Equivalent Circuit Model of Dual Active Bridge Converter
Equivalent Circuit Model of Dual Active Bridge Converter
Equivalent Circuit Model of Dual Active Bridge Converter
Converter
Anping Tong1,2, Lijun Hang3, Guojie Li1,2, Jingzhou Xu4
1. Key Laboratory of Control of Power Transmission and Conversion (Shanghai Jiao Tong University), Ministry of Education
2. Dept. of Electrical and Electronic Engineering, Shanghai Jiao Tong University, Shanghai 200240, China
3. Hangzhou Dianzi University, Hangzhou, China
4. Nanjing Power Supply Company, Nanjing, China
[email protected], [email protected], [email protected], [email protected]
Abstract—This paper presents a full-order equivalent circuit sampled-data modeling method. Since the assumption of
model of Dual Active Bridge (DAB) converter. The proposed “small” perturbation at the equilibrium point is not used in the
equivalent circuit facilitates the in-depth knowledge on the derivation of this model, it can retain the global characteristic
physical insight of DAB converter and can precisely retain both of the original system at both open-loop operation and closed-
the steady state and dynamic behavior of the original system. loop operation. Finally, the simulations and experiments are
This equivalent circuit can be divided into two parts: ac branch used to verify the effectiveness of the proposed equivalent
and dc branch. Both ac branch and dc branch are analyzed in circuit model.
detail to derive a more accurate static equation of DAB converter
with considering the parasitic resistance in conducting path.
Finally, the theoretical results are verified by the simulations and II. DERIVATION OF THE EQUIVALENT CIRCUIT MODEL OF DAB
experiments.
CONVERTER
Keywords—DAB converter; equivalent circuit; discrete time
modeling method; single phase shift modulation scheme
Fig. 1 depicts the topology of DAB converter that consists
of two single phase full bridges and a magnetic tank. The
magnetic tank usually includes an inductor and a high
I. INTRODUCTION frequency transformer. The vp(t) and vs(t) denotes the ac
The Dual Active Bridge Isolated Bidirectional dc/dc voltage of primary and secondary full bridges (i.e., H1 and H2),
(DAB)[1] converter has been widely used in various respectively. M denotes the voltage transfer ratio, where M =
applications[2], [3], such as battery management systems[4], (n×vo/vi).
H1 H2
electric vehicles[5], solid state transformer[6], and etc.. Magnetic tank
S1 S3 Q1 Q3
Proper analytical models of power converter are essential +
iL Ls
+
vs
for controller design and component selection. An analytical, vi
vp
- - Co vo RL
n:1
and circuit-oriented model that can capture the global S2 S4 Q2 Q4
characteristic of DAB converter helps to reveal its physical
essence and facilitates to design high performance controller. Fig. 1. Basic topology of DAB converter.
The main difficulty in modeling DAB converter is that the
current of the inductor is purely ac, thus, the input-output The modulation schemes that are commonly employed to
transfer function obtained by State-space Averaging (SSA) regulate the output voltage and transferred power of DAB
method[7], [8] cannot predict the dynamics of DAB converter include: single phase shift (SPS) modulation,
converter[5], [9]. The sampled-data modeling method[10]–[12] extended phase shift (EPS) modulation, Dual phase shift (DPS)
and the generalized average modeling method[9], [13] are modulation, and triple phase shift (TPS) modulation. This
commonly used to analyze the converter’s dynamic behavior. paper focuses on the DAB converter under SPS control, which
However, both these two modeling methods are too is most widely used. The SPS modulation scheme has only one
complicated for practical use. And the resulting models provide control variable—the phase shift ratio between the H1 and H2.
little physical insight of DAB converter. Fig. 2 illustrates the corresponding gate driving signal of all
It is well known that the equivalent circuit model is a switching devices, and the waveform of vp(t), vs(t) and iL(t).
