IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 70, NO.

1, JANUARY 2023 441

A New Virtual Inductance Control Method for

Frequency Stabilization of Grid-Forming Virtual
Synchronous Generators
Yaqian Yang , Jiazhu Xu , Member, IEEE, Chang Li , Weiming Zhang,
Qiuwei Wu , Senior Member, IEEE, Ming Wen, and Frede Blaabjerg , Fellow, IEEE

Abstract—Frequency stabilization is the premise of guar- issues, more and more distributed renewable energy generations
anteeing grid-friendly integration of virtual synchronous are penetrating into modern power systems [1]. Low-inertia and
generator (VSG). Based on that premise, this article, fo- poor-damping are remarkable characteristics of modern power
cused on frequency stability, establishes the small signal
model of grid-forming VSG system. At first, the mechanism systems due to the grid-integration of renewable energy. That
of frequency oscillation occurring in system of active- and characteristic directly leads to fast rate of change of frequency
reactive-power coupling is analyzed by the defined feed- (RoCoF) and inevitable frequency nadir (FN) [2], which threat-
back effect factor in this article. Besides, a new dynamic ens secure operation of the ac power systems.
model is established for identifying dynamic interaction of Virtual synchronous generator (VSG) control, which emu-
voltage magnitudes and frequency by means of feedback
effect. The analytical results of established models agree lates the inertia and damping characteristics of conventional
with that of eigenvalue analysis. Furthermore, a new virtual synchronous generator (SG), has been an emerging technique
inductance control strategy is proposed to mitigate the un- to improve inertia and damping of the system [3]. At present,
stable oscillation of frequency and powers, enhance damp- the study of VSG are mainly focused on two aspects, i.e., control
ing performance, and improve stability margins. Unlike the [4]–[7], and stabilization [8]–[16].
conventional virtual inductance control, which is reliable on
dual-loop control framework, the proposed virtual induc- For example, a virtual inertia control strategy was proposed
tance control in this article is based on principle of energy to provide inertia for ac power grid and improve FN [4]. How-
conservation and can be applicable for the grid-forming ever, this type of inertial control links dc voltage coupled with
inverter without inner dual-loop control structure. Finally, frequency, which is at the cost of drop of dc voltage. To rectify
the proposed modeling as well as virtual inductance control
that drawback, a new compensator was proposed to alleviate
method is experimentally verified.
the offset of dc voltage [5]. To overcome the shortcomings
Index Terms—Feedback effect, frequency offset (FO), fre- originated from fixed control parameters, an adaptive inertial
quency stabilization, rate of change of frequency (ROCOF), control strategy was proposed to improve the stability of virtual
virtual inductance control, virtual synchronous generator
(VSG).
synchronous machine (VSM) under various grid strengths [6]–
[7]. Besides of this, an improved virtual inertia control was pro-
I. INTRODUCTION posed to suppress frequency fluctuation of three-phase voltage
NVIRONMENTAL pollution as well as energy crisis is source inverters [8]. No matter how the improved VSG control is
E getting more and more attentions. To cope with those implemented, the core idea is focused on the emulation of rotor
swing equation of conventional SG. However, the premise that
VSG provide inertia and damping support for the system is that
Manuscript received 2 August 2021; revised 23 October 2021 and VSG itself or grid-integration VSG system can maintain stable
15 December 2021; accepted 18 January 2022. Date of publication 9
February 2022; date of current version 22 August 2022. This work operation.
was supported in part by the China Scholarship Council under Grant For example, a new modeling and parameters design method-
202006130038. (Corresponding author: Jiazhu Xu.) ology have been developed for VSG to guarantee secure oper-
Yaqian Yang and Jiazhu Xu are with the Hunan University, Changsha
410082, China (e-mail: [email protected]; [email protected]). ation of the system [9]. However, it is not involved in voltage
Chang Li and Qiuwei Wu are with the Technical University of stability or frequency stability. To fill in this gap, voltage and
Denmark, 2800 Copenhagen, Denmark (e-mail: [email protected]; frequency stability analysis were discussed with a grid-forming
[email protected]).
Weiming Zhang is with the Shenzhen Power Supply Company, Ltd., VSG [10]. However, the proposed magnitude-phase motion
Shenzhen 518000, China (e-mail: [email protected]). equation (MPME) simplifies the model and does not consider
Ming Wen is with the Economic Research Institute, State Grid the reactive power control coupling dynamics, which plays a
Hunan Electric Power Company, Changsha 410082, China (e-mail:
[email protected]). predominant role in the system stability.
Frede Blaabjerg is with the Aalborg University, 9220 Aalborg, Den- It was pointed out that grid-forming VSG can maintain
mark (e-mail: [email protected]). good stability even when attaching to weak grid [13], [14].
Color versions of one or more figures in this article are available at
https://doi.org/10.1109/TIE.2022.3148749. However, the stability mechanism is clarified from impedance
Digital Object Identifier 10.1109/TIE.2022.3148749 model. The study in this article shows that there is a risk
0278-0046 © 2022 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.

Authorized licensed use limited to: China Three Gorges University. Downloaded on August 31,2023 at 08:40:58 UTC from IEEE Xplore. Restrictions apply.
442 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 70, NO. 1, JANUARY 2023

