Aalborg Universitet

A Comparative Study of Grid-Following and Grid-Forming Control Schemes in Power

Electronic-Based Power Systems

Gao, Xian; Zhou, Dao; Anvari-Moghaddam, Amjad; Blaabjerg, Frede

Published in:
Power Electronics and Drives

DOI (link to publication from Publisher):

10.2478/pead-2023-0001

Creative Commons License

CC BY-NC-ND 4.0

Publication date:
2023

Document Version
Publisher's PDF, also known as Version of record

Link to publication from Aalborg University

Citation for published version (APA):

Gao, X., Zhou, D., Anvari-Moghaddam, A., & Blaabjerg, F. (2023). A Comparative Study of Grid-Following and
Grid-Forming Control Schemes in Power Electronic-Based Power Systems. Power Electronics and Drives, 8(1),
1-20. https://doi.org/10.2478/pead-2023-0001

General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners
and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

- Users may download and print one copy of any publication from the public portal for the purpose of private study or research.
- You may not further distribute the material or use it for any profit-making activity or commercial gain
- You may freely distribute the URL identifying the publication in the public portal -
Take down policy
If you believe that this document breaches copyright please contact us at [email protected] providing details, and we will remove access to
the work immediately and investigate your claim.
Power Electronics and Drives
Volume 8(43), 2023   DOI: 10.2478/pead-2023-0001

Power Electronics and Drives

A Comparative Study of Grid-Following and

Grid-Forming Control Schemes in
Power Electronic-Based Power Systems

Research paper

Xian Gao* , Dao Zhou , Amjad Anvari-Moghaddam , Frede Blaabjerg

AAU Energy, Aalborg University, Aalborg, Denmark

Received: 31, October 2022; Accepted: 30, November 2022

Abstract: A
 long with de-carbonisation, the penetration of power electronic converters has increased, and the power system has become a
power electronic-based power system. In such a situation, the stability of the power system faces great challenges. In the event of
a large disturbance, the power grid will lack the ability to maintain a stable voltage and frequency. In order to improve the stability of
the power grid, the traditional grid-following (GFL) control is needed to be converted to the grid-forming (GFM) control. This paper
analyses the control schemes of the GFL and GFM converters by investigating their state-space models, and the eigenvalue trajectories
of both control schemes are shown to analyse the stability of the systems. Moreover, a case study is exemplified to compare the
performance of the two control strategies while responding to frequency disturbances. Finally, a time-domain simulation model of a 15
kW grid-connected converter is built in Matlab/Simulink to benchmark the performance of the GFL and GFM converters under different
working conditions. The result reveals that the GFL converter may encounter some instabilities when applied in power electronic-based
systems, while the GFM converter is more suitable for the weak power grid.

Keywords: grid-following control • grid-forming control • state-space model • stability analysis • grid strength

1. Introduction
Global energy demand has been increasing over the past few decades. As a promising candidate, renewable
energy is developing rapidly to cope with potential energy crises. Renewable energy systems will play an even more
important role in future electricity production.
However, renewable energy resources are coupled to the power grid through power electronic converters,
which respond quickly and do not provide any moment of inertia (Fang et al., 2018; Peng et al., 2019). Today’s
power grids rely on synchronous generators to generate mechanical inertia. These generators are very large
and are synchronised with each other tightly, so small disturbances from the load or the generation process
cannot influence the frequency stability of the power grid. However, distributed renewable energy systems are
usually connected to the power grid using grid-following (GFL) converters instead of conventional synchronous
generators. The goal of the GFL converter is to simply lock and track the grid frequency. GFL converters typically
operate at their rated output power and do not respond to deviations in grid frequency as do synchronous
generators (Lasseter et al., 2020). With the increasing popularity of renewable energy systems, many large central
power plants are being phased out. Ultimately, the moment of inertia and damping of the entire grid to which the
renewable energy system is connected are decreasing, resulting in a weaker ability of the grid to handle sudden
deviations of the frequency. The characteristics of a low-inertia system have a great influence on the stability of
the power grid. In order to improve the stability of a distributed power grid with highly penetrated renewable energy
systems, the grid-forming (GFM) control strategy has emerged (Rocabert et al., 2012). One of the most popular

* Email: [email protected]
1

Open Access. © 2022 Gao et al., published by Sciendo. This work is licensed under the Creative
Commons Attribution NonCommercial-NoDerivatives 4.0 License.
 Comparison of grid-following and grid-forming converters