powerful tool for the analysis of power converters and can be According to Fig. 2, DT is the time to turn on of Q1 and Q4,
used to examine all the transfer functions. A full-order where D[א0,1]. Four stages are generated according to the
equivalent circuit model, which can precisely retain both the ON/OFF state of switching devices within one switching
steady state and dynamic characteristics of DAB converter, is period (i.e., 0~2T)
developed in this paper. In contrast with the conventional Stage 1 [0~DT]: S1, S4, Q2, Q3 ON; S2, S3, Q1, Q4 OFF;
method that builds equivalent circuit model by averaging the
switching devices, the proposed model is derived based on the Stage 2 [DT~T]: S1, S4, Q1, Q4 ON; S2, S3, Q2, Q3 OFF;
This work was supported by the National Key Research and Development Stage 3 [T~(1+D)T]: S2, S3, Q1, Q4 ON; S1, S4, Q2, Q3 OFF;
Program of China (2016YFB0900201), by the Science Development Program
of state grid Jiangsu electric power company (J2017075); and by the National Stage 4 [T +DT ~2T]: S2, S3, Q2, Q3 ON; S1, S4, Q1, Q4 OFF;
Natural Science Foundation of China (NSFC) (51777049).
l-)))
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on September 20,2024 at 06:27:28 UTC from IEEE Xplore. Restrictions apply.
Stage 1 Stage 2 Stage 3 Stage 4 Stage 1
xˆ s
S 1S 2 WD s >0 1@ sI ASSA
1
BSSA 0 (4)
S 4S 3 vˆi s dˆ 0
V1 Note that the transfer function (4) indicate that the
vp disturbance of input voltage does not have any influence on the
output voltage. This conclusion is not in accordance with
Q 1Q 2 experiments and simulations. Therefore, the SSA model of
Q 4Q 3 DAB converter cannot reproduce the dynamic characteristic of
V2 actual circuit.
vs
t
(n-1)th x[n-1] nth Switching Period x[n] (n+1)th
iL 0 DT T T+DT 2T
Stage 4 Stage 1 Stage 2 Stage 3 Stage 4 Stage 1
Conventional Proposed
x[n]=F(x[n-1],D,T,u)
Sampling Points Sampling Points
Fig. 2. Typical Operating waveforms of SPS modulation scheme. Fig. 3. State variables iteration relationship for conventional discrete
time model
For each stage, the system state equations is given by
To capture the dynamic characteristic of DAB converter,
dx discrete time modeling method can be used. Fig. 3 illustrates
Ai x Bi vi i ^1, 2,3, 4` (1) the iteration functions of state variables for traditional discrete
dt
T
time modeling method. Generally, the iterative function of one
where, x ª¬iL t , vo t º¼ , and stage is written as
Ti D
³ e AiW dW Bi vi
Ai Ti D
§ Res ·
1 § Res 1 · xn ,i fi xn,i 1 , D e (5)
0
¨ L ¸
Ls ¨ L Ls ¸¸
¨ s ¸ A A ¨ s By setting the moments for turning on the S1 and S4 as the
A1 A4
¨ 1 1 ¸
2 3
¨ 1 1 ¸ sampling point marked by red arrow in Fig. 2, the iterative
¨ ¸ ¨ ¸ mapping between two sampling point is given by[10], [12]
© Co Co RL ¹ © Co Co RL ¹
§ 1 · § 1 · x > n@ Fc x > n 1@ , D, T , vi Gc D x > n 1@ H c D vi (6)
B1 B2 ¨ Ls ¸ B3 B4 ¨¨ Ls ¸¸
¨ ¸ where,
¨ 0 ¸ ¨ 0 ¸
© ¹ © ¹ Gc ( D) e A4 (1 D )T e A3 DT e A2 (1 D )T e A1DT
Res is the equivalent series resistance in the conducting path,
DT
it consists of the ON resistors of switching devices and the H c ( D) e A4 (1 D )T e A3 DT e A2 (1 D )T ³ e A1t dtB1
ESR of the inductor and high frequency transformer. In this 0
(1 D )T DT
case, the turn ratio of high frequency transformer n is assumed e A4 (1 D )T e A3 DT ³ e A2 t dtB2 e A4 (1 D )T ³ e A3t dtB3
to be 1 in this paper. On the assumption of lossless converter 0 0
(1 D )T
(i.e., Res=0), the transferred power, PT, is given by [5] ³ e A4 t dtB4
0
vi vo D 1 D This model is of full-order. Numeric computation, which is
PT , D > 0, 0.5@ (2) too complicated for practical use, is indispensable for
2 f s Ls
calculating (6) due to the complicated matrix function.
