Comparison of Virtual Synchronous Generators

Dynamic Responses
Sara Yahia Altahir Mohamed Xiangwu Yan
School of Electrical & Electronic Engineering Electrical Engineering Department
North China Electric Power University North China Electric Power University
Beijing, China Baoding, China
[email protected] [email protected]

Abstract—Recently, researchers have proposed different Now, the VSG models recognized in literature can be
types of virtual synchronous generators and the performance has divided into several levels. According to the external
been investigated. The virtual synchronous generator has characteristics of the inverter and its ability to support the
proposed with the different order such as the second order, the islanding mode and grid-connected mode, VSGs can be
third order, the fifth order and the seventh order. However, the divided into two categories, VSG controlled-voltage source
second-order model is mainly used to mimic the synchronous and VSG controlled-current source [8].
generator characteristics. The motivation for applying the virtual
synchronous generator is its ability to contribute to the frequency Based on the nature of the reference signal out from the VSG,
control and its ability to share the load demand with other virtual three main groups are identified; current references, voltage
synchronous generators in the same system. This paper firstly references, and power references [8]. In continuation, The
introduces a comprehensive comparison of the dynamic VSG has proposed with different order such as the second
performance of two types of virtual synchronous generators and order, the third order, the fifth order and the seventh order.
provides the possible results of the different types. The dynamic However, the second-order voltage-source voltage-reference is
behaviors are compared under the parallel operation in a commonly used to mimic the synchronous generator [9].
microgrid. Secondly, the paper develops a dynamic response
control method for multi-parallel virtual synchronous On the basis of related literature, this paper introduces a
generators. The control method has verified through computer comprehensive comparison of the dynamic performance of
simulation. two types of VSGs and provides the possible results of the
different types. The dynamic behaviors are compared under
Keywords— virtual synchronous generator; parallel operation; the parallel operation in a microgrid. Then, a dynamic
virtual inertia; damping; dynamic response response Control method to avoid the asymmetry in the
control level is constructed and the effectiveness of the model
I. INTRODUCTION is verified by simulation. The method is suitable for
application because it’s simple and effective.
Recently, the traditional power system is changing from
centralized power generation based-synchronous generators The rest of this paper is organized as follows: in section II,
located in power stations to decentralized generation, with the a comparison of two types of VSG models is presented.
increase of the number of distributed energy resources (DER) Simulation results are included in section III with the
such as wind, photovoltaic, and other forms of clean discussion. In section IV, a new method is developed to
energy[1]. Most of the DERs use power electronic device such control the dynamic performance of the paralleled VSGs. The
as inverter to interface with the grid [2]. Compared to the performance of the new idea is investigated by simulation
synchronous generator which can support the grid frequency results in section V. Finally, the conclusion is presented in
by providing enough inertia and damping support, the DER section VI.
based-inverters have negative impacts on the system stability
due to the lack of sufficient damping and equivalent inertia Distributed Generator
[2]. To solve this problem, the virtual synchronous generator PCC
Line impedance
(VSG) control technology is proposed [3]. The concept of the
VSG is based on emulating the synchronous generator Vdc U I
dynamic characteristics.
Power
References [4]-[7], were the leading in this research. Calculation
Later, researchers have proposed different approaches for
providing additional inertia and increasing the system Pe Qe
stability. The structure diagram of the VSG is presented as in Voltage and Vref VSG
Pref
Fig.1. Now, the VSG has regarded as a promising technology
PWM
Current Control Control Qref
and started to become the main control method of DERs.
Fig. 1. Typical structure of VSG.
 II. VIRTUAL SYNCHRONOUS GENERATOR B. VSG Type#1
Often, it is convenient to use the term D (ω − ω 0 ) to
A. Fundamentals
emulate the damping effect when the grid is strong [15]-[17].
According to the basic principle of synchronous In islanding mode, this term can be also used to provide a
generators, the imbalance between the power produced and the frequency reference to ensure that the VSG frequency is
power consumed could produce amount of power depend stabilized at ω0 when reach the steady state, where ω0 is equal
upon the energy stored in the rotation mass [10]. This power is 50 or 60 Hz. A block diagram of the swing equation
given by the swing equation (1) [11], [12], where J is the with D (ω − ω 0 ) intended for VSG is depicted in Fig. 2. Then
moment of inertia, ω is the rotating frequency of the
(4) will be in the following form.
generator, Pm is the mechanical power, and Pe is the
electromagnetic power. This equation is considered as the core dω 1
J = Tref − Te − D (ω − ω 0 ) + (ω − ω ) (5)
of the VSG. dt ωm 0
dω (1)
Jω = Pm − Pe
dt This equation shows that, the effect of the frequency droop
is equivalent to the damping effect. However, the equivalent
In general, P − ω and Q − V droop function is used to frequency drooping coefficient ( 1 ω m ) is far larger than the
emulate the primary frequency regulation and voltage damping coefficient [18]. That means the damping coefficient
regulation of a synchronous generator [2], which are given as will affect the droop coefficient and results in power sharing
ω = ω 0 − m ( Pm − Pref ) (2) error [19]. However, if the frequency droop is not used, the
damping coefficient D can behave as a frequency droop
(
U = U 0 − n Q − Q ref ) (3) coefficient [20].

