IEEE POWER ELECTRONICS REGULAR PAPER

Transient Stability Improvement of Grid-Forming

Converters through Voltage Amplitude Regulation
and Reactive Power Injection
Wenjia Si, Student Member, IEEE and Jingyang Fang, Senior Member, IEEE

Abstract−Grid-forming converters, which provide frequency controllable voltage sources emerge as a promising solution
and voltage support to the grid, have been identified as an ena- [4], [5]. Grid-forming converters have the capability to sup-
bling technology in modern power systems. However, the unsta- port grid operation similarly to SGs, allowing a seamless tran-
ble operation of grid-forming converters under large disturb- sition between grid-tied and islanded operation [6].
ances poses a significant challenge. To address the challenge, this
article presents an in-depth analysis of the transient stability of
Although grid-forming converters are the enabling units in
grid-forming converters and proposes two effective methods to modern power systems, the relevant technology is still in its
improve their transient stability. First, it is revealed that the ac- infancy. There are many stability problems faced by grid-
tive power injection and absorption of grid-forming converters forming converters. The stability of grid-forming converters
are asymmetrical. Subsequently, the methods of voltage ampli- encompasses two main aspects—small-signal and large-signal
tude regulation and reactive power injection are proposed to stability [7], [8]. These two aspects represent the ability of
improve the transient stability of grid-forming converters. By grid-forming converters to maintain synchronization with the
means of phase portraits, the effect of transient stability im- grid when subjected to a small or a large disturbance, respec-
provement by the proposed methods are clearly described. Fur- tively. By use of the well-established tools based on linear
thermore, a detailed quantitative analysis is performed, and the
critical values of voltage amplitude and reactive power in transi-
system theory, the small-signal stability of grid-forming con-
ent periods are derived to guide controller design. Finally, simu- verters has extensively been studied. For example, reference
lation and experimental results are provided to validate the effec- [9] derives a detailed state-space model, comprising the mod-
tiveness of the proposed methods. els of system plants, measurements, current and voltage con-
trollers, etc. References [10]−[12] propose small-signal stabil-
Index terms—Grid-forming converters, transient stability, ity improvement methods through virtual impedance control,
voltage amplitude regulation, reactive power injection. phase compensation, and cross-coupling control schemes, re-
spectively. However, the large-signal stability touching more
I. INTRODUCTION complex and non-linear analyses requires further exploration.
In particular, the transient stability refers to a type of large-
To reduce carbon footprints and meet the growing energy
signal stability that associated with active power transfer and
demand, the trend of an increased renewable energy penetra-
its limitation. The transient stability of grid-forming convert-
tion in power systems is expected to continue. Generally, re-
ers during faults remains as a hot research topic [13], [14].
newable energies are coupled to the grid through grid-tied
Transient stability problems can arise from grid voltage
power electronic converters that operate as current sources,
sags, transmission lines faults, and load swings, etc. Several
which are named as grid-following converters [1], [2]. How-
metrices play a vital role on the transient stability, which in-
ever, the increasing integration of grid-following converters
clude inertia, damping factor, active power reference, trans-
presents challenges to the stability and control of power sys-
mission line impedance, voltage amplitude, and reactive pow-
tems. This is because mainstream grid-following converters
er reference. Inertia and damping factor affect the dynamic
are unable to offer essential services such as grid formulation,
response process (e.g., the response speed) during transients
inertia, and damping that are provided by synchronous genera-
[15]. The active power reference determines the existence of
tors (SGs) [3]. Thus, grid-forming converters operating as
operating points after a fault. If the active power reference
Manuscript received March 3, 2023; revised May 25, 2023; accepted exceeds the limitation, the system will lose operating points
June 24, 2023. This work was supported in part by Jinan Municipal Gov- and become unstable [16]. Transmission lines impedance,
ernment under Grant 20 New University Items, Grant Number 202228069, voltage amplitude, and reactive power reference all affect the
in part by the National Natural Science Foundation of China under Grant
transient stability of grid-forming converters by changing the
Excellent Young Scientists Funding Program (Overseas), Grant Number
20221017-9, in part by the Guangdong Basic and Applied Basic Research active power limitation.
Foundation under Grant Guangdong-Shenzhen Joint Youth Fund, Grant With the above-mentioned metrices in mind, several meth-
Number 2022A1515110422, in part by the Ministry of Science and Tech- ods have been reported to enhance the transient stability of
nology, PRC under Grant High-End Foreign Expert Recruitment Program, grid-forming converters. For example, reference [17] indicates
Grant Number G2022150006L, and in part by the Shandong University
under Grant Distinguished Young Scholars Project, Grant Number that increasing the damping factor can improve the transient
31400082260522. stability. However, it only reveals the negative effect of iner-
Wenjia Si and Jingyang Fang are with the School of Control Science tia—increasing the overshoot of the power angle. Instead, ref-
and Engineering, Shandong University, Jinan 250100, China (e-mails:
erence [18] points out the positive effect of inertia—increasing
[email protected]; [email protected]).
IEEE POWER ELECTRONICS REGULAR PAPER

