seamless transition, the on Power Electronics, vol. 27, no. 1, pp. 174-188, Jan. 2012.
magnetizing current is utilized, and a power control law is [11] F. Liu, X. Sun, J. Feng, J. Wu and X. Li, “The improved dual active
proposed. This optimized control law also has low RMS current bridge converter with a modified phase shift and variable frequency
control,” 2018 IEEE Applied Power Electronics Conference and
and maximum power transmission capability. Finally, Exposition (APEC), San Antonio, TX, 2018, pp. 814-819.
experimental results with a 250kHz, 500W laboratory [12] S. Chakraborty and S. Chattopadhyay, “Fully ZVS, Minimum RMS
prototype demonstrate that the proposed solutions are effective. Current Operation of the Dual-Active Half-Bridge Converter Using
In addition to ZVS achievement, high power density is also Closed-Loop Three-Degree-of-Freedom Control,” in IEEE Transactions
on Power Electronics, vol. 33, no. 12, pp. 10188-10199, Dec. 2018.
an important demand in some engineering applications. GaN [13] H. Shi et al., “Minimum-Backflow-Power Scheme of DAB-Based
based devices are feature of low on-resistance, low junction Solid-State Transformer With Extended-Phase-Shift Control,” in IEEE
capacitance and very small turn-off energy. They are not only Transactions on Industry Applications, vol. 54, no. 4, pp. 3483-3496,
beneficial to achieve ZVS and improve efficiency for DAB July-Aug. 2018.
[14] B. Long et al., “Light Load Efficiency Improvement Method for
converter, but also helpful for power density improvement. To GaN-Based DAB Converter with Hybrid Discontinuous Current Mode,”
increase the power density, future modifications can be made to 2019 10th International Conference on Power Electronics and ECCE
redesign the magnetics using plane transformer, and also to Asia (ICPE 2019 - ECCE Asia), Busan, Korea (South), 2019, pp.
redesign 3-D structure to make the prototype more compact. 2415-2423.
[15] G. Oggier, G. O. García and A. R. Oliva, “Modulation strategy to operate
Also, the impact of turns ratio n on the close-loop system the dual active bridge DC-DC converter under soft switching in the whole
performance and the designing of the PI controller are worth of operating range,” in IEEE Transactions on Power Electronics, vol. 26, no.
further study in the future. 4, pp. 1228-1236, April 2011.
[16] Z. Guo, “Modulation Scheme of Dual Active Bridge Converter for
Seamless Transitions in multi Working Modes Compromising ZVS and
REFERENCES Conduction Loss,” in IEEE Transactions on Industrial Electronics. doi:
[1] B. Zhao, Q. Song, W. Liu and Y. Sun, “Overview of Dual-Active-Bridge 10.1109/TIE.2019.2945270
Isolated Bidirectional DC–DC Converter for High-Frequency-Link [17] W. Choi, K. Rho and B. Cho, “Fundamental Duty Modulation of
Power-Conversion System,” in IEEE Transactions on Power Electronics, Dual-Active-Bridge Converter for Wide-Range Operation,” in IEEE
vol. 29, no. 8, pp. 4091-4106, Aug. 2014. Transactions on Power Electronics, vol. 31, no. 6, pp. 4048-4064, June
[2] R. P. Twiname, D. J. Thrimawithana, U. K. Madawala and C. A. Baguley, 2016.
“A Dual-Active Bridge Topology With a Tuned CLC Network,” in IEEE [18] M. Yaqoob, K. Loo and Y. M. Lai, “Fully Soft-Switched
Transactions on Power Electronics, vol. 30, no. 12, pp. 6543-6550, Dec. Dual-Active-Bridge Series-Resonant Converter With
2015. Switched-Impedance-Based Power Control,” in IEEE Transactions on
[3] S. Poshtkouhi and O. Trescases, “Flyback Mode for Improved Power Electronics, vol. 33, no. 11, pp. 9267-9281, Nov. 2018.
Low-Power Efficiency in the Dual-Active-Bridge Converter for [19] Y. Shen, X. Sun, W. Li, X. Wu and B. Wang, “A Modified Dual Active
Bidirectional PV Microinverters With Integrated Storage,” in IEEE Bridge Converter With Hybrid Phase-Shift Control for Wide Input
Transactions on Industry Applications, vol. 51, no. 4, pp. 3316-3324, Voltage Range,” in IEEE Transactions on Power Electronics, vol. 31, no.
July-Aug. 2015. 10, pp. 6884-6900, Oct. 2016.
[4] S. Inoue and H. Akagi, “A bidirectional DC–DC converter for an energy [20] Z. Guo, D. Sha and K. Song, “Output-Series Connected Dual Active
storage system with galvanic isolation,” in IEEE Transactions on Power Bridge Converters for Zero-Voltage Switching Throughout Full Load
Electronics, vol. 22, no. 6, pp. 2299-2306, Nov. 2007. Range by Employing Auxiliary LC Networks,” in IEEE Transactions on
[5] M. N. Kheraluwala, R. W. Gascoigne, D. M. Divan and E. D. Baumann, Power Electronics, vol. 34, no. 6, pp. 5549-5562, June 2019.
“Performance characterization of a high-power dual active bridge [21] Z. Qin, Y. Shen, P. C. Loh, H. Wang and F. Blaabjerg, “A Dual Active
DC-to-DC converter,” in IEEE Transactions on Industry Applications, Bridge Converter With an Extended High-Efficiency Range by DC
vol. 28, no. 6, pp. 1294-1301, Nov.-Dec. 1992. Blocking Capacitor Voltage Control,” in IEEE Transactions on Power
[6] M. H. Kheraluwala and R. W. De Doncker, “Single phase unity power Electronics, vol. 33, no. 7, pp. 5949-5966, July 2018.
factor control for dual active bridge converter,” Conference Record of the [22] H. Wen, B. Su and W. Xiao, “Design and performance evaluation of a
1993 IEEE Industry Applications Conference Twenty-Eighth IAS Annual bidirectional isolated dc-dc converter with extended dual-phase shift
Meeting, Toronto, Ontario, Canada, 1993, pp. 909-916 vol.2. scheme,” in IET Power Electronics, vol. 6, no. 5, pp. 914-924, May 2013.
[7] C. Mi, H. Bai, C. Wang and S. Gargies, “Operation, design and control of [23] G. G. Oggier, G. O. GarcÍa and A. R. Oliva, “Switching Control Strategy
dual H-bridge-based isolated bidirectional DC-DC converter,” in IET to Minimize Dual Active Bridge Converter Losses,” in IEEE
Power Electronics, vol. 1, no. 4, pp. 507-517, December 2008. Transactions on Power Electronics, vol. 24, no. 7, pp. 1826-1838, July
[8] B. Zhao, Q. Yu and W. Sun, “Extended-Phase-Shift Control of Isolated 2009.

