4648 IEEE TRANSACTIONS ON SMART GRID, VOL. 10, NO.

4, JULY 2019

Transient Angle Stability of Virtual Synchronous

Generators Using Lyapunov’s Direct Method
Zhikang Shuai , Senior Member, IEEE, Chao Shen, Xuan Liu , Member, IEEE,
Zuyi Li , Senior Member, IEEE, and Z. John Shen, Fellow, IEEE

Abstract—With an increasing number of distributed energy main control strategies, droop control and virtual synchronous
resources integrated into the power system, inverters need to control, were proposed to address these issues [3]–[7].
take on the corresponding responsibility for the security and sta- Droop control scheme can realize power sharing for the
bility of the system. Virtual synchronous generators (VSGs) are
proposed to mimic dynamic characteristics of traditional rota- parallel operation of inverters without the necessity of critical
tional synchronous generators (RSGs) to compensate for the communication links [3], [4]. This advantage can facilitate the
loss of inertia and reserve capacity. Similar to RSGs, VSGs plug- and-play of DERs. Moreover, the limited inertia provided
will experience transient angle instability under certain condi- by the power filter can also prevent frequency and voltage from
tions, which likely threatens the system security. In this paper, abrupt change, especially in the islanded mode [5]. Compared
transient angle stability of a VSG is investigated by Lyapunov’s
direct method. The deteriorative effect of reactive power control with droop-controlled inverters, virtual synchronous genera-
loop on transient angle stability is first analyzed and then volt- tors (VSGs) can compensate the loss of inertia and the rotation
age variation is incorporated into an approximate Lyapunov’s reserve capacity of the system by mimicking the output char-
direct method. In this method, the inverter internal voltage is acteristics of RSGs [6], [7]. From the perspective of the power
treated as a parameter rather than a state variable. Moreover, grid, inverters controlled as VSGs have an operating charac-
the influence of different parameters on transient angle stability
is studied. Finally, an enhanced control strategy is presented to teristic that is much closer to that of the conventional RSGs.
improve the transient angle stability by adjusting the reference Thus, VSGs attract great attention and are widely studied in
power. Numerical simulation results are presented to validate the recent years.
effectiveness of the proposed method and the enhanced control. Similar to RSGs, VSGs will experience severe transient sta-
Index Terms—Virtual synchronous generator, transient angle bility problems, like transient angle stability (TAS) problem,
stability, Lyapunov’s direct method, reactive power control which results from the introduction of the correlated large
loop. inertia. Many literatures are dedicated to model reduction,
control advancement and dynamic stability analysis with
VSGs [8]–[11]. These studies focus on the improvements of
operation characteristics and small signal stability analysis of
I. I NTRODUCTION VSGs. It is reported in [12] that there is a quasi-steady angle
ISTRIBUTED energy resources (DERs) are integrated
D to the power grid through power electronic converters to
realize energy conversion. In general, power electronic con-
stability problem with voltage-controlled converters. In [13],
the similarity between the dynamic responses of the dc side
of inverters and the rotor motion equation of RSGs is dis-
verters are thermal capacity limited and have less inertia com- cussed and the equivalent mathematical model is built. While
pared to traditional rotational synchronous generators (RSGs). the transient stability of a VSG remains to be explored.
This would make them unable to withstand network distur- Since there are similarities between droop-controlled invert-
bances and bring considerable challenges to the secure oper- ers and VSGs in some aspects [11], the investigation on
ation of the system, especially in high proportion power elec- the transient instability of droop-controlled inverters may be
tronic system, such as in islanded microgrid [1], [2]. Thus, two used for reference [14]–[20]. A two-order swing equation of
a droop-controlled inverter is established and a Lyapunov can-
Manuscript received January 2, 2018; revised April 30, 2018 and July 18, didate function is thus derived to explore transient stability
2018; accepted August 14, 2018. Date of publication August 22, 2018; date of
current version June 19, 2019. This work was supported in part by the National problems [15]. For the application of the derived function
Natural Science Foundation of China under Grant 51622702, and in part by the in distribution networks, the system damping and the line
Higher National Excellent Doctoral Dissertation of Special Funds under Grant resistors are considered [16]. The effect of the current lim-
201441. Paper no. TSG-00005-2018. (Corresponding author: Zhikang Shuai.)
Z. Shuai, C. Shen, and X. Liu are with the College of Electrical iter on the transient instability mechanism of droop-controlled
and Information Engineering, Hunan University, Changsha 410082, inverters was analyzed in [17] and [18]. In these studies, the
China (e-mail: [email protected]; [email protected]; effect of reactive power control loop has not been consid-
[email protected]).
Z. Li and Z. J. Shen are with the Department of Electrical and Computer ered yet, so the transient response of inverters has not been
Engineering, Illinois Institute of Technology, Chicago, IL 60616 USA (e-mail: fully examined and the complicate instability mechanism of
[email protected]; [email protected]). droop-controlled inverters in some specific cases has not been
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org. thoroughly explained. A method to study the transient sta-
Digital Object Identifier 10.1109/TSG.2018.2866122 bility of droop-controlled inverters is introduced by solving
1949-3053 c 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: STATE UNIV NY BINGHAMTON. Downloaded on September 22,2023 at 23:01:53 UTC from IEEE Xplore. Restrictions apply.
SHUAI et al.: TAS OF VSGs USING LYAPUNOV’S DIRECT METHOD 4649

the formulated Linear Matrix Inequality (LMIs) based on the

multi-variable Popov criterion [19], [20]. However, it lacks
in the detailed explanation of the instability mechanism of
droop-controlled inverters. Moreover, large virtual inertia in
VSGs would make their transient response different from that
of droop-controlled inverters.
Transient stability issues of conventional RSGs have
been extensively studied [21]–[23]. It was pointed out
in [21] and [22] that transient angle instability results from
sustained imbalance between the electromagnetic and mechan-
ical torque in the RSGs. Different from RSGs, VSGs can
regulate output active power as well as reactive power through Fig. 1. The main topology and control block of a VSG.
active and reactive power control loops. During transient
period, like grounded fault, there would be a sudden increase
in the output reactive power of an inverter, which would lead
to an abrupt decrease in the internal voltage amplitude of
the inverter through reactive power control loop. This will
reduce transient angle stability margin of VSGs. Traditional
approaches to estimate transient stability of RSGs are no
longer viable to VSGs. Thus, it is necessary to investigate
on transient stability problems of VSGs and special attention
needs to be paid to the effect of the reactive power control
loop on system’s transient stability.
Due to possible threats to the safe operation of power system
results from the transient instability of VSGs, a series of fault
ride through control strategies were proposed [8], [24]–[26].
These control strategies mainly aim at limiting the fault cur-
rent, but lack of discussions on the improvement of the Fig. 2. Control loops of a VSG, (a) Active power control loop, (b) Reactive
transient stability of VSGs. It is found in this paper that power control loop.
transient angle instability of a VSG is inherently caused by
the imbalance between the reference active power and output
active power. For system’s stable operation, the improvement
II. T RANSIENT A NGLE S TABILITY A NALYSIS OF VSG S
of the transient stability and the limitation of the fault current
must be both considered at the same time. A. Principle of a VSG Control
In this paper, transient instability mechanism of VSGs is The main topology of a VSG connected to a three-phase
investigated and an approximate Lyapunov’s direct method is grid is shown in Fig. 1. It includes three inverter legs oper-
proposed to estimate the transient stability of VSGs, which ated using pulse width modulation (PWM) to transfer power
takes the effect of reactive power control loop into account. from dc side to ac side and LCL filters to reduce the voltage
Then, an enhanced control strategy is proposed to improve and current ripple. LT is the transformer equivalent inductor
the transient angle stability by adjusting the reference active and Ll represents for the equivalent line inductor (also includes
power. The main contributions of this paper are summarized Rl ). VSG control can be divided into two parts: active power
as follows: control loop (swing equation) and reactive power control loop
1) The deteriorative effect of reactive power control loop (Q-V droop). The active power control loop is responsible
on transient angle stability is illustrated through P-δ curve, to mimic mechanical rotor motion of RSGs, as shown in
and the voltage variation has been incorporated into an Fig. 2 (a). The reactive power control loop is used to regulate
approximate Lyapunov’s direct method. output reactive power of a VSG, which is shown in Fig. 2 (b).
2) The effects of different parameters on the attraction Active power control loop and reactive power control loop
region of VSGs are investigated. produce phase angle and voltage magnitude respectively to
3) An enhanced control strategy is proposed to improve the generate reference sine waves, which are then delivered
transient stability of VSGs, and the corresponding mechanism to the PWM module to produce trigger signals. A virtual
has been analyzed. speed governor is added into active power control loop to
The rest of the paper is organized as follows. In Section II, adjust the reference active power according to the system
the transient angle stability analysis of a VSG is presented, frequency [11].
and the instability mechanism is illustrated. In Section III, an The state equations of the whole system are represented
approximate Lyapunov’s direct method is proposed to estimate on the common reference frame of the three-phase power
transient stability of the system and the effects of different grid. The axis (D-Q) is the common reference frame rotat-
parameters on the attraction region are analyzed. The enhanced ing at a frequency ωg , where axis (d-q) is the reference frame
control is introduced in Section IV. Conclusions are made in of a VSG, as shown in Fig. 3. δ is the transformation angle
Section VI. of a VSG with respect to the common reference frame, which
Authorized licensed use limited to: STATE UNIV NY BINGHAMTON. Downloaded on September 22,2023 at 23:01:53 UTC from IEEE Xplore. Restrictions apply.
4650 IEEE TRANSACTIONS ON SMART GRID, VOL. 10, NO. 4, JULY 2019

