Modeling and Control of Dual Active Bridge

Converter with Two Control Loops and Output Filter

Thanh Lich Nguyen and Gerd Griepentrog Van Trang Phung

Institute for Power Electronics and Control of Drives Department of Electrical Engineering and Computer Sciences
Technical University of Darmstadt University of Siegen,
Darmstadt, Germany Siegen, Germany

Abstract—This paper deals with modeling and control strategy to ensure a flexible exchange of power inside a nanogrid and
for a dual active bridge (DAB) converter. The generalized average among different nanogrids.
modeling method is used to obtain the state-space representation
of the DAB, in which the DC and the fundamental harmonic in the
Fourier series expansion of state variables are considered.
Transfer functions from control-to-output are determined, which
will be used to define two controllers for the output current and
the output voltage in a cascaded control structure. Simulations and
experiments will be used to validate the operation of the proposed
method.

Keywords—dual active bridge converter, average model, phase

shift modulation, control.

I. INTRODUCTION
Microgrids are becoming a potential solution for combining
distributed generation units such as photovoltaics panels, wind
turbines, and energy storage systems [1][2][3]. The concept of
nanogrids was introduced in [4] to indicate a simple and small
scale microgrid. Nanogrids are often used for single nodes, e.g. (a)
houses and small buildings. Most nanogrids are connected to
the utility grid. However in island mode, nanogrids that are not
connected to the utility grid may suffer from an unbalance of
the generated electrical energy and the load. One of the
solutions to this problem as depicted in Fig. 1 is connecting
several nanogrids together to enable them to share the electrical
energy [4].
In the schematic shown in Fig. 1(a), the interface of
(b)
interconnection depicted in the red box is required to have a
bidirectional power transfer capability and galvanic isolation.
Fig. 1. The structure of a DC microgrid; (a) The modular structure of a DC
The dual active bridge (DAB) topology is often selected for the
nanogrid; (b) microgrid made of interconnected DC nanogrids [4]
interface converter as it fulfills the aforementioned
requirements. It can be seen in Fig. 1(a) that converters In terms of modeling, much effort has focused on the DAB
corresponding to different sources inside a nanogrid and the converter. Full-order modeling and dynamic analysis of a
interface converter share a common DC bus. To manage phase-shift DAB converter in both the continuous-time domain
electrical energy between the sources, the DC/DC converters and the discrete-time domain are presented in [5] and [6],
are regarded as current sources where output currents fed to the respectively. Both approaches consider the leakage inductance
common DC bus are under control. By regulating the direction current of the high-frequency (HF) transformer as a state
of the output currents with respect to the DC bus, each converter variable. The two methods were proven successful in modeling
can work as either a generator or a load resulting in a flexible the DAB and designing a single closed loop control structure
energy exchange between the sources. This paper focusses on for the output voltage. Compared to reduced-order models
the modeling and control of the interface converter where a introduced in [7] the full-order models ensure higher accuracy
DAB is adapted. Both the output current and output voltage of since more variables are considered.
the DAB are controlled in a cascaded two-control-loop structure

l-))) 
 In addition to the single voltage control loop, a cascaded III. GENERALIZED MODEL OF DAB CONVERTER
control structure has been developed for the DAB converter, in
which an inner current control loop is inserted within an outer A. Fourier interpolation
voltage cascade [8]. Generally, the current flowing in the As the generalized average modeling method is based upon
windings of the HF transformer is to be controlled by the current the Fourier interpolation of time-dependent variables, for
control loop. An alternative selection for the current control convenience, some important properties of the Fourier
loop associates with the output current which is fed to the output interpolation will be presented hereafter.
voltage as it was proposed and presented in [9]. This approach The Fourier series expansion of a periodic signal ( ) in one
is based on modeling of the DAB in the discrete-time domain
fundamental period − ≤ ≤ is given by [5][10]:
and was developed with the intention to implement it on a
digital signal processing unit (DSP).
Based on a generalized average method mentioned in [5], ( )= 〈 〉 ( )∙e (2)
this paper presents a full-order continuous-time state-space
model of a DAB with a modification in the topology of the DAB
where = 2π = 2π/ ; 〈 〉 is the coefficient of the
such that a LC filter is added before the output DC link for
harmonic
meeting the electromagnetic interference (EMI) requirement.
Additionally, a cascaded control structure is developed for the
output current and the output voltage rather than just a single 1
voltage control loop. 〈 〉 ( )= ( ) ∙e (3)