− ω0
Where ω0 is the rated angular velocity, U0 is the reference D
voltage amplitude, m is the frequency droop coefficient, n is +
the voltage droop coefficient, Pref and Qref are the rated active ω − Δω 1 Pm 1 Tm + − 1 1 θ
and reactive power, respectively. + m + ω − Js s
Te
The incorporation of (1) and (2) can ensure that the VSG ω
ω0 Pref
generates the desired active power and at the same time
maintains the system frequency stability. The significance of
(3) will not be discussed in this paper. Then, the swing Fig. 2. VSG type#1.
equation can be expressed as
C. VSG Type#2
dω 1 (4)
J = Tref − Te + (ω − ω ) In grid connected system, it is important that the VSG can
dt ωm 0 adjust its output power according to the grid frequency ω g . In
this case the VSG needs to employ a phase-locked loop (PLL)
Where Tref = Pref ω and Te = Pe ω . By (4), the frequency
to provide ω g [19], [21], and [22]. Then the rated frequency
dynamics can be determined and the active load demand can
ω0 needs to be replaced by ω g for producing the damping
be shared with other VSGs of the same type [13], [14].
power as in (6). In grid connected mode, when use this term,
It is often desirable to add the damping term to the swing the damping toque is equal to zero at steady state and does not
equation for the purpose of damping the low frequency influence the output power. In islanding mode ω g is the
oscillation and to ensure that the VSG frequency is the same
with the grid frequency [8]. To that point, several kinds of angular frequency of PCC ω PCC . The configuration of this
damping terms can be added to (4) to program the damping type of VSG is shown in Fig. 3. However this type of VSG is
character into the control of the inverter. not suitable for the weak power grid.
When the damping term needs to be added to the swing dω 1
J = Tref − Te − D (ω − ω g ) + (ω − ω ) (6)
equation, a damping coefficient D must be chosen. dt ωm 0
Theoretically, if D is multiplied by the speed difference Δω ,
the result is called electrical damping. This damping is ωg ug
associated with the damper windings [9]. If D is multiplied by D PLL
−
the angular speed of the VSG voltage in electrical radians per
ω − Δω 1 −
second, the result is called mechanical damping. This damping Pm 1 Tm + 1 1 θ
is associated with the mechanical-friction [8]. However, this m + ω − Js s
+
damping do not cause much effect compared to the electrical Te
damping, and do not much considered. According to the ω
ω0 Pref
electrical damping term being used, the VSG models in
literature can be even more divided into two types.
Fig. 3. VSG type#2.
 III. COMPARISION OF VSG TYPE#1 AND TYPE #2
P1 , Q1 PCC
For the purpose of comparison, we study the dynamic Ls1 U ∠δ
behaviors of VSG type#1 and type#2 considering two 1 1
R1 , L1
identical VSGs (same capacity and parameters) connected in C s1
parallel as shown in Fig. 4. VSG 1
Matlab simulations are conducted in order to make the P2 , Q2
comparison evident by waveforms. The parameters for Ls 2 U ∠δ R2 , L2
simulation are given in Table I. two cases are considered. 2 2 Load
Cs 2
Case 1: the two VSGs are of type#1 and the results show VSG 2 U g ∠δ g
in Fig. 5.
Case 2: the two VSGs are of type#2 and the results show Fig. 4. Two paralleled VSG with a common load.
in Fig.6. state.
The dynamic behaviors are compared under the parallel It is well known that, if the frequency droop coefficients
operation because it is recognized that the change in the are inversely proportional to the capacity of the VSGs, the
parameters D, J or m could influence both the dynamic load power will be shared by the same capacity ratio.
response of specific VSG and other VSGs in parallel with it. However, this fact has not been confirmed in Fig. 5(b). The
given active powers of the two VSGs in Fig. 6(b) are
In Fig. 5(a), the difference of the dynamic response after
dramatically in the ratio of 2:1.
running the simulation between VSG1 and VSG2 is caused
due to the difference of their damping coefficients. The value When comparing Fig. 6(a) and (b), it is recognized that the
of frequency droop coefficients and virtual inertias are the damping coefficient and frequency droop coefficient have
same. This result emphasizes the fact that the effect of the totally different effects. In addition, the meaning of the
frequency droop is equivalent to the damping effect. However, damping term and frequency droop is very understandable
it can be seen from Fig. 6(a) that, the power waveforms tended from the figures.
to be different just at the beginning, and converge at the steady
Power (W)

