Theoretical Analysis of Bidirectional Bifurcation

Phenomena in DC Cascaded Converter Systems

Li Ding Chikong Tse
Department of Electrical Engineering Department of Electrical Engineering
City University of Hong Kong SAR, China City University of Hong Kong SAR, China
Email: [email protected] Email: [email protected]

Abstract—DC-DC converters are major components of a DC through the search of the bifurcation points. Furthermore, by
cascaded converter system. The converters interface with external varying the system’s parameters, the effect of such parameters
power inputs, internal DC buses and loadings of subsystems. on the stable operating region of the system can be investigated
Stability of these constituent converters under all operating
conditions is vital to the operation of the whole system. Current [12]. In [13], the bifurcation phenomena of a DC cascaded
research results show that the system’s stability can be guaran- converter with multi-load converters have been studied through
teed when the system’s parameters are within specific ranges, the discrete-time mapping model. The results show that the
which are similar to the results obtained from standalone DC- system may exhibit fast-scale, slow-scale and coexisting fast-
DC converters. However, DC cascaded systems are high order scale and slow-scale instabilities under varying input voltage.
systems and the dynamic behaviors are more complex than
their standalone counterparts. In this paper, we identify and For the DC cascaded converter with averaged current-mode
analyze the bidirectional bifurcation phenomena in the current- control, the system may exhibit border collision bifurcation
mode controlled DC cascaded system. Based on the derived under the specific parameter changes [14]. For the DC cas-
averaged model, the bidirectional bifurcation phenomena and caded system with the renewable energy source input, the
the associated mechanism will be studied in depth. Finally, the system may exhibit catastrophic bifurcation under an abrupt
validity of the analysis is verified by laboratory experiments.
Index Terms—DC cascaded system, bidirectional bifurcation,
load change situation [15]. For design-oriented proposes, the
averaged model, design-oriented analysis above works have also provided the stability boundaries in the
practical parameter space. The stability boundaries in these
I. I NTRODUCTION prior works divide the parameter space into a stable operating
region and an unstable operating region. This implies that the
DC cascaded converter systems have been widely used system can operate stable if some of the circuit parameters
in micro-grids, electric vehicles, commercial communications are in specific closed or open ranges. In addition, a large
systems, computers, and other applications because of the high capacitance can reduce the peak value of the output impedance
reliability, ease of standardization, expansion, maintenance and of the source converter, which can help the system satisfy
hot plug capability [1]–[3]. Within a DC cascaded converter the impedance-based stability criteria more easily [16], [17].
system, power sources and loads are connected with inter- Therefore, engineers tend to increase the magnitude of the
facing power converters via a DC bus. In this system, the output capacitor of the source converter when the system
subsystems may interact with each other via the DC bus and exhibits oscillation. However, the DC cascaded system is a
such interactions may destabilize the whole system [4], [5]. high order system, and the movement of the eigenvalues of
Much research effort has been devoted to identify the the system with the variation of the circuit’s parameters may
stability of the DC cascaded converter system. The most be different from conventional standalone DC-DC converters.
common method is to apply the impedance-based criterion In recent years, bidirectional bifurcation have been found in
which is proposed by Middlebrook in 1976 [6]. The criterion the DC cascaded system [18]. However, the analysis of this
essentially states that the stability of the cascaded system phenomenon has been rarely reported.
can be guaranteed if both the source converter and the load In this paper, the bidirectional Hopf bifurcation phenomena
converter are stable individually and the magnitude of the are discovered in the DC cascaded system. Such bifurcation
output impedance of the source converter is smaller than the phenomena will be analyzed through an averaged model and
magnitude of the input impedance of the load converter within the movement of the eigenvalues will be shown under a certain
the entire frequency range. choice of parameter range, which shows the complex dynamics
In the past few years, the nonlinear phenomena in the of the DC cascaded system. Besides, the mechanism of the
DC cascaded system have attracted much attention, and some bidirectional bifurcation will be discussed based on the derived
investigations have focused on the bifurcation behaviors in averaged model.
such systems [7]–[11]. The bifurcation analysis method is an This paper is organized as follows. In Section II, the
effective way to analyze the stability of the system. Essentially, cascaded system under study will be introduced. Then, the bi-
the stable operating region of the system can be obtained furcation phenomena and the analysis based on the movement
 Source Load Si Li rLi
E converter converter RL vo

vdi isensei
Di Ci vCi

Fig. 1. Typical structure of the DC cascaded system.

