Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
4, DECEMBER 2017
Abstract— In this paper, a new method for power system the system with moderate to large size. The impedance-based
stability analysis is introduced. The method is based on injection method is an alternative method, which decomposes the
of a small voltage or current in an arbitrary point of a power system into a source and load impedance equivalent [1]–[3].
system. The apparent impedance is then defined as the ratio
between the voltage and current at the injection point. It is Stability can be analyzed by applying the Nyquist criterion
shown that the apparent impedance can be used to estimate the to the ratio between source and load impedances. The main
eigenvalues of the system that are observable from the injection advantage of the impedance-based method is that the stability
point. The eigenvalues are obtained by applying system identifi- can be analyzed from measurement/simulations in a single
cation techniques to the measured set of apparent impedances. point in the system. Hence, this can be viewed as a black-box
The method is similar to the well-established impedance-based
stability analysis based on source and load impedance models. approach. In order to perform such analysis, a disturbance
However, while the source/load impedance ratio is viewed as or perturbation must normally be injected into the system.
the minor-loop gain, the apparent impedance can be viewed as A drawback of the impedance-based method is that the
a closed-loop transfer function. It can also be expressed as the stability margin depends on the interface point location, i.e.,
parallel connection of the source and load impedance. It is shown, the point which defines the source and load subsystem.
in this paper, how the system eigenvalues can be extracted based
on a set of apparent impedance values. The apparent impedance This paper proposes a new method for stability analysis
holds, therefore, complementary information compared with the called the apparent impedance method. The approach is based
existing impedance-based stability analysis. The method can also on measurements in a single point similar to the above-
be used as a tool to validate analytically derived state-space mentioned impedance-based analysis. However, instead of
models. In this paper, the method is presented as a simulation obtaining the source and load impedances separately, only
tool, while further work will extend it to include experimental
setups. Two case studies are presented to illustrate the method: the equivalent impedance seen from the injection source must
1) a dc case with a buck converter feeding a constant power load be identified. This impedance is defined as the apparent
and 2) a three-phase grid-connected voltage source converter impedance and represents a closed-loop transfer function in
with a current controller and a phase lock loop. The estimated the system. Consequently, the system state-space model can be
(apparent) eigenvalues of the studied systems are equal to those estimated based on sampled values of the apparent impedance.
obtained from the analytic state-space model.
Since this is a closed-loop transfer function, the eigenvalues of
Index Terms— Impedance-based analysis, power system the estimated state-space model are independent of injection
stability analysis, state-space modeling, vector fitting (VF). location. Well-established system identification tools can be
used for the estimation. In this paper, the vector fiting (VF)
I. I NTRODUCTION
and matrix fitting (MF) methods are applied [4].
Fig. 1. General dc power system partitioned into two subsystem with shunt
injection. Top: circuit diagram. Bottom: block diagram. Fig. 2. General three-phase power system in the dq-domain with shunt
injection. Top: circuit diagram. Bottom: block diagram.
4) obtain a continuous impedance model from a discrete The apparent impedance is defined as
set of impedance values. Vinj (s)
The presented method will give an accurate validation Z a (s) = (1)
Iinj (s)
by comparing each eigenvalue individually, and this will
also give hints toward where, in the modeling, the error is where Vinj and Iinj are defined in Fig. 1. Uppercase letters are
located. used to indicate Laplace domain and frequency domain in this
The last application is useful when interpreting paper. The following expressions can then be obtained for Z a
Bode and Nyquist plots obtained from the simulated or by circuit analysis applied to Fig. 1:
measured impedance values. The state-space model obtained Z1 Z2
from vector and MF methods can be recalculated into a Z a = Z 1 ||Z 2 = . (2)
Z1 + Z2
continuous impedance model, which will give smooth curves.
