This article has been accepted for publication in IEEE Transactions on Transportation Electrification.

This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TTE.2022.3229076

A Model Predictive Control Scheme without Current

Sensor of Dual Active Bridge DC-DC Converters:
Improving Dynamic Performance and Reducing
Hardware Cost
Yaru Deng, Student Member, IEEE, Wensheng Song, Senior Member, IEEE, Shuai Yin, Ming Zhong,
Kexin Yang, and Xiaoyun Feng

Abstract—Dual active bridge DC-DC (DAB) converters have a topology, convenient switching of power flow, 3-H (high
wide range of applications. However, the stability of DAB frequency, high efficiency, and high-power density), and soft
converters may be destroyed by the output voltage disturbance, switching performance [3]. It is widely used as the interface
which is caused by variations in input voltage and load conditions. between different DC-bus voltage levels and has broad
To improve dynamic performance of DAB converters, many
application prospects, such as energy storage systems [4]-[5],
optimization schemes are proposed and require additional voltage
and current sensors. In this paper, a model predictive control DC microgrids [6], vehicle-to-grid applications (V2G) [7],
(MPC) method is proposed to achieve fast dynamic performance railway traction system [8], and electric vehicle (EV) [9]-[10],
in load-current sensorless condition, which leads to cost reduction as shown in Fig. 1. Actually, in the power conversion
of the hardware system. The proposed MPC method is insensitive application, one of the main difficulties is the input voltage
to circuit parameters and easy to implement. Moreover, combined fluctuation and the varying-load operation. The disturbances
with inductor current stress optimization (CSO), the proposed may cause output voltage ripple and adversely affect the
method can achieve multi-object optimization of DAB converters, stability of the DAB system. Thereby, rapid dynamic response
including improving dynamic response, reducing the cost of the and robustness of DAB converters are essential requirements to
hardware system, etc. Firstly, the transmission power model under
achieve the mature application of DAB converters [11]. For
dual-phase-shift (DPS) of DAB converters is analyzed. Then, some
typical methods, such as the PI control, the load current example, pulsed power loads such as electromagnetics are
feedforward (LCFF) control, and the sliding mode based direct connected to power electronic systems through DAB
power control (SM-DPC), are introduced. Finally, an experiment converters, which are expected to have faster dynamic response
phototype of DAB converters is developed, and comprehensive [12].
experiments with resistive load and 3-phase inverter load are AC Bus

employed to verify the correctness and effectiveness of the

DC-AC
proposed method.
Renewable Energy

AC-DC DC-DC
Index Terms—dual active DC-DC converters (DAB), model WGE Energy Storage System

predictive control (MPC), load current sensor, dynamic response DC-DC DC-DC Converter DC-AC

performance Fuel cell

Railway Traction System

AC-DC DC-DC

PV
DC Bus DC Bus Electric Automobile
I. INTRODUCTION
Fig. 1. The application of DAB converters.
The DC-DC converter is an important equipment for electric
energy conversion, which can complete the power conversion To improve the dynamic performance of the DAB converter,
between various levels of DC-bus [1]-[2]. Dual active bridge some optimizing methods have been proposed [13]-[27], such
DC-DC (DAB) converters have the advantages of symmetrical as the conventional PI control, and direct inductor current
control [13]. However, the dynamic response of DAB
Manuscript received July 25, 2022; revised September 23, 2022; accepted converters is limited with traditional PI and PI-based control
December 04, 2022. Date of publication XXX, XXXX. This work is supported
by the National Natural Science Foundation of China under Grant 52022084
schemes [14], and these methods are always sensitive to
and 61733015, and in part by Sichuan Youth Science and Technology parameters, and most of them require offline calculation. An
Innovation Research Team under Grant 22CXTD0055. Recommended for adaptive PI scheme with the dynamic linearized harmonic
publication by Associate Editor XXXXXX. (Corresponding authors: model is proposed in [15]. But the linearized harmonic model
Wensheng Song and Xiaoyun Feng).
Y. Deng, W. Song, S. Yin, K. Yang and X. Feng are with the School of is relatively complex and difficult to be implemented in the
Electrical Engineering, Southwest Jiaotong University, Chengdu, 611756, digital controller. Besides, a linear-quadratic regulator control
China (e-mail: yaru_deng_swjtu@[Link]; songwsh@[Link]; based on linear matrix inequalities (LMI-LQR) [16], and
ys2434984249@[Link]; ykxin@[Link]; fengxy@[Link]).
M. Zhong was with the School of Electrical Engineering, Southwest sliding-mode-based direct power control (SM-DPC) [17] are
Jiaotong University, Chengdu, 611756, China, and he is currently with Huawei proposed to improve the dynamic performance. However, the
Technologies Co., Ltd., Chengdu, 518129, China (e-mail: output current sensor and input voltage sensor have been
m_zhong_2018@[Link]).
introduced, which leads to the increase of hardware cost. In

© 2022 IEEE. Personal use is permitted, but republication/redistribution requires IEEE [Link] [Link] for more information.
Authorized licensed use limited to: UNIVERSIDAD DE VIGO. Downloaded on April 15,2023 at [Link] UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for publication in IEEE Transactions on Transportation Electrification. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TTE.2022.3229076

renewable energy systems and the railway locomotive traction predictive control (MDCS-MPC) scheme is proposed for pulsed
application, to realize multi-level voltage conversion and power load and boosting the dynamic capability of DAB
reduce the withstand voltage level of power devices, a large converters [35], but it is extremely complex and difficult for
amount of DAB converters or multi-module cascaded DAB high-frequency applications. A digital predictive current
converters are required [18]. The hardware cost caused by the control is proposed in [36], which requires AC current sampling
additional voltage or current sensors, and computation burden and heavy computation burden. Furthermore, a nonlinear model
of the processor is another main difficulty of DAB converters. predictive control method is proposed in [37], which is
Based on the strategies of direct inductor current control, a insensitive to parameters. In contrast, the conventional model
fast transient boundary control (FTBC) [19] is proposed to predictive control, such as FCS-MPC and CCS-MPC are
achieve excellent dynamic performance. However, the FTBC unsuitable for DAB converters, and the applications have not
requires five sensors for input voltage, output voltage, output been so intensively explored. Most of the existing MPC
current, and primary-side and secondary-side inductor current. schemes applied in DAB converters are sensitive to system
A simplified load-feedforward control design is proposed in parameters and difficult to implement because of the abundant
[20], which can enhance the dynamic response. But the output computation burden [38]-[39].
voltage, output current, and inductor current sensors are To solve the two main difficulties of DAB converters:
required. Then, the output voltage tracking and dynamic improving dynamic performance and reducing hardware cost, a
performance have been improved with the load current MPC method without current sensor is proposed in this paper.
feedforward (LCFF) control method [21], but it shows poor The transmission power predictive model of the DAB converter
performance with parameter uncertainty, such as disturbances under dual phase shift (DPS) modulation is developed. When
of the input voltage and the output current. Besides, because of input voltage/load steps, the output voltage behavior can be
output voltage and current sensors, the system hardware cost is predicted without any current sensor. This paper is organized as
still high. Based on direct power control (DPC) [22], a virtual- follows, Firstly, the operating principle of DPS modulation and
direct power control (VDPC) method [23] is proposed to the transmission power model are analyzed, besides the current
improve the effectiveness of the power-based control methods stress optimization (CSO) model of the DAB converter is
The inductor current sensor is eliminated and the dynamic discussed in Section Ⅱ. Secondly, in Section Ⅲ, some general
performance under load disturbance is improved. But the load control methods, such as the conventional PI control, load
current measured is not accurate in a switching period, which current feedforward (LCFF) [21], and the sliding mode based
will affect the performance of the DAB converter, and a large direct power control (SM-DPC) [17], are analyzed for
capacitor is essential to avoid the ripple. comparison. The operating principles and characteristics of
According to [24], the existing schemes proposed to improve these three control schemes are introduced and compared in
the dynamic performance of DAB converters can be divided detail. In addition, a current sensorless-based model predictive
into parallel structure and series structure. However, regardless control scheme for DAB converters is proposed in Section Ⅳ,
of the structure, the load current sensor is indispensable. Thus, and the cost function of transmission power is established.
to decrease the hardware cost caused by the load current sensor, Moreover, the CSO is combined with the proposed method,
a current sensor-less control method [25] is proposed, and the which can ensure excellent dynamic and steady-state
nonlinear disturbance observer is applied to obtain excellent performance for current sensorless applications. Then, in
dynamic performance with current sensorless. Similarly, a Section Ⅴ, an experimental prototype of the DAB converter is
disturbance-observer-based control (DOBC) method is developed and a comparison experiment is employed to verify
proposed in [26], which has high computational complexity and the effectiveness of the proposed method. Finally, the
is difficult to implement. However, in these two nonlinear conclusions are summarized in Section VI. And the main
control schemes, a positive feedback value is introduced into contributions of the paper are as follows:
the control value, which will lower model accuracy of DAB 1) When input voltage or load suddenly changes, the
converters. The robustness and the dynamic performance of the proposed MPC method can achieve fast dynamic
DAB converter are improved with sliding mode control (SMC) performance in load-current sensorless condition,
method [27]. However, chattering and degradation of tracking which leads to cost reduction of the hardware system.
performance are the problems brought by SMC. 2) The proposed MPC method shows strong stability for
Model predictive control (MPC) has been widely applied in load disturbances and robustness for parameter
power electronic converters [28]-[29], which has a rapid sensitivity, which is insensitive to circuit parameters.
dynamic response and simple implementation. A finite control And, the results of comparison experiments reflect the
set-model predictive control (FCS-MPC) is proposed for AC effectiveness of the proposed MPC for large
power conversion in [30]-[31], which can achieve a fast disturbances and parameters mismatch.
dynamic response. However, the FCS-MPC is not feasible to 3) The proposed MPC method is easy to implement in
apply in DAB converters directly [32]-[33]. A continuous low-complexity, and can avoid the chattering effect and
control set-model predictive control (CCS-MPC) is proposed in degradation of tracking performance, which are serious
[34], which can avoid voltage overshoot in the transient process. issues in DAB converters with SMC control.
Notwithstanding, these methods are not suitable for DAB 4) The proposed MPC method can combine with CSO and
converters. Then, a moving-discretized-control-set model multiple phase shift modulation for efficiency

