Received: 31 August 2021 Revised: 16 November 2021 Accepted: 30 December 2021 IET Renewable Power Generation

DOI: 10.1049/rpg2.12398

ORIGINAL RESEARCH PAPER

Virtual oscillator-based methods for grid-forming inverter control:

A review

Sima Azizi Aghdam Mohammed Agamy

Department of Electrical and Computer Abstract

Engineering, University at Albany, State University
of New York, Albany, New York, USA
In inverter-dominant power systems, grid-forming (GFM) inverters regulate voltage and
frequency. To construct GFM inverters, conventionally, various control methods based
Correspondence on synchronous machine emulation or droop characteristics have been employed. How-
Mohammed Agamy, Department of Electrical and ever, recently, inverters are regulated by Virtual Oscillator Control (VOC) to emulate the
Computer Engineering, University at Albany, State
dynamics of a weakly nonlinear oscillators. In contrast to droop control, VOC is a time-
University of New York, Albany, NY, USA.
Email: [email protected] domain controller that enables interconnected inverters to stabilize arbitrary initial con-
ditions to a synchronized sinusoidal limit cycle rapidly. The fast synchronization, accu-
Funding information rate power sharing in grid-connected and islanded modes, simple and robust implementa-
New York State Research and Development Author- tion make it a promising candidate for inverter-dominant power systems. Moreover, VOC
ity (NYSERDA), Grant/Award Number: 137936
removes the computational burden within machine emulation-based methods and pro-
vides an approach to measure the frequency without employing phase-locked loop (PLL).
Hence, to leverage the advantages of this method and expand its application in power sys-
tem control, this paper reviews different VOC implementations. Mainly, this paper focuses
on the studies related to the fundamental theory of oscillators and design process. Addi-
tionally, there are discussions on the stability analysis of VOC-based systems, harmonics
and pre-synchronization. Challenges on the compatibility with heterogeneous controllers,
fault ride-through and other limitations of VOC including non-inertial response and cou-
pled power control are investigated and solutions are outlined.

1 INTRODUCTION methods, [6] compared different methods including machine

emulation-based method [7–9], droop-based control [10–13],
Over recent years, to provide more clean and accessible energy, power synchronization control [14], and matching control [3,
power system supply sources have changed towards utiliz- 15]. Other GFM methods such as VOC [16], H∞ ∕H2 - based
ing renewable energy sources (RES) [1]. As a consequence, control [17, 18], ViSync[19] and frequency-shaping approach
interfacing devices and corresponding control approaches [20] were discussed in [21]. It is shown that while using machine
have significantly altered [2]. In the new configurations, bulky emulation-based approaches can enable GFM operation, the
rotational synchronous generators (SGs) are replaced by power exact replication of SGs requires a complicated model with high
converters, which have limited inherent inertia, restricted volt- mathematical operations. For instance, analysing a high order
age and frequency regulations [3, 4]. Thus, providing accurate model of SG is complicated and can yield to numerical instabil-
synchronization and power sharing among various generators ity, time delays in the control and signal processing algorithms
and guarantee a stable operation require utilizing enhanced [3, 22, 23]. Besides, there is a significant difference between the
control methods [2, 4]. In this regard, various methods have feature of the static inverters with fast dynamics and relatively
been investigated trying to mitigate the undesirable impact of slow SGs. This issue is still challenging and need more studies to
mentioned transition. Among them, GFM control methods know whether this emulation is accurate and functional [24]. To
are more popular as they can inherently form the voltage reduce the complexity of the machine emulation-based meth-
phasor and inertial response [5]. To evaluate various GFM ods, swing equation-based approaches and matching control

This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the
original work is properly cited, the use is non-commercial and no modifications or adaptations are made.
© 2022 The Authors. IET Renewable Power Generation published by John Wiley & Sons Ltd on behalf of The Institution of Engineering and Technology

IET Renew. Power Gener. 2022;16:835–855. wileyonlinelibrary.com/iet-rpg 835

836 AZIZI AGHDAM AND AGAMY

FIGURE 1 Grid-forming inverter control approaches

are used which may lead to power and frequency oscillations as they do not need power calculations and power filters. The
[2]. Droop-based methods which adjust voltage and frequency efficiency of the VOC-based inverter in power-sharing among
by tuning active and reactive power [25] are widely employed parallel-connected photovoltaics is discussed [41]. The two-
in power system control. Although the efficacy of this method stage converter-based PV generation is used combined with
has been proved, the voltage dynamics are often neglected modified VOC and cascade sliding mode control. With this pro-
[23]. Besides, the synchronization of systems is only local as posed control configuration, the power-sharing in proportion
the phasor dynamics admit multiple equilibria on the torus, to the inverter power rating and maximum power point tracking
corresponding to different relative angle configurations [26, is guaranteed autonomously without the need of energy storage
27]. Recently, a novel method, VOC, is used for controlling the systems. Further applications of VOC in the single-phase and
GFM inverters, which attracts significant attention. The supe- three-phase in islanded mode were investigated in [42, 43].
riority of VOC performance compared to conventional GFM However, for grid-connected mode application, [44] indicated
controllers has shown in [10, 11, 28, 29]. More discussions on that these oscillators need to be modified as they have several
the GFM controllers can be found in [21]. Figure 1 summarizes undesirable features. Van der Pol, for instance, contains high
and demonstrates the main GFM inverter control approaches. level of harmonics and it is incompatible with secondary con-
To leverage the positive features of VOC, this work focuses on trol [44]. Additionally, they are not dispatchable [38, 40]. Hence,
the various arrangement of VOC. The basic idea of VOC is [44] introduced an enhanced design that can eliminate the issue
using the dynamic of nonlinear oscillators originated from syn- with harmonics and compatibility. [45] altered the VOC design
chronization of coupled oscillators [30]. As it was proven that by adding a coefficient enabling power control. In addition, a
there is an analogy between coupled oscillators network and top-down approach inspired by consensus strategies and flock-
power inverters [30, 31], this method is generalized and used ing was introduced [46–48]. As an example of consensus-based
in power system control. The initial controller used Kuramoto approach, [23, 49, 50] discussed dispatchable virtual oscillator
model and implemented in polar coordinates with fixed mag- control (dVOC) and verified that it provides almost global
nitude assumption [32]. However, further research, [33, 34], asymptotic Moreover, the authors of [37] proposed Andronov–
investigated the modified versions of the Kuramoto model of Hopf Oscillator (AHO) which is similar to dVOC approach
coupled oscillators. Recently, the idea of coupled oscillators while providing simple and explicit design. Although dVOC
has been implemented with Liénard type oscillators such as and AHO-based controllers provide a suitable GFM operation,
Van der Pol (VdP) and Dead-Zone Oscillator (DZO) [35, 36]. their performance are limited during fault. Hence, in recent
These oscillators can provide periodic, self-sustained, and stable attempt, authors in [16] proposed an unified virtual oscillator
oscillations [37] and they require less computational operation control (uVOC) which added fault-ride through (FRT). It
as they depend on the instantaneous measurements [38]. There- also provided grid synchronization without using PLL, which
fore, faster transient response compared to droop control can extends the application of VOC in future power system control.
be achieved [39]. These communication-free oscillators, also, As discussed, VO-based methods can enhance future power
provide fast synchronization and accurate power- sharing [40] system control by providing distinguished features. These
AZIZI AGHDAM AND AGAMY 837

TABLE 1 Benefits and key features of virtual oscillator-based control Moreover, the control loop interactions leading the coupled
Benefits Key features power control is discussed and the related works are reviewed.
It is noteworthy to mention that the scope of this work is
Fast synchronization and ∙ No need for explicit power calculations
limited to the study of VO-based approaches used in GFM
accurate power sharing [42, and power filters [44]
44] control and other existing VOs and GFM controllers are not
considered.
Instantaneous response [51] ∙ Time-domain based control law [38]
∙ No need for using the low-pass filter [38] The remainder of paper is organized as follows: Section 2
provides the analysis of VO theory and synchronization con-
No need for PLL to measure ∙ Embedded power-frequency equations in
the frequency [16] VOC can replace the PLL [16, 37, 50] cept. In Section 3, the focus is more on various oscillators
structures and their parameters design. It will be followed by
Provide a droop-like voltage ∙ Embedded droop-based equations in VOC
and frequency response [52] in steady state is equal to droop principles a detailed comparison between the droop control and VOC in
[10, 11, 39] Section 4. The operation consideration including stability anal-
Less computational burden ∙ Control law is based on the inverter output ysis, pre-synchronization and harmonics are discussed in Sec-
[45] current instead of power calculations [52] tion 5. Section 6 reviews the implementation challenges as well
∙ No need for modeling the high order as relative proposed solutions. Finally, concluding remarks and
synchronous machine like machine direction for future work are outlined in Section 7.
emulation-based approaches [23, 49]
Robust controller [40] ∙ Independent of filter parameter variations
[40]
∙ Independent of type of load [40]
2 CONCEPT OF VIRTUAL
∙ Independent of number of inverters and OSCILLATORS AND SYNCHRONIZATION
their initial conditions [40]
Straightforward and simple ∙ Combined voltage tracking control with 2.1 Synchronization of virtual oscillators
design [42] primary control [52]
∙ No need for specific external reference of Synchronization of coupled oscillators has been studied in var-
voltage and frequency or any ious areas such as nature, science and engineering [53–56]. The
trigonometric functions [42]
synchronization of oscillators are obtained when phase locking
and consensus among corresponding states of coupled dynam-
ical system are achieved [30]. In general, a coupled oscillator
features and corresponding benefits are summarized in Table 1. network can be described by a set of heterogeneous oscilla-
However, in advance to applying them in real applications, tors and a graph describing the oscillators interaction. Based on
different aspects of them still need more studies. In this regard, this description, coupled oscillators dynamic equation can be
this work provides a broad survey on applied VOC in power defined by single-state or multi-state dynamic model [57–59]. A
system control to point out challenges and directions for future dynamic model of a network with N oscillators, each character-
investigation. Unlike the previous works which have focused ized by a phase angle 𝜃i ∈ S1 and a natural rotation frequency
on the various applications of VOC in GFM inverter control, 𝜔i ∈ ℝ2 is given by
this work aims to provide a systematic review on VOC. To this
end, it compares different VOC arrangements and highlight ∑
N
the merits and weaknesses of these implementations. It, also, 𝜃̇ i = wi − ai j sin(𝜃i − 𝜃 j ), (1)
studies the various parameters of VOC control law and their j =1
impacts on the system stability. Besides, the previous inves-
tigations considered several assumptions such as constant dc where the dynamic of each isolated oscillator is 𝜃̇ i = 𝜔i for
voltage, pure inductive and resistive networks and so on which i, j ∈ 1, … , N , and the interaction function between neighbor-
are not valid in special cases. The main contribution of this ing oscillators is ai j sin(𝜃i − 𝜃 j ), where ai j is a weight showing
work is preparing a background to challenge the validity of coupling strength for each undirected edge {i, j } [60]. [60], also,
these assumptions and robustness of controller against design presented the mechanical analog of the coupled oscillators and
and network parameter variations, type of load and operation springs. It is shown that there is a competition between the
modes. Furthermore, operational challenges and the main exist- external driving forces 𝜔i and the internal restoring torques
ing issues within the VOC implementation are comprehensively ai j sin(𝜃i − 𝜃 j ). Therefore, coupled oscillator dynamics (1) is a
studied. It will show the lack of inertial response, control loop trade-off between each oscillator’s tendency to align with its
interactions and incapability to provide FRT degrade the con- natural frequency 𝜔i and the synchronization enforcing cou-
troller performance. Various solutions for these deficiencies are pling ai j sin(𝜃i − 𝜃 j ) with its neighbors. In this case, a strong
outlined. Among several issues within VOC implementation, coupled and homogeneous network is amenable to synchro-
the lack of inertial response is specifically studied in this work. nization, where all frequencies 𝜔i or even all phases 𝜃i become
This is because providing inertial response is vital for GFM aligned. However, a weakly coupled and strongly heterogeneous
controller applied in inverter-dominated power system (low-
inertia system). A three-phase modified dVOC-based inverter is 1
S 1 = [0, 2𝜋] denotes the unit circle.
simulated and the frequency response enhancement is shown. 2
The set of real numbers is denoted by ℝ.
838 AZIZI AGHDAM AND AGAMY