This will make it easier to accurately identify resonances and Note that, these expressions do not depend on the type of
to accurately evaluate the behavior close to the (−1, 0)-point subsystem equivalent (Thevenin versus Norton), since we are
in the Nyquist plot. disregarding the impact of the voltage and current sources
In this paper, the apparent impedance method is presented during small-signal analysis.
mainly as a simulation tool. However, it can be extended
to also perform online identification of state-space models III. S TABILITY A NALYSIS BY THE A PPARENT I MPEDANCE
in experimental setups. Assuming that accurate impedance Apparent impedance stability analysis is a small-signal
measurement equipment is available, the implementation will method. The objective is to estimate all eigenvalues of the
be simple and with low computational requirements. system based only on sampled values of Z a . The stability
This paper extends a previous work by Rygg et al. [5]. Here, analysis follows directly by evaluating these eigenvalues in the
the method was defined for dc systems, and it was proved complex plane. A flowchart of the methodology is presented
that both the shunt current and series voltage injections are in Fig. 3.
applicable. Furthermore, a case-study analysis demonstrated
that the apparent impedance eigenvalues do not depend on
injection location. This paper extends the method to include A. DC Systems
three-phase systems, and the dq-domain is utilized for this In the small-signal analysis of dc systems, it is assumed that
task. Only shunt current injection is assumed, but the same the entire system can be represented by a linearized state-space
results can be obtained by series voltage injection. model
sx = Ax + Bu
II. A PPARENT I MPEDANCE D EFINITION
y = C x + (D + s E)u (3)
The definition of the apparent impedance assumes an injec-
tion of voltage or current at some point in the power system. where x is the vector of n states and u is the single input
In Fig. 1, an injection is applied to a dc system composed by to the system, while y is the single output. A is the n × n
two subsystems. The injection point separates the system into state matrix, B is an n × 1-vector, and C is a 1 × n-vector,
two subsystems (1 and 2), here represented by their frequency- while D and E are scalars. The term E is required to represent
dependent Thevenin and Norton equivalents. nonproper transfer functions (see Section V and Appendix B
1476 IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 5, NO. 4, DECEMBER 2017
IV. O BTAINING A PPARENT I MPEDANCE remarked that the injection signals are not synchronized with
F ROM S IMULATIONS the grid voltage v inj during simulation; hence, the d-axis
Obtaining impedance values by simulation can be achieved of the injection signal has an arbitrary phase shift with
by the most time-domain analysis tools. The idea is to inject a respect to the grid voltage. Instead, the current and voltage
small disturbance in the interface point as shown in Fig. 1. The measurements are aligned with the fundamental component
disturbance can contain a single frequency (single tone), or be of the grid voltage during postprocessing. Consequently, the
composed by several frequencies (multitone). In this paper, q-axis component of i inj,3 ph,1 and the d-axis component of
only multitone is applied, but identical results can be obtained i inj,3 ph,2 will not be zero during postprocessing, since the
by single-tone analysis. d-axis is aligned to a different angle.
The methodology is shown by the flowchart in Fig. 3. The next step is to simulate the system with the above-
Both the dc and three-phase systems are considered. The defined injection signals applied. Time-domain response of
methodology is based on selecting a frequency vector ftab = v inj and i inj is then stored and transformed into frequency
[ f 1 f 2 . . . f n ] for which the apparent impedance should be domain by the FFT. For the three-phase case, the responses
estimated. The next step is to synthesize an injection signal as are transformed to the dq-domain before applying the FFT.
the sum of sinusoids at these frequencies The apparent impedance can then be calculated based on the
frequency domain data. For the dc case, this is simply the
n
ratio of injection point voltage by injection current (1). For
i inj,DC (t) = Imag sin (wtab (i )t) (11) the three-phase case, it is necessary to combine the responses
i=1
from the two injection signals as derived in [15]. The fol-
where ωtab = 2π ftab . lowing equation can be used to find the impedance matrix
For three-phase systems, the apparent impedance is a (see [15, eq. (19)]):
2 × 2 matrix in the dq-domain. Impedance measurements by −1
Vinj1,d Vinj2,d Iinj1,d Iinj2,d
frequency sweeps are well established in previous work [15]. Za,dq =
Both the dq-domain and sequence domain are applied for Vinj1,q Vinj2,q Iinj1,q Iinj2,q
−1
this task, while this paper is focusing on the dq-domain. = Vinj,mat Iinj,mat . (13)
A comparison of different injection techniques is out of the
After this step, the apparent impedance is identified at all
scope of this paper, and it is remarked that all techniques
frequencies in f tab , and this is the required input data for the
will give the same result when the assumption of a time
invariant (TI) system is satisfied. system identification process described in Section V.