© 2022 IEEE. Personal use is permitted, but republication/redistribution requires IEEE [Link] [Link] for more information.
Authorized licensed use limited to: UNIVERSIDAD DE VIGO. Downloaded on April 15,2023 at [Link] UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for publication in IEEE Transactions on Transportation Electrification. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TTE.2022.3229076

optimization. In other words, the proposed MPC To establish the state equation of the DAB converter, the
method is suitable for various phase shift modulations. output capacitor voltage Uo and inductor current iL are selected
as state variables. Based on the voltage and current law of
II. FUNDAMENTAL OF DAB AND CURRENT STRESS Kirchhoff, the model of the DAB converter is
OPTIMIZATION STRATEGY di
L L = uab − Nucd，Co
dU o
= i2 − io (1)
dt dt
A. Fundamental of DAB Converter under DPS Modulation The voltage conversion ratio is defined as K =Ui / NUo. When
As shown in Fig. 2, the topology of the DAB converter K ≥1, as shown in Fig. 3, the transmission power model P and
mainly consists of two symmetrical full bridges and a high- the inductor current peak value Ip of the DAB converter under
frequency isolation transformer. Ui is the input voltage and Uo DPS modulation [40] can be expressed as follows, respectively.
is the output voltage. uab and ucd are the terminal voltages at the  NTU iU o  1 2
 P = 2 L   (1 −  ) − 2   (   ) (2)
input and output terminals of the transformer, respectively. N is  

the transformer voltage ratio. L represents the auxiliary  P = NTU iU o   (1 −  ) − 1  2  (   )
 2L  2 
inductance and the leakage inductance of the transformer, iL is NTU o
I p = max{ iL (t ) } =  2 + ( K − 1)(1 −  ) (3)
the inductor current of L, and iS is the input current. Ci and Co 4L 
represent the input and output capacitors, respectively. i2 is the To simplify the calculation and facilitate the analysis of the
output current of the secondary-side H-bridge and io is the model of the DAB converter, the normal values of the
output current. transmission power and the inductor current are defined as
i2 io NTU iU o NTU o
is PB = ，I B = (4)
Q1 Q4 Q5 Q8 8L 8L
L N:1
iL
Ui Ci ucd Co R Uo Based on (2)-(4), the per-unit values of transmission power
uab
P,pu and inductor current peak Ip,pu of the DAB converter under
Q2 Q3 Q6 Q7
DPS modulation can be derived as
4 (1 −  ) − 2   
 2

Fig. 2. The topology of the DAB converter. P,pu =  (5)

4 (1 −  ) − 2   
2

Q1 Q2 Q 1 Q2
I p,pu = 4 + 2 ( K -1)(1 −  ) (6)
φT/2 t φT/2 t
Q3 Q4 Q3 Q4

Q5 Q6
t t B. Current Stress Optimization Strategy of DAB
Q5 Q6
θT/2 t θT/2
Q7 Q8
t
Q 7 Q8
t The CSO strategy, also known as the inductor current peak
uab
t uab
t optimization, can effectively improve the power transmission
ucd ucd
efficiency of the DAB converter. To reduce the transmission
t t
power loss and switching loss, based on the transmission power
iL iL
T t T t model in (5) and the inductor current peak in (6), the CSO
(a) φ ≤ θ (b) θ ≤ φ
solution is expressed as
Fig. 3. The timing diagram of the DAB converter under DPS modulation. min I p,pu ( , )，s.t.P,pu ( , ) = Pset (7)
 ,

Phase shift modulation is one of the classical control methods Where Pset is the desired per-unit value of transmission
of the DAB converter, such as single phase shift (SPS), power. Then, the Lagrange function of the CSO strategy of the
extended phase shift (EPS), DPS, and triple phase shift (TPS). DAB converter can be defined as
The degree of optimization and functional diversity of the DAB E = I p,pu +  ( P,pu − Pset ) (8)
converter is positively related to the number of controllable Where E is the Lagrange function, and λ is the Lagrange
phase shift variables [2]. Combined with the characteristics of multiplier. The per-unit value of the inductor current peak is the
phase shift modulation, DPS is chosen to achieve CSO and objective function, and the per-unit value of transmission power
efficiency improvement of the DAB converter in this paper. is the equivalent constraint.
The switching pulse, square wave voltage uab and ucd, and When the first derivative of (8) is equal to zero, the inductor
inductor current iL are shown in Fig. 3. θ represents the outer peak current reaches its extreme value. In this case, the inner
phase shift ratio and φ is the inner phase shift ratio, T is the phase φ can be represented by the outer phase θ, and its
switching period. relational expression is
TABLE I  ( K − 1)(1 − 2 )
   (9)
INDEX TERMS  2
=
4
Symbols Definitions Symbols Definitions 1 −  −  
 ( K − 1)
2
Ui / Uo Input / Output voltage iL Inductor current
Input / Output terminal Secondary-side H-bridge Based on the transmission power model and CSO strategy,
uab / ucd i2
voltage current
iS / io Input / Output current R Load resistance
the optimal phase shift ratio of the DAB converter under DPS
L Auxiliary inductor N Transformer ratio can be derived as (10)-(11), which consists of K and the per-
Input/Output filter Switching period / unit value of transmission power P,pu.
Ci / Co T / fs
capacitor frequency

© 2022 IEEE. Personal use is permitted, but republication/redistribution requires IEEE [Link] [Link] for more information.
Authorized licensed use limited to: UNIVERSIDAD DE VIGO. Downloaded on April 15,2023 at [Link] UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for publication in IEEE Transactions on Transportation Electrification. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TTE.2022.3229076

 P,pu ( K − 1) simplest method to regulate the output voltage. The

 =
 2 ( K + 3) K +2 K − 3
2
(10) conventional feedback control is easy to implement and the
 0  P,pu 
 = 1 − P,pu ( K − 1) − 2 P,pu 2K 2 hardware cost is low because only one output voltage sensor is

 2 ( K + 3 ) ( K − 1) ( K + 3) required. The PI controller is used to minimize the steady-state
 1 − P,pu
error, but the dynamic performance is terrible.
1
 = −
 2 2 ( K 2 − 2 K + 3) K 2 + 2K − 3 (11) B. Load Current Feedforward Control
  P,pu  1
 = ( K − 1) 1 − P,pu 2K 2
To further improve the dynamic performance of the DAB
 2 ( K 2 − 2 K + 3)
 converter, PI controller with the load current feedforward is
Subsequently, the minimum per-unit value of the inductor added to improve the dynamic performance of the DAB
peak current is converter through measurement and compensation in the
 K 2 + 2K − 3 presence of disturbances. LCFF is the classical one of the
 2P,pu ( K − 1)( K + 3) 0  P,pu 
I p,pu = 2K 2 (12) feedforward control methods, which is shown in Fig. 5.
2 K − 2(1 − P )( K 2 − 2 K + 3) K 2 + 2K − 3
 P,pu  1
Load current
kIo
Feedforward θ
Δ
 ,pu
2K 2 U*o PI θ0 θ
Cotroller Dual- Q1-Q4