II. THE CONVENTIONAL DAB CONVERTER

From (3) coefficients 〈 〉 and 〈 〉 satisfy
The circuit diagram of a conventional DAB converter is
shown in Fig. 2. It consists of two H-bridges, two DC capacitors ℜ{〈 〉 } = ℜ{〈 〉 }; ℑ{〈 〉 } = −ℑ{〈 〉 } (4)
and a HF transformer. The HF transformer provides galvanic
isolation between two voltage levels. Each H-bridge comprises where the nomenclatures "ℜ" and "ℑ" denote the real and the
four switching devices along with anti-parallel freewheeling imaginary parts of the complex coefficients, respectively.
diodes.
The derivative of the coefficient for variable ( ) is [11]
d d
〈 〉 ( )=〈 〉 ( )−j 〈 〉 ( ) (5)
d d
The proof of (5) can be obtained directly by using the
Fourier representation (2). The coefficient of the product of
two variables and is
Fig. 2. Bidirectional dual active bridge converter

There are three commonly used methods to control the 〈 〉 = 〈 〉 〈 〉 (6)

DAB converter: 1) the control of phase shift between the two
bridges; 2) the duty ratios of switching devices and 3) the According to (4) the coefficients corresponding to = 1
switching frequency [5][9]. The method based on phase-shift and = −1 are complex conjugates, and if only harmonics up
control is adapted in this work because of its simplicity and its to the 1st order are taken into account, the product of zeroth
effectiveness. In the phase-shift modulation technique, the two coefficient terms can be approximated by
H-bridges are driven with a constant duty cycle of 50%. The
power flow is controlled by varying the phase shift between two 〈 〉 =〈 〉 〈 〉 +
square wave voltages generated by the two H-bridges. The flow (7)
direction of power is determined by the sign of the phase shift. 2(ℜ{〈 〉 }ℜ{〈 〉 } + ℑ{〈 〉 }ℑ{〈 〉 })
The transferred power is
and the 1st coefficient terms can be approximated by

= (1 − ) (1) ℜ{〈 〉 } = 〈 〉 ℜ{〈 〉 } + ℜ{〈 〉 }〈 〉 (8)

2
where = / is the turn ratio from the primary side to the ℑ{〈 〉 } = 〈 〉 ℑ{〈 〉 } + ℑ{〈 〉 }〈 〉 (9)
secondary side of the HF transformer; and are the input
and output voltages, respectively; = /π is the phase shift B. Gereralized model of DAB
ratio with being the phase shift; is the switching frequency; Fig. 3 demonstrates a simplified DAB referred to the
= / + is the leakage inductance referred to the secondary side, where = / is the input voltage referred
output side where and are the leakage inductances in to the secondary side, = is the output voltage. The two
the input and output sides, respectively. bridges produce square wave voltages and at the terminals
of the HF transformer. The amplitudes of and are equal to


the DC voltage and , respectively. and are the
equivalent inductance and resistance of the HF transformer; with ( ) = 1, ≤ < + (11)
2 2 2
is the load resistance. Differing from the conventional DAB
−1, otherwise
depicted in Fig. 2, the DAB shown in Fig. 3 has a LC output
filter characterized by an inductor , a DC resistance of the To model the DAB converter, the output voltage , the
inductor, and a capacitor for the purposes of meeting the EMI transformer current , the filter voltage and the filter current
requirement. The calculation of the parameters of the LC filter are considered as state variables. The mathematical
is presented in [12]. description of the DAB can be obtained by following the
Kirchoffs’s law