The currents and voltages are transformed into the
dq-domain by applying Park transform with phase angle V. E STIMATING S TATE -S PACE M ODELS BY
equal to θ (t) = θ1 + ω1 t. The initial condition θ1 is set V ECTOR F ITTING AND M ATRIX F ITTING
equal to the phase A fundamental voltage angle. This
angle is identified during postprocessing by taking the fast VF is a well-established method for rational approxi-
fourier transform (FFT) of the voltage v inj . Consequently, mation in the frequency domain using poles and residues
a phase lock loop (PLL) is not needed to perform the [4], [18], [19]. The method is able to estimate a state-space
injection. The method for calculating the impedance matrix model to a measured or computed transfer function based on
is based on the work in [15], and it is necessary to combine curve fitting. VF is widely applied in many engineering fields,
two linear independent injection signals as explained from high-voltage power systems to microwave systems and
in [15] and [16]. A possible combination of injection signals high-speed electronics. A MATLAB implementation of the
is method is available online [20].
⎡ ⎤ The input to the VF algorithm is as follows:
sin (ωtab (i )t) cos
(ω1 t + 0) 1) a set of measured/simulated apparent impedance values
⎢ 2π ⎥
n
⎢sin (ωtab (i )t) cos ω1 t − ⎥ function values Z a1 , Z a2 ...Z an taken at the frequencies
i inj,3 ph,1 (t) = ⎢
Imag ⎢ 3 ⎥
⎣ ⎥ f 1 , f 2 ... f n ;
i=1 2π ⎦ 2) the order of the resulting state-space model. The
sin (ωtab (i )t) cos ω1 t +
3 maximum possible value is the number of impedance
⎡ ⎤
sin (ωtab (i )t) cos (ω
1 t + 0) values (n).
⎢ 2π ⎥ The output of VF is then the state-space model represented
n
⎢sin (ωtab (i )t) cos ω1 t + ⎥
i inj,3 ph,2 (t) = ⎢
Imag ⎢ 3 ⎥ by A, B, C, D, and E (3). The model is also expressed on the
⎥
i=1 ⎣ 2π ⎦ pole-residue form (see Appendix C for details).
sin (ωtab (i )t) cos ω1 t − The fitting algorithm will identify the system eigenvalues
3
(12) with high accuracy when the system under study is TI. Still,
it must be remarked that the resulting state variables cannot
where ω1 is the fundamental frequency and the three elements be related to any physical quantities in the system when a
in the vector are the a, b, and c phase components. With pure black-box approach is adopted. This is a consequence of
this choice of injection signals, i inj,3 ph,1 is a pure d-axis having a black-box approach in general, and not a weakness
component, while i inj,3 ph,2 is a pure q-axis component. It is of the specific method.
1478 IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 5, NO. 4, DECEMBER 2017
matrix to be independent of the others. More information on The system is represented in MATLAB Simulink Simscape
MF can be found in Appendix D. Power Systems using a switched (detailed) model. The simu-
lation time step is fixed and equal to Tsim = 1 μs.
A. Selecting the Model Order B. Analytical State-Space Model
By circuit analysis, the state-space model of the system in
An important parameter in the system identification process
Fig. 5 is derived in (15) and (16). The state-space model is
is the model order. An mth order model will give m states and
without any elements B, C, D, and E, since only the eigen-
m eigenvalues. The number of states in the system is unknown
values of matrix A are relevant for the stability analysis. The
by the apparent impedance method, since this is a black-box
CPL has been linearized around its operation point denoted
approach. A methodology presented in Fig. 4 is proposed
by the stationary voltage VC2 . The applied parameter values
to identify insignificant states and eigenvalues. It is known
are given in Table I
that when the residue Ri divided by eigenvalue λi has small ⎡ ⎤ ⎡ ⎤
absolute value for the state i, this state does not contribute i L1 i L1
⎢v C1 ⎥ ⎢v C1 ⎥
to the measured response [4]. Consequently, by gradually ⎢ ⎥ ⎢ ⎥
reducing model order until all eigenvalues are significant, s⎢ ⎥ ⎢ ⎥
⎢ i L2 ⎥ = A ⎢ i L2 ⎥ (15)
the correct number of states will be identified. A thresh- ⎣v C2 ⎦ ⎣v C2 ⎦
old must also be selected for this detection. In this paper, v C2
˜ v C2
˜
RYGG AND MOLINAS: SMALL-SIGNAL METHOD FOR STABILITY ANALYSIS OF POWER ELECTRONIC-BASED SYSTEMS 1479
Fig. 5. Schematic of the dc case study system, including the injection source (iinj ).
