3758 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS—I: REGULAR PAPERS, VOL. 68, NO.

9, SEPTEMBER 2021

Stability Assessment for Multi-Infeed

Grid-Connected VSCs Modeled in
the Admittance Matrix Form
Luis Orellana , Luis Sainz, Eduardo Prieto-Araujo , Member, IEEE, and Oriol Gomis-Bellmunt , Fellow, IEEE

Abstract— The increasing use of power electronics converters interaction with components of the traditional power sys-
to integrate renewable energy sources has been a subject of tem such as synchronous generators, power transformers and
concern due to the resonance oscillatory phenomena caused by transmission lines. One of the most important problems is
their interaction with poorly damped AC networks. Early studies
are focused on assessing the controller influence of a single the oscillatory phenomena caused by the interaction between
converter connected to simple networks, and they are no longer the VSC control and the grid. These oscillations can lead
representative for existing systems. Lately, studies of multi-infeed to instabilities specially in poorly damped networks [1]–[3].
grid-connected converters are of particular interest, and their There are non-damped cases when the system maintains a
main aim is to apply traditional criteria and identify their sustained oscillation due to non-linearities such as saturation
difficulties in the stability assessment. An extension of traditional
criteria is commonly proposed as a result of these analysis, and limiters [4], [5].
but they can be burdensome for large and complex power The commonly used methods to model grid-connected VSC
systems. The present work addresses this issue by proposing a systems to study the oscillatory phenomena are the state-space
simple criterion to assess the stability of large power systems and impedance-based modelling approaches [6], [7]. The first
with high-penetration of power converters. The criterion has its one represents the system as a set of linear equations in
origin in the mode analysis and positive-net damping stability
criteria, and it addresses the stability in the frequency domain the time domain, but it requires detailed information of the
by studying the eigenvalues magnitude and real component of control code which is possibly not available. On the other
dynamic models in the admittance matrix form. Its effectiveness hand, impedance-based modelling approach is based on the
is tested in two case studies developed in Matlab/Simulink which impedance characterization of the system (e.g., detailed knowl-
compare it with traditional criteria, proving its simplicity. edge about the converts is not needed) which can be expressed
Index Terms— Stability analysis, grid-connected converter, as a transfer function in the s-domain. Stability criteria such
multi-infeed, nodal admittance matrix, generalized Nyquist cri- as the Nyquist stability criterion [8], [9], impedance-based
terion, frequency domain analysis. stability criterion [10]–[12], and the positive-net damping
stability criterion (PND) [13], [14] have been used to study
I. I NTRODUCTION the controller influence of a single VSC over the stability of
simple networks in the frequency domain.
T HE use of grid-connected power converters has been
increasing due to the need to connect large renewable
energy resources to the AC power network. These resources
Stability studies of multi-infeed VSC-based AC grids are
currently of great interest and different approaches to assess
are typically connected to the AC grid by means of voltage stability of the nodal admittance matrix in the s-domain
source converter technology (VSC), which play an impor- by using modal analysis are presented in [15]–[17], and
tant role in the transmission system development. However, in the frequency domain based on the generalized Nyquist
VSCs also bring new challenges and problems due to the criterion (GNC) in [18]–[20]. These studies use the nodal
admittance matrix modelling approach, an enhancement of
Manuscript received November 25, 2020; revised April 3, 2021 and June 4, impedance-based modelling methods, as it is simple and
2021; accepted June 15, 2021. Date of publication July 7, 2021; date of current
version August 10, 2021. This work was supported in part by the European powerful when characterizing multi-infeed large power sys-
Union’s Horizon 2020 research and innovation programme through the Marie tems. In [15], the stability of multiple-input and multiple-
Skłodowska-Curie under Grant 765585, in part by the FEDER/Ministerio de output (MIMO) systems is assessed by looking at the nodal
Ciencia, Innovación y Universidades–Agencia Estatal de Investigación under
Project RTI2018-095429-B-I00, and in part by the ICREA Academia program. admittance matrix poles; the contribution in [16] studies the
This article was recommended by Associate Editor F. M. Neri. (Corresponding zeros of the nodal admittance or loop impedance matrix deter-
author: Luis Orellana.) minant; the stability is assessed with the damping coefficient
Luis Orellana, Eduardo Prieto-Araujo, and Oriol Gomis-Bellmunt are with
the Centre d’Innovació Tecnològica en Convertidors Estàtics i Acciona- and the negative-resistance effect of the resonance modes
ments, Departament d’Enginyeria Elèctrica, Universitat Politècnica de in [17]; however, all mentioned studies are conducted in the
Catalunya, 08028 Barcelona, Spain (e-mail: [email protected]; eduardo. s-domain, and there is a preference in the industry to work
[email protected]; [email protected]).
Luis Sainz is with the Department of Electric Engineering, ETS with measurements in the frequency domain as it allows the
d’Enginyeria Industrial de Barcelona, Universitat Politecnica de Catalunya, use of black-box models.
08028 Barcelona, Spain (e-mail: [email protected]). Typically, the stability of the nodal admittance matrix is
Color versions of one or more figures in this article are available at
https://doi.org/10.1109/TCSI.2021.3091938. assessed in the frequency domain with the GNC which extends
Digital Object Identifier 10.1109/TCSI.2021.3091938 the Nyquist criterion for single-input and single-output (SISO)
1549-8328 © 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.

Authorized licensed use limited to: UNIVERSITAT POLITECNICA DE CATALUNYA. Downloaded on October 26,2023 at 18:54:39 UTC from IEEE Xplore. Restrictions apply.
 ORELLANA et al.: STABILITY ASSESSMENT FOR MULTI-INFEED GRID-CONNECTED VSCs 3759

to MIMO dynamic systems as introduced in [8]. For example,

the stability of MIMO systems is assessed with the GNC by
using the impedance-based approach of the nodal admittance
matrix form for a three-phase meshed and balance power
system in [18], for hybrid AC/DC grids in [19], and for
large-scale multiconverter systems in [20]. However, studies
in [21], [22] show that GNC may lead to a wrong stability
conclusion due to misleading associations at the time of
deriving the closed-loop transfer function. Further drawbacks
of using GNC were identified during the present study, such
as computational efforts (i.e., time and memory) at the time
of evaluating the open-loop in the [− j ∞, + j ∞] frequency
range in order to contour the unstable poles located in the
right half-plane (RHP) for high-order admittance matrices.
Additionally, the analysis of large networks is challenging due
to the numerous Nyquist curves of eigenvalues.
The resonance mode analysis (RMA), introduced in [23],
helps to identify harmonic resonance modes for systems in
the admittance matrix form; however, no stability criterion is
proposed. The letter in [24] uses the “peak-picking” method
and the “circle fit” method introduced in [25] to analyse
these resonance modes obtained from measurement data, yet
no contribution about addressing the stability beyond the
traditional criteria was made.
To address the above concerns in traditional frequency
domain stability criteria, this paper contributes with a new sta-
bility criterion, called as positive-mode damping (PMD) stabil-
ity criterion, which is based on the RMA and the PND stability
criteria. The features of the proposed criterion are summarized Fig. 1. Grid-connected VSC. (a) Schematic diagram control structure.
(b) Block diagram small-signal model.
as follows:
1) the oscillatory modes can be characterized in the fre-
commonly used closed-loop stability criteria in the frequency
quency domain;
domain.
2) it does not require detailed information;
3) the system stability can be assessed from experimental
measurements (e.g., black-box models); II. G RID -C ONNECTED VSC M ODELLING
4) it is not affected by aggregation of system elements; Fig. 1(a) displays the control structure of a generic
5) it is easy to use; grid-connected VSC. The model is an averaged three-phase
6) it requires less effort to calculate and evaluate than converter, which uses vector control strategy with a cascaded
traditional stability criteria; controller to control active and reactive power [26].
7) its application can be programmed. The dynamics of the VSC can be modelled by both
state-space and impedance-based modelling approaches [6],
The effectiveness of proposed stability criterion is tested in [7]. The small-signal model for stability studies at the point
two case studies implemented in Matlab/Simulink, comparing of common coupling (PCC), as displayed in Fig. 1(a), can be
its results with the corresponding eigenvalue analysis and the formulated to state-space equations as
GNC criterion application over the same system. The first
case study is built to demonstrate that the proposed criterion ẋ(t) = Ax(t) + Bu(t) u(t) = v qd (t)
provides a correct stability assessment while others methods y(t) = Cx(t) + Du(t) y(t) = i c−qd (t), (1)
fail. It is composed by three grouping options, where each
one studies the closed-loop stability of the system, in the where x(t), u(t) and y(t) are the states, input and output of
admittance matrix form, by diving it into two subsystems. the system state-space representation; and by a two-by-two
The instability condition is the same for all grouping options, impedance matrix, where each of its elements is a transfer
but the approach to build each of the subsystems is different function in the s-domain as follows
 