Fig. 3. Reference frame transformation.

is defined as power angle

dδ
= ω − ωg (1)
dt
Fig. 4. Equivalent circuit of a VSG connected to power grid, (a) without
The swing equation of the VSG control block is considering reactive power control, (b) considering reactive power control.
d2 θ
J = T ∗ − Tem − Dp (ω − ω0 ) (2)
dt2
where J represents the virtual inertia and Dp denotes the active power control loop ranges from a few milliseconds to tens of
damping coefficient. ω0 is the reference angular speed and milliseconds.
T em is the output torque, which can be calculated by Tem = For a RSG, the studied time scale of a transient angle
P/ω0 . The reference active power P∗ is generated by the speed stability problem is from hundreds of milliseconds to a few
governor, which can be calculated as follows seconds [21]. It is of the same order for a VSG since a smaller
inertia in a VSG incorporates a less damping ability compared
P∗ = P0 − Kp (ω − ω0 ) (3) to a RSG. Thus, it is known that the reactive power control
Combining equation (1)-(3), we get loop would reach the steady state as soon as possible during
the transient period. From this point of view, the derivative
d2 δ equation of inverter’s internal voltage E in (5) can be turned
J = T0 − Tem − D(ω − ω0 ) (4)
dt2 into a quasi-steady equation, which means that E can be treated
where T0 = P0 /ω0 and D = Dp + Kp /ω0 , D is the equiva- as a parameter variable rather than a state variable. Based on
lent damping. The mathematical equations of reactive power this assumption, a VSG model in (4)-(6) can be simplified into
control loop is a two-order derivative equation plus a quasi-steady equation,
as shown below
dE
= Q0 − Q − Dq (E − E0 )
K (5)
dt d2 δ E2 G + EVg B sin δ + EVg G cos δ
J = T0 − − D(ω − ω0 )
where K is the voltage integral coefficient and Dq represents dt2 ω0
the Q-V coefficient. E0 and Q0 are the reference voltage and (8)
reference reactive power respectively. 0 = Q0 − E2 B − EVg B cos δ + EVg G sin δ − Dq (E − E0 )
As shown in [9], the dynamic characteristics of the inductor
and the capacitor can be ignored since they response much (9)
faster than the power control loop. Utilizing the output voltage
and phase of a VSG, active and reactive powers transferred B. Deteriorative Effect of Reactive Power Control Loop on
between a VSG and the grid should be as following: Transient Angle Stability
P = E2 G − EVg B sin δ − EVg G cos δ The equivalent equation of a VSG connected to an infi-
Q = E B + EVg B cos δ − EVg G sin δ
2
(6) nite bus is derived in part A of this section. It is known
that a voltage-controlled inverter can be represented by a volt-
where G = R/(R2 + X 2 ) and B = −X/(R2 + X 2 ). X and R age source in order to study the transient characteristic [27].
are the inductance and resistance of the transfer impedance Based on this assumption, the equivalent circuit can be estab-
between a VSG and the power grid respectively. lished in Fig. 4 (b), where V g is the equivalent RMS voltage
The block diagram of the reactive power control loop is of the grid, and Z represents the equivalent line impedance.
shown in Fig. 2 (b). Similar to the active power control loop, Take the grid voltage angle as a reference, the angle differ-
the time constant of the reactive power control loop can be ence between a VSG and the grid is the power angle δ. To
calculated by better illustrate theeffect of the reactive power control loop
on transient angle stability, the equivalent model of a VSG
τv = K/Dq (7)
with constant inverter’s internal voltage (without considering
It is known that in RSGs, the process of establishing reactive power control loop) is established in Fig. 4(a) as
field flux is much faster than the motion of the mechanical a comparison. Only power angle δ is controllable through P-f
rotor [21]. For a VSG, the time constant τv of the reactive droop loop in Fig. 4(a).

Authorized licensed use limited to: STATE UNIV NY BINGHAMTON. Downloaded on September 22,2023 at 23:01:53 UTC from IEEE Xplore. Restrictions apply.
SHUAI et al.: TAS OF VSGs USING LYAPUNOV’S DIRECT METHOD 4651