(i.e. with strongly dissimilar natural frequencies) network does in the output inductance, and measurement of terminal quan-
not display any coherent behavior. Generalizing this concept tities or controller computational error [67]. In [68–70], it is
and using it in power system applications can help present the shown that when internal voltage in parallel inverters with pas-
analogy between coupled oscillator network and power system sive loads are different from each other, there will be a periodic
components, that is, inverters. As shown in [61], the swing equa- current circulation that decays to zero as the terminal voltages
tion of the power system and non-uniform Kuramoto-oscillator in the system of parallel inverters synchronize. To have an accu-
are equivalent. Hence, by leveraging the feature of a nonlinear rate synchronization among a set of inverters, [42] proposed
oscillator, the frequency and voltage of the power system can be control based on DZO. Moreover, this issue was overcome
K
controlled. To this end, (1) is rearranged by replacing ai j = by applying a non-linear control variable in S-Z module of the
N
given as Kuramoto model of coupled oscillator virtual oscillator in which the phase difference between the
voltage and current is used to determine the output value of
K ∑
N the oscillator [68]. The S-Z module in the virtual oscillator is a
𝜃̇ i = wi − sin(𝜃i − 𝜃 j ), (2) module with dual characteristics of controllable current source
N j =1
and controllable resistor, which is connected to a resonant
circuit. The mathematical equations and analysis can be found
where K is coupling gain of N oscillators. The description in in [68]. Thus, the phase lag of the current caused by the circu-
[62] emphasised that Kuramoto model of oscillators are locally lating current will be eliminated. Similarly, [69] added Source-
exponentially synchronized regarding all to all connectivity with Resistance (SR) module to adjust amplitude and frequency and
different wi when K > Kcritical ; where Kcritical is a certain thresh- compensate load current disturbance. The current of SR is
old that satisfies synchronization condition. To define Kcritical , controlled by output voltage of virtual oscillator. Furthermore,
Lyapunov approach was used to ensure stability by appropri- a distinctive method for synchronization of parallel inverters
ate choice of K [30]. In other investigations [63]-[64], synchro- was introduced [67], which uses the virtual-based oscillators
nization conditions for coupled oscillators using passivity the- to generate carrier reference unlike previous methods which
ory were discussed. In presence of passive elements in these produce modulation reference for synchronization of paral-
methods, the formulation of a storage function and power net- leled inverter. It is confirmed that this method can eliminate
works will be difficult. Hence, in [40], 2 input-output stability the switching frequency circulating currents in addition to
methods was used. This method facilitates the analysis of model fundamental one, which has been suppressed by modulation
with storage functions. Furthermore, synchronization of multi- waveform.
ple inverters based on droop control was analyzed with equiv-
alent Kuramoto-oscillator models and sufficient stability condi-
tions [33],[65]. While this conventional method has been used 3 VIRTUAL OSCILLATOR CONTROL
in power system control, it cannot provide very fast synchro- METHODS
nization (less than the 10 cycles) [66]. The reason is that non-
sinusoidal signals can dominate the transient behavior in a stan- In this section, oscillators that have been used in the power
dard phasor-domain [66]. To enhance the synchronization of system control in either grid-following or grid-forming modes
multiple inverters, control methods based on virtual oscillator will be presented. Generally, they can be classified into two
(e.g. Liénard oscillators) was proposed in [42],[40]-[38], which main groups: Liénard oscillators including VdP and DZO ,
is independent of the number of oscillators (inverters) and the and consensus-based approaches such as dVOC, AHO and
load impedance feeding by inverters. This communication-less uVOC.
method, also, does not require computation of the real and reac-
tive power output which increases the speed of synchroniza-
tion. Additionally, it is shown that employing this method min- 3.1 Van der pol oscillator
imizes system frequency deviation from the nominal value and
enable accurate synchronization among parallel inverters. The One of the well-known methods to emulate non-linear dynam-
importance of precise synchronization in multiple-inverter net- ics of virtual oscillator control is Van der Pol oscillator(VdP).
works stems from the need to eliminate the circulating currents This oscillator is defined by a second-order model of oscil-
among inverters, which reduces the efficiency of the system. In lator [38]. The general equation of VdP which evolves in
the following section the impact of circulating current caused by time according to the second-order differential equation is
inaccurate synchronization among inverters will be presented in given by
more detail.
d 2x dx
2
− 𝜇(1 − x 2 ) + x = 0, (3)
dt dt
2.2 Circulating current among parallel
inverters where x denotes the position coordinate and it is a function of
the time t. 𝜇 represents a scalar parameter indicating the non-
Circulating current among parallel inverters originates from linearity and the strength of the damping [71]. Leveraging Lié-
inaccurate synchronization of inverter output voltage, mismatch nard’s theorem to prove that the system has a limit cycle, and
AZIZI AGHDAM AND AGAMY 839

ditions based on Lyapunov approach. The results show the

sufficient conditions based on the Lyapunov approach for the
two models are analogous and a framework for synchroniza-
tion of second order coupled oscillators can be constructed. The
incompatibility of nonlinear oscillators with conventional linear-
control design is addressed in [38], where the sinusoidal steady-
state and harmonic behavior of such oscillators are extracted
by averaging and perturbation-based nonlinear analysis meth-
ods. The analysis led to design procedure guidelines which
are described in the next section. Furthermore, by using peri-
odic averaging method, the droop laws are intrinsically embed-
ded within a slower time scale in the nonlinear dynamics of
VdP oscillators are extracted [39]. The comparison of perfor-
mance of non-linear oscillators and other grid-forming con-
trollers exhibited the efficiency of dynamic responses of the
virtual oscillator. However, there are critical limitations in con-
ventional VOC to be used in grid-connected mode. [44] pro-
FIGURE 2 A three-phase inverter with Van der Pol controller [38] posed a compatible hierarchical control structure that enables
operation and seamless transition between islanded and grid-
connected modes. In [72], it is demonstrated that the output
voltage drops due the inverter’s filter. Hence, a dual-loop con-
rewriting (3) based on the Liénard transformation leads to trol architecture that allows inverters with current controllers to
( 1 ) be re-purposed for voltage control mode operation with VOC is
ẋ = 𝛾 x − x 3 − y , (4a) proposed. This modification compensate the voltage drop and
3
extend the applicability of the VOC to settings where a cur-
1 rent control is used. Furthermore, [73] presented a decentral-
ẏ = x, (4b) ized control strategy based on VdP that yields switch interleav-
𝛾
ing for parallel-connected dc-dc buck converters. It was shown
where ẋ and ẏ are state variables and 𝛾 is a parameter which this controller compared to novel distributed methods requires
defines the controller. VdP is utilized for control of single-phase no communication, and enhance reliability, modularity and cost.
and three-phase inverters in islanded mode [38], [45]. To study
the VdP structure, a three-phase inverter-based controlled by
VdP is shown in Figure 2. As shown, the oscillator is composed 3.1.1 Parameter selection procedure for VdP
of harmonic oscillator with inductance and capacitance of the
oscillating circuit L and C, respectively, a negative-conductance As described in the previous section, to achieve a unit limit cycle,
-𝜎 and a cubic voltage-dependent current source. The current virtual-oscillator parameters must be determined such that a
consumed by the cubic voltage-dependent current source is set of ac performance specifications is met. Towards this end,
given by 𝛼 𝜈c3 , where 𝛼 is a positive constant and virtual capac- firstly, physical-inverter terminal voltage and output current,
itance voltage vc is the oscillator state variable. Another state and those of the virtual-oscillator circuit are scaled by kv and
variable that can show the dynamic of oscillator is virtual induc- ki . To standardize design, kv is chosen such that when capacitor
tance current i̇L [38]. According these variables, the dynamic voltage is 1V RMS, the inverter-terminal voltage is equal to the
equation of VdP oscillator can be defined as open-circuit voltage (Voc ). Furthermore, ki is defined such that
when the output current (io) is 1A, the inverter is loaded to rated
[ ] active power Prated [38]. Hence, kv and ki are obtained as
1 v3
v̇ c = −𝛼 2 + 𝜎v − kv iL − kv ki i , (5a)
C kv Vmin
kv = Voc ki = N ph , (6)
v Prated
i̇L = . (5b)
kv L
where N ph denotes the number of phases and Vmin is minimum
where kv and ki are scaling factors. kv scales vc to produce allowable voltage. Based on Voc and Vmin , control parameters of
the signal v, control signal of the pulse width modulation VdP, 𝛼 and 𝜎, can be defined
PWM , and ki normalizes inverter output current i. The equa-
tion to obtain kv and ki will be explained in detail in next Voc Voc2 2𝜎
𝜎= 𝛼= . (7)
section. Vmin Voc2 − V 2 3
min
[61] expanded the idea of synchronization of coupled oscil-
lators and studied two benchmark oscillators including VdP Furthermore, to define L and C, as described in [38], a boundary
and Fitzhugh Nagumo and the sufficient synchronization con- for permissible C based on the dynamic response and harmonic
840 AZIZI AGHDAM AND AGAMY

analysis is defined. For instance, Cmin and Cmax are restricted to

meet the specified rise time trise , maximum allowable angular fre-
quency |Δwmax |, acceptable harmonic level described by 𝛿3∶1 max
,
the ratio of amplitudes of the third harmonic to the fundamen-
tal, and etc. as follows

1 Voc Qrated
Cmin ∶= , (8)
2|Δwmax | Vmin Prated

𝜎
Cmin1 ∶= max , (9)
8wnom 𝛿3∶1 FIGURE 3 (a) Dead-zone function f (v), (b) the nonlinear
voltage-dependent current source g(v) [36]
trise,max
Cmax ∶= 𝜎, (10)
6 imum slope of the voltage-dependent current source function
g(Vosc ), which is given as
max(Cmin , Cmin1 ) ≤ C ≤ Cmax . (11)
g(Vosc ) = f (Vosc ) − 𝜎Vosc . (15)
The rated power including rated reactive power, Qrated , and Prated ,
Vmin , Voc , 𝜎 and nominal angular frequency, wnom , also, impact on The functions f(⋅) and g(⋅) are indicated in Figures 3a and 3b,
the range of proper C . Thus, the allowable capacitor range can respectively. The solid lines indicate the DZO related func-
be obtained by applying restriction factors and (8)–(11). Then, tions and dashed ones illustrate VdP benchmark functions. Fig-
the value of L can be selected in such a way that the resonant ure 3 confirms that despite the difference between parame-
frequency matches the nominal frequency wnom . ters of DZO and VdP, both methods have similar approaches
to establish a decentralized control of inverters as they are
1 governed by approximately the same functions. The electrical
L= 2
. (12)
wnom C circuit-equivalent of three-phase power system with nonlinear
DZO is shown in Figure 4.
Besides the aforementioned requirements to choose the proper As shown, similar to VdP oscillator the digital controller con-
L and C , they should be selected such that near sinusoidal oscil- sists of a nonlinear voltage-dependent current source defined in
lations are ensured as well as providing fast response. To these (15) and parallel RLC circuit. To build the command signals to
ends, another design parameter that is quasi-harmonic limit 𝜖 is control the inverter, the nonlinear dynamic equations with two
defined. It can arrange the relationship of L and C as states, Vosc and the inductor current iL , are constructed as
√
1 [ ( 1) ]
𝜖=
L
. (13) V̇ osc = Vosc 𝜎 − − f (Vosc ) − iL − iosc
C C R
(16)
1
The smaller value of 𝜖 exhibits near-sinusoidal oscillations, i̇L = V .
L osc
which can be achieved at the expense of a sluggish dynamic
response. Therefore, the parameter design trades off between
dynamic response and harmonic content.

3.2 Dead-zone oscillator

Another configuration of VOC is DZO which is inspired by

VdP oscillator [74], [75]. DZO relies on the dead-zone function
f (Vosc ), defined as follows

⎧
⎪2𝜎(Vocs − 𝜙), Vocs > 𝜙
⎪
f (Vocs ) = ⎨0, |Vocs | ≤ 𝜙 , (14)
⎪
⎪2𝜎(Vocs + 𝜙), Vocs < −𝜙
⎩

where Vosc denotes the DZO terminal voltage, and 𝜙 determines

the interval that Vosc is zero. In this arrangement, 2𝜎 is the max- FIGURE 4 A three-phase inverter with the DZO [36]
AZIZI AGHDAM AND AGAMY 841