between them, leading to possible wrong stability assess- Z vsc−qq (s) Z vsc−qd (s)
v qd = i c−qd , (2)
ment conclusions. The second case has two examples in Z vsc−dq (s) Z vsc−dd (s)
order to show the simplicity of the proposed method for   
Zvsc (s)
assessing the stability of large and complex networks. The
visualisation of results and the computation effort for a large where v qd = [v q v d and i c−qd = [i c−q i c−d ]T .
]T
network is compared between the proposed criterion and A comparison between both small-signal modelling

Authorized licensed use limited to: UNIVERSITAT POLITECNICA DE CATALUNYA. Downloaded on October 26,2023 at 18:54:39 UTC from IEEE Xplore. Restrictions apply.
 3760 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS—I: REGULAR PAPERS, VOL. 68, NO. 9, SEPTEMBER 2021

by using eigenvalue analysis. On the other hand, the sta-

bility in impedance-based models is typically assessed with
closed-loop stability criteria, that studies the open-loop formed
by the ratio of the two subsystems’ impedance partitioned
accordingly to each criterion [10], [11]. The stability of
MIMO systems is typically assessed applying the GNC to the
open-loop [8].

A. Eigenvalue Analysis
Stability of the multi-infeed grid-connected VSCs in Fig. 2
can be studied from state-space equations in (1), where
u(t) = i n (t) and y(t) = v(t) are the input and output
variables of the system state-space. Stability can be addressed
Fig. 2. Schematic diagram of multi-infeed grid-connected VSCs. by obtaining the eigenvalues λi = σi ± j ωi of the state-space
matrix calculated as |A − λI| = 0. The stability criterion is
approaches in a local reference is described in [27] for built on the real and imaginary part: the real part σi = Re{λi }
further information. represents the system damping and the imaginary ωi = Im{λi }
When two or more VSCs are connected to the AC grid, the frequency of oscillation. The system is unstable if it con-
they can no longer be in a local reference. In other words, tains any eigenvalue in the RHP (i.e., σ0 > 0 means oscillatory
they all should be referenced to a reference or slack bus in instability of the frequency mode ω0 ) [32]. Considering (4),
the AC network, as detailed in [28] (see the Appendix for the expression in (1) can be rearranged as
more details). v = [C(sI − A)−1 B + D] i n , (5)
In the impedance-modelling approach, the converter   
ZT (s)
impedance can be easily added to the network nodal admit-
tance matrix by means of its admittance Yvsc (s) = (Zvsc (s))−1 where ZT (s) = YT (s)−1
is the impedance matrix transfer
as other YRL series or YC shunt connected elements by function which can be expressed from (5) in the following
applying the voltage node method, form
 −1   1 1
R + Ls ωL Cs Cω ZT (s) = C[adj(sI − A)]B + D = ZTb (s), (6)
YRL (s) = YC (s) = . |sI − A| D(s)
−ωL R + Ls −Cω Cs
(3) being adj(sI − A) the adjoint matrix of (sI − A) and ZTb (s)
the adjoint matrix of ZT (s).
III. M ULTI -I NFEED G RID -C ONNECTED VSCs M ODELLING It can be noted from (6) that the poles of ZT (s) are the roots
of the denominator D(s) = |sI − A|, namely the eigenvalues
Fig. 2 shows a schematic diagram used to repre-
λi of the state-space matrix A [33], [34]. Therefore, system
sent multi-infeed VSC-based AC grids as carried out in [15],
stability can be assessed by either the eigenvalues λi of A as
[19]–[21], [29]–[31]. The network is characterized by its
well as the poles of ZT (s).
admittance matrix YN (s), and the voltages and currents at its
Eigenvalue analysis is a simple stability criterion which
terminals (i.e., n the number of buses) are v = [v 1 . . . v n ]T
helps to analyse the stability of large systems in a short time.
and i = [i 1 . . . i n ]T . The external elements (e.g., VSC
However, it requires detailed information in order to model
converters) connected at the network buses are represented by
real systems, which sometimes is not available (e.g., control
their Norton equivalent circuits formed by the Norton currents
structure and parameters of power converters).
i n = [i n1 . . . . i nn ]T with their corresponding impedance
connected in parallel.
The relation between voltages and currents at the AC grid B. Generalized Nyquist Criterion
terminals is expressed as The expression in (4) can be rewritten as

i = YN (s)v v = (I + L(s))−1 ZN (s)i n L(s) = ZN (s)YS (s), (7)
i = (YN (s) + YS (s))v = YT (s)v, (4)
i = i n − YS (s)v n
where ZN (s) = (YN (s))−1 and I is an nth order identity
where YT (s) and YS (s) are the nodal admittance matrix matrix.
transfer functions of the full system and the external elements’ If the open-loop L(s) does not have any RHP poles,
impedance which can be easily obtained by applying the the stability of the closed-loop system in (7) can be assessed
voltage node method. by the GNC, which extends the traditional Nyquist criterion
for SISO systems to Nyquist curves of the eigenvalues of
IV. S TABILITY A SSESSMENT C RITERIA L(s) [8], [29]. This means that system stability is assessed by
A small-signal stability analysis can be performed using counting the clockwise encirclements of the eigenvalues λni of
both state-space and impedance-based dynamic models. L(s) around the (−1, j 0) point. This is valid for multi-infeed
In state-space representations, stability is commonly assessed grid-connected VSCs modelling in (7) because the network

Authorized licensed use limited to: UNIVERSITAT POLITECNICA DE CATALUNYA. Downloaded on October 26,2023 at 18:54:39 UTC from IEEE Xplore. Restrictions apply.
 ORELLANA et al.: STABILITY ASSESSMENT FOR MULTI-INFEED GRID-CONNECTED VSCs 3761