(a) (b)

317

316.5 VSG1

316
VSG2

315.5

315

314.5

314
0 0.5 1 1.5
Time (s)
(c) (d)
Fig.5. power and frequency responces of case 1 (a) and (c) when the ratio D1: D2=15:30, (b) and (d) when m1: m2=50:100.
 10000 12000
VSG1
10000
8000

8000
VSG1
Power (W)

6000
6000
VSG2

4000
4000

PVSG1 =P VSG2 VSG2

2000
2000 P VSG1 =2P VSG2
0

0 -2000
0 0.5 1 1.5 0 0.5 1 1.5
Time (s) Time (s)
(a) (b)

(c) (d)
Fig.6. power and frequency responces of case 2 (a) and (c) when the ratio D1: D2=15:30, (b) and (d) when m1: m2=50:100.

As can be observed from Fig. 5(c) and Fig. 6(c), in the

former, the VSG frequency follows ω0 with accuracy of 99.6, IV. SUGGESTED DYNAMIC RESPONSE CONTROL
but, in the later the VSG frequency just follow the grid After recognizing that, the absence of the symmetry in the
frequency ω g , which in this case is determined by the two damping coefficients could affect the dynamics, it was very
VSGs. The same thing can be observed when comparing Fig. useful to re-design the damping term such that the asymmetry
5(d) and Fig. 6(d). The above comparison is summarized in can be avoided.
Table II. As we know, in islanding mode ω g is decided by all
paralleled VSGs and defined as the angular frequency of PCC.
TABLE I. SIMULATION PARAMETERS For convenient one can write ω g of the model in Fig. 4 as

parameters value unit H1ωs1 + H 2 ωs 2

ω g = ωPCC = (7)
H1 + H 2
Pref / Rated power 15 kW
U / Ac voltage 220 V Where ωsi and H are the stator angular frequency of and
Ls / Filter inductor 10 mH inertia constant, respectively [11], [12]. When H1 and H2 are
Cs / Filter capacitor 200 μF equal then,
fs / Switching frequency 5 kHz
ω0 / Rated frequency ωs1 + ωs 2 2
=  ωsi 2 = ωsi
314.16 rad/s
ω g = ωPCC = (8)
J / Virtual inertia 0.1 kg.m2 2 i =1
R / Line resistance 0.1 Ω
L / Line inductance 3 mH Then, the swing equation considering the expression in (8)
is as expressed in (9). Note that this is still the exact version of
ZLoad / Load power 15/6 kW / kVar
the swing equation
 dω 2 of each VSG will include n parts.
J = Tm − Te − D (ω −  ωsi 2) (9)
dt i =1
dω n
J = Tm − Te − D ( (ωri − ωsi ) n) (12)
assuming the damping term for every VSG can be designed dt i =1
such that
The accessibility of the neighboring frequency depends on
dω 2 2 presence of communication links.
J = Tm − Te − D ( ωri 2 −  ωsi 2) (10)
dt i =1 i =1 It is worth noting that, by using this method, a strong
coupling between the virtual rotors is introduced. This
n coupling makes achieving the proper power sharing possible
Where  ωri n is called system frequency ωsys and it during the whole process. The next section presents the results
i =1
to verify the new method. At the end of the section, a comparison
replaces ω . Then, after rewrite (10) has been made.
dω ω − ωs1 ωr 2 − ωs 2
J = Tm − Te − D ( r1 + ) (11) ωg ωr1
dt 2 2
− ω
Dj ΔωΣ +  ri
As can be seen from (11), the damping torque of the VSG N 
consists of 2 parts; the first part represents the torque produced +  
by the damper as a result of the rotor deviation for specific ωrn
Tm + − 1 1 ωrj 1 θ
VSG. The second part describes the damping torque which
can be produced by the damping winding in proportional to − J s s
the deviation in the rotor of the neighboring VSG (mutual
effect). This method is shown in Fig. 7. Te
Similarly, for n number of VSGs in parallel, the damping term
Fig. 7. Proposed dynamic response control.