of divergent oscillation appearing, when VSG is attached

to power grid of resistance-inductance impedance, due to
negative damping and positive feedback formed in the sys-
tem. To fill in the gaps, this article builds a more com-
plete MPME model to clarify the cause of frequency instabil-
ity. Besides, the physical significance is discussed in various
perspectives.
Although there have been several virtual impedance control
strategies put forward to mitigate the oscillation of inverter
or grid-forming VSG [15]–[17], all of them are realized by
formulating the control links between current signal and voltage
signal. However, dual-loop control structure is necessary for Fig. 1. Topology and control strategy of grid-forming VSG.
the implementation of virtual impedance. Therefore, to the best
knowledge of the author, virtual impedance control is seldom II. MECHANISM ANALYSIS OF FREQUENCY OSCILLATION OF
proposed for grid-forming VSG without dual-loop control struc- GRID-FORMING VSG
ture. Therefore, this article fills in the gaps.
The major contributions of this article are organized as In this section, mechanisms of both frequency oscillation and
follows. power coupling are discussed by the small signal modeling.
1) Based on the established small signal modeling of grid-
forming VSG system, the motion dynamics of both Ro- A. Power Coupling Model of Grid-Forming VSG System
CoF and frequency offset (FO) can be judged by feedback In this part, small signal modeling is developed for analysis
effect. It is found that frequency stability of grid-forming of frequency oscillation and power couplings. Fig. 1 shows the
VSG system is closely related with the transients of typical grid-forming VSG [10], which includes the topology
RoCoF and FO. And the feedback effect brought by inter- and control strategy. Lv means introduced virtual inductance,
nal relationship between RoCoF and FO gives physical which will be discussed in later sections. As the dc link voltage
insights to the frequency stability. is maintained by energy storage system, it is regarded as a
2) A new coupling model between voltage magnitude and constant value. The grid-forming VSG is connected to power
frequency is established to clarify the feedback interac- grid via transmission lines and grid impedance. Total impedance
tion between voltage and frequency of grid-forming VSG of lines and grid is denoted as Lg and Rg . Lf and Rf represent the
system, and the results of this analytical model agree with inductance and parasitic resistance of L filter, respectively. vabc
that of eigenvalue analysis. and iabc are three-phase voltage at the point of common coupling
3) A new virtual inductance control strategy is put forward and the three-phase current through L-filter, respectively.
to mitigate divergent oscillation of VSG frequency and According to Fig. 1, the active- and reactive-power can be
powers, and enhance damping performance of the system. obtained in dq frame
It should be noted that this type of virtual inductance 
is different from conventional virtual inductance control, Pe = 32 (ed id + eq iq ) s+ω
ωL
L
(1)
which is dependent on the dual-loop control architecture. Qe = 32 (−ed iq + eq id ) s+ω
ωL
L
The proposed virtual inductance control is based on prin-
ciple of energy conservations. It can be applied for VSG where ω L is the cutoff frequency of low pass filter, and the ed , eq ,
control structure without inner dual-loop control in this id , iq are the voltage and current of d- and q-axis, respectively.
article. Besides, the effectiveness of the proposed virtual In dq frame, the voltage and current can be represented by

inductance control strategy is demonstrated by vector ed = (R + Ls)id − ωn Liq
analysis and eigenvalue analysis. (2)
4) The proposed control algorithm and methodology can be eq = ωn Lid + (R + Ls)iq
⎧
analogously applicable to dc systems, which is our present ⎪ (R+Ls)(E cos δe −Ug )+ωn LE sin δe
⎨ id = (R+Ls)2 +(ωn L)2
focus.
(3)
The rest of this article are organized as follows. Section ⎪
⎩ iq = −ωn L(E cos δe −U2g )+(R+Ls)E sin δe
2
II discusses mechanism analysis of frequency oscillation in (R+Ls) +(ω L) n

grid-forming VSG system by the established small signal where R = Rf +Rg , L = Lf +Lg .
model. The dynamic behaviors of RoCoF and FO are studied The VSG control dynamics, which emulates the inertia and
by identifying feedback effect. And dynamic coupling model damping of conventional SG, can be expressed as
is established to clarify dynamic interaction between voltage
magnitudes and frequency of VSG. Section III introduces the Pref − Pe = Dp (ωe − ωn ) + Jp ωn (ωe − ωn )s (4a)
proposed virtual inductance control strategy to mitigate oscil- Qref − Qe = Jq (E − Uref )s (4b)
lation and improve damping performance. Section IV gives
the experiments validations. Finally, Section V concludes this 1
δe = ωe (4c)
article. s

Authorized licensed use limited to: China Three Gorges University. Downloaded on August 31,2023 at 08:40:58 UTC from IEEE Xplore. Restrictions apply.
YANG et al.: NEW VIRTUAL INDUCTANCE CONTROL METHOD FOR FREQUENCY STABILIZATION OF GRID-FORMING VSGs 443

initial to define the ith moment of Δω e as Δω e_i , thereby state

of ith moment of ΔE_i is dependent on the effect of NEω (s)
with disturbance of Δω e_i . ΔE_i will in turn impose effect on
state of Δω E through MωE (s). Therefore, the state of Δω e
will be accordingly updated, referred to as Δω e_i+1 . Therefore,
either positive feedback effect or negative feedback effect can
be identified by comparing the updated state Δω e_i+1 with the
earlier state Δω e_i . Moreover, the motion trajectory of VSG
frequency can be judged by comparison between renewed state
Δω e_i+1 and the original state Δω e_i . The coupling relation-
ships between the renewed state and initial state can be described
mathematically as follows:
Fig. 2. Closed-loop control system of VSG system.  ΔE_i
NEω (s) = Δω e_i
(9)
where Pref , Qref are the reference active- and reactive-power, MωE (s) = Δω e_i+1
ΔE_i
Pe , Qe are the output powers of the inverter. It should be noted
Δωe_i+1 = MωE (s)NEω (s)Δωe_i = GM N (s)Δωe_i (10)
that (4a) and (4c) mean frequency support for the system, and
(4b) represents voltage support for the system. where GMN (s) is the transfer function, which can depict the
Combing the equation of (1)–(3), the relationships among feedback effect on state variable Δω e and, thus, predict the
active power, reactive power, magnitudes, and phase angle of motion trajectory of VSG frequency, i.e.,
output voltage of inverter can be written as
GM N (s)
Pe = f (δe , E) (5a) 
GP E (s)GQδ (s)/ Jq s2 (Jp ωn s + Dp )
Qe = g(δe , E). (5b) = .
[1 + GP δ (s)/(Jp ωn s2 + Dp s)] (1 + GQE (s)/Jq s)
Linearizing(5) around the equilibrium points, one can obtain (11)
that To better identify the motion state of VSG frequency, k is

