IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 4, NO.

1, MARCH 2016 93

Stability Analysis and Dynamic Performance

Evaluation of a Power Electronics-Based DC
Distribution System With Active Stabilizer
Mehdi Karbalaye Zadeh, Roghayeh Gavagsaz-Ghoachani, Serge Pierfederici,
Babak Nahid-Mobarakeh, Senior Member, IEEE, and Marta Molinas, Member, IEEE

Abstract— The stability of dc microgrids is influenced by energy sources, e.g., photovoltaic and fuel cell, or energy stor-
the nonlinear behavior of the converter-controlled loads with age devices, e.g., battery and supercapacitor, as well as modern
constant power characteristics. This nonlinear dynamic, in line power electronic loads, by eliminating the redundant energy
with reduced size input filters used in the embedded power
distribution of transportation systems, has a degrading effect conversion [3], [4]. Moreover, the development of dc micro-
on the dynamic performance and stability of the system. This grids can overcome the practical issues of ac networks,
paper presents the complete stability analysis of a dc distribution such as frequency synchronization, reactive power flow, inrush
system composed of power electronics-based source and load. current of transformers, and current harmonics [5], [6].
For this objective, a discrete-time dynamic model is developed Therefore, such small-scale dc power systems, namely, dc
and is applied to the studied system. Here, for the first time,
the dynamic effect of the load controller is taken into the nanogrid, are introduced as key technologies for the smart grid
system model. In the studied system, the load converter is development toward embracing more distributed renewable
connected in cascade with the rest of the system, and is equipped sources, and providing higher efficiency power distribution to
with an input LC filter, representing a common distributed future building electric systems [7]–[12].
architecture for the transportation applications. The controller In addition to the smart grid applications, dc microgrid is
of the source converter employs an active stabilizer to extend
the stability margin of the system. This stabilizer uses high- an optimized solution for the next generation transportation
pass-filtered voltage of the dc bus following with a proportional systems, such as electric ships, more electric aircraft (MEA),
compensator. The dynamic behavior of the controller and the and advanced automotive systems [13]–[17]. In this direc-
stabilizer is experimented through a series of laboratory tests. tion, mechanical actuators in aircrafts are replaced by
Different load dynamics are implemented to demonstrate the converter-controlled electric drives, which are supplied by a
impact of slow and fast load dynamics on the stability of the
system. dc microgrid [18]. However, the stability of dc power systems
is a major concern, due to the nonlinear characteristics of the
Index Terms— DC–DC power converters, power electronics, power electronics systems, which can lead to unstable oscilla-
power system dynamics, stability analysis, transportation.
tions propagating in the dc system [19]. These oscillations may
be initiated either by the interactions between the controllers
I. I NTRODUCTION and the nonlinear switching effect of the converters [20] or
by the constant power dynamic of tightly regulated active
A CTIVE dc power systems are embedded in hybrid
ac/dc networks, in the form of dc microgrids, providing
the efficient and reliable energy distribution [1], [2]. The
loads [21]. Such loads are usually controlled by pulsewidth
modulated (PWM) inverters or dc–dc converters connected
in cascade to the rest of the system. The control bandwidth
dc microgrids facilitate integrating the emerging dc renewable
of such load converters is sufficiently high to ensure that
Manuscript received May 17, 2015; revised August 19, 2015; accepted the regulated load power is not influenced by fluctuations in
September 17, 2015. Date of publication September 29, 2015; date of current dc bus voltage [17]. In stability studies, such power electronic-
version January 29, 2016. Recommended for publication by Associate Editor based loads have been usually simplified as constant power
Johan H. R. Enslin.
M. Karbalaye Zadeh is with the Department of Electric Power Engineering, load (CPL), introducing negative impedance characteristics at
Norwegian University of Science and Technology, Trondheim 7491, the load terminals [21]–[30]. Though this model is valid for the
Norway (e-mail: mehdi.zadeh@ntnu.no). overall stability of the system, dynamic of the load converter
R. Gavagsaz-Ghoachani, S. Pierfederici, and B. Nahid-Mobarakeh are with
the Groupe de Recherche en Électrotechnique et Électronique de Nancy, Ecole is still not considered.
Nationale Supérieure d’Electricité et de Mécanique, Université de Lorraine, The schematic of a dc microgrid is shown in Fig. 1. In such
Nancy 54000, France (e-mail: roghayeh.gavagsazghoachani@univ-lorraine.fr; distribution systems, LC filters are added to the point of load
serge.pierfederici@univ-lorraine.fr; babak.nahidmobarakeh@univ-lorraine.fr).
M. Molinas is with the Department of Engineering Cybernetics, Norwegian converters, not only to reject the load oscillations, but also to
University of Science and Technology, Trondheim 7491, Norway (e-mail: meet power quality standards considering the harmonic effects
marta.molinas@elkraft.ntnu.no). and electromagnetic interference [31]. The stability of CPL,
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org. in a large extent, depends on the input filter. Therefore, to
Digital Object Identifier 10.1109/JESTPE.2015.2484218 meet the stability requirements, designers usually use big filter
2168-6777 © 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
94 IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 4, NO. 1, MARCH 2016