In general, DZO control was formulated for voltage-source Moreover, the effectiveness of virtual oscillator- based con-
inverters [29]; whereas [76] investigated the current-source troller in the providing an equal power contribution of each
inverter to deal with protection issues. To do this, a virtual inverter in power system is shown in [42]. The relationship of
admittance which simulates the output current produced by a rated power of jth or kth inverter can be represented by
voltage difference across a component that programmed in the
controller is used. The current source implementation outper- Pk kj
forms the voltage source since the effective output impedance = ∀ k, j = 1, … , N , (18)
Pj kk
can be adjusted by tuning virtual admittance while in the volt-
age source implementation it needs to change the core oscillator k j and kk are output-filter scaling parameters. The output cur-
dynamics or inverter filter physically. The requirement for this rent of each inverter is scaled by k j ki and kk ki . Thus, to enhance
implementation is having a current control bandwidth much the power sharing among heterogeneous inverters with vari-
higher than the base controller. In this case, according to the ous rated power, k j and kk can be defined to balance differ-
singular perturbations theorem converter can be assumed as an ent rated powers. This modification guarantees equal power
ideal current source. It is concluded that the implementation sharing among inverters without physically changing the fil-
based on these conditions leads to a current-controlled amplifier ter impedance at the output of an inverter. Therefore, the rel-
oscillator system which is equivalent to voltage sourced ampli- ative power contribution of an inverter under DZO control
fier. Therefore, the basic oscillator performance, synchroniza- can be altered by changing controller coefficients. Furthermore,
tion of multiple inverters, responses to load changes, and droop the proposed filter-design strategy also ensures that the rel-
characteristics will enhance. Additionally, the extension of DZO ative distortion contributions of the individual inverter out-
from single-phase to three-phase was demonstrated. put currents are matched, such that the THD of the parallel
system is essentially the same as the THD of each individual
inverter [42]. Furthermore, recently, [77] tried to introduce a
3.2.1 Parameter selection procedure for DZO more comprehensive design procedure for VOC to guarantee
the asymptotic stability. This procedure also takes into account
The parameters selection of DZO including linear oscillator pre-synchronization requirements for the case an inverter is
with (R, L, C), and scaling factors are similar to specifications added/removed.
provided in 3.1. DZO non-linear oscillator parameters, 𝜙 and
𝜎 must be defined such that a stable and unique limit cycle
is achieved. To this end, the value of 𝜙 can be tuned with an 3.3 Dispatchable virtual oscillator
open-circuit test to ensure that the inverter output voltage does
1
not exceed the maximum load voltage. In addition, 𝜎 > and As shown in previous sections, VdP and DZO are capable
√ R
to enhance the synchronization and equal power sharing in
L 1
𝜖= (𝜎 − ), the phase plot resembles a unit circle, which islanded grid [39]. However, the utilization of them in grid-
C R
implies that the voltage oscillation approximates an ideal sinu- connected mode faces some limitations. First, these approaches
soidal in the time domain [42],[40]. To ensure the global asymp- do not provide dispachability inherently due to their structures
totic synchronization conditions, in a network with N number [45]. This is because the oscillator has only one input (feedback
of inverters, DZO parameters must satisfy the following equa- of i) and no input to define power set-points. In this case, con-
tion troller is not able to have any control on power and inverter
behaves as a grid-following inverter. Moreover, VdP oscillator
‖ ki k−1
v Z f ( jw)Zosc ( jw) ‖
‖
‖ contains the third harmonic which restricted the application
sup ‖ −1 ‖ 𝜎 < 1, (17)
w∈R ‖ ‖
‖ ki kv Z f ( jw) + Zosc ( jw) ‖ 2 in grid-connected mode[44],[78]. Hence, to provide dispatch-
ablity feature and suppress harmonics, [44] proposed modi-
where Zosc is the impedance of the DZO resonant RLC cir- fied Liénard oscillator. Additionally, dVOC which inherits the
cuit and Z f is a linear reference filter impedance. The condi- consensus-like dynamic features and guarantees the network of
tion in (17) confirms that the synchronization based on this inverters is almost globally asymptotic stable with respect to
controller is independent of load type and number of inverters set-points satisfying ac power-flow is introduced [49],[23],[50].
[42]. Meanwhile, (17) implies the controller may not be robust [50] showed that dVOC is dispatchable, as its control law con-
to filter impedance or VO impedance variations. To analyse this, tains power set-point that enable system operators to spec-
[40], specifically examined the robustness of VOC to parameter ify power set-points for each inverter. It is, also, confirmed
variations by measuring the error in voltage synchronization. In that dVOC operation reduces to the aforementioned oscillators
[40], the factors causing parameter variations were introduced as when power set-points are not regarded [50]. Schematic in Fig-
small round-off or truncation errors in programming of param- ure 5 indicates a three-phase inverter controlled by dVOC. As
eters with microcontroller or filter impedance variations due to it can be seen, dVOC does not contain explicit term of typ-
hardware degradation, etc. It was concluded that small varia- ical oscillator (L, C ) and is governed by mathematical equa-
tions in parameters leads to negligible error in power sharing tion duplicating conventional oscillator. This mathematical anal-
and circulating currents among inverters decay to zero. ysis considering essential conditions to reach the almost global
842 AZIZI AGHDAM AND AGAMY

tion was manipulated and Di was defined as

[ ∗ ]
1 pi qi∗
Di = ∗2 R(𝜃) , (22)
vi −qi∗ p∗i

where R(𝜃) is rotating matrix, which is represented as given

[ ]
cos(𝜃) −sin(𝜃)
R(𝜃) ∶= , (23)
sin(𝜃) cos(𝜃)

where 𝜃 ∶= tan−1 (𝜌w0 ). As it can be seen, Di depends on

the ratio of inductance-resistance value (𝜌) and reference val-
ues instead of the information of line and angle set-point.
Besides, by defining Di , dispatchability feature is added to
controller without the need for extra loop.
∙ Voltage magnitude correction or voltage magnitude feedback
FIGURE 5 A three-phase inverter with the dVOC [50] provides a minimum error between reference voltage magni-
tude and measured voltage magnitude and can be formalized
as
asymptotic stablity is comprehensively explained in [49],[23]. In [ ] [ ]
v𝛼 𝜂𝛼𝜙v𝛼
this work, only the fundamental control law of dVOC will be 𝜇𝜙(vi )vi = 𝜂𝛼𝜙 = , (24)
discussed. As shown in [50], to generate a voltage signal based v𝛽 𝜂𝛼𝜙v𝛽
on the pre-defined magnitude and frequency, the control law of
dVOC is constructed as follows where 𝜇 defines convergence speed to steady-state. It is a
multiplication of 𝜂 and another controller parameter 𝛼, which
V̇ = w0 Jvi + 𝜂(Di vi − R(𝜃)i0, i ) + 𝜇𝜙(vi )vi . (19) denotes amplitude regulation gain. 𝜙(vi ) represents normal-
ized quadratic voltage magnitude error and it operates as a
This equation can be expressed by three terms as: voltage regulator in SG. 𝜙(vi ) is defined as
vi∗2 − ||vi ||2
∙ A harmonic oscillator which is the fundamental part of 𝜙(vi ) = . (25)
dVOC and generates a voltage signal based on the desirable vi∗2
amplitude vi∗ and nominal angular frequency w0 as
In addition to minimizing error, this term can be used in the
transition from grid-forming to grid-following mode. In this
w0 Jvi , (20)
case, the operation mode can be altered by changing 𝜇. As
√ outlined in [16], when 𝜇=0 grid-forming operation converts
where w0 = ( LC )−1 is defined as nominal angular fre- to grid-following operation. In this case, due to not have any
quency, J := R(𝜋∕2), 90√ o rotation matrix representing the
term in (19) to adjust voltage, the converter cannot maintain
complex imaginary unit −1 into ℝ2 . vi is the inverter out- the desired output voltage in absence of a grid. However, this
put voltage which will be equal to vi∗ in the steady-state. transition may be needed during disturbances for protection
∙ Phase correction or synchronization feedback is defined as purposes.
[ ]
𝜂 Di vi −R(𝜃)i0,i . (21) From (19), it is concluded that the basic idea of this con-
troller is minimizing voltage magnitude error and phase error,
This term tries to minimise the phase between inverter out- measured by voltage magnitude and synchronization feedback,
put current i0 and reference current, calculated based on the respectively, and operates with the conventional harmonic oscil-
active and reactive power set-points p∗i and qi∗ , respectively. lator frequency w0 and voltage vi at steady-state [50]. To have a
In other words, it can provide tracking power set-points. Syn- precise performance and meet stability considerations, the selec-
chronization gain 𝜂 plays a key rule in the stability of entire tion of controller parameters is expanded in the following sub-
system as well as precise performance of controller. Its design section.
procedure will be outlined in the following section. To set
the reference powers, p∗i and qi∗ , which should be consis-
tent with power-flow equations, the information of line and 3.3.1 Parameter selection procedure for dVOC
angle set-point are required. This information is not known
by the inverter operator. Hence, with a decentralized con- In this section, a guideline for choosing 𝜂 and 𝛼 which guaran-
trol method, as shown in detail in [49], the power-flow equa- tees the stability of the system are represented. A main criterion
AZIZI AGHDAM AND AGAMY 843

ation since AHO controller is dispatchable. It can track the pre-

specified active power, reactive power, voltage and frequency
set-points in either grid-connected and islanded. Additionally, it
can be applicable to three-phase implementations due to having
symmetric and planar dynamics, and intrinsically embed orthog-
onal signals [37]. Similar to other conventional oscillators, AHO
consists of linear oscillator part with natural frequency and two
negative conductance -𝜎 and 𝜖2 𝜎 which systematically inject
energy into the circuit to sustain oscillations. Also, a nonlinear
state dependent voltage and current sources, denoted by gv and
gi [80] are defined as

gv = 𝜖𝛼(vc2 + 𝜖2 iL2 )𝜖iL gi = 𝛼(vc2 + 𝜖2 iL2 ). (27)

Dynamic equation of AHO by considering the capacitor voltage

FIGURE 6 A three-phase inverter with the AHO [80] vc and scaled inductor current iL defined as follows

1
v̇ c = [−i − gi + 𝜎vC − u1 ],
to satisfy stability in dVOC-based inverter is considering the C L
(28)
sufficient time separation between inverters dynamic and trans- 1
i̇L = [vc − gv + 𝜎𝜖(𝜖iL ) − 𝜖u2 ],
mission line dynamics [23], [24]. While line dynamics could be L
ignored in conventional power system due to the slow dynamic
∗
of generators, it leads to instability in inverter-dominant system where [u1 , u2 ]= ki R(𝜙)(i𝛼𝛽 − i𝛼𝛽 ) and i𝛼𝛽 is the measured
[24]. In [24], to consider the transmission line effect on dVOC inverter output current represented in 𝛼𝛽 stationary framework.
performance and the stability analysis, singular perturbation the- ∗
Reference currents (i𝛼𝛽 ) can be defined as
ory was used as the standard quasi-steady-state approximation
cannot be applicable. Besides, in [23], [24], a boundary for 𝛼 2
and 𝜂 values based on specified conditions were outlined which i𝛼∗ = (v p∗ + v𝛽 q∗ ), (29)
3||v𝛼𝛽 || 𝛼
is summarized as following
2
i𝛽∗ = (v p∗ − v𝛼 q∗ ), (30)
∑
N v ∗j 1
∗2
vmin 3||v𝛼𝛽 || 𝛽
||Y jk |||1 − cos(𝜃∗jk )| + 𝛼 < (1 + cos(𝜃∗ )) ∗2 𝜆2 (L)
j =1
vk∗ 2 vmax √
where ||v𝛼𝛽 || = v𝛼2 + v𝛽2 is the magnitude of modulation sig-
c
− c𝜂 < , nals [v𝛼 , v𝛽 ], which can be obtained from
𝜌||𝛾||(c + 5||𝜅 − ||)
(26)
𝜁 kv ki
v̇ 𝛼 = 2 − ||v ||2 )v − w v −
(2Vnom (i − i ∗ ), (31)
where for all k, j ∈ N , ||Y jk || is the line admittances. 𝜃∗jk is k2v
𝛼𝛽 𝛼 nom 𝛽
C 𝛽 𝛽
𝜋
steady-state angle, 𝜃∗ ∈ [0, ] shows the existence of a maxi-
2 𝜁 kv ki
mal steady-state angle such that |𝜃∗jk | ≤ 𝜃∗ , vmin
∗2 ∗2 are
and vmax v̇ 𝛽 = 2 − ||v ||2 )v + w v −
(2Vnom 𝛼𝛽 𝛽 nom 𝛼 (i − i ∗ ), (32)
k2v C 𝛼 𝛼
the smallest and largest magnitude set-points, respectively. 𝜅 is
defined as diag(Di ) and 𝜆2 (L) is the second smallest eigenvalue
where 𝜁 denotes the convergence speed to steady state.
of the graph Laplacian L. The stability margin is represented by
c ∈ ℝ > 0 and ||𝛾||=||R(𝜃)T ||.
3.4.1 Parameter selection procedure for AHO
3.4 Andronov–Hopf oscillator
Similar to Lineard oscillator (VdP and DZO) design, to define
appropriate C and 𝜁 for AHO, an acceptable range for C an 𝜁
Another approach which is similar to dVOC (consensus-based
can be defined based on maximum power transition time 𝜏max ,
approach) is AHO. This oscillator emulates the dynamics of max
maximum voltage rise time trise , minimum voltage Vmin , maxi-
Andronov–Hopf systems [79] and uses the local inverter-
mum allowable frequency deviation |Δw|max , and nominal oscil-
level objectives which yields a straightforward design [37]. The
lation amplitude Xnom .
schematic of three-phase inverter based on AHO is illustrated in
Figure 6. This design contains explicit terms to define desirable
1 1
dynamic response such as voltage rise time and power transition Cmin ∶= √ , (33)
2
(2)Vmin |Δw|max
time. Similar to dVOC, AHO can be used in grid-forming oper-
844 AZIZI AGHDAM AND AGAMY

2
3Vnom
Cmax ∶= 𝜏 , (34)
ZSrated max

where Z is the total impedance including the line and filter

impedance. To define a sufficient convergence speed to steady
state 𝜁, as described in [37], trise
max
and Xnom are set to compute
𝜁min .

