Large-Signal Stability Analysis of DC Distribution
Large-Signal Stability Analysis of DC Distribution
Abstract—DC-DC converters are major components of individually and the magnitude of the output impedance of
the DC distribution systems. The converters interface with the source converter is smaller than the magnitude of the input
external power inputs, internal DC buses and loadings impedance of the load converter within the entire frequency
of subsystems. The interfacing DC-DC converters should
be stable locally and globally under mutual interactions range. Subsequently, many impedance based criteria [8]–[12]
through a DC bus within the system. Current research have been developed aiming to narrow the forbidden range
efforts have focused on the analysis of the stability of the given by Middlebrook’s criterion and increase the degree of
DC distribution system subject to small-signal disturbance. freedom of a design.
However, in practice, the system routinely operates un- The above impedance-based criteria are derived from the
der large-signal disturbances, such as when an additional
subsystem is turned on after being connected to the DC small-signal stability analysis, which permits fast calculation,
bus. In this scenario, the small-signal model may fail to but can be inaccurate for large-signal operations. Therefore,
fully describe the dynamics of the system. In this paper, some large-signal analysis methods have been proposed. The
we identify and analyze the bifurcation process when the behavior of a DC cascaded system under large-signal dis-
system undergoes abrupt load changes. According to the turbance has been described and analyzed by phase-plane
nonlinear operation of the interconnected system, a large-
signal stability criterion is derived. This criterion is simple, analysis [13]–[15]. In these previous works, the system’s dif-
and can be easily extended to multiple connected converter ferential equations are graphically solved (plotted), providing
systems. The criterion is also consistent with the result the trajectories on the phase plane. Specifically, the graphical
from bifurcation analysis. Finally, the validity of the pro- and experimental results show that the DC bus voltage and the
posed criterion is verified by the full-circuit simulations and output voltage of the system may collapse under sudden load
the experimental works.
changes. The phase-plane analysis is explicit and suitable for
Index Terms—Bifurcation analysis, constant power load, numerical simulations. However, detailed internal parameters
DC distribution system, design-oriented analysis, large- should be specified and different parameters produce different
signal disturbance, stability criterion
trajectories. Also, it does not readily generate an analytical and
general relationship between the parameters and the system’s
I. I NTRODUCTION stability.
Apart from the phase-plane analysis, Lyapunov-based meth-
D C distribution systems are widely used in micro-grids,
electric vehicles, communication systems and other
power supply applications [1]–[3]. Within a DC distribution
ods are the effective and widely used to analyze the large-
signal stability of the system. In reference [16], the stability of
a three-phase two-level power converter under different time
system, power sources and loads are connected with interfac-
scales has been analyzed based on the Lyapunov function.
ing power converters via a DC bus. In this system, interacting
The Lyapunov stability theorem is also applied to the DC-
power converters should be designed for a stable operation
DC converters to obtain the stability of global asymptotic
according to some system design criteria [4]–[6].
conditions [17].
The stability criteria of DC distribution systems can be de-
However, for more complex power electronics systems, it is
rived from small-signal and large-signal points of view. For the
difficult to find the Lyapunov function. Some methods, such as
small-signal stability analysis, the first criterion was proposed
the Takagi-Sugeno multi-modeling [18] and Brayton-Moser’s
by Middlebrook in 1976 [7] for a DC cascaded system which
mixed potential function theory [19], have been developed to
is the ancestor of today’s micro-grid system. According to
generate the Lyapunov function. The Brayton-Moser mixed
Middlebrook’s criterion, the stability of a system of two DC-
potential function theory can be used to analyze the large-
DC converters in cascaded connection can be guaranteed if
signal stability of the nonlinear circuits and obtain the analyt-
both the source converter and the load converter are stable
ical solution of the stable operating region [20]. In reference
[21], a large-signal stability criterion has been derived with
Manuscript received Month xx, 2022; revised Month xx, xxxx; ac- mixed potential theory to analyze the catastrophic bifurcation
cepted Month x, xxxx. This work is supported by Hong Kong Research
Grant Council under GRF 112071/21E. phenomenon of the photovoltaic-battery hybrid power system
L. Ding is with the Department of Electronic and Information Engi- under large-signal disturbance. In reference [22], based on an
neering, The Hong Kong Polytechnic University, Hunghom, Hong Kong. equivalent gyrator model of the buck converter, a criterion
(Email: [email protected]).