to study the effect of the load controller on the dynamic

performance and the stability of the system. Therefore, using
the proposed model, a more accurate pattern of stability can
be achieved. On the other hand, the dynamic behavior of
the system is analyzed with the bifurcation diagrams. Using
the complete model of the system, the qualitative studies
with the bifurcation analysis can demonstrate how the char-
acteristics of the system change as a function of a control
parameter.
The proposed model is applied to a simple cascade archi-
Fig. 1. Schematic of a dc distribution system for transportation applications. tecture, comprised of a source converter regulating the dc bus
voltage, a load converter, and an input filter. It is regarded
a generic system configuration for the dc distribution sys-
capacitors, when inductance cannot be reduced as it is defined tems, particularly for transportation applications. The studied
by the length of the wire. system includes the main elements of distributed system
However, the increased capacitance is not preferred, architecture, and can be generalized to any such dc systems,
particularly for MEA, where size and weight of the system which can represent many applications, e.g., local dc power
components should be minimized. To control the degrading systems in MEA or battery-fed dc–dc converters in electric
effect of CPLs, several techniques have been proposed, such vehicles.
as passive damping, virtual capacitor and resistor, and active One main objective of this paper is to identify the unstable
stabilization [32]–[36]. The damping of converter-fed electric operating points and the root cause of instabilities under vari-
drive has been increased by adding resistance to the input filter, ation of control parameters. For this objective, the state-space
but this solution is not applicable for transportation systems model of the complete system is established using an exact
because of the power losses on the filter resistance [32], [33]. discrete-time method developed for PWM converters [38].
Linear and nonlinear active stabilization methods are reported After developing the system model, the Jacobian matrix of the
in [34]–[36]. system is used to calculate the discrete-time eigenvalues [39],
In [34], the system damping is increased by adding a and system bifurcations are also studied using a discrete
virtual capacitance to the controller. In this method, the control Poincare formulation [40]. The dynamic behavior of the
structure is modified to establish a nonlinear control stabilizer system can then be studied using bifurcation diagrams, and
system. This, consequently, increases the complexity of the the asymptotic stability of the system is investigated using the
control system. A nonlinear centralized active stabilizer, for eigenvalues.
global stability of a dc microgrid, was proposed in [35]. The rest of this paper is organized as follows. The studied
However, the centralized approach needs a large amount of system, the controller, and the stabilizer are described
data processing [37]. A distributed load-side stabilizer is in Section II. Section III analyzes the dynamic performance
presented in [36] as a multiagent approach. In that work, an and the stability of the system, where stable and unstable
individual linear compensator is used for each load. operating points are predicted. In Section IV, the laboratory
In [22], we proposed a source-side stabilizer to increase the experiments are used to study the performance of the stabilizer
maximum stable power supplied by a dc–dc buck converter to and the dynamic effect of slow and fast loads. Finally, we
a CPL. This stabilizer is implemented with a simple structure conclude this paper in Section V.
and uses high-pass-filtered voltage of the dc bus. Here, unlike
the previous research, the stabilizer is implemented on the
II. S YSTEM R EPRESENTATION AND C ONTROLLER
source converter. This is advantageous over the load-side
techniques, since it is not affected by the load controllers. An electrical diagram of the studied system is shown
Moreover, this method does not change or modify the control in Fig. 2. The source converter is a buck dc–dc converter,
structure, and hence, it can be added as a stabilizing block which supplies the dc bus with a regulated voltage Vs . The
to different types of controllers (i.e., linear or nonlinear). load converter is a boost dc–dc converter cascade connected
The application of the proposed method is also generalized to the dc grid through an input LC filter. In this scheme, the
to a multiload dc power system in [19]. In that work, the filter represents either input filter of the load converter or stray
system model is established based on the proposed discrete- inductance of the cable.
time method, and the asymptotic stability of the dc distribution The parameters r f , C f , and L f are the resistance, capac-
system is then analyzed using the system eigenvalues. The itance, and inductance of the filter, respectively; C is the
impact of the filter sizing on the critical power is also studied, dc bus capacitance; L 1 and r1 are the series inductance and
and the stabilizer is used to compensate the degrading effect resistance of the source converter; and L 2 and r2 indicate the
of the reduced size filter. However, the dynamics of the load inductance and resistance of the load converter. A dc source
controllers are still not taken into the system analysis, since provides dc voltage Ve . Here, i L 1 , i L 2 , and i dc are inductor
the load model is reduced to an ideal CPL. current of the source, load current, and dc current, respectively.
In this paper, for the first time, the load converter is Vs , Vcf , and Vo are dc bus voltage, filter voltage, and the output
taken into the discrete-time model of the system, in order voltage of load converter, respectively. The system model is
KARBALAYE ZADEH et al.: STABILITY ANALYSIS AND DYNAMIC PERFORMANCE EVALUATION 95

Fig. 2. Electric diagram of the studied dc power system.

Fig. 3. Block diagram of the control system.

then described by the differential equations of the physical f s : switching frequency), and are used to generate the control
system as follows: commands.
⎧ The measured and sampled values of the system currents
⎪
⎪
di L 1
=
1
(u s · Ve − Vs − r1 · i L 1 ) and voltages are given in the form of i L 1 _mes , i L 2 _mes , i dc_mes ,
⎨
dt L1 Vs_mes , Vcf_mes , and Vo_mes with mes referred to measured
(1)
⎪
⎪ d Vs 1 values. The load controller regulates its output power P.
⎩ = (i L 1 − i dc )
Hence, the power control block translates the reference load
⎧ dt C
⎪
⎪
di dc 1 power Pref to a reference current [i 2,ref = (Pref /Vcf_mes )]
⎨ dt = L (Vs − Vcf − r f · i dc ) used by the load’s current controller. Reference current of the
f
(2)
⎪
⎪ d V cf 1 source converter is generated by the voltage controller. The
⎩ = (i dc − i L 2 ) voltage controller is functioned to regulate the dc bus voltage
dt Cf
⎧ by a PI compensator as follows:
⎪ di L 2 1
⎪
⎨ dt = L (Vcf − r2 · i L 2 − (1 − u l )Vo )
2 K iv
  (3) Cv (s) = K pv + . (4)
⎪
⎪ d V 1 Vo s
⎩ o = (1 − u l )i L 2 − .
dt Co R Both source and load current controllers are designed based
on a switching surface depending on the current error. It forces
the inductor current, by driving the system trajectories to a
A. Voltage and Current Controller
desired surface. The block diagram of the source controller and
The block diagram of the control system is shown in Fig. 3. the stabilizer are shown in Fig. 4, and the current controller
The source controller is comprised of a voltage control loop, of the load converter is presented in Fig. 5: K i1 and λ1 are
a current control loop, and the stabilizer. The voltage controller control coefficients of the source; K i2 and λ2 are the control
uses a linear proportional–integral (PI) compensator. The coefficients related to the load controller. The calculation of the
current controller is established based on the indirect sliding- control coefficients is detailed in [43]. These coefficients allow
mode approach, and the control command is derived from tuning the dynamic response of the controller. K i1 and K i2
equivalent control law [41], [42]. The converter controllers define the bandwidth of the source and the load controller,
are implemented in digital form to be applied to a dSPACE respectively. By applying the desired switching surface, the
digital control card in the laboratory. duty cycle is derived and is used to generate the switching
Then, the control variables are sampled by a sam- command. Here, Ds and Dl are the duty cycle of the source
pling rate equal to the switching frequency (T = (1/ f s ), and the load converter, respectively.
96 IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 4, NO. 1, MARCH 2016