...
Phase-Shift
III. GENERAL CONTROL METHODS OF DAB CONVERTERS Uo Current Stress ᵠ Modulation Q5-Q8

...
Ui Optimization
Eq. (9)
There are various optimization control methods of DAB
Fig. 5. The control diagram of the LCFF scheme.
introduced in Section Ⅰ. However, the existing optimization
methods all aim at the efficiency improvement of DAB Based on the control diagram of LCFF in Fig. 5, the phase
converters. This section mainly analyzes the dynamic response shift θ between the primary and secondary bridges can be
performance of the existing optimization control methods, such expressed as follows:
as the conventional feedback control, the LCFF control, the  = o +  (14)
SM-DPC control, and the MPC control. To compare the Where Δθ is the compensation phase shift of the feedforward
dynamic performance improvement of the DAB converter control method, which can be expressed as
under different control methods, the conventional feedback  = kI o (15)
control, the LCFF control, and the SM-DPC control are selected where k is the feedforward ratio and Io is the load current. The
for comparative experiments with the proposed method. Next, dynamic performance of the DAB converter can be further
the operating principle and main characteristics of the improved with the feedforward method under the load change.
conventional feedback control and the LCFF control, and the Combined feedforward plus feedback control can improve
SM-DPC control are analyzed. the dynamic performance over simple feedback control as the
A. The Conventional Feedback Control disturbance can be measured and counterbalanced before it
affects the process output. The LCFF control can significantly
The conventional feedback control is the simplest method reduce the output impedance and present a perfect load current
with output voltage tracking. As shown in Fig. 4, the phase shift disturbance rejection capability. However, its performance
ratio θ, between the primary and secondary bridges is obtained deteriorates quickly when there are parameter uncertainties.
from the proportional-integral (PI) controller output, the input
of the PI controller is the error of the desired and actual value C. Sliding Mode Based Direct Power Control
of the output voltage. Then, the inner phase shift ratio φ is Furthermore, there more nonlinear control methods are
derived from (9). applied to DAB, the SM-DPC method is presented in [17] to
U*o PI θ
Cotroller Dual- Q1-Q4 improve the performance of DAB, such as accurate reference
...

Phase-Shift
Uo Current Stress
ᵠ Q5-Q8
tracking and fast dynamic response. The control block diagram
Modulation
...

Ui Optimization of the SM-DPC is shown as Fig. 6.

Eq. (9) Io
Sliding Mode Based
θ
Fig. 4. The control diagram of the conventional feedback control. U*o
Direct Power Control Dual- Q1-Q4
...

Eq.(16)-Eq.(18)
Phase-Shift
According to the conventional feedback control principle and Uo ᵠ Modulation Q5-Q8
...

Current Stress
the relationship of the optimal phase shift ratios of DAB as (9). Ui
Optimization Eq. (9)
The optimal phase shift ratios φ and θ can be expressed as Fig. 6. The control diagram of the SM-DPC scheme.
  1  *
 =  kp + s ki  (U o − U o ) According to the control diagram SM-DPC in Fig. 6, the
  
  ( K − 1)(1 − 2 ) (13) phase shift θ between the primary and secondary bridges can be
   
 =  2 expressed as follows:
 4
1 −  −    PSM ( K − 1) K 2 +2 K − 3
 ( K − 1)
  = 0  PSM 
2
  2 ( K + 3) 2K 2 (16)
Where kp and ki are the proportional and integral coefficients  1 1 − PSM K 2 +2 K − 3
 = −  PSM  1
of the PI controller, respectively. The conventional feedback  2 2 ( K 2 − 2 K + 3) 2K 2

control is a PI controller with a direct feedback portion, the Where PSM is the virtual power deduced by the equivalent

© 2022 IEEE. Personal use is permitted, but republication/redistribution requires IEEE [Link] [Link] for more information.
Authorized licensed use limited to: UNIVERSIDAD DE VIGO. Downloaded on April 15,2023 at [Link] UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for publication in IEEE Transactions on Transportation Electrification. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TTE.2022.3229076

control of the sliding mode control (SMC). φ > θ, the analysis of the MPC is the same as the analysis when
  φ < θ, which is no longer described in this paper. Besides, the
= o + 2 (U o* − U o ) + 3  (U o* − U o ) dt (17)
i
PSM
Co 1 1
control block diagram of the MPC is shown in Fig. 7.
The coefficients α1, α2, and α3 can be selected by Ackerman’s Io
Model Predictive
Control θ
formula [41] as (18), and the fBW is the desired bandwidth of the U*o Dual- Q1-Q4

...
Eq.(23)&Eq.(25)
Phase-Shift
closed-loop system. Uo ᵠ Modulation Q5-Q8

...
Current Stress
2  (18)
= 4 f BW , 3 = 4 2 f BW
2 Ui
Optimization Eq. (9)
1 1
Fig. 7. The control diagram of the MPC.
The SM-DPC is obtained by combining sliding mode control
and direct power control, which can improve the dynamic The predictive current control with two current sensors can
response and robustness of DAB converters. However, the also improve the dynamic performance. Under the same control
oscillation of the output voltage and inductor current, and the bandwidth, the output impedance of the predictive current
output current sensor are also needed. control is lower than that of the feedback-only control [42]. But
the predictive current control method may not be feasible for
D. Model Predictive Control
high-frequency applications, since ac sampling and
To further improve the dynamic performance of the DAB, computation are required for each switching cycle, which leads
the model predictive control (MPC) utilizes the predictive to the requirement of a high-speed digital processor. To reduce
model and the past deviations to predict the future behaviors of the sensitivity to circuit parameters and the cost of hardware
the converter. A cost function is adopted to define the deviation design while improving the dynamic performance of DAB
between the predicted control variable and its desired value, converters, a model predictive model control without current
which can be minimized by selecting the appropriate phase- sensor is proposed in this paper.
shift angle of the DC-DC converter, to achieve fast dynamic
response performance. IV. MODEL PREDICTIVE CONTROL WITHOUT CURRENT
Based on (2) and (9), the transmission power is expressed as SENSOR
 NTU iU o  K 2 − 2 K + 3 ( K − 1)    
2

P =   (1 −  ) − (19) A. Analysis and Evolution of the proposed MPC

 2 L  2 8 
  

 NTU iU o  3

4 3 1 2

Based on the analysis of section Ⅲ-C, to improve the
 P =  + −   
2L  ( K − 1) 2 
2

  dynamic performance of DAB in current sensorless condition,
According to (1) and (19), when φ < θ, the dynamic model of the current sensorless-based model predictive power control is
the output voltage can be expressed as proposed in this paper, which is named the proposed MPC. To
dU U NTU  K − 2 K + 3 ( K − 1)  (20)
2 2 overcome the challenges of MPC and predictive current control,
+ = o
 o
 1− − i
( )
dt RCo 2 LCo  2 8 
 such as the high hardware cost, the transmission power is
The differential terms are discretized with forward Euler as chosen as the control object of the DAB converter. Then, the
dU U ( k + 1) − U ( k ) (21) cost function of the proposed MPC consists of the discrete
= o o o