C. K. Tse is with the Department of Electrical Engineering, City Uni- derived from the mixed potential theory has been used to
versity of Hong Kong, Kowloon, Hong Kong. (Email: [email protected]). study the large-signal stability of a current-mode controlled
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
10
II. B IFURCATION P HENOMENON AND A NALYSIS
6
A typical structure of a DC distribution system with one 7
vo1(V)
15
method can be extended to any other types of converters. In
0
this system, the bus voltage is regulated by Bs . All converters 0.15 0.16 0.17 0.18 0.19 0.2
t(s)
are controlled with three independent voltage-mode controller
circuits, as shown in Fig. 2. The maximum duty cycle of Fig. 3. Voltage collapse on the input port of the load converters and the
these three converters are clamped at 0.9. The power source E output voltage of B1 after B2 is connected to the DC bus.
shown in Fig. 1 is 24 V. Also, rLine = 0.6 Ω and LLine = 5 µH
account for the wire impedance from the source converter to
load converters. The values of the circuit components used
in the full circuit simulation are given in Table I, where TABLE I
g = Ra /Rf , τ = Ra Ca and Kv = Rd /(Rd + Rc ). In the C IRCUIT C OMPONENTS
following sections, the subscripts 1, 2 and s in the symbols Component B1 B2 Bs Component B1 B2 Bs
are used to represent B1 , B2 and Bs , respectively. L / µH 470 220 220 g 19.6 0.02 19.6
C / µF 680 680 680 τ /s 0.1 0.1 0.01
rL / Ω 0.5 0.5 0.5 Kv 0.84 0.17 0.36
A. Bifurcation Phenomenon R/Ω 6 30 – Vref / V 5 5 5
vC / V 6 30 14 f / kHz 16 16 16
In practice, the source converter and the load converters
are decentralized in a DC distribution system and the load
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
converters may not connect to the DC bus at the same time. TABLE II
Without loss of generality, we assume that Sl1 is turned on C IRCUIT O PERATING S TATES
before t = 0 s, which means that source converter Bs and State Ss Ds S1 D1 S2 D2
load converters B1 are stable. After t = 0.15 s, Sl2 is turned
1 off on off on off on
on and B2 is connected to the DC bus.
2 off on off on on off
Fig. 3 shows the transient waveforms of the system when
3 off on on off off on
Sl2 is turned on some time after Sl1 is turned on. After B2
4 off on on off on off
is connected to the DC bus, it can be observed that the input
5 on off off on off on
voltage of the load converters (vbusl ) drops to 7 V. This voltage
6 on off off on on off
drop will also make the voltage input to B1 and B2 a bit lower
7 on off on off off on
or even to a point lower than the required output of B1 and 8 on off on off on off
B2 . So, the output voltage of B1 drops to less than 6 V, and
the output voltage of B2 cannot reach its desired output value
which is 30 V. The instability is irreversible, and the system
Then, the discrete-time model that describes the dynamics
is stuck in the abnormal state.
of the system can be derived from equation (3). Suppose the
switching period is T . Denote x(nT ) = xn (0), or simply
B. Bifurcation Analysis xn for brevity, which is the initial value at the beginning of
In general, bifurcations can be classified into continuous and switching period n. Within a period T , the state-j equation
discontinuous bifurcations, depending on whether the states of given by (3) describes the system starting from time (nT +
the system are varying continuously or discontinuously. The P of Tj = τj −τj−1 , where
τj−1 ) to (nT +τj ) for a time interval
cause of this bifurcation is that there is a structural change in τ0 = 0 and τ8 = T , such that j Tj = T . Using equation
the system as the parameters of the system are varied through (3), the value of xn (τj ) by the end of this state-j is given by:
the critical point, and such a bifurcation may cause undesirable
or even catastrophic consequences as the state variables may xn (τj ) = Nj (Tj ) xn (τj−1 ) + (Nj (Tj ) − I) A−1
j Bj E, (4)
exhibit undesirably wide excursion in the state space causing
damage to some system components [24]. where I is an 9 × 9 identity matrix, and Nj (ξ) is the
As shown in Fig. 3, when B2 is connected to the DC bus, corresponding system matrices given by
the system becomes unstable. The input voltage of the load
∞
converters collapses, and excessive power is consumed by the X 1 k k
Nj (ξ) = eAj ξ = I + A ξ . (5)
wire resistor. The output voltage of the load converters fall k! j
k=1
below the desired values. Meanwhile, the duty cycles of the
load converters are fixed at the maximum values, which cause Equation (4) can be rewritten as xn (τj ) = fj (xn (τj−1 )).