ΔPe = GP δ (s)Δδe + GP E (s)ΔE defined as feedback effect factor (FEF), i.e.,
(6)
ΔQe = GQδ (s)Δδe + GQE (s)ΔE
Δωe_i+1 |ω=ωCFP = kΔωe_i |ω=ωCFP (12)
where GPδ (s), GPE (s), GQδ (s), GQE (s) are the coupling terms
between voltage magnitude as well as phase angle and powers. where the parameter k is the magnitude of GMN (s) at the critical
According to the developed small signal model, closed-loop frequency point (CFP). Thus, k can be recalculated as

control system of active power-frequency is shown in Fig. 2. A > 1 (A > 0)
k = 10 20 (13)
< 1 (A < 0)
B. Dynamic Interactive Behavior and Feedback Effect
Between Frequency and Magnitude of VSG Output where A represents the magnitude of transfer function GMN (s)
Voltage as the unit of Decibel (dB).
By rearranging (12), the RoCoF can be formulated as
According to Fig. 2, the internal relationship from ΔE to Δω e
(ΔE divided by Δω e ) can be defined as Mω E (s). Likewise, the Δωe_i+1 Δωe_i
=k
relationship from Δω e to ΔE can be defined as NEω (s). By Δt Δt
substituting (5a) into (4a) and combing (4c), Mω E (s) is derived ⇓
by linearization dωe_i+1 dωe_i
=k . (14)
Δωe −GP E (s)/(Jp ωn s + Dp ) dt dt
MωE (s) = = . (7)
ΔE [1 + GP δ (s)/(Jp ωn s2 + Dp s)] By considering (13), a unified expression is derived to clarify
the dynamic characteristics of RoCoF, i.e.,
Likewise, substituting (5b) into (4b), NEω (s) can be obtained
as follows: dωe_n dωe_0
= kn (15)
ΔE −GQδ (s)/Jq s2 dt

dt
NEω (s) = = . (8)
Δωe 1 + GQE (s)/Jq s +∞ (k > 1)
lim k n = . (16)
n→+∞ 0 (k < 1)
It can be seen that (7) and (8) formulate the feedback loop
to identify the dynamic interactions between frequency (Δω e ) It can be seen in (16) that it forms positive feedback effect
and voltage magnitudes (ΔE). To more intuitively identify when k>1. And the RoCoF gets greater and greater, indicating
the dynamic interactive behavior between Δω e and ΔE, it is that divergent oscillation appears in the system. It is because
significant to define the ith moment and the (i+1)th moment that divergence oscillation of frequency indicates both RoCoF
of the state variables, e.g., Δω e and ΔE. Provided that it is and FO get larger and larger periodically. Vice versa, it forms

Authorized licensed use limited to: China Three Gorges University. Downloaded on August 31,2023 at 08:40:58 UTC from IEEE Xplore. Restrictions apply.
444 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 70, NO. 1, JANUARY 2023

Fig. 3. Eigenvalue loci and bode diagram with various Rg . (a) Domi-
nant root locus. (b) GMN (s). Fig. 4. Active power closed-loop transfer function block diagram.

TABLE I
SYSTEM SPECIFICATIONS the greater size of Rg is poorer for the stability of the system
and has a larger risk of subsynchronous oscillation. Fig. 3(b)
shows bode diagram of GMN (s), from which it can be seen that a
smaller Rg means lower magnitude of FEF. The lower magnitude
of FEF means lower risk of positive feedback effect, and this
will improve stability margins of the system. It can be inferred
that smaller Rg can make the system more stable, and frequency
stability is enhanced. According to the analysis of (14) and (16)–
(18), it forms a negative feedback effect when FEF is below 1
(0 dB). And in this case, both RoCoF and FO are declined over
time and gradually tend to be zero with the negative feedback
effect. Therefore, it can be inferred that it forms a negative
feedback when Rg = 0.8 Ω.
However, the magnitude of GMN (s) is greater than zero at the
CFP when Rg = 1.2 Ω and Rg = 1.6 Ω, divergent oscillation (in-
stability) of active power and frequency will appear. Therefore,
the proposed modeling with feedback effect is well illustrated
negative feedback effect when k<1. And the oscillation will by bodes analysis and eigenvalues.
be dampened and frequency gets convergent to an equilibrium To further discuss the frequency dynamics as well as system
point. stability, by considering (6), (8), and Fig. 2, an active power
In addition, the FO (δω e ) is defined as the frequency difference closed-loop control diagram is shown in Fig. 4. The whole
between the VSG frequency and the grid frequency, and the ith feedback paths include three parts, i.e., power-angle-related part
FO can be derived as shown (PARP) (hδ (s)), voltage-coupling-related part (VCRP) (hE (s))
i−1
ti+1 and damping part (hD (s)). The outputs of the three feedback
dωe loops can be regarded as three equivalent active powers (i.e.,
δωe_i = dt (i = 0, 1, 2 · · · n · · · + ∞). (17)
i=0 t
dt ΔPδ , ΔPE , and ΔPD ) that interact with the virtual rotor, as
i
depicted in Fig. 4. The dynamic behaviors of the virtual rotor
Another index to identify frequency stability is relative fre- are determined by the composite power (i.e., ΔPΣ = ΔPδ +
quency offset (RFO). For any positive integers i, j, and i > j, if ΔPE + ΔPD ), which can be used to evaluate the stability of the
RFO (Rω e ) is greater than zero, thereby the VSG frequency will system.
be divergent and cannot return back to equilibrium point due to Dynamic behaviors of system can be seen as superposition of
positive feedback effect, and vice versa dynamics of the feedback powers of three items, i.e.,
Rωe = δωe_i − δωe_ j (i > 0, j > 0, i > j) . (18) ΔP  = hδ (s)Δδe + hE (s)Δδe + hD (s)Δδe (19)
     
According to the abovementioned discussions, FEF (k) im- ΔPδ ΔPE ΔPD
poses a great effect on the motion trajectory of RoCoF and where the transfer functions hδ (s), hE (s), and hD (s) are given by
RFO. The FO gets larger and larger when FEF>1 that forms a ⎧
positive feedback effect. However, FO gets smaller and smaller ⎪
⎪ hδ (s) = ΔP
Δδe = GP δ (s)
δ

⎨
when FEF<1, and negative feedback effect is formed to stabilize −GQδ (s)GP E (s)/Jq s
hE (s) = ΔP ΔPE ΔE Δωe
Δδe = ΔE ∗ Δωe ∗ Δδe =
E
.
VSG frequency. ⎪
⎪
1+GQE (s)/Jq s
⎩ ΔPD
To validate the proposed modeling, Fig. 3 obtains the results of hD (s) = Δδe = Dp s
eigenvalue analysis and bode diagram of GMN (s) with different (20)
Rg . The system parameters are shown in Table I. In Fig. 3(a), the The amplitude and phase of three transfer functions at critical
dominant eigenvalues are moving from left-half plane towards frequency point can be derived, and accordingly, the vector
to the right-half plane as Rg gets enlarged, which indicates that diagrams of the active powers are carried out to investigate the