3
𝜁min ∶= max . (35)
2Xnom trise

Moreover, by further analysis and comparing dVOC and AHO

dynamic equations, the analogy of between the coefficients can
be explored, which helps us to define the dVOC gain (𝜂 , 𝛼 and
𝜇) as

kv k i 𝜁
𝜂= , 𝛼= , 𝜇 = 𝜁. (36)
C 𝜂 FIGURE 7 A three-phase inverter with the uVOC [16]

tal frequency positive- and negative-sequence components of

3.5 Unified virtual oscillator unbalanced or distorted grids. This enhanced synchronization
capability, combined with fast overcurrent limiting, facilitates
The results in [49], [24], [50], verified the GFM functionali- ride-through of both symmetric and asymmetric faults.
ties of dVOC or AHO for inverter-dominant power system.
However, these approaches are incapable to provide fault ride-
through (FRT) operation. Hence, to extend their applications, 3.5.1 Parameter selection procedure for uVOC
[16] offers a novel virtual oscillator controller namely uVOC.
This approach uses the fundamental dVOC characteristics and As uVOC follows the principle of dVOC and AHO, its design of
proposed a method which facilitated synchronization with an parameters is similar to the procedure outlined for dVOC. More
arbitrarily low grid voltage and fast over-current limiting. The detail of the design of this approach can be found in [16]. Sim-
control law of the uVOC is taken as ilarly, based on the desired steady-state response, the controller
𝜋
parameters 𝜂 and 𝜇 for 𝜙 = can be defined as following:
V̇ = w0 Jvi + 𝜂(i0 − i )e j 𝜙 + 𝜇(vi∗2 − vi2 ). (37) 2

Similar to (19), the uVOC control law consists of three part: har- N ph Δwmax Vmax
2
𝜂= , (38a)
monic oscillator, synchronization feedback as well as magnitude Prated
correction term. The magnitude correction term is exerted radi-
ally along the d-axis on the voltage vector and it is opposed by 2 𝜂 Qrated
𝜇= [ ], (38b)
either the error in real power output or the reactive power out- 2
N ph (2Vmax − Vnom
2 2
) − Vnom
4

put depending on the selection of 𝜙 [16]. A three-phase voltage

source converter (VSC) using uVOC is indicated in Figure 7. where the parameter definitions and restriction is similar to 3.4.
Another prominent feature of this design is that it does not So far, various oscillator implementations have been discussed.
require to switch back to back-up controller during fault like It can be concluded that all oscillators follow the same principle
other grid-forming method. Moreover, the design of this oscil- which is generating a limit cycle to control the power system
lator guarantees seamless transition between grid forming and non-linearly. The structure of all oscillators is comprised of a lin-
grid following operation by adding synchronization loop with- ear part, resonant RLC circuit and nonlinear current dependant
out PLL. Furthermore, it was shown that uVOC, like previous voltage source. The main difference between oscillators are the
oscillators is an instantaneous droop controller. Thus, it does nonlinear function that drives the oscillator. A brief summary of
not require any quasi steady-state or phasor approximation. In the main features of these approaches are presented in Table 2.
[81], uVOC is used in implementation of a multi-port converter As shown, VdP oscillator results in the generation of third order
(MPC) consisting of a grid forming port, a DC port, and a split- harmonics and it is not dispatchable. Thus, in grid-forming
phase-AC port which can be used as building blocks in hybrid application VdP and DZO should be modified or replaced by
AC-DC systems such as community microgrids enabling intra- consensus-based oscillators. While consensus-based oscillators
community and inter-community energy sharing. [82] investi- are dispatchable, except the uVOC, other oscillators are not
gated a double synchronous uVOC for grid-forming voltage equipped with fault-ride through capability and they can not be
source converters to achieve synchronization to the fundamen- reliable controller when the system is subjected to significant
AZIZI AGHDAM AND AGAMY 845

TABLE 2 Comparison of different oscillators

Features VdP DZO AHO dVOC[16] uVOC[16]

Dispatachability NO NO YES YES YES

Harmonics compensation NO YES YES YES YES
Virtual inertia emulation NO NO NO NO NO
Fault-ride-through NO NO NO NO YES
Extra loop for voltage regulation NO NO NO NO NO
Access to explicit term for L and C in design process YES YES YES NO NO

disturbance. Also, they can not provide inertia emulation. This

issue will be addressed with more details in section 6.1. Another
common point among all oscillator-based approaches is that
they are equivalent to droop control in steady-state with their
control law linearization by considering specific assumptions.
This feature making the oscillators to be compatible with legacy
system is scrutinized in following section.

4 COMPARISON OF DROOP CONTROL

AND VIRTUAL OSCILLATOR CONTROL

This section compares VOC and droop control. [39], [83–86]

show that these two methods have wide applications in syn-
chronization and power sharing between different inverters. [6]
investigated the efficiency of droop-based controller as a volt-
age and frequency regulator in the GFM operation. It was con- FIGURE 9 A three-phase inverter with droop control [80]
firmed that this method is compatible with legacy systems and
enables proper power sharing between inverters in both grid-
connected or islanded mode [38]. According to droop prin-
nominal value of frequency, voltage, real and reactive power,
ciple, in inductive networks, droop control linearly trades off
respectively. m p and nq are the droop coefficients. Figure 8 visu-
frequency deviation versus active power. Voltage deviation will
alizes the above equations. This method is based on phasor
be eliminated by applying appropriate reactive-power. To these
analysis and used the average active and reactive power to con-
ends, the droop law is given by
struct the control states governing inverters; whereas VOC is
𝜔 = 𝜔∗ + m p (P ∗ − P ), (39a) a time domain approach and reacts to instantaneous measure-
ment. Figure 9 depicts that droop controller requires low-pass
filter (LPF) to smooth the power fluctuations as well as real
v = v ∗ + nq (Q ∗ − Q), (39b) and reactive power computations. In contrast, VOC relies on
inverter output current to track the power set-points and does
where 𝜔, v, P and Q measure the frequency, voltage, real power not need LPF explicitly [26]. Using LPF in droop controllers
and reactive power, respectively. 𝜔∗ , v ∗ , P ∗ and Q ∗ denote the restrict the bandwidth of the controller as they need to sup-
press low frequency harmonics. Thus, droop control reacts with
a delay compared to VOC which does not require LPF. There-
fore, there is trade-off between the controller.
Further analysis of the VOC dynamic equation and leverag-
ing the method of periodic averaging in [39] confirm that droop
laws are intrinsically embedded in the nonlinear dynamics of
virtual oscillator within a slower time scale. The mathematical
analysis and derivation of droop relation are represented in [10,
11, 26]. According to these analyses, parameter of each oscilla-
tor explained in Section 3, are rewritten with respect to droop
coefficients and summarized in Table 3. It is worthwhile to men-
FIGURE 8 Droop characteristics (a) P − 𝜔 droop control. (b) Q − V tion that to extract these coefficients a set of assumptions that
droop control eliminate non-linearity are considered. For example, in dVOC, a
846 AZIZI AGHDAM AND AGAMY

TABLE 3 Extracting droop coefficients from virtual oscillator-based method

VdP [44] DZO AHO dVOC [50] UVOC [16]

kv ki kv ki kv ki 𝜂 𝜂
mp = mp = mp = mp = mp =
2CV ∗2 2CV ∗2 2CV ∗2 2V ∗2 2V ∗2
( )−1 2 − V ∗2 ) − V ∗4 ]
kk 𝛽V ∗3 kv ki kv ki 1 [2(Vmax
mq = − v i V ∗ − mq = − ( f (V ∗ ))−1 mq = mq = mq =
2𝜎 3 2𝜎 2𝜁CV ∗2 2𝛼V ∗2 2Qrated

small voltage magnitude deviation is assumed (||vi || ≈ vi∗ ) [50]. must be considered to guarantee the almost global asymptotic
Based on the analogy between VOC and droop control, two stability of oscillator. In other words, considering a sufficient
methods can be designed to produce similar steady-state char- time-scale separation between the inverter dynamics and line
acteristics; whereas their dynamic performance can be differ- dynamics to avoid any possible interaction between dynamics
ent [26]. [85] analyzed the time responses of the two methods of line and controller is crucial. Moreover, [35] focused on the
in more specific way. It compares the linearized form of two network topology impact and outlined several conditions for
models as well as the original models. With linearized models, ensuring the global asymptotic synchronization of Liénard-type
it was indicated that two methods have different modes while nonlinear oscillators in uniform Linear Time Invariant (LTI)
they have similar dominant modes which determine the equal electrical networks. A uniform network implies a networks that
transient response. The comparison, also, covers the original has identical per-unit-length impedances or identical resistance-
model of droop control and averaged VOC. In this case, similar inductance ratio (R∕L) for the interconnecting lines and ignor-
transient response was achieved though the steady state perfor- ing shunt elements. It is worth noting that purely resistive and
mance is different. The main reason is that the original VOC purely inductive networks are special cases of the uniform net-
dynamics have harmonic oscillations in steady state caused by works. The authors in [88] expanded the previous investigation
the non-sinusoidal characteristics of the virtual oscillator. Fur- on the global asymptotic synchronization and used heteroge-
thermore, [87] and [51] compare the time response of virtual neous LTI electrical networks as a framework. This research has
oscillator and droop controlled microgrids in the small-signal more reliability as the real networks are heterogeneous network
framework. [51] emphasised the effect of frequency and ampli- with different R∕L. It is concluded that VOC within a set of
tude regulation ranges on the efficiency of different approaches conditions can provide stability regardless of the network topol-
in terms of time response and energy consumption. It was ogy. Further research, [80], measures the impact of filter param-
shown when the terminal voltage frequency regulation range eters on the stability of system controlled by AHO. Based on
is large enough (i.e. Δ f ≥ 0.2Hz), VOC dynamic responses is the eigenvalue analysis, it is indicated that to guarantee the GFM
faster than droop control. Thus, in this case, virtual oscillator- stability, the filter ratio of resistance-inductance R f ∕L f , must
based controller reduces the transient energy losses as it reacts be above the critical value of R f ∕L f . For example, in a purely
faster. In lower frequency regulation range, however, droop- inductive network, the system is unstable as it has two right hand
based control showed better response as the voltage amplitude side eigenvalues while increasing R f ∕L f stabilize the system.
range increases. [51], also, showed the results are valid for dif- In addition to considering line dynamics and filter impedance,
ferent loads such as resistive load, a constant power load and a it is essential to have consistent primary response of controller
nonlinear load. in complex networks consisting of DGs with different primary
responses such as synchronous generators, droop-control and
VOC-based units. Due to having fast primary response of
5 OPERATIONAL CONSIDERATIONS controllers (i.e. VOC-based inverters), it may lead to impact the
stability of entire system [89]. This issue will be more challeng-
5.1 Stability analysis ing if the power capacity of GFM inverters by faster controller
is significant. In addition, in [80], the impact of the power refer-
The small-signal stability assessment and the sufficient con- ence on small-signal stability was discussed and showed that P ∗
ditions for stability of GFM inverters governed by VOC was has little impact on small-signal stability given P ∗ does not reach
discussed comprehensively in literature. In [42, 44, 49], the the maximum transferable power. While the majority of investi-
stability of virtual oscillator-based controller and its parameter gations focus on small-signal stability analysis, the behaviour of
selection criteria for ensuring synchronization and almost global system when subjected to large-signal disturbance have not been
stability were outlined. Particularly, [24] studied the effect of widely studied. Recently, in [90], transient stability of dVOC-
transmission line dynamics on GFM control for inverter-based based inverter in presence of severe transient disturbance
ac power systems. It was discussed that the electromagnetic was investigated and compared to droop controller. In this
transients of the transmission lines were negligible in the research, a first-order nonlinear power angle (𝛿) dynamic equa-
generator-dominant power system due to the slow dynamics of tion was derived from the non-uniform oscillator dynamics. To
the system. However, in inverter-based power system, ignoring assess the stability of the system in [90], the equilibrium point,
these dynamics leads to system instability. Hence, through the which implies the point that P ∗ = Pmax , is defined. This point
design of control gains, the dynamic model of transmission lines exists provided that P ∗ is less than the Pmax , maximum power
AZIZI AGHDAM AND AGAMY 847