(a)
of the eigenvalues obtained through the averaged model will
isensei
be given. In Section III, the mechanism of the bifurcation phe- Caci Raci
Ravi
vCi
Rfci Cavi
nomena will be discussed. Section IV provides experimental vaci R1i
vavi Rfvi
verifications of the key results. Finally, Section V concludes vdi E/A
E/A R2i
this paper. Vrefi
Vrampi
II. B IFURCATION P HENOMENA AND A NALYSIS
(b)
The DC cascaded converter system studied in this paper
is shown in Fig. 1. For illustrative purposes, both the source Fig. 2. (a) Buck converter circuit as the source converter and the load
converter; (b) controller circuit for each converter.
and the load converters are constructed using buck converters.
Source converter B1 regulates the power from source E to the
input of load converter B2 . Also, load converter B2 regulates 17
power to the load RL . Each converter as shown in Fig. 2(a) 16.5

contains an inductor Li with a loss resistor rLi , a capacitor 16

Ci , a diode Di and a switch Si , where i = 1, 2. Switch Si 15.5

of Bi is controlled using the controller shown in Fig. 2(b), 15

vC1(V)
14.5
which adopts the common inner-loop average current control 14
and outer-loop voltage control. Each compensation network 13.5

employs the usual proportional-integral compensation. The 13

switching frequency is 50 kHz. The values of the circuit 12.5

components are given in Table I, where gv = Rav /Rf v , 12

100 200 300 400
C1(µF)
500 600

τv = Rf v Cav , Kv = R2 /(R1 + R2 ), gc = Rac /Rf c ,

(a)
τc = Rf c Cac . In the following sections, subscripts 1, 2 in the
elements’ symbols are used to represent the source converter 8.6

and the load converter, respectively. 8.55

8.5
Capacitance C1 is firstly selected as the bifurcation param-
8.45
eter and the bifurcation diagram is shown in Fig. 3(a). Here, 8.4
the bifurcation diagram is developed for C1 within 100 µF
vC2(V)

8.35

and 650 µF. From Fig. 3(a), it can be found that the system 8.3

can be stable only in the range 310 µF < C1 < 450 µF. The 8.25

8.2
result indicates that the system may become unstable if the
8.15
value of the DC link capacitor C1 is either too large or two 8.1
100 150 200 250 300 350 400 450
small. Therefore, increasing the value of the DC link capacitor L2(µH)

cannot ensure the stable operation of the DC cascaded system. (b)

Then, inductance L2 is selected as the bifurcation parameter
and the bifurcation diagram is shown in Fig. 3(b). Here, the Fig. 3. Bifurcation diagrams. (a) C1 as the bifurcation parameter; (b) L2 as
the bifurcation parameter.
bifurcation diagram is developed for L2 within 100 µH and
450 µH. From Fig. 3(b), it can be found that the system is
kept stable only in the range 160 µH < L2 < 300 µH. The
TABLE I
result indicates that the system may become unstable if the C IRCUIT C OMPONENTS
value of L2 is either too large or too small.
From Fig. 3, it can be observed that as the bifurcation Component B1 B2 Component B1 B2
parameters increase, two bifurcation regions on both sides of L / µH 220 220 C / µF 470 470
the stable operating region have emerged. Such a bifurcation rL / Ω 0.1 0.1 RL 4
phenomenon can be regarded as bidirectional bifurcation gv 0.1 0.15 gc 1 0.5
which is rarely reported and analyzed in prior works. Detailed τv / ms 1 1 τc / ms 0.15 0.1
analysis of this type of the bifurcation will be given in the Kv 0.38 0.6 Vref / V 5 5
next section.
 4
x 10
1 8000