Authorized licensed use limited to: China Three Gorges University. Downloaded on August 31,2023 at 08:40:58 UTC from IEEE Xplore. Restrictions apply.
YANG et al.: NEW VIRTUAL INDUCTANCE CONTROL METHOD FOR FREQUENCY STABILIZATION OF GRID-FORMING VSGs 445

Fig. 5. Bode diagrams of hδ (s), hE (s), hD (s), hΣ (s) and the corre-
sponding vector diagram. (a) Rg = 1.2 Ω. (b) Rg = 0.8 Ω. (c) Cor-
responding vector diagram of (a). (d) Corresponding vector diagram Fig. 6. Root locus of the studied system considering various Rg and
of (b). Lg , (a) Rg changing from 0.3 to 1.6 Ω. (b) Lg varying from 6.6 to
13.2 mH.

stability of the system. It should be emphasized that Fig. 5(a) and

(b) shows the results of transfer functions of hδ (s), hE (s), and with Rg gradually increased from 0.3 to 1.6 Ω with Lg chosen
hD (s) with Rg = 0.8 Ω and Rg = 1.2 Ω, respectively. Fig. 5(c) as 8.8 mH. It can be seen that the coupling effect between
and (d) presents the corresponding vectors diagram. ΔPδ , ΔPE , active- and reactive-power becomes ever-increasing stronger.
ΔPD , and ΔPΣ represent PARP, VCRP, damping part, and total From partial enlarged view, a pair of dominant eigenvalues
power [18]. moves from the left half-plane to right half-plane, eventually
Note that the phase of total active power at CFP is the arrives at unstable regions. Eigenvalue analysis results coincide
focus of this article. As for synchronous machine, the feedback with that of vector analysis. Fig. 6(b) shows the root locus
active power can be divided into two items: damping power with variations of Lg from 6.6 to 13.2 mH with Rg = 1.2 Ω.
and synchronizing power, which are proportional to Δω e and The dominant eigenvalues move towards to the left half-plane
Δδ e , separately. That vectors, which are in the first quadrant as Lg gets greater, which indicates that the greater size of Lg
means that the system is stable because of positive damping enhance the stability of grid-forming VSG. In general, no matter
power and synchronizing power. From Fig. 5(c), the synthetic increasing Lg or decreasing Rg aims to lessen coupling effect
vector is located in the fourth quadrant with Rg = 1.2 Ω. between active- and reactive-power, which can improve stability
By decomposing the synthetic power ΔPΣ into damping and margins of the system.
synchronizing ones, the negative damping power exists, which
will trigger the oscillation instability. Different from Fig. 5(c) III. PROPOSED VIRTUAL INDUCTANCE CONTROL STRATEGY
FOR MITIGATION OF FREQUENCY OSCILLATION OF
and (d) shows the synthetic vector ΔPΣ , which provides positive
damping to suppress the frequency oscillation and compel the GRID-FORMING VSG
system frequency convergent to an equilibrium point. It can be In this section, a virtual inductance feed-forward control is
inferred that the grid resistance plays an important role in the proposed to enhance the stability of grid-forming VSG system,
stability of the system. Besides, greater grid resistance indicates thereby frequency oscillation of grid-forming VSG is mitigated
the stronger coupling effect between active- and reactive-power, by the virtual inductance control as well.
which will aggravate stability of grid-forming VSG.
A. Operation Principle of Virtual Inductance Control
C. Influence of Grid Parameters on Power Stability and
As discussed earlier, increase of inductance alleviates the
Frequency Oscillation coupling effect between active- and reactive-power. In order
To reveal the influence of grid parameters on stability of to achieve good dynamic behavior, a virtual inductance, which
grid-forming VSG system, eigenvalues with various Rg and Lg is emulates the real inductance, is developed. It should be noted
developed in Fig. 6. From Fig. 6(a), the eigenvalues movement that the virtual inductance is implemented by the control loop,

Authorized licensed use limited to: China Three Gorges University. Downloaded on August 31,2023 at 08:40:58 UTC from IEEE Xplore. Restrictions apply.
446 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 70, NO. 1, JANUARY 2023

Fig. 8. Root locus diagram of the system with virtual inductance Lv

varying from 0 to 17.6 mH.

Fig. 7. Small signal diagram closed loop system of grid-forming VSG

with virtual inductance control.

which does not add hardware costs. Besides, the principle as

well as architecture of virtual inductance is different from that
in [19] and [20], since that no architecture of dual-loop control
is contained in this type of VSG. The equivalent power with Fig. 9. Frequency responses of GMN_vir (s) with different virtual
inductances.
consideration of virtual inductance can be rewritten as