TABLE I
PARAMETER D ATA A PPLIED IN THE S IMULATION M ODEL
where
⎡ R +R 1 Rc1 ⎤
1 c1
− − 0 0
⎢ L1 L1 L1 ⎥
⎢ ⎥ Fig. 6. Steady-state waveforms of iinj and v inj . Injection frequencies given
⎢ 1 1 ⎥
⎢ 0 − 0 0 ⎥ by (17).
⎢ C1 C1 ⎥
⎢ ⎥
⎢ Rc1 1 R1 + R2 + Rc1 1 ⎥
⎢ − − 0 ⎥
A=⎢ ⎢ ⎥.
L2 L L2 L2 ⎥
⎢ P ⎥∗
⎢ 1 ⎥
⎢ 0 0 0 2 ⎥
⎢ C2 C2 VC2 ⎥
⎢ 1 ⎥
⎢ 1 ⎥
⎣ 0 0 0 − ⎦
τ τ
(16)
The eigenvalues λanalytic of the state matrix A have been
calculated in MATLAB using the data in Table I. They are
compared with the eigenvalues obtained by the proposed
method in Section VI-D.
C. Illustration of Methodology
In this paper, a multitone signal composed by eight frequen-
Fig. 7. Harmonic components of iinj and v inj with injection at f tab (17).
cies is injected. The frequencies are logarithmically spaced in Obtained by FFT applied to Fig. 6.
the range between 2 and 2000 Hz as
√
f tab = [2, 6, 14, 38, 104, 278, 746, 2000] H z. (17) rms of (1/ 2) · 8 = 2.8 mA. This is approximately 0.5 % of
the average load current.
A time-domain simulation of the steady-state operation is As highlighted in Fig. 3, the signals in Fig. 6 are the only
presented in Fig. 6. The injected current i inj and the intersec- information needed to perform the stability analysis. First,
tion point voltage v inj are indicated in the plot. The amplitude FFT is applied to v inj and i inj, and the result is presented
of each injected frequency component is 0.5 mA, giving a total in Fig. 7. Since shunt current is applied to this example,
1480 IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 5, NO. 4, DECEMBER 2017
TABLE II
C OMPARISON OF A NALYTICAL E IGENVALUES W ITH THE A PPARENT
Fig. 8. Impedance plots obtained from the dc case system. "x" represents E IGENVALUES λa O BTAINED AT THE DC C ASE S YSTEM .
the simulated values at f tab (17), while the solid lines represent VF. ζ : D AMPING R ATIO AND f : O SCILLATION F REQUENCY
sx = Ax
s s s s c c s s
x = [x pwm1,d , x pwm1,q , x pwm2,d , x pwm2,q , i cd , i cq , v pd , v pq , ṽ pd , ṽ pq , i gd , i gq , γdc , γqc , γPLL , θ ]T
⎡ 4 4 K p Vdc Kf 1 Vdc ⎤
− ω1 − 2 0 − − 0 0 0 0 0 q 0 0 α1
⎢ Tc Tc 2Tc Tc Tc 2Tc ⎥
⎢ 4 4 K K V 1 V ⎥
⎢−ω1 − 0 − 2
f
−
p dc
0 0 0 0 0 0
dc
0 α2 ⎥
⎢ Tc Tc Tc 2Tc Tc 2Tc ⎥
⎢ ⎥
⎢ 1 0 0 ω1 0 0 0 0 0 0 0 0 0 0 0 0 ⎥
⎢ 0 1 −ω1 0 0 0 0 0 0 0 0 0 0 0 0 0 ⎥
⎢ ⎥
⎢ 2 4 K ad + Rc 1 K ad Vcq ⎥
0
⎢− 0 − ω1 − − ⎥
⎢ L c 0 Tc L c L L
0 0 0
Lc
0 0 0 0
L0 c ⎥
⎢ c
K ad + Rc