d ( ) ( ) ( )
= − (12)
d

d ( ) ( ) ( ) ( ) ( )
=− ( )+ − (13)
d
Fig. 3. Simplified DAB converter referred to the secondary side
d ( ) ( ) ( ) ( )
= − (14)
According to the phase shift modulation technique, the d
switching mode ( ) of the left H-bridge, ( ) of the right H-
bridge and the transformer current in one period are depicted in d ( ) ( ) ( )
Fig. 4 [5]. The waveforms are derived under the assumption that =− − ( )+ (15)
d
the transformers magnetizing current is ignored, the switching
devices are ideal and have no voltage drop. As the output voltage , the filter voltage and the output
current are DC in the steady state, only DC components of the
three state variables are considered. On the contrary, the
transformer current is a pure alternating current (AC). For
simplicity, only the 1st harmonic of is taken into account. By
using the determination of the DC and the 1st harmonic given in
(5) and (7)-(9), the state-space differential equations of the
DAB are
d〈 〉 1 1
= 〈 〉 − 〈 〉 (16)
d
d(ℜ{〈 〉 })
=− ℜ{〈 〉 } + ℑ{〈 〉 }
d
1
+ (〈 〉 ℜ{〈 〉 } + ℜ{〈 〉 }〈 〉 )
Fig. 4. Waveforms during one period

The voltage at the transformer input side can only have 1

− (〈 〉 ℜ{〈 〉 } + ℜ{〈 〉 }〈 〉 ) (17)
two states: 1) + when switches and are ON; and 2)
− when switches and are ON [5]. Therefore,
d(ℑ{〈 〉 })
( )= ( ) ( ) =− ℑ{〈 〉 } − ℜ{〈 〉 }
d
1
1, 0≤ < (10) + (〈 〉 ℑ{〈 〉 } + ℑ{〈 〉 }〈 〉 )
with ( )= 2
−1, ≤ < 1
2 − (〈 〉 ℑ{〈 〉 } + ℑ{〈 〉 }〈 〉 ) (18)
where = 1/ , ( ) is the input voltage with taking into
account the perturbation, 〈 〉 = . d〈 〉 1 2
= 〈 〉 〈 〉 + ℜ{〈 〉 }ℜ{〈 〉 }
Similarly, the terminal voltage at the secondary side is d
( )= ( ) ( ) 2 1
+ ℑ{〈 〉 }ℑ{〈 〉 } − 〈 〉 (19)


 d〈 〉 1 1 ≈ sin( π) ∙ 〈 〉 + sin( π) ∙ ∆〈 〉
=− 〈 〉 − 〈 〉 + 〈 〉 (20) (30)
d +cos( π) ∙ ∆ π ∙ 〈 〉
The DC and the 1st harmonic of ( ) and ( ) can be Denoting
calculated theoretically according to (3) and Fig. 4.
= [∆〈 〉 , ∆(ℜ{〈 〉 }), ∆(ℑ{〈 〉 }), ∆〈 〉 , ∆〈 〉 ]
2 as the state variables vector and = ∙ as the output
〈 〉 = 0; ℜ{〈 〉 } = 0; ℑ{〈 〉 } = − (21)
π variables where is the 5 × 5 identity matrix, the small-signal
state-space representation of the DAB converter is
2sin( π)
〈 〉 = 0; ℜ{〈 〉 } = − ; d
π = ∙ + ∙∆ (31)
(22) d
2cos( π) = ∙
ℑ{〈 〉 } = −
π
where
Substituting (21) and (22) into (16)-(20), the state-space
model of the DAB converter in a matrix form is given in (23). 1 1
⎡− 0 0 0 ⎤
IV. SMALL SIGNAL MODEL OF THE DAB CONVERTER ⎢ ⎥
⎢ 2sin( π) ⎥
The state-space representation of the DAB given in (23) ⎢ 0 −
π
0 ⎥
allows the determination of the small-signal control-to-output ⎢ ⎥
transfer functions, which demonstrates the dynamic responses 2cos( π)
=⎢ 0 − − 0 ⎥
of the converter from a small deviation of the system in the ⎢ π ⎥
control signal. Since the phase-shift modulation technique is ⎢ 4sin( π) 4cos( π) 1⎥
adapted, the control signal is the phase-shift ratio . The ⎢ 0 − − 0 − ⎥
⎢ π π ⎥
deviation of the control signal and the state variables are defined
by ⎢ −1 0 0
1
− ⎥
⎣ ⎦
= +∆ (24) 0
⎡ 2cos( π)〈 〉 ⎤
〈 〉 = 〈 〉 + ∆〈 〉 (25) ⎢ ⎥
⎢ ⎥
⎢ 2sin( π)〈 〉 ⎥
ℜ{〈 〉 } = ℜ{〈 〉 } + ∆(ℜ{〈 〉 }) (26) =⎢ − ⎥
⎢ ⎥
ℑ{〈 〉 } = ℑ{〈 〉 } + ∆(ℑ{〈 〉 }) (27) ⎢ 4 [ℑ{〈 〉 }sin( π) − ℜ{〈 〉 }cos( π)]⎥
⎢ ⎥
〈 〉 =〈 〉 + ∆〈 〉 (28) ⎣ 0 ⎦