ZN (s) is passive and the external components in YS (s) are where the diagonal terms of Z are called modal
individual subsystems, which in stand-alone operation are impedances λzi .
stable (i.e., ZN (s) and YS (s) do not have any RHP poles). Parallel resonance phenomena is associated with the singu-
However, this might not be valid anymore if some of these larity of YT ( j ω) which happens when one of its eigenvalues
individual subsystems come from an aggregation of part of the λ yi approaches 0. The resonance modes can also be identified
network containing VSCs. This aggregation could be unstable from peaks values at the magnitude modal impedance |λzi |
due to the interaction of the grouped VSCs and network curves in the frequency domain [23].
passive components, introducing RHP poles in YS (s) [21]. It must be noted that the poles of ZT (s) (i.e., the eigenvalues
On the other hand, part of the drawbacks identified in eigen- of the state-space matrix A) are the same as the poles of the
value analysis, such as the required knowledge of the com- modal impedances of Z (s),
plete control structure, can be solved by using a GNC-based
1 1
stability analysis. Stability can be assessed from frequency Z (s) = TZTb (s)L = Zb (s)
dependent models provided by manufactures (e.g., converter D(s) D(s)
impedance curves) which are obtained from numeric simula- Zb (s) = TZTb (s)L = [λzb1 λzb2 . . . λzbn ]I. (11)
tion or experimental measurements. However, applying GNC,
Therefore, the stability of the system in (10) can be assessed
wrong stability conclusions might be made due to order-
cancellation, open-loop RHP poles, and improper minor-loop with the diagonal matrix Z , and the analysis can be carried
gain or impedance ratio have been identified in [21], [22]. out independently for each modal impedance λzi as a SISO
system by applying the PND stability criterion to each λzi [13].
These modal impedances can be expressed as,
V. P OSITIVE M ODE DAMPING C RITERION
It is well-known that instabilities are related to low-damped λzbi ( j ωx )
λzi ( j ωx ) =
network resonances. This has been proved for a single 
ip
grid-connected VSC with the PND stability criterion in [8], ( j ωx − pi )( j ωx − pi∗ )
i=1
[13], [14] which evaluates the damping of the SISO transfer
G( j ωx ) G( j ωx )
function Z T (s) at resonance frequencies. It is stated that a = ∗ = 2 , (12)
system is stable if and only if the damping is positive at ( j ωx − p0 )( j ωx − p0 ) σ0 + ω0 − ω2x − j 2σ0 ωx
2

these resonance frequencies, i.e., Re{Z T ( j ωx )} > 0. It is where p0 = σ0 ± j ω0 is pair of complex conjugate poles of
also worth mentioning that the PND stability criterion in [13] λzi corresponding to a certain system oscillatory mode which
evaluates the closed-loop function (i.e., Z T ( j ωx )) to address match with eigenvalues λ0 of the state-space matrix A, and
the stability; therefore, it is not affected by misleading asso- G( j ωx ) is a polynomial expression representing the rest of
ciations of the system elements as it is the case of the the terms of λzi .
Nyquist criterion, which evaluates the open-loop function [21], It can be observed that the modal impedance in (12) will
[22]. The proposed stability approach, called positive-mode be maximum or have a peak value at the oscillation frequency
damping (PMD) stability criterion, extends the PND stability (i.e., ωx ≈ ω0 ) in the case of a poorly damped oscillatory
criterion to multi-infeed grid-connected VSCs by means of the mode (i.e., |σ0 |  |ω0 |), which is the main concern in
RMA. academia and industry due to the following reasons: (a) a
The RMA provides an effective tool for evaluating the power system maintains stable operation for strongly damped
resonances of networks modelled in the admittance matrix modes with large negative σi ; and (b) monotonic instability
form [23], by addressing the statement in (4) as caused by large positive σi occurs less often in power sys-
v = YT ( j ωx )i n YT ( j ωx ) = LY T, (8) tems. In these cases, the growing oscillations caused by large
positive σ0 are sustained due to saturation and limiters non-
where YT ( j ωx ) is the system admittance matrix at frequency linearities [4], [5].
ωx ; v and i n are the nodal voltage and current injection vectors; If ωx is within the small neighbourhood of ω0 , G( j ωx ) ≈
L and T are the right and left eigenvector matrices; and Y G( j ω0 ) = G r + j G x where G r and G x are constant complex
is the diagonal eigenvalue matrix, numbers dependent on ω0 [30]. Thus, λzi can be further
⎡ ⎤
λ y1 0 . . . 0 expressed as
⎢ 0 λ y2 . . . 0 ⎥
⎢ ⎥ σ02 + ω02 − ω2x + j 2ωx σ02
⎢ . . . . ⎥ λzi ( j ωx ) = (G r + j G x )
Y = ⎢ ⎢ ⎥. (9)
⎢ . . . . ⎥⎥
(σ02 + ω02 − ω2x )2 + (2σ0 ωx )2
⎣ . . . . ⎦ = λzi,r (ωx ) + j λzi,x (ωx ) (13)
0 0 . . . λ yn
where
It must be noted that the inverse diagonal eigenvalue matrix
Y is the diagonal eigenvalue matrix of ZT ( j ωx ) in (5), (σ02 + ω02 − ω2x )G r − 2ωx σ02 G x
λzi,r (ωx ) =
ZT ( j ωx ) = (YT ( j ωx ))−1 = LZ T (σ02 + ω02 − ω2x )2 + (2σ0 ωx )2
1 (σ02 + ω02 − ω2x )G x + 2ωx σ02 G r
Z = [λz1 λz2 . . . λzn ]I λzi = , (10) λzi,x (ωx ) = . (14)
λ yi (σ02 + ω02 − ω2x )2 + (2σ0 ωx )2

Authorized licensed use limited to: UNIVERSITAT POLITECNICA DE CATALUNYA. Downloaded on October 26,2023 at 18:54:39 UTC from IEEE Xplore. Restrictions apply.
 3762 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS—I: REGULAR PAPERS, VOL. 68, NO. 9, SEPTEMBER 2021

The oscillatory resonance occurs at zero-crossing frequen-

cies of λzi,x , i.e., λzi,x (ωx ) = 0,

2G r σ0 ± (2G r σ0 )2 + 4(σ02 + ω02 )G 2x
ωx1,x2 = , (15)
2G x
where the feasible solutions correspond to positive zero-
crossing frequency values with the largest magnitude [30].
In case of poorly damped oscillatory modes where
|σ0 |  |ω0 |, it implies that ωx approximately matches with
the frequency of the oscillatory mode ω0 , i.e., ωx ≈ ω0 . Thus,
the real part of λzi at ωx can be approximated as
−2G x ω0 σ0 −G x
λz0r (ωx ≈ ω0 ) ≈ = = k x σ0 , (16)
2ω0 σ0 (2ω0 σ0 )2
where k x is the slope of λz0x at ωx ≈ ω0 , i.e.,
 
∂λzi x (ω)
kx =
∂ω ω=ωx
−8ω3x G x σ02 − 8ω2x G r σ03 −G x
≈ ≈ . (17)
16ω04 σ04 2ω0 σ02
According to the above, the PMD stability criterion is
summarized as follows,
PMD stability criterion: multi-infeed grid-connected VSCs
systems are stable (i.e., σ0 < 0) if and only if,
(i) k x > 0 and λzi,r < 0; or (ii) k x > 0 and λzi,r < 0
at resonant frequencies ωx for all local maximums or peak
values of |λzi ( j ω)| (i = 1 to n).
The condition k x > 0 indicates that λzi,x (ω) passes through
zero-axis at ωx from a capacitive area to an inductive area;
and the condition k x < 0 means that λzi,x (ω) passes through
zero-axis at ωx from an inductive area to a capacitive area.
The second condition usually occurs at peak resonance
points for inductive (i.e., positive λzi,x (ω) values which
increase in line with the frequency, j ωL) and capacitive
(i.e., negative λzi,x (ω) values which decrease as long as the
frequency increases, − j/(ωC)) behaviour, which is associated
Fig. 3. Testing networks single-line diagrams. (a) Case study I.
to physical elements in conventional power systems. However, (b) Case study II(a). (c) Case study II(b).
the control structure of power converter can also produce
different inductive and capacitive behaviour (i.e., influence in parallel to a grid-equivalent impedance as shown
of the outer loops and the PLL) which is not related to any Fig. 3(a).
physical element of the system as observed in [35] for PMSG • Case study II addresses the issue of assessing the stability
based wind farms in weak AC networks in the subsynchronous of a large power system in the frequency domain by
frequency range. In this case, both conditions, (i) and (ii), studying two networks.
might be considered for σ0 < 0 at the peak resonance points. (a) The testing network of case study I is taken a step
In the harmonic range as studied in Section III, the imagi- ahead by completing the string configuration with a 2 km
nary part of the VSC output impedance is not strongly affected cable between converters. An additional VSC is also
by the control structure and keeps the inductive behaviour connected in string as displayed in Fig. 3(b).
produced by its filter, L c . In this case, the condition (ii) might (b) A larger and more complex system than previous
happen for σ0 < 0 at the peak resonance points, which is the study networks as the modified IEEE 14 bus system
usual case in traditional electrical power systems. (Fig. 3(c)) introduced in [36] is used to complete the
study.
VI. C ASE S TUDY
The construction of the nodal admittance matrix closed-loop
The previously described stability criteria is tested in two of case study I and II(a) is derived in this work, and the
study cases: dynamic models are verified in the s-domain by comparing the
• Case study I studies the effect of misleading association poles and zeros of the impedance matrix, ZT (s) = (YT (s))−1 ,
when dividing the system to assess the closed-loop stabil- with the eigenvalues of linear state-space models. The
ity. The study network consists of two VSCs connected state-space models are validated with time domain simulations