The conventional State-space Averaging (SSA) method [8], Moreover, it cannot be used to derive an accurate equivalent
[14] is commonly used to derivate the small-signal transfer circuit of DAB converter.
function of power converter and to build corresponding circuit
oriented model. However, for DAB converter, the following To overcome the drawback of extent models, the time point
input matrix and state matrix are derived for turning on the Q1 and Q4 can be chosen as the sampling
point to build the proposed model. Therefore, iterative
§ Res · relationship within one sampling period is depicted as in Fig. 4.
¨ L 0 ¸
§0·
A SSA ¨ s ¸ B
¨ ¸ (3)
¨ 1 ¸
SSA
© 0¹
¨ 0 ¸
© Co RL ¹
Since all elements of B SSA is zero at all operating point, the
disturbance transfer matrix is
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on September 20,2024 at 06:27:28 UTC from IEEE Xplore. Restrictions apply.
t t
(n-1)th x[n-1] nth Switching Period x[n] (n+1)th (n-1)th x[n-1] nth Switching Period x[n] (n+1)th
0 DT T T+DT 2T 2T+DT 0 TíDT T 2TíDT 2T
Stage 1 Stage 2 Stage 3 Stage 4 Stage 1 Stage 2 StageEStage
2 Stage
1 3 StageEStage
4 Stage
2 1
x[n]=F`(x[n-1],D,T,u) x[n]=Fs(x[n-1],D,T,vi)
Fig. 4. State variables iteration relationship for the proposed Fig. 5. Simplified iterative relationship for the proposed sampled-data
modeling method modeling method.
Based on Fig. 4, the iterative mapping between two Comparing with the massive computational efforts needed
sampling point is expressed as to investigate the conventional model (6), the proposed model
saves the amount of calculation via two approaches:
x > n@ F c x > n 1@ , D, T , vi G c D x > n 1@ H c D vi (7)
1) G c( D) is invariant within the whole operation range, while
where,
Gc D of has different value at every operating point;
G c( D) e A1 DT
e A4 (1 D )T
e A3 DT
e A2 (1 D )T
2) only two stages need to be computed in one sampling
(1 D )T
H c( D) ³ period.
A1 DT A4 (1 D )T A3 DT
e e e e A2t dtB2
0
DT (1 D )T
³ e dtB3 e A1DT ³
A4 (1 D )T
e A1 DT
e A3t
e A4t dtB4 1
0 0
DT 1
³ A1t
e dtB1 ] S(t)
0
According to (1), A1= A 4 and A 2= A 3, the G c D can be í1
simplified as DT T DT+T 2T 2T+DT
G c( D) e A1DT e A4 (1 D )T e A3 DT e A2 (1 D )T e A1T e A2T (8) Fig. 6. The waveform of S(t)
Via the transformation as (8), the state matrix G c( D)
becomes a constant matrix that only depends on the circuit The sampled-data model can be built through the piecewise
parameters, and it is decoupled with control signal D state-space description, and vice versa, we can also derive
c state-space function from the sample-data model. Therefore, by
(i.e., wG ( D) 0 ). If input matrix H c D can be
wD applying the G c( D) and H Sc D , a unified state equations is
transformed into the following form as (9), we can build an given by
equivalent circuit based on G c( D) and H Sc D .