〈 〉 = 〈 〉 + ∆〈 〉 (29)
The calculation of ℜ{〈 〉 } and ℑ{〈 〉 } complements the
where steady states are represented by capital variables, large- state-space equation (31). First, the DAB converter is run at a
signal states by lower case variables and small signal states steady state with a predefined value of = . The leakage
by ∆. It is realized that (23) includes multiplications of control inductance current is obtained and the Fast-Fourier
input and state variables. One example of such multiplications Transform (FFT) is applied to determine the real part ℜ{〈 〉 }
is derived in the following as an example [5]. and the imaginary part ℑ{〈 〉 } of the first harmonic.
sin( π) 〈 〉 = sin[( + ∆ )π](〈 〉 + ∆〈 〉 )

1 1
⎡− 0 0 0 ⎤
⎢ ⎥
⎢ 2sin( π) ⎥ 0
〈 〉 0 − 0 ⎥ 〈 〉
⎡ ⎤ ⎢ π ⎡ ⎤ ⎡ 0 ⎤
ℜ{〈 〉 } ⎢ ⎥ ⎢ℜ{〈 〉 }
d⎢ ⎥ 2cos( π) ⎥ ⎢ 2 ⎥
ℑ{〈 〉 }⎥ = ⎢ 0 − − 0 ⎥ × ⎢ ℑ{〈 〉 } ⎥ + ⎢− ⎥ [〈 〉 ] (23)
d ⎢ ⎢ π ⎥ ⎢ π ⎥
⎢ 〈 〉 ⎥ ⎢ ⎢ 〈 〉 ⎥ ⎢ 0 ⎥
4sin( π) 4cos( π) 1⎥
⎣ 〈 〉 ⎦ ⎢ 0 − − 0 − ⎥ ⎣ 〈 〉 ⎦ ⎣ ⎦
π π 0
⎢ ⎥
⎢ −1 0 0
1
− ⎥
⎣ ⎦


 V. CLOSED-LOOP CONTROL OF DAB CONVERTER proportional and integral gains as listed in TABLE I are used in
According to the state-space representation (31), the both simulation and experiment.
control-to- output transfer functions are defined by (32). TABLE I. PARAMETERS OF CONTROLLERS

= , , , _ , =( ∙ − ) (32) Voltage controller Current controller

where , , , _ , and are the transfer functions = 1.2 (A/V) = 0.001 (1/A)
from the small signal ∆〈 〉 , ∆(ℜ{〈 〉 }), ∆(ℑ{〈 〉 }), ∆〈 〉 , = 140 (A/Vs) = 100 (1/As)
and ∆〈 〉 to the control small signal ∆ , respectively. From
those functions, represents the relationship between a
change of the phase shift ratio ∆ and the output voltage and
that between ∆ and the current flowing into the filter.

Fig. 5. Control scheme for DAB

The cascaded control structure with two control loops for

the DAB converter is depicted in Fig. 5. The inner cascade
allows the control of the output current and is superimposed
by the outer cascade aiming to control the output voltage . In
Fig. 6. Bode plot of the open-loop transfer function of the current loop
this control structure, the delay term due to the control delay is
modelled by a first-order lag element ( ) with the time
constant = 2. , where = 20 μs is the switching period
corresponding to the switching frequency = 50 kHz.
1
( )= (33)
1+ .
The transfer functions of the proportional-integral (PI)
controllers for the current and the voltage loops are given by
(34) and (35), respectively.