the critical clearing time, but the proposed parameter-tunning

method based on an on-off switch is inaccurate. In [19] and vgdc Grid-Forming Converter vgabc Grid
[20], the transient stability of grid-forming converters can be Lgi igiabc Lgg iggabc Ls vsabc
Rs
improved by reducing the active power reference during the +
grid voltage sag. Nevertheless, this causes a reduction in ac-
tive power transfer, which further deteriorates frequency regu- vgfabc +
lation and even leads to blackouts. Reference [21] proposes a Sgr
dq-frame asymmetrical virtual transient impedance control PWM Cgf - PCC
scheme, which may cause instability due to impedance cou- Measure and Control
PWM Modulator
plings.
Transient stability improvement methods associated with igiabc vgfabc
qg_ref qg
voltage amplitude and reactive power are emerging, but there dq0 θg_ref abc θg_ref abc θg_ref Reactive
exist research gaps. For instance, reference [22] puts forward Power
abc dq0 dq0
two voltage boosters for improving the transient stability of Control
systems fed by 100% grid-forming converters. However, it igi_dq vgf_dq
vgf_ref_d
lacks a quantitative analysis and experimental verifications. In Current i gi_ref_dq Voltage
addition, an on-off behavior instead of smooth voltage regula- Control Control
tion suffers from inaccuracy. Reference [23] reveals that reac- 0
tive power control can deteriorate the system transient stability Stage 4 θg Active pg
Sgr Power
by creating a transient voltage drop. However, this article fo- Stage 3 δg Power pg_ref qg Calculation
cuses on analyses instead of solutions. Reference [24] propos- Sync. Control
es a method for improving the transient stability via a frequen- θg_ref Stage 1 3 θs vgfabc iggabc
cy-reactive power feedforward control. However, this method PLL
vgabc
increases control complexity, as it requires a phase-locked
loop (PLL) to detect the frequency in real time. Moreover, the Fig. 1. The system configuration of a grid-forming converter.
frequency feedforward leads to a more pronounced coupling the grid through an LCL filter formed by Lgi, Cgf, and Lgg.
between active power and reactive power control, which can Generally, we model the power grid as a serial connection of
cause other instability. resistors Rs, inductors Ls, and voltage sources vsabc [25]. vgdc
This article takes into account both the injection and absorp- denotes the dc bus voltage of the grid-forming converter. vgfabc
tion of active power and proposes two methods—voltage and igiabc represent the capacitor voltages and the converter-
regulation and reactive power injection, to improve the transi- side currents, respectively. vgabc and iggabc stand for the voltages
ent stability of grid-forming converters. In addition, a detailed at the point of common coupling (PCC) and the grid-injected
quantitative analysis is performed. The benefits of the pro- currents, respectively.
posed methods are listed as follows. In addition, Fig. 1 illustrates the measure and control block
(1) Significantly improve the transient stability of grid- diagrams of the grid-forming converter system. The subscript
forming converters. dq denotes the signal in the synchronous frame after the Park’s
(2) Easy to implement and adjust in real time. transformation, and ref represents the reference signal. The
(3) Keep active power transfer unchanged. inner current controller aims to damp the LCL resonance and
(4) No additional hardware is required. limit converter currents for protection purposes [26]. The mid-
(5) Provide smooth regulation of the voltage amplitude or dle voltage controller aims to tightly regulate the amplitude
reactive power injection. and the phase angle of the capacitor voltages vgfabc. Generally,
The rest of this article is organized as follows. Section II the outer power controllers target at the active power and the
describes the structure of grid-forming converters and the reactive power, tracking their references. The active power pg
mechanism of transient instability. In Section III, the proposed and the reactive power qg are calculated from vgabc and iggabc. It
transient stability improvement method through voltage regu- should be noted that the reactive power controller can be re-
lation is disclosed and detailed. Section IV introduces the tran- placed by a voltage controller, which provides accurate and
sient stability improvement method, which exploits reactive fast voltage control [6]. In other words, the grid-forming con-
power injection. Section V verifies the proposed two methods verter can operate in either the voltage control or the reactive
through simulation and experimental results. Finally, Section power control mode.
VI concludes this article. In addition, the phase angle of the grid-forming converter is
determined by the active power controller. In Fig. 1, θg_ref is
II. TRANSIENT STABILITY OF GRID-FORMING CONVERTERS the reference phase angle used for frame transformations. θg
denotes the real phase angle of the grid-forming converter. θs
A. The Structure and Simplified Model of Grid-Forming Con- is the phase angle provided by a PLL, which tracks the phase
verters angle of the grid voltage. The phase angle difference between
the grid-forming converter and the grid is represented as δg,
Fig. 1 shows a system configuration of a grid-forming con- which is called the power angle. As shown, the process of grid
verter. As shown, the grid-forming converter is connected to synchronization includes four stages.
IEEE POWER ELECTRONICS REGULAR PAPER