GFM control strategies is the application of virtual synchronous generators (VSGs), which enables the converter
to imitate the behaviour of conventional synchronous generators to improve the inertia and damping of the power
grid (Chen et al., 2020).
This paper describes the above problems and discusses the performance of the GFL and GFM control systems
under various working conditions. The main contributions of this paper can be summarised as follows. (1) Detailed
illustrations of the typical structures of the GFL and GFM controls are given, and the state-space models of both
the GFL and GFM controls are built. (2) A comparison between the GFL and GFM controls under various working
conditions is done. (3) The simulation results show that the smaller the short circuit ratio (SCR) becomes, the
greater is the impact that the load fluctuation has on the power grid and that the GFM control is more suitable for a
weak power grid.
The remaining part of the paper is organised as follows. Section 2 describes the conceptual differences between
the GFL and the GFM controls. In Sections 3 and 4, the control diagrams of the GFL and GFM controls are illustrated
in detail and the state-space models of both the GFL and GFM controls are built. In Section 5, a simulation model of
a grid-connected converter is built in Matlab/Simulink to compare the different performances of the GFL and GFM
controls under various working conditions to verify the expected outcomes. Finally, some conclusions are drawn in
Section 6.

2. Comparison Between the GFL Converter and the GFM Converter

Power converters can be divided into GFL converters and GFM converters according to their control schemes. The
GFL converter can be simply represented as a controlled current source, as shown in Figure 1a, where Zg is the
equivalent grid impedance; Vg ∠ δg is the grid voltage vector; Ig ∠ φg is the converter output currents. It usually uses
a phase-locked loop (PLL) to track the phase angle of the voltages at the point of common coupling (PCC) to ensure
that the converter is synchronised to the power grid (Wang et al., 2020). The GFL converter achieves the target
injected active and reactive power by adjusting the currents injected into the power grid. However, it cannot provide
regulation of the frequency and voltage for the grid directly, which acquires the frequency and voltage references
from an additional voltage source or the power grid (Du et al., 2020). Therefore, the GFL converter cannot operate
in an islanded mode and cannot handle the problems of frequency disturbances.
Instead, the GFM converter can be simply represented as a controlled voltage source, as shown in Figure 1b,
where E ∠ δ is the inverter ouput voltage vector. Some GFM control strategies do not require a PLL to track the
phase angle of the voltages at the PCC and can imitate the conventional synchronous generators to achieve self-
synchronisation with the power grid (Rosso et al., 2021). Compared to the conventional synchronous generators that
provide frequency regulation by the stored rotating energy in the rotor, GFM converters can adjust their output faster
to deal with frequency disturbances in the power grid (Pattabiraman et al., 2018). Therefore, the GFM converters
are suitable for operation in an islanded mode.
In addition, some works in the literature demonstrate that the use of a PLL may affect the stable operation of the
power converters in weak grids (Lasseter et al., 2020; Rosso et al., 2021). That is because the GFL converter tracks
the voltages at the PCC, which is easily affected by the output currents in weak grids. On the contrary, the GFM

Vg ∠ δ g Vg∠ δg
Grid Grid
Zg Zg

I g∠ϕg E∠ δ

(a) (b)
Fig. 1. Simplified representation of power converters: (a) GFL converter; (b) GFM converter. GFL, grid-following; GFM, grid-forming.

2
Gao et al.

Control mode Advantages Disadvantages

GFL converter 1. Quick regulation 1. Lack of frequency and voltage regulation

2. Simple control structure 2. Unable to operate in an islanded mode
3. Instability in weak grids
GFM converter 1. Able to operate in an islanded mode 1. Instability in stiff grids
2. Provides the regulation of frequency and voltage 2. Easily susceptible to overload

GFL, grid-following; GFM, grid-forming.

Table 1. Comparison of GFL and GFM converters.

converter can achieve self-synchronisation based on the output active power, which allows it to synchronise to weak
grids. However, in a stiff grid, a little phase difference between the converter voltages and grid voltages may lead to
large fluctuations of active power, resulting in overload. Table 1 summarises the advantages and disadvantages of
the GFL and GFM converters (Gao et al., 2021).