Authorized licensed use limited to: UNIVERSITAT POLITECNICA DE CATALUNYA. Downloaded on October 26,2023 at 18:54:39 UTC from IEEE Xplore. Restrictions apply.
 ORELLANA et al.: STABILITY ASSESSMENT FOR MULTI-INFEED GRID-CONNECTED VSCs 3763

Fig. 4. Case study I, grouping options. (a) GO1. (b) GO2. (c) GO3.

TABLE I where there is a good match with non-linear Simulink and

S YSTEM PARAMETERS linearized state-space models for stable (i.e., figures are not
included for the sake of space) and unstable conditions
in Fig 5(a) and Fig 5(b). Instability in the system occurs in the
time domain simulation at approximately 1190 Hz when the
time delay of VSC2 increases up to 0.5 times τsw as displayed
Fig 5(a).
1) Grouping Option 1: The nodal admittance matrix in
the grouping option 1 (GO1) is constructed according to
Section III and displayed in Fig. 4(a). The grid equivalent
is connected at bus 1, and two VSCs at buses 2 and 3, respec-
tively. The network passive elements are grouped in YN and
by comparing the results obtained with the ones of non-linear external elements which can cause instability by interacting
Simulink models. No state-space model was developed for with resonances of the system in YS (i.e., no open-loop
case study II(b) because the process can be long and complex. RHP poles). The system closed-loop as expressed in (7) is
Nevertheless, this example allows to verify the usefulness of conformed by
the proposed stability criterion in large networks and compare ⎡ ⎤−1
its performance to the GNC. Ytl1 + Ytl2 − Ytl1 − Ytl2
The stability assessment with the PMD stability crite- ZN = ⎣ −Ytl1 Ytl1 + Ycc1 02×2 ⎦ (18)
rion is compared with the results of eigenvalue analysis −Ytl2 02×2 Ytl2 + Ycc2
and the GNC in linearized state-space and impedance-based
and
Matlab models for case study I and II(a), and only in an ⎡ ⎤
impedance-based model for case study II(b). The unstable res- Yg 02×2 02×2
onance modes oscillation frequency of linear models is further YS = ⎣02×2 Yvsc1 02×2 ⎦ , (19)
verified for all study cases with time domain simulations of 02×2 02×2 Yvsc2
non-linear Simulink models. The system (complemented with
data from [37]) and control parameters for both networks can where Ytl1 , Ytl2 , Ycc1 , Ycc2 , Yvsc1 , Yvsc2 and Yg are 2 by
be found in Table I. 2 matrices, which can be expressed in the frequency domain
The time delay is calculated with the expression or s-domain as in (3). The inputs and outputs of the sys-
τ f d = qd τsw as described in [8]. A qd = 0.25 was initially tem are i n = [i ng−q i ng−d 02×1 02×1 ]T and v =
considered for all converters in all study cases. The switching [v t mv−q v t mv−d v tl1−q v tl1−d v tl2−q v tl2−d ]T
period τsw = 1/ f sw is determined for fsw = 2 kHz. respectively.
Instability happens when the time delay is modified (i.e., In Fig. 5, the stability is assessed with s-domain and
varying qd ) [38]. frequency domain stability criteria. A pair complex conjugate
poles in the RHP can be noticed in Fig. 5(b) at f 0 = 1192 Hz
(i.e., ω0 = 7488 = 2π f 0 ) by evaluating the system impedance
A. Case Study I matrix ZT (s). These poles match with the eigenvalues of the
The testing network is a 3 bus system with 2 converters state-space representation of ZT as described in the eigenvalue
(i.e., VSC1 and and VSC2) connected to a network equivalent analysis section. The instability can be further confirmed
impedance (Fig 3(a)). The system stability is assessed for in Fig. 5(c), where the λn3 Nyquist curve of L encircles the
three grouping options. Each of them associates some elements (−1, j 0) point in the clockwise direction. It is worth men-
in ZN and YS of the closed-loop function with a different tioning that other Nyquist curves such as λn6 seem to encircle
approach as displayed in Fig 4. The nodal admittance matrix (−1, j 0) but by zooming around the critical point no encir-
model has been verified in the time domain and the s-domain, clement was observed.

Authorized licensed use limited to: UNIVERSITAT POLITECNICA DE CATALUNYA. Downloaded on October 26,2023 at 18:54:39 UTC from IEEE Xplore. Restrictions apply.
 3764 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS—I: REGULAR PAPERS, VOL. 68, NO. 9, SEPTEMBER 2021

Fig. 5. Stability assessment GO1. (a) Time domain simulation. (b) Eigenvalue analysis. (c) GNC.

and
 
Yg + Ybvsc2 02×2
YS = , (21)
02×2 Yvsc1
where the inputs and outputs of the system are i n =
[i ng−q i ng−d 02×1 ]T and v = [v t mv−q v t mv−d
v tl1−q v tl1−d ]T .
The stability of GO2 is assessed with the GNC in Fig. 7(a).
The λn3 Nyquist curve encircles the (−1, j 0) point two
times but in counterclockwise direction. On the other hand,
a closed-up view around the critical point shows that λn2 is not
enclosing it. In GO1, the system was identified as unstable for
all three criteria, but the GNC criterion fails to predict stability
in this grouping option. This is caused by the open-loop RHP
poles introduced by Yvsc2 b in YS .
The PMD stability criterion is tested in GO2 as displayed
Fig. 6. Stability assessment GO1, PMD stability criterion. in Fig. 7(b). There is a peak in one of the magnitude of the
eigenvalues curves |λz2 | at 1192 Hz where it matches with its
negative real part, Re{λz2 }, as it was obtained in GO1. The
The stability assessment with the PMD stability criterion is instability was once again confirmed by the PMD stability
displayed in Fig. 6. The modal impedance magnitude curve criterion, and the result was not affected by the RHP poles of
λz5 in the frequency domain has a peak at 1192 [Hz] where YS (s). The PMD stability criterion does not fail in the stability
its real part is negative, confirming once more the instability assessment of the GO2 because the stability criteria is applied
of ZT . The stability assessment in GO1 agrees for all stability to the closed-loop transfer function of the system.
criteria. 3) Grouping Option 3: The instability in the grouping
2) Grouping Option 2: The grouping option 2 (GO2) stud- option 3 (GO3) is caused by VSC2 as in previous grouping
ies the same network under the same instability conditions options, but the elements Ycc1 and Ytl1 are grouped into Yvsc1
as GO1. However, it merges Ycc2 and Ytl2 into Yvsc2 as as displayed in Fig. 4(c). The order of the system matrix is
displayed in Fig. 4(b). The new converter admittance is also reduced, and the closed-loop is composed by
b
Yvsc2 = (Zbvsc2 )−1 = [(Zvsc2 //Zcc2 ) + Ztl2 ]−1 which enables  −1
the possibility of interaction between the converter controller Ytl2 −Ytl2
ZN = (22)
with resonant circuits from the network (e.g., parallel res- −Ytl2 Ytl2 + Ycc2
onances caused by the transformer inductance Ztl2 and the and
shunt capacitor of the converter filter Zcc2 ) within Yvsc2 b
.  
Yg + Ybvsc1 02×2
In consequence, the system matrix order is reduced (i.e., YS = , (23)
02×2 Yvsc2
the number of buses of the network is reduced from three
to two), and the closed-loop is conformed by b
where Yvsc1 = (Zbvsc1 )−1 = [(Zvsc1//Zcc1 ) + Ytl1 ]−1 ,
 −1 and the inputs and outputs of the system are i n =
Ytl1 −Ytl1 [i ng−q i ng−d 02×1 ]T and v = [v t mv−q v t mv−d
ZN = (20)
−Ytl1 Ytl1 + Ycc1 v tl2−q v tl2−d ]T .