0278-0046 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
Authorized licensed use limited to: Carleton University. Downloaded on August 26,2020 at [Link] UTC from IEEE Xplore. Restrictions apply.
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/TIE.2020.3014562, IEEE
Transactions on Industrial Electronics

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS

[24] F. An, W. Song and K. Yang, “Optimised power control with extended Yao Sun (M'13) received the B.S., M.S. and
phase shift in dual-active-bridge dc–dc converters,” in Electronics Letters, Ph.D. degrees from the School of Information
vol. 54, no. 10, pp. 651-653, 17 5 2018. Science and Engineering, Central South
[25] S. S. Shah, V. M. Iyer and S. Bhattacharya, “Exact Solution of ZVS University, Changsha, China, in 2004, 2007 and
Boundaries and AC-Port Currents in Dual Active Bridge Type DC–DC 2010, respectively. He is currently with the
Converters,” in IEEE Transactions on Power Electronics, vol. 34, no. 6, School of Information Science and Engineering,
pp. 5043-5047, June 2019. Central South University, China, as a professor.
[26] H. Tao, A. Kotsopoulos, J. L. Duarte and M. A. M. Hendrix, His research interests include matrix
“Transformer-Coupled Multiport ZVS Bidirectional DC–DC Converter converter, micro-grid and wind energy
With Wide Input Range,” in IEEE Transactions on Power Electronics, conversion system
vol. 23, no. 2, pp. 771-781, March 2008.
[27] M. Zhu, C. Shao, S. Wang, L. Hang, Y. He and S. Fan, “System Design of
Dual Active Bridge (DAB) Converter Based on GaN HEMT Device,” Mei Su was born in Hunan, China, in 1967. She
2019 22nd International Conference on Electrical Machines and Systems received the B.S., M.S. and Ph.D. degrees from
(ICEMS), Harbin, China, 2019, pp. 1-6. the School of Automation, Central South
[28] O. M. Hebala, A. A. Aboushady, K. H. Ahmed and I. Abdelsalam, University, Changsha, China, in 1989, 1992 and
“Generic Closed-Loop Controller for Power Regulation in Dual Active 2005, respectively. Since 2006, she has been a
Bridge DC–DC Converter With Current Stress Minimization,” in IEEE School of Automation Central South University.
Transactions on Industrial Electronics, vol. 66, no. 6, pp. 4468-4478, Her research interests include matrix
June 2019. converter, adjustable speed drives, and wind
energy conversion system.

Guo Xu (M'15) received the B.S. degree in

electrical engineering and automation and the
Ph.D. degree from the Beijing Institute of
Technology, Beijing, China, in 2012 and 2018,
respectively. From 2016 to 2017, he was a
Visiting Student with the Center for Power
Electronics System, Virginia Polytechnic Institute
and State University, Blacksburg, VA, USA.
Since 2018, he has been with the School of
Automation, Central South University, Changsha,
China, where he is currently an Associate Professor.
His research interests include modeling and control of power
electronics converters, high-efficiency power conversion, and magnetic
integration in power converters.

Liting Li was born in Guangdong, China, in 1996.

She received the B.S. degree in school of
electrical and information engineering from
Changsha University of Science and Technology,
Changsha, China, in 2019. She is currently
pursuing a M.S. degree in School of Automation
from Central South University, Changsha, China.
Her current research include topology and
controlling of power converters and high
efficiency power conversion.

Xiaoying Chen was born in Fujian, China, in

1995. He received the B.S. degree in electronic
engineering from Central South University,
Changsha, China, in 2017, where he is currently
working toward the Ph.D. degree. His research
interests include modeling and control of power
electronics converters and high-efficiency power
conversion.

Yonglu Liu (S’16) was born in Chongqing, China, in

1989. He received the B.S., M.S., and Ph.D. degrees in
electrical engineering from Central South University,
Changsha, China, in 2012, 2015, and 2017
respectively. He has been an Associate Professor with
School of Automation, Central South University,
China.
His research interests include power electronics and
renewable energy power conversion systems.

0278-0046 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
Authorized licensed use limited to: Carleton University. Downloaded on August 26,2020 at [Link] UTC from IEEE Xplore. Restrictions apply.