Comparison of Simulation of Dual-Active Bridge in

Different Simulators and Using Different Simulation

Methods
Shiyuan Yin Suman Debnath Qianxue Xia Shilpa Marti
School of Electrical and Power Electronics System School of Electrical and Rivian
Computer Eng. Integration Group Computer Eng. Irvine, USA
Georgia Institute of Technology Oak Ridge National Laboratory Georgia Institute of Technology [email protected]
Atlanta, USA Oak Ridge, USA Atlanta, USA
[email protected] [email protected] [email protected]

Phani Marthi Maryam Saeedifard

Power Electronics System School of Electrical and
Integration Group Computer Eng.
Oak Ridge National Laboratory Georgia Institute of Technology
Oak Ridge, USA Atlanta, USA
[email protected] [email protected]

Abstract— The accuracy of states measured from simulation of DC conversion applications [2-4] with bidirectional power flow
a dual-active bridge (DAB) converter with an open-loop controller and zero-voltage-switching capability.
has been used for comparing simulation software and algorithms.
The accuracy of the measured states is sensitive to simulation The design and control of the DAB converter has been
settings and parameters chosen. In this paper, the simulation of widely discussed in the literature. As one of the classical control
the DAB converter with open-loop control is run in the methods, the open-loop control is often applied in some cases
MATLAB/Simscape Electrical, PLECS, and PSCAD software due to simplicity, and in the preliminary stage of hardware test
environments, with various simulation settings. The simulation of DAB converters. In [5-6], the voltage gain of the DAB
results are also compared with a new simulation algorithm converter at steady-state has been derived, which helps calculate
proposed for DAB simulations. Different simulation settings are the output voltage with open-loop control.
evaluated, including different solver types, operating conditions,
and the on-state resistance values of the MOSFET switches. The Simulation software is an important tool to help predict or
simulation results are compared with the theoretical calculations. validate the performance of the system. However, in the
It is shown that unexpected divergence and DC offset may occur technical literature, it has never been reported that the DAB
in some cases. This paper provides guidelines to simulate the DAB open-loop simulation is very sensitive to simulation settings.
converter with open-loop control. Hence, the selection of simulation software and settings are very
important for getting accurate results.
Keywords—DAB, dual-active bridge converter, simulator,
MATLAB, PLECS, PSCAD In this paper, the operation of the DAB converter with open-
loop control is analyzed. Then, the simulation studies are
I. INTRODUCTION conducted in the MATLAB/Simscape Electrical, PLECS, and
PSCAD software environments with different simulation
The dual-active bridge (DAB) converter topology was first
settings and system parameters. The simulation results are
proposed in 1991 [1], and since then has become popular in DC-
analyzed utilizing the metrics defined in [7]. Finally, the
guidelines to simulate the DAB converter with open-loop
Research sponsored by Solar Energy Technologies Office of U.S. control are provided. Moreover, the performance of the fast
Department of Energy. This material is based upon work supported by the U.S. electromagnetic transient (EMT) simulation algorithms
Department of Energy's Office of Energy Efficiency and Renewable Energy proposed for DAB simulations [8] is also evaluated in this paper.
(EERE) under the Solar Energy Technologies Office Award Number 34019.
The views expressed herein do not necessarily represent the views of the U.S. II. OPERATION PRINCIPLE OF DAB WITH OPEN-LOOP CONTROL
Department of Energy or the United States Government.
This manuscript has been authored by UT-Battelle, LLC under Contract The circuit schematic of a single-phase DAB converter is
No.DE-EE0002064 with the U.S. Department of Energy. The United States shown in Fig. 1. The DAB converter has two H-bridges
Government retains and the publisher, by accepting the article for publication, interfaced through a high-frequency transformer in which the
acknowledges that the United States Government retains a non-exclusive, paid- leakage inductance acts as the main energy transfer medium. V1
up, irrevocable, world-wide license to publish or reproduce the published form
of this manuscript, or allow others to do so, for United States Government
and V2 are the DC input and output voltages of the DAB
purposes. The Department of Energy will provide public access to these results converter, n is the turns ratio of the transformer, Llk represents
of federally sponsored research in accordance with the DOE Public Access
Plan(http://energy.gov/downloads/doe-public-access-plan).