10000

8000

6000
Power (W)

4000
P VSG1 =P VSG2

2000
VSG1

0 VSG2

0 0.5 1 1.5
Time (s)
(a) (b)

(c) (d)
Fig.8. power and frequency responces of suggested control method (a) and (c) when the ratio D1: D2=15:30, (b) and (d) when m1: m2=50:100.
 V. SIMULATION [2] X. Hou, H. Han, C. Zhong, W. Yuan, M. Yi, and Y. Chen,
“Improvement of transient stability in inverter-based AC microgrid via
In this section, the same scenarios are performed for the adaptive virtual inertia,” in Energy Conversion Congress and Exposition
two VSGs after applying the control method described in IV. (ECCE), Sep 2016, pp. 1-6.
[3] Q.C. Zhong, and G. Weiss, “Synchronverters: Inverters that mimic
By this method, smooth responses are obtained and the synchronous generators,” IEEE Transactions on Industrial Electronics,
output active power is seems to be independent of the vol.58, no.4, Apr 2011, pp. 1259-67.
damping coefficients inconsistency as shown in Fig. 8 (a). [4] H.P. Beck, and R. Hesse, “Virtual synchronous machine,” in Electrical
Fig. 8(c) shows the frequency dynamics. It is noticed that, the Power Quality and Utilisation (EPQU), Oct 2007, pp. 1-6.
accuracy of the VSG frequency is improved from 99.6 by [5] T.V. Van, K. Visscher, J. Diaz, V. Karapanos, A. Woyte, et al., “Virtual
VSG type#1 to 99.98 by this method. From Fig. 8(b) and (c), synchronous generator: An element of future grids,” In Innovative Smart
it can be seen that the active power and frequency take a Grid Technologies Conference Europe, Oct 2010, pp. 1-7.
longer time than usual to reach their steady-state values. [6] M.P.N. van Wesenbeeck, M.P.N. xde Haan, P. Varela, and K. Visscher,
“Grid tied converter with virtual kinetic storage,” in PowerTech, IEEE
Nevertheless, the power sharing is accurately obtained at the Bucharest, 2009, p. 1–7.
end. [7] K. Sakimoto, Y. Miura, and T. Ise, “Stabilization of a power system
with a distributed generator by a virtual synchronous generator
TABLE II. COMPARISION OF DAMPING METHODS FOR VSG function,” in Power Electronics and ECCE Asia (ICPE & ECCE), IEEE
8th International Conference on, May 2011, pp. 1498-1505.
[8] X. Meng, Z. Liu, J. Liu, T. Wu, S. Wang, B. Liu, “Comparison between
Damping term merit demerit virtual synchronous generator and droop controlled inverter,” in Power
Electronics Conference (SPEC), Dec 2016, pp. 1-6.
1. PLL free model. [9] S. D'Arco, and J.A. Suul, “Virtual synchronous machines-Classification
2. Self-droop characteristics if Similar behavior by of implementations and analysis of equivalence to droop controllers for
D(ω − ω0 ) the governor is not used. the damping and microgrids,” in PowerTech (POWERTECH), , Jun 2013, pp. 1-7.
3. ω follows the reference ω0 droop. [10] H. Bevrani, T. Ise, and Y. Miura, “Virtual synchronous generators: A
survey and new perspectives,” in ternational Journal of Electrical Power
1. PLL is & Energy Systems, vol.54, Jan 2014, pp. 244-54.
necessitating. [11] P.M. Anderson, and A.A. Fouad, Power system control and stability.
Clear and separate damping and 2. ω follows ω g , John Wiley & Sons, Sep 2008.
D(ω − ωg ) droop features.
which is determined [12] K.R. Padiyar, Power system dynamics. BS publications, 2008.