transferred between inverter and infinite bus. Pmax is obtained tem. This is utilized for synchronization purpose only and will
from be deactivate in normal situation [16],[76]. Also, this method
is used for decoupling voltage and phase regulation loops [44],
VE [16]. These research emphasis that adding a virtual impedance
Pmax = 3 . (40)
XT does not affect the asymptotic synchronization or the stability
property of the network as the emulated resistance has identical
As it can be seen Pmax depends on inverter voltage V and effect as that of a real resistance in the network in the frequency
infinite bus voltage E as well as total reactance connecting range of interest. Furthermore, [91] proposed a novel method
inverter and infinite bus XT . When the system is subjected for pre-synchronization of AHO-based inverter. This investi-
to a large-signal disturbance (i.e. transmission line outage or gation unlike existing methods, alter the conventional AHO
load disturbance), according (40), V or XT may change which dynamics by adding comparative states, which define the time
impact on the Pmax and equilibrium point. In this case, system of grid connection, instead of using virtual impedance emu-
behaves in two different ways. The first situation is that equilib- lation. In [92], a hierarchical control structure for a smooth
rium point exists after disturbance, but 𝛿 could not reach this transition of microgrids from islanded mode to grid-connected
point before hitting the critical value. Therefore, the system mode was investigated. The authors proposed a control struc-
loses synchronism and will be unstable. On the contrary, if ture uses an average consensus protocol for the secondary con-
the system reaches the equilibrium point before exceeding the trol layer and AHO control as a primary controller. It was shown
critical angle, it can stabilize in steady state. To explain rea- that the proposed method can improve the power-sharing accu-
son causing instability, we can refer to incapability of dVOC to racy among different inverters in normal operation and pre-
retain a constant voltage regulation and the coupled reactive and synchronization provide a seamless transition from islanded to
active power control. In this case, consider a situation when the grid-connected mode.
system faces a remarkable step in load and voltage significantly
sags. As a consequence, the Pmax decreases and it may lead to sit-
uation that system will have no equilibrium point as P ∗ > Pmax . 5.3 Harmonics
Therefore, the voltage should be tightly maintained around
predefined voltage, which can be achieved by a well-designed The existence of harmonics in the non-linear virtual oscilla-
controller gains. It was shown that dVOC controller with dif- tor may restrict the application of VOC in the power system
ferent parameters have different droop characteristics that leads control, in particular GFM inverters control. [78] studied the
to different steady state voltage and power angles. Hence, con- impact of VOC in terms of harmonics on the GFM and GFL
sidering safe margin in control design to eliminate the adverse modes. It was shown that in the islanded mode, harmonics
effect of reactive power on the stability is crucial. More discus- caused by VOC can be bounded by appropriate design of con-
sions on the impact of coupled power control can be found in troller parameters. However, in grid-connected mode it can be
Section 6.2. more complicated as the grid parameters also influence control.
Additionally, in grid-tied the existence of harmonics may lead
to considerable circulating current between grid and inverter. It
5.2 Pre-synchronization of VOC-based also leads to inaccurate synchronization among inverters as well
inverter as inverter and grid. Hence, the harmonics should be considered
and mitigated properly. [78] investigated the source of harmon-
Generally, in addition to the stability analysis, an important ics such as natural delay in controller, non-linear terms in VOC,
issue that can influence the power system performance is pre- non-linear system components (e.g. non-linear load). The focus
synchronization of inverter voltage vinv and grid-side voltage of [78] is on the impact of VOC structure on the harmonics. It,
vg before switch connection. Ignoring synchronization leads to also, showed the harmonics mitigation by controller design. To
significant current spike which damage the hardware or trig- this end, the dynamic equation of VdP (5) is used and rewritten
ger the trip incorrectly [76]. Hence, considering this issue in as
VOC-based inverter is critical as these controllers are not pro-
tected by current limiter like current-controlled inverters. To v̈ − 𝜖𝜎(1 − 𝛽v 2 )v̇ + v = 0. (41)
deal with this issue, one approach is obtaining some degree
of synchronization before closing the switch by adjusting the As discussed in [44], the non-linear term in the VdP leads
state of oscillator to match grid oscillation [76]. For example, to low-order harmonic, particularly, 3rd-order harmonic. [38]
in [44], the state variables varied by altering oscillator capacitor proposed design based on operating in quasi-harmonic lim-
C and voltage scale factor kv . The second approach for limit- its harmonic 𝜖 → 0 and choosing appropriate value for L and
ing the current appearing during breaker connection, is apply- C . High value for C reduces the harmonic content signifi-
ing a virtual impedance. This approach emulates the impedance cantly while prolongs the response time of oscillator. Thus,
that can suppress surge in the current caused by differences there is a trade-off between response times, that is, rise time
between vg and vinv . This impedance does not have any impact and harmonics content. [38, 44], analysed the dynamic equa-
on the real current or voltage and overall stability of the sys- tion of VdP and expressed the relations of these parameters as
848 AZIZI AGHDAM AND AGAMY

follows. attempt, the conventional VOC is modified and a novel SMC

control is designed using the high-pass filter technique and lin-
𝜖𝜎
𝛿3∶1 = , (42) earized super twist algorithm.
8
6 6C
trise ≅ ≅ . (43)
w ∗ 𝜎𝜖 𝜎 6 LIMITATION OF VIRTUAL
OSCILLATOR-BASED CONTROLLER
Based on this approach harmonics can be bounded in a desir-
able range in islanded mode while they cannot be suppressed As discussed in the previous sections, VOC can benefit the
in grid-tied mode. Therefore, [44] utilized notch filters and GFM inverter control as it grants unique features such as faster
showed the efficiency of them for grid-tied application. How- dynamic response, accurate power-sharing and synchronization
ever, using of them in the islanded mode is not recommended etc. However, its implementation faces several limitations which
as it may affect the synchronization speed of the oscillators. It is restrict the employing of them in future inverter-dominant sys-
worthwhile noting that these filters cannot be effective to sup- tems. Thus, this section tries to address those challenges and
press harmonics arising from network-side voltage. [78] inves- outline possible solutions.
tigated the issue of harmonics in system controlled by VOC
and proposed a method based on the virtual impedance Zvir
selective harmonic approach. Moreover, the impact of high fre- 6.1 Inertial response
quency dynamics of oscillator-based VSC considering the con-
troller delay and the effect of PWM, and higher order output As synchronous generators have been replaced by static power
filter is considered. Consequently, distinction is made between converters, the level of inertia in power systems has decreased
converter-side and network-side current feedback for controller remarkably [3]. In low-inertia, providing frequency stability is
implementation while using an LCL filter. It was shown that significant concern [2]. Because in the low-inertia power sys-
network-side current feedback can lead better response in har- tems, the imbalance in the generation power and demand power
monic mitigation based on Zvir . In contrast to these aforemen- leads to rapid and significant frequency deviation leading to
tioned works that focus on the VOC as the source of distortion, system instability [3]. To solve this issue the GFM operation
[93–95] showed the significant capability of harmonics rejection of inverters was introduced, which forms voltage phasor and
of VOC. It was approved that the oscillator voltage retains low operates synchronously to the grid voltage [6]. This operation
THD regardless the type of load (i.e. linear or nonlinear load), can inherently provide inertial response. In contrast, the grid-
due to its limit cycle nature. The reason is that the limit cycle, following inverters cannot provide inertia since there is a delay
which is designed to rotate in fundamental frequency, can only in injection power due to frequency measurement. Although
been influenced by the fundamental component of current. As VOC is classified as a GFM approach, it does not provide iner-
VOC itself has strong load decoupling capability without any tial response. In [6], the dynamic equation of VdP oscillator-
additional controller such as multiple Proportional-Resonant based controller is compared to the 3rd-order model of the syn-
(PR) controllers. Therefore, the control design will be much chronous machine (simplified model). It was shown that VOC
simpler with less digital processor’s computational burden. Fur- does not have state variable representing inertia time constant.
thermore, they analysed the harmonic in the point of com- As a result, this method constructs the voltage phasor, but there
mon coupling (PCC) and proposed a method similar to [78],[96] is no virtual inertia property to resists against changes. This defi-
based on a frequency adaptive virtual impedance scheme that is ciency will be more challenging in ultra low-inertia systems [3]
compatible with VOC. The results confirmed PCC voltage har- and restricts the application of VOC. Therefore, to take the
monics mitigation and proper harmonic power sharing. To over- advantages of VOC and extend its application, in this work, a
come sub-harmonics in the output voltage caused by frequency method which synthesizes virtual inertia (VI) to conventional
variation in grid-connected mode, [97] proposed a modified dVOC is proposed.
structure of VOC enabling frequency compensation scheme. To this end, inverter must inject/absorb the amount of
Moreover, most recent investigations introduced dVOC and power, PVI , which equals to difference between generation
AHP which have less harmonic content in comparison to VdP- power and demand power, to compensate frequency deviation
based control. [78] mentioned that virtual-impedance based har- caused by power imbalance. Consequently, the power set-point
monic mitigation, which is used for VdP, is also applicable for for dVOC will be updated by adding PVI to previous set-point.
consensus-based methods. While VOC has mainly studied in ac The schematic of enhanced dVOC, which is enable to provide
system, [98], conducted a research on the VOC of distributed VI, is illustrated in Figure 10. To obtain PVI , the following equa-
power filters to attenuate the dc bus ripples. The design is based tion (44), which is derived from the swing equation [2], is used.
on the concept of synchronization and power filter is controlled
to emulate a current mode oscillator to synchronize with the
d Δw
disturbance current on the dc bus. Furthermore, [99] proposed PVI = KI + KD Δw, (44)
dt
a novel method that combines sliding mode control (SMC) and
VOC to eliminate voltage harmonics and unbalance caused by
non-linear and unbalanced loads in a three-phase system. In this Δw = w − wo, (45)
AZIZI AGHDAM AND AGAMY 849

where k is design parameters. Based on (44)-(47), PVI can be

obtained, which is zero in normal situation. When disturbance
occurs, however, the output of VI controller will be updated and
leads to a new dVOC power set-point as

∗ = P∗ + P .
Pnew (49)
VI

Thus, the inverter can react to frequency deviation rapidly and

compensate it. As shown in Figure 11, adding VI control to the
conventional dVOC improves the minimum frequency when a
large transient, that is, significant step-load (1pu) at t = 20s is
applied to the system. Similarly, [103] added VI capability to
conventional AHO by adding a proportional derivative (PD)
controller to the VOC regulator. The equivalent inertia con-
stant of the VOC inverter is obtained by combination with
the swing equation of the SG. Also, the effect of altering the
equivalent VOC inertia constant using small-signal analysis and
FIGURE 10 A three-phase dVOC-based inverter with virtual inertia
experiment was investigated. This approach provides inertial
response for AHO-based controller with constant inertia coeffi-
d Δw cient. Although the efficiency of constant VI control was proved
where w denotes the measured frequency. Δw and repre-
dt in [103], this design leads to longer settling time, power oscilla-
sent the change in frequency and the corresponding rate-of-
tions and more exchanged power during disturbance when iner-
change of frequency (RoCoF ), respectively. KI , inertia coeffi-
tia coefficient is relatively large [104]. Conversely, applying low
cient, is responsible for smoothing the slope of changes while
inertia coefficient decreases the impact of VI control. There-
KD , damping inertia affects minimum/maximum frequency.
fore, to provide a more efficient VI control based on VOC,
According to (44), to obtain PVI , w needs to be measured.
an adaptive design is added to the controller in this work. The
This measurement was obtained from PLL in [101] which led to
adaptive function distinguishes different intervals and update
instability in weak grid. However, in this work, PLL-based fre-
the inertia and damping coefficients regarding the sign of fre-
quency measurement is replaced by using conventional droop
quency deviation and its correspondent slope. Therefore, larger
equation embedded in VOC control law
KI and smaller KD will be applied to system when frequency
deviates from nominal values (Δw.Δw ̇ > 0) to slow down the
w = wo − k p (Pout − P ∗ ), (46) RoCoF and arrests frequency deviation to hit the lower value.
Conversely, larger KD and smaller KI will be applied to system to
where Pout and P ∗ are the inverter output power and reference accelerate system dynamics when the frequency returns back to
power, respectively. k p is droop coefficient which in dVOC- the nominal frequency (Δw.Δw ̇ ≤ 0). These objectives are for-
based control can be defined as mulated as follows
𝜂
kp = . (47) ( )
||V𝛼𝛽 ||2 ⎧ d Δw d Δw d Δw
⎪ 1 + k 1 K0,I Δw > 0, | |>M
dt dt dt
KI = ⎨ , (50a)
d Δw d Δw
As PLL is replaced by power-frequency equation (46), the ⎪k2 K0,I Δw ≤ 0, | |>M
injected delay of PLL for frequency measurement is removed. ⎩ dt dt
Besides, in contrast to grid-following control, VOC does not
require power calculation and measuring of the output volt-
⎧(1 − k Δw)K Δw
d Δw
> 0, |
d Δw
|>M
age of invert. Thus, it does not apply any delay which restricts ⎪ 3 0,D dt dt
inertial response. The impact of proposed method in frequency KD = ⎨ , (50b)
d Δw d Δw
response improvement is shown in Figure 11. ⎪(1 + k3 Δw)K0,D Δw ≤ 0, | |>M
⎩ dt dt
Another challenge with VI control implementation is the
need for derivative term in PVI calculation. Using derivative
often induced noise in the system leads to inaccurate response where K0,I and K0,D are the initial value of inertia and damp-
[2]. Hence, the derivative term can be replaced by algebraic ing coefficients, respectively. k1 , k2 , k3 are adaptive loop gains
equation, (48), inspired by [102]. Future investigation will study and M is a parameter determining the dead -band for VI loop
the effect of derivative term substitution by algebraic equations. response. To obtain K0,I and K0,D , conventional droop-based
equation method (39) is used as follows
√
d Δw −KI + KI2 − 4kΔw(KD Δw − (Pout − P ∗ ) K0,D ΔP
= , (48) 𝜏f = , K0,D = . (51)
dt 2kΔw K0,I wΔw
850 AZIZI AGHDAM AND AGAMY