Authorized licensed use limited to: UNIVERSITAT POLITECNICA DE CATALUNYA. Downloaded on October 26,2023 at 18:54:39 UTC from IEEE Xplore. Restrictions apply.
 ORELLANA et al.: STABILITY ASSESSMENT FOR MULTI-INFEED GRID-CONNECTED VSCs 3765

Fig. 7. Stability assessment GO2. (a) GNC. (b) PMD stability criterion. Fig. 8. Stability assessment GO3. (a) GNC. (b) PMD stability criterion.

When addressing the stability of ZT by evaluating the L more challenging due to the number of eigenvalue Nyquist
with the GNC in Fig. 8(a), the system instability was identified curves.
as GO1. The λn2 Nyquist curve encircles the critical point 1) Example (a): The network of case study I is extended to
(−1, j 0) in the clockwise direction. In GO3, no open-loop add more complexity to the study. This network is composed
RHP poles were observed in comparison to GO2, because by three VSCs connected in string configuration to a grid
VSC2 is not merged with any grid component and the insta- equivalent, as displayed in Fig. 3(b).
bility occurs due to the interaction between VSC2 and the The system is a 7 bus network where the grid equivalent is
grid. connected at bus 1, and the three converters are connected at
GO3 was also assessed with the PMD stability criterion buses 3, 5 and 7, respectively, as illustrated in Fig. 9(a). Each
in Fig. 8(b). The assessment matches the GNC one. Again, converter is separated from another by a 2 km cable modelled
the instability was found at 1192 Hz as GO1 and GO2, where with a PI section equivalent Ycli , which is composed by a RL
|λz2 | in the frequency domain has a peak where its Re{λz2 } is section Yrli and a shunt capacitor Yccli on both ends. The
negative. elements of the network are associated as stated in Section III
This case study proves the effectiveness of the proposed to construct the closed-loop representation of ZT . The matrices
stability criterion PMD and the methodology in Section III. ZN and YS are detailed in (24) and (25), as shown at the
The malpractice at the moment of associating the elements bottom of the next page, where Yrl1−2 = Yrl1 + Yrl2 ;
of the network into two subsystems in order to study the Yccl1−2 = Yccl1 + Yccl2 ; Yrl2−3 = Yrl2 + Yrl3 ;
closed-loop stability of ZT could lead to misleading stability Yccl2−3 = Yccl2 + Yccl3 ; and the inputs and outputs
conclusions during the application of the GNC. are i n = [i ng−q i ng−d 01×2 01×2 01×2 ]T and v =
[v t mv−q v t mv−d v tl1−q v tl1−d v tl2−q v tl2−d 
v tl3−q v tl3−d ]T .
B. Case Study II The dynamic model has been verified with time domain
The effectiveness of the proposed method can be fur- simulations and in the s-domain, where there is a good agree-
ther tested in two larger networks than the one from case ment with the results of non-linear Simulink and linearized
study I, example (a) and (b), where the GNC becomes state-space models as displayed in Fig 9(b) and Fig 9(c).

Authorized licensed use limited to: UNIVERSITAT POLITECNICA DE CATALUNYA. Downloaded on October 26,2023 at 18:54:39 UTC from IEEE Xplore. Restrictions apply.
 3766 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS—I: REGULAR PAPERS, VOL. 68, NO. 9, SEPTEMBER 2021

Fig. 9. Stability assessment case study II(a). (a) ZN & YS . (b) Time domain simulation. (c) Eigenvalue analysis. (d) GNC. (e) PMD stability criterion.

In this example, the instability of the system is caused by match the eigenvalues of the system state-space matrix. The
VSC2 when the τ f d is increased up to qd = 0.45 times τsw . instability is confirmed in the frequency domain in Fig. 9(d),
The stability of ZT is assessed in Fig. 9 with frequency and where the λn10 Nyquist curve of L encircles the critical point
s-domain criteria. In Fig. 9(c), ZT (s) is evaluated in the (−1, j 0) in the clockwise direction. It is worth mentioning that
s-domain, and a pair complex conjugate poles in the RHP can assessing stability was challenging and time consuming due
be noticed at f 0 = 1192 Hz (i.e., ω0 = 7487 = 2π f0 ) which to the number of eigenvalues to be studied. For instance, λn9

⎡ ⎤−1
Yrl1 + Yccl1 − Yrl1 02×2 02×2 02×2 02×2 02×2
⎢ −Yrl1 Ytl1 + Yrl1−2 + Yccl1−2 − Ytl1 − Yrl2 02×2 − Ytl1 02×2 ⎥
⎢ ⎥
⎢ 02×2 − Ytl1 Ytl1 + Ycc1 02×2 02×2 02×2 02×2 ⎥
⎢ ⎥
ZN = ⎢
⎢ 02×2 − Yrl2 02×2 Ytl2 + Yrl2−3 + Yccl2−3 − Ytl2 − Yrl3 02×2 ⎥
⎥ (24)
⎢ 02×2 02×2 02×2 − Ytl2 Ytl2 + Ycc2 02×2 02×2 ⎥
⎢ ⎥
⎣ 02×2 02×2 02×2 − Yrl3 02×2 Ytl3 + Yrl3 + Yccl3 − Ytl3 ⎦
02×2 02×2 02×2 02×2 02×2 − Ytl2 Ytl3 + Ycc3
⎡ ⎤
Yg 02×2 02×2 02×2 02×2 02×2 02×2
⎢02×2 02×2 02×2 02×2 02×2 02×2 02×2 ⎥
⎢ ⎥
⎢02×2 02×2 Yvsc1 02×2 02×2 02×2 02×2 ⎥
⎢ ⎥
YS = ⎢
⎢02×2 02×2 02×2 02×2 02×2 02×2 02×2 ⎥ ⎥ (25)
⎢02×2 02×2 02×2 02×2 Yvsc2 02×2 02×2 ⎥
⎢ ⎥
⎣02×2 02×2 02×2 02×2 02×2 02×2 02×2 ⎦
02×2 02×2 02×2 02×2 02×2 02×2 Yvsc3

Authorized licensed use limited to: UNIVERSITAT POLITECNICA DE CATALUNYA. Downloaded on October 26,2023 at 18:54:39 UTC from IEEE Xplore. Restrictions apply.
 ORELLANA et al.: STABILITY ASSESSMENT FOR MULTI-INFEED GRID-CONNECTED VSCs 3767

Fig. 10. Stability assessment case study II(b) GO1. (a) ZN & YS , (b) Time domain simulation. (c) Eigenvalue analysis. (d) GNC.

encircles the critical point but in counterclockwise direction,

and other eigenvalue Nyquist curves λn11 to λn14 follow a
similar trajectory as λn10 , but they do not encircle the critical
point.
The stability assessment with the PMD stability criterion
is displayed in Fig. 9(e). The magnitude of λz13 curve in the
frequency domain has a peak at 1192 [Hz] where its real part
is negative, confirming ZT instability. It can also be spotted
a mode at 1441 Hz with larger magnitude than the unstable
mode. However, its real part is positive which makes it stable.
This larger magnitude means that the real part is close to
zero and this eigenvalue may be a candidate to lead system to
instability.