dx
T T Aeq x Beq vi (11)
H Sc D e A1T
³ e dW B1 ³ e A1W dW B 2
A2W
(9) dt
0 0
where,
By combining (7) and (9), the analytic expressions for B1
and B 2 are given as § 4 Res S t · § 2 D 1 S t ·
¨ ¸ ¨ ¸
¨ Ls Ls ¸ ¨ Ls ¸
Aeq Beq
T
e A2W dW B1 ¨ S t 1 ¸ ¨ D(1 D) ¸
³0
e A3 DT A21 e A2 (1 D )T I B1 A31 e A3 DT I B2
¨ ¸ ¨¨ 2 f L C ¸¸
¨ C Co RL ¸¹
T © o © ¹
³ e A1W dW B 2
s s o
e A1 DT A41 e A4 (1 D )T I B2 A11 e A1 DT I B1 (10)
0 Note the S(t) is a fixed ac function whose the waveform is
§ 2D 1 · § 1 2D · shown in Fig. 6.
¨ Ls ¸ ¨ Ls ¸
The equivalent circuit is developed according to the
B1 | ¨ ¸ B | ¨ ¸
following rules: the off-diagonal elements of Aeq denotes state
¨ D(1 D) ¸ 2
¨ D(1 D) ¸
¨ ¸ ¨ ¸ variable controlled source; the elements of Beq represents input
© 2 f s Ls Co ¹ © 2 f s Ls Co ¹ variable controlled source. Therefore, the circuit oriented
According to G c( D) and H Sc D , a simplified iterative model of DAB converter is built as in Fig. 8.
relationship that contains only two stages in one sampling
period is illustrated as in Fig. 5.
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on September 20,2024 at 06:27:28 UTC from IEEE Xplore. Restrictions apply.
(1-2d)iLS(t)
La Res iL
iLS(t) B. AC Branch
CCCS CCCS
+ +
The ac branch of the proposed equivalent circuit consists of
vg
Auxiliary
Cs vo RL the controlled sources that include S(t). This branch can be
Power Flow
used to take into account the effects induced by Res. Since S(t)
VCVS VCVS is half-period symmetric (i.e., S(t)=íS(t+T)), only the first half
VCVS (1-2d)vgS(t) voS(t)
(1-d)dvo/2fsLa
VCCS switching cycle needs to be analyzed. At steady-state, it is
Dominated Power Flow (1-d)dvg/2fsLa
reasonable to neglect the ripple of output voltage vo, then, iL(t)
Fig. 7. The proposed equivalent circuit of DAB converter. satisfies the following equation within the first half switching
cycle
Note that, being different from the traditional equivalent diL 1
circuit model that is derived by averaging the switching devices, 1 2 D vg vo Res iL k aiL (13)
this model is obtained based on sampled-data model. Therefore, dt L
the proposed equivalent circuit captures the dynamics of where,
original system at sampling points. This phenomenon is further Res
1
illustrated in Section IV. k 1 2 D vg vo a
L L
Due to iL(t)= íiL(t+T) at steady state, the solution of (13) is
III. ANALYSIS OF THE EQUIVALENT CIRCUI expressed as
As in Fig. 7, there are two kinds of controlled source, one
kind contains the ac function, S(t), and the other kind are 2b b
iL t e at (14)
independent with S(t). The whole equivalent circuit, therefore, a 1 e aT
a
can be divided into two parts: the ac branch and the dc branch.
Therefore, the averaged output current of the ac controlled
(1-2d)iLS(t)
La Res iL
iLS(t) source, iL(t)×S(t), is
CCCS CCCS
+ + 2b e aT 1
I = 1 T i t S t dt = b
vg Cs vo RL ac
T ³0 L aT
a e 1 T a
2
(15)
VCVS VCVS
VCVS (1-2d)vgS(t) voS(t) According to the equivalent circuit shown in Fig. 0, the
VCCS
(1-d)dvo/2fsLa (1-d)dvg/2fsLa output capacitor, Co, is fed by two controlled current source,
DC Branch one is of dc branch, and the other belongs to ac branch. By
applying the principle of charge-balance to the output capacitor
(a) over one switching cycle, we have
AC Branch
2k e aT 1 k D 1 D T vo
(1-2d)iLS(t)
La Res iL
iLS(t)
aT
(16)
CCCS CCCS a e2
1 a L RL
+ +
Cs vo RL
Therefore, the equation to calculate the output voltage with
vg
considering the influence of Res is expressed as
VCVS VCVS
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on September 20,2024 at 06:27:28 UTC from IEEE Xplore. Restrictions apply.