high current stress on the switching devices. In this way, xn+1 = xn (T ) = xn (τ8 ) can be determined
When a load converter is plugged to the DC bus, there iteratively from xn (0) = xn using equation (4). Thus, in
is a large-signal disturbance on the dc bus line. In order to general, we have
investigate the bifurcation phenomenon of the system during
the transient, the discrete-time mapping model has to be xn+1 = f (xn ). (6)
established. In this system, all the subsystems are designed
for operation in continuous conduction mode (CCM). A state To complete the derivation, we have to find the relation
vector x containing three 3-dim column vectors xs , x1 and among the duty cycles of the subsystems and the state vari-
x2 for converters Bs , B1 and B2 , respectively, is chosen as ables xn . According to Fig. 2, switch Si will be turned off
follows: ∆
when si (xn din ) = (vpi − vrampi ) is zero as:
T
x = [xs x1 x2 ] , (1)
∆
where s (xn dn ) = vp − vramp = 0. (7)
T
xs = [iLs vCs vas ] , The equilibrium point XQ and the corresponding duty cycle
T DQ of the system can be found by determining the steady-state
x1 = [iL1 vC1 va1 ] , (2)
T solutions. Using the discrete-time model developed earlier, the
x2 = [iL2 vC2 va2 ] .
steady-state variables and the duty cycles can be found by
Without loss of generality, we can assume that the subsys- putting xn+1 = xn = XQ and dn = DQ . In the steady state,
tems share and synchronize a common period T , having eight the capacitor voltages equal the bus voltage and the output
operating states as described in Table II. The state-j equation, of the system, i.e., VCs = Vbus , VC1 = Vo1 and VC2 = Vo2 .
where j = 1, 2, 3, 4, 5, 6, 7, 8, within a period is given by Defining
ẋ = Aj x + Bj E state j. (3) T
XQ = [ILs Vbus VAs IL1 Vo1 VA1 IL2 Vo2 VA2 ] , (8)
Expressions of Aj and Bj can be readily found and are
omitted here. and solving for XQ and DQ , we get
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
1 1
Im(λ)
Im(λ)
0 0
p
Vbus − Vbus 2 − 4P r
Line
ILs = , (9c) −0.5 −0.5
2rLine
Vbusl = Vbus − ILs rLine , (9d) −1
−1 −0.5 0 0.5 1
−1
0.8 0.9 1
Re(λ) Re(λ)
Vbus
DQs = , (9e) (a) (b)
E
Vo1 1 1
IL1 = , (9f)
R1 0.5 0.5
rL1
Vo1 1 + R1
Im(λ)
Im(λ)
0
DQ1 = , (9g) 0
Vbusl
Vo2 −0.5 −0.5
IL2 = , (9h)
R2 (1 − DQ2 )
√ −1
−1 −0.5 0
Re(λ)
0.5 1
−1
0.8 0.9 1 1.1
1 − b + b2 − 4c Re(λ)
rLine
reference value. Meanwhile, B1 may enter the accommodation
region since the a voltage drop occurs in vbusl .
From the foregoing discussion and the previous full-circuit vbus vbusl Req1 Req2
simulation results, we can see that there is a complex interac-
tion among the subsystems through the non-ideal DC bus when
the DC distribution system is under large-signal disturbance Fig. 6. Equivalent circuit of the system under large-signal disturbance.
such as an abrupt load change, and such disturbance may cause
the voltage of the DC bus to collapse and thus would affect
the stability of the whole system. Meanwhile, the collapse of
the bus voltage may cause undesirable or even catastrophic
Vbusl1 = Vo1 (R1 + rL1 )/(Dm1 R1 ). (11)
consequences as the current magnitude on the DC bus have a
sudden increase which may damage the system’s components.