( )= + (34)

( )= + (35)

where , are the gains of the proportional and integral parts

of the current controller while , correspond to the
voltage controller.
The open-loop transfer functions ( ) from to , and
Fig. 7. Bode plot of the open-loop transfer function of the voltage loop
( ) from to , are defined by:
The Bode diagrams of the open-loop transfer functions of
( )= ( ) ( ) ( ) (36) the current and the voltage control loops are shown in Fig. 6 and
Fig. 7, respectively. It is observed that the gain crossover
( ) ( ) ( ) ( ) frequencies of the current loop is 100 Hz and that of the voltage
( )= (37)
1+ ( ) ( ) ( ) loop is 70 Hz. The bandwidth of the two control loops are quite
low due to the resonant point at the cut-off frequency of the out
The parameters of the PI controllers in the inner and the filter. In order to increase the bandwidth, one possible solution
outer loops are tuned by using the SISO tool in Matlab along is to increase the proportional gain of the PI controller.
with the procedure presented in [13]. The obtained values of the However, the increase in the gain decreases both the gain


margin and phase margin that might cause destabilization of the
current control loop. Another solution is to eliminate the TABLE II. PARAMETERS OF DAB CONVERTERS
resonant frequency by providing a damping network at the
output side of the converter such as that as introduced in [12]. Parameters Symbol Value Unit

VI. SIMULATION AND EXPERIMENTAL RESULTS Input voltage 100 V

Simulation and experimental models were built to validate Output voltage 48 V

the operation of the proposed method. The parameters of the
DAB converter are given in TABLE II. Fig. 8 depicts the Power 400 W
simulation results corresponding to step changes of resistive
loads. It is observed in Fig. 8 that the output voltage is kept Transformer ratio : 16:8
stable with only a small overshoot and undershoot with respect
to the changes in the load. The load is increased at = 0.07 s Total inductance @50kHz ′
t
60 μH
leading to a decrease in the output voltage and an increase in
the output current . The undershoot of the voltage is 1 V or 2% Total resistance @50kHz ′ 110 mΩ
t
of the reference value while the settling time is around 20 ms.
The settling time depends on the sizes of the load changes and Power semi. devices − IRFP260MPBF MOSFET
the output capacitor. A similar phenomenon can be seen at =
0.11 s when the load is suddenly decreased. Switching frequency 50 kHz

Output capacitor 2200 μF

Filter capacitor 40 μF

Filter inductor 10 μH

Filter resistor 16 mΩ

Fig. 8. Simulation results of DAB converter

Fig. 9 shows the schematic of the DAB converter built in

the laboratory. A TMS320F28335 DSP is used to execute the Fig. 9. System diagram
control algorithm and generate gate signals at a switching
frequency of 50 kHz. As the leakage inductance of the HF Experimental results are presented in Fig. 10 - Fig. 12. Fig.
transformer referred to the primary side is 5 μH, an additional 10 shows the responses of the output voltage , the output
inductor is used to make sure that the total inductance referred current and the load current corresponding to a step
to the primary side is 60 μH. Litz wire AWG is used for both change of the resistive load. The undershoot of the voltage is
the HF transformer and the inductor. around 1 V or 2% of the reference value and the settling time
The measurement circuit is equipped with two voltage is approximately 20 ms. It is observed that the experimental
transducers LEM LV-25P to measure the input and output DC results coincide with those obtained in the simulation model.
voltages. Two current transducers LEM LA-25-NP are used to The results corresponding to step down of the load are shown
measure the input and output currents as they are shown in Fig. in Fig. 11. Once again, a similarity between the experiments
9. In order to suppress conducted EMI, the filter inductor with and the simulations can be realized while comparing Fig. 8 and
the value of =10 μH is used. The DC resistance of the filter Fig. 11. Fig. 12 shows the terminal voltages of the HF
inductor is =16 mΩ. transformer and the current flowing in the primary winding.


 variable are taken into account. From the state-space
representation, control-to-output transfer functions were
Vo (20V/div) derived to allow the design of a cascaded control structure
(2V/div) dedicated to the control of the output current and the output
voltage. Verification of the developed model was examined by
if (1A/div) both simulation and experiment. The obtained results show
good transient responses of the output current as well as the
output voltage. The ongoing work considers the application of
iR (1A/div) the developed control strategy to manage electrical power
exchange among different nanogrids working in the island
mode.
(10ms/div)