c
Vcd ⎥
⎢ 2 4 1 K ad ⎥
⎢ 0 − 0 −ω1 − 0 − 0 0 0 0 0 0 ⎥
⎢ Lc Tc L c Lc Lc Lc Lc ⎥
⎢ 1 1 ⎥
⎢ 0 0 0 0 0 0 ω1 0 0 − 0 0 0 0 0 ⎥
⎢ Cc CC ⎥
⎢ 1 1 ⎥
⎢ 0 0 0 0 q0 −ω1 0 0 0 0 − 0 0 0 0 ⎥
⎢ ⎥
A =⎢
⎢
Cc
1 1
Cc
Vpq ⎥
0
⎥
⎢ 0 0 0 0 0 0 0 − 0 0 0 0 0 0 ⎥
⎢ τv τv τv 0 ⎥
⎢ Vpd ⎥
⎢ 1 1 ⎥
⎢ 0 0 0 0 0 0 0 0 − 0 0 0 0 0 − ⎥
⎢ τv τv τv ⎥
⎢ 1 Rth ⎥
⎢ 0 0 0 0 0 0 0 0 0 − ω1 0 0 0 0 ⎥
⎢ L th L th ⎥
⎢ 1 R ⎥
⎢ 0 0 0 0 0 0 0 0 0 −ω1 −
th
0 0 0 0 ⎥
⎢ ⎥
⎢ L th L th ⎥
⎢ Kp Icq K p ⎥
0
⎢ 0 0 0 0 − 0 0 0 0 0 0 0 0 0 0 − ⎥
⎢ Ti Ti ⎥
⎢ Icd K p ⎥
0
⎢ Kp ⎥
⎢ 0 0 0 0 0 − 0 0 0 0 0 0 0 0 0 ⎥
⎣ Ti Ti ⎦
0 0 0 0 0 0 0 K i,PLL 0 0 0 0 0 0 0 α3
0 0 0 0 0 0 0 K p,PLL 0 0 0 0 0 0 1 α4
− 12 Icq
0 K V + K I0
p dc f cd
α1 =
Tc
1 0
2 Icd K p Vdc + K f Icq
0
α2 =
Tc
α3 = −K i,PLL Vpd
0
0
α4 = −K p,PLL Vpd (18)
Fig. 10. Schematic of three-phase case-study system, including the injection source (iinj ). Left: Buck converter. Right: CPL.
Fig. 12. Comparison of analytical and apparent eigenvalues for the three-
phase case. Left: full overview. Right: zoomed-in view of origin.
the 2×2 dq impedance matrix Zdq is the transfer function, [7] N. Bottrell, M. Prodanovic, and T. C. Green, “Dynamic stability of a
which yields microgrid with an active load,” IEEE Trans. Power Electron., vol. 28,
no. 11, pp. 5107–5119, Nov. 2013.
Z dd (s) Z dq (s) [8] S. D’Arco, J. A. Suul, and O. B. Fosso, “Automatic tuning of cas-
Zdq (s) = = C (sI − A)−1 B + D + sE
Z qd (s) Z qq (s) caded controllers for power converters using eigenvalue parametric
sensitivities,” IEEE Trans. Ind. Appl., vol. 51, no. 2, pp. 1743–1753,
n
Ri Mar. 2015.
= + D + sE (23) [9] S. Hiti, V. Vlatkovic, D. Borojevic, and F. C. Y. Lee, “A new control
s − λi algorithm for three-phase PWM buck rectifier with input displacement
i=1
factor compensation,” IEEE Trans. Power Electron., vol. 9, no. 2,
where Ri is the 2×2 residue matrix corresponding to the pp. 173–180, Mar. 1994.
i th eigenvalue λi . When defining a norm to evaluate if an [10] S. Hiti, D. Boroyevich, and C. Cuadros, “Small-signal model-
ing and control of three-phase PWM converters,” in Proc. Conf.
eigenvalue is significant or not, the following condition is Rec. IEEE Ind. Appl. Soc. Annu. Meet., vol. 2. Oct. 1994,
proposed: pp. 1143–1150.
||Ri ||2 [11] K. D. T. Ngo, “Low frequency characterization of PWM convert-
> 10−4 (24) ers,” IEEE Trans. Power Electron., vol. PE-1, no. 4, pp. 223–230,
|λi | Oct. 1986.