TABLE II. COMPARISON OF THE OUTPUT VOLTAGE OBTAINED BY EXPERIMENTS, SIMULATIONS, IDEAL MODEL, AND PROPOSED MODEL
TABLE I. CIRCUIT PARAMETERS OF THE PROTOTYPE experimental results and the simulations of detailed switching
circuit are substantially different from the predictions of the
Items Descriptions Specifications lossless model. In contrast, the output voltages calculated by
vi Input Voltage 30V (17) match the experiments and the simulations very well.
fs Switching Frequency 20kHz From these results, the validity of the proposed equivalent
Auxiliary Inductor 19.7ȝH+2.1ȝH
Inductor Magnetic Material Sendust alloy
circuit can be demonstrated.
core material To further illustrate the accuracy of the full-order
Turn Ratio 1:1
Transformer Magnetic Material Ferrite core
equivalent circuit introduced in this paper, Fig. 10 compares
material the simulating waveforms of this model with that of the
MOSFET CSD18536KTT detailed switching circuit. The PI controller is used to tract the
reference voltage, and the parameters are as follow: kp=1, ki=5,
RL=10ȍ, Res=40mȍ, Vin=50V, Vref=40V, Ls=50ȝH, n=1:1,
The Drain-to-Source On Resistance of MOSFET is 1.3 mȍ; Co=300ȝF. In this case, the closed-loop system is unstable. And
the total copper resistance of the inductor and high frequency the sampling points are also marked by red dots in Fig. 10(b).
transformer is 69.5 mȍ; and extra resistors, Res,e=105.3 mȍ, As the figures show, the proposed model can precisely retain
are connected in series with the inductor. In summary, the oscillations of the actual circuit at sampling points.
Res 1.3 u 4 69.5 105.3 m: 181m: . In practical
implementation, the dead-time between the gate driving signal
of the upper switching device and that of the lower switching
device is needed. This dead-time is measured by oscilloscopes
and is also considered in simulations of switching circuit and
theoretical calculations.
(a)
(a)
(b)
Fig. 10. Comparison between the simulating results of the equivalent
circuit model and that of the detailed switching circuit at unstable
operation: (a) waveforms of output voltage; (b) waveforms of
inductor current.
(b)
Fig. 9. (a) Comparison of the experimental results, the simulations, V. CONCLUSION
the lossless model, and the proposed model. (b) Experimental
waveforms of DAB converter with RL=80ȍ. This paper presents a novel circuit oriented model of DAB
converter. The proposed model is built by using the discrete
The experimental results are visualized in Fig. 9(a) and the time modeling method rather than by averaging the state
data of the measurement results is listed in Table II. Fig. 9(b) variables, and small signal assumption is not needed. Therefore,
depicts the experimental waveforms. As listed in Table II, the this equivalent circuit of DAB converter can reproduce the
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on September 20,2024 at 06:27:28 UTC from IEEE Xplore. Restrictions apply.
steady state response and large signal transient response of the [7] R. D. Middlebrook and S. Cuk, “A general unified approach to
original system for both open loop and close loop operation. modelling switching-converter power stages,” in 1976 IEEE Power
Electronics Specialists Conference, 1976, pp. 18–34.