following inverters [38]. In GFM mode, virtual oscillators such

as VdP or DZO cannot be utilized since their structures do
not inherently provide power control and they lack the com-
patibility with secondary control and grid-synchronization [44].
As a result, there is no control on the line stress, losses. To
employ VOC in more efficient way and applying it in grid-
connected mode, [44] suggested a method in which VOC-based
controller is modified to define tertiary control to track the
real and reactive power references by adding the correspond-
FIGURE 11 The impact of VI control on the frequency response [100] ing loops. [105] expanded VOC applications in grid-connected
mode by introducing a method in which a decentralized adaptive
battery energy storage system (BESS) with a primary control
scheme is used. Moreover, [52], [106] used Proportional Inte-
gral (PI) controllers to regulate the output power by tuning the
current feedback gain of VOC. In the mentioned approaches,
adjusting the current feedback gain affects the inverter terminal
voltage magnitude, which exhibits the cross-coupling control of
powers. This is because VOC utilizes the time domain dynam-
ics to control the inverter and does not have explicit active
and reactive power control loops like droop-based control [44],
[90]. In [45], the control power of VdP oscillator and coupled-
FIGURE 12 The impact of control methods on the RoCoF [100]
power control effect was investigated. It is shown that although
using PI control in power control is a simple and functional, any
change in reference active or reactive power causes the interfer-
ing impact on the two loops, which leads to significant fluctua-
tion in reactive/active power outputs. Hence, a decoupling con-
trol is required to reduce this interaction which directly impact
the system performance. To this end, the authors in [45] pro-
posed a novel method by introducing a complex parameter K
determined to track and control a given reference power. To
remove the coupled power control, an extra loop is applied
which decouple real and reactive power by cross-coupling terms
Rx and Lx . [45] outlined the exact value of Rx and Lx is difficult
to acquire, because of the nonlinear relationship between K and
the power output.
FIGURE 13 Exchanged power during step-load=1pu [100]
Besides, it complicated the controller design and imple-
mentation. In more recent investigation, [107], the authors,
Figures 11–13 show the efficiency of the proposed method simultaneously regulated both active and reactive power in a
in presence of large transient. As it can be seen the adaptive single-phase system controlled by VdP. They proposed using
VI control improves the RoCoF significantly and reduces the PI controller to apply variable kv and ki to provide desired
amount of exchanged power. However, the nadir frequency is power references. Also, the recommended using ramp-change
lower than non-adaptive control. To justify this it is important instead of a step-change in the reference power set-point or
to notice that additional considerations such as defining the pri- decline the PI controller’s response to remove the coupling
ority of objectives in addition to conditions (50) is needed. For effect among variables. Further research [49]-[50], reported
instance, the focus of our design is on the exchanged power dVOC which is synthesized in a top-down system-level design
reduction as well as deceleration of frequency changes with lim- procedure and provides power control based on arbitrary
iting frequency deviation. Although the proposed method can set-points. Similarly, another dispatchable oscillators benefiting
alleviate the issues with non-inertial response of VOC, for uti- from AHO is discussed in [37]. Compared to conventional
lizing VO-based approaches in low-inertia system further stud- VOC, consensus-based approaches are simpler controllers and
ies on the optimal exchanged power and optimal VOC design provide dispatchable power without extra loop. However, the
are required. coupling effect of active and reactive power in these methods
have still existed. This issue is discussed by reviewing the sim-
ulation of a three-phase inverter controlled by AHO from [37]
6.2 Power control in Figure 14. As Figure 14.a indicates, a step change in active
power set-point yields to unwanted change in reactive power.
In conventional applications, VOC is mostly used for syn- Since there is no compensation for reactive power in AHO-
chronization and power sharing adjustment of multiple grid- based controller, voltage amplitude will change. The Figure 14.b
AZIZI AGHDAM AND AGAMY 851

TABLE 4 Main limitations of basic virtual oscillator control and proposed

solutions

Limitations Solutions

Lack of effective fault ∙ Introducing uVOC and fast over-current

handling capability limiting via a series compensator [16]
[16]
Lack of power control in ∙ Active and reactive power set-points tracking
VdP and DZO [45, [45, 52]
52, 108] ∙ Introducing consensus-based controller [37, 49,
50]
∙ Introducing modified Vdp and DZO [44, 45]
Grid synchronization ∙ Providing seamless transient between GFM and
[16, 44] GFL mode by using conventional relay to
control the connection switch
∙ Employing virtual impedance method and
uVOC [16]
Lack of third order ∙ Using Notch filter in grid-connected mode [44]
compensation in VdP ∙ Using consensus-based oscillator [16, 37, 49, 50]
[44, 78] ∙ Using uVOC and virtual impedance-based
selective harmonic current suppression method
FIGURE 14 Example simulation of power control interaction in [78]
AHO-based inverter [37]
Interaction of power ∙ Considering pure inductive or resistive network
loop control [44, 90] [37, 44, 50, 108]
∙ Using emulation of virtual impedance [16, 44]
confirms this issue within VOC. Thus, this controller can reg-
Lack of inertial response ∙ Modifying dVOC and AHO by adding external
ulate the frequency accurately by injecting/absorbing active
[6] virtual inertia control loop [100, 103]
power in inductive grid, but it cannot removes the control
Compatibility with other ∙ Tuning the control parameters such that all the
loop interaction and voltage deviation. In recent work, [108]
controller [44] units exhibit identical proportional load-sharing
addressed the issue within reactive power control in AHO- in steady-state [111]
based inverter and introduced a modified AHO providing
compensation for voltage. This is achieved by using the power
flow equations and an integral control law to determine the
value of the added voltage set-point. Therefore, the inverter deficiency severely limits practical application of these con-
output voltage can be regulated according to the power flow. trollers in GFM operation. Recently, [16] has proposed a struc-
Also, the steady-state power tracking errors are eliminated. ture enabling FRT by synchronization with arbitrarily low grid
Furthermore, this issue can relatively be alleviated by putting a voltage and fast over-current limiting. Besides, the compari-
limitation on reference active power or operation points which son of uVOC and other GFM methods such as power syn-
violates the controller robustness feature. Another solution chronization control (PSC) and droop-based control demon-
for power decoupling is considering the network structure strated that unlike these methods, uVOC does not require a
as pure inductive/ resistive, which may not be valid in many backup PLL for fault management. Moreover, the advantages of
cases [2]. Coupled P and Q control, also has impact on the uVOC would be more significant in over-current limiting com-
transient stability [90]. This research demonstrates the reac- pared to droop-based GFM converters. The reason is that in
tive power regulation in dVOC in presence of a disturbance droop-based method, over-current protection may be achieved
causes a voltage drop and deteriorates the transient stability. by directly limiting the reference to the inner current control
To control the adverse effect of coupling, dVOC parame- loop which eventually may lead to loss of synchronization and
ters are designed to have a tight voltage regulation which instability [109]. Furthermore, in the recent investigation on
diminishes the controller robustness. Up to this point, the dVOC, [110] investigated adding reference-current limiting to
negative impact of power control loop interactions has not well a reduced-model of dVOC which decreases the computational
addressed in the VO-based controller. Thus, there is a need burden as well as increases the reliability of the system.
for further studies considering this issue in more detail.This
challenge and other outlined challenges within basic VOC
and the corresponding proposed solutions are represented in 6.4 Heterogeneous controllers
Table 4.
Another important factor in VO-based controllers’ utilization
in power system control is the compatibility of them with the
6.3 Fault management legacy system. [44] discussed this issue. It outlined that in a net-
work consisting of droop -based and VOC-based units as well
Another challenge with VOC implementation is that there is as synchronous generators with different primary responses, to
no compatible FRT strategy for VO-based controllers. This avoid system instability primary control response time should
852 AZIZI AGHDAM AND AGAMY

be assessed. This issue will be more important when the capac- dVOC and uVOC have different representations. The impact
ity of VOC-based DGs is significant compared to the rest of of the controller parameters on the stability and the dynamic
system [44]. Faster primary response provided by part of the response of the system were outlined.
system (i.e. VOC-based unit) may destabilize the entire of the Also, studies of different VO-based approaches show
system [89]. The majority of research related to VOC, the iden- the functionality of consensus-based oscillators compared to
tical controllers are considered. However, in [112], the perfor- VdP and DZO as they are dispatchable. More specifically,
mance of VOC-based inverter connecting to conventional SG, the inverter output harmonics for the different approaches
which is controlled based on conventional droop controller, were discussed and it was concluded that consensus-based
was investigated. Moreover, [111] scrutinized multiple invert- approaches have lower harmonics compared to VdP. Besides,
ers operating in parallel with heterogeneous control techniques. VOC control law was compared to droop control and it was
The authors analyzed a system consisting of two parallel con- concluded that they have the same characteristics at steady state.
nected inverters including droop-controlled inverter and VO- The embedded droop principals of VOC were extracted and
controlled inverter. To guarantee a proportional power shar- used as substitution of PLL for frequency measurement. The
ing between the two parallel-connected inverters controlled operational consideration and limitations of these oscillators
by various control techniques, parameters of both the droop- were presented. In order to employ these controllers in real
control and VOC-parameters should be adjusted such that all applications further investigations are required. The effects of
the units exhibit identical proportional load-sharing in steady changes in inductance to resistance ratios, transformer param-
state. In addition to power sharing studies in a system with het- eter tolerances and degradation, and other network dynamic
erogeneous controller, [113] investigated the small-signal sta- behaviors on the operation of VO-based controllers need to
bility study of a coupled synchronous generator and VOC- be evaluated. Also, the robustness of system against controller
based inverter system. System’s small-signal stability at different parameter variations requires further studies. Additionally, as
inverter penetration levels and the impact of rotational inertia, VOC dynamic response is fast, studies of the impact of such
reactive power support, and filter parameters on stability were response on component stresses and hence the reliability and
investigated. The results highlight the possible incompatibility lifetime of inverters and associated filters are necessary. Fur-
of various controller in the system and it emphasised on the thermore, to improve the VOC performance and utilizing it in
need to develop next generation stabilizing grid-forming con- large-scale system, intelligent approaches such as model predic-
trollers. In [114], N parallel-connected VOC inverters with het- tive control (MPC) and artificial intelligence (AI) can be com-
erogeneous power ratings are considered. It was shown the het- bined with VOC.
erogeneous filter parameters and power rating do not impact the
performance of VOC. However, the performance of VOC with ACKNOWLEDGEMENTS
heterogeneous controllers is challenging and need to be consid- This work was supported in part by the New York State
ered. Research and Development Authority (NYSERDA) under
Grant 137936.

7 CONCLUDING REMARKS AND

DATA AVAILABILITY STATEMENT
DIRECTIONS FOR FUTURE RESEARCH
Data sharing not applicable – no new data generated, or the
article describes entirely theoretical research.
This paper presented a review of different VO-based strate-
gies in the GFM operation. The analogy between power system
components and conventional oscillators behaviour motivated CONFLICT OF INTEREST
power engineers to use virtual oscillators for control in power The authors have declared no conflict of interest.
systems. It was shown that VOC enhances the synchroniza-
tion speed and power sharing accuracy as it does not need LPF ORCID
or power calculation. Additionally, in VOC, primary control is Sima Azizi Aghdam https://orcid.org/0000-0002-1835-0661
combined with voltage control loop which simplifies the imple-
mentation. Moreover, VOC provides a wide region of opera- REFERENCES
tion compared to droop-based control, which operates in the 1. U.s. energy information administration. annual energy outlook 2017; u.s.
limited area close to synchronous speed. VOC can outperform department of energy: Washington, dc, usa, (2017). accessed on 21 June
2017
machine-based emulation methods, which need high level of
2. Tamrakar, U., Shrestha, D., Maharjan, M., Bhattarai, B.P., Hansen, T.M.,
numerical analysis where high-order machine is emulated. Also, Tonkoski, R.: Virtual inertia: Current trends and future directions. Appl.
machine-based method may not be suitable for inverter control Sci. 7(7), 654 (2017). Available from: https://www.mdpi.com/2076-
conventional generators have features different from inverters. 3417/7/7/654
In this work, different oscillators utilized in power system con- 3. Milano, F., Dörfler, F., Hug, G., Hill, D.J., Verbič, G.: Foundations
and challenges of low-inertia systems (invited paper). In: 2018 Power
trol and the main criteria for their design were investigated. It
Systems Computation Conference (PSCC), pp. 1–25. IEEE, Piscataway
was shown that all approaches consist of two linear and nonlin- (2018)
ear oscillator implemented on the digital controller. In all meth- 4. Bevrani, H., Ise, T., Miura, Y.: Virtual synchronous generators: A sur-
ods the explicit terms for defining L and C are available while vey and new perspectives. Int. J. Electr. Power Energy Syst. 54, 244–254
AZIZI AGHDAM AND AGAMY 853