Stage 1: Prior to grid connection, the relay Sgr is open. θg_ref pg_ref pg_ref_pu Δωg_pu(Δfg_pu) ωg_pu θg
= θs. In this case, the grid phase angle is locked by the PLL. 1/S0 Kpf ω0/s
+ + +
Stage 2: The converter regulates the capacitor voltages vgfabc, pg_pu 1
which follow the grid voltages.
Stage 3: When vgfabc are identical to vgabc, the converter is 1/S0 δg
ω0/s
connected to the grid by closing Sgr. During the initial period pg
of grid connection, θg_ref = δg + θs. During this stage, the phase
(a) The p−f droop control
angle is formed by both the power angle δg and the PLL phase
angle θs.
Stage 4: After the converter successfully synchronizes with qg_ref qg_ref_pu Δvg_pu Δvg vgd_ref
the grid, the PLL is no longer necessary as long as the active 1/S0 Kqv V0
+ + +
power is well regulated [27]. Therefore, θg_ref = θg, the phase qg_pu
Vg_ref
angle is provided independently by the power controller.
Before proceeding to details, we first simplify the model of
1/S0
grid-forming converters with some assumptions, which are qg
based on stable grid connection and given as follows.
Assumption 1: Due to the quick and precise voltage control, (b) The q−v droop control
the grid-forming converter is simplified into a controllable ac Fig. 3. Block diagrams of the droop control.
voltage source, with the ignorance of inner-loop dynamics. subscript pu represents the per unit value. S0, ω0, and V0 are the
Assumption 2: The dynamics caused by line impedances are nominal power, angular frequency, and voltage of the grid-
ignored. forming converter, respectively. Kpf and Kqv represent the
Assumption 3: Three-phase grid-forming converters are
droop gains of the active power control and the reactive power
balanced and symmetrical.
Assumption 4: Over-current protection is not triggered. control, respectively. fg and ωg are the frequency and the angu-
With these assumptions, a simplified model of two voltage lar frequency, respectively. Vg_ref denotes the voltage ampli-
sources is constructed, as illustrated in Fig. 2. The grid- tude reference of the grid-forming converter. vgd represents the
forming converter voltage is represented as a phasor Vg0, d-axis voltage in the synchronous frame.
where Vg is the phase voltage amplitude and is chosen as the As shown in Fig. 3(a) and Fig. 3(b), the p−f and q−v control
reference of zero radian. The grid voltage and current are de- equations can be written as
noted by Vs∠−δg and Igg∠φ, respectively. Rgs and Xgs are the d g / dt = 0 K pf ( pg_ref / S0 − pg / S0 ) (3)
resistance and reactance between the grid-forming converter vgd_ref = V0 K qv (qg_ref / S0 − qg / S0 ) + vg_ref (4)
and the grid, respectively. Based on this simplified model, the
three-phase active power pg and reactive power qg transmitted Table I. Parameters used in simulation and experimental tests.
from Vg to Vs can be derived as Descriptions Symbols Values
3[Vg (Vg − Vs cos  g ) Rgs + VgVs sin  g X gs ]
pg = (1) Rated power S0 5000 VA
2( X gs 2 + Rgs 2 ) Rated voltage V0 110 Vrms
Grid voltage amplitude Vs 110 Vrms
3[Vg (Vg − Vs cos  g ) X gs − VgVs sin  g Rgs ]
qg = (2) Converter voltage reference Vg_ref 110 Vrms
2( X gs 2 + Rgs 2 )
DC-link voltage vgdc 350 V
B. Power Control Grid angular frequency ω0 314 rad/s

As the afore-mentioned simplified model ignores the inner- Line inductance Lgs 13 mH
loop current and voltage control, we focus on the outer-loop Line resistor Rgs 0.6 Ω
power control henceforth. The droop control, emerged before p-f droop gain Kpf 0.15
1990, is a well-established scheme known for its simplicity q-v droop gain Kqv 1
and decentralized power sharing ability and used here [28].
The droop control mainly refers to the droop relationship
between the active power and frequency (p−f) as well as the pg
pg_max pg
reactive power and voltage (q−v). Fig. 3 details the droop con-
trol. The symbol Δ stands for the parameter change, and the pg_ref
a b
Igg φ jXgs Rgs δ'g_pmax
δg_pmax δg δg
b' a' p'g_ref Rgs=0
Vg 0 Vs δg p'g_max Rgs>0
(a) pg−δg curve with equilibrium points (b) pg−δg curves with different Rgs