dt T predicted per-unit value of transmission power P*(k+1) and the

Substituting (21) into (20), the output voltage value at the desired transmission power Pset(k).
next sampling interval can be expressed as J ( k + 1) = ( P* ( k + 1) -Pset ( k ) )
2
(26)
TI ( k ) (22)
U ( k + 1) = U ( k ) − + m ( k ) (1 −  ( k ) ) − n
o
o o
Co Where the discrete predicted per-unit value of the
Where Ui (k) is the input voltage at the kth sampling interval, transmission power P*(k+1) is
θ (k) is the phase-shift ratio at the kth sampling interval. Io (k) is 4 ( k + 1) (1 −  ( k + 1) ) − 2 2 ( k + 1)  ( k + 1)   ( k + 1)
P ( k + 1) =  (27)
4 ( k + 1) (1 −  ( k + 1) ) − 2 ( k + 1)  ( k + 1)   ( k + 1)
2
the load current at the kth sampling interval, and Uo (k) and Uo
(k+1) are the output voltage at the kth and (k+1)th sampling And the discrete desired transmission power Pset (k) is
interval, respectively. And m and n are shown in (23). expressed as
m=
( K − 2K + 3) NT U ( k ) ，n = ( K − 1) NT U ( k )
2 2
(23)
i
2 2
i
Pset ( k ) = (kp + i ) (U o* ( k ) − U o ( k ) ) d ( k )
k (28)
4 LCo 16 LCo s
The Uo (k+1) is an expression consisting of Uo (k), θ (k), and In (28), the Δd(k) =U*o(k) /Uo(k) and the Δd(k)[0.5,1.5].
circuit parameters. To minimize the error between the desired The difference between P*(k+1) and Pset(k) is minimum when
and physical output voltage, the cost function is defined as the cost function is equal to the minimum value, which means
J ( k ) = U o ( k + 1) − U o* ( k ) 
2
(24) that the DAB converter achieves accurate power tracking
capability at this time. Thus, the minimum value of the cost
Where U*o (k) is the desired output voltage, when the J (k) is
function J(k+1) in (26) can be solved by (29).
minimized, the optimal phase-shift ratio can be deduced as
J (k + 1) (29)
1 U ( k ) − U ( k ) + TI ( k ) C + n =0
*
1
 = − − * o (25) o o o  (k + 1)
2 4 m
The optimized θ(k+1) can be expressed as
As shown in (25), the optimal phase-shift is related to the
input voltage, circuit parameters, and output current. The MPC
can improve the dynamic performance but it is very sensitive to
circuit parameters and the load current sensor is needed. When

© 2022 IEEE. Personal use is permitted, but republication/redistribution requires IEEE [Link] [Link] for more information.
Authorized licensed use limited to: UNIVERSIDAD DE VIGO. Downloaded on April 15,2023 at [Link] UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for publication in IEEE Transactions on Transportation Electrification. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TTE.2022.3229076

 K 2 − 2K + 1 + 2P ( k ) K 2 +2 K − 3  = a + bPset (35)
 set
 Pset ( k )  1
 4 ( K − 2 K + 3)
2
2K 2 (30) K 2 − 2K + 1 2d
 ( k + 1) =  a= ，b = (36)
 1 K − ( K − 3) − ( 24 − 16 K ) Pset ( k )
2
K 2 +2 K − 3 4 ( K 2 − 2 K + 3) 4 ( K 2 − 2 K + 3)
 + 0  Pset ( k ) 
2 2(3 − 2 K ) 2K 2 Eq. (35) is expanded by Taylor’s formula at steady-state
As shown in (30), the optimal θ(k+1) is related to K, Pset (k), point A, and the high-order term is ignored
and Δd(k). In other words, the optimal θ(k+1) is decided by the b b2 (37)
 = + (P − P ) − (P − P )
2

2 a + bPs ( )
s set s 3 set s
operation condition of the DAB converter. 4 a + bPs
The operating principle of the proposed MPC method is
Where θs and Ps are the values of the steady-state point A.
shown as Fig. 8. Firstly, the voltage sensors collect the input
Then, the disturbance of the phase-shift ratio can be deduced as
and output voltage signal, Ui(k) and Uo(k) respectively, at the  
beginning of the kth sampling period interval, which is used to ˆ = 
b
+
b2  Pˆ (38)

 2 a + bPs 4 ( )
3

calculate K. Then, the discrete desired output power Pset (k) is a + bPs 
 
obtained by multiplying the output of the PI controller by Δd(k), And the disturbance of the transmission power is the output
where the PI controller is used to compensate for the deadtime of PI controller, which is
effect, model error, sampling error, etc. Secondly, the optimal  k 
Pˆ =  kp + i  (U o* − U o ) (39)
outer phase shift θ(k+1) is derived by (30), which is the  s
expression of the proposed MPC method. Besides, the optimal Then, the transfer function Gc (s)of the proposed MPC is
inner phase shift ratio φ(k+1) is obtained from (9), which is the  
 k 
Gc ( s ) =  kp + i   
b b2 (40)
expression of the CSO strategy under DPS modulation. Finally, + 
 s   2 a + bPs 4
( )
3
a + bPs 
the optimal phase shift ratios are converted into switch signals,  
which control the MOSFET devices in the DAB converter to Based on the analysis, the closed-loop feedback control
achieve power transmission and performance improvement. system of the proposed MPC method is shown as Fig. 9. A
is io sampling/conversion delay e-Tds caused by digital control is
Q1 Q4 Q5 Q8
n:
Ui
iL L
1
Co R Uo
considered, and the closed-loop transfer function of the
Ci uab ucd
proposed MPC is obtained as (41).
G ( s ) = Gc ( s ) Gvd ( s ) e −Td s
Q2 Q3 Q6 Q7

Q1-Q4 Signals ... ... Q5-Q8 Signals  b 

  (41)
Dual-Phase-Shift Modulation NTRU i ( K 2 − 2 K + 3) (1 − 2 )  ki   2 a + bP2
*
 e −Td s
=  kp +   + b 
θ(k+1) φ(k+1) 4 L (1 + sRCo )  s 

( ) 
3
 4 a + bP 
*
Current Stress Optimization
Eq. (9)  
Ui(k)
θ(k+1)
Δ d(k)=U*o(k)/Uo(k)
 −y 
the proposed MPC ×
PI
Uo(k)
G (s) =
(
NTRU i 3 ( K + 1)  − 2 ( K − 1) 

)
ki  2 z x − yPs

 −Td s (42)
 kp +   y2 e
4 L ( K − 1)(1 + sRCo )
Eq. (30) × Cotroller
U*o(k)  s +
 3 
Fig. 8. The control diagram of the proposed MPC.  4 z x − yPs
 ( ) 


B. Stability Analysis Similarly, when φ > θ, the closed-loop transfer function of

the proposed MPC is obtained as (42). And, the x, y, z in (42)
To make the proposed MPC method more consummate, the can be expressed as
stability analysis of the proposed method is accomplished.
x = ( K − 3) ，y = 24 − 16 K，z = 6 − 4 K
2
(43)
When φ ≤ θ, a small signal perturbation of the output voltage
and phase-shift ratio shown in (31) is introduced in (20), and Based on the transfer function of the proposed MPC method
the model with a small signal perturbation of the output voltage and the experimental parameters shown in Table Ⅱ, the bode
is expressed as (32). diagram is illustrated in Fig. 10. Besides, kp = 0.04 and ki = 5,
U o = U o + uˆo，  =  + ˆ (31) the gain crossover frequency ωc = 216 rad/s and the PM = 71.6°
and ωc = 14000 rad/s and the PM = 89.7°, respectively. So, the
NTU i  K 2 − 2 K + 3
) − ( K 8− 1)

(
2
d Uo
+
Uo
=   − 
2
 (32)
dt RCo 2 LCo  2  proposed MPC method is stable.

U*o Pset s Uo
Gc e-Td Gvd
According to (32) and ignoring the steady-state and high-
order disturbance term, the small signal model of the output
voltage can be derived as Fig. 9. The block diagram of the proposed MPC.
duˆo uˆ NTU i ( K 2 − 2 K + 3) (1 − 2 ) (33)
+ o = ˆ
dt RCo 4 LCo
The control-output transfer function Gvd (s) can be deduced
with Laplace transform.
NTRU ( K − 2 K + 3) (1 − 2 ) (34)
2

G (s) =
i

4 L (1 + sRCo )
vd

To reflect the influence of output voltage disturbance on the

proposed MPC method. Based on (30), the optimal phase-shift
ratio θ can be expressed as (a) φ ≤ θ (b) φ > θ
Fig. 10. The bode diagram of the proposed MPC under DPS modulation.