P_Lv = 32 (ed_Lv id_Lv + eq_Lv iq_Lv ) s+ω
ωL
L
3
(21)
Q_Lv = 2 (−ed_Lv iq _Lv + eq_Lv id_Lv ) s+ωL
ωL Fig. 8 shows the eigenvalue analysis result with virtual in-
ductance varying from 0 to 17.6 mH. Critical oscillatory modes
where id_Lv , iq_Lv is the d- and q-axis current with introduction
move toward to the left half-plane as virtual inductance gets
of virtual inductance, i.e.,
⎧ enlarged. It indicates that increase of virtual inductance can
⎨ id_Lv = [R+(Lv +L)s](E cos δe −U g )+ωn (Lv +L)E sin δe
strengthen damping performance and enhance the system sta-
[R+(Lv +L)s]2 +[ωn (Lv +L)]2
. bility.
⎩ iq_L = −ωn (Lv +L)(E cos δe −U2g )+[R+(Lv +L)s]E sin δe
v [R+(Lv +L)s] +[ωn (Lv +L)]2 Analogous to (7) and (8), the relationship between Δω e and
(22) ΔE, with introducing virtual inductance, can be reformulated as
Based on the energy conservation, the difference of powers
Δωe_vir
between with virtual inductance and without virtual inductance MωE_vir (s) =
can be derived as ΔE_vir
 −[GP E (s) + KP E_Lv (s)]
Pvir = P_Lv − Pe = m(δe , E) =
. (23) Jp ωn s + Dp + [GP δ (s) + KP δ_Lv (s)]/s
Qvir = Q_Lv − Qe = n(δe , E)
(25)
Linearizing (23) yields
 ΔE_vir
NEω_vir (s) =
ΔPvir = KP δ_Lv (s)Δδe + KP E_Lv (s)ΔE Δωe_vir
(24)
ΔQvir = KQδ_Lv (s)Δδe + KQE_Lv (s)ΔE −[GQδ (s) + KQδ_Lv (s)]/s
= . (26)
where ΔPvir and ΔQvir represent the control instructions feed Jq s + GQE (s) + KQE_Lv (s)
into the active-power control loop and reactive-power control Therefore, replacing the relationship between Δω E_i+1 and
loop, respectively. Δω E_i in (27), one can derive that
Therefore, effect of virtual inductance can be realized by the
derived control blocks in (24), as shown in Fig. 7. GM N _vir (s) = MωE_vir (s)NEω_vir (s). (27)
Fig. 9 displays the frequency response result of GMN_vir (s)
B. Discussion on Virtual Inductance to Dampen with changing various virtual inductances. The magnitude of
Frequency Oscillation
GMN_vir (s) is greater than 0 dB when Lv = 0. It means that
This part explores the principle of virtual inductance for system loses stabilization in this case, and it forms a positive
damping frequency oscillation of the system. Besides, it shows feedback effect that will destabilize frequency. However, as the
the positive damping effect brought by the proposed virtual virtual inductance Lv adding, the magnitude of GMN_vir (s) is
inductance control method. always below 0 dB, which ensures good stability margin of

Authorized licensed use limited to: China Three Gorges University. Downloaded on August 31,2023 at 08:40:58 UTC from IEEE Xplore. Restrictions apply.
YANG et al.: NEW VIRTUAL INDUCTANCE CONTROL METHOD FOR FREQUENCY STABILIZATION OF GRID-FORMING VSGs 447

Fig. 10. Bode diagrams of virtual active power in two parts with
different Lv .
Fig. 11. Active power closed-loop transfer function block diagram with
virtual inductance.

the system. In addition, it can be inferred that greater virtual

inductance makes system more robust, and the negative feedback
effect brought by virtual inductance gets stronger.

C. Physical Significance of Proposed Virtual Inductance

Control for Mitigation of System Oscillation
According to the abovementioned discussions, negative
damping as well as positive feedback effect of the system is the
predominant cause of divergence of system frequency. Hence,
this part will introduce physical significance of the proposed
virtual inductance for mitigation of system oscillation.
From the view of attributes of the powers, the feedback active
power with consideration of virtual inductance can be divided
into two parts, hvir_δ (s) and hvir_E (s), i.e.,
⎧ ΔPvir_δ
⎪
⎪ hvir_δ (s) = Δδe = KP δ_Lv
⎨ Fig. 12. Frequency responses of feedback power transfer functions
ΔPvir_E
hvir_E (s) = . and the corresponding vector diagram with Lv = 8.8 mH. (a) Frequency
⎪
⎪ Δδe
⎩ response. (b) Vector illustration.
= [GP δ (NEω_vir −NEω )+KP δ_Lv NEω_vir ]/s
(28)
Fig. 10 plots the frequency response results of hvir_δ (s) and active powers are formulated as
hvir_E (s) with various sizes of virtual inductance Lv (Lv1 = ⎧ 

4.4 mH, Lv2 = 8.8 mH, Lv3 = 13.2 mH, Lv4 = 17.6 mH). ⎪
⎪ hδ  (s) = ΔP
Δδe = GP δ (s) + KP δ_Lv (s)
δ
⎪
⎨
It can be seen that the magnitude of hvir_δ (s) is higher than  −(GQδ (s)+KQδ_Lv (s))(GP E (s)+KP E_Lv (s))
hE  (s) = ΔP E
= .
that of hvir_E (s) in the selected frequency range, no matter ⎪
⎪
Δδ e Jq s+(GQE (s)+KQE_Lv (s))
⎪
⎩ h  (s) = ΔPD = D s 
what size of Lv , which illustrates that hvir_δ (s) plays a dom- D Δδe p
inant role in stabilization of the system. The synthetic active (31)
power-angle transfer function with virtual inductance can be Consequently, (30) can be rearranged as
calculated by
ΔP  = hδ  (s)Δδe + hE  (s)Δδe + hD  (s)Δδe . (32)
hvir (s) = hvir_δ (s) + hvir_E (s). (29)      
  
ΔPδ ΔPE ΔPD

The small signal control system with virtual inductance con- Fig. 12 shows the frequency responses of hδ ’(s), hD ’(s),
trol is presented, as shown in Fig. 11. Therefore, the synthetic hE ’(s), hΣ ’(s) and vector diagrams, which reflects the impact
powers can be derived as of virtual inductance on dynamic behavior of the grid-forming
VSG system. Vectors are characterized by information of phase
ΔP  = hδ (s)Δδe + hE (s)Δδe + hD (s)Δδe + hvir (s)Δδe . and magnitude at CFP. The synthetic power is located in the first
       
ΔPδ ΔPE ΔPD ΔPvir quadrant, which illustrates the proposed virtual inductance con-
(30) trol provides damping support for the system. Besides, negative
In order to illustrate how the virtual inductance impacts the damping power caused by strong power coupling effect can be
dynamics of active power, the newly formed three feedbacks offset by the effect of virtual inductance.

Authorized licensed use limited to: China Three Gorges University. Downloaded on August 31,2023 at 08:40:58 UTC from IEEE Xplore. Restrictions apply.
448 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 70, NO. 1, JANUARY 2023

Fig. 13. Frequency responses and vector plot of the four active
power feedback parts with different virtual inductance. (a) Frequency
responses of hvir (s). (b) Vector diagram of active power components of
ΔPδ , ΔPE , ΔPD , ΔPvir , and ΔPΣ .

Fig. 15. Current response with and without proposed control. (a) With-
out proposed control (Rg /Xg = 0.43). (b) With proposed control (Rg /Xg
= 0.43). (c) Without proposed control (Rg /Xg = 0.29). (d) With proposed
control (Rg /Xg = 0.29).

Fig. 14. Simulation results of the frequency with different virtual

inductance.