2) Example (b): The network size and complexity is further

extended in this example. A commonly used network in the
literature as the IEEE 14 modified bus system (Fig 3(c))
introduced in [36] is used. The network is modelled according
to the reference, with the exception of filters, the converter, Fig. 11. Stability assessment case study II(b) GO1, PMD stability criterion.
and the SVC. The filters are not modelled because they do not
contribute significantly to the main resonance modes according in the system happens when the VSC time delay increases up
to [23]. The converter and the SVC (Static Var Compensator), to 0.6 times the switching period of VSC1.
which are connected to bus 3 and 8, are modelled as VSCs and There are two grouping options in this example. First,
scaled according to the reference data. The system instability in GO1, the system nodal admittance matrix is constructed as
Authorized licensed use limited to: UNIVERSITAT POLITECNICA DE CATALUNYA. Downloaded on October 26,2023 at 18:54:39 UTC from IEEE Xplore. Restrictions apply.
 3768 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS—I: REGULAR PAPERS, VOL. 68, NO. 9, SEPTEMBER 2021

Fig. 12. Stability assessment case study II(b) GO2. (a) ZN & YS , (b) GNC all eigenvalues. (c) GNC λn13 .

described in Section III. Then, in GO2, some system elements has a peak at 888 Hz in magnitude and the real part is
are associated in order to cause RHP in the open loop, which negative.
is a condition for the GNC to fail. b) Grouping option 2: In the GO2, the system compo-
a) Grouping option 1: In the GO1, the network is built nents are associated in order to cause RHP in the open-loop
according to Section III (i.e., passive components in ZN and L as shown in Fig 12(a). VSC1, the filter Zcc1 , and the
external elements to avoid interaction with resonances of the transformer Ytl1 , connected at bus bar 15, are merged into
system in YS as displayed in Fig 10(a). According to [36], b
Yvsc1 = (Zbvsc1 )−1 = [Zvsc1 //(Zcc1 + Ytl1 )]−1 . The bus bar
VSC1 is connected to bus 3 through a transformer; therefore, b
15 is no longer needed (i.e., Yvsc1 is directly connected to bus
a new bus, 15, is added for this purpose. 3); therefore, the ZT order is reduced from 30 to 28.
In the time domain simulation, the non-linear Simulink The stability of ZT for GO2 is assessed in the fre-
model becomes unstable when T f d = 0.6 Tsw of VSC1 as quency domain with the GNC and PMD stability criterion.
displayed in Fig 10(b). The oscillatory unstable resonance is In Fig 12(b), the large number of Nyquist curves makes it
graphically determined at approximately 885 Hz. The insta- difficult to address the stability straight away. For example,
bility is verified with the pole-zero plot of the linear ZT (s) the Nyquist curve, λn13 , encircles the critical point after
in the s-domain in Fig 10(c), where there is a pair of poles making many turns close to the critical point, but in counter-
in the RHP with an oscillating frequency of 888 Hz. In the clockwise direction as displayed in Fig 12(c). The GNC in this
frequency domain, the instability is corroborated with the case fails to predict the stability. On the other hand, in Fig 13,
GNC where λn9 curve encircles the critical point in clockwise the PMD stability criterion easily predicts the stability at the
direction as shown in Fig 10(d). Similarly, the proposed point where there is a peak at 888 Hz in the magnitude of one
PMD stability criterion predicts the stability in Fig 11 at the eigenvalues magnitude of ZT ( j ω), λz17 , and the real part
the point where one of the eigenvalues of ZT ( j ω), λz20 , of it is negative.

Authorized licensed use limited to: UNIVERSITAT POLITECNICA DE CATALUNYA. Downloaded on October 26,2023 at 18:54:39 UTC from IEEE Xplore. Restrictions apply.
 ORELLANA et al.: STABILITY ASSESSMENT FOR MULTI-INFEED GRID-CONNECTED VSCs 3769

according to state-space and impedance-based modelling tech-

niques. Finally, the stability assessment results are com-
pared with those from the eigenvalue analysis and the GNC
criterion.
This criterion has proven to be a powerful tool compared to
eigenvalue analysis and the GNC, because it has the following
relevant characteristics for a stability criterion: (i) frequency
characterization of the closed-loop unstable oscillatory modes;
(ii) does not require detailed information (e.g., use of measure-
ments from black-box models); (iii) not sensitive to associa-
tions of systems elements; (iv) simple to evaluate and less
computation effort (i.e., short computation time and memory
usage); and (v) visually friendly and physical interpretation of
results.
The PMD stability criterion has been successfully tested
when addressing the stability of large power system with
high-penetration of power electronics converters. However,
some future work was identified during the present work
such as:
Fig. 13. Stability assessment case study II(b) GO2, PMD stability criterion. • working with experimental measurements and black-box
models;
These are the findings of applying the PMD stability crite- • examining the possibility to define stability margins;
rion to large and complex networks. • determining the influence of the operation point in lin-
• The instability and the resonance mode frequency can earized models over the stability;
be easily identified through eigenvalue |λzi ( j ω)| and • studying and quantifying the power system nodes contri-
Re{λzi ( j ω)} curves for high-order ZT ( j ω) matrices com- bution over the unstable resonance modes;
pared to GNC. They provide a simpler visualisation • and identifying how the network elements of the system
of the physical interpretation of oscillatory modes and affect the instability.
instabilities.
• The PMD stability criterion works with the complete
nodal admittance matrix YT ( j ω), this is the reason why A PPENDIX
it is not affected by the aggregation of system elements A VSC M ODELLING
in comparison with the GNC, which only studies the The grid-connected VSC small-signal model displayed
open-loop L. in Fig. 1 can be formulated as:
• The fact of working in the frequency domain allows to
use black-box models or models based in measurements
which can be added to YT ( j ω) in the frequency domain. A. Outer Loop Controller
• Computation effort has also been reduced, because The outer loop controls active power with the q-component
c
(i re
for the PMD stability criterion it is only needed to f −q ) and reactive power with the d-component
c
(i re
plot the frequency range of interest. On the other f −d ). The current references can be defined as:
hand, the GNC requires a higher computation effort i re
c
f −q = −Folp p and i re f −d = −Folq q , where
c c c

(e.g., store larger matrix arrays in memory and longer Folp = k p−olp + ki−olp /s and Folq = k p−olq + ki−olq /s. The
processing time) by plotting a wider frequency range small signal active power (pc ) and reactive power (q c ) are
{ω ∈ R | −∞ < ω < +∞}. the following,
3 c
VII. C ONCLUSION pc = [(i v c + v 0−q
c c
i c−q c
+ i c0−d v dc
2 c0−q q
The present work contributes with the PMD stability cri- c
+ v 0−d c
i c−d )] (26)
terion to address the drawbacks found in commonly used 3
stability criteria to study multi-infeed grid-connected VSCs. q c = [(i c0−d
c
v qc − v 0−d
c
i c−q
c
− i c0−q
c
v dc
2
The PMD stability criterion can only be used in power systems + v 0−q
c
i c−d
c
)], (27)
which satisfy the condition σ0  ω0 , which is the main
concern in power system stability studies. where v 0−q
c
, v 0−d
c
, i c0−q
c
, i c0−d
c
are voltages and currents at the
The article starts by testing the effectiveness of the PMD linearization point.
stability criterion on a three-bus simple network with two
grid-connected VSCs. The complexity of the testing net-
works is then increased from a seven-bus network with three B. Inner Loop Controller
grid-connected VSCs to a 14 bus system with two VSCs The small-signal voltage modulated by the converter
c
(v c−re
connected. The networks are linearized in Matlab/Simulink f ) and the voltage measurement (V ) can be