the criterion for transient angle stability intuitively [28]. In

pre-fault steady state, the system operates at point δ s . Then,
a grounded fault occurs and the operating point moves from
curve I to curve II. Under this condition, power imbalance is
introduced and δ increases according to (19) (called acceler-
ation process). When the fault is cleared at δc , the operating
point moves from curve II to curve I. The output active power
is larger than the reference active power and δ decreases to
reach for a stable equilibrium point (called deceleration pro-
cess). δmax represents the maximum power angle the system
would reach after the fault is cleared. Without considering the
energy loss caused by the resistor and damping windings, δmax
can be calculated by the condition that the acceleration area S1
equals to the deceleration area S2 . The criterion for transient
angle stability is that δmax < δ u .
When the effect of reactive power control loop is con-
sidered, the equivalent circuit in Fig. 4 (b) can be used to
describe the system. As we know that a grounded fault would
lead to a sudden increase in the output reactive power of
a VSG because of the voltage amplitude difference between
a VSG and the grid. According to the principle of reac-
tive power control loop in Fig. 2 (b), inverter’s internal
voltage E would decrease to a certain value, which is man-
aged by the coefficient Dq . When the fault is cleared at δc ,
inverter’s internal voltage cannot immediately restore to the
rated value, not only due to the existence of the integrator,
but also for the reason that the operating point δc is far away
from δ s . Thus, the maximum values Pmax1 and Pmax2 of P-δ
Fig. 5. P-δ curve during transient period, (a) constant inverter’s internal
voltage, (b) considering the effect of reactive power control loop.
curves I and II will decrease to Pmax1 and Pmax2 respec-
tively, as shown in Fig. 5 (b). The decrease in the maximum
value of P-δ curve will lead to an increase in the accelera-
To analyze system transient response of a VSG in P-δ curve, tion area during the fault and a decrease in the deceleration
equation (8) can be turned into area after the fault is cleared, which may result in the max-
imum power angle δmax larger than δ u at the same fault
d2 δ EVg
J = T − sin δ t − D(ω − ω0 ) (10) clearing time. This is called the deteriorative effect of reac-
dt2 |Z|ω0 tive power control loop on transient angle stability. It should
where T  = T0 -E2 G/ω0 and δ t = δ-ϕ, Z and ϕ are the line be noted that though inverter’s internal voltage keeps chang-
impedance and the impedance angle, respectively. In transmis- ing, which is caused by the variation of the power angle δ
sion grid, resistors are usually ignored because the inductance during the fault, it remains a lower value compared to the
X is much larger than the resistor R. Under this condition, rated value.
the transferred active power of a VSG can be represented as
P≈EVsinδ/X, which is commonly used in the conventional
power system. However, the resistor cannot be ignored in the III. T RANSIENT A NGLE S TABILITY E STIMATION
distribution network, which would make the active power is W ITH LYAPUNOV ’ S D IRECT M ETHOD
not an odd function of the power angle. So we adopt the trans-
A. Derivation of the Proposed Lyapunov’s Direct Method
lation of the coordinate to change the equation (11) into an
odd function, thus the P-δ curve in Fig. 5 can be used to Since the deteriorative effect of reactive power control
analyze the transient power angel issues in the distribution loop on transient stability margin is analyzed, this part will
network. give a new Lyapunov’s direct method to estimate transient
So based on the coordinate translation, the corresponding stability considering reactive power control loop. To evalu-
relationship between active power P and power angle δ of the ate the transient stability margin, state variables in (8) need
model in Fig. 4 (a) is shown in Fig. 5 (a). Curve I describes to be moved from the equilibrium (δ s , ωg ) to (0, 0). This
the relationship between P and δ of the pre-fault and post-fault is necessary for Lyapunov theorem which is typically for-
systems while curve II is the P-δ curve of the system under mulated for a system with the equilibrium point at zero.
a fault. There are two equilibrium points denoted by δ s and Thus, define
δ u where power balance can be reached. δ s represents the sta- 
ble equilibrium point and δ u is the unstable equilibrium point. x1 = δ − δ s
(11)
Extended equal area criterion (EEAC) can be used to explain x2 = ω − ωg

Authorized licensed use limited to: STATE UNIV NY BINGHAMTON. Downloaded on September 22,2023 at 23:01:53 UTC from IEEE Xplore. Restrictions apply.
4652 IEEE TRANSACTIONS ON SMART GRID, VOL. 10, NO. 4, JULY 2019

Then, derivative equations in (8) turn into parameter E in (14) and (15) since the voltage variation is
 not included in the derived Lyapunov function. However, it is
dx1
dt = x2 extremely difficult to construct the Lyapunov function when
EVg B EVg G
J dxdt2 = T − ω0 sin(x1 + δ s ) − ω0 cos(x1 + δ s ) − Dx2 voltage dynamic is considered [23], [31].
(12) The objective of this section is not to show how we can
successfully tackle a hard problem of constructing an effective
where T = T  − D(ωg − ω0 ). Lyapunov function for highly nonlinear equations of (4)-(5).
First integral method can be used to derive a candidate Instead, this difficulty will be overcome by translating changes
Lyapunov function for a two-order system [29], which satisfies of E from a state variation to a parameter variation. Basic ideas

n of the proposed method are presented below.
∂fi
=0 (13) Mathematically speaking, transient stability analysis by
∂xi Lyapunov’s direct method is to investigate whether the state-
i=1

By neglecting the damping term Dx2 , the system can be variables of the system belong to the attraction domain at
seen as a conservative system, and the detailed process to con- the time of the fault clearance. It means that the post-fault
struct a Lyapunov candidate function can be found in [29]. The system state must be known ahead of time. Thus, the numer-
damping term Dx2 can be incorporated into the derived func- ical integration of the faulted system is inherently needed to
tion as shown in [30]. So the Lyapunov function V(x1 , x2 ) of obtain the initial value of the post-fault system. In the conven-
the system can be expressed in the form as tional power system, system state variables, like frequency and
voltage phase angle, are monitored to evaluate the operation
1 2 EVg B   s   condition of the system online [21]. For the system consisted
V(x1 , x2 ) = Jx − Tx1 + cos δ + x1 − cos δ s
2 2 ω0 with VSGs, state variables, like voltage amplitude, voltage
EVg G   s   phase angle and frequency, can be obtained from the controller
− sin δ + x1 − sin δ s
ω0 directly. Once these necessary state variables are achieved, the
D2 2 operation condition of the system during the transient period
+ Dλx1 x2 + λx (14) can be evaluated based on the proposed method.
2J 1
Since the mathematical model of a VSG connected to an
where λ is a constant (0 < λ < 1), Jx22 /2 represents the vir- infinite bus has been reduced into a two-order derivative equa-
tual rotor kinetic energy and Tx1 represents the rotor potential tion plus a quasi-steady equation in (8)-(9), the candidate
energy relative to the reference frame. The third term and the Lyapunov function (14) can still be used to evaluate the tran-
fourth term evaluate the magnetic stored energy and the dissi- sient stability of a VSG. It means for each determined E, (14)
pated energy of the line impedance. The fifth and sixth terms can guarantee the asymptotical stability for state variables
are the approximate dissipated energy of virtual damping. The within the attraction domain. However, the s.e.p., u.e.p., and
positive definite of the proposed Lyapunov function and the the critical energy Vc would not be a constant value for a given
semi- negative definite of dV/dt are given in Appendix B. fault condition. Instead, δ s , δ u and Vc need to be updated with
As a suitable Lyapunov function has been constructed, the a new E according to (9) at each time step.
procedure for determining the critical fault clearance time Based on the reduced model in (8)-(9), the algorithm of the
based on Lyapunov direct method is presented as follows: proposed approach is summarized as follows:
Step 1: Power-flow solution of the pre-fault system under Step 1: Power-flow solution of the pre-fault system under
steady state. steady state.
Step 2: Obtain the stable and unstable equilibrium point Step 2: Calculate the initial value of the inverter’s internal
(s.e.p. and u.e.p.) of the post-fault system, which voltage E of the faulted system according to (9).
are denoted by (δ s , ωg ) and (δ u , ωg ) respectively. Step 3: Obtain s.e.p. and u.e.p. of the post-fault system,
Step 3: Calculating the critical energy value Vc by which are denoted by (δ s , ωg ) and (δ u , ωg )
substituting (δ u , ωg ) into Lyapunov function. respectively.
Step 4: Execute forward numerical integration of the Step 4: Calculating the critical energy value Vc by
faulted system with the pre-fault steady-state substituting (δ u , ωg ) into the Lyapunov function
operating point as initial conditions. At each time with E calculated by step 2.
step, calculate V(δ, ω) to compare with Vc . The Step 5: Execute forward numerical integration of the
time the value of V(δ, ω) equals to Vc is defined faulted system by (8)-(9) with the pre-fault
the critical clearing time tc . steady-state operating point as initial conditions.
The unstable equilibrium point of the studied system can be At each time step, we calculate E by (9) and then
calculated by δ u = π − δ s . Then, the critical value Vc can be recalculate δ s = δ s (E), δ u = δ u (E) to obtain Vc (δ s ,
computed from δ u and δ s δ u , 0, E). The time V(δ, ω, E) equals to Vc is called
  the critical clearing time.
Vc = V δ s , δ u , ωg (15)
It is observed from the proposed method that additional
The Lyapunov function without considering the effect of computation burden is needed to recalculate δ s , δ u and Vc
reactive power control loop has been derived above. It should for each inverter’s internal voltage E during transient period.
be noted here that V as well as Vc are not relevant to the It means that the proposed method depends on the voltage

Authorized licensed use limited to: STATE UNIV NY BINGHAMTON. Downloaded on September 22,2023 at 23:01:53 UTC from IEEE Xplore. Restrictions apply.
SHUAI et al.: TAS OF VSGs USING LYAPUNOV’S DIRECT METHOD 4653

disturbed trajectory which determines the critical energy value

Vc . However, as discussed above, state variables are inherently
needed to determine the stable condition of the system [21].
As shown in step 5, the proposed method only needs another
variable, inverter’s internal voltage E, to recalculate the critical
energy. Moreover, the proposed method avoids the difficulty in
finding the unstable equilibrium point, which is crucial impor-
tant in the system considering voltage dynamics. Because
stable equilibrium point (δ s , ωg ) can be easily obtained by
assigning the right side of equation (12) to 0 for each deter-
mined E, the unstable equilibrium point δ u can be calculated
by δ u = π − δ s . Equations to solve for δ s , δ u and Vc are all
algebraic equations, which do not need much storage space. Fig. 6. Estimated attraction domain of the studied system.