0.8
6000
0.6
4000
0.4
2000
0.2

Im(λ)

Im(λ)
0 0

−0.2
−2000
−0.4
−4000
−0.6
−6000
−0.8

−1 −8000
−2000 −1500 −1000 −500 0 500 −2000 −1500 −1000 −500 0 500
Re(λ) Re(λ)

(a) (b)

(c) (d)

Fig. 4. Loci of the eigenvalues. (a) With the variation of C1 ; (b) with the variation of L2 ; (c) enlargement of (a); (d) enlargement of (b).

A. Bifurcation Analysis TABLE II

C IRCUIT O PERATING S TATES
Fig. 3, it can be observed that the bifurcation phenomenon
found in the DC cascaded system is a slow-scale bifurcation. State S1 D1 S2 D2
Therefore, the averaged model can be used to analyze this 1 on off on off
phenomenon. A summary of the procedure for developing the 2 on off off on
averaged model is given as follows. 3 off on on off
The cascaded system shown in Fig. 1 is an 8th-order system. 4 off on off on
The state vector x containing two 4-dim column vectors
x1 and x2 for the source converter and the load converter,
respectively, is chosen as follows: Expressions of Aj and Bj can be readily found and are
x = [x1 x2 ]
T
(1) omitted here.
The essential step in the modeling is to ”average” out
where the system matrices, yielding the following continuous-time
T
xi = [iLi vCi vavi vaci ] for i = 1, 2. (2) averaged model:
The two converters are designed for a two-stage CCM
F (x) = ẋ = Am x + Bm E. (4)
operation. Without loss of generality, we can assume that
the cascaded converters share a common period T and are Here, the coefficient matrixs Am and Bm of the averaged
synchronized, with four operating states as shown in Table II. model are the weighted average values of Aj and Bj (1 ≤
The first step in establishing the averaged model of the j ≤ 4), respectively, which means that:
system is to write down the state equations in different
4
switching modes. Since both two converters operate in CCM, X
Am = (dj − dj−1 )Aj , (5)
the state sequence Zj (for j = 1 . . . 4) of the states observed
j=1
within a switching period are Z1 = (1, 2, 4) or Z2 = (1, 3, 4).
The state-j equation, where j = 1, 2, 3, 4, within a period is
4
given by X
Bm = (dj − dj−1 )Bj . (6)
ẋ = Aj x + Bj E state j. (3) j=1
 vC1: [500 mV/div]
vC1: [500 mV/div]

vC2: [500 mV/div]

vC2: [500 mV/div]

Time : [1 ms /div]
Time : [1 ms /div]

(a) (b)

vC1: [500 mV/div] vC1: [500 mV/div]

vC2: [500 mV/div] vC2: [500 mV/div]

Time : [1 ms /div] Time : [1 ms /div]

(c) (d)

Fig. 5. Experimental waveforms for (a) C1 = 150 µF , (b) C1 = 570 µF , (c) L2 = 120 µH and (d) L2 = 400 µH.