Vbusl2 = Vo2 [(1 − Dm2 )2 R2 + rL2 ]/((1 − Dm2 )R2 ). (12)
B. Stable Operation Requirements of the System Here, Dm1 and Dm2 represent the maximum duty cycle
values of B1 and B2 , respectively. Vbusl1 and Vbusl2 represent
If the system can maintain stability under large-signal the permissible minimum input voltage values of B1 and
disturbance, it can finally return to the desired operating point. B2 , respectively. For simplicity, only the equivalent series
For this case, B2 can start-up successfully and B1 can recover resistances of the inductors are taken into account, with other
to its normal operating state. This also means that all the parasitic parameters such as equivalent series resistances of the
load converters can maintain a constant power load (CPL) capacitors all set to zero. Detailed derivation of the equations
characteristic. Since all the load converters have a duty cycle (11) and (12) can be found in a prior work [25].
limit (Dm ), in order to ensure the load converters maintaining From the foregoing discussion, the steady-state value of the
the CPL characteristic, the duty cycle should always lie within load converters’ input voltage is adopted to inspect the large-
the range (0, Dm ). Thus, the input voltage of each load signal stability of the system. Thus, it is important to analyze
converter should be greater than a permissible minimum value the allocation of Vbus on the wire impedance and the input
to ensure that the system can return to the desired operating port of the load converters.
point. Otherwise, the duty cycle will saturate and stay at Dm , Under the critical condition, the whole system can be
and the load converters will operate in an open-loop condition simplified as shown in Fig. 6. In this model, Bs is equivalent
and behave as resistive loads rather than CPLs. to a voltage source, and Vbus is the output voltage of Bs .
Equivalent resistances Req1 and Req2 are the DC values of
C. Concept of the Criterion input resistances of the load converters which are operating at
their maximum duty cycles. Here, the line inductance is not
From the foregoing discussion, it can be concluded that the included in the simplified model. The reason is that the it only
input voltage of the load converters can be used as an indicator affects the dynamic of the system, and have no effect on the
to assess the large-signal stability of the system. Here, only calculation of the steady-state operating point of the system.
the steady-state values need to be considered and the detailed
According to equations (11) and (12), each load converter
dynamics such as oscillation can be ignored. This is because
has its permissible minimum input voltage because the output
the load converters may fail to maintain the CPL characteristic
voltages of the load converters are different. Since the load
and can be judged operating in an undesired region if their in-
converters are in parallel connection, an overall permissible
put voltages are below the permissible minimum value. Thus,
minimum input voltage Vbusl min should be greater than the
it is reasonable to derive a criterion using a set of algebraic
biggest one. Thus, Vbusl min is defined as
equations instead of differential equations to investigate the
system’s stability under large-signal disturbance. Vbusl min ≥ max {Vbusl1 , Vbusl2 } . (13)
It should be pointed that under an abrupt load change
situation, the duty cycle of a load converter which was stable According to Fig. 5 and equations (11), (12) and (13), the
originally, e.g., B1 , may not reach the maximum value during large-signal stability criterion can be given as:
the accommodation process. This is because the initial value
of the energy storage element is not zero if the converter is Req vbus
Vbusl min =
stable originally. However, for simplification, we will consider Req + rLine
the critical condition corresponding to all the load converters
Vo1 Dm1 Req1 Vo2 Req2
operating at the maximum duty cycle during the adjusting ≥ max , . (14)
R1 (1 − Dm2 )R2
transition.
Here Req = Req1 ||Req2 ; and Req1 = RD 1 +rL1
2 and Req2 =
m1
(1 − Dm2 )2 R2 + rL2 represent the DC values of the input
D. Derivation of the Criterion impedances of B1 and B2 , respectively. It is obviously that
When B1 and B2 operate in the critical condition, the the parameters used to calculate the boundary of the system
relation of the output voltages and the converter’s parameters are the load, equivalent series resistance of the inductor, output
in the steady state can be defined as voltage and the maximum duty cycle of each load converter,
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
Sln
Load
converter n Rn von
(Bn) Fig. 8. Equivalent circuit of the system under large-signal disturbance
(N load converters).