3. Control Structure of a GFL Converter

The GFL converter is widely used in distributed renewable energy systems. When grid-connected power converters
are used as the interface between the renewable energy systems and power grids, most grid-connected converters
with GFL control consist of a PLL unit and double-loop vector control. The GFL converter uses a PLL to track the
phase angle of the voltages at the PCC and adopts a vector control strategy to adjust the active and reactive
currents that are injected into the power grid. This paper adopts active - reactive power control (PQ control) as an
outer control loop. For this control strategy, the outer loop is used to adjust the active and reactive power injected
into the power grid, while the inner current is used to adjust the converter currents according to the reference values
set by the outer power loop (Sangwongwanich et al., 2018).
The system topology of a GFL converter with the PQ control is shown in Figure 2, where udc is the DC-link
voltage; ua, ub and uc are the converter output voltages, respectively; ia, ib and ic are the converter currents;
upcca, upccb and upccc are the voltages at the PCC; uga, ugb and ugc are the grid voltages; iga, igb and igc are the grid
currents; iCa, iCb and iCc are the capacitor currents; Lf and Cf are the inductance and capacitance, respectively,
of the LC filter; Lg and Rg are the grid impedance and resistance, respectively; Pe and Qe are the output
active power and reactive power, respectively; Pref and Qref are the references of active and reactive power,
respectively.

Inverter PCC Power Grid

Lf upcca Rg Lg
ua ia iga uga
udc ub ib upccb igb ugb
uc ic upccc igc ugc
iCa iCb iCc
igabc
upcc
SPWM Cf abc θ PLL load
dq
upcc
abc θ
dq id iq igabc
θ abc iCd iCq Power
dq calculation
KC KC
urq urd Pe Qe
- + - + idref - + Pref
PI PI
- + - + iqref - + Qref
PI PI

Fig. 2. The topology of a GFL converter with the PQ control. GFL, grid-following; PCC, point of common coupling; PI, proportional integral; PLL,
phase-locked loop; SPWM, sinusoidal pulse width modulation.

3
 Comparison of grid-following and grid-forming converters

3.1. The PLL unit

The PLL unit enables the converter to synchronise to the power grid. Because of the existence of the PLL, when
small-signal perturbations are added, there is a small error qpll – qs between the control synchronising frame (defined
by the PLL) (Dong et al., 2015; Yang et al., 2019) and the actual system synchronising frame (defined by the PCC
voltage). The actual system synchronising frame can be converted to the control synchronising frame using a
rotation matrix Tθ:

 cos(q pll − q s ) sin(q pll − q s )   1 q pll − q s 

Tq =  ≈ . (1)
 − sin(q pll − q s ) cos( q pll − q s )   q s − q pll 1 

The transformations between the two reference frames can be expressed as follows (Xie et al., 2019):

 xdc   xds   1 q pll − q s   xds 

  = TDq   ≈   s  ; (2)
c
 xq   xqs   q s − q pll 1   xq 

 xds   xdc   1 q s − q pll   xdc 

  = TDq −1   ≈   c  , (3)
 xqs   xqc   q pll − q s 1   xq 
where the subscripts d and q represent the d-axis and q-axis components of a variable; the variables in the actual
system synchronising frame are marked with the superscript s, while the variables in the control synchronising
frame are marked with the superscript c.
Adding small-signal perturbations to Eqs (2) and (3), they can be rewritten as follows:

 X dc 0 + Dxdc   1 q pll + Dq − q s   X s + Dx s 
 =   d0 d
;
 (
 X qc0 + Dxqc   q s − q pll + Dq ) 1   X qs0 + Dxqs 
 
(4)

 X ds 0 + Dxds  
 =
1 ( 
)
q s − q pll + Dq   X dc 0 + Dxdc 
, (5)
 X qs0 + Dxqs   q pll + Dq − q s 1   X qc0 + Dxqc 
  
where the subscript ‘0’ denotes the steady-state values, and the prefix ∆ denotes the small-signal disturbance of
variables.
Because the control synchronising frame is aligned with the actual system synchronising frame in the steady
state, it means that the angle error qpll – qs is zero (Wen et al., 2016). Deleting the steady-state values, the linearised
small-signal model of transformations can be given as follows:

 Dxdc   Dxds   X q 0 
 = +  ⋅ Dq ; (6)
 Dxqc   Dxqs   − X d 0 

 Dxds   Dxdc   X q 0 
 = −  ⋅ Dq . (7)
 Dxqs   Dxqc   − X d 0 

The control scheme of the PLL is shown in Figure 3, where TPLL is the proportional integral (PI) controller of the
c
PLL; s is the Laplace variable. According to Figure 3, the relationship between upccq and q can be expressed as
follows:

c
TPLL ( s ) c  1  u pccq ,
q PLL = u pccq =  k pPLL + kiPLL  (8)
s  s s

where kpPLL and kiPLL are the proportional and integral parameters, respectively, for the PLL unit.