The average modeling method eliminates the piecewise III. D ISCUSSION OF THE M ECHANISM OF THE
characteristic of the state equations, and the analytical expres- B IDIRECTIONAL B IFURCATION
sion of dj can be derived based on the equation (4).
From the previous analysis, it can be found that the bidi-
B. Derivation of the Jacobian rectional bifurcation may occur if one pair of complex eigen-
values cross the imaginary axis twice or two pairs of different
Setting all the time derivatives of the state variables in
complex eigenvalues cross the imaginary axis alternatively as
equation (4) to 0, we can obtain the equilibrium point xe of
the parameters vary. For the conventional standalone converter
the system, and its Jacobian of the neighborhood can also be
which usually adopts a voltage-mode control or current-mode
obtained. The stability of the system can be determined by
control, the system is a third or fourth order system. The
investigating the eigenvalues of the Jacobian.
eigenvalues of the Jacobian matrix usually contains both real
Using the parameters listed in Table I, the loci of the eigenvalues and complex eigenvalues. This implies that the Ja-
eigenvalues of the system can be generated, as shown in Fig. cobian matrix does not have two pairs of complex eigenvalues.
4. From Figs. 4 (a) and (c), it can be observed that with the Therefore, the phenomenon that two pairs of different complex
value of the DC link capacitor C1 increasing, a pair of complex eigenvalues cross the imaginary axis alternatively does not
eigenvalues may cross the imaginary axis from the right half occur. In addition, for the conventional standalone converter,
plane to the left half plane, and then cross the imaginary axis its complex eigenvalue can be written as λ = a±jb, where ’a’
again to the right half plane. The trajectories of the eigenvalues and ’b’ can be represented by the circuit parameters. If these
indicate that the system may occur Hopf bifurcation if the complex eigenvalues cross the imaginary axis, then we have
value of C1 is either too large or too small. a = 0. However, usually, a unique value exists satisfying this
From the above the analysis, it can be found that, when condition for the conventional standalone converter. Therefore,
C1 is selected as the bifurcation parameter, one pair of it is not possible for the complex eigenvalue to cross the
the complex eigenvalues cross the imaginary axis twice as imaginary axis twice. Because of the page limit, detailed
the bifurcation parameter changes. Therefore, a bidirectional derivation will be omitted here.
bifurcation phenomenon occurs.
On the other hand, when L2 is selected as the bifurcation IV. V ERIFICATION BY S IMULATION AND E XPERIMENT
parameter, the bidirectional bifurcation will occur because two
pairs of different complex eigenvalues cross the imaginary axis A DC cascaded system has been constructed to verify the
alternatively with the increasing of the value of L2 , as shown unstable phenomenon analyzed in the previous sections. The
in Figs. 4 (b) and (d). parameters shown in Table I are used. Here, the DC link
Therefore, two types of the bidirectional bifurcation phe- capacitor C1 and the inductor of the load converter L2 are
nomena can be observed in the DC cascaded system and the chosen as the bifurcation parameters. Typical AC components
mechanisms of these two types of bidirectional bifurcation of the waveforms are shown in Fig. 5, which match the
phenomena are different. theoretical analysis given in the previous sections.
 V. C ONCLUSION [5] S. Singh, A. R. Gautam, and D. Fulwani, “Constant power loads and
their effects in dc distributed power systems: A review,” Renew. Sustain.
This paper investigates a common DC cascaded converter Energy Rev., vol. 72, pp. 407–421, Jan. 2017.
system consisting of two averaged current-mode controlled [6] R. D. Middlebrook, “Input filter considerations in design and application
buck converters. The bidirectional bifurcation phenomena have of switching regulators,” in Proc. IEEE IEEE IAS Annual Meeting, 1976,
pp. 366-382
been discovered in this system. The bifurcation phenomena
[7] X. Xiong, C. K. Tse, and X. Ruan, “Bifurcation analysis of standalone