Fig. 4. Block diagram of the source controller and the stabilizer.

TABLE I
VALUES OF THE S YSTEM C OMPONENTS AND C ONTROL PARAMETERS

Fig. 5. Current controller of the load converter.

B. Stabilization System
The stabilizer is implemented based on a high-pass filter
followed by a proportional compensator. This filter enables
preserving the system dynamics while rejecting the voltage
oscillations. Then, an additional duty cycle is produced by the
stabilizer, which is used to damp the system oscillations. Here,
the cutoff frequency of the filter is ωsf , which corresponds to
the voltage oscillations. K stab is stabilizer’s gain, which will be
tuned with respect to the dynamic and transient performance Different case studies are carried out to study the effect
of the system. Finally, the stabilizing duty cycle Dstab is of different control parameters on the system stability. The
added to the main duty cycle of the current controller D0 values of the filter parameters as well as the parameters
(see Fig. 4). The switching command (u s ) is then generated of the outer loop control, such as K pv and K iv , are given
through a PWM. in Table I.

III. S TABILITY A NALYSIS A. Bifurcation Analysis

The stability of the dc power system, as shown in Fig. 2, A bifurcation corresponds to the loss of system stability in a
is analyzed in this section. For this objective, a discrete-time specific operating point, which is called bifurcation point [40].
state-space model of the complete system is developed based Beyond this point, the system variables bifurcate, as imposed
on the discrete-time method that we first proposed in [22]. by the system dynamics [40]. Here, the bifurcation diagrams
Here, both the source and the load dynamics are taken into are derived from the sufficient data sets of the state variables
the system model. Consequently, the Jacobian matrix of the dc resulting from the analytical model of the system. This is
power system is established, and the system’s eigenvalues are repeated for the different values of the bifurcation parameter.
calculated: the eigenvalues are used to identify the stable and The dynamic analysis is performed in two cases: 1) with the
unstable operating points. The stability pattern of the system change of the load power Pref and 2) with changing the control
can then be derived based on the stable and unstable operating bandwidth.
points. Moreover, the dynamic behavior of the system is stud- 1) Effect of the Load Power: As the first step, the reference
ied using the bifurcation analysis [44]. Bifurcation diagrams power of the load converter Pref is changed, and the system
are used for qualitative studies of changes in the system’s dynamic is studied with the bifurcation maps. In this paper,
characteristics as a function of a bifurcation parameter, which the parameters of the source converter’s current controller
may be a passive element or a control parameter [45], [46]. are set to K i1 = λ1 = 1500: the control parameters of the
Here, the control parameters, including the load source converter are K i2 = λ2 = 4000, and the stabilizer
power (Pref ) and the control bandwidth of the source and is initially deactivated (K stab = 0). The bifurcation diagrams
load converters (K i1 and K i2 ), are the bifurcation parameters. for the inductor currents of the source converter (i L 1 ) and
The dynamic properties of the dc power system can then the load converter (i L 2 ) are shown in Fig. 6(a) and (b),
be investigated as the function of the control bandwidth. respectively. When the reference power exceeds a certain
KARBALAYE ZADEH et al.: STABILITY ANALYSIS AND DYNAMIC PERFORMANCE EVALUATION 97

Fig. 6. Bifurcation diagram of the currents with the variation of Pref , Fig. 7. Bifurcation diagram of the currents with the variation of Pref ,
without stabilizer (K stab = 0). (a) Inductor current of the source converter i L 1 . with stabilizer (K stab = 4). (a) Inductor current of the source converter i L 1 .
(b) Inductor current of the load converter i L 2 . (b) Inductor current of the load converter i L 2 .