B. Numerical Validation of the Proposed Lyapunov’s

Direct Method
Since an approximate Lyapunov’s direct method has been
proposed in part A of this section, simulation results based
on MATLAB/Simulink and PSCAD/EMTDC would be given
in this part to validate the feasibility and effectiveness of the
proposed method. First, the effect of reactive power control
loop on the attraction region of the system would be analyzed
to prove the correctness of theoretical analysis. Then, two sim-
ulation cases would be carried out to verify the feasibility of
the proposed method in predicting the transient stability of the
system.
1) The Effect of Reactive Power Control Loop on the
Attraction Region: As analyzed in Section II-B, the reac-
tive power control loop will lead to the decrease of the
inverter’s internal voltage during the transient period, which
would deteriorate the transient angle stability of VSGs. The
system shown in Fig. 1 is taken as the research object, and the
attraction region is evaluated. The operation condition is that
a three-phase grounded fault occurs and the grid voltage V g
is decreased to 30% of the rated value at 1.5s. And after the
fault is cleared, V g recovers to the rated value. The detailed
parameters are shown in Table I of Appendix A.
The changes of the estimated attraction domain at different
time are shown in Fig. 6. Vcr represents the attraction domain
of the studied system without considering the effect of reac-
Fig. 7. Lyapunov’s direct method to estimate system stability, (a) Stable
tive power control loop. Vcr1 and Vcr2 represent the attraction condition when fault is cleared at 2s, (b) unstable condition when fault is
domain of the studied system at 2s and 2.12s respectively. It cleared at 2.12s.
can be observed that the attraction region without considering
the effect of reactive power control is much bigger than the Fig. 7 shows Lyapunov’s direct method in determining tran-
actual attraction domain of the system. This result is consis- sient stability. Vp , Vk and Vt are the system potential energy,
tent with the theoretical analysis. In this situation, a system kinetic energy and total energy respectively. Vcrf and Vcrp
judged to be stable can probably go unstable if the effect of represent the critical energies of the system with and without
reactive power control is not considered. This will lead to seri- considering the effect of reactive power control loop respec-
ous error in transient stability assessment. As time changes, tively. It can be observed from both Fig. 7 (a) and (b) that
the attraction domain of the studied system would be reduced Vcrp is much bigger than Vcrf , which is in accordance with
due to the decrease of the inverter’s internal voltage, which is the curves Vcr and Vcr1 in Fig. 6. In both conditions, the total
indicated by Vcr1 and Vcr2 in Fig. 6. energy Vt does not exceed the critical energy Vcrp . Thus, the
2) A VSG Connected to an Infinite Bus: First, a VSG con- system could be judged to be stable without the consideration
nected to a three-phase voltage source is used as a study of the effect of reactive power control loop. However when
system, as shown in Fig. 1, to validate the effectiveness of the proposed Lyapunov’s direct method is adopted, the esti-
the proposed Lyapunov’s direct method in predicting tran- mated result is different: the critical energy Vcrf is 62 at 2s, as
sient angle stability. The operation condition is the same as shown in Fig. 7 (a), while its value is 51 at 2.12s, as shown
the aforementioned ones in this section. in Fig. 7 (b), which is identical to Vcr1 and Vcr2 in Fig. 6.

Authorized licensed use limited to: STATE UNIV NY BINGHAMTON. Downloaded on September 22,2023 at 23:01:53 UTC from IEEE Xplore. Restrictions apply.
4654 IEEE TRANSACTIONS ON SMART GRID, VOL. 10, NO. 4, JULY 2019

Fig. 8. Angular speed versus power angle trajectory when fault is cleared
at 2s and 2.12s respectively. Fig. 10. Topology of two-VSGs connected to an infinite bus.

Fig. 11. Lyapunov’s direct method to estimate system stability when the
fault is cleared at 2.3s.

Fig. 9. Three-phase voltage and current simulation results, (a) stable condi-
tion when fault is cleared at 2s, (b) unstable condition when fault is cleared
at 2.12s. Fig. 12. Angular speed versus power angle trajectory of VSG1 and
VSG2 when fault is cleared at 2.3s.

It can be observed that when the fault is cleared at 2s, the

total energy Vt does not reach the critical energy Vcrf , which Fig. 9 gives corresponding three-phase voltage and current
means the system is stable. When the fault is cleared at 2.12s, simulation results under certain conditions. iabc is the output
the total energy Vt exceeds the critical energy, which means current and vabc represents the voltage at the LC filter. Output
the system is unstable in this condition. current and voltage oscillate at a huge magnitude when the
Numerical simulations are applied to validate the cor- fault is cleared at 2.12s, while the system can return to the
rectness of the proposed Lyapunov’s direct method in steady state when fault is cleared at 2s. The critical fault clear-
MATLAB/Simulink, and the results are shown in Fig. 8. When ing time suggested by the proposed Lyapunov’s direct method
the fault is cleared at 2s, power angle δ returns to the stable is 2.1s for the studied system. So we can conclude that reac-
equilibrium point. When the fault is cleared at 2.12s, system tive power control loop has great effect on the transient angle
state variables diverge and cannot reach a equilibrium point. stability of VSGs and the proposed method is able to predict
The simulation result conflicts with the estimate result with- the system stability.
out considering the effect of reactive power control loop but 3) Two-VSGs Connected to an Infinite Bus: To further
it is consistent with the proposed Lyapunov’s direct method. validate the effectiveness of the proposed Lyapunov’s direct

Authorized licensed use limited to: STATE UNIV NY BINGHAMTON. Downloaded on September 22,2023 at 23:01:53 UTC from IEEE Xplore. Restrictions apply.
SHUAI et al.: TAS OF VSGs USING LYAPUNOV’S DIRECT METHOD 4655

Fig. 13. Estimated attraction domain with the variation of reference active Fig. 14. Estimated attraction domain with the variation of damping
power. coefficient.