© 2022 IEEE. Personal use is permitted, but republication/redistribution requires IEEE [Link] [Link] for more information.
Authorized licensed use limited to: UNIVERSIDAD DE VIGO. Downloaded on April 15,2023 at [Link] UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for publication in IEEE Transactions on Transportation Electrification. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TTE.2022.3229076

C. Dynamic Analysis 11(b), respectively. Finally, the results of the comparison

experiments are analyzed and the conclusion is given.
The traditional method uses the PI controller as the controller. TABLE Ⅱ
The dynamic performance of the DAB converter under PI PARAMETERS OF THE PHYSICAL PROTOTYPE
control is low due to the large inertia part. The proposed MPC-
Parameters Parameters Parameters Parameters
DPS control can predict the output power with respect to the Ui=65/75V L=208/178μH fs=10kHz R=20/30/40/80Ω
variation of the input voltage and desired voltage in the next Uo=60V Co=1120μF Ci=1120μF N : 1=1 : 1
switching cycle according to the sampling information at the TABLE Ⅲ
current moment. According to (26), when the load changes or PARAMETERS OF THE 3-PHASE INVERTER
the input voltage changes, the proposed MPC-DPS control can Parameters Values Parameters Values
get the phase shift ratio that makes the converter respond fastest. Input voltage Vin=60V Load resistance R=10/10/10Ω
Thus, it is clear that the proposed MPC-DPS control is realized q-axis current i*q=5/3A Load inductance L=8/8/8mH
by combing the nonlinear characteristic of DAB converters Switching frequency fs=10kHz IGBTs IHW50N65
itself according to (26)-(30), which is also the reason that the
dynamic performance of the proposed MPC-DPS control is
better than other linear control methods (Eg. PI control). In
addition, when the output voltage is far away from the desired
value, the controller always adjusts the converter with the
greatest effort (φ reaches the boundary) to track the output
voltage with the desired value again.
(a) DAB physical prototype (b) 3-phase inverter prototype.
D. Disturbance Analysis Fig. 11. Photo of the experimental physical prototype.
Model predictive is sensitive to the circuit parameters, such A. Comparison of Steady-State Performance
as switching frequency, inductance and capacitance of the DAB
converter. However, there is a deviation between the theoretical The comparison of steady-state performance is based on the
values of inductance and capacitance. Because of the deviation, PI-SPS, the LCFF-DPS, and the proposed MPC-DPS. When
the model of DAB coverters are inaccurate, which will affect K=65/60, the transformer terminal voltage and inductor current
the performance of traditional MPC methods. The transmission waveforms of the DAB converter under the PI-DPS and the
power is chosen as the control object of the proposed MPC, the proposed MPC-DPS are shown in Fig. 12. The efficiency under
detailed analysis is shown in section Ⅳ-A, and the phase-shift the PI-SPS and the proposed MPC-DPS is 95.19% and 95.32%,
ratios are expressed as (30) and (9). The phase-shift ratios are respectively. And the value of inductor current peak under the
only related to the input voltage and output voltage. To reflect PI-SPS and the proposed MPC-DPS are almost equivalent. It is
the effect of circuit parameters mismatch on the control because the inner phase shift φ(k+1) calculated by the proposed
performance of the proposed MPC method, the experimental MPC-DPS is very close to zero when K is close to 1. In this case,
comparison of the proposed MPC method under different it can be considered that the efficiency and inductor current of
inductance is carried out in section Ⅴ-C. the proposed MPC under DPS is the same as that under SPS.
However, when the input voltage is 80V and the desired
V. EXPERIMENTAL ANALYSIS output voltage is still 60V, K=80/60 is far from 1, the inner
phase shift φ(k) obtained by the proposed MPC-DPS method is
A scale-down physical experimental prototype of the DAB not zero. As shown in Fig. 13, the inductor current peak value
converter with TMS320F28335 controller is shown in Fig. of the DAB converter under the PI-SPS is 7.50A, and the value
11(a). Moreover, the electrical circuit parameters of the under the proposed MPC-DPS is 5.50A. The current stress can
physical prototype are shown in TABLE Ⅱ. The physical be effectively reduced under the proposed MPC-DPS.
prototype is built to test the effectiveness of the proposed MPC. Analogously, the efficiency of the PI-SPS and the proposed
First, the traditional PI control with DPS modulation (the PI- MPC-DPS is 88.2% and 92.6% respectively. In summary, the
DPS), the LCFF with DPS modulation (the LCFF-DPS), the proposed MPC can solve different phase shifts according to the
SM-DPC control with DPS modulation (the SM-DPC), and the different operating points of the DAB converter, and realize the
proposed MPC method with DPS modulation (the proposed application of various phase shift modulation methods.
MPC-DPS) are chosen for experiment comparison. Then, the
comparison experiments of steady-state and dynamic Uab Uab
performance are carried out among the four control methods.
Ucd Ucd
However, the load is not the resistive load in the practical
application, which is much more complicated than a simple
resistance load. A 3-phase inverter is built as the load of the iL iL

DAB converter. And due to the limitation of the laboratory

(a) (b)
equipment, a Y-connection resistance-inductance load is used Fig. 12. Experimental waveforms when K=65/60, (a) the PI-SPS, (b) the
to instead of a real motor as the inverter load. The parameters proposed MPC-DPS.
and photo of the 3-phase inverter are given in Table. Ⅲ and Fig.

© 2022 IEEE. Personal use is permitted, but republication/redistribution requires IEEE [Link] [Link] for more information.
Authorized licensed use limited to: UNIVERSIDAD DE VIGO. Downloaded on April 15,2023 at [Link] UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for publication in IEEE Transactions on Transportation Electrification. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TTE.2022.3229076

DPS are invisible and can be ignored, of which the output

Uab Uab voltage remains almost constant. Thus, the proposed MPC
Ucd Ucd method can greatly improve the dynamic response performance
iL
iL compared with the existing optimization methods.

Ui Ui
6.75V 316ms 188ms
` Uo Uo 5.25V
(a) (b) 282ms 6.00V 158ms 5.00V
Fig. 13. Experimental waveforms when K=80/60, (a) the PI-SPS (b) the Io Io
proposed MPC-DPS. iL iL

Under different operation conditions, the efficiency of the 7.60A 7.55A

DAB converter under the PI-SPS, the LCFF-DPS, and the (a) (b)
proposed MPC-DPS is listed in TABLE Ⅳ. When K=65/60, the
inner phase shift ratio φ(k) is approximately equal to zero, Ui Ui

which means that the modulation scheme of the proposed MPC Uo 4.25V 174ms
Uo

becomes SPS modulation in all load situations. In this case, the 162ms 3.75V

Io Io
effect of CSO is almost negligible. However, when K=75/80, iL iL
the modulation scheme of the proposed MPC is DPS
modulation in R=30 Ω and R=40 Ω situations, the effect of CSO 7.50A 7.45A

is remarkable. When K deviates from unity, the inductor current (c) (d)
Fig. 14. Experimental waveforms when input voltage step change 65/75V, (a)
peak can be greatly reduced due to the proposed MPC method, the PI-DPS, (b) the LCFF-DPS, (c) the SM-DPC, (d) the proposed MPC-
which is consistent with the theoretical analysis. DPS.
TABLE Ⅳ To verify the effectiveness of the proposed method under
EFFICIENCY UNDER DIFFERENT OPERATION CONDITIONS
load step change conditions, the experimental parameters are
Control Method Operation Condition Efficiency
set as Ui = 65V and U*o = 60V. The load resistance steps up
R=20Ω 95.19%
K=65/60 R=30Ω 96.50% from 66% to 100% and then steps back to 66% and the load
The PI-SPS
R=40Ω 96.55% resistance steps up from 50% to 100% and then steps down back
R=30Ω 95.52% to 50%, which are shown in Fig. 15 and Fig. 16, respectively.
K=75/60
R=40Ω 95.14%
K=80/60 R=20Ω 88.20%
Compared with the other three methods, the proposed MPC-
R=20Ω 95.20% DPS can achieve the shortest settling time and lowest voltage
K=65/60 R=30Ω 96.58% overshoot under load step change, the settling time and
R=40Ω 96.58% overshoot are invisible and can be ignored. Besides, compared
The LCFF-DPS
R=30Ω 95.78%
K=75/60
R=40Ω 94.94%
the SM-DPC as shown in Fig. 15(c) and Fig. 16(c), the proposed
K=80/60 R=20Ω 91.34% MPC-DPS can avoid the chattering effect introduced by the
R=20Ω 95.32% SM-DPC. Thus, the proposed MPC method can greatly
K=65/60 R=30Ω 96.51% improve the dynamic response performance compared with the
R=40Ω 96.88%
The proposed MPC-DPS
R=30Ω 95.92% existing optimization methods.
K=75/60
R=40Ω 95.05%
K=80/60 R=20Ω 92.60% Ui Ui
418ms 19.75V 290ms 18.50V
Uo Uo
B. Comparison of Dynamic Performance 15.00V 304ms 14.25V 237ms
Io Io
The comparison experiment of the PI with DPS modulation iL iL