Fig. 2. Simplified model of the grid-forming converter and the grid. Fig. 4. pg−δg curves with equilibrium points and different Rgs.
IEEE POWER ELECTRONICS REGULAR PAPER

In addition, by removing the blocks inside the dashed box indicates that the grid-forming converter is able to maintain
in Fig. 3(b), the reactive power control mode is switched into stable.
the voltage control mode. Subsequently, we can regulate the
output voltage of the grid-forming converter by modifying III. TRANSIENT STABILITY IMPROVEMENT THROUGH VOLT-
vg_ref. AGE AMPLITUDE REGULATION

C. pg−δg Curves
A. Fundamental Principle
Based on (1), pg−δg curves are plotted in Fig. 4 with the pa-
rameters in Table I. It is worth noting that the grid-forming Section II introduces two types of transient instability (with
converter integrated with energy storage can absorb active or without equilibrium points), which depend on the extent of
power. Therefore, pg−δg curves involve two cases: faults. A serious fault significantly reduces the active power
Case I: pg > 0, which represents that the grid-forming con- transfer limitation, leading to the loss of equilibrium points.
verter outputs active power to the grid. By increasing the active power transfer limitation, we can re-
Case II: pg < 0, which represents that the grid-forming con- establish equilibrium points and ensure stable operation. No-
verter absorbs active power from the grid. tably, from (1), the active power is directly affected by the
In Fig. 4(a), pg_ref and p'g_ref stand for the active power refer- grid-forming converter voltage Vg.
ences in Case I and Case II, respectively. pg_max and p'g_max As depicted in Fig. 6, when the grid-forming converter op-
denote the active power transfer limitations in Case I and Case erates in the voltage control mode and the grid voltage Vs
II, respectively. δg_pmax and δ'g_pmax represent the boundary drops from 110 Vrms to 15 Vrms, Vg drops in response to Vs.
power angles. a, a', b, and b' are the equilibrium points. The Accordingly, the pg−δg curve after a grid voltage sag is shown
shadow area represents the stable operating region. as Curve I. By increasing Vg, Curves II, III, and IV are ob-
Fig. 4(b) shows the two curves with Rgs = 0 and Rgs > 0. The tained. Clearly, the active power transfer limitation is in-
active power is consumed by the transmission line resistance creased with the increase of Vg.
Rgs, leading to an asymmetrical pg−δg curve. In Case I, the The phase portraits shown in Fig. 7 are graphical solutions
grid-forming converter should output extra active power to Rgs, of (3), providing simpler and more intuitive results. At each
which increases the limitation of active power transfer. In con- point, the changing trend of δg is determined by its derivative
trast, in Case II, Rgs absorbs a portion of the active power, re- dδg/dt. As shown in Fig. 7, with the increase of Vg, the curve
sulting in a reduction of the active power transfer limitation. gets closer to the dotted line, where dδg/dt = 0. Therefore,
However, according to the authors’ knowledge, existing re- voltage regulation is an effective method to improve transient
search only focuses on transient stability problems in Case I, stability of the grid-forming converter. In Figs. 6 and 7, the
which is inadequate. This paper fills this gap by investigating main parameters used are based on Table I. In addition, pg_ref
the transient stability of grid-forming converters with both and p'g_ref are set to 260 W and –175 W, respectively.
active power injection and absorption.
B. Quantitative Analysis of Voltage Amplitude Regulation
D. Mechanism of Transient Instability pg
After a fault (e.g., the grid voltage sag) occurs, the limitation pg_ref IV
III
of active power transfer decreases, as illustrated in Fig. 5. II
There are basically two types of transient instability, e.g., with I
or without equilibrium points. δ'g_pmax
Referring to Fig. 5, if no equilibrium point exists after a δg_pmax δg
fault, pg will always be either less than pg_ref or greater than Vg=15 V
p'g_ref. Per (3), this results in either dδg/dt>0 or dδg/dt<0, caus- p'g_ref Vg=19 V
ing δg to monotonically increase or decrease towards +∞ or –∞. Vg=20 V
Vg=21 V
Consequently, the grid-forming converter loses synchroniza-
tion with the grid. If equilibrium points exist where dδg/dt = 0, Fig. 6. pg−δg curves with voltage amplitude regulation.
δg may reach into an equilibrium point and stay on it, which 0.6
Vg=15 V III IV
0
Vg=19 V
pg Before fault
dδg/dt(Hz)
dδg/dt(Hz)