where ||(·)||2 is the 2-norm of a matrix (other matrix norms [12] Z. Yao, P. G. Therond, and B. Davat, “Frequency characteristic of AC
power system,” in Proc. IFAC World Congr., Jul. 1993, pp. 737–740.
can also be applied). [13] R. Burgos, D. Boroyevich, F. Wang, K. Karimi, and G. Francis,
“On the AC stability of high power factor three-phase rectifiers,”
E. Definition of a Passive System in Proc. IEEE Energy Convers. Congr. Expo. (ECCE), Sep. 2010,
pp. 2047–2054.
The definition from [27] is adopted in this paper. A passive [14] B. Wen, D. Dong, D. Boroyevich, R. Burgos, P. Mattavelli, and Z. Shen,
system is defined as a system that is not able to generate “Impedance-based analysis of grid-synchronization stability for three-
phase paralleled converters,” IEEE Trans. Power Electron., vol. 31, no. 1,
energy on its own. Mathematically, this can be formulated by pp. 26–38, Jan. 2015.
representing the system with transfer matrix H(s) [e.g., the [15] G. Francis, R. Burgos, D. Boroyevich, F. Wang, and K. Karimi,
2×2 impedance matrix Zdq (s)]. If the system is passive, the “An algorithm and implementation system for measuring impedance in
the D-Q domain,” in Proc. IEEE Energy Convers. Congr. Expo. (ECCE),
following condition holds: Sep. 2011, pp. 3221–3228.
λi ≥ 0 ∀λi ∈ (H( j ω) + H( j ω) H ) ω∈R (25) [16] Y. A. Familiant, J. Huang, K. A. Corzine, and M. Belkhayat, “New
techniques for measuring impedance characteristics of three-phase
where H( j ω) H is the conjugate transpose (Hermitian trans- AC power systems,” IEEE Trans. Power Electron., vol. 24, no. 7,
pp. 1802–1810, Jul. 2009.
pose) of H( j ω). The condition is that all eigenvalues of [17] M. Cespedes and J. Sun, “Three-phase impedance measurement for sys-
H( j ω) + H( j ω) H are positive for all frequencies ω. tem stability analysis,” in Proc. IEEE 14th Workshop Control Modeling
For an SISO system H ( j ω), the condition is simplified Power Electron. (COMPEL), Jun. 2013, pp. 1–6.
[18] B. Gustavsen, “Improving the pole relocating properties of vector
to requiring −90° ≤ H ( j ω) ≤ 90°. This is a more fitting,” IEEE Trans. Power Del., vol. 21, no. 3, pp. 1587–1592,
intuitive definition of passivity, since a negative real part of Jul. 2006.
the terminal characteristic is associated with the production of [19] D. Deschrijver, M. Mrozowski, T. Dhaene, and D. D. Zutter, “Macro-
modeling of multiport systems using a fast implementation of the vector
energy. fitting method,” IEEE Microw. Wireless Compon. Lett., vol. 18, no. 6,
pp. 383–385, Jun. 2008.
ACKNOWLEDGMENT [20] The Vector Fitting. Accessed on Feb. 19, 2016. [Online]. Available:
https://www.sintef.no/projectweb/vectfit/
The authors would like to thank Dr. B. Gustavsen at [21] L. Harnefors, X. Wang, A. G. Yepes, and F. Blaabjerg, “Passivity-
SINTEF Energy Research for developing the dedicated ver- based stability assessment of grid-connected VSCs—An overview,”
sion of the Matrix Fitting routine that was used in this IEEE J. Emerg. Sel. Topics Power Electron., vol. 4, no. 1,
pp. 116–125, Mar. 2016.
paper.
[22] L. Harnefors, R. Finger, X. Wang, H. Bai, and F. Blaabjerg, “VSC
input-admittance modeling and analysis above the Nyquist frequency
R EFERENCES for passivity-based stability assessment,” IEEE Trans. Ind. Electron.,
[1] R. D. Middlebrook, “Input filter considerations in design and application vol. 64, no. 8, pp. 6362–6370, Aug. 2017.
of switching regulators,” in Proc. IEEE Ind. Appl. Soc. Annu. Meet., [23] X. Wang, F. Blaabjerg, and P. C. Loh, “Passivity-based stability
1976, pp. 366–382. analysis and damping injection for multiparalleled VSCs with LCL
[2] M. Belkhayat, “Stability criteria for AC power systems with regulated filters,” IEEE Trans. Power Electron., vol. 32, no. 11, pp. 8922–8935,
loads,” Ph.D thesis, Purdue Univ., West Lafayette, IN, USA, 1997. Nov. 2017.