method, a grid-connected microgrid that consists of two VSGs Fig. 13. It is observed that the attraction region decreases
is established in MATLAB/Simulink. The system topologyis with the increase of the reference active power, which can be
shown in Fig. 10, and the relevant detailed parameters are explained by P-δ curve in Fig. 5: a large reference active power
given in Table II of Appendix A. The control strategy shown would lead to a large acceleration area during fault period and
in Fig. 2 is adopted for VSG1 and VSG2. As shown in Fig. 10, a small deceleration area during the post-fault period. This
there is a fault occurs to load 3 at 1.5s, and the equivalent would deteriorate the system transient stability, so it is easier
grid voltage decrease to 27% of the rated value. The fault is for a VSG with large capacity to lose transient stability.
cleared at 2.3s and the equivalent grid voltage recovers to the 2) Effect of Damping Coefficient D: Fig. 14 gives the esti-
rated value. mated attraction domain with the variation of the damping
Fig. 11 shows the result of the proposed Lyapunov’s direct coefficient D (It should be noted that D equals the sum of Dp
method in predicting the transient stability of the studied and Kp /ω0 ). From the theoretical analysis in Section II, a large
system. The result illustrates that the total energy Vt1 of damping coefficient would consume more energy during tran-
VSG1 does not reach the critical energy value Vcrf 1 , which sient period and would slow down the increase of the power
means VSG1 can return to the steady-state when the fault is angle δ, as shown by equation (12). So a large damping coeffi-
cleared at 2.3s. However, it can be seen that the total energy cient is beneficial for the system transient angle stability. It can
Vt2 of VSG2 goes beyond the critical energy value Vcrf 2 , be observed from Fig. 14 that the estimated attraction domain
which means VSG2 cannot return to the equilibrium point increases as the damping coefficient increases, which agree
when the fault is cleared at 2.3s. Simulation results are shown with the theoretical analysis. However, a large damping coef-
in Fig. 12, and it is observed that VSG1 can return to the equi- ficient may lead to the performance degradation in terms of
librium point when the fault is cleared at 2.3s. While the state frequency control of a VSG. Thus, there is a tradeoff between
variables of VSG2 diverge and cannot reach a stable point. the capability to maintain the stability during transient periods
This demonstrates the effectiveness of the proposed method and the frequency deviation in steady-state.
in predicting system stability with two VSGs. Similarly, the 3) Effect of Line Impedance Z: Four different parameters,
proposed method is also applicable for a system consisted of Z1 = 0.01 + j0.15pu, Z2 = 0.02 + j0.3pu, Z3 = 0.03 + j0.45pu
multi-VSGs. and Z4 = 0.1 + j0.28pu, are chosen to identify the effect of
the line impedance on the attraction region of the system.
C. The Effect of Different Parameters on Attraction Region Z 1 , Z 2 and Z 3 have the same impedance angle, but dif-
Since an approximate Lyapunov’s direct method has been ferent amplitudes. Z 2 and Z 4 have the same amplitude but
proposed as shown in part A of this section, the effect of different impedance angles. As indicated by equation (11),
different parameters on the attraction region of VSGs- domi- the line transmission capacity is in inverse proportion to the
nated system would be analyzed in this part. As analyzed in line impedance, which means large impedance may drive the
Section II-B, transient angle instability of a VSG is inherently system into the instability situation during transient periods.
caused by the imbalance between the reference active power Illustrated intuitively from curves of Z 1 , Z 2 and Z 3 in Fig. 15,
and the output active power. So the line impedance, damping it can be concluded that the larger the line impedance ampli-
coefficient and reference active power, which have effects on tude is, the smaller the attraction region of the system is. And
the response of the swing equation, would be taken as vari- the comparison between Z 2 and Z 4 implies that a smaller
ables. Moreover, Q-V droop coefficient, which determines the impedance angle can increase the attraction region of the
dynamic of the inverter’s internal voltage, would also be taken system. This is because the resistor can increase the dissi-
into account in this part. pated energy of the system as shown in equation (14), which
1) Effect of Reference Active Power P0 : The reference would help to maintain the system stability.
active power P0 varies from 10kW to 30kW to identify its 4) Effect of Q-V Droop Coefficient Dq : The effect of Q-V
effect on the attraction region, and the result is shown in droop coefficient Dq on the attraction region of the system

Authorized licensed use limited to: STATE UNIV NY BINGHAMTON. Downloaded on September 22,2023 at 23:01:53 UTC from IEEE Xplore. Restrictions apply.
4656 IEEE TRANSACTIONS ON SMART GRID, VOL. 10, NO. 4, JULY 2019

Fig. 17. Implementation of the enhanced control.

Fig. 15. Estimated attraction domain with the variation of line impedance.

Fig. 18. P-δ curve during transient period with the enhanced control.

Fig. 16. Estimated attraction domain with the variation of Q-V droop As pointed out in [21] that sustained imbalance between
coefficient.
input active power and output active power is the reason for
the angle instability, fault ride through control can limit output
power which results in increased power imbalance. This would
is shown in Fig. 16. During the transient period, inverter’s further increase the acceleration area, as shown in Fig. 5 (b),
internal voltage would decrease according to the Q-V droop and reduce the stability margin.
curve. And the larger the Q-V droop coefficient Dq is, the Though inverters’ thermal capacity is limited, they can be
less inverter’s internal voltage drops. A large inverter’s internal controlled flexibly to avoid the occurrence of transient angle
voltage is beneficial to the transient stability of a VSG, as instability. For traditional RSGs, power input can be regulated
shown in Fig. 6. It can also be observed from Fig. 16 that the by adjusting governors to improve transient stability [21]. For
attraction domain of the system increases with the increase of VSGs, the reference active power is physically equivalent to
the Dq . The result validates the correctness of the theoretical the input power in RSGs. Based on the instability mechanism
analysis. analysis of a VSG, the transient stability margin of a VSG can
be improved by reducing the reference active power according
to the severity of the fault.
IV. C ONTROL S TRATEGY FOR E NHANCING
Fig. 17 gives the implemental block diagram for the
T RANSIENT A NGLE S TABILITY
enhanced control by adding a power compensation loop. The
A. A New Method to Enhance Transient Angle Stability grid voltage is detected to determine whether there is a fault
Before introducing the new method, the conflict between the or not. When the grid voltage decreases to a certain value V set ,
requirement to decrease the output current determined by the an additional torque T would be produced to decrease the
limited thermal capacities of switch devices and the increase reference torque. The relationship between the voltage drop
of the output power required to avoid power imbalance needs level and the additional torque is regulated by
to be analyzed. It is known that inverters are flexible and con-  
T = M Vrated − Vg (16)
trollable so that they can be controlled to mimic dynamic
characteristics of RSGs. However, one of the major differ- where M is the proportional coefficient. V g is the amplitude of
ences between these two energy conversion devices is that the grid voltage. V rated represents the rated voltage magnitude,
inverters have limited thermal capacities [27]. During transient and its value is 311V in the studied system. M needs to be
period, inverters will export a large current which probably tuned to ensure the transient stability of the system. Generally,
burns switch devices. Thus, fault ride through control is nec- M can be calculated by M = T/ V.
essary for inverters to limit the output current during transient Fig. 18 shows the effectiveness of the proposed control in
period, like virtual impedance control [24], [25]. improving transient angle stability by P-δ curve. It should be

Authorized licensed use limited to: STATE UNIV NY BINGHAMTON. Downloaded on September 22,2023 at 23:01:53 UTC from IEEE Xplore. Restrictions apply.
SHUAI et al.: TAS OF VSGs USING LYAPUNOV’S DIRECT METHOD 4657

Fig. 19. Estimated attraction domain of the studied system at 2.12s.

Fig. 20. Lyapunov’s direct method to estimate system stability with enhanced
control when fault is cleared at 2.12s.