1
0.9
vbus: [20 V/div]
0.8
0.7
vbusl: [20 V/div]
0.6
rLine (Ω)
0.5 vo1: [5 V/div]
0.4
0.3
Turn on Sl2 vo2: [20 V/div]
0.2
0.1
0
0.5 0.6 0.7
D
0.8 0.9 1 Time: [4 ms/div]
m2
14.5 14.5
vbus(V)
14 14
13.5 13.5 vbusl: [20 V/div]
14 16
vo1: [5 V/div]
vbusl(V)
vbusl(V)
14
10
12
6 10
7 6.5 Turn on Sl2 vo2: [20 V/div]
vo1(V)
vo1(V)
6 6
5 5.5
30 30 Time: [2 ms/div]
vo2(V)
vo2(V)
15 15
0 0
0.15 0.2 0.25 0.3 0.35 0.4 0.15 0.2 0.25 0.3
t(s) t(s) (b)
(a) (b)
Fig. 11. Transient waveforms (a) for rLine = 0.6 Ω (without soft-
start), showing unstable system under large-signal disturbance; (b) for
Fig. 10. Transient waveforms (a) for rLine = 1 Ω (with soft-start),
rLine = 0.5 Ω (without soft-start), showing stable system under large-
showing unstable system under large-signal disturbance; (b) for rLine =
signal disturbance.
0.8 Ω (with soft-start), showing stable system under large-signal distur-
bance.
Unstable Unstable
vo1: [5 V/div] 2
rLine (Ω)
rLine (Ω)
1.5
1.5
Turn on Sl2 vo2: [20 V/div] Stable
Stable
1
1
Fig. 14. Operating boundaries for the system in (R2 , rLine ) parameter
vbus: [20 V/div]
space (a) without and (b) with soft-start.
vbusl: [20 V/div]
vo1: [5 V/div]
system can operate normally after B2 is connected to the DC
Turn on Sl2 vo2: [20 V/div] bus, verifying the effectiveness of the proposed criterion on
predicting the stability condition of the system with soft-start
Time: [20 ms/div] routine under large-signal disturbance.
According to the proposed criterion and Fig. 9, it can be
(b)
found that the maximum value of rLine will decrease if we
increase the soft-start duration to limit the equivalent duty
Fig. 12. Transient waveforms (a) for rLine = 1 Ω (with soft-start), cycle of B2 in the start-up process to below 0.79. Here,
showing unstable system under large-signal disturbance; (b) for rLine = we adjust the soft-start parameters to set the equivalent duty
0.8 Ω (with soft-start), showing stable system under large-signal distur-
bance. cycle of B2 to about 0.7. It can be predicted that the system
is unstable under large-signal disturbance if rLine = 0.7 Ω.
vbus: [20 V/div] Fig. 13 shows the experimental result. It can be observed that
B1 cannot recover to its original operating region while B2
vbusl: [20 V/div]
cannot reach its desired operating point. This agrees with the
vo1: [5 V/div] prediction based on the proposed criterion.
Turn on Sl2 vo2: [20 V/div]
C. Stability Boundary of the system
Time: [40 ms/div] Stable transient operating boundaries in the (R2 , rLine )
parameter space are compared with the theoretical analysis,
as shown in Fig. 12. The boundary of the theoretical analysis
Fig. 13. Transient waveforms for rLine = 0.7 Ω (with soft-start and the
equivalent maximum duty cycle is about 0.7), showing unstable system shown in Fig. 14(a) can be obtained by varying the value of
under large-signal disturbance. R2 and calculating the results according to inequality (14).
In Fig. 14(b), it can be obtained by varying the value of R2 ,
substituting the duty cycle value of B2 from 0 to the maximum
signal disturbance can be predicted by investigating the steady- value in inequality (14) and plotting the relationship between
state value of the input voltage of the load converters. Dm2 and rLine like the Fig. 9, and finding out the maximum
value ofrLine . In Fig. 14, the stable regions corresponding
B. Soft-Start Switching Subsystems Verification to the system can operate at the desired region under large-
signal disturbance, while the unstable regions correspond to
In the closed-loop controlled system with soft-start routine, the system working in an undesirable operating regions under
however, the duty cycle of B2 increases slowly after B2 large-signal disturbance. It can be found that the results of
is plugged to the DC bus. Fig. 12(a) shows the system the experiment and the full-circuit simulation are in good
which includes a soft-start routine with rLine = 1 Ω and agreement with the theoretical analysis.