(2014). Available from: https://www.sciencedirect.com/science/article/ 25. Li, Y., Vilathgamuwa, D.M., Loh, P.C.: Design, analysis, and real-time test-
pii/S0142061513003062 ing of a controller for multibus microgrid system. IEEE Trans. Power
5. Matevosyan, J., Badrzadeh, B., Prevost, T., Quitmann, E., Ramasubra- Electron. 19, 1195–1204 (2004)
manian, D., Urdal, H., et al.: Grid-forming inverters: Are they the key 26. Johnson, B., Rodriguez, M., Sinha, M., Dhople, S.: Comparison of virtual
for high renewable penetration?. IEEE Power Energy Mag. 17(6), 89–98 oscillator and droop control. In: 2017 IEEE 18th Workshop on Control
(2019) and Modeling for Power Electronics (COMPEL), pp. 1–6. IEEE, Piscat-
6. Unruh, P., Nuschke, M., Strauß, P., Welck, F.: Overview on grid- away (2017)
forming inverter control methods. Energies 13(10), 2589 (2020). Avail- 27. Brouillon, J.S., Colombino, M., Groß, D., Dörfler, F.: The effect of
able from: https://www.mdpi.com/1996-1073/13/10/2589 transmission-line dynamics on a globally synchronizing controller for
7. Zhong, Q.C., Weiss, G.: Synchronverters: Inverters that mimic syn- power inverters. In: 2018 European Control Conference (ECC), pp.
chronous generators. IEEE Trans. Ind. Electron. 58(4), 1259–1267 2242–2247. IEEE, Piscataway (2018)
(2011) 28. Sinha, M., Dhople, S., Johnson, B., Ainsworth, N., Dörfler, F.: Nonlinear
8. Beck, H.P., Hesse, R.: Virtual synchronous machine. 2007 9th Interna- supersets to droop control. In: 2015 IEEE 16th Workshop on Control
tional Conference on Electrical Power Quality and Utilisation, pp. 1–6. and Modeling for Power Electronics (COMPEL), pp. 1–6. IEEE, Piscat-
IEEE, Piscataway (2007) away (2015)
9. D’Arco, S., Suul, J.A., Fosso, O.B.: Control system tuning and stabil- 29. Dhople, S.V., Johnson, B.B., Hamadeh, A.O.: Virtual oscillator control
ity analysis of Virtual Synchronous Machines. In: 2013 IEEE Energy for voltage source inverters. In: 2013 51st Annual Allerton Conference
Conversion Congress and Exposition, pp. 2664–2671. IEEE, Piscataway on Communication, Control, and Computing (Allerton), pp. 1359–1363.
(2013) IEEE, Piscataway (2013)
10. Guerrero, J.M., de Vicuna, L.G., Matas, J., Castilla, M., Miret, J.: A wire- 30. Joshi, S.K., Sen, S., Kar, I.N.: Synchronization of coupled oscilla-
less controller to enhance dynamic performance of parallel inverters in tor dynamics. IFAC-PapersOnLine 49(1), 320–325 (2016). Avail-
distributed generation systems. IEEE Trans. Power Electron. 19(5), 1205- able from: https://www.sciencedirect.com/science/article/pii/
1213 (2004) S2405896316300738
11. De Brabandere, K., Bolsens, B., Van den Keybus, J., Woyte, A., Driesen, J., 31. Dörfler, F., Bullo, F.: Exploring synchronization in complex oscillator net-
Belmans, R.: A voltage and frequency droop control method for parallel works. In: 2012 IEEE 51st IEEE Conference on Decision and Control
inverters. IEEE Trans. Power Electron. 22(4), 1107–1115 (2007) (CDC), pp. 7157–7170. IEEE, Piscataway (2012)
12. Chen, Y., Hesse, R., Turschner, D., Beck, H.P.: Comparison of meth- 32. Chopra, N., Spong, M.W.: On synchronization of kuramoto oscillators.
ods for implementing virtual synchronous machine on inverters. Renew. In: Proceedings of the 44th IEEE Conference on Decision and Control,
Energy Power Quality J., 1(10), pp. 734–739 (2012) pp. 3916–3922. IEEE, Piscataway (2005)
13. Majumder, R., Chaudhuri, B., Ghosh, A., Majumder, R., Ledwich, G., 33. Simpson Porco, J.W., Dörfler, F., Bullo, F.: Synchronization and power
Zare, F.: Improvement of stability and load sharing in an autonomous sharing for droop-controlled inverters in islanded microgrids. Automat-
microgrid using supplementary droop control loop. IEEE Trans. Power ica 49(9), 2603–2611 (2013). Available from: https://www.sciencedirect.
Syst. 25(2), 796–808 (2010) com/science/article/pii/S0005109813002884
14. Zhang, L., Harnefors, L., Nee, H.P.: Power-synchronization control of 34. Dörfler, F., Bullo, F.: Synchronization and transient stability in power
grid-connected voltage-source converters. IEEE Trans. Power Syst. 25, networks and non-uniform kuramoto oscillators. In: Proceedings of the
809–820 (2010) American Control Conference. IEEE, Piscataway (2009)
15. Grid-Friendly Matching of Synchronous Machines by Tapping into the 35. Sinha, M., Dörfler, F., Johnson, B., Dhople, S.: Synchronization of liénard-
DC Storage. type oscillators in uniform electrical networks. In: 2016 American Control
16. Awal, M.A., Husain, I.: Unified virtual oscillator control for grid-forming Conference, pp. 4311–4316. IEEE, Piscataway (2016)
and grid-following converters. IEEE J. Emerg. Sel. Topics Power Elec- 36. Tôrres, L.A.B., Hespanha, J.P., Moehlis, J.: Synchronization of identical
tron. 9(4), 4573-4586 (2021) oscillators coupled through a symmetric network with dynamics: A con-
17. Kammer, C., Karimi, A.: Decentralized and distributed transient control structive approach with applications to parallel operation of inverters.
for microgrids. IEEE Trans. Control Syst. Technol. 27(1), 311–322 (2019) IEEE Trans. Autom. Control 60(12), 3226–3241 (2015)
18. Huang, L., Xin, H., Dorfler, F.: H∞ −Control of grid-connected convert- 37. Lu, M., Dutta, S., Purba, V., Dhople, S., Johnson, B.: A grid-compatible
ers: design, objectives and decentralized stability certificates. IEEE Trans. virtual oscillator controller: Analysis and design. In: 2019 IEEE Energy
Smart Grid 11(5), 3805–3816 (2020). Available from: https://doi.org/10. Conversion Congress and Exposition (ECCE), pp. 2643–2649. IEEE,
1109/TSG.2020.2984946 Piscataway (2019)
19. Huang, L., Xin, H., Wang, Z., Wu, K., Wang, H., Hu, J., et al.: A vir- 38. Johnson, B.B., Sinha, M., Ainsworth, N.G., Dörfler, F., Dhople, S.V.: Syn-
tual synchronous control for voltage-source converters utilizing dynam- thesizing virtual oscillators to control islanded inverters. IEEE Trans.
ics of DC-link capacitor to realize self-synchronization. IEEE J. Emerg. Power Electron. 31(8), 6002–6015 (2016)
Sel. Topics Power Electron. 5(4), 1565–1577 (2017) 39. Sinha, M., Dörfler, F., Johnson, B.B., Dhople, S.V.: Uncovering droop
20. Jiang, Y., Bernstein, A., Vorobev, P., Mallada, E.: Grid-forming frequency control laws embedded within the nonlinear dynamics of van der
shaping control for low-inertia power systems. IEEE Control Syst. Lett. pol oscillators. IEEE Trans. Control Network Syst. 4(2), 347–358
5(6), 1988–1993 (2021) (2017)
21. Rathnayake, D.B., Akrami, M., Phurailatpam, C., Me, S.P., Hadavi, S., 40. Johnson, B.B., Dhople, S.V., Hamadeh, A.O., Krein, P.T.: Synchronization
Jayasinghe, G., et al.: Grid forming inverter modeling, control, and appli- of nonlinear oscillators in an lti electrical power network. IEEE Trans.
cations. IEEE Access 9, 114781–114807 (2021) Circuits Syst. I Regul. Pap. 61(3), 834–844 (2014)
22. Denis, G., Prevost, T., Panciatici, P., Kestelyn, X., Colas, F., Guillaud, 41. Thai, T.T., Raisz, D., Monti, A., Htut, H.M.: A load power
X.: Review on potential strategies for transmission grid operations based sharing strategy for virtual oscillator-based photovoltaics genera-
on power electronics interfaced voltage sources. In: 2015 IEEE Power tion in islanded microgrid. IFAC-PapersOnLine 53(2), 12906–12911
Energy Society General Meeting, pp. 1-5. IEEE, Piscataway (2015) (2020). Available from: https://www.sciencedirect.com/science/article/
23. Colombino, M., Groß, D., Dörfler, F.: Global phase and voltage syn- pii/S2405896320327671
chronization for power inverters: A decentralized consensus-inspired 42. Johnson, B.B., Dhople, S.V., Hamadeh, A.O., Krein, P.T.: Synchroniza-
approach. In: 2017 IEEE 56th Annual Conference on Decision and Con- tion of parallel single-phase inverters with virtual oscillator control. IEEE
trol (CDC), pp. 5690–5695. IEEE, Piscataway (2017) Trans. Power Electron. 29(11), 6124–6138 (2014)
24. Groß, D., Colombino, M., Brouillon, J.S., Dörfler, F.: The effect of 43. Johnson, B.B., Dhople, S.V., Cale, J.L., Hamadeh, A.O., Krein, P.T.:
transmission-line dynamics on grid-forming dispatchable virtual oscilla- Oscillator-based inverter control for islanded three-phase microgrids.
tor control. IEEE Trans. Control Network Syst. 6(3), 1148–1160 (2019) IEEE J. Photovoltaics 4(1), 387–395 (2014)
854 AZIZI AGHDAM AND AGAMY

44. Awal, M.A., Yu, H., Tu, H., Lukic, S.M., Husain, I.: Hierarchical control 65. Simpson Porco, J.W., Dörfler, F., Bullo, F.: Droop-controlled invert-
for virtual oscillator based grid-connected and islanded microgrids. IEEE ers are kuramoto oscillators. IFAC Proc. Vol. 45(26), 264–269
Trans. Power Electron. 35, 988–1001 (2020) (2012). Available from: https://www.sciencedirect.com/science/article/
45. Raisz, D., Thai, T.T., Monti, A.: Power control of virtual oscillator con- pii/S1474667015348448
trolled inverters in grid-connected mode. IEEE Trans. Power Electron. 66. Tôrres, L.A.B., Hespanha, J.P., Moehlis, J.: Synchronization of identical
34(6), 5916–5926 (2019) oscillators coupled through a symmetric network with dynamics: A con-
46. Sepulchre, R., Paley, D., Leonard, N.: Stabilization of planar collective structive approach with applications to parallel operation of inverters.
motion: All-to-all communication. IEEE Trans. Autom. Control 52, 811– IEEE Trans. Autom. Control 60(12), 3226–3241 (2015)
824 (2007) 67. Hu, J., Ma, H.: Synchronization of the carrier wave of parallel three-phase
47. Jadbabaie, A., Lin, J., Morse, A.S.: Coordination of groups of mobile inverters with virtual oscillator control. IEEE Trans. Power Electron.
autonomous agents using nearest neighbor rules. IEEE Trans. Autom. 32(10), 7998–8007 (2017)
Control 48(6), 988–1001 (2003) 68. Shi, X., Huang, Q., Jing, S., Yi, J.: Performance analysis of parallel inverter
48. Cucker, F., Smale, S.: Emergent behavior in flocks. IEEE Trans. Autom. based on virtual oscillator control. In: 2018 IEEE Innovative Smart
Control 52(5), 852–862 (2007) Grid Technologies - Asia (ISGT Asia), pp. 318–322. IEEE, Piscataway
49. Colombino, M., Groß, D., Brouillon, J.S., Dörfler, F.: Global phase and (2018)
magnitude synchronization of coupled oscillators with application to the 69. Tu, Y., Su, J., Du, Y., Yang, X., Wang, S.: Improved virtual oscillator con-
control of grid-forming power inverters. IEEE Trans. Autom. Control trol for parallel operation inverter systems. In: 2016 IEEE 8th Inter-
64(11), 4496–4511 (2019) national Power Electronics and Motion Control Conference (IPEMC-
50. Seo, G.S., Colombino, M., Subotic, I., Johnson, B., Groß, D., Dör- ECCE Asia), pp. 2063–2067. IEEE, Piscataway (2016)
fler, F.: Dispatchable virtual oscillator control for decentralized inverter- 70. Li, M., Wei, B., Xie, P., Tan, S., Guerrero, J.M., Vasquez, J.C.: Syn-
dominated power systems: Analysis and experiments. In: 2019 IEEE chronization and current sharing for nonlinear-oscillator-based invert-
Applied Power Electronics Conference and Exposition (APEC), pp. 561– ers in islanded three-phase microgrid. In: 2019 IEEE Energy Conver-
566. IEEE, Piscataway (2019) sion Congress and Exposition (ECCE), pp. 2566–2571. IEEE, Piscat-
51. Shi, Z., Li, J., Nurdin, H.I., Fletcher, J.E.: Comparison of virtual oscilla- away (2019)
tor and droop controlled islanded three-phase microgrids. IEEE Trans. 71. Hu, K., wai Chung, K.: On the stability analysis of a pair of van der pol
Energy Convers. 34(4), 1769–1780 (2019) oscillators with delayed self-connection, position and velocity couplings.
52. Ali, M., Nurdin, H.I., Fletcher, J.E.: Output power regulation of a virtual AIP Adv. 3, 112118 (2013)
oscillator controlled inverter. In: 2018 IEEE 18th International Power 72. Lu, M., Seo, G.S., Sinha, M., Rodriguez, F., Dhople, S., Johnson, B.: Adap-
Electronics and Motion Control Conference (PEMC), pp. 1085–1090. tation of commercial current-controlled inverters for operation with vir-
IEEE, Piscataway (2018) tual oscillator control. In: 2019 IEEE Applied Power Electronics Confer-
53. Strogatz, S.H.: Nonlinear Dynamics and Chaos: With Applications to ence and Exposition (APEC), pp. 3427–3432. IEEE, Piscataway (2019)
Physics, Biology, Chemistry and Engineering. Westview Press, Boulder 73. Sinha, M., Poon, J., Johnson, B.B., Rodriguez, M., Dhople, S.V.: Decen-
(2000) tralized interleaving of parallel-connected buck converters. IEEE Trans.
54. Fan, C.X., Jiang, G.P., Jiang, F.H.: Synchronization between two com- Power Electron. 34(5), 4993–5006 (2019)
plex dynamical networks using scalar signals under pinning control. IEEE 74. Stan, G.B.: Global analysis and synthesis of oscillations: A dissipativity
Trans. Circuits Syst. I Regul. Pap. 57(11), 2991–2998 (2010) approach (2005)
55. Varela, F., Lachaux, J.P., Rodriguez, E., Martinerie, J.: The brainweb: Phase 75. Stan, G.B., Sepulchre, R.: Analysis of interconnected oscillators by dissi-
synchronization and large-scale integration. Nat. Rev. Neurosci. 2, 229– pativity theory. IEEE Trans. Autom. Control 52(2), 256–270 (2007)
239 (2001) 76. Opila, D.F., Kintzley, K., Shabshab, S.C., Phillips, S.: Virtual oscillator
56. Jadbabaie, A., Lin, J., Morse, A.S.: Coordination of groups of mobile control of equivalent voltage-sourced and current-controlled power con-
autonomous agents using nearest neighbor rules. IEEE Trans. Autom. verters. Energies 12, 298 (2019)
Control 48(6), 988–1001 (2003) 77. Costa, D.A., Tôrres, L.A.B., Silva, S.M., Conti, A.D., Brandão, D.I.: Param-
57. Kuramoto, Y.: Self-entrainment of a population of coupled non-linear eter selection for the virtual oscillator control applied to microgrids. Ener-
oscillators. In: Araki, H. (ed.) International Symposium on Mathematical gies 14, 1818 (2021)
Problems in Theoretical Physics. pp. 420-422. Springer, Berlin Heidelberg 78. Awal, M.A., Yu, H., Husain, I., Yu, W., Lukic, S.M.: Selective harmonic
(1975) current rejection for virtual oscillator controlled grid-forming voltage
58. Hamadeh, A., Stan, G.B., Sepulchre, R., Goncalves, J.: Global state syn- source converters. IEEE Trans. Power Electron. 35(8), 8805–8818 (2020)
chronization in networks of cyclic feedback systems. IEEE Trans. Autom. 79. Kuznetsov, Y.A.: Andronov-Hopf bifurcation. Scholarpedia 1(10), 1858
Control 57(2), 478–483 (2012) (2006). revision #81241
59. Bharath, R., Slotine, J.J.E.: Nonlinear observer design and synchroniza- 80. Lu, M., Purba, V., Dhople, S., Johnson, B.: Comparison of droop con-
tion analysis for classical models of neural oscillators. arXiv:1310.0479 trol and virtual oscillator control realized by Andronov-Hopf dynamics.
(2013) In: IECON 2020 The 46th Annual Conference of the IEEE Industrial
60. Dörfler, F., Bullo, F.: Synchronization in complex networks of Electronics Society, pp. 4051–4056. IEEE, Piscataway (2020)
phase oscillators: A survey. Automatica 50(6), 1539–1564 (2014). 81. Awal, M.A., Bipu, M.R.H., Chen, S., Khan, M., Yu, W., Husain, I.: A grid-
Available from: https://www.sciencedirect.com/science/article/pii/ forming multi-port converter using unified virtual oscillator control. In:
S0005109814001423 2020 IEEE Energy Conversion Congress and Exposition (ECCE), pp.
61. Joshi, S.K., Sen, S., Kar, I.N.: Synchronization of coupled oscilla- 4300–4307. IEEE, Piscataway (2020)
tor dynamics. IFAC-PapersOnLine 49(1), 320–325 (2016). Avail- 82. Awal, M.A., Rachi, M.R.K., Yu, H., Husain, I., Lukic, S.: Double syn-
able from: https://www.sciencedirect.com/science/article/pii/ chronous unified virtual oscillator control for asymmetrical fault ride-
S2405896316300738 through in grid-forming voltage source converters (2021)
62. Chopra, N., Spong, M.W.: On exponential synchronization of kuramoto 83. Sinha, M., Dörfler, F., Johnson, B.B., Dhople, S.V.: Virtual oscillator con-
oscillators. IEEE Trans. Autom. Control 54(2), 353–357 (2009) trol subsumes droop control. In: 2015 American Control Conference
63. Arcak, M.: Passivity as a design tool for group coordination. IEEE Trans. (ACC), pp. 2353-2358. IEEE, Piscataway (2015)
Autom. Control 52(8), 1380–1390 (2007) 84. Awal, M.A., Yu, H., Lukic, S., Husain, I.: Droop and oscillator based grid-
64. Tôrres, L.A.B., Hespanha, J.P., Moehlis, J.: Power supply synchronization forming converter controls: A comparative performance analysis. Front.
without communication. In: 2012 IEEE Power and Energy Society Gen- Energy Res. 8, 168 (2020). Available from: https://www.frontiersin.org/
eral Meeting, pp. 1–6. IEEE, Piscataway (2012) article/10.3389/fenrg.2020.00168
AZIZI AGHDAM AND AGAMY 855