value, namely, critical power P0 , bifurcation occurs and the

system loses the stability. In this case, the critical power is
observed ∼600 W. Beyond this point (Pref > P0 ), limit cycle
oscillations appear in the bifurcation map. This phenomenon
corresponds to Hopf or Neimark bifurcation [44]. Then, the
stabilizing signal is enabled with K stab = 4, and the system
model is established with the same control parameters as the
previous study. Again, the system variables result from the
system model, and bifurcation maps of the inductor currents
are shown in Fig. 7. By enabling the stabilizer, the bifurcation
point moves forward toward the reference power, and the Fig. 8. Bifurcation diagram of the inductor current i L 1 with the change
new critical power appears ∼1100 W. This confirms that the of K i1 , Pref = 700 W, K i2 = 4000, and K stab = 0.
maximum stable power delivered by the dc system is increased
by enabling the stabilizer.
2) Effect of the Control Bandwidth: As the next step of
analysis, the effect of the control parameters of the source
and load converters on the dynamic behavior of the system
is investigated using the bifurcation maps. In this case, the
reference power is fixed at Pref = 700 W, and the stabilizer
is deactivated (K stab = 0). First, the control parameter of the
source controller K i1 is changed, which defines the bandwidth
of the current controller. The controller of the source converter
is designed such that the bandwidth of the voltage controller
is sufficiently lower than the current controller. Therefore, the
dynamics of the system are defined by the current controller. Fig. 9. Bifurcation diagram of the inductor current i L 1 with the change
of K i2 , Pref = 700 W, K i1 = 1500, and K stab = 0.
In this case, the bifurcation map of the source current is shown
in Fig. 8. It is observed that the stability margin of the system
can be increased by increasing the control bandwidth of the Second, the bandwidth of the load controller K i2 is changed
source converter. However, the increased bandwidth is not as the bifurcation parameter. The effect of increasing the load
recommended because of its negative effect on the transient converter’s control bandwidth is investigated with the bifur-
behavior of the system. cation map, as shown in Fig. 9. As seen from the map, the
98 IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 4, NO. 1, MARCH 2016

Fig. 10. Stability pattern of the system with the change of K i2 and Pref : Fig. 13. Stability pattern of the system with the change of Pref and K stab :
K i1 = 2000 and K stab = 0. Black circles: stable operating points. K i1 = 1500 and K i2 = 3500. Black circles: stable operating points.
Red crosses: unstable operating points. Red crosses: unstable operating points.

Fig. 14. Experimental setup.

Fig. 11. Stability pattern of the system with the change of Pref and K i2 :
K i1 = 1500 and K stab = 0. Black circles: stable operating points.
Red crosses: unstable operating points. Jacobian matrix are calculated; then, stable and unstable
operating points are identified, and are used to draw the
stability pattern of the system [39]. In this paper, the system
analysis is repeated with changing the reference power of
the load converter (Pref ). The influence of the load controller
is also studied with the change of the control bandwidth.
Three different scenarios are investigated corresponding to
three different sets of control parameters.
In the first scenario, the bandwidth of the source controller
is fixed at K i1 = λ1 = 2000 and the stabilizer gain K stab = 0.
In this case, the bandwidth of the source is relatively high,
namely, fast controller. The dynamic of the load controller then
changes from K i2 = 3400 to K i2 = 5400, and the stability
pattern results, as shown in Fig. 10. The stable and unstable
operating points are shown with black circles and red crosses,
Fig. 12. Stability pattern of the system with the change of Pref and K i2 : respectively. The stability pattern has two variables, such as
K i1 = 1500, with the stabilizer (K stab = 4). Black circles: stable operating
points. Red crosses: unstable operating points. The maximum stable power is reference power Pref and K i2 . This is observed that lower
increased by ∼80%, compared with the case without stabilizer. load power can be controlled in the system with faster load
controller. However, the change of the stability margin for
faster load controller has the opposite effect from the source different control bandwidths is not uniform, since the system
controller. It means that the stability margin of the system is model is nonlinear.
reduced by increasing the bandwidth of the load controller. In the second scenario, the bandwidth of the source con-
In fact, the very fast load controller behaves like CPL [22]. troller is reduced to K i1 = λ1 = 1500 (and K stab = 0).
In this case, the controller’s dynamic is similar to the study
B. Eigenvalue Analysis of Section III-A. Again, the stability pattern of the system
The asymptotic stability of the system is analytically results from the eigenvalues, with the change of Pref and K i2 ,
investigated through the identification of instabilities in the as shown in Fig. 11. This is observed that the stability margin
steady state of the system: eigenvalues of the system’s of the system is reduced for all the values of K i2 , compared
KARBALAYE ZADEH et al.: STABILITY ANALYSIS AND DYNAMIC PERFORMANCE EVALUATION 99

Fig. 16. Experimentally generated phase portrait of system variables

i L 1 and vs with the change of Pref , without stabilizer (K stab = 0). (a) Stable
Fig. 15. Experimental waveforms of voltage and currents with behavior with the fast load dynamic. (b) Instability with slow load dynamic.
Pref = 500 W → 1000 and without stabilizer (K stab = 0). (a) Fast load
dynamic. (b) Slow load dynamic. The system loses stability with the slow
load ramp, while the fast load ramp prevents the system from the unstable
oscillations.

with the pattern of Fig. 10. In fact, the system dynamic

is associated with the ratio of the bandwidth of the source
controller with those of the load controller (K i2 /K i1 ). This
means that when the load bandwidth is sufficiently higher than
the source controller, lower power can be delivered to the load
through the dc power system. However, the control bandwidths
of the load converters are practically much higher than the
source, in order to make the regulated load power independent
of the dc bus oscillations. Therefore, this is important to
consider the fast load controllers.
Experimental step response of the source current. Results
are compared in the case of without stabilizer( ) and with
stabilizer with different values of . (a) Dynamic oscillations.
(b) Transientresponse.
In the third scenario, the stabilization signal is applied to Fig. 17. Experimentally generated phase portrait of system variables
the controller, in order to increase the stability margin of i L 1 and vs with the change of Pref , with stabilizer (K stab = 4). (a) Fast
load dynamic. (b) Slow load dynamic.
the system. Then, the second control scenario is repeated
with K stab = 4 (K i1 = λ1 = 1500). Again, the stability IV. E XPERIMENTAL R ESULTS
pattern of the system is constructed, and is shown in Fig. 12.
In this case, the stable load power is increased for all the The dynamic behavior of the studied system is tested on a
values of K i2 . This was predicted by the bifurcation diagram laboratory setup, including the source and the load converters,
for a single control parameter. as shown in Fig. 14. The source converter is fed by a program-
Finally, the effect of the stabilizer gain on the critical power mable dc source supplying the dc voltage Ve . The maximum
is investigated using the stability pattern of the system, as current delivered by the source is limited to 10 A. The con-
shown in Fig. 13. In this paper, the reference power Pref and verters are built with Semikron trench insulated-gate bipolar
the stabilizer gain K stab change, while the control bandwidth transistor, and the converter controllers are implemented
is constant: K i1 = 1500 and K i2 = 3500. According to the with a dSPACE real-time processor (DS1103). The switching
system analysis, the critical power is increased by increasing frequency is 10 kHz, and the sampling rate of the controller
the stabilizer gain. The qualitative behavior of the stabilizer is is equal to the switching frequency. In the measurement unit,
studied using the laboratory tests in Section IV. the wide frequency Tektronix current probes are used with
100 IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 4, NO. 1, MARCH 2016