0.4 Vg=20 V
After fault Vg=21 V –0.2
pg_ref 0.2 I Vg=15 V
I Vg=19 V
δ'g_pmax
III II –0.4 Vg=20 V
0 IV II
Vg=21 V
δg_pmax δg –0.6
0 1 2 3 –3 –2 –1 0
p'g_ref δg(rad) δg(rad)
(a) Phase portrait in Case I (b) Phase portrait in Case II
Fig. 5. pg−δg curves before and after faults. Fig. 7. Phase portraits with voltage amplitude regulation.
IEEE POWER ELECTRONICS REGULAR PAPER

pg (δg_pmax,pg_max) 0
pg_ref Rgs=0 Ω
Rgs=0.3 Ω
–0.5

p'g_max(kW)
2 Rgs=0.6 Ω

pg_max(kW)
δ'g_pmax
δg_pmax –1 Rgs=0 Ω
δg 1
Rgs=0.3 Ω
p'g_ref –1.5 Rgs=0.6 Ω
(δ'g_pmax,p'g_max) 0
0 20 40 60 80 100 0 20 40 60 80 100
Fig. 8. pg−δg curve with critically stable states.
Vg(V) Vg(V)
As illustrated in Fig. 8, there exists a critically stable state
between stable and unstable states, where a single equilibrium (a) pg_max−Vg curves in Case I (b) p'g_max−Vg curves in Case II
point exists. In the critically stable state, the active power
Fig. 9. pg_max-Vg and pg_max-Vg curves with different Rgs.
transfer limitation pg_max or p'g_max is equal to its reference pg_ref
or p'g_ref, respectively. We denote the voltage amplitude of the Thus, the validity of the quantitative analysis method in this
grid-forming converter in the critically stable state as Vgc. As section is proven.
previously discussed, the transient stability improves as Vg C. Effect of Line Resistances on Voltage Amplitude Control
increases. To prevent transient instability, Vg should be greater
than Vgc. Hence, it is important to derive Vgc. To further explore the effect of voltage regulation on tran-
The active power transfer limitation pg_max or p'g_max is the sient stability improvement, we plot pg_max-Vg curves for Case
extremum of the power-angle curve. By setting dpg/dδg = 0, I and p'g_max-Vg curves for Case II, as shown in Fig. 9. The
the power angle δg_pmax or δ'g_pmax at the extremum point can be derivatives of pg_max-Vg and p'g_max-Vg curves are derived as
derived as dpg_max 3Vs (sin  g_pmax X gs − cos  g_pmax Rgs ) 3RgsVg
= + (11)
arctan(− X gs / Rgs ) + π,Rgs  0 2( X gs + Rgs ) 2( X gs 2 + Rgs 2 )
2 2
dVg
 g_pmax =  (5)
 π / 2, Rgs = 0 
dpg_max 
3Vs (sin  g_pmax 
X gs − cos  g_pmax Rgs ) 3RgsVg
= + (12)
arctan(− X gs / Rgs ),Rgs  0 dVg 2( X gs + Rgs )
2 2
2( X gs 2 + Rgs 2 )

 g_pmax = (6) From (5) and (6), it can be deduced that δg_pmax ∈ [π/2, π)
− π / 2, Rgs = 0
and δg_pmax ∈ [−π/2, 0). Therefore, from (11) and (12), we
Subsequently, according to (1), the power transfer limitations
have
can be expressed as
pg_max = pg_ref  3Vs (sin  g_pmax X gs − cos  g_pmax Rgs )
 0
 2( X gs 2 + Rgs 2 )
3[VgcVs sin  g_pmax X gs + Vgc (Vgc − Vs cos  g_pmax ) Rgs ] (7)
=  3V (sin   
2( X gs + Rgs )
2 2  s g_pmax X gs − cos  g_pmax Rgs )
 0 (13)
   2( X gs + Rgs )
2 2
pg_max = pg_ref
 3RgsVg

3[VgcVs sin  g_pmax  ) Rgs ]
X gs + Vgc (Vgc − Vs cos  g_pmax (8)  0
=  2( X gs 2 + Rgs 2 )
2( X gs + Rgs )
2 2