Moreover, an improved steady-state models of DAB converter
[8] R. W. Erickson and D. Maksimovic, Fundamentals of power
is derived based on the proposed model. Finally, simulations electronics. Springer Science & Business Media, 2007.
and experiments confirm the correctness of theoretical results. [9] Hengsi Qin and J. W. Kimball, “Generalized Average Modeling of
Dual Active Bridge DC–DC Converter,” IEEE Trans. Power
Electron., vol. 27, no. 4, pp. 2078–2084, Apr. 2012.
[10] D. Costinett, R. Zane, and D. Maksimovic, “Discrete-time small-
REFERENCES signal modeling of a 1 MHz efficiency-optimized dual active bridge
converter with varying load,” in Control and Modeling for Power
Electronics (COMPEL), 2012 IEEE 13th Workshop on, 2012, pp. 1–7.
[1] M. N. Kheraluwala, R. W. Gascoigne, D. M. Divan, and E. D. [11] D. Costinett, “Reduced order discrete time modeling of ZVS transition
Baumann, “Performance characterization of a high-power dual active dynamics in the dual active bridge converter,” in 2015 IEEE Applied
bridge DC-to-DC converter,” Ind. Appl. IEEE Trans. On, vol. 28, no. Power Electronics Conference and Exposition (APEC), 2015, pp.
6, pp. 1294–1301, 1992. 365–370.
[2] R. M. Burkart and J. W. Kolar, “Comparative Ș-ȡ-ı Pareto [12] L. Shi, W. Lei, J. Huang, Z. Li, Y. Cui, and Y. Wang, “Full discrete-
Optimization of Si and SiC Multilevel Dual-Active-Bridge Topologies time modeling and stability analysis of the digital controlled dual
With Wide Input Voltage Range,” IEEE Trans. Power Electron., vol. active bridge converter,” in 2016 IEEE 8th International Power
32, no. 7, pp. 5258–5270, Jul. 2017. Electronics and Motion Control Conference (IPEMC-ECCE Asia),
[3] S. Inoue and H. Akagi, “A Bidirectional Isolated DC-DC Converter as 2016, pp. 3813–3817.
a Core Circuit of the Next-Generation Medium-Voltage Power [13] Z. Li, Y. Wang, L. Shi, J. Huang, Y. Cui, and W. Lei, “Generalized
Conversion System,” IEEE Trans. Power Electron., vol. 22, no. 2, pp. averaging modeling and control strategy for three-phase dual-active-
535–542, Mar. 2007. bridge DC-DC converters with three control variables,” in 2017 IEEE
[4] N. M. L. Tan, T. Abe, and H. Akagi, “Design and Performance of a Applied Power Electronics Conference and Exposition (APEC), 2017,
Bidirectional Isolated DC-DC Converter for a Battery Energy Storage pp. 1078–1084.
System,” IEEE Trans. Power Electron., vol. 27, no. 3, pp. 1237–1248, [14] L. Corradini, D. Maksimoviü, P. Mattavelli, and R. Zane, Digital
Mar. 2012. Control of High-Frequency Switched-Mode Power Converters.
[5] F. Krismer and J. W. Kolar, “Accurate Small-Signal Model for the Hoboken, NJ, USA: John Wiley & Sons, Inc, 2015.
Digital Control of an Automotive Bidirectional Dual Active Bridge,” [15] F. Zhang, M. M. U. Rehman, R. Zane, and D. Maksimoviü,
IEEE Trans. Power Electron., vol. 24, no. 12, pp. 2756–2768, Dec. “Improved steady-state model of the dual-active-bridge converter,” in
2009. 2015 IEEE Energy Conversion Congress and Exposition (ECCE),
[6] X. She, X. Yu, F. Wang, and A. Q. Huang, “Design and 2015, pp. 630–636.
Demonstration of a 3.6-kV-120-V/10-kVA Solid-State Transformer
for Smart Grid Application,” IEEE Trans. Power Electron., vol. 29,
no. 8, pp. 3982–3996, Aug. 2014.
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on September 20,2024 at 06:27:28 UTC from IEEE Xplore. Restrictions apply.