(the PI-DPS), the LCFF-DPS, the SM-DPC, and the proposed

11.35A 11.30A
MPC-DPS is implemented to certificate the effectiveness of the
(a) (b)
proposed method in dynamic performance. The comparison
consists of input voltage step change and load step change
Ui Ui
experiments under the three control methods mentioned before. Uo 168ms 3.50V Uo
As shown in TABLE Ⅱ, the desired value of output voltage is 4.00V 148ms

U*o = 60V, and the input voltage is set to step between 65V and Io Io
iL
75V. The load resistance is 20Ω, 30Ω and 40Ω, and 20Ω is the iL

rated load resistance, which is recorded as 100% load situation. 11.25A 9.35A

The 30Ω, 40Ω, and 80Ω is the 66%, 50%, and 25%load (c) (d)
situation, respectively. Fig. 15. Experimental waveforms when load step change 66%/100%, (a) the
PI-DPS, (b) the LCFF-DPS, (c) the SM-DPC, (d) the proposed MPC-DPS.
Fig. 14 shows the experimental waveforms of the DAB
converter when the input voltage steps up from 65V to 75V and
back to 65V when R = 30Ω. Compared with the other three
methods, the settling time and overshoot of the proposed MPC-

© 2022 IEEE. Personal use is permitted, but republication/redistribution requires IEEE [Link] [Link] for more information.
Authorized licensed use limited to: UNIVERSIDAD DE VIGO. Downloaded on April 15,2023 at [Link] UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for publication in IEEE Transactions on Transportation Electrification. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TTE.2022.3229076

is changed but the theoretical value is unchanged to realize the

Ui Ui
comparison. The results of the comparison show that the
616ms 36.50V Uo 192ms 27.00V
Uo
proposed MPC method is not sensitive to the circuit parameter,
21.25V 424ms 19.20V 294ms

Io Io
and the dynamic performance of the proposed MPC is excellent
iL iL under different inductances.
15.15A 12.95A
Ui Ui
(a) (b) Uo 12ms Uo 14ms

8ms 11ms
Ui Ui Io Io
252ms 10.00V 12ms iL
Uo Uo iL
9.00V 340ms 8ms

Io Io
iL iL (a) (b)
Fig. 18. Experimental waveforms when load step change 50%/100%, (a)
11.40A 9.60A L=208μH, (b) L=178μH.
(c) (d)
Fig. 16. Experimental waveforms when load step change 50%/100%, (a) the D. Comparison Experiments Under Inverter Load
PI-DPS, (b) the LCFF-DPS, (c) the SM-DPC, (d) the proposed MPC-DPS.
To validate the control performance of the proposed MPC
The comparison experiment of the PI-DPS, the LCFF-DPS, method with a complicated load situation, the dynamic response
the SM-DPC, and the proposed MPC-DPS is implemented to experiment was carried out with the 3-phase inverter as the load
certify the effectiveness of the proposed MPC method in light of the DAB converter. Because the 3-phase Y-connection load
load conditions. And the experimental comparison is carried out cannot be simultaneously changed, the dynamic response
and the parameters are set as Ui = 65V and U*o = 60V. The load experiment of the DAB-inverter platform is carried out by
resistance steps between 50% and 25% load, Fig. 17 shows changing the controlled q-axis current i*q of the inverter from
experiment waveforms of the DAB converter. Compared with 5A to 3A. The parameters of the 3-phase inverter are shown in
the other three methods, the settling time and overshoot of the TABLE Ⅲ, and the experimental parameters of the DAB
proposed MPC-DPS are invisible and can be ignored, of which converter are set as Ui = 65V, U*o = 60V, L=208μH. The steady-
the output voltage remains almost constant. Thus, the proposed state waveforms of the phase voltage UA and phase current iA
MPC method can greatly improve the dynamic response in the inverter is shown in Fig. 19. When the controlled q-axis
performance compared with the existing optimization methods. current i*q of the inverter change from 5A to 3A, the peak-peak
And, under the SM-DPC method, there is oscillation of the value of the phase current iA is 6.90A and 4.35A, respectively.
output voltage and inductor current, and the output current And the 3-phase current of the inverter is a symmetrical 3-phase
sensor is also needed. sine wave. The dynamic waveforms of the output voltage Uo of
the DAB converter and the phase current iA of the 3-phase
Ui Ui
inverter under different control methods are given in Fig. 20.
236ms 13.25V Uo 108ms 9.25V
Uo
11.00V 286ms 8.75V 142ms
The settling time and overshoot of the PI-DPS are 150ms and
Io Io 7.75V, the settling time and overshoot of the LCFF-DPS are
iL iL 93ms and 9.00V, the settling time and overshoot of the SM-
5.55A 5.45A DPC are 52ms, 5.50V, and the settling time of the proposed
(a) (b) MPC-DPS are 13ms, and the output voltage of DAB remains
almost constant under the proposed MPC method. The
Ui Ui experiment results have verified that the proposed MPC method
Uo 80ms Uo can maintain excellent dynamic response under complicated
96ms
load situations, such as inverter load. To present the control
Io Io
iL iL
performance of the proposed MPC method under different load
4.50A 4.35A
situations, the settling time and overshoot are summarized in
TABLE Ⅴ and TABLE Ⅵ.
(c) (d)
Fig. 17. Experimental waveforms when load step change 50%/25%, (a) the PI- TABLE Ⅴ
DPS, (b) the LCFF-DPS, (c) the SM-DPC, (d) the proposed MPC-DPS. SETTLING TIME COMPARISON UNDER DIFFERENT OPERATION CONDITIONS

C. Parameter Sensitivity Experiments Operation PI-DPS LCFF-DPS SM-DPC MPC-DPS

Ui = 65/75V, R
282/316ms 158/188ms 162/174ms negligible
The experimental comparison of the proposed MPC is = 30Ω
Ui = 65V,
carried out, and the experimental parameters are set as Ui = 65V, 236/286ms 108/142ms 80/96ms negligible
R = 80/40Ω
U*o = 60V, L=208/178μH, and the load resistance steps up from Ui = 65V,
418/304ms 290/237ms 168/148ms negligible
50% to 100% then steps down back to 50%, as shown in Fig. R = 30/20Ω
18. The model of the proposed MPC is derived based on Ui = 65V,
616/424ms 192/294ms 252/340ms 12/8ms
R = 40/20Ω
L=208μH, regardless of whether the physical inductance is Ui = 65V,
150ms 93ms 52ms 13ms
208μH or 178μH. In other words, the physical inductance value i*q = 5/3A

© 2022 IEEE. Personal use is permitted, but republication/redistribution requires IEEE [Link] [Link] for more information.
Authorized licensed use limited to: UNIVERSIDAD DE VIGO. Downloaded on April 15,2023 at [Link] UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for publication in IEEE Transactions on Transportation Electrification. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TTE.2022.3229076

TABLE Ⅵ performance. When the input voltage step, resistive

OVERSHOOT COMPARISON UNDER DIFFERENT OPERATION CONDITIONS
load step, and 3-phase inverter load step, the output
Operation PI-DPS LCFF-DPS SM-DPC MPC-DPS voltage almost remains constant, without overshoot
Ui = 65/75V, voltage and fast transient response.
6.75/6.00V 5.25/5.00V 4.25/3.75V Zero
R = 30Ω
Ui = 65V,
(2) Compared with the SM-DPC, the dynamic performance
11.00/13.25V 8.75/9.25V Zero Zero of the DAB converter is improved, while the hardware
R = 80/40Ω
Ui = 65V, cost is decreased by reducing the current sensor. The
15.00/19.75V 14.25/18.50V 4.00/3.50V Zero
R = 30/20Ω
Ui = 65V,
proposed MPC is easy to implement, and can improve
21.25/36.50V 19.20/27.00V 9.00/10.00V Zero the dynamic response without current sensor. In
R = 40/20Ω
Ui = 65V,
7.75V 9.00V 5.50V Zero addition, it can avoid chattering and degradation of
i*q = 5/3A tracking performance problems brought by SMC.
(3) The stability analysis for load disturbances and
robustness for parameter sensitivity is demonstrated.
The proposed MPC method has the advantage of
UA UA
excellent dynamic behavior and parameters insensitive.
6.90A 4.35A
The results of parameter sensitivity experiments show
iA iA that the dynamic performance of the proposed MPC is
excellent under different inductances.
(a) (b)
Fig. 19. Experimental waveforms of the phase voltage UA and phase current iA,
(4) The proposed MPC method has strong flexibility and
* *
(a) i q = 5A, (b) i q = 3A. adaptability in practical engineering applications. It can
achieve excellent dynamic performance and operate
Uo
without current sensor, which can decrease the
Uo 7.75V 9.00V
150ms 93ms hardware cost.
iA iA