Authorized licensed use limited to: UNIVERSITAT POLITECNICA DE CATALUNYA. Downloaded on October 26,2023 at 18:54:39 UTC from IEEE Xplore. Restrictions apply.
 3770 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS—I: REGULAR PAPERS, VOL. 68, NO. 9, SEPTEMBER 2021

expressed like [3] J. Sun et al., “Renewable energy transmission by HVDC across the
 c   c   c  continent: System challenges and opportunities,” CSEE J. Power Energy
vre f −q v h−q i re f −q Syst., vol. 3, no. 4, pp. 353–364, Dec. 2017.
= − F il
vre
c
f −d v h−d
c i re
c
f −d
[4] S. Shah, P. Koralewicz, V. Gevorgian, and R. Wallen, “Large-signal
  c  impedance modeling of three-phase voltage source converters,” in Proc.
Fil −ωL c i c−q 44th Annu. Conf. IEEE Ind. Electron. Soc. (IECON). Piscataway, NJ,
+ (28) USA: IEEE, Oct. 2018, pp. 4033–4038.
ωL c Fil i c−d
c
[5] T. Wu, Q. Jiang, J. Shair, H. Mao, and X. Xie, “Inclusion of current
limiter nonlinearity in the characteristic analysis of sustained subsyn-
and
       chronous oscillations in grid-connected PMSGs,” IEEE Trans. Energy
v q v c−q R + Lcs ωL c i c−q Convers., early access, Dec. 16, 2020, doi: 10.1109/TEC.2020.3045296.
= + c , [6] E. A. A. Coelho, P. C. Cortizo, and P. F. D. Garcia, “Small-signal stabil-
v d v c−d −ωL c Rc + L c s i c−d ity for parallel-connected inverters in stand-alone AC supply systems,”
(29) IEEE Trans. Ind. Appl., vol. 38, no. 2, pp. 533–542, 2002.
[7] L. Harnefors, M. Bongiorno, and S. Lundberg, “Input-admittance cal-
where [v h−qc v h−d
c ]T = H [v c v c ]T and [v c
v q d c−q
culation and shaping for controlled voltage-source converters,” IEEE
Trans. Ind. Electron., vol. 54, no. 6, pp. 3323–3334, Dec. 2007.
v c−d
c
]T = FD [vre c
f −q v re f −d ]. Hv and FD are the
c
[8] L. Harnefors, “Modeling of three-phase dynamic systems using complex
first-order feed-forward filter and the fifth-order Padé approx- transfer functions and transfer matrices,” IEEE Trans. Ind. Electron.,
imant delay, respectively. The PI controller Fil = k p−il + vol. 54, no. 4, pp. 2239–2248, Aug. 2007.
[9] L. Harnefors, L. Zhang, and M. Bongiorno, “Frequency-domain
Lc Rc
ki−il /s and its gains are k p−il = and ki−il = [39]. passivity-based current controller design,” IET Power Electron., vol. 1,
τil τil no. 4, p. 455, 2008.
[10] J. Sun, “Impedance-based stability criterion for grid-connected invert-
ers,” IEEE Trans. Power Electron., vol. 26, no. 11, pp. 3075–3078,
C. Phase-Locked Loop Nov. 2011.
The small-signal angle can be obtained with the following [11] M. Cespedes and J. Sun, “Impedance modeling and analysis of grid-
expression θ = −Fpll v dc , where Fpll = k p− pll + ki− pll /s. connected voltage-source converters,” IEEE Trans. Power Electron.,
vol. 29, no. 3, pp. 1254–1261, Mar. 2014.
The rotation from a local to a global reference (T−1 qd ) and [12] X. Zhang, X. Ruan, and C. K. Tse, “Impedance-based local stability
vice-versa (Tvqd ) can be modelled as criterion for DC distributed power systems,” IEEE Trans. Circuits Syst. I,
 c     Reg. Papers, vol. 62, no. 3, pp. 916–925, Mar. 2015.
v c−q cos θ0 − sin θ0 v c−q [13] L. Sainz, M. Cheah-Mane, L. Monjo, J. Liang, and O. Gomis-Bellmunt,
= “Positive-net-damping stability criterion in grid-connected VSC sys-
v c−d
c sin θ0 cos θ0 v c−d
  tems,” IEEE J. Emerg. Sel. Topics Power Electron., vol. 5, no. 4,
−v c−q0 sin θ0 − v c−d0 cos θ0 pp. 1499–1512, Dec. 2017.
+ θ [14] M. Cheah-Mane, L. Sainz, J. Liang, N. Jenkins, and C. E. Ugalde-Loo,
v c−q0 cos θ0 − v c−d0 sin θ0 “Criterion for the electrical resonance stability of offshore wind power
(30) plants connected through HVDC links,” IEEE Trans. Power Syst.,
vol. 32, no. 6, pp. 4579–4589, Nov. 2017.
and [15] E. Ebrahimzadeh, F. Blaabjerg, X. Wang, and C. L. Bak, “Harmonic
     stability and resonance analysis in large PMSG-based wind power
v q cos θ0 sin θ0 v qc plants,” IEEE Trans. Sustain. Energy, vol. 9, no. 1, pp. 12–23, Jan. 2018.
=
v d − sin θ0 cos θ0 v dc [16] Y. Zhan, X. Xie, H. Liu, H. Liu, and Y. Li, “Frequency-domain
  modal analysis of the oscillatory stability of power systems with high-
−v q0
c
sin θ0 + v d0
c
cos θ0 penetration renewables,” IEEE Trans. Sustain. Energy, vol. 10, no. 3,
+ θ, (31) pp. 1534–1543, Jul. 2019.
−v q0
c
cos θ0 − v d0
c
sin θ0 [17] F. Xing, Z. Xu, Z. Zhang, Y. Dan, and Y. Zhu, “Resonance stability
analysis of large-scale wind power bases with type-IV wind generators,”
where θ = θ0 − θ0c (i.e., difference between the Energies, vol. 13, no. 19, p. 5220, Oct. 2020. [Online]. Available:
small-signal angle in the global reference and the angle in https://www.mdpi.com/1996-1073/13/19/5220
the local reference) [28]. The expression in (30) is also used [18] X. Wang, F. Blaabjerg, and W. Wu, “Modeling and analysis of harmonic
stability in an AC power-electronics-based power system,” IEEE Trans.
to change the converter current (Tiqd ) from a global to a local Power Electron., vol. 29, no. 12, pp. 6421–6432, Dec. 2014.
reference. [19] J. Pedra, L. Sainz, and L. Monjo, “Three-port small signal admittance-
based model of VSCs for studies of multi-terminal HVDC hybrid
AC/DC transmission grids,” IEEE Trans. Power Syst., vol. 36, no. 1,
ACKNOWLEDGMENT pp. 732–743, Jan. 2021.
This work was supported in part by the European Union’s [20] Y. Li et al., “Stability analysis and location optimization method for
multiconverter power systems based on nodal admittance matrix,” IEEE
Horizon 2020 research and innovation programme through the J. Emerg. Sel. Topics Power Electron., vol. 9, no. 1, pp. 529–538,
Marie Skłodowska-Curie under Grant 765585. This document Feb. 2021.
reflects only the author’s views; the European Commission [21] C. Zhang, M. Molinas, A. Rygg, and X. Cai, “Impedance-based analysis
of interconnected power electronics systems: Impedance network mod-
is not responsible for any use that may be made of the eling and comparative studies of stability criteria,” IEEE J. Emerg. Sel.
information it contains. Topics Power Electron., vol. 8, no. 3, pp. 2520–2533, Sep. 2020.
[22] Y. Liao and X. Wang, “Impedance-based stability analysis for inter-
connected converter systems with open-loop RHP poles,” IEEE Trans.
R EFERENCES Power Electron., vol. 35, no. 4, pp. 4388–4397, Apr. 2020.
[1] G. Revel, A. E. Leon, D. M. Alonso, and J. L. Moiola, “Dynamics [23] W. Xu, Z. Huang, Y. Cui, and H. Wang, “Harmonic resonance mode
and stability analysis of a power system with a PMSG-based wind farm analysis,” IEEE Trans. Power Del., vol. 20, no. 2, pp. 1182–1190,
performing ancillary services,” IEEE Trans. Circuits Syst. I, Reg. Papers, Apr. 2005.
vol. 61, no. 7, pp. 2182–2193, Jul. 2014. [24] S.-F. Chou, X. Wang, and F. Blaabjerg, “Frequency-domain modal
[2] C. Buchhagen, C. Rauscher, A. Menze, and J. Jung, “BorWin1—First analysis for power-electronic-based power systems,” IEEE Trans. Power
experiences with harmonic interactions in converter dominated grids,” in Electron., vol. 36, no. 5, pp. 4910–4914, May 2021.
Proc. Int. ETG Congr. Die Energiewende-Blueprints New Energy Age, [25] F. Zhi-Fang and H. Jimin, Modal Analysis. Amsterdam,
2015, pp. 27–33. The Netherlands: Elsevier, 2001.