LLine = 5 µH. Initially, Sl1 is on and Sl2 is off. After
converters B1 and Bs have reached their steady states, Sl2
is turned on. According to Fig. 8, when the system adopts V. C ONCLUSION
soft-start, and the equivalent duty cycle of B2 in the start-up A bifurcation phenomenon has been found for the DC
process is adjusted to be equal to 0.79, if rLine > 0.91 Ω, cascaded power system with multi-load converters. Essentially,
the system will lose stability under large-signal disturbance. the input voltage of the load converters collapses suddenly
From Fig. 12(a), it can be observed that vbus can be kept at the when an extra load converter connected to a non-ideal DC
regulated value after B2 is connected to the DC bus, but vo2 bus of the system. The root cause of this phenomenon is the
cannot reach its regulated value of 30 V and vo1 collapses and saturation of the duty cycles of the load converters and the
cannot recover to its normal operating region. Fig. 12(b) shows existence of the wire resistance of the DC bus. The voltage
the same system with rLine = 0.8 Ω. It can be found that the dropped across on the wire resistance may cause the input
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
voltage of the load converters to go below the permissible [9] C. M. Wildrick, F. C. Lee, B. H. Cho, and B. Choi, “A method of
minimum value and the duty cycles of the load converters defining the load impedance specification for a stable distributed power
system,” IEEE Trans. Power Electron., vol. 10, DOI 10.1109/63.387992,
will be fixed at the maximum value. In this situation, the load no. 3, pp. 280–285, May 1995.
converters fail to maintain the constant power load charac- [10] P. Huynh and B. H. Cho, “A new methodology for the stability analysis
teristic and behave as resistive loads. In addition, the output of large-scale power electronics systems,” IEEE Trans. Circ. Syst. I,
Fund. Theory Appl., vol. 45, DOI 10.1109/81.669060, no. 4, pp. 377–
voltages of the load converters drop due to the saturation of 385, Apr. 1998.
duty cycles. Thus, the system cannot operate at the desired [11] X. Feng, J. Liu, and F. C. Lee, “Impedance specifications for stable DC
operating point, signifying an instability has occurred under distributed power systems,” IEEE Trans. Power Electron., vol. 17, DOI
10.1109/63.988825, no. 2, pp. 157–162, Feb. 2002.
large-signal disturbance. A common practice of improving [12] S. Vesti, T. Suntio, J. Oliver, R. Prieto, and J. Cobos, “Impedance-
the system’s stability performance under this situation is to based stability and transient-performance assessment applying max-
introduce the soft-start control. Stability boundaries have been imum peak criteria,” IEEE Trans. Power Electron., vol. 28, DOI
10.1109/TPEL.2012.2220157, no. 5, pp. 2099–2104, May 2013.
identified for the system, clearly suggesting how the large- [13] C. Rivetta and G. A. Williamson, “Large-signal analysis of a dc-dc buck
signal stability of the system could be affected by the variation power converter operating with constant power load,” in Proc. IEEE Ind.
of some selected parameters. Electron. Soc. Meeting, vol. 1. DOI 10.1109/IECON.2003.1280073, pp.
732–737, Nov. 2003.
In this paper, we report an instability phenomenon in the [14] C. Rivetta and G. A. Williamson, “Global behaviour analysis of a dc-
DC distribution system and identify the cause of the instability. dc boost power converter operating with constant power load,” in Proc.
The physical origin of the unstable phenomenon is exposed IEEE Int. Symp. Circ. Syst., vol. 5, DOI 10.1109/ISCAS.2004.1329968,
pp. 956–959, May 2004.
through the bifurcation analysis. By investigating the adjusting [15] F. Zhang and Y. Yan, “Start-up process and step response of a dc–dc
transition and the stable operation requirements of the system converter loaded by constant power loads,” IEEE Trans. Ind. Electron.,
under large-signal disturbance, a simple and effective large- vol. 58, DOI 10.1109/TIE.2010.2045316, no. 1, pp. 298–304, Jan. 2011.
[16] F. Umbrı́a, J. Aracil, and F. Gordillo, “Singular perturbation stability
signal criterion is proposed. The concept of the proposed analysis of three phase two-level power converters,” in Proc. 18th
criterion is that the large-signal stability of the system can be Mediterr. Conf. Control Autom. (MED), Marrakesh, Morocco, DOI
judged by investigating the DC operating point of the load 10.1109/MED.2010.5547605, pp. 123–128, Jun. 2010.