REFERENCES
[1] R. W. A. A. De Doncker, D. M. Divan and M. H. Kheraluwala, "A three-
phase soft-switched high-power-density DC/DC converter for high power
(a) (b) applications," IEEE Trans. Ind. Appl., vol. 27, no. 1, pp. 63-73, Jan.-Feb.
1991.
[2] N. Hou and Y. W. Li, "Overview and comparison of modulation and
Uo 5.50V Uo 1.75V
control strategies for non-resonant single-phase dual-active-bridge DC-
52ms 13ms
DC converter, " IEEE Trans. Power Electron., vol. 35, no. 3, pp. 3148–
iA iA
3172, Mar. 2020.
[3] B. Zhao, Q. Song, and W. Liu, "Overview of dual-active-bridge isolated
bidirectional DC-DC converter for high-frequency-link power-conversion
system, " IEEE Trans. Power Electron., vol. 29, no. 8, pp. 4091-4106. Nov.
2014.
(c) (d) [4] M. A. H. Rafi and J. Bauman, "Optimal control of semi-dual active bridge
Fig. 20. Experimental waveforms when q-aix current change 5/3A, (a) the PI- DC/DC converter with wide voltage gain in a fast-charging station with
DPS control, (b) the LCFF-DPS control, (c) the SM-DPC control (d) the battery energy storage," IEEE Trans. Transp. Electrification, vol. 8, no. 3,
proposed MPC-DPS control. pp. 3164-3176, Sep. 2022.
[5] N. Hou and Y. Li, "Communication-free power management strategy for
VI. CONCLUSION the multiple DAB-based energy storage system in islanded DC microgrid,"
IEEE Trans. Power Electron., vol. 36, no. 4, pp. 4828-4838, Apr. 2021.
In this paper, a model predictive control without current [6] X. She, X. Yu, F. Wang and A. Q. Huang, "Design and demonstration of
a 3.6-kV–120-V/10-kVA solid-state transformer for smart grid
sensor under DPS is proposed to improve the dynamic application," IEEE Trans. Power Electron., vol. 29, no. 8, pp. 3982-3996,
performance and reduce hardware cost of DAB converters. The Aug. 2014.
predictive model of transmission power under DPS modulation [7] D. Das, N. Weise, K. Basu, R. Baranwal and N. Mohan, "A bidirectional
has been developed, and the optimal phase shift has been soft-switched DAB-based single-stage three-phase AC–DC converter for
V2G application," IEEE Trans. Transp. Electrification, vol. 5, no. 1, pp.
obtained from the CSO and cost function minimization. Besides 186-199, Mar. 2019.
the stability and parameter sensitivity are analyzed. Firstly, the [8] N. H. Baars, J. Everts, H. Huisman, J. L. Duarte and E. A. Lomonova, "A
proposed MPC method combined with CSO can ensure high 80-kW isolated DC–DC converter for railway applications," IEEE Trans.
Power Electron., vol. 30, no. 12, pp. 6639-6647, Dec. 2015.
efficiency of the DAB converter. Then, the dynamic [9] Y. Park, S. Chakraborty and A. Khaligh, "DAB converter for EV onboard
performance of the DAB converter has been significantly chargers using bare-die SIC MOSFETs and leakage-integrated planar
improved without current sensor, which can decrease the transformer," IEEE Trans. Transp. Electrification, vol. 8, no. 1, pp. 209-
224, Mar. 2022
hardware cost. The comparison experiments of the PI-DPS, the [10] S. Luo, F. Wu and G. Wang, "Effect of dead band and transient actions
LCFF-DPS, the SM-DPC, and the proposed MPC-DPS have on CTPS modulation for DAB DC–DC converter and solutions," IEEE
been carried out under resistive load and 3-phase inverter load Trans. Transp. Electrification, vol. 7, no. 3, pp. 949-957, Sep. 2021.
[11] K. Takagi and H. Fujita, "Dynamic control and performance of a dual
condition. The conclusions are summarized as follows. active-bridge DC–DC converter, " IEEE Trans. Power Electron., vol. 33,
(1) Compared with the PI-DPS and LCFF-DPS, the no. 9, pp. 7858–7866, Sep. 2018.
proposed MPC scheme can achieves excellent dynamic

© 2022 IEEE. Personal use is permitted, but republication/redistribution requires IEEE [Link] [Link] for more information.
Authorized licensed use limited to: UNIVERSIDAD DE VIGO. Downloaded on April 15,2023 at [Link] UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for publication in IEEE Transactions on Transportation Electrification. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TTE.2022.3229076