Fig. 13 shows the results of bodes as well as vector diagrams

with varying size of virtual inductance (Lv1 = 4.4 mH, Lv2 =
8.8 mH, Lv3 = 13.2 mH, Lv4 = 17.6 mH). It shows that the
greater virtual inductance results in stronger damping effect and
negative feedback effect that can stabilize system. The effec-
tiveness of the proposed virtual inductance control is reflected.
Thereby the frequency stability is enhanced by the virtual in-
ductance control. Fig. 16. StarSim experimental setups.

IV. SIMULATION AND EXPERIMENT RESULTS leads to poorer stability of the system with weaker damping
To verify the effectiveness of the proposed virtual inductance performance. The oscillation in Fig. 15(a) is divergent, but os-
control method, Scenarios in different value of virtual induc- cillation in Fig. 15(c) is convergent due to the different Rg /Xg , but
tance are built. Simulation results of frequency characteristic the dynamic characteristics are all improved with the proposed
are displayed in Fig. 14. The attenuation speeds of the frequency virtual inductance control in Fig. 15(b) and (d). Generally, in all
with virtual inductance being 8.8 mH are much faster than that of the case studies, the proposed virtual inductance can take good
of virtual inductance being 4.4 mH. It inferred that the larger effect in suppressing oscillation with larger inertia and stronger
size of virtual inductance can make the oscillation attenuated damping.
much faster and provide stronger damping for the system. In this section, the experiments are developed to demonstrate
Specifically, the stability as well as dynamic performance of the proposed analysis framework as well as the proposed virtual
the grid-forming VSG in this article is not only related with the inductance control strategy. In Fig. 16, StarSim experimental
short circuit ratio, but also concerned with the ratio of Rg /Xg . setups include: oscilloscope, host PC, Rapid control prototype
As can be seen in Fig. 15(a) and (c), higher level of Rg /Xg (RCP), IO board, and Hardware-in-the-loop (HIL) model. The

Authorized licensed use limited to: China Three Gorges University. Downloaded on August 31,2023 at 08:40:58 UTC from IEEE Xplore. Restrictions apply.
YANG et al.: NEW VIRTUAL INDUCTANCE CONTROL METHOD FOR FREQUENCY STABILIZATION OF GRID-FORMING VSGs 449

TABLE II
PARAMETERS OF GRID IMPEDANCE AND VIRTUAL INDUCTANCE

Fig. 18. Experiment results of P, Q, f, and iabc of case III. (a) Power
and frequency performance. (b) Grid-connected current.

Rg changed from 0.3 to 1.1 Ω, and a magnitude-equal oscillation

occurs where the oscillation frequency is 101.5 rad/s, which
validates the theoretical analysis. Compared to case I, a greater
size of virtual inductance is introduced to test the effectiveness
of control strategy in case II. In Fig. 17(c) and (d), it can be
seen that oscillation is attenuated much faster and convergence
to equilibrium point with much faster response. It can be inferred
that larger virtual inductance can provide more damping for the
system to mitigate oscillation instability.
As for case III, Fig. 18 shows that the greater Rg results
in larger magnitude of oscillation. With the virtual inductance
control strategy put into operation, oscillation of the system
gets convergent and finally be stable at an equilibrium point.
It can be inferred that the proposed virtual inductance can
provide damping support, mitigate the coupling effect between
active- and reactive-power, and improve stability margins of the
grid-forming VSG system. However, the transient response is
lower by comparing to case II, it indicates that the higher power
coupling would decrease the speed of stable recovery.

Fig. 17. Experiment results of P, Q, f and iabc with different Lv . V. CONCLUSION

(a) Power and frequency waveforms with Lv = 4.4 mH. (b) Grid-
connected current waveform with Lv = 4.4 mH. (c) Power and frequency This article established small-signal closed-loop active power
waveforms with Lv = 8.8 mH. (d) Grid-connected current waveform with
Lv = 8.8 mH.
model, which is used for identification of frequency stability of
grid-forming VSG system. It is shown that the VSG system
has a risk of oscillation instability due to negative damping,
RCP upload the controller built in the MATLAB Simulink and it can be eliminated by the proposed virtual inductance
model. And the HIL is connected to the hardware of the Simulink control method. Besides, a new coupling model was developed
model. The HIL-RCP based StarSim can validate the proposed to clarify the dynamic interaction between voltage magnitude
control algorithm rapidly in real time. The parameters of grid and frequency of the system. By judging the size of FEF, motion
impedance and virtual inductance in those cases are given in trajectory of RoCoF and FO can be known. Specifically, when
Table II. the size of FEF is below one, it forms negative feedback that can
Fig. 17(a) and (b) shows the experiment results of the fre- stabilize system, and both RoCoF and FO finally tend to be zero,
quency, grid-connected current, active-, and reactive-power with vice versa. Moreover, a new virtual inductance control strategy
the proposed virtual inductance method of case I. It shows that was proposed to dampen oscillation of grid-forming VSG by
the system becomes unstable with disturbance of grid resistance increasing positive damping power, and the effectiveness is

Authorized licensed use limited to: China Three Gorges University. Downloaded on August 31,2023 at 08:40:58 UTC from IEEE Xplore. Restrictions apply.
450 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 70, NO. 1, JANUARY 2023