noted that T = P/ω0 . It can be seen that an additional

torque T would decrease the reference torque as well as the
acceleration area S1 . When the fault is cleared and the grid
voltage restores to the rated value, T is set to zero to make
sure that a VSG can deliver the rated power in steady-state
according to the P-f droop curve. The decrease of accelera-
tion area S1 can improve the transient stability of the studied
system.
To incorporate the enhanced control to Lyapunov’s direct
method, there is no need to change the proposed method in
Section III-B because the enhanced control has no effect on
the post-fault system. Once the fault is cleared, the reference
active power would return to its normal value to guaran-
tee the power regulation in steady-state. As we know that Fig. 21. Angular speed versus power angle trajectory with or without
the derived Lyapunov function is based on the post-fault enhanced control when fault is cleared at 2.12s.
system, and the initial state of the system state variables
is needed to determine the system stability. The proposed
method would change the fault trajectory of the studied that the proposed enhanced control can improve the attraction
system so that the initial state of the post-fault system would domain of the studied system. It should be noted that a proper
be changed. Thus, equation (14) is used to calculate the compensation coefficient M is chosen to make P equals to
system energy, and the initial state variables of the post-fault 5kW in this condition.
system with the proposed method can be obtained with the The effectiveness of the enhanced control to improve the
following equation: transient stability is demonstrated in simulation results. The
 result on the transient stability of the system predicted by the
E2 G+EVg B sin δ+EVg G cos δ
J ddt2δ = T0 − T −
2
ω0 − D(ωi − ω0 )
proposed Lyapunov’s direct method is shown in Fig. 20, where
0 = Q0 − E2 B − EVg B cos δ + EVg G sin δ − Dq (E − E0 ) Vcrf , Vt , Vp and Vk represent the critical energy, the total
(17) energy, the potential energy and the kinetic energy respec-
tively. The subscript 3 and 4 denote the conditions without
where T is introduced as a input torque compensation com-
and with the proposed enhanced control. It can be seen that
ponent. When the initial state of the post-fault system is
Vcrf 4 is larger than Vcrf 3 , which is consistent with the differ-
determined by equation (17), system stability can be evaluated
ence of the attraction region in Fig. 19. When the proposed
with the proposed Lyapunov’s direct method.
enhanced control is adopted, Vt4 is much lower than Vcrf 4 at
2.12s, so the system is predicted to be stable. However, the
B. Numerical Validation of the Proposed Method total energy Vt3 has already exceeded Vcrf 3 at 2.12s when
In this part, Fig. 1 is used as the studied system and detailed the proposed method is not adopted, which means the system
parameters are shown in Table I Appendix A. The operation has lost stability in this condition. Simulation results on angu-
condition is the same as that in Section III-B. lar speed versus angle difference are given in Fig. 21. It is
As analyzed in part A of this section, the transient stability observed that the system with enhanced control can return to
margin of a VSG can be improved by decreasing the refer- the stable equilibrium point, while it loses stability without the
ence active power during the transient period. The validation proposed method at 2.12s. Furthermore, it can be seen that the
of the theoretical analysis is provided in Fig. 19, where Vcr2 proposed control can change the fault trajectory of the studied
and Vcr2 are the estimated attraction domain with and with- system and the initial state of the post-fault system would be
out the enhanced control at 2.12s, respectively. It can be seen changed. This is because the enhanced control slows down the

Authorized licensed use limited to: STATE UNIV NY BINGHAMTON. Downloaded on September 22,2023 at 23:01:53 UTC from IEEE Xplore. Restrictions apply.
4658 IEEE TRANSACTIONS ON SMART GRID, VOL. 10, NO. 4, JULY 2019

Fig. 22. Inner control block diagram of VSGs.

Fig. 24. Three-phase voltage and current simulation results when fault is
cleared at 2.3s.

Fig. 25. Implementation of VI-FCL control block diagram.

Fig. 23. Angular speed versus power angle trajectory when fault is cleared
at 2.3s (with and without inner control loops).

acceleration process by balancing the reference active power

and the output active power.

V. D ISCUSSION Fig. 26. Equivalent circuit of a VSG connected to power grid with virtual
impedance fault current limiting control strategy.
A. The Effect of Inner Control Loops
To validate the effect of inner control loops on the transient
response of VSGs, the inner control block diagram, which is
implemented in the abc reference frame, is shown in Fig. 22.
The outer voltage control loop is used to produce the reference
current and deliver them to the inner current control loop.
To track the sinusoidal signals, PR controllers are adopted to
realize zero steady-state error control [32], [33]. The system
operation condition is the same with Fig. 9 and parameters of
the inner control loops are shown in Table I. The three- phase
grounded fault occurs at 1.5s and it is cleared at 2.3s.
As predicted by the proposed Lyapunov’s direct method in
Fig. 8, the critical fault clearing time for the VSGs without Fig. 27. Three-phase voltage and current simulation results when fault is
cleared at 2s.
inner control loops is 2.1s. It can be observed from Fig. 23,
when the fault is cleared at 2.3s, the VSGs with inner con-
trol loops can return to the equilibrium point while the VSGs is taken as FRT control strategy and its effect on transient
without inner control loops lose stability. The three-phase volt- angle stability of VSGs would be analyzed through simula-
age and current curves with inner control loops are shown in tion results. The basic control loops are given in Fig. 25 and
Fig. 24. It can be concluded from the simulation results that the equivalent circuit is shown in Fig. 26. It should be noted
the inner control loops can help to stabilize the transient sta- that the designed operation condition is the same with Fig. 9.
bility of the system since the equivalent impedance and the The three-phase grounded fault occurs at 1.5s and it is cleared
inverter’s internal voltage have been changed by the intro- at 2s. According to the voltage drop depth, the added virtual
duction of inner control loops [27]. The detailed theoretical impedance Zv equals to 0.05+j0.2 p.u.
analysis would be carried out in the future researches. It can be observed from Fig. 27 that the magnitude of fault
current is limited to about two times of the rated value due
B. The Effect of Virtual Impedance Fault Current Limiting to the increased line impedance added by VI-FCL control
Method (VI-FCL) strategy. As illustrated in Fig. 15 that large impedance would
As mentioned in Section IV part A, fault ride through reduce the attraction region of VSGs. So when VI-FCL con-
control strategy (FRT) is widely used to limit inrush fault cur- trol strategy is triggered during fault period, the system would
rent of VSGs to protect semiconductor devices [34]. In this be more easily driven into instability region. Compared with
part, virtual impedance fault current limiting method (VI-FCL) Fig. 9 (a), the system equipped with VI-FCL control strategy

Authorized licensed use limited to: STATE UNIV NY BINGHAMTON. Downloaded on September 22,2023 at 23:01:53 UTC from IEEE Xplore. Restrictions apply.
SHUAI et al.: TAS OF VSGs USING LYAPUNOV’S DIRECT METHOD 4659

TABLE I TABLE II
PARAMETERS OF THE S TUDIED S YSTEMS PARAMETERS OF THE G RID -C ONNECTED S YSTEMS

A PPENDIX B
Positive Definite of Derived Lyapunov Function: The
derived Lyapunov function in (14) can be classified into

loses stability at the same fault clearing time (2s), which is V(x1 , x2 ) = VKE + VPE + VDE (B1)
consistent with the theoretical analysis. It is clear that V(0,0)=0 and VKE = Jx22 /2 ≥ 0. Thus, we need
to prove that VPE and VDE are positive.
VI. C ONCLUSION
EVg B   s  
This paper investigates the transient angle stability of VSGs VPE = −Tx1 + cos δ + x1 − cos δ s
ω0
by taking the reactive power control loop into account. The
EVg G   s  
transient instability mechanism is revealed and the insta- − sin δ + x1 − sin δ s
bility phenomenon is illustrated. Moreover, an approximate ω0
EVg x1     
Lyapunov’s direct method is adopted to evaluate the transient = sin δ s − ϕ + u − sin δ s − ϕ du (B2)
stability of the system. Based on the mechanism analysis, |Z|ω0 0
a new control strategy is proposed to improve transient sta- Define δ ts = δ s -ϕ. Then equation (B2) can be changed into
bility of the system. The main conclusions can be drawn as
EVg x1    
follows: VPE = sin δ ts + u − sin δ ts du (B3)
1) The reactive power control loop will reduce inverter’s |Z|ω0 0
internal voltage through Q-V droop curve during transient
When π − 2δ ts ≥ x1 ≥ 0, sin(δ ts + x1 ) ≥ sin δ ts . sin(δ ts +
period, and this may drive the system into instability situation
u) − sin δ ts and du are all positive, so VPE ≥ 0. When 0 ≥
easily.
x1 ≥-π -2δ ts , sin(δ ts + x1 ) ≤ sin δ ts . sin(δ ts + u)-sinδ ts and du
2) Inverter’s internal voltage can be treated as a parameter
are all negative, so VPE ≥ 0. Thus, we can prove that when
variable rather than a state variable in transient angle stability
π -2δ ts ≥ x1 ≥ -π -2δ ts , VPE ≥ 0.
analysis. The proposed approximate Lyapunov’s direct method
To prove that VDE is positive definite, we can obtain first
can predict the transient stability of the system.
integral of motion by first multiplying both sides of the sec-
3) A VSG with large capacity intends to lose transient
ond equation of (12) by x1 and integrating with respect to
angle stability during fault period. The attraction region of the
time t
system increases with the increase of the damping coefficient
t t
and the Q-V droop coefficient. Large line impedance would •
J x2 x1 dt + Dx2 x1 dt
reduce the stability margin of the system, and the resistor is 0 0
beneficial for the transient stability for VSGs. EVg t   
4) Transient angle instability of a VSG is inherently caused = sin δ ts + x1 − sin δ ts x1 dt (B4)
|Z|ω0 0
by the imbalance between the reference active power and the
output power. By decreasing the reference active power during The left side of equation (B4) can be rewritten as follows:
the fault period, transient stability margin can be improved. t
• • t t
•2
J x2 x1 dt = J x1 x1 − J x1 dt
0 0 0
A PPENDIX A t
= Jx2 x1 − J x22 dt (B5)
See Tables I and II. 0