[12] M. Farhadi and O. A. Mohammed, "Performance enhancement of actively [33] O. Yade, J.-Y. Gauthier, X. Lin-Shi, M. Gendrin, and A. Zaoui,
controlled hybrid DC microgrid incorporating pulsed load," IEEE Trans. “Modulation strategy for a dual active bridge converter using model
Ind. Appl., vol. 51, no. 5, pp. 3570- 3578, Sept.-Oct. 2015. predictive control,” Proc. IEEE Int. Symp. Predictive Control Elect.
[13] M. Leng, G. Zhou, S. Zhou, K. Zhang, and S. Xu, “Stability analysis for Drives Power Electron., pp. 15–20, 2015.
peak current-mode controlled buck LED driver based on discrete-time [34] F. M. Oettmeier, J. Neely, S. Pekarek, R. DeCarlo, and K. Uthaichana,
modeling,” IEEE J. Emerg. Sel. Topics Power Electron., vol. 6, no. 3, pp. “MPC of switching in a boost converter using a hybrid state model with a
1567–1580, Sep. 2018. sliding mode observer,” IEEE Trans. Ind. Electron., vol. 56, no. 9, pp.
[14] N. Hou, W. Song, and M. Wu, “Minimum-current-stress scheme of dual 3453–3466, Sep. 2009.
active bridge DC–DC converter with unified phase-shift control,” IEEE [35] L. Chen, S. Shao, Q. Xiao, L. Tarisciotti, P. W. Wheeler and T. Dragičević,
Trans. Power Electron., vol. 31, no. 12, pp. 8552–8561, Dec. 2016. "Model predictive control for dual-active-bridge converters supplying
[15] D. Segaran, B. P. McGrath, and D. G. Holmes, “Adaptive dynamic control pulsed power loads in naval DC micro-grids," IEEE Trans. Power
of a bi-directional DC-DC converter,” Proc. IEEE Energy Convers. Congr. Electron., vol. 35, no. 2, pp. 1957- 1966, Feb. 2020.
Expo., Atlanta, GA, USA, pp. 1442–1449, Sep. 2010. [36] S. Dutta, S. Hazra, and S. Bhattacharya, “A digital predictive current-
[16] P. Xia, H. Shi, H. Wen, Q. Bu, Y. Hu and Y. Yang, "Robust LMI-LQR mode controller for a single-phase high-frequency transformer-isolated
control for dual-active-bridge DC–DC converters with high parameter dual active bridge DC-to-DC converter,” IEEE Trans. Ind. Electron., vol.
uncertainties," IEEE Trans. Transp. Electrification, vol. 6, no. 1, pp. 131- 63, no. 9, pp. 5943–5952, Sep. 2016.
145, Mar. 2020. [37] F. An, W. Song, K. Yang, N. Hou, and J. Ma, "Improved dynamic
[17] K. Li, Y. Yang, S. -C. Tan and R. S. -Y. Hui, "Sliding-mode-based direct performance of dual active bridge DC–DC converters using MPC
power control of dual-active-bridge DC-DC converters," 2019 IEEE scheme," IET Power Electron., vol. 11, no. 11, pp. 1756–1765, Aug. 2018.
Applied Power Electronics Conference and Exposition (APEC), Anaheim, [38] D. Zhao, K. Shen, L. Chen, Z. Wang, W. Liu, T. Yang, and P. Wheeler,
CA, USA, pp. 188-192, Mar. 2019. "Improved active damping stabilization of DAB converter interfaced
[18] F. An, W. Song, B. Yu and K. Yang, "Model predictive control with aircraft DC microgrids using neural network-based model predictive
power self-balancing of the output parallel DAB DC–DC converters in control," IEEE Trans. Transp. Electrification, vol. 8, no. 2, pp. 1541-1552,
power electronic traction transformer," IEEE J. Emerg. Sel. Topics Power Jun. 2022.
Electron., vol. 6, no. 4, pp. 1806-1818, Dec. 2018. [39] J. Saeed and A. Hasan, “Unit prediction horizon binary search-based
[19] G. G. Oggier, M. Ordonez, J. M. Galvez, and F. Luchino, “Fast transient model predictive control of full-bridge DC–DC converter,” IEEE Trans.
boundary control and steady-state operation of the dual active bridge Control Syst. Technol., vol. 26, no. 2, pp. 463-474, Mar. 2018.
converter using the natural switching surface,” IEEE Trans. Power [40] B. Zhao, Q. Song, and W. Liu. "Power characterization of isolated
Electron., vol. 29, no. 2, pp. 946–957, Feb. 2014. bidirectional dual-active-bridge DC-DC converter with dual phase shift
[20] Z. Shan, J. Jatskevich, H. H. -C. Iu and T. Fernando, "simplified load- control, " IEEE Trans. Power Electron., vol. 27, no. 9, pp. 4172-4176, Feb.
feedforward control design for dual-active-bridge converters with current- 2012.
mode modulation," IEEE J. Emerg. Sel. Topics Power Electron., vol. 6, [41] L. Martmez-Salamero, J. Calvente, R. Giral, A. Poveda, and E. Fossas,
no. 4, pp. 2073-2085, Dec. 2018. “Analysis of a bidirectional coupled-inductor Cuk converter operating in
[21] R. Redl and N. O. Sokal, “Near-optimum dynamic regulation of dc–dc sliding mode,” IEEE Trans. Circuits Syst. I Fundam. Theory Appl., vol. 45,
converters using feed-forward of output current and input voltage with no. 4, pp. 355–363, Apr. 1998.
current-mode control,” IEEE Trans. Power Electron., vol. PE-1, no. 3, pp. [42] S. Shao, L. Chen, Z. Shan, F. Gao, H. Chen, D. Sha, and T. Dragiˇcevi´c,
181–192, Jul. 1986. "Modeling and advanced control of dual-active-bridge DC–DC converters:
[22] S. Lin, X. Li, C. Sun, and Y. Tang, “Fast transient control for power a review," IEEE Trans. Power Electron., vol. 37, no. 2, pp. 1524-1547,
adjustment in a dual-active-bridge converter,” Electron. Lett., vol. 53, pp. Feb. 2022.
1130–1132, Aug. 2017.
[23] W. Song, N. Hou, and M. Wu, "Virtual direct power control scheme of
Yaru Deng (Student Member, IEEE) was
dual active bridge DC-DC converters for fast dynamic response, " IEEE
Trans. Power Electron., vol. 33, no. 2, pp. 1750- 1759, Feb. 2018. born in Yuexi, Sichuan, China, in 1996. She
[24] N. Hou, L. Ding, P. Gunawardena, Y. Zhang and Y. W. Li, "a received the B.S. degree in electrical
comprehensive comparison of two fast-dynamic control structures for the engineering from Southwest Jiaotong
DAB DC–DC converter," IEEE Trans. Power Electron., vol. 37, no. 6, pp.
University, Chengdu, in 2019, where she is
6488-6500, Jun. 2022.
[25] Y. Wu, M. H. Mahmud, Y. Zhao and H. A. Mantooth, "Uncertainty and currently pursuing the PhD. degree in
disturbance estimator-based robust tracking control for dual-active-bridge electrical engineering. Her current research
converters," IEEE Trans. Transp. Electrification, vol. 6, no. 4, pp. 1791- interests include optimization control methods
1800, Dec. 2020.
of power electronic transformers and dc distribution systems.
[26] M. Ali, M. Yaqoob, L. Cao and K. H. Loo, "Disturbance-observer- based
DC-bus voltage control for ripple mitigation and improved dynamic
response in two-stage single-phase inverter system," IEEE Trans. Ind. Wensheng Song (Senior Member, IEEE)
Electron., vol. 66, no. 9, pp. 6836- 6845, Sep. 2019. received the B.S. degree in electronic and
[27] T. Siew-Chong, Y.M. Lai, C.K. Tse, "A unified approach to the design of information engineering and the Ph.D. degree
PWM-based sliding-mode voltage controllers for basic DC–DC converters
in electrical engineering from Southwest
in continuous conduction mode," IEEE Trans Circuits Syst I Regul. Pap,
vol. 53, no. 8, pp. 1816–1827, Aug. 2006. Jiaotong University, Chengdu, China, in 2006
[28] J. Ma, W. Song, X. Wang, F. Blaabjerg and X. Feng, "Low-complexity and 2011, respectively. From September
model predictive control of single-phase three-level rectifiers with 2009 to September 2010, he was a Visiting
unbalanced load," IEEE Trans. Power Electron., vol. 33, no. 10, pp. 8936-
Scholar with the Department of Electrical
8947, Oct. 2018.
[29] B. Yu, W. Song, J. Li, B. Li and M. S. R. Saeed, "Improved finite control Engineering and Computer Science, University of California at
set model predictive current control for five-phase VSIs," IEEE Trans. Irvine, Irvine, CA, USA. From July 2015 to December 2015, he
Power Electron., vol. 36, no. 6, pp. 7038-7048, Jun. 2021. was a Visiting Scholar with the University of Alberta,
[30] T. Dragicevic, “Model-predictive control of power converters for robust
Edmonton, AB, Canada. He is currently a Full Professor with
and fast operation of AC microgrids,” IEEE Trans. Power Electron., vol.
33, no. 7, pp. 6304–6317, Jul. 2018. the School of Electrical Engineering, Southwest Jiaotong
[31] L. Tarisciotti, A. Formentini, A. Gaeta, M. Degano, P. Zanchetta, R. University. His current research interests include power
Rabbeni, and M. Pucci, “Model-predictive control for shunt active filters electronics, motor drives, railway traction drive systems, and
with fixed switching frequency,” IEEE Trans. Ind. Appl., vol. 53, no. 1,
multilevel converters.
pp. 296–304, Jan. 2017.
[32] J. Holtz, "Predictive finite-state control—when to use and when not,"
IEEE Trans. Power Electron., vol. 37, no. 4, pp. 4225-4232, Apr. 2022.

© 2022 IEEE. Personal use is permitted, but republication/redistribution requires IEEE [Link] [Link] for more information.
Authorized licensed use limited to: UNIVERSIDAD DE VIGO. Downloaded on April 15,2023 at [Link] UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for publication in IEEE Transactions on Transportation Electrification. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TTE.2022.3229076

Shuai Yin was born in Changsha, Hunan,

China, in 1998. He received the B.S. degree in
electrical engineering from Changsha
University of Science & Technology, in 2020.
He is currently pursuing the M.S. degree in
electrical engineering from Southwest
Jiaotong University, Chengdu. His current
research interests include optimization control methods of
power electronic transformers and dc distribution systems.

Ming Zhong was born in Chengdu, Sichuan,

China, in 1996. He received the B.S. and M.S.
degrees in electrical engineering from
Southwest Jiaotong University, Chengdu, in
2018 and 2021. His current research interests
include optimization control methods of power
electronic transformers and dc distribution
systems.

Kexin Yang received the M.S. degree in

electrical engineering in 2019 from Southwest
Jiaotong University, Chengdu, China, where
he is currently working toward the Ph.D.
degree in electrical engineering. His current
research interests include power electronics
and the reliability of power device.

Xiaoyun Feng received the B.S., M.S., and

Ph.D. degrees in electrical engineering from
Southwest Jiaotong University, Chengdu,
China, in 1983, 1988, and 2001, respectively.
From 1998 to 1999, she was a Visiting Scholar
in University of Tokyo, Tokyo, Japan. Since
1983, she has been with the School of
Electrical Engineering, Southwest Jiaotong
University, where she is currently a Full Professor. Her major
research interests include railway optimizing operation, real-
time control of train operation, energy-efficient driving,
electrical traction converters, and motor drive systems.

© 2022 IEEE. Personal use is permitted, but republication/redistribution requires IEEE [Link] [Link] for more information.
Authorized licensed use limited to: UNIVERSIDAD DE VIGO. Downloaded on April 15,2023 at [Link] UTC from IEEE Xplore. Restrictions apply.