When Rgs = 0, both dpg_max/dVg and dp'g_max/dVg are constants,

By solving (7) and (8), Vgc can be derived as
which refer to the two straight lines in Fig. 9. When Rgs ≠ 0,
[ V 2 ( R cos 
g_pmax − X gs sin  g_pmax ) + (8 / 3) pg_ref Rgs ( X gs + Rgs )
2 2 2
 s gs according to (11), dpg_max/dVg increases as Vg increases. As
Vgc = +Vs ( Rgs cos  g_pmax − X gs sin  g_pmax )] / 2 Rgs , Rgs  0 illustrated in Fig. 9(a), pg_max increases more rapidly with the
 increase of Vg. Obviously, increasing Rgs will strengthen the
(2 pg_ref X gs ) / 3Vs , Rgs = 0
effect of transient stability improvement through voltage am-
(9) plitude regulation in Case I. However, from (12), dp'g_max/dVg
or is negative at beginning and then becomes positive as Vg in-
[ V 2 ( R cos    
g_pmax − X gs sin  g_pmax ) + (8 / 3) pg_ref Rgs ( X gs + Rgs )
2 2 2 creases. Hence, increasing Rgs will weaken the effect of transi-
 s gs ent stability improvement through voltage amplitude regula-
Vgc = +Vs ( Rgs cos  g_pmax 
− X gs sin  g_pmax )] / 2 Rgs , Rgs  0
 tion in Case II, as shown in Fig. 9(b). Notably, a considerable
 − (2 p  X ) / 3V , R = 0
g_ref gs s gs resistance, such as Rgs = 0.6 Ω (where Xgs = 4 Ω), will signifi-
(10) cantly deteriorates the effect of transient stability improvement.
By solving (9) and (10) with the parameters in Table I and Therefore, in Case II, it is advisable to reduce the resistance of
setting pg_ref and p'g_ref to 260 W and −175 W, respectively, transmission lines to mitigate its adverse impact. Additionally,
Vgc in Case I and Case II can be calculated as 19.98 V and a virtual positive/negative resistance can be employed together
19.88 V, respectively. As illustrated in Fig. 6 and Fig. 7, in with the proposed method in Case I/II to further improve the
both Case I and Case II, Curves I and II with Vg < Vgc are effect of voltage amplitude regulation. However, due to page
unstable, while Curves III and IV with Vg > Vgc are stable. limits, these techniques will not be explored in this paper.
IEEE POWER ELECTRONICS REGULAR PAPER

pg_ref=200 W
pg_ref=260 W
IV. TRANSIENT STABILITY IMPROVEMENT THROUGH REACTIVE pg_ref=300 W –1
2.5
POWER INJECTION
–1.5

δg(rad)
2 c1' c2' c3'

δg(rad)
c1 c2 c3
A. Fundamental Principle 1.5 –2 p'g_ref = –150 W
1 p'g_ref = –175 W
Per (4), there is a strong correlation between reactive power –2.5 p'g_ref = –200 W
and voltage amplitudes. It is evident that injecting reactive 0.5
power is able to increase the voltage amplitude. Thus, based 0 200 400 600 800 0 200 400 600 800
qg_ref(Var) qg_ref(Var)
on the analysis in Section III, this section proposes transient
stability improvement through reactive power injection. (a) qgc in Case I (b) qgc in Case II
By substituting (2) into (4) and considering that Vg = vgd_ref, Fig. 12. Critical reactive power qgc obtained by the graphical method.
the relationship between Vg and δ can be derived as avoid transient instability. Therefore, reactive power injection
Vg ( g , qg_ref ) = is an effective method to improve the transient stability of the
3Vs (cos  g X gs + sin  g Rgs )V0 kqv − 2 S0 ( X gs 2 + Rgs 2 ) grid-forming converter.
(14)
6 X gsV0 kqv B. Quantitative Analysis of Reactive Power Injection
[2 S0 ( X gs 2 + Rgs 2 ) − 3Vs (cos  g X gs + sin  g Rgs )V0 kqv ]2 Similar to voltage amplitude regulation, there exists a criti-
+ cally stable state between stable and unstable states. We repre-
+24 X gsV0 kqv ( X gs 2 + Rgs 2 )( S0 vg_ref + qg_ref V0 kqv ) sent the reactive power of the grid-forming converter in the
6 X gsV0 kqv critically stable state as qgc. In the reactive power control mode,
By substituting (14) into (1), pg−δg curves are obtained, as the formula describing the relationship between pg, δg, and
plotted in Fig. 10. As the grid voltage Vs drops from 110 Vrms qg_ref is complicated. To avoid complex calculations, we use
to 15 Vrms, the voltage amplitude reference Vg_ref shown in the graphical method to derive qgc. By substituting pg = pg_ref
Fig. 3(b) drops with Vs. Curve I is the pg−δg curve after the into (1) as a constraint and combining with (14), the relation-
grid voltage sag. By increasing qg_ref, Curves II, III, IV are ship between δg and qg_ref can be written as
obtained. The main parameters used in the analysis are listed pg_ref =
in Table I. In addition, pg_ref and p'g_ref are 260 W and –175 W, 3{Vg ( g , qg_ref )Vs sin  g X gs + Vg ( g , qg_ref )[Vg ( g , qg_ref ) − Vs cos  g ]Rgs }
respectively. 2( X gs 2 + Rgs 2 )
As shown in Fig. 10 and Fig. 11, with the increase of qg_ref,
(15)
the limitation of active power transfer increases. When qg_ref is
Per (15), the δg−qg_ref curves can be plotted, as depicted in
large enough, the equilibrium points are re-established to
Fig. 12. Based on the previous analysis, there exists only one
pg equilibrium point at the critically stable state, which is char-
pg_ref
IV acterized by only one intersection point between qg_ref = C
III
II (where C is a constant) and the δg−qg_ref curve, as the dots
I shown in Fig. 12. The abscissas of the dots represent the val-
ues of qgc. The main parameters used in Fig. 12 are given in
δg_pmax δg Table I. As shown in Fig. 12, qgc ≈ 490 Var when pg_ref = 260
qg_ref = 0 Var W, and qgc ≈ 440 Var when p'g_ref = −175 W. The calculated
qg_ref = 400 Var
p'g_ref qg_ref = 500 Var values of qgc are consistent with the results shown in Fig. 10
qg_ref = 600 Var and Fig. 11. In both Case I and Case II, Curves I and II with
Fig. 10. pg−δg curves with reactive power injection. qg_ref < qgc correspond to unstable systems, while Curves III
and IV with qg_ref > qgc are stable.
qg_ref=0 Var IV III
0.5 qg_ref=400 Var 0 Rgs=0 Ω –0.2
Rgs=0.3 Ω
p'g_max(kW)