validated by the FEF analysis. In this framework, it does not [7] D. Li, Q. Zhu, S. Lin, and X. Y. Bian, “A self-adaptive inertia and damping
rely on dual-loop control structure, which is different from combination control of VSG to support frequency stability,” IEEE Trans.
Energy Convers., vol. 32, no. 1, pp. 397–398, Mar. 2017.
the conventional virtual inductance control. What is more, it [8] J. Fang, P. Lin, H. Li, Y. Yang, and Y. Tang, “An improved virtual
is implemented based on energy conservation, and it can be inertia control for three-phase voltage source converters connected to a
adaptive to the power converter with another types of filters, like weak grid,” IEEE Trans. Power Electron., vol. 34, no. 9, pp. 8660–8670,
Sep. 2019.
LC filters, and LCL filters. Besides, the LC filter is as similar as [9] H. Wu et al., “Small-Signal modeling and parameters design for virtual
that of LCL filter since that the grid-side L filter of LCL filter synchronous generators,” IEEE Trans. Ind. Electron., vol. 63, no. 7,
can be integrated into the grid impedance as a whole. All of the pp. 4292–4303, Jul. 2016.
[10] C. Li, Y. Yang, Y. Cao, L. Wang, and F. Blaabjerg, “Frequency and voltage
analytical results of established models agree with the results stability analysis of Grid-forming virtual synchronous generator attached
of eigenvalue analysis. Finally, both proposed analytical model to weak grid,” IEEE J. Emerg. Sel. Topics Power Electron., to be published,
and virtual inductance control are verified experimentally. doi: 10.1109/JESTPE.2020.3041698.
[11] C. Li, Y. Li, Y. Cao, H. Zhu, C. Rehtanz, and U. Häger, “Virtual syn-
chronous generator control for damping DC-Side resonance of VSC-
APPENDIX MTDC system,” IEEE J. Emerg. Sel. Topics Power Electron., vol. 6, no. 3,
pp. 1054–1064, Sep. 2018.
Combining (4a) and (5a) yields [12] M. Saeedian, R. Sangrody, M. Shahparasti, S. Taheri, and E. Poures-
maeil, “Grid-Following DVI-Based converter operating in weak grids for
Pref − f (δe , E) = Dp (ωe − ωn ) + Jp ωn (ωE − ωn )s. (A1) enhancing frequency stability,” IEEE Trans. Power Del., vol. 37, no. 1,
pp. 338–348, Feb. 2022, doi: 10.1109/TPWRD.2021.3059898.
Linearizing (4c), (A1) and combining (6), then (A2) can be [13] W. Wu et al., “Sequence-impedance-based stability comparison between
obtained VSGs and traditional grid-connected inverters,” IEEE Trans. Power Elec-
tron., vol. 34, no. 1, pp. 46–52, Jan. 2019.
ΔPref − (GP δ (s) 1s Δωe + GP E (s)ΔE) [14] J. Guo et al., “Analysis and mitigation of low-frequency interactions
(A2) between the source and load virtual synchronous machine in an Islanded
= Dp (Δωe − Δωn ) + Jp ωn s(Δωe − Δωn ). microgrid,” IEEE Trans. Ind. Electron., vol. 69, no. 4, pp. 3732–3742,
Apr. 2022, doi: 10.1109/TIE.2021.3075847.
According to (A2), the relationship from ΔE to Δω e can be [15] X. Liang, C. Andalib-Bin-Karim, W. Li, M. Mitolo, and M. N. S. K.
derived as Shabbir, “Adaptive virtual impedance-based reactive power sharing in
virtual synchronous generator controlled microgrids,” IEEE Trans. Ind.
Δωe −GP E (s)/(Jp ωn s + Dp )
MωE (s) = = . (A3) Appl., vol. 57, no. 1, pp. 46–60, Jan./Feb. 2021.
ΔE [1 + GP δ (s)/(Jp ωn s2 + Dp s)] [16] Y. Hu, Y. Shao, R. Yang, X. Long, and G. Chen, “A configurable virtual
impedance method for grid-connected virtual synchronous generator to
Similarly, combining (4b) and (5b) yields improve the quality of output current,” IEEE J. Emerg. Sel. Topics Power
Electron., vol. 8, no. 3, pp. 2404–2419, Sep. 2020.
Qref − g(δe , E) = Jq (E − Uref )s. (A4) [17] G. Lou, Q. Yang, W. Gu, X. Quan, J. M. Guerrero, and S. Li, “Analysis
and design of hybrid harmonic suppression scheme for VSG considering
Linearizing (4c), (A4) and combining (6), then (A5) can be nonlinear loads and distorted grid,” IEEE Trans. Energy Convers., vol. 36,
obtained no. 4, pp. 3096–3107, Dec. 2021, doi: 10.1109/TEC.2021.3063607.
[18] P. Kundur, Power System Stability and Control. New York, NY, USA:
1 McGraw-Hill, 1994.
ΔQref −(GQδ (s) Δωe +GQE (s)ΔE) = Jq (ΔE −ΔUref )s. [19] K. Wang and X. Yuan, “Stability analysis of the virtual induc-
s tance for LCL filtered Droop-controlled Grid-connected inverters,”
(A5)
IEEE J. Emerg. Sel. Topics Power Electron., to be published,
According to (A5), the relationship from Δω e to ΔE can be doi: 10.1109/JESTPE.2021.3058810.
derived as [20] Y. Hu, Y. Shao, R. Yang, X. Long, and G. Chen, “A configurable virtual
impedance method for grid-connected virtual synchronous generator to
ΔE −GQδ (s)/Jq s2 improve the quality of output current,” IEEE J. Emerg. Sel. Topics Power
NEω (s) = = . (A6)
Δωe 1 + GQE (s)/Jq s Electron., vol. 8, no. 3, pp. 2404–2419, Sep. 2020.

Yaqian Yang received the master’s degree in

REFERENCES electrical engineering from the Changsha Uni-
[1] F. Blaabjerg, M. Liserre, and K. Ma, “Power electronics converters for versity of Science and Technology, Changsha,
wind turbine systems,” IEEE Trans. Ind. Appl., vol. 48, no. 2, pp. 708–719, China, in 2018. She is currently working toward
Mar./Apr. 2012. the Ph.D. degree in electrical engineering with
[2] C. Li, Y. Yang, N. Mijatovic, and T. Dragicevic, “Frequency stability as- Hunan University, Changsha.
sessment of Grid-forming VSG in framework of MPME with feedforward Her current research interests include stabi-
decoupling control strategy,” IEEE Trans. Ind. Electron., vol. 69, no. 7, lization and control of grid-forming virtual syn-
pp. 6903–6913, Jul. 2022, doi: 10.1109/TIE.2021.3099236. chronous generators, synthetic inertial applied
[3] Y. Yang, C. Li, J. Xu, F. Blaabjerg, and T. Dragičević, “Virtual inertia in hybrid ac–dc microgrids.
control strategy for improving damping performance of DC microgrid with
negative feedback effect,” IEEE J. Emerg. Sel. Topics Power Electron., Jiazhu Xu (Member, IEEE) was born in
vol. 9, no. 2, pp. 1241–1257, Apr. 2021. Chuzhou, China, in 1980. He received the B.S.,
[4] J. Fang, H. Li, Y. Tang, and F. Blaabjerg, “Distributed power system virtual M.S., and Ph.D. degrees in electrical engineer-
inertia implemented by grid-connected power converters,” IEEE Trans. ing from Hunan University, Changsha, China, in
Power Electron., vol. 33, no. 10, pp. 8488–8499, Oct. 2018. 2002, 2004, and 2007, respectively.
[5] M. Saeedian, B. Eskandari, S. Taheri, M. Hinkkanen, and E. Pouresmaeil, Since 2008, he has been an Associate Pro-
“A control technique based on distributed virtual inertia for high pene- fessor with the College of Electrical and Infor-
tration of renewable energies under weak grid conditions,” IEEE Syst. J., mation Engineering, Hunan University, where
vol. 15, no. 2, pp. 1825–1834, Jun. 2021. he became a Professor in 2017. His current
[6] L. Huang et al., “An adaptive inertia control to improve stability of virtual research interests include ac–dc power conver-
synchronous machines under various power grid strength,” in Proc. IEEE sion theory, locomotive traction power supply
Power Energy Soc. Gen. Meeting, Atlanta, GA, USA, 2019, pp. 1–5. systems, and optimization of modern electrical equipment.