85. Shi, Z., Nurdin, H.I., Fletcher, J.E., Li, J.: Similarities between virtual 102. Hou, X., Sun, Y., Zhang, X., Lu, J., Wang, P., Guerrero, J.M.: Improve-
oscillator controlled and droop controlled three-phase inverters. In: 2018 ment of frequency regulation in vsg-based ac microgrid via adap-
IEEE 18th International Power Electronics and Motion Control Confer- tive virtual inertia. IEEE Trans. Power Electron. 35(2), 1589–1602
ence (PEMC), pp. 434–439. IEEE, Piscataway (2018) (2020)
86. Gurugubelli, V., Ghosh, A., Panda, A.K., Rudra, S.: Implementation and 103. Li, J., Fletcher, J.E., Holmes, D.G., McGrath, B.P.: Developing a machine
comparison of droop control, virtual synchronous machine, and virtual equivalent inertial response for a virtual oscillator controlled inverter in
oscillator control for parallel inverters in standalone microgrid. Int. Trans. a machine-inverter based microgrid. In: 2020 IEEE Energy Conversion
Electr. Energy Syst. 31(5), e12859 (2021) Congress and Exposition (ECCE), pp. 4314–4321. IEEE, Piscataway
87. Shi, Z., Li, J., Nurdin, H., Fletcher, J.: Transient response comparison of (2020)
virtual oscillator controlled and droop controlled three-phase inverters 104. Meng, J., Wang, Y., Fu, C., Wang, H.: Adaptive virtual inertia control of
under load changes. IET Gener. Transm. Distrib. 14, (2020) distributed generator for dynamic frequency support in microgrid. In:
88. Sinha, M., Dörfler, F., Johnson, B.B., Dhople, S.V.: Synchronization of 2016 IEEE Energy Conversion Congress and Exposition (ECCE), pp.
liénard-type oscillators in heterogenous electrical networks. In: 2018 1–5. IEEE, Piscataway (2016)
Indian Control Conference (ICC), pp. 240–245. IEEE, Piscataway (2018) 105. Hazra, P., Hadidi, R., Makram, E.: Dynamic study of virtual oscillator con-
89. Paquette, A.D., Divan, D.M.: Virtual impedance current limiting for trolled inverter based distributed energy source. In: 2015 North American
inverters in microgrids with synchronous generators. IEEE Trans. Ind. Power Symposium (NAPS), pp. 1–6. IEEE, Piscataway (2015)
Appl. 51(2), 1630–1638 (2015) 106. Muhammad, A., Nurdin, H.I., Fletcher, J.E.: Simultaneous regulation of
90. Yu, H., Awal, M.A., Tu, H., Husain, I., Lukic, S.: Comparative tran- active and reactive output power of parallel-connected virtual oscillator
sient stability assessment of droop and dispatchable virtual oscillator controlled inverters. In: IECON 2018-44th Annual Conference of the
controlled grid-connected inverters. IEEE Trans. Power Electron. 36(2), IEEE Industrial Electronics Society, pp. 4051–4056. IEEE, Piscataway
2119–2130 (2021) (2018)
91. Lu, M., Dutta, S., Purba, V., Dhople, S., Johnson, B.: A Pre- 107. Ali, M., Nurdin, H.I., Fletcher, J.E.: Dispatchable virtual oscillator con-
synchronization Strategy for Grid-forming Virtual Oscillator Controlled trol for single-phase islanded inverters: Analysis and experiments. IEEE
Inverters. In: 2020 IEEE Energy Conversion Congress and Exposition Trans. Ind. Electron. 68(6), 4812–4826 (2021)
(ECCE), pp. 4308–4313. IEEE, Piscataway (2020) 108. Heins, T., Tran, T., Raisz, D., Monti, A.: Power control of Andronov-
92. Sowa, I., Tran, T.T., Heins, T., Raisz, D., Monti, A.: An average consen- Hopf oscillator based distributed generation in grid-connected micro-
sus algorithm for seamless synchronization of Andronov-Hopf oscillator grids. In: Sattler, K.U., Nguyen, D.C., Vu, N.P., Long, B.T., Puta, H.
based multi-bus microgrids. IEEE Access 9, 90441–90454 (2021) (eds.) Advances in Engineering Research and Application, pp. 675–687.
93. Yu, H., Awal, M.A., Tu, H., Du, Y., Lukic, S., Husain, I.: A virtual Springer International Publishing, Cham (2021)
impedance scheme for voltage harmonics suppression in virtual oscillator 109. Paquette, A.D., Divan, D.M.: Virtual impedance current limiting for
controlled islanded microgrids. In: 2020 IEEE Applied Power Electron- inverters in microgrids with synchronous generators. IEEE Trans. Ind.
ics Conference and Exposition (APEC), pp. 609–615. IEEE, Piscataway Appl. 51(2), 1630–1638 (2015)
(2020) 110. Ajala, O., Lu, M., Johnson, B., Dhople, S., Domínguez García, A.: Model
94. Luo, S., Wu, W., Koutroulis, E., Chung, H.S.H., Blaabjerg, F.: A new reduction for inverters with current limiting and dispatchable virtual
virtual oscillator control without third-harmonics injection for DC/AC oscillator control. arXiv:2105.09754 (2021)
inverter. IEEE Trans. Power Electron. 36(9), 10879–10888 (2021) 111. Ali, M., Sahoo, A., Nurdin, H.I., Ravishankar, J., Fletcher, J.E.: On
95. Thai, T.T., Raisz, D., Monti, A., Ponci, F., Ahmadifar, A.: Voltage har- the power sharing dynamics of parallel-connected virtual oscillator-
monic reduction using virtual oscillator based inverters in islanded micro- controlled and droop-controlled inverters in an ac microgrid. In: IECON
grids. In: 2018 18th International Conference on Harmonics and Quality 2019-45th Annual Conference of the IEEE Industrial Electronics Soci-
of Power (ICHQP), pp. 1–6. IEEE, Piscataway (2018) ety, vol. 1, pp. 3931–3936. IEEE, Piscataway (2019)
96. Lin, J.L.: Virtual oscillator control of distributed power filters for selective 112. Tayyebi, A., Groß, D., Anta, A., Kupzog, F., Dörfler, F.: Interac-
ripple attenuation in dc systems. IEEE Trans. Power Electron. 36, 8552– tions of grid-forming power converters and synchronous machines.
8560 (2021) arXiv:1902.10750 (2019)
97. Ashwin, K., Umanand, L., Subba Reddy, B.: Simulation and analysis of 113. Siraj Khan, M.M., Lin, Y., Johnson, B., Sinha, M., Dhople, S.: Stability
virtual oscillator control for parallel connected single phase inverters. Assessment of a System Comprising a Single Machine and a Virtual Oscil-
In: 2020 IEEE International Conference on Electronics, Computing and lator Controlled Inverter with Scalable Ratings. In: IECON 2018-44th
Communication Technologies (CONECCT), pp. 1–6. IEEE, Piscataway Annual Conference of the IEEE Industrial Electronics Society, pp. 4057–
(2020) 4062. IEEE, Piscataway (2018)
98. Duarte, J.L., Roes, M., Xu, X., Wang, F.: Virtual oscillator control 114. Khan, M., Lin, Y., Johnson, B., Purba, V., Sinha, M., Dhople, S.: A
as applied to dc microgrids with multiple sources. In: 2020 IEEE reduced-order aggregated model for parallel inverter systems with virtual
9th International Power Electronics and Motion Control Conference oscillator control. In: 2018 IEEE Workshop on Control and Modeling
(IPEMC2020-ECCE Asia), pp. 2007–2011. IEEE, Piscataway (2020) for Power Electronics (COMPEL), pp. 1–6. IEEE, Piscataway (2018)
99. Tran, T., Raisz, D., Monti, A.: Harmonic and unbalanced voltage compen-
sation with VOC-based three-phase four-leg inverters in islanded micro-
grids. IET Power Electron. 13, 2281–2292 (2020)
100. Aghdam, S.A., Agamy, M.: Adaptive virtual inertia synthesis via enhanced
dispatchable virtual oscillator controlled grid-tied inverters. In: IECON How to cite this article: Azizi Aghdam, S., Agamy,
2021-47th Annual Conference of the IEEE Industrial Electronics Soci- M.: Virtual oscillator-based methods for grid-forming
ety, pp. 1–6. IEEE, Piscataway (2021) inverter control: A review. IET Renew. Power Gener. 16,
101. D’Arco, S., Suul, J.A.: Equivalence of virtual synchronous machines and
frequency-droops for converter-based microgrids. IEEE Trans. Smart
835–855 (2022). https://doi.org/10.1049/rpg2.12398
Grid 5(1), 394–395 (2014)