[17] F. S. Garcia, J. A. Pomilio, G. S. Deaecto, and J. C. Geromel,
converters’ input voltage and the detailed dynamics of the “Analysis and control of dc-dc converters based on lyapunov stability
system can be omitted. The proposed criterion can be extended theory,” in Proc. IEEE Energy Convers. Congr. Expo. (ECCE), DOI
to apply in large-scale DC distribution systems. In addition, the 10.1109/ECCE.2009.5316085, pp. 2920–2927, Sep. 2009.
[18] D. Marx, P. Magne, B. Nahid-Mobarakeh, S. Pierfederici, and B. Davat,
effect of the soft-start control is also taken into consideration “Large signal stability analysis tools in dc power systems with constant
in the proposed criterion, and the analytical results show that power loads and variable power loads-a review,” IEEE Trans. Power
the proposed criterion can also predict the large-signal stability Electron., vol. 27, DOI 10.1109/TPEL.2011.2170202, no. 4, pp. 1773–
1787, Apr. 2012.
of the system with soft-start control and provide a sufficient [19] R. K. Brayton and J. K. Moser, “A theory of nonlinear networks-I,”
criterion to guarantee the system’s stability under large-signal Quart. Appl. Math., vol. 22, no. 1, pp. 1–33, 1964.
disturbance. The results are presented in design-oriented forms [20] L. Weiss, W. Mathis, and L. Trajkovic, “A generalization of Brayton-
Moser’s mixed potential function,” IEEE Trans. Circ. Syst. I, Fund.
to facilitate the identification of the parameter range that Theory Appl., vol. 45, DOI 10.1109/81.669065, no. 4, pp. 423–427,
ensures system’s stability under large-signal disturbance. Apr. 1998.
[21] M. Huang, H. Ji, J. Sun, L. Wei, and X. Zha, “Bifurcation-
R EFERENCES based stability analysis of photovoltaic-battery hybrid power sys-
tem,” IEEE J. Emerging Sel. Topics Power Electron., vol. 5, DOI
[1] S. Luo and I. Batarseh, “A review of distributed power systems part 10.1109/JESTPE.2017.2681125, no. 3, pp. 1055–1067, Sep. 2017.
I: DC distributed power system,” IEEE Aerosp. Electron. Syst. Mag., [22] W. Du, J. Zhang, Y. Zhang, and Z. Qian, “Stability criterion for cascaded
vol. 20, DOI 10.1109/MAES.2005.1499272, no. 8, pp. 5–16, Aug. 2005. system with constant power load,” IEEE Trans. Power Electron., vol. 28,
[2] S. Luo and I. Batarseh, “A review of distributed power systems Part II. DOI 10.1109/TPEL.2012.2211619, no. 4, pp. 1843–1851, Apr. 2012.
high frequency AC distributed power systems,” IEEE Aerosp. Electron. [23] X. Liu, Y. Zhou, W. Zhang, and S. Ma, “Stability criteria for constant
Syst. Mag., vol. 21, DOI 10.1109/MAES.2006.1662037, no. 6, pp. 5–14, power loads with multistage LC filters,” IEEE Trans. Veh. Technol.,
Jun. 2006. vol. 60, DOI 10.1109/TVT.2011.2148133, no. 5, pp. 2042–2049, Jun.
[3] R. Haroun, A. Cid-Pastor, A. El Aroudi, and L. Martinez-Salamero, 2011.
“Synthesis of canonical elements for power processing in dc distribution [24] C. K. Tse, Complex behavior of switching power converters. Boca Raton,
systems using cascaded converters and sliding-mode control,” IEEE FL, USA: CRC press, 2003.
Trans. Power Electron., vol. 29, DOI 10.1109/TPEL.2013.2261093, [25] Y. Chen, C. K. Tse, S. S. Qiu, L. Lindenmüller, and W. Schwarz,
no. 3, pp. 1366–1381, Mar. 2014. “Coexisting fast-scale and slow-scale instability in current-mode con-
[4] A. Riccobono and E. Santi, “Comprehensive review of stability criteria trolled DC/DC converters: Analysis, simulation and experimental
for dc power distribution systems,” IEEE Trans. Ind. Appl., vol. 50, DOI results.” IEEE Trans. Circ. Syst. I, Reg. Papers, vol. 55, DOI
10.1109/TIA.2014.2309800, no. 5, pp. 3525–3535, May 2014. 10.1109/TCSI.2008.923282, no. 10, pp. 3335–3348, Nov. 2008.