XXX-X-XXXX-XXXX-X/XX/$XX.00 ©20XX IEEE

the leakage inductance of the transformer referred to high- A. MATLAB/Simscape Electrical
voltage side, vp and vs are the primary and secondary voltages of In MATLAB/Simscape Electrical, a powergui block with
the transformer, and iLlk is the leakage inductor current. the discrete simulation option is chosen for the power electronics
Assuming an ideal converter without any power loss, the blocks' simulation as the continuous option will bring an initial
output power is expressed by: state error. The sample time is set to 1e-7 s, which is one tenth
of the level of the time delay Td. The simulation results obtained
𝑛𝑉 𝑉2
𝑃1 = 2𝑓 𝐿1 2 𝜑(𝜋 − 𝜑), (1) from MATLAB/Simscape Electrical are summarized in Table
𝑠 𝑙𝑘 π
III.
where φ is the phase shift angle between the two H-bridges and
fs is the switching frequency. TABLE I. SPECIFICATIONS OF THE SIMULATED DAB CONVERTER

When the high-voltage side is connected to a resistive load Parameter Value

RL, the output power is: Rated power 150 kW
𝑉2 Transformer turns ratio 1:n 1:2
𝑃𝑜 = 𝑅2 . (2)
𝐿 Switching frequency fs 10 kHz

Let P1=Po, therefore, by combining (1) and (2), the output Leakage inductor Llk 100 μH
voltage V2 can be solved as: Output Capacitor Co 7.7 mF
𝑛𝑉 𝑅𝐿
𝑉2 = 2𝑓 𝐿1 2 𝜑(𝜋 − 𝜑). (3) TABLE II. TEST CASES
𝑠 𝑙𝑘 𝜋
Calculated
Input Resistive Time
No. output voltage
voltage V1 load RL delay Td
V2
1 800 V 17.07 Ω 7 μs 1644 V
2 50 V 100 Ω 7 μs 602 V

Fig. 1. The circuit schematic of the single-phase DAB converter.

III. SIMULATION OF THE DAB CONVERTER UNDER OPEN-LOOP

CONTROL

A DAB converter simulation model (as shown in Fig. 1) is

built in the MATLAB/Simscape Electrical, PLECS, and
PSCAD software environments. The simulation results are
compared with the corresponding theoretical values of the
output voltage from (3).

When the single-phase shift control is applied, the PWM

signals for the eight switches are shown in Fig. 2. The same
PWM signals with 50% duty cycle are sent to switches S1 and S4
in the primary H-bridge. Switches S2 and S3 are sharing the same
gate signals, which are complementary to the signals sent to Fig. 2. PWM signals for the DAB converter with the single-phase shift
switches S1 and S4. The switching signals to S5 and S8 in the control.
secondary H-bridge are delayed by Td with respect to the
corresponding signals to S1 and S4. The switching signals to In Table III, Ron is the on-state resistance of the MOSFET.
switches S6 and S7 are complementary to signals sent to switches When Ron is set at 1e-9 Ω, which symbols a loss-free device,
S5 and S8. The relationship between Td and φ is: simulation cannot reach a stable state in both cases. That means
that the inductor current iLlk and output voltage V2 are diverging,
𝜑 = 2𝜋𝑇𝑑 𝑓𝑠 . (4) as shown in Fig. 3. When Ron is set to 8 mΩ, the simulation can
settle down, as seen in Fig. 4. However, it makes the output
The parameters for the simulated DAB converter are listed voltage V2 lower than its calculated value from (3) as losses are
in Table I. Two different operating conditions, as listed in Table not accounted in the theoretical calculations. The low-power test
II, are simulated. The first test case is the designed nominal (presented in Fig. 5) has a higher steady-state error with respect
operation of the DAB converter, while the second one is a low- to the calculated value from (3) than the nominal operation. The
power operating condition of the DAB converter. The output simulation always takes a very long time, i.e., above 9 hours.
voltages are calculated according to (3) for both cases, as listed
in Table II.
 Considering the significant simulation time cost,
MATLAB/Simscape Electrical is not recommended to run the
simulation of the DAB converter with the open-loop control.