Authorized licensed use limited to: STATE UNIV NY BINGHAMTON. Downloaded on September 22,2023 at 23:01:53 UTC from IEEE Xplore. Restrictions apply.
4660 IEEE TRANSACTIONS ON SMART GRID, VOL. 10, NO. 4, JULY 2019

t 1 2
Dx2 x1 dt = Dx (B6) By substituting (B12) and (B13) into (B11), we get
0 2 1
dV Dλ EVg   
By substituting equations (B5) and (B6) into (B4), (B4) can = −(1 − λ)Dx22 + sin δ ts − sin δ ts + x1 x1
dt J |Z|ω0
be rewritten as (B16)
t 1
Jx2 x1 − J x22 dt + Dx12 Since 0< λ <1, the first part of (B16) is negative and
0 2
t  ts     
EVg sin δ ts − sin δ ts + x1 x1 < 0 (B17)
= sin δ + x1 − sin δ ts x1 dt (B7)
|Z|ω0 0
for π -2δ ts ≥ x1 ≥-π -2δ ts as proved in the previous part.
From the above equation, we can obtain the following Equation (B16) says that Lyapunov V(x1 ,x2 ) meets the semi-
relationship by multiplying D in both sides negativeness of its time derivative for all λ ∈(0,1) unless the
t system goes out of the stable region π -2δts ≥ x1 ≥-π -2δts .
D x22 dt
0
D2 2 D EVg t    R EFERENCES
= Dx2 x1 + x − sin δ ts + x1 − sin δ ts x1 dt
2J 1 J |Z|ω0 0 [1] J. M. Guerrero et al., “Distributed generation: Toward a new energy
(B8) paradigm,” IEEE Ind. Electron. Mag., vol. 4, no. 1, pp. 52–64,
Mar. 2010.
[2] Z. Shuai et al., “Microgrid stability: Classification and a review,” Renew.
The last term of (B8) can be included into the previous two Sustain. Energy Rev., vol. 58, pp. 167–179, May 2016.
terms by a constant λ, which is defined as damping coefficient. [3] Y. Li, D. M. Vilathgamuwa, and P. C. Loh, “Design, analysis, and real-
time testing of a controller for multibus microgrid system,” IEEE Trans.
t D2 2 Power Electron., vol. 19, no. 5, pp. 1195–1204, Sep. 2004.
D x22 dt = Dλx1 x2 + λx (B9) [4] N. Pogaku, M. Prodanovic, and T. C. Green, “Modeling, analysis and
0 2J 1 testing of autonomous operation of an inverter-based microgrid,” IEEE
Trans. Power Electron., vol. 22, no. 2, pp. 613–625, Mar. 2007.
It is known that the left side of the equation (B9) is positive, [5] J. M. Guerrero, L. Hang, and J. Uceda, “Control of distributed unin-
so we get that terruptible power supply systems,” IEEE Trans. Ind. Electron., vol. 55,
no. 8, pp. 2845–2859, Aug. 2008.
D2 2 [6] S. D’Arco and J. A. Suul, “Equivalence of virtual synchronous machines
VDE = Dλx1 x2 + λx ≥ 0 (B10) and frequency-droops for converter-based MicroGrids,” IEEE Trans.
2J 1 Smart Grid, vol. 5, no. 1, pp. 394–395, Jan. 2014.
Thus, we can prove that V(x1 , x2 ) is positive definite. [7] Q. C. Zhong and G. Weiss, “Synchronverters: Inverters that mimic
synchronous generators,” IEEE Trans. Ind. Electron., vol. 58, no. 4,
Semi-Negative Definite of dV/dt: The negativeness of time pp. 1259–1267, Apr. 2011.
derivative of the derived Lyapunov function can be proved by [8] Z. Shuai, W. Huang, C. Shen, J. Ge, and Z. J. Shen, “Characteristics
using chain rule for partial derivatives, which can be evaluated and restraining method of fast transient inrush fault currents in syn-
chronverters,” IEEE Trans. Ind. Electron., vol. 64, no. 9, pp. 7487–7497,
as follows Sep. 2017.
dV ∂V dx1 ∂V dx2 [9] Z. Shuai, Y. Hu, Y. Peng, C. Tu, and Z. J. Shen, “Dynamic stability anal-
= + (B11) ysis of synchronverter-dominated microgrid based on bifurcation theory,”
dt ∂x1 dt ∂x2 dt IEEE Trans. Ind. Electron., vol. 64, no. 9, pp. 7467–7477, Sep. 2017.
[10] X. Guo, W. Liu, and Z. Lu, “Flexible power regulation and current-
The results of partial derivative are limited control of grid-connected inverter under unbalanced grid voltage
∂V EVg B  s  faults,” IEEE Trans. Ind. Electron., vol. 64, no. 9, pp. 7425–7432,
= −T − sin δ + x1 Sep. 2017.
∂x1 ω0 [11] J. Liu, Y. Miura, and T. Ise, “Comparison of dynamic characteristics
  between virtual synchronous generator and droop control in inverter-
EVg G D2
− cos δ s + x1 + Dλx2 + λx1 based distributed generators,” IEEE Trans. Power Electron., vol. 31,
ω0 J no. 5, pp. 3600–3611, May 2016.
[12] H. Lin, C. Jia, J. M. Guerrero, and J. C. Vasquez, “Angle stability
EVg    D2
=− sin δ ts − sin δ ts + x1 + Dλx2 + λx1 analysis for voltage-controlled converters,” IEEE Trans. Ind. Electron.,
|Z|ω0 J vol. 64, no. 8, pp. 6265–6275, Aug. 2017.
(B12) [13] L. Xiong et al., “Static synchronous generator model: A new per-
spective to investigate dynamic characteristics and stability issues of
∂V grid-tied PWM inverter,” IEEE Trans. Power Electron., vol. 31, no. 9,
= Jx2 + Dλx1 (B13)
∂x2 pp. 6264–6280, Sep. 2016.
[14] D. Marx, P. Magne, B. Nahid-Mobarakeh, S. Pierfederici, and B. Davat,
By substituting (B12) and (B13) into (B11), we get “Large signal stability analysis tools in DC power systems with constant
power loads and variable power loads—A review,” IEEE Trans. Power
•
dV EVg x1    •
Electron., vol. 27, no. 4, pp. 1773–1787, Apr. 2012.
=− sin δ ts − sin δ ts + x1 + Dλx2 x1 [15] P. Hart and B. Lesieutre, “Energy function for a grid-tied, droop-
dt |Z|ω0 controlled inverter,” in Proc. North Amer. Power Symp., 2014, pp. 1–6.
[16] Q. Sun, Y. Zhang, H. He, D. Ma, and H. Zhang, “A novel energy
D2 • • •
+ λx1 x1 +Jx2 x2 +Dλx1 x2 (B14) function-based stability evaluation and nonlinear control approach for
J energy Internet,” IEEE Trans. Smart Grid, vol. 8, no. 3, pp. 1195–1210,
May 2017.
The second equation of (12) can be rewritten as [17] H. Xin, L. Huang, L. Zhang, Z. Wang, and J. Hu, “Synchronous
instability mechanism of P-f droop-controlled voltage source converter
• EVg    D caused by current saturation,” IEEE Trans. Power Syst., vol. 31, no. 6,
x2 = sin δ ts − sin δ ts + x1 − x2 (B15)
|Z|ω0 J J pp. 5206–5207, Nov. 2016.