Fig. 18. Experimental step response of the source current. Results are compared in the case of without stabilizer (K stab = 0) and with stabilizer with different
values of K stab . (a) Dynamic oscillations. (b) Transient response.

the bandwidth up to 100 MHz. Then, the PWM command data are used to draw the 3-D phase portrait for the case of
is generated by dSPACE. In the experimental circuit, a buck fast load change, corresponding to Fig. 15(a). In this case,
dc converter is used as the source converter, and a boost dc the system trajectory moves toward a spiral pattern, and then
converter acts as the load converter. An LC filter is added to settles down in the vicinity of the stable operating point.
the point of the load converter in order to reduce the harmonic On the contrary, Fig. 16(b) shows the experimental phase
effects. The load converter (boost converter) regulates its portrait for the slow load change, corresponding to the exper-
output power with the capability of dynamic changes, to iment of Fig. 15(b). Here, the state of the system moves from
emulate the constant power dynamic of the load. Stepwise and a stable equilibrium point to unstable limit cycle oscillations,
ramped changes are implemented in the reference power of the and cannot move back to the stable point. In other words, the
load, allowing slow and fast load changes. The parameters of system trajectory has a torus limit cycle attractor, meaning
the system components used for the experimental tests are the that the trajectory is attracted to periodic orbits on a torus
same as those used for the simulations. surface [44]. The periodic orbits in Fig. 16(b) correspond to
A series of experimental tests is performed to study the Hopf or Neimark bifurcation, as predicted by the bifurcation
system dynamics: the impact of fast and slow load change map in the section III (see Fig. 6).
on the dynamic behavior of the system is studied, the experi- Finally, the stabilizer is activated with K stab = 4, and the
mental analysis is repeated with the stabilization system, and experimental tests are repeated with Pref = 500 W → 1000 W,
the influence of the stabilizer’s gain K stab on the dynamic and as shown in Fig. 17: the system trajectory for the fast load
transient behavior of the system is investigated. change is presented in Fig. 17(a), and the result of the slow
In the first test, the stabilizer is switched OFF, and the load change is shown in Fig. 17(b). Clearly, the system
reference power of the load converter is formed to simulate trajectory undergoes transients and moves back to a stable
the desired load dynamic. The voltage of the dc bus is fixed equilibrium, for the two cases. The experimental results are in
at Vsref = 150 V. The reference power is increased abruptly agreement with the bifurcation analysis.
from a stable operating point Pref = 500 W to an unstable As the next laboratory test, we investigate the performance
operating point Pref = 1000 W, and is reduced back to the of the stabilizer and the influence of changing the gain K stab on
initial value with a ramp. The slope of the decreasing ramp system transients. For this objective, the load power is changed
defines how the dynamic of the load power is fast. stepwise from Pref = 500 W → 1000 W, and the performance
First, a fast ramp is applied to the load power, and the sys- of the controller is compared with and without stabilizer. The
tem response is shown in Fig. 15(a). The green line (P) shows experimental data are used to plot the waveform of the inductor
the reference power. i dc , i L 2 , and vs are dc current, inductor current (i L 1 ), as shown in Fig. 18.
current of the load, and dc voltage oscillation, respectively. The First, the stabilizer gain (K stab ) is set to zero, and the
dc part of voltage Vs is removed, and only the voltage oscilla- system gets unstable oscillations, which is shown by the
tion (vs ) is kept, for better visibility. As shown in Fig. 15(a), blue dashed line. Then, the stabilizer is activated, and
the fast load change prevents the system characteristics from the experiment is repeated with different stabilizer gains:
the limit cycle oscillations. Therefore, the system stays stable, K stab = 2.25, 4, and 5.5. According to the experimental
and the power controller follows the reference without failure. results, the system can tolerate the load change, and the current
Second, the same step (Pref = 500 W → 1000 W) is applied oscillations are damped.
with a slow reducing ramp, and the system response is shown However, the system damping and the frequency of
in Fig. 15(b): obviously, the system loses the stability. In fact, oscillations depend on K stab . Therefore, adjusting K stab
a large disturbance can destabilize the system variables if it allows achieving sufficient overshoot and settling time.
remains for the sufficient time duration. Indeed, a higher K stab provides a relatively higher stability
For the given scenario, the experimental phase portraits are margin. However, the increased gain increases the over-
also used to study the trajectory of state variables i L 1 and vs as shoot, which should be limited to meet the quality standards.
a function of the load power P. In Fig. 16(a), the experimental This also increases the transient energy and power.
KARBALAYE ZADEH et al.: STABILITY ANALYSIS AND DYNAMIC PERFORMANCE EVALUATION 101