4
Gao et al.

u pcca
abc θPLL
u pccb u cpccq
TPLL ω 1
u pccc
dq
+ s
0

Fig. 3. Control scheme of the PLL. PLL, phase-locked loop.

The linearised small-signal model of the PLL unit can be derived as follows:

c
TPLL ( s ) c  1  Du pccq .
Dq = Du pccq =  k pPLL + kiPLL  (9)
s  s s

According to Eqs (6) and (9), Dq can be given as

Dq =
s
(
TPLL ( s )
Du spccq − U pccd 0Dq . ) (10)

In the time domain, Eq. (10) can be written as follows:

d Dxq
= Du cpccq = Du spccq − U pccd 0 Dq ; (11)
dt

d Dq
dt
(
= kiPLL Dxq +k pPLL Du spccq − U pccd 0Dq . ) (12)

Therefore, the state-space model of the PLL unit can be expressed as follows:

D xPLL = APLL DxPLL + BPLL DxLCL ,


(13)

where

T
DxPLL =  Dxq Dq  ; (14)

T
s  ;
DxLCL =  Dids Diqs Du spccd Du spccq s
Digd Digq (15)
 

 0 −U pccd 0
APLL =  ; (16)
 kiPLL − k pPLLU pccd 0 

0 0 0 1 0 0
BPLL =  ;
 0 0 0 k pPLL 0 0  (17)

3.2. Power control loop

The outer power control loop is used to regulate the active and reactive power injected into the power grid. In
the synchronous d-q reference frame, the instantaneous active and reactive power are decoupled and can be
expressed as follows:

 3 c
( c c
 p = 2 u pccd igd + u pccqigq
c

.
)
 (18)
 2
(
 q = 3 uc i c − uc i c
pccq gd pccd gq )
5
 Comparison of grid-following and grid-forming converters

The average active and reactive power can be given as follows:

 wc
 Pe = s + w p
 c , (19)

Q = w c
q
 e s + wc

where wc is the cut-off frequency of the low-pass filter (LPF) of the measured active and reactive power.
According to Eqs (18) and (19), the linearised small-signal equations can be expressed as


D xcal = Acal Dxcal + Bcal DxLCL , (20)

where

T
Dxcal =  DPe DQe  ; (21)

 − wc 0 
Acal =  ; (22)
 0 − wc 

 3 3 3 3 
0 0 2 wc I gd 0 2
wc I gq 0
2
wcU pccd 0
2
wcU pccq 0 
Bcal = . (23)
0 0 − 3 w I 3
wc I gd 0
3
wcU pccq 0
3
− wcU pccd 0 
 2
c gq 0
2 2 2 

The outputs of the power loop serve as the references for the inner current loop. The outputs of the power loop
can be given as follows:

idref

c
( )(
= k pPQ + kiPQ / s Pref − Pe
,
)
c (24)
i
 qref = − k (
pPQ + kiPQ / s Q )(
ref − Qe )
where kpPQ and kiPQ represent the proportional and integral coefficients, respectively, of the power control loop.
Two state-space variables gd and gq are introduced as shown below:

 d gd
 dt = Pref − Pe
 . (25)
 d gd = Q − Q
ref e
 dt

Combining Eqs (24) and (25), the linearised small-signal model can be given as follows:


D xPQ = BPQ1Dxcal ; (26)

DyPQ = CPQ DxPQ + DPQ Dxcal , (27)

where

T
DxPQ =  Dg d Dg q  ; (28)

T
c
DyPQ =  Didref c  ;
Diqref (29)
 
6
Gao et al.

 −1 0 
BPQ1 =  ; (30)
 0 −1

 kiPQ 0 
CPQ =  ; (31)
 0 − kiPQ 

 − k pPQ 0 
DPQ =  . (32)
 0 k pPQ 

3.3. Current loop

The inner current control loop is used to regulate the converter currents to follow the references set by the outer
power loop and also to protect the power conversion. It is worthwhile to note that a virtual resistance-based active
damping method is applied here to reduce the resonant peak without sacrificing the efficiency of the converter
(Adapa and John, 2018; Dahono et al., 2001). The capacitor current is fed back into the current control loop. The
outputs of the current control loop can be given as follows:

u c
 rdref ( c
)(
= k pc + kic / s idref c
)
− igd c
− wg Lg igq ( )
− KC idc − igd
c