are first observed through the bifurcation diagrams with the photovoltaic-battery hybrid power system,” IEEE Trans. Circ. Syst. I,
variation of the DC link capacitor and the inductor of the load Reg. Papers, vol. 60, no. 5, pp. 1354–1365, May 2013.
converter. Based on an averaged model, the mechanism of the [8] A. Kwasinski and C. Onwuchekwa, “Dynamic behavior and stabilization
of dc microgrids with instantaneous constant-power loads,” IEEE Trans.
bidirectional bifurcation has been analyzed and two types of Power Electron., vol. 26, no. 3, pp 822–834, 2011.
bidirectional bifurcations are identified. Type one bifurcation [9] X. Yue, D. Boroyevich, F. C. Lee, F. Chen, R. Burgos, and F. Zhuo,
will occur since one pair of complex eigenvalues cross the “Beat frequency oscillation analysis for power electronic converters in dc
imaginary axis twice as the DC link capacitor varies. Type nanogrid based on crossed frequency output impedance matrix model,”
IEEE Trans. Power Electron., vol. 33, no. 4, pp. 3052–3064, Apr. 2018.
two bifurcation will occur since two pairs of different complex [10] X. Li, X. Ruan, X. Xiong, Q. Jin, and K. T. Chi, “Stability issue of
eigenvalues cross the imaginary axis alternatively with the cascaded systems with consideration of switching ripple interaction,”
variation of the inductor of the load converter. The results IEEE Trans. Power Electron., vol. 34, no. 7, pp. 7040–7052, Jul. 2019.
[11] G. Zhou, Y. Li, M. Leng, Z. Gong, and G. Xu, “Stabilizing effect of
indicate that DC cascaded systems exhibit more complex load converter in cascaded system considering ripple interaction,” IEEE
behaviors than standalone DC-DC converters. The stable oper- Trans. Circ. Syst. II: Express Briefs, vol. 68, no. 1, pp. 296–300, Jan.
ating regions and the unstable operating regions may emerge 2021.
alternately with the variation of the parameters, which makes [12] C. K. Tse, Complex behavior of switching power converters. Boca Raton,
FL, USA: CRC press, 2003.
the design of the cascaded systems more complicated. Finally, [13] F. Xie, B. Zhang, D. Qiu, and Y. Jiang, “Non-linear dynamic behaviours
the two types of bidirectional bifurcation are verified through of dc cascaded converters system with multi-load converters,” IET Power
experimental measurements. Electron., vol. 9, no. 6, pp. 1093–1102, may 2016.
[14] L. Ding, S.-C. Wong, and K. T. Chi, “Bifurcation analysis of a current-
R EFERENCES mode-controlled dc cascaded system and applications to design,” IEEE
J. Emerging Sel. Topics Power Electron., vol. 8, no. 4, pp. 3214–3224,
[1] S. Luo and I. Batarseh, “A review of distributed power systems part
Dec. 2020.
I: DC distributed power system,” IEEE Aerosp. Electron. Syst. Mag.,
vol. 20, no. 8, pp. 5–16, Aug. 2005. [15] M. Huang, H. Ji, J. Sun, L. Wei, and X. Zha, “Bifurcation-based stability
[2] S. Luo and I. Batarseh, “A review of distributed power systems Part II. analysis of photovoltaic-battery hybrid power system,” IEEE J. Emerging
high frequency AC distributed power systems,” IEEE Aerosp. Electron. Sel. Topics Power Electron., vol. 5, no. 3, pp. 1055–1067, Sep. 2017.
Syst. Mag., vol. 21, no. 6, pp. 5–14, Jun. 2006. [16] S. Abe, M. Hirokawa, M. Shoyama, and T. Ninomiya, “Optimal bus
[3] R. Haroun, A. Cid-Pastor, A. El Aroudi, and L. Martinez-Salamero, capacitance design for system stability in on-board distributed power
“Synthesis of canonical elements for power processing in dc distribution architecture,” in Proc. EPE-PEMC, Sep. 2008, pp. 393–399.
systems using cascaded converters and sliding-mode control,” IEEE [17] X. Zhang, X. Ruan, and C. K. Tse, “Impedance-based local stability
Trans. Power Electron., vol. 29, no. 3, pp. 1366–1381, Mar. 2014. criterion for dc distributed power systems,” IEEE Trans. Circ. Syst. I,
[4] S. Liu, P. X. Liu, and X. Wang, “Stability analysis of grid-interfacing Reg. Papers, vol. 62, no. 3, pp. 916–925, Mar. 2015.
inverter control in distribution systems with multiple photovoltaic-based [18] X. Zhang, T. Wang, H. Bao, Y. Hu, and B. Bao, “Stability effect of
distributed generators,” IEEE Trans. Ind. Electron., vol. 63, no. 12, pp. load converter on source converter in a cascaded buck converter,” IEEE
7339–7348, May 2016. Trans. Power Electron., vol. 38, no. 1, pp. 604–618, Jan. 2022.