TABLE III. SIMULATION RESULTS IN MATLAB/SIMSCAPE ELECTRICAL

Steady-
Duration Time Fully
Solver Test state
of Ron taken to stabl
type case error of
simulation simulate e?
V2
1e-9 33874.73
No \
Ω s
1
34448.72
8 mΩ Yes -0.37%
s
5s Discrete
1e-9 32607.45
No \
Ω s
2
33689.95
8 mΩ Yes -11.31%
s
(a)

(b)
(a)
Fig. 4. Simulated waveforms under test case 1 and Ron = 8 mΩ in
MATLAB/Simscape Electrical (stable) (a) 5 s time span, and (b) zoomed in
waveforms.

(b)
Fig. 3. Simulated waveforms under test case 1 and Ron = 1e-9 Ω in Fig. 5. Simulated waveforms under test case 2 and Ron = 8 mΩ in
MATLAB/Simscape Electrical (unstable) (a) 5 s time span, and (b) zoomed in MATLAB/Simscape Electrical with 5 s time span (stable).
waveforms.
B. PLECS
In PLECS, there are two available solver types. When the
simulation runs with a fixed-step and discrete solver selection,
the fixed-step size is set to 1e-7 s. When the simulation runs with
a variable-step and auto solver selection, the max step size is set
as 1e-7 s with the relative tolerance of 1e-5 s. The simulation
results obtained from PLECS are summarized in Table IV.
Like MATLAB/Simscape Electrical, when Ron is set at 1e-9
Ω, both iLlk and V2 are diverging from their nominal operating
points in Fig. 6. When the simulation runs with the variable-step
solver, the simulation cannot stay fully stable even when Ron is
set to 8 mΩ. In this case, as shown in Fig. 7, the DC offset in iLlk
changes over time. An unstable leakage current, even with the
higher Ron, is a possibility.
Hence, if PLECS is being used, only the fixed-step solver
with Ron in milli-ohms is recommended to run the simulation of
the DAB converter with the open-loop control, as the case
shown in Fig. 8. The simulation in PLECS takes less time to
simulate when compared to MATLAB/Simscape Electrical.
From Tables III and IV, the steady-state errors are very similar
between the stable simulations in MATLAB/Simscape Fig. 8. Simulated waveforms under test case 1 and Ron = 8 mΩ using fixed-
Electrical and PLECS. step solver in PLECS with 5 s time span (stable).

TABLE IV. SIMULATION RESULTS IN PLECS

Steady-
Duration Time Fully
Solver Test state
of Ron taken to stable
type case error of
simulation simulate ?
V2
1e-9 Ω 159.28 s No \
1
Fixed- 8 mΩ 156.96 s Yes -0.30%
step 1e-9 Ω 205.75 s No \
2
8 mΩ 162.43 s Yes -12.31%
5s
1e-9 Ω 345.21 s No \
1
Variabl 8 mΩ 346.67 s No \
e-step 1e-9 Ω 345.21 s No \
2
8 mΩ 395.25 s No \