 , (33)
c
urqref
 ( c
)(
= k pc + kic / s iqref c
)
− igq c
+ wg Lg igd ( )
− KC iqc − igq
c

where wg is the grid frequency; kpc and kic represent the proportional and integral coefficients, respectively, of the
inner current control loop; KC is the proportionality coefficient of the capacitor current feedback.
Two state-space variables φd and φq are defined as follows:

 d jd c c
 dt = idref − igd
 . (34)
 d jq = i c − i c
 dt qref gq

Combining Eqs (5), (33) and (34), the linearised state-space model of the current loop can be given as follows:


D xc = Bc1DxPLL + Bc 2DxLCL + DyPQ ; (35)

Dycs = Cc Dxc + Dc1Dydc + Dc 2DxLCL + Dc 4DxPLL , (36)

where

T
Dxc =  Djd Djq  ; (37)

T
s
Dycs =  Durd s  ;
Durq (38)
 

0 −I gd 0 
Bc1 =  ; (39)
0 I gq0 
 

0 0 0 0 −1 0 
Bc 2 =  ; (40)
0 0 0 0 0 −1

7
 Comparison of grid-following and grid-forming converters

 kic 0
Cc =  ;
0 kic  (41)

 k pc 0 
Dc1 =  ; (42)
 0 k pc 

 − KC 0 0 0 − k pc + KC − wg Lg 
Dc 2 =  ; (43)
 0 − KC 0 0 wg Lg − k pc + KC 

0
Dc 4 = 
( KC − k pc ) I gq0 + wg Lg I gd 0 − KC Iq0 − U rq0  . (44)
0
 (k pc − KC ) I gd 0 + wg Lg I gq0 + KC Id 0 + U rd 0 
Because of the sampling delay caused by digital control and the transmission delay caused by the digital pulse
width modulation (PWM) unit (Wen et al., 2016), the delay unit needs to be added to the model, which can be given
as follows:

  − wd 0   wd 0 
D xdel =  Dx + Dy = Adel Dxdel + Bdel Dycs ;
 0 − wd  del  0 wd  cs (45)

1 0   Dud 
s
Dydel =     = Dxdel , (46)
s
0 1   Duq 

where wd = 1/Td, and Td is the total delay time of the system.

3.4. Electrical system model (power grid)

According to Kirchhoff’s voltage law, the state-space model of the electrical system shown in Figure 2 can be
expressed as shown in Eq. (47) (Wu et al., 2020):

  Du gd
s 
D xLCL = ALCL DxLCL + BLCL1Dydel + BLCL 2   + BLCL3Dw , (47)
 Du gq
s 
 

where

 0 w −1 / L f 0 0 0 
 
 −w 0 0 −1 / L f 0 0 
1 / C 0 0 w −1 / C f 0 
=  
f (48)
ALCL ;
0 1/ Cf −w 0 0 −1 / C f 
 
 0 0 1 / Lg 0 − Rg / Lg w 
 
 0 0 0 1 / Lg −w − Rg / Lg 

T
1 / L f 0 0 0 0 0
BLCL1 =   ; (49)
 0 1/ Lf 0 0 0 0 

8
Gao et al.

T
0 0 0 0 −1 / Lg 0 
BLCL 2 =  ; (50)
0 0 0 0 0 −1 / Lg 

T
BLCL3 =  I q 0 − I d 0 U pccq 0 −U pccd 0 I gq 0 − I gd 0  . (51)

3.5. The state-space model of a GFL converter

Based on the above analysis, the state-space model of the GFL converter can be expressed as follows:

  Du gd
s 
D x sys _ GFL = Asys _ GFL Dxsys _ GFL + B1   + B2 Dw , (52)
 Du gq
s 
 

where Dxsys _ GFL is a 16 × 1 matrix, Asys _ GFL is a 16 × 16 matrix, B1 is a 16 × 2 matrix, and B2 is a 16 × 1 matrix.

T
Dxsys _ GFL =  DxPLL Dxcal DxPQ Dxc Dxdel DxLCL  ; (53)

 APLL [0]2× 2 [0]2× 2 [0]2× 2 [0]2× 2 BPLL 

 
 [0]2 × 2 Acal [0]2× 2 [0]2× 2 [0]2× 2 Bcal 
 
 [0] BPQ1 [0]2× 2 [0]2× 2 [0]2× 2 [0]2×6 ; (54)
Asys _ GFL =  2× 2 

Bc1 DPQ CPQ [0]2× 2 [0]2× 2 Bc 2 
 Bdel Dc 4 Bdel Dc1DPQ Bdel Dc1CPQ Bdel Cc Adel Bdel Dc 2 
 