REFERENCES
[1] J. M. Guerrero, M. Chandorkar, T. L. Lee, and P. C. Loh, “Advanced
Fig. 10. Experimental result for load current step up from 0.96 A to 2.88 A Control Architectures for Intelligent Microgrids #x2014;Part I:
Decentralized and Hierarchical Control,” IEEE Trans. Ind. Electron.,
vol. 60, no. 4, pp. 1254–1262, Apr. 2013.
[2] J. J. Justo, F. Mwasilu, J. Lee, and J.-W. Jung, “AC-microgrids versus
DC-microgrids with distributed energy resources: A review,” Renew.
Vo (20V/div) Sustain. Energy Rev., vol. 24, pp. 387–405, Aug. 2013.
[3] T. Dragičević, X. Lu, J. C. Vasquez, and J. M. Guerrero, “DC Microgrids
(2V/div)
#x2014;Part II: A Review of Power Architectures, Applications, and
Standardization Issues,” IEEE Trans. Power Electron., vol. 31, no. 5,
pp. 3528–3549, May 2016.
if (1A/div) [4] A. Werth, N. Kitamura, and K. Tanaka, “Conceptual Study for Open
Energy Systems: Distributed Energy Network Using Interconnected
DC Nanogrids,” IEEE Trans. Smart Grid, vol. 6, no. 4, pp. 1621–1630,
Jul. 2015.
(10ms/div)
iR (1A/div) [5] H. Qin and J. W. Kimball, “Generalized Average Modeling of Dual
Active Bridge DC #x2013;DC Converter,” IEEE Trans. Power
Electron., vol. 27, no. 4, pp. 2078–2084, Apr. 2012.
[6] C. Zhao, S. D. Round, and J. W. Kolar, “Full-order averaging modelling
of zero-voltage-switching phase-shift bidirectional DC-DC
converters,” IET Power Electron., vol. 3, no. 3, pp. 400–410, May 2010.
[7] H. Bai, C. Mi, C. Wang, and S. Gargies, “The dynamic model and
Fig. 11. Experimental result for load current step down from 2.88 A to 0.96 A
hybrid phase-shift control of a dual-active-bridge converter,” in 2008
34th Annual Conference of IEEE Industrial Electronics, 2008, pp.
2840–2845.
[8] G. D. Demetriades and H. P. Nee, “Dynamic modeling of the Dual-
Active Bridge topology for high-power applications,” in 2008 IEEE
Power Electronics Specialists Conference, 2008, pp. 457–464.
vac1 (100V/div) [9] F. Krismer and J. W. Kolar, “Accurate Small-Signal Model for the
Digital Control of an Automotive Bidirectional Dual Active Bridge,”
IEEE Trans. Power Electron., vol. 24, no. 12, pp. 2756–2768, Dec.
2009.
[10] V. A. Caliskan, O. C. Verghese, and A. M. Stankovic, “Multifrequency
averaging of DC/DC converters,” IEEE Trans. Power Electron., vol.
14, no. 1, pp. 124–133, Jan. 1999.
vac2 (100V/div) [11] S. R. Sanders, J. M. Noworolski, X. Z. Liu, and G. C. Verghese,
“Generalized averaging method for power conversion circuits,” IEEE
(5μs/div) Trans. Power Electron., vol. 6, no. 2, pp. 251–259, Apr. 1991.
[12] F. Krismer, “Modeling and Optimization of Bidirectional Dual Active
it (1A/div) Bridge DC–DC Converter Topologies,” Ph.D. dissertation, Power
Electronic Systems Laboratory (ETH Zürich), 2010.
[13] G. D. Demetriades, “On small-signal analysis and control of the single
and the Dual Active Bridge topologies,” Ph.D. dissertation, Royal
Fig. 12. Switching wave form for load current of 0.96 A
Institute of Technology, Department of Electrical Engineering,
Electrical Machines and Power Electronics, KTH, Stockholm, 2005.

VII. CONCLUSION
In this paper, the generalized average method was used to
model a DAB converter where the DC component of the DC
state variables and the fundamental harmonic of the AC state