Consequently, an optimal gain should be adjusted based on [9] Y. Tian, Z. Chen, F. Deng, X. Sun, and Y. Hu, “Active power and
the load properties. dc voltage coordinative control for cascaded dc–ac converter with
bidirectional power application,” IEEE Trans. Power Electron., vol. 30,
no. 10, pp. 5911–5925, Oct. 2015.
V. C ONCLUSION [10] X. Lu, J. M. Guerrero, K. Sun, and J. C. Vasquez, “An improved droop
control method for dc microgrids based on low bandwidth communi-
This paper has presented the stability analysis of a dc power cation with dc bus voltage restoration and enhanced current sharing
system for transportation applications: the influence of the accuracy,” IEEE Trans. Power Electron., vol. 29, no. 4, pp. 1800–1812,
Apr. 2014.
load characteristics, control parameters, and the stabilization
[11] T. Dragicevic, J. M. Guerrero, J. C. Vasquez, and D. Skrlec, “Supervi-
system on the dynamic behavior of the dc system has been sory control of an adaptive-droop regulated dc microgrid with battery
studied using the theoretical and experimental analyses. In the management capability,” IEEE Trans. Power Electron., vol. 29, no. 2,
theoretical analysis, the dynamic behavior of the system has pp. 695–706, Feb. 2014.
[12] A. P. N. Tahim, D. J. Pagano, E. Lenz, and V. Stramosk, “Mod-
been investigated using bifurcation diagrams as well as an eling and stability analysis of islanded dc microgrids under droop
analytical stability pattern of the system eigenvalues. Here, control,” IEEE Trans. Power Electron., vol. 30, no. 8, pp. 4597–4607,
the dynamic model of the load converter has been taken into Aug. 2015.
[13] M. K. Zadeh, B. Zahedi, M. Molinas, and L. E. Norum, “Centralized
the system analysis. The proposed method has been applied stabilizer for marine dc microgrid,” in Proc. 39th Annu. Conf. IEEE Ind.
to a simple dc distribution system with a cascade architecture, Electron. Soc. (IECON), Nov. 2013, pp. 3359–3363.
as a generic system model. [14] M. Farasat, A. Arabali, and A. M. Trzynadlowski, “Flexible-voltage
dc-bus operation for reduction of switching losses in all-electric
According to the theoretical results, the stability margin and ship power systems,” IEEE Trans. Power Electron., vol. 29, no. 11,
the dynamic performance of the system are influenced by both pp. 6151–6161, Nov. 2014.
source and load controllers. Faster load reduces the stable [15] H. Zhang, F. Mollet, C. Saudemont, and B. Robyns, “Experimental
validation of energy storage system management strategies for a local dc
power, while faster source controller increases the stability distribution system of more electric aircraft,” IEEE Trans. Ind. Electron.,
margin. Load shaping is also an important factor, which vol. 57, no. 12, pp. 3905–3916, Dec. 2010.
influences the load dynamics and, consequently, the stability [16] B. Sarlioglu and C. T. Morris, “More electric aircraft: Review, chal-
lenges, and opportunities for commercial transport aircraft,” IEEE Trans.
of the system, as has been validated by the laboratory tests. Transp. Electrification, vol. 1, no. 1, pp. 54–64, Jun. 2015.
Moreover, the impact of the stabilizer on the system stability [17] A. Khaligh, “Realization of parasitics in stability of DC–DC converters
has been investigated in the steady state and transients. Based loaded by constant power loads in advanced multiconverter automotive
on the experimental results, the stabilizer gain can be adjusted systems,” IEEE Trans. Ind. Electron., vol. 55, no. 6, pp. 2295–2305,
Jun. 2008.
to provide the required system damping. This defines the tran- [18] P. Magne, B. Nahid-Mobarakeh, and S. Pierfederici, “Dynamic consid-
sient characteristics of the system response, such as overshoot eration of dc microgrids with constant power loads and active damping
and settling time. system—A design method for fault-tolerant stabilizing system,” IEEE
J. Emerg. Sel. Topics Power Electron., vol. 2, no. 3, pp. 562–570,
The stability margin of the system may also be affected Sep. 2014.
by multiloads in larger microgrids. However, the dynamic [19] M. K. Zadeh, R. Gavagsaz-Ghoachani, J.-P. Martin, S. Pierfederici,