 [0]6 × 2 [0]6× 2 [0]6× 2 [0]6× 2 BLCL1 ALCL 

[0] 
B1 =  10×2  ; (55)
 BLCL 2 

[0] 
B2 =  10 ×1  . (56)
 BLCL3 

The SCR is usually defined to measure the strength of the power grid. According to the Institute of Electrical
and Electronics Engineers (IEEE) Standard 1204-1997 (IEEE Std 1204-1997, 1997), when the SCR <2, the power
grid is considered very weak. Alternatively, when the SCR >3, the power grid is considered strong. The definition of
SCR can be expressed as follows:

2
S SC 1.5U g ,
SCR = = (57)
SN SN Z g

where Ssc is the short-circuit apparent power of the power grid; SN is the rated apparent power of the inverter; Ug is
the amplitude of the grid voltage; and Zg is the amplitude of grid impedance.
In this paper, the SCR is changed by varying the values of the grid impedance Zg. Long transmission lines will
give large grid impedances and lead to a low SCR.
A simulation model of a 15-kW grid-connected converter is built in Matlab/Simulink and the key parameters of
the case study system are listed in Table 2 (Gao et al., 2022).
According to the parameters listed in Table 2, the eigenvalue trajectory of the GFL converter can be plotted under
various SCR values. When the SCR decreases from 7.5 to 1.0, the eigenvalue trajectory of the GFL converter is
plotted in Figure 4. When the SCR decreases, the eigenvalues l1 and l2 move towards the right half plane, which
means that the system loses stability. When the grid impedance is 20 mH and the corresponding SCR is 1.5, the
system is in critical stability. When the grid impedance is >20 mH and the SCR is <1.5, the system becomes unstable.
9
 Comparison of grid-following and grid-forming converters

Parameter Description Values

Grid
Ug Grid voltage 220 V
fg Grid frequency 50 Hz
Lg Grid impedance 4–30 mH
Rg Grid resistance 0.2–1.5 W
Converter
udc DC-link voltage 700 V
Lf Impedance of LC filter 3 mH
Cf Capacitance of LC filter 20 mF
Pref Reference of active power 15 kW
Qref Reference of reactive power 0 kVar
fs Switching frequency 20 kHz
fsa Sampling frequency 20 kHz
Control parameters for the GFL converter
wpll Bandwidth of PLL 176 rad/s
wPQ Bandwidth of power loop 78 rad/s
wi Bandwidth of current loop 1,683 rad/s
KC P controller parameter of capacitor current feedback 15
wc Cut-off frequency of low-pass filter 100 rad/s
Control parameters for the GFM converter
wvsg Bandwidth of VSG loop 14 rad/s
wu Bandwidth of voltage loop 200 rad/s
wi Bandwidth of current loop 1,683 rad/s
KC P controller parameter of capacitor current feedback 15
wc Cut-off frequency of low-pass filter 100 rad/s
D Damping coefficient 25
J Virtual inertia 0.2 kg/m2
ku Q-U loop coefficient 200
kq Integrity coefficient 0.1

GFL, grid-following; PLL, phase-locked loop; VSG, virtual synchronous generator.

Table 2. Parameters of a 15-kW grid-connected converter.

4. Control Structure of a GFM Converter

With the continuously increasing popularity of distributed renewable energy systems, the moment of inertia and
damping of the entire power grid are decreasing. The stability of the power grid is at risk. Furthermore, the power
of renewable energy systems injected into the distribution grid is dependent on the environmental conditions, which
aggravates the instability of the power grid. In recent years, VSG control technology, which is one of the GFM control
strategies, has attracted increasing attention because of its ability to imitate the inertia and damping characteristics
of synchronous generators. Therefore, the power grid will have a strong ability to cope with the power fluctuations
from the renewable energy systems and the load. Since this paper mainly focusses on grid-connected converters,
to simplify the analysis, the DC side is just represented as an ideal DC source. The topology of a GFM converter
with VSG control is shown in Figure 5, where Em and q represent the amplitude and phase angle of reference
voltage, respectively.

4.1. VSG algorithm loop

The VSG algorithm is mainly composed of a power-frequency controller and an excitation controller. The power-
frequency controller enables the converters to imitate the characteristics of synchronous generators. The excitation

10
Gao et al.

100 1
0
-100
-5850 -5650

6 100 λ1
Imaginary Axis (×103 )

Initial Eigenvalues
3 Final Eigenvalues

0 -200 λ2
3
-3 0
-3
-16 -8 0
-6 -500
0 -5 -2.5 0 -200 -100 0
Real Axis (×103 )

Fig. 4. Eigenvalue trajectory of the GFL converter with SCR changing from 7.5 to 1.0 and using the parameters in Table 2. GFL, grid-following; SCR,
short-circuit ratio.