C. PSCAD
In the PSCAD, the solution time step is set to 1e-7 s. The
Fig. 6. Simulated waveforms under test case 1 and Ron = 1e-9 Ω using fixed-
step solver in PLECS with 5 s time span (unstable). simulation results obtained from the PSCAD are summarized in
Table V.
Contrasting the results generated from simulations in
MATLAB/Simscape Electrical and PLECS, even when Ron is
set being 1e-9 Ω (ideal component), the simulations can reach
the steady-state. In the first test case, the simulated output
voltage V2 is the same as its calculated value, as shown in Table
V and Fig. 9. In the second test case, the simulated output
voltage V2 is very close to the calculated value, but there is a DC
offset (around 100 A) in iLlk. The DC offset is unexpected and is
shown in Fig. 10. When Ron is 8 mΩ, both simulated test cases
can stay stable, and no DC offset in the leakage inductor current.
The steady-state error of the first test case is the same as that in
the MATLAB/Simscape Electrical, while the steady-state error
of the second test case (as shown in Fig. 11) is the smaller than
that in the MATLAB/Simscape Electrical and PLECS.
In PSCAD, the time taken to simulate is within the similar
range as PLECS. Thus, the PSCAD is recommended for the
Fig. 7. Simulated waveforms under test case 1 and Ron = 8 mΩ using DAB converter open-loop control simulation for both ideal loss-
variable-step solver in PLECS with 5 s time span (unstable). free devices and practical lossy devices, considering the stable
performance and the low time cost.
 TABLE V. SIMULATION RESULTS IN PSCAD
Steady-
Duration Time
Solver Test Fully state
of Ron taken to
type case stable? error of
simulation simulate
V2
1e-9
397.33 s Yes 0
Ω
1
8
259.03 s Yes -0.37%
mΩ
Yes, but
5s \
1e-9 inaccuracies
429.20 s -0.99%
Ω observed in
2
iLlk
8
410.06 s Yes -9.97%
mΩ

D. Fast Electromagnetic Transient (EMT) Simulation

Fig. 9. Simulated waveforms under test case 1 and Ron = 1e-9 Ω in PSCAD Algorithms
with 5 s time span (stable). Simulation algorithms have been proposed in [8] to speed up
the simulation of the DAB converter. By utilizing the hybrid
discretization method, relaxation techniques, and the event-
driven interpolating method, the simulation based on the
proposed algorithms can maintain accuracy with an extended
simulation time step.
The simulation algorithm in [8] is upgraded to a multi-order
hybrid discretization method, with the leakage current in
transformer being discretized using a backward Euler method
with second-order approximation of the external inputs to the
dynamics. In the modified algorithm, the time-step is set to 1e-
6 s, while the simulation interpolation is set to 1e-7 s. The
simulation results are summarized in Table VI.
It is found that the simulations are stable under all test
conditions, and Ron does not have any impact. For test case 1, the
simulated V2 is the same as the calculated value, regardless of
the value of Ron, as shown in Fig. 12. For test case 2, the
Fig. 10. Simulated waveforms under test case 2 and Ron = 1e-9 Ω in PSCAD
simulated V2 is 1.73% lower than the calculated value, as seen
with 5 s time span (stable, but iLlk contains DC offset). in Fig. 13. Thus, the simulation algorithms good stability and
accuracy.