[5] S. Liu, P. X. Liu, and X. Wang, “Stability analysis of grid-interfacing [26] P. Griffith, “Designing switching voltage regulators with the TL494,”
inverter control in distribution systems with multiple photovoltaic- Texas Instruments Application Report. Available: http://focus. ti.
based distributed generators,” IEEE Trans. Ind. Electron., vol. 63, DOI com/lit/an/slva001d/slva001d. pdf, 2005.
10.1109/TIE.2016.2592864, no. 12, pp. 7339–7348, May 2016.
[6] S. Singh, A. R. Gautam, and D. Fulwani, “Constant power loads and
their effects in dc distributed power systems: A review,” Renew. Sustain.
Energy Rev., vol. 72, DOI 10.1016/j.rser.2017.01.027, pp. 407–421, Jan.
2017.
[7] R. D. Middlebrook, “Input filter considerations in design and application
of switching regulators,” in Proc. IEEE IEEE IAS Annual Meeting, 1976,
pp. 366-382
[8] S. Schulz, B. Cho, and F. Lee, “Design considerations for a distributed
power system,” in Proc. IEEE Power Electron. Spec. Conf., San Antonio,
TX, DOI 10.1109/PESC.1990.131244, pp. 611–617, Jun. 1990.
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
Li Ding (Student Member) received the B.E. de- Chi K. Tse (M’90–SM’97–F’06) received the
gree in information engineering from the Guang- BEng (Hons) degree with first class honors in
dong University of Technology, Guangzhou, electrical engineering and the PhD degree from
China, in 2011, the M.E. degree in signal and the University of Melbourne, Australia, in 1987
information processing from the South China and 1991, respectively.
University of Technology, Guangzhou, in 2014. He is presently Chair Professor of Electrical
He is currently pursuing the Ph.D. degree in Engineering at City University of Hong Kong,
power electronics with the Hong Kong Polytech- Hong Kong. Prior to joining City University of
nic University, Hong Kong. His current research Hong Kong in October 2019, he was with
interests include modeling and analysis power Hong Kong Polytechnic University, with which he
electronic systems and study the complex be- served as Head of the Department of Electronic
havior in power electronic circuits. and Information Engineering from 2005 to 2012. His research interests
include power electronics, nonlinear systems and complex network ap-
plications. He was awarded a number of research and industry awards,
including Prize Paper Awards by IEEE T RANSACTIONS O N P OWER E LEC -
TRONICS in 2001, 2015 and 2017, RISP Journal of Signal Processing
Best Paper Award in 2014, Best paper Award by International Journal
of Circuit Theory and Applications in 2003, two Gold Medals at the
International Inventions Exhibition in Geneva in 2009 and 2013, a Silver
Medal at the International Invention Innovation Competition in Canada
in 2016, a Grand Prize and Gold Medal at the Silicon Valley Interna-
tional Invention Festival in 2019, and a number of recognitions by the
academic and research communities, including honorary professorship
by several Chinese and Australian universities, Chang Jiang Scholar
Chair Professorship, IEEE Distinguished Lectureship, Distinguished Re-
search Fellowship by the University of Calgary, Gledden Fellowship and
International Distinguished Professorship-at-Large by the University of
Western Australia. While with the Hong Kong Polytechnic University, he
received the President’s Award for Outstanding Research Performance
twice, Faculty Research Grant Achievement Award twice, Faculty Best
Researcher Award, and several teaching awards.
Dr. Tse serves and has served as Editor-in-Chief for the IEEE T RANS -
ACTIONS ON C IRCUITS AND S YSTEMS II (2016-2019), IEEE Circuits and
Systems Magazine (2012-2015), Editor-in-Chief of IEEE Circuits and
Systems Society Newsletter (since 2007), Associate Editor for three
IEEE Journal/Transactions, Editor for International Journal of Circuit
Theory and Applications, and is on the editorial boards of a few
other journals. He currently chairs the steering committee for the IEEE
Transactions on Network Science and Engineering. He also serves as
panel member of Hong Kong Research Grants Council, and member of
several professional and government committees.