[3] J. Sun, “Small-signal methods for AC distributed power systems— [24] L. Harnefors, M. Bongiorno, and S. Lundberg, “Input-admittance
A review,” IEEE Trans. Power Electron., vol. 24, no. 11, pp. 2545–2554, calculation and shaping for controlled voltage-source converters,”
Nov. 2009. IEEE Trans. Ind. Electron., vol. 54, no. 6, pp. 3323–3334,
[4] B. Gustavsen and A. Semlyen, “Rational approximation of frequency Dec. 2007.
domain responses by vector fitting,” IEEE Trans. Power Del., vol. 14, [25] C. A. Desoer and Y.-T. Wang, “On the generalized Nyquist stability
no. 3, pp. 1052–1061, Jul. 1999. criterion,” IEEE Trans. Autom. Control, vol. AC-25, no. 2, pp. 187–196,
[5] A. Rygg, M. Amin, M. Molinas, and B. Gustavsen, “Apparent impedance Apr. 1980.
analysis: A new method for power system stability analysis,” in Proc. [26] J. Kwon, X. Wang, F. Blaabjerg, C. L. Bak, V.-S. Sularea, and C. Busca,
IEEE 17th Workshop Control Modeling Power Electron. (COMPEL), “Harmonic interaction analysis in a grid-connected converter using
Jun. 2016, pp. 1–7. harmonic state-space (HSS) modeling,” IEEE Trans. Power Electron.,
[6] F. Katiraei, M. R. Iravani, and P. W. Lehn, “Small-signal dynamic model vol. 32, no. 9, pp. 6823–6835, Sep. 2016.
of a micro-grid including conventional and electronically interfaced [27] S. Grivet-Talocia and B. Gustavsen, Passive Macromodeling:
distributed resources,” IET Generat., Transmiss. Distrib., vol. 1, no. 3, Theory and Applications, vol. 239. Hoboken, NJ, USA: Wiley,
pp. 369–378, May 2007. 2015.
1486 IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 5, NO. 4, DECEMBER 2017
Atle Rygg received the M.Sc. degree in electri- Marta Molinas (M’94) received the Diploma
cal engineering from the Norwegian University of degree in electromechanical engineering from
Science and Technology, Trondheim, Norway, in the National University of Asuncion, Asunción,
2011, where he is currently working toward the Paraguay, in 1992, the M.E. degree from Ryukyu
Ph.D. degree with the Department of Engineering University, Nishihara, Japan, in 1997, and the D.Eng.
Cybernetics. degree from the Tokyo Institute of Technology,
From 2011 to 2015, he was a Research Sci- Tokyo, Japan, in 2000.
entist in power electronics with SINTEF Energy She was a Guest Researcher with the University
Research, Trondheim. His current research interests of Padova, Padua, Italy, in 1998. From 2004 to
include impedance-based stability analysis of power 2007, she was a Post-Doctoral Researcher with
electronic systems, where the aim is to contribute the Norwegian University of Science and Technol-
to the fundamental understanding in this family of methods. ogy (NTNU), Trondheim, Norway, where she was a Professor with the
Department of Electric Power Engineering from 2008 to 2014. She is currently
a Professor with the Department of Engineering Cybernetics, NTNU. Her
current research interests include the stability of power electronics systems,
harmonics, instantaneous frequency, and nonstationary signals from the human
and the machine.
Dr. Molinas was an AdCom Member of the IEEE Power Electronics Society
from 2009 to 2011. She is an Associate Editor of the IEEE J OURNAL OF
E MERGING AND S ELECTED T OPICS IN P OWER E LECTRONICS Journal and
the IEEE T RANSACTIONS ON P OWER E LECTRONICS , and an Editor of the
IEEE T RANSACTIONS ON E NERGY C ONVERSION.