qg_ref=500 Var –0.1 1.2

dδg/dt(Hz)

0.4 II
qg_ref=600 Var Rgs=0.6 Ω
dδg/dt(Hz)

I
pg_max(kW)

0.3 –0.2 –0.4

I 0.8
0.2 –0.3 qg_ref=0 Var
III qg_ref=400 Var –0.6 Rgs=0 Ω
0.1 II –0.4 0.4
IV qg_ref=500 Var Rgs=0.3 Ω
0 –0.5 qg_ref=600 Var –0.8 Rgs=0.6 Ω
0
0 1 2 3 –3 –2 –1 0 0 1 2 3 4 5 6 0 1 2 3 4 5 6
δg(rad) δg(rad) qg_ref(kVar) qg_ref(kVar)
(a) Phase portrait in Case I (b) Phase portrait in Case II (a) pg_max−qg_ref curves in Case I (b) p'g_max−qg_ref curves in Case II
Fig. 11. Phase portraits with reactive power injection. Fig. 13. pg_max−qg_ref and p'g_max−qg_ref curves with different Rgs.
IEEE POWER ELECTRONICS REGULAR PAPER

C. Effect of Line Resistances on Reactive Power Injection drop with vs. Low-frequency oscillations are observed in all
the waveforms of pg, δg, and igg, which indicates that an unsta-
To further analyze the influence of reactive power injec-
ble transient response is triggered by the grid voltage sag. To
tion on active power transfer limitation, pg_max−qg_ref and
improve the transient stability, vgf is regulated to 20 V after a
p'g_max−qg_ref curves are drawn with the parameters in Table I,
fault occurs, which is a suggested value in Section III. The
as illustrated in Fig. 13. In Case I, increasing Rgs enhances
relevant simulation results are shown in Fig. 14(b). As shown,
the effect of transient stability improvement through reactive
no oscillation is observed during the transient process, indicat-
power injection, as shown in Fig. 13(a). In Case II, increas-
ing an improvement in transient stability. The simulation re-
ing Rgs deteriorates the effect of transient stability improve-
sults confirm the theoretical analysis on voltage regulation.
ment through reactive power injection [see Fig. 13(b)]. The
The simulation results, illustrated in Fig. 15(a), show that
conclusion drawn is analogous to that of voltage amplitude
when the grid-forming converter is operating in the reactive
regulation. Likely, virtual impedance methods can also be
power control mode with qg_ref = 0 Var, the grid voltage drops
used to strengthen the effect of reactive power injection.
from 110 Vrms to 15 Vrms, and the converter voltage refer-
V. SIMULATION AND EXPERIMENTAL RESULTS ence Vg_ref drops following vs to 15 V. Oscillations of qg, pg, δg,
and igg occur, demonstrating that the grid-forming converter
We perform simulation and experimental studies in this sec- loses synchronization with the grid. As shown in Fig. 15(b),
tion for verification purposes. The simulation and experi- this instability can be removed by increasing qg_ref to 500 Var
mental results clearly verify the effectiveness of the proposed after a fault occurs, which is a suggested value in Section IV.
methods for transient stability improvement. These simulation results confirm the theoretical analysis on
the reactive power injection.
A. Simulation Results
B. Experimental Results
We built a MATLAB/Simulink (R2022a) model with the
system schematic and control structure in Fig. 1, the system We conducted experiments for further verification purposes,
and control parameters in Table I, and the active power refer- and a photo of the experimental setup is presented in Fig.16.
ence set to 260 W. The grid-forming converter was implemented by a three-phase
As illustrated in Fig. 14(a), when the grid-forming converter converter experiment platform, which includes the power
operates in the voltage control mode, the grid voltage vs drops module, control module, sampling module, etc. The control
from 110 Vrms to 15 Vrms, and the converter voltages vgf module was implemented by a digital signal processor (Texas
vs drops vs drops
vs : [50 V/div]
vs : [50 V/div]
qg : [400 Var/div]

pg : [200 W/div] pg : [200 W/div]