Inverter PCC Power Grid

Lf upcca Rg Lg
ua ia iga uga
udc ub ib upccb igb ugb
uc ic upccc igc ugc
iCa iCb iCc
upcc igabc
Cf Power
SPWM calculation load

Pe Qe
igabc Pref
VSG
θ abc abc θ algorithm
dq dq Qref
iCd iCq θ abc Em θ
dq
abc
K C KC id iq dq
urq urd - upccdref -
- + + idref + upccd upcc
PI PI
dq
- + - + iqref upccdref - upccq abc θ
PI PI

Fig. 5. The topology of a GFM converter with the VSG control. GFM, grid-forming; PCC, point of common coupling; PI, proportional integral; SPWM,
sinusoidal pulse width modulation; VSG, virtual synchronous generators.

controller achieves the droop control between the voltage and the reactive power in the steady state. The two
controllers can be given as follows (Peng et al., 2020; Zhong and Weiss, 2011):

 d w Pref Pe
 J
dt
=
w
− − D w − wg
w ;
( )
 (58)
 dq = w
 dt

1 dEm
kq dt
= Qref − Qe + ku U N − U pcc , ( ) (59)

11
 Comparison of grid-following and grid-forming converters

where J is the moment of inertia; D is the damping coefficient; UN is the rated voltage; Upcc is the amplitude of the
voltage at the PCC; kq is the integrity coefficient; ku is the voltage droop coefficient.
According to Eqs (58) and (59), the linearised small-signal model can be obtained and the state-space model of
the VSG algorithm loop can be given as follows:


D xVSG = AVSG DxVSG + BVSG1Dxcal + BVSG 2DxLCL , (60)

where

T
DxVSG =  Dw DEdref  ; (61)

− D / J 0
AVSG =  ; (62)
 0 0 

 −1 / J wg 0 
BVSG1 =  ; (63)
 0 − kq 

0 0 0 0 0 0
BVSG 2 =  . (64)
0 0 − kq ku 0 0 0 

4.2. Voltage control loop

The voltage control loop is used to regulate the voltages at the PCC to follow the references set by the VSG
algorithm loop. The outputs of the voltage control loop are given as follows:

 dref (
i c = k + k / s E
pu iu )( c
) c
dref − u pccd − w g C f u pccq
,
 (65)
c
iqref
 ( )( )
= k pu + kiu / s 0 − u cpccq + wg C f u cpccd

where kpu and kiu represent the proportional and integral coefficients, respectively, of the voltage control loop.
In order to express the equations simply, two state-space variables md and mq are given as in Eq. (66):

 d md c
 dt = Edref − u pccq
 . (66)
 d mq = −u c
 dt pccq

Combining Eqs (65) and (66), the state-space model of the voltage loop can be expressed as
follows:


D xu = Bu1DxVSG + Bu 2 DxLCL ; (67)

Dyu = Cu Dxu + Du1DxVSG + Du 2 DxLCL , (68)

where

T
Dxu =  Dmd Dmq  ; (69)

T
c
Dyu =  Didref c  ;
Diqref (70)
 

12
Gao et al.

0 1 
Bu1 =  ; (71)
0 0 

0 0 −1 0 0 0 
Bu 2 =  ; (72)
0 0 0 −1 0 0 

 kiu 0
Cu =  ; (73)
0 kiu 

0 k pu 
Du1 =  ; (74)
0 0 

0 0 − k pu − wg C f 0 0
Du 2 =  . (75)
0 0 wg C f − k pu 0 0 

4.3. Current control loop

The current control loop of the GFM control is the same as that of the GFL control, so it will not be described here
again. The state-space model of the current control loop can be expressed as follows:


D xc = Bc 2DxLCL + Dyu ; (76)

Dyc = Cc Dxc + Dc1Dyu + Dc 2DxLCL ; (77)


D xdel = Adel Dxdel + Bdel Dyc . (78)

4.4. Electrical system model (power grid)

The electrical system model of the GFM converter is similar to that of the GFL converter, which is illustrated in
Section 3. However, the value of Dw is obtained from the VSG algorithm loop. The state-space model of the
electrical system can be given as in Eq. (79):

  Du gd
s 
D xLCL = ALCL DxLCL + BLCL1Dydel + BLCL 2   + BLCL 4DxVSG , (79)
 Du gq
s 
 

where BLCL4 is a 16 × 2 matrix:

BLCL 4 =  BLCL3 [0]6×1  . (80)

4.5. The state-space model of a GFM converter

Based on the above analysis, the state-space model of the GFM converter can be expressed as follows:

  Du gd
s 
D x sys _ GFM = Asys _ GFM Dxsys _ GFM + B1  , (81)
 Du gq
s 
 

where Dxsys _ GFM is a 16 × 1 matrix, Asys _ GFM is a 16 × 16 matrix.