Authorized licensed use limited to: China Three Gorges University. Downloaded on August 31,2023 at 08:40:58 UTC from IEEE Xplore. Restrictions apply.
YANG et al.: NEW VIRTUAL INDUCTANCE CONTROL METHOD FOR FREQUENCY STABILIZATION OF GRID-FORMING VSGs 451

Chang Li received received the bachelor’s de- Ming Wen was born in Hubei, China, 1981. He
gree from Guizhou University, Guiyang, China, received the M.S. and Ph.D. degrees in electri-
in 2014 and the Ph.D. degree from Hunan Uni- cal engineering from Hunan University, Chang-
versity, Changsha, China, in 2020, both in elec- sha, China, in 2006 and 2010, respectively.
trical engineering. He has been a Senior Engineer with State
Since October 2020, he has been a Postdoc- Grid Hunan Electric Power Company Limited
toral Researcher with the Technical University Economical and Technical Research Institute.
of Denmark, Lyngby, Denmark. His current re- His research interests include energy internet
search interests include power electronics inter- demand forecasting, electricity price, smart grid,
faced power systems. renewable energy accommodation, and power
system planning.

Weiming Zhang received the bachelor’s degree

in electrical engineering and automation from
Jiangsu University, Zhenjiang, China, in 2019, Frede Blaabjerg (Fellow, IEEE) received the
and the master’s degree in electrical engineer- Ph.D. degree in electrical engineering from
ing from Hunan University, Changsha, China, in Aalborg University, Aalborg, Denmark, in 1995.
2021. He was with ABB-Scandia, Randers,
Since 2021, he has been with Shenzhen Denmark, from 1987 to 1988. He became
Power Supply Company, Ltd., Shenzhen, China. an Assistant Professor in 1992, an Associate
His current research interests include design Professor in 1996, and a Full Professor of Power
and control of dc distribution system. Electronics and Drives in 1998. In 2017, he
was a Villum Investigator. He is honoris causa
with University Politehnica Timisoara, Romania
and Tallinn Technical University, Estonia. He
Qiuwei Wu (Senior Member, IEEE) received the has authored and coauthored more than 600 journal papers in the
Ph.D. degree in power system engineering from fields of power electronics and its applications. He has coauthored four
Nanyang Technological University, Singapore, monographs and is the editor of ten books in power electronics and its
in 2009. applications. His current research interests include power electronics
He was a Senior R&D Engineer with Ves- and its applications such as in wind turbines, PV systems, reliability,
tas Technology R&D Singapore Pvt., Ltd., from harmonics, and adjustable speed drives.
March 2008 to October 2009. He was work- Prof. Blaabjerg was the recipient of 33 IEEE prize paper awards,
ing with the Department of Electrical Engi- IEEE PELS Distinguished Service Award in 2009, the EPE-PEMC
neering, Technical University of Denmark from Council Award in 2010, IEEE William E. Newell Power Electronics
November 2009 to August 2021 (Postdoc Award 2014, the Villum Kann Rasmussen Research Award 2014, the
November 2009–October 2010, Assistant Pro- Global Energy Prize in 2019, and the 2020 IEEE Edison Medal. He
fessor November 2010–August 2013, Associate Professor September was the Editor-in-Chief of IEEE TRANSACTIONS ON POWER ELECTRONICS
2013–August 2021). He is a tenured Associate Professor with Tsinghua- from 2006 to 2012. He has been Distinguished Lecturer for IEEE
Berkeley Shenzhen Institute, Tsinghua Shenzhen International Grad- Power Electronics Society from 2005 to 2007 and for IEEE Industry
uate School, Tsinghua University. He was a Visiting Scholar with the Applications Society from 2010 to 2011 as well as 2017 to 2018. In 2019
Department of Industrial Engineering and Operations Research, Uni- and 2020, he was a President of IEEE Power Electronics Society. He
versity of California, Berkeley, from February 2012 to May 2012 funded has been Vice-President of the Danish Academy of Technical Sciences.
by the Danish Agency for Science, Technology and Innovation, Den- He was nominated in 2014–2020 by Thomson Reuters to be between
mark. He was a visiting professor named by Y. Xue, an Academician the most 250 cited researchers in Engineering in the world.
of the Chinese Academy of Engineering, Shandong University, China,
from November 2015 to October 2017. He was a visiting scholar with
the School of Engineering and Applied Sciences, Harvard University
from November 2017 to October 2018 funded by the Otto Mønsted
Fond. His research interests include decentralized/distributed optimal
operation and control of power systems with high penetration of renew-
ables, including distributed wind power modeling and control, decen-
tralized/distributed congestion management, voltage control and load
restoration of active distribution networks, and decentralized/distributed
optimal operation of integrated energy systems.
Dr. Wu is an Editor of IEEE TRANSACTIONS ON POWER SYSTEMS and
IEEE Power Engineering Letters. He is also the deputy Editor-in-Chief
and an Associate Editor of International Journal of Electrical Power and
Energy Systems, and Journal of Modern Power Systems and Clean
Energy. He is a subject editor for IET Generation, Transmission and
Distribution and IET Renewable Power Generation.

Authorized licensed use limited to: China Three Gorges University. Downloaded on August 31,2023 at 08:40:58 UTC from IEEE Xplore. Restrictions apply.