Authorized licensed use limited to: STATE UNIV NY BINGHAMTON. Downloaded on September 22,2023 at 23:01:53 UTC from IEEE Xplore. Restrictions apply.
SHUAI et al.: TAS OF VSGs USING LYAPUNOV’S DIRECT METHOD 4661

[18] L. Huang et al., “Transient stability analysis and control design Chao Shen received the B.S. degree in elec-
of droop-controlled voltage source converters considering cur- trical engineering from the College of Electrical
rent limitation,” IEEE Trans. Smart Grid, to be published, and Information Engineering, Hunan University,
doi: 10.1109/TSG.2017.2749259. Changsha, China, in 2015, where he is currently pur-
[19] M. Kabalan, P. Singh, and D. Niebur, “Nonlinear Lyapunov stability suing the Ph.D. degree. His research interests include
analysis of seven models of a DC/AC droop controlled inverter con- mathematical modeling, transient stability analysis,
nected to an infinite bus,” IEEE Trans. Smart Grid, to be published, and control of microgrid system.
doi: 10.1109/TSG.2017.2752146.
[20] Y. Zhang and L. Xie, “A transient stability assessment framework in
power electronic-interfaced distribution systems,” IEEE Trans. Power
Syst., vol. 31, no. 6, pp. 5106–5114, Nov. 2016.
[21] P. Kundur, Power System Stability and Control. New York, NY, USA:
McGraw-Hill Educ., 1994.
[22] P. Kundur et al., “Definition and classification of power system stability
IEEE/CIGRE joint task force on stability terms and definitions,” IEEE
Trans. Power Syst., vol. 19, no. 3, pp. 1387–1401, Aug. 2004.
[23] H. Sasaki, “An approximate incorporation of field flux decay into tran-
sient stability analyses of multimachine power systems by the second Xuan Liu (M’14) received the B.S. and M.S.
method of Lyapunov,” IEEE Trans. Power App. Syst., vol. PAS-98, no. 2, degrees in electrical engineering from Sichuan
pp. 473–483, Mar. 1979. University, China, in 2008 and 2011, respectively,
[24] J. He and Y. W. Li, “Analysis, design, and implementation of virtual and the Ph.D. degree in electrical engineering from
impedance for power electronics interfaced distributed generation,” IEEE the Illinois Institute of Technology, Chicago, IL,
Trans. Ind. Appl., vol. 47, no. 6, pp. 2525–2538, Nov./Dec. 2011. USA, in 2015. He is currently a Professor with the
[25] A. D. Paquette and D. M. Divan, “Virtual impedance current limiting College of Electrical and Information Engineering,
for inverters in microgrids with synchronous generators,” IEEE Trans. Hunan University, China. His research interests
Ind. Appl., vol. 51, no. 2, pp. 1630–1638, Mar./Apr. 2015. include smart grid security and operation and eco-
[26] A. H. Etemadi and R. Iravani, “Overcurrent and overload protection of nomics of power systems.
directly voltage-controlled distributed resources in a microgrid,” IEEE
Trans. Ind. Electron., vol. 60, no. 12, pp. 5629–5638, Dec. 2013.
[27] Z. Shuai, C. Shen, X. Yin, X. Liu, and J. Shen, “Fault analysis of
inverter-interfaced distributed generators with different control schemes,”
IEEE Trans. Power Del., vol. 33, no. 3, pp. 1223–1235, Jun. 2018.
[28] Y. Xue, T. Van Custem, and M. Ribbens-Pavella, “Extended equal area
criterion justifications, generalizations, applications,” IEEE Trans. Power
Syst., vol. 4, no. 1, pp. 44–52, Feb. 1989.
[29] N. G. Bretas and L. F. C. Alberto, “Lyapunov function for power systems
with transfer conductances: Extension of the invariance principle,” IEEE
Trans. Power Syst., vol. 18, no. 2, pp. 769–777, May 2003. Zuyi Li (SM’09) received the B.S. and M.S.
[30] Y.-H. Moon, B.-K. Choi, and T.-H. Roh, “Estimating the domain of degrees in electrical engineering from Shanghai
attraction for power systems via a group of damping-reflected energy Jiaotong University, Shanghai, China, in 1995 and
functions,” Automatica, vol. 36, no. 3, pp. 419–425, Mar. 2000. 1998, respectively, and the Ph.D. degree in elec-
[31] H. Khalil, Nonlinear Systems. Englewood Cliffs, NJ, USA: trical engineering from the Illinois Institute of
Prentice-Hall, 1996. Technology, Chicago, IL, USA, in 2002, where he
[32] J. Fang, Y. Tang, H. Li, and X. Li, “A battery/ultracapacitor hybrid is currently a Professor with the Electrical and
energy storage system for implementing the power management of vir- Computer Engineering Department. His research
tual synchronous generators,” IEEE Trans. Power Electron., vol. 33, interests include economic and secure operation
no. 4, pp. 2820–2824, Apr. 2018. of electric power systems, cyber security in smart
[33] D. N. Zmood and D. G. Holmes, “Stationary frame current regulation grid, renewable energy integration, electric demand
of PWM inverters with zero steady-state error,” IEEE Trans. Power management of data centers, and power system protection.
Electron., vol. 18, no. 3, pp. 814–822, May 2003.
[34] X. Lu, J. Wang, J. M. Guerrero, and D. Zhao, “Virtual-impedance-
based fault current limiters for inverter dominated AC microgrids,” IEEE
Trans. Smart Grid, vol. 9, no. 3, pp. 1599–1612, May 2018.

Zhikang Shuai (S’09–M’10–SM’17) received the

B.S. and Ph.D. degrees in electrical engineering
from the College of Electrical and Information Z. John Shen (S’89–M’94–SM’01–F’11) received
Engineering, Hunan University, Changsha, China, the B.S. degree in electrical engineering from
in 2005 and 2011, respectively. He was with Tsinghua University, China, in 1987 and the M.S.
Hunan University, as an Assistant Professor from and Ph.D. degrees in electrical engineering from
2009 and 2012, an Associate Professor in 2013, Rensselaer Polytechnic Institute, Troy, NY, USA, in
and a Professor in 2014. His research interests 1991 and 1994, respectively.
include power quality control, power electronics, He was a Faculty Member with the University of
and microgrid stability analysis and control. He Michigan–Dearborn, USA, in 1999 and 2004, and
was a recipient of the 2010 National Scientific and the University of Central Florida, USA, in 2004 and
Technological Awards of China, the 2012 Hunan Technological Invention 2012. He joined the Illinois Institute of Technology,
Awards of China, and the 2007 Scientific and Technological Awards from USA, in 2013 as a Grainger Chair Professor in elec-
the National Mechanical Industry Association of China. He is an Associate trical and power engineering. He has also held a courtesy professorship
Editor of the IEEE J OURNAL OF E MERGING AND S ELECTED T OPICS IN with Hunan University, China, since 2007, and with Zhejiang University,
P OWER E LECTRONICS, CSEE Journal of Power and Energy Systems, and the China, since 2013. His research interests include power electronics and power
Chinese Journal of Electrical Engineering. semiconductor devices.

Authorized licensed use limited to: STATE UNIV NY BINGHAMTON. Downloaded on September 22,2023 at 23:01:53 UTC from IEEE Xplore. Restrictions apply.