behavior of the system is mainly influenced by the interaction B. Nahid-Mobarakeh, and M. Molinas, “Discrete-time modelling, sta-
bility analysis, and active stabilization of dc distribution systems with
between the source and the load converter. In the proposed constant power loads,” in Proc. IEEE Appl. Power Electron. Conf.
model, the main system components are included into the Expo. (APEC), Mar. 2015, pp. 323–329.
discrete-time model of the system. Therefore, the stability [20] E. R. Vilamitjana, A. El Aroudi, and E. Alarcón, Chaos in Switching
Converters for Power Management Designing for Prediction and
analysis performed in this paper can be generalized to larger Control. New York, NY, USA: Springer-Verlag, 2012.
microgrids with multiple loads. [21] A. Emadi, B. Fahimi, and M. Ehsani, “On the concept of negative
impedance instability in the more electric aircraft power systems with
constant power loads,” SAE Tech. Paper 1999-01-2545, 1999.
R EFERENCES [22] M. K. Zadeh, G.-G. Roghayeh, S. Pierfederici, N.-M. Babak, and
[1] P. C. Loh, D. Li, Y. K. Chai, and F. Blaabjerg, “Autonomous operation of M. Molinas, “A discrete-time tool to analyze the stability of weakly
hybrid microgrid with ac and dc subgrids,” IEEE Trans. Power Electron., filtered active front-end PWM converters,” in Proc. IEEE Transp.
vol. 28, no. 5, pp. 2214–2223, May 2013. Electrification Conf. Expo. (ITEC), Jun. 2014, pp. 1–7.
[2] Y. Gu, W. Li, and X. He, “Frequency-coordinating virtual impedance for [23] A. Emadi, A. Khaligh, C. H. Rivetta, and G. A. Williamson, “Constant
autonomous power management of dc microgrid,” IEEE Trans. Power power loads and negative impedance instability in automotive systems:
Electron., vol. 30, no. 4, pp. 2328–2337, Apr. 2015. Definition, modeling, stability, and control of power electronic con-
[3] V. Nasirian, S. Moayedi, A. Davoudi, and F. L. Lewis, “Distributed verters and motor drives,” IEEE Trans. Veh. Technol., vol. 55, no. 4,
cooperative control of dc microgrids,” IEEE Trans. Power Electron., pp. 1112–1125, Jul. 2006.
vol. 30, no. 4, pp. 2288–2303, Apr. 2015. [24] A. M. Rahimi and A. Emadi, “An analytical investigation of dc/dc power
[4] A. Kwasinski and C. N. Onwuchekwa, “Dynamic behavior and stabiliza- electronic converters with constant power loads in vehicular power
tion of dc microgrids with instantaneous constant-power loads,” IEEE systems,” IEEE Trans. Veh. Technol., vol. 58, no. 6, pp. 2689–2702,
Trans. Power Electron., vol. 26, no. 3, pp. 822–834, Mar. 2011. Jul. 2009.
[5] X. Wang, F. Blaabjerg, and W. Wu, “Modeling and analysis of harmonic [25] M. K. Zadeh, M. Amin, J. A. Suul, M. Molina, and O. B. Fosso,
stability in an ac power-electronics-based power system,” IEEE Trans. “Small-signal stability study of the Cigré DC grid test system with
Power Electron., vol. 29, no. 12, pp. 6421–6432, Dec. 2014. analysis of participation factors and parameter sensitivity of oscillatory
[6] R. S. Balog, W. W. Weaver, and P. T. Krein, “The load as an energy modes,” in Proc. 18th Power Syst. Comput. Conf. (PSCC), Wrocław,
asset in a distributed dc SmartGrid architecture,” IEEE Trans. Smart Poland, Aug. 2014, pp. 1–8.
Grid, vol. 3, no. 1, pp. 253–260, Mar. 2012. [26] W. Du, J. Zhang, Y. Zhang, and Z. Qian, “Stability criterion for cascaded
[7] F. Blaabjerg, Z. Chen, and S. B. Kjaer, “Power electronics as efficient system with constant power load,” IEEE Trans. Power Electron., vol. 28,
interface in dispersed power generation systems,” IEEE Trans. Power no. 4, pp. 1843–1851, Apr. 2013.
Electron., vol. 19, no. 5, pp. 1184–1194, Sep. 2004. [27] D. Marx, P. Magne, B. Nahid-Mobarakeh, S. Pierfederici, and B. Davat,
[8] D. Boroyevich, I. Cvetkovic, R. Burgos, and D. Dong, “Intergrid: “Large signal stability analysis tools in dc power systems with constant
A future electronic energy network?” IEEE J. Emerg. Sel. Topics Power power loads and variable power loads—A review,” IEEE Trans. Power
Electron., vol. 1, no. 3, pp. 127–138, Sep. 2013. Electron., vol. 27, no. 4, pp. 1773–1787, Apr. 2012.
102 IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 4, NO. 1, MARCH 2016