δg : [2 rad/div]
δg : [2 rad/div]
igg : [10 A/div]
igg : [10 A/div]
Time : [2 s/div] Time : [2 s/div]
(a) Without voltage amplitude regulation (a) Without reactive power injection
vs drops vs drops
vs : [50 V/div] vs : [50 V/div]

qg : [400 Var/div]
pg : [200 W/div]
pg : [200 W/div]
δg : [2 rad/div]
δg : [2 rad/div]
igg : [10 A/div] igg : [10 A/div]

Time : [2 s/div] Time : [2 s/div]

(b) With voltage amplitude regulation (b) With reactive power injection
Fig. 14. Simulated transient response of the grid-forming converter with and Fig. 15. Simulated transient response of the grid-forming converter with and
without voltage amplitude regulation, including the grid voltage vs, active without reactive power injection, including the grid voltage vs, reactive
power pg, power angle δg, and current igg. power qg, active power pg, power angle δg, and current igg.
IEEE POWER ELECTRONICS REGULAR PAPER

Obviously, all the experimental results are in accordance with

Oscilloscope DC power supply
theoretical analyses and simulation results.

VI. CONCLUSIONS

This article has proposed two methods—voltage regulation

and reactive power injection, to improve the transient stability
of grid-forming converters, and token into account both the
Grid-forming conerver AC power supply injection and absorption of active power. The proposed meth-
ods can easily be implemented and significantly improve the
Fig. 16. Photo of the experimental setup. transient stability of grid-forming converters. The effect of
Instruments TMS320F28379D). The grid-forming converter transient stability improvement through the proposed methods
was fed by one dc power supply (ITECH IT6526), and its out- are quantitatively revealed in detail and intuitively described
put was tied to the grid via an LCL filter. Three-phase induc- through phase portraits. The critical values of voltage ampli-
tors and resistors emulated the grid impedance Xgs and Rgs, tude and reactive power in transient periods are calculated to
respectively. An ac programmable power supply (ITECH guide parameter design. All the theoretical analysis has been
IT7625) provided the grid voltage vs. An oscilloscope (LeCroy corroborated by simulations and experimental results.
HDO8108) captured all the experimental waveforms.
REFERENCES
The system and control parameters of experiments are con- [1] F. Blaabjerg, R. Teodorescu, M. Liserre, and A. V. Timbus, “Overview
sistent with those in the simulation. Fig. 17(a) and Fig. 17 (b) of control and grid synchronization for distributed power generation sys-
depict the experimental results of transient response when the tems,” IEEE Trans. Ind. Electron., vol. 53, no. 5, pp. 1398–1409, Oct.
grid voltage drops from 110 Vrms to 15 Vrms, without voltage 2006.
[2] J. Rocabert, A. Luna, F. Blaabjerg, and P. Rodríguez, “Control of power
amplitude regulation and with voltage amplitude regulated to converters in AC microgrids,” IEEE Trans. Power Electron., vol. 27, no.
20 V, respectively. Likewise, Fig. 18(a) and (b) show the ex- 11, pp. 4734–4749, Nov. 2012.
perimental results of transient response during the grid voltage [3] P. Kundur. Power System Stability and Control. New York, NY, USA:
drops from 110 Vrms to 15 Vrms, without reactive power in- McGraw-Hill, 1994.
[4] U. Markovic, O. Stanojev, P. Aristidou, and G. Hug, “Partial grid form-
jection and with 500 Var reactive power injection, respectively. ing concept for 100% inverter-based transmission systems,” in Proc.
It is evident that voltage amplitude regulation and reactive IEEE Power & Energy Society General Meeting, Portland, OR, USA,
power injection significantly enhance the transient stability. Aug. 2018.
vs drops vs drops
vs : [50 V/div] vs : [50 V/div]
qg : [400 Var/div]
pg : [200 W/div]
pg : [200 W/div]
δg : [2 rad/div]
δg : [2 rad/div]

igg : [10 A/div] igg : [10 A/div]

Time : [2 s/div] Time : [2 s/div]

(a) Without voltage amplitude regulation (a) Without reactive power injection

vs drops vs drops
vs : [50 V/div] vs : [50 V/div]

pg : [200 W/div] qg : [400 Var/div]

pg : [200 W/div]
δg : [2 rad/div]
δg : [2 rad/div]
igg : [10 A/div] igg : [10 A/div]

Time : [2 s/div] Time : [2 s/div]

(b) With voltage amplitude regulation (b) With reactive power injection
Fig. 17. Experimental transient response of the grid-forming converter with Fig. 18. Experimental transient response of the grid-forming converter with
and without voltage amplitude regulation, including the grid voltage vs, and without reactive power injection, including the grid voltage vs, reactive
active power pg, power angle δg, and current igg. power qg, active power pg, power angle δg, and current igg.