by the load in [13] B. Zhang, X. Yan, S.Y. Altahir, “Control design and small-signal
islanded mode. modeling of multi-parallel virtual synchronous generators,” in
1. Separate damping and droop. Compatibility, Power Electronics and Power Engineering (CPE-
Use of PLL and POWERENG), Apr 2017, pp. 471-476.
D(ωsys − ωg ) 2. Both ω and ω g are controlled
communication [14] Q.C. Zhong, and T. Hornik, Control of power inverters in renewable
very well. links energy and smart grid integration. John Wiley & Sons, vol.97, Nov
3. Robust dynamics. 2012.
[15] H. Wu, X. Ruan, D. Yang, X. Chen, W. Zhao, Z. Lv, et al., “Small-
VI. CONCLUTION signal modeling and parameters design for virtual synchronous
generators,” IEEE Transactions on Industrial Electronics, vol.63, no.7,
This paper introduces a comprehensive comparison of the Jul 2016, pp. 4292-303.
dynamic performance of two types of VSGs and provides the [16] D. Li, Q. Zhu, S. Lin, and X.Y. Bian, “A Self-Adaptive Inertia and
possible results of the different types. The dynamic behaviors Damping Combination Control of VSG to Support Frequency Stability,”
are compared under the parallel operation in an islanded IEEE Transactions on Energy Conversion, vol.32, no.1, Mar 2017, pp.
397-8.
microgrid. Also, a VSG model that can be applied in the
[17] L. Bin, Z. Lin, Y. Xirui, Z. Chen, L. Jinhong, and X. Bao, “New control
parallel system of VSGs to get the best dynamics is introduced scheme of power decoupling based on virtual synchronous generator,” in
and the effectiveness of the method is verified through Power and Energy Conference at Illinois (PECI), Feb 2016, pp. 1-8.
simulation by using MATLAB. [18] J. Liu, Y. Miura, and T. Ise, “Comparison of dynamic characteristics
between virtual synchronous generator and droop control in inverter-
based distributed generators,” IEEE Transactions on Power Electronics,
ACKNOWLEDGMENT vol.31, no.5, May 2016, pp. 3600-11.
This paper was supported by the National High-tech R&D [19] M. Li, Y. Wang, N. Xu, Y. Liu, W. Wang, et al., “A novel virtual
Program (863 Program) of China (2015AA050603), and synchronous generator control strategy based on improved swing
Science and Technology Program of State Grid (PD71-17- equation emulating and power decoupling method,” in Energy
Conversion Congress and Exposition (ECCE), Sep 2016, pp. 1-7.
008).
[20] Q.C. Zhong, and G. Weiss, “Static synchronous generators for
distributed generation and renewable energy,” in Power Systems
REFERENCES Conference and Exposition, PSCE'09, IEEE/PES, Mar 2009, pp. 1-6.
[1] T.L. Vandoorn, B. Meersman, J.D. De Kooning, and L. Vandevelde, [21] T. Shintai, Y. Miura, and T. Ise, “Oscillation damping of a distributed
“Directly-coupled synchronous generators with converter behavior in generator using a virtual synchronous generator,” IEEE transactions on
islanded microgrids,” IEEE Transactions on Power Systems, vol.27, power delivery, vol.29, no.2, Apr 2014, pp. 668-76.
no.3, Aug 2012, pp. 1395-406. [22] J. Meng, X. Shi, Y. Wang, and C. Fu, “A virtual synchronous generator
control strategy for distributed generation,” in Electricity Distribution
(CICED), Sep 2014, pp.495-498.