[28] A. Griffo and J. Wang, “Large signal stability analysis of ‘more electric’ Roghayeh Gavagsaz-Ghoachani received the
aircraft power systems with constant power loads,” IEEE Trans. Aerosp. M.Sc. degree in electrical engineering from the
Electron. Syst., vol. 48, no. 1, pp. 477–489, Jan. 2012. Institut National Polytechnique de Lorraine, Nancy,
[29] M. Belkhayat, R. Cooley, and A. Witulski, “Large signal stability France, in 2007, and the Ph.D. degree in electrical
criteria for distributed systems with constant power loads,” in Proc. engineering from the Université de Lorraine, Nancy,
26th Annu. IEEE Power Electron. Specialists Conf. (PESC), Jun. 1995, in 2012.
pp. 1333–1338. She is currently a Researcher with the Groupe
[30] N. Bottrell, M. Prodanovic, and T. C. Green, “Dynamic stability of a de Recherche en Electrotechnique et Electronique
microgrid with an active load,” IEEE Trans. Power Electron., vol. 28, de Nancy, Université de Lorraine, and also an
no. 11, pp. 5107–5119, Nov. 2013. Associate Professor with the Department of
[31] S. Y. Erich and W. M. Polivka, “Input filter design criteria for current- Renewable Energies, Shahid Beheshti University,
programmed regulators,” IEEE Trans. Power Electron., vol. 7, no. 1, Tehran, Iran. Her current research interests include the stability study and
pp. 143–151, Jan. 1992. control of power electronics systems.
[32] M. Cespedes, L. Xing, and J. Sun, “Constant-power load system
stabilization by passive damping,” IEEE Trans. Power Electron., vol. 26,
no. 7, pp. 1832–1836, Jul. 2011.
[33] L. Xing, F. Feng, and J. Sun, “Optimal damping of EMI filter input
impedance,” IEEE Trans. Ind. Appl., vol. 47, no. 3, pp. 1432–1440, Serge Pierfederici received the Dipl.-Ing.
May/Jun. 2011. degree in electrical engineering from the Institu
[34] P. Magne, D. Marx, B. Nahid-Mobarakeh, and S. Pierfederici, National Polytechnique de Lorraine (INPL), Ecole
Nationale Supérieure d’Electricité et de Mécanique,
“Large-signal stabilization of a dc-link supplying a constant power load
using a virtual capacitor: Impact on the domain of attraction,” IEEE Université de Lorraine, Nancy, France, in 1994, and
Trans. Ind. Appl., vol. 48, no. 3, pp. 878–887, May/Jun. 2012. the Ph.D. degree from INPL, in 1998.
[35] P. Magne, B. Nahid-Mobarakeh, and S. Pierfederici, “General active He has been with INPL since 1999, where he
global stabilization of multiloads dc-power networks,” IEEE Trans. is currently a Full Professor. His current research
Power Electron., vol. 27, no. 4, pp. 1788–1798, Apr. 2012. interests include stability study of distributed
[36] E. Jamshidpour, B. Nahid-Mobarakeh, P. Poure, S. Pierfederici, power systems and modeling and control of power
electronic converters.
F. Meibody-Tabar, and S. Saadate, “Distributed active resonance sup-
pression in hybrid dc power systems under unbalanced load conditions,”
IEEE Trans. Power Electron., vol. 28, no. 4, pp. 1833–1842, Apr. 2013.
[37] M. K. Zadeh, R. Parseh, M. Molinas, and K. Kansanen, “Modeling
and simulation of wireless communication based robust controller Babak Nahid-Mobarakeh (M’05–SM’12) received
for multi-converter systems,” in Proc. IEEE Int. Conf. Smart Grid the Ph.D. degree in electrical engineering from the
Commun. (SmartGridComm), Oct. 2013, pp. 738–743. Institut National Polytechnique de Lorraine, Nancy,
[38] M. K. Zadeh, L.-M. Saublet, R. Gavagsaz-Ghoachani, J.-P. Martin, France, in 2001.
S. Pierfederici, and B. Nahid-Mobarakeh, “A new discrete-time mod- He was with the Centre de Robotique,
elling of PWM converters for stability analysis of DC microgrid,” Electrotechnique et Automatique, University of
in Proc. Int. Conf. Modeling Simulation Electr. Mach., Converters, Picardie, Amiens, France, as an Assistant Professor,
Syst. (Electrimacs), 2014. from 2001 to 2006. In 2006, he joined the
[39] S. Banerjee and G. C. Verghese, Eds., Nonlinear Phenomena in Power Ecole Superieure d’Electricite et de Mecanique,
Electronics. Piscataway, NJ, USA: IEEE Press, 1999. Université de Lorraine, Nancy, where he is currently
[40] A. H. Nayfeh and B. Balachandran, Applied Nonlinear Dynamics: an Associate Professor. He is also with the Groupe
Analytical, Computational and Experimental Methods. New York, NY, de Recherche en Electrotechnique et Electronique de Nancy. He has authored
USA: Wiley, 2008. or co-authored over 100 international journals and conference papers. His
[41] E. Monmasson, Ed., Power Electronic Converters: PWM Strategies and current research interests include nonlinear and robust control techniques
Current Control Techniques. New York, NY, USA: Wiley, 2013. applied to electric systems, fault detection and fault tolerant control of power
[42] S.-C. Tan, Y. M. Lai, and C. K. Tse, “Indirect sliding mode control systems, and stabilization of microgrids.
of power converters via double integral sliding surface,” IEEE Trans.
Power Electron., vol. 23, no. 2, pp. 600–611, Mar. 2008.
[43] A. Lachichi, S. Pierfederici, J.-P. Martin, and B. Davat, “Study of a
hybrid fixed frequency current controller suitable for dc–dc applica-
tions,” IEEE Trans. Power Electron., vol. 23, no. 3, pp. 1437–1448, Marta Molinas (M’94) received the Diploma
May 2008. degree in electromechanical engineering from
[44] C. K. Tse, Complex Behavior of Switching Power Converters. the National University of Asuncion, Asuncion,
Boca Raton, FL, USA: CRC Press, 2003. Paraguay, in 1992, the M.Eng. degree from
[45] M. K. Zadeh, R. Parseh, M. Molinas, and K. Kansanen, “Bifurcation Ryukyu University, Okinawa, Japan, in 1997, and
in PWM converter-based systems with wireless communication-based the D.Eng. degree from the Tokyo Institute of
current controller,” in Proc. 4th IEEE/PES Innov. Smart Grid Technol. Technology, Tokyo, Japan, in 2000.
Eur., Oct. 2013, pp. 1–5. She was a Guest Researcher with the
[46] S. K. Mazumder, A. H. Nayfeh, and D. Boroyevich, “Theoretical University of Padua, Padua, Italy, in 1998. From
and experimental investigation of the fast- and slow-scale instabilities 2004 to 2007, she was a Post-Doctoral Researcher
of a dc–dc converter,” IEEE Trans. Power Electron., vol. 16, no. 2, with the Norwegian University of Science and
pp. 201–216, Mar. 2001. Technology (NTNU), Trondheim, Norway, and a Professor with the
Department of Electric Power Engineering, NTNU, from 2008 to 2014.
In 2014, she was with Columbia University, New York, NY, USA, where
Mehdi Karbalaye Zadeh received the M.Sc. degree she was involved in microgrids for developing regions. She is currently with
in electrical engineering from the University of the Department of Engineering Cybernetics, NTNU, as a Full Professor.
Tehran, Tehran, Iran, in 2010. He is currently pur- From 2008 to 2009, she was a Japan Society for the Promotion of Science
suing the Ph.D. degree with the Department of Research Fellow for ten months with the Energy Technology Research
Electric Power Engineering, Norwegian University Institute, National Institute of Advanced Industrial Science and Technology,
of Science and Technology, Trondheim, Norway. Tsukuba, Japan. Her current research interests include stability of power
electronics systems, harmonic and oscillatory phenomena, and nonstationary
His current research interests include control and
stability analysis of power electronic systems, smart signals from human and machines.
grid, and dc power distribution. Dr. Molinas has been an AdCom Member of the IEEE Power Electronics
Society. She is a Reviewer of the IEEE T RANSACTIONS ON I NDUSTRIAL
E LECTRONICS and the IEEE T RANSACTIONS ON P OWER E LECTRONICS .