Authorized licensed use limited to: UNIVERSITAT POLITECNICA DE CATALUNYA. Downloaded on October 26,2023 at 18:54:39 UTC from IEEE Xplore. Restrictions apply.
 ORELLANA et al.: STABILITY ASSESSMENT FOR MULTI-INFEED GRID-CONNECTED VSCs 3771

[26] L. Xu, D. Zhi, and L. Yao, “Direct power control of grid connected Luis Sainz received the M.S. degree in indus-
voltage source converters,” in Proc. IEEE Power Eng. Soc. Gen. Meeting, trial engineering and the Ph.D. degree in engineer-
Jun. 2007, pp. 1–6. ing from the Universitat Politècnica de Catalunya
[27] L. Orellana, M. Cheah-Mane, E. Prieto-Araujo, L. Sainz, and (UPC), Barcelona, Spain, in 1990 and 1995, respec-
O. Gomis-Bellmut, “On resonance instabilities in VSCs connected to tively. Since 1991, he has been with the Electrical
weak grids,” in Proc. Symp. Aalborg, 2019. Engineering Department, UPC, where he was an
[28] A. Rygg, M. Molinas, E. Unamuno, C. Zhang, and X. Cai, “A simple Associate Professor from 1997 to 2020. He is cur-
method for shifting local dq impedance models to a global reference rently a Professor with UPC. His research interests
frame for stability analysis,” Jun. 2017, arXiv:1706.08313. [Online]. include power quality, and power system stability in
Available: https://arxiv.org/abs/1706.08313 grid-connected VSCs.
[29] M. Amin and M. Molinas, “Small-signal stability assessment of power
electronics based power systems: A discussion of impedance- and
eigenvalue-based methods,” IEEE Trans. Ind. Appl., vol. 53, no. 5,
pp. 5014–5030, Sep. 2017.
[30] H. Liu, X. Xie, and W. Liu, “An oscillatory stability criterion based
on the unified dq-frame impedance network model for power systems
with high-penetration renewables,” IEEE Trans. Power Systems, vol. 33,
no. 3, pp. 3472–3485, May 2018.
[31] H. Liu and X. Xie, “Impedance network modeling and quantitative
stability analysis of sub-/super-synchronous oscillations for large-scale
wind power systems,” IEEE Access, vol. 6, pp. 34431–34438, 2018.
[32] P. P. Kundur, N. J. Balu, and M. G. Lauby, Power System Stability and
Control. New York, NY, USA: McGraw-Hill, 1994.
[33] J. M. Maciejowski and E. L. Dagless, Multivariable Feedback Eduardo Prieto-Araujo (Member, IEEE) received
Design (Electronic Systems Engineering Series). Reading, MA, USA: the degree in industrial engineering from the School
Addison-Wesley, 1989. of Industrial Engineering of Barcelona (ETSEIB),
[34] C. T. Chen, D. Chen, and K. Firm, Linear System Theory and Design Technical University of Catalonia (UPC), Barcelona,
(Oxford Series in Electrical and Computer Engineering). Oxford, U.K.: Spain, in 2011, and the Ph.D. degree in electri-
Oxford Univ. Press, 1999. cal engineering from UPC in 2016. He joined the
[35] H. Liu et al., “Subsynchronous interaction between direct-drive PMSG CITCEA-UPC research group in 2010. He is cur-
based wind farms and weak AC networks,” IEEE Trans. Power Syst., rently a Serra Húnter Lecturer with the Electrical
vol. 32, no. 6, pp. 4708–4720, Nov. 2017. Engineering Department, UPC. His research inter-
[36] R. Abu-hashim et al., “Test systems for harmonics modeling and ests include renewable generation systems, control
simulation,” IEEE Trans. Power Del., vol. 14, no. 2, pp. 579–583, of power converters for HVDC applications, inter-
Apr. 1999. action analysis between converters, and power electronics dominated power
[37] Deliverable 3.1: Detailed Functional Requirements to WPPs. systems.
PROMOTioN, 2016, p. 34.
[38] L. Sainz, L. Monjo, M. Cheah-Mane, J. Liang, O. Gomis-Bellmunt, and
E. Prieto-Araujo, “Admittance study of grid-connected VSCs for har-
monic oscillatory instabilities,” IET Gener., Transmiss. Distrib., vol. 13,
no. 18, pp. 4049–4060, Sep. 2019.
[39] L. Harnefors and H.-P. Nee, “Model-based current control of AC
machines using the internal model control method,” IEEE Trans. Ind.
Appl., vol. 34, no. 1, pp. 133–141, Jan. 1998.

Luis Orellana received the B.Sc. degree in electro- Oriol Gomis-Bellmunt (Fellow, IEEE) received the
mechanical engineering from Universidad Privada degree in industrial engineering from the School
Boliviana (UPB), Bolivia, in 2012, and the M.Sc. of Industrial Engineering of Barcelona (ETSEIB),
degree in electrical energy systems from Cardiff Technical University of Catalonia (UPC), Barcelona,
University, Cardiff, U.K., in 2017. He is currently Spain, in 2001, and the Ph.D. degree in electrical
pursuing the Ph.D. degree in electrical engineering engineering from UPC in 2007. In 1999, he joined
with the Universitat Politècnica de Catalunya (UPC), Engitrol S.L., where he worked as a Project Engineer
Barcelona, Spain. Since 2018, he has been a part with the Automation and Control Industry. Since
of InnoDC (Innovative tools for offshore wind and 2004, he has been with the Electrical Engineering
DC grids) project at the CITCEA research group. Department, UPC, where he is currently a Professor
He has more than three years’ experience elaborating and participates with the CITCEA-UPC research
high-voltage electricity projects and analysing power systems at the Empresa group. Since 2020, he has also been an ICREA Academia Researcher. His
Nacional de Electricidad (ENDE), Bolivia. His research interests include research interests include the fields linked with electrical machines, power
power system stability and grid integration of renewable energies. electronics, and renewable energy integration in power systems.

Authorized licensed use limited to: UNIVERSITAT POLITECNICA DE CATALUNYA. Downloaded on October 26,2023 at 18:54:39 UTC from IEEE Xplore. Restrictions apply.