Fig. 11. Simulated waveforms under test case 1 and Ron = 8 mΩ in PSCAD
with 5 s time span (stable).
Fig. 12. Simulated waveforms under test case 1 and Ron = 1e-9 Ω based on the
modified simulation algorithms in [8] with 5 s time span (stable).
 simulation settings and operating conditions. It is shown that the
simulation results are impacted by the selection of software,
solver type, and on-state resistance Ron of the power MOSFET.
Unexpected divergence and DC offset may occur in some cases.
This paper provides guidelines to simulate the DAB
converter with open-loop control. MATLAB/Simscape
Electrical is not recommended as the time taken to simulate is
too long, and Ron cannot be too small to avoid unstable
simulation. PLECS can run the simulation very fast, but Ron
cannot be too small either. The fixed-step solver has a better
performance than the variable-step solver in PLECS. PSCAD is
good to simulate the ideal device and match the output voltage
calculation even though unexpected DC component may appear
in the inductor current under low power and low current
operations. The time taken to simulate in PSCAD is in the
similar range as PLECS. The performance of the EMT
simulation algorithms proposed in [8] is also evaluated in this
Fig. 13. Simulated waveforms under test case 2 and Ron = 1e-9 Ω based on the
modified simulation algorithms in [8] with 5 s time span (stable). paper. The algorithms show significant speed-up effect, and
have good accuracy for simulating with ideal components. It is
TABLE VI. SIMULATION RESULTS BASED ON THE EMT SIMULATION a good choice for simulating the system which includes a large
ALGORITHMS number of DAB converters.
Steady-
Duration
Solver Test
Time
Fully state
ACKNOWLEDGMENT
of Ron taken to
simulation
type case
simulate
stable? error Authors would also like to thank Dr. John Seuss from SETO
of V2 for overseeing the project developments and providing
1e-9 Ω 30.38 s Yes 0 guidance.
1
8 mΩ 32.27 s Yes 0
5s \
1e-9 Ω 33.84 s Yes -1.73% REFERENCES
2
8 mΩ 35.36 s Yes -1.73%
[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-
The time taken to simulate the algorithms is much shorter power applications,” IEEE Trans. on Industry Applications, vol. 27, no.
than other previous discussed simulators, as shown in Fig. 14. 1, pp. 63-73, Jan.-Feb. 1991.
The speed-up function of the simulation algorithms in [8] is [2] B. Zhao, Q. Song, W. Liu and Y. Sun, “Overview of Dual-Active-Bridge
significant. Isolated Bidirectional DC–DC Converter for High-Frequency-Link
Power-Conversion System,” IEEE Trans. on Power Electronics, vol. 29,
no. 8, pp. 4091-4106, Aug. 2014.
[3] F. Krismer and J. W. Kolar, “Efficiency-Optimized High-Current Dual
Active Bridge Converter for Automotive Applications,” IEEE Trans. on
Industrial Electronics, vol. 59, no. 7, pp. 2745-2760, July 2012.
[4] S. P. Engel, M. Stieneker, N. Soltau, S. Rabiee, H. Stagge and R. W. De
Doncker, “Comparison of the Modular Multilevel DC Converter and the
Dual-Active Bridge Converter for Power Conversion in HVDC and
MVDC Grids,” IEEE Trans. on Power Electronics, vol. 30, no. 1, pp.
124-137, Jan. 2015.
[5] A. R. Rodríguez Alonso, J. Sebastian, D. G. Lamar, M. M. Hernando and
A. Vazquez, “An overall study of a Dual Active Bridge for bidirectional
DC/DC conversion,” IEEE Energy Conversion Congress and Exposition
(ECCE), pp. 1129-1135, Sept. 2010.
[6] C. Mi, H. Bai, C. Wang, and S. Gargies, “Operation, design and control
of dual H-bridge-based isolated bidirectional DC-DC converter,” IET
Power Electron., vol. 1, no. 4, pp. 507–5017, 2008.
[7] S. Debnath, et al, “High Penetration Power Electronics Grid: Modeling
Fig. 14. The comparison of the time taken to simulate (in decimal logarithmic and Simulation Gap Analysis”, ORNL/TM-2020/1580, 2020.
scale) by using different simulators or the simulation algorithmes. https://www.osti.gov/biblio/1649067.
[8] Q. Xia, S. Debnath, P. R. V. Marthi, S. Marti and M. Saeedifard, “High-
IV. CONCLUSION Fidelity Models and Fast EMT Simulation Algorithms for Isolated Multi-
port Autonomous Reconfigurable Solar power plant (MARS),” IEEE 12th
The simulation of the DAB converter with the open-loop International Symposium on Power Electronics for Distributed
control has been run in the MATLAB/Simscape Electrical, Generation Systems (PEDG), pp. 1-7, 2021.
PLECS and PSCAD software environments, with different