13
 Comparison of grid-following and grid-forming converters

Initial Eigenvalues 300 ×10-4

6 7
Final Eigenvalues
0
Imaginary Axis (×103 )

3 150 -7
22 -13.21 -13.19
0.1
0 0 0
×10-5 0
1.5 -0.1
-3 0 -13 -10
-2 -22
-150 ×10-4
-249.985 -249.984 8
-6 0 0
-8
-12 -9 -6 -3 -300 -10.03 -10.01
-2000 0 2000
Real Axis (×104 )

Fig. 6. Eigenvalue trajectory of the GFM converter with SCR changing from 7.5 to 1.0 and using the parameters in Table 2. GFM, grid-forming;
SCR, short-circuit ratio.

T
Dxsys _ GFM =  Dxcal DxVSG Dxu Dxc Dxdel DxLCL  ; (82)

 Acal [0]2× 2 [0]2× 2 [0]2× 2 [0]2× 2 Bcal 

 
 BVSG1 AVSG [0]2× 2 [0]2× 2 [0]2× 2 BVSG 2 
 
[0] Bu1 [0]2× 2 [0]2× 2 [0]2× 2 Bu 2 .
Asys _ GFM =  2× 2  (83)
[ ]2 × 2 [0]2× 2 [0]2× 2
0 Du1 Cu Bc 2 + Du 2

[0] Bdel Dc1Du1 Bdel Dc1Cu Bdel Cc Adel Bdel Dc 2 + Bdel Dc1Du 2 
 2× 2 
[0]6 × 2 BLCL 4 [0]6× 2 [0]6× 2 BLCL1 ALCL 

According to the parameters shown in Table 2 in Section 3, the eigenvalue trajectory of the GFM converter can
be plotted. In Figure 6, the eigenvalue trajectory of the GFM converter is plotted when the grid impedance increases
from 4 mH to 30 mH and the corresponding SCR decreases from 7.5 to 1.0. It can be seen that when the SCR
decreases, all the eigenvalues are still in the left half plane, which means that the system stays stable.

5. Simulation Results: Case Study

In order to demonstrate the aforementioned control schemes, a simulation model of 15 kW grid-connected converter
is built in Matlab/Simulink. The SCR of the case study is 7.5. The parameters are shown in Table 2 in Section 3.
The case of loss of generation and load shed are chosen to compare the different performances of the GFL and
GFM controls. In order to discuss the different dynamic responses of the GFL and GFM controls under operation at
different SCRs, a simulation is carried out in the case of load increase.
According to the IEEE 1547-2018 standard (IEEE Std 1547-2018, 2018), the rated frequency is 60.0 Hz and the
normal operating range of the system frequency is from 58.8 Hz to 62.0 Hz. Converting to the per-unit (pu) values,
the normal operating deviation of the frequency is from -0.02 pu to + 0.03 pu. Furthermore, the rate of change of
frequency (RoCoF) is limited to 3.0 Hz/s. Therefore, in this case study, the normal frequency range is from 48.3 Hz
to 51.5 Hz and the RoCoF is limited to 2.5 Hz/s.

5.1. Performance under loss of generation

During the period spanning 2–3 s, the loss of generation happens. The reference and feedback of output active
power, reactive power and grid currents, as well as the output voltages under the d-q axis for the GFL control, are
shown in Figure 7. It can be seen that both the outer power loop and inner current loop can work well.

14
Gao et al.

(a)

(b)

(c)
Fig. 7. Controllers of the GFL converter under loss of generation: (a) Outer power loop; (b) output voltages; (c) inner current loop. GFL, grid-following.

According to the definition of the complex power, the active power Pe and the reactive power Qe flowing between
the PCC and the power grid can be expressed as follows:

 3U pcc 2 cos a − 3U pccU g cos ( a + d)

 Pe =
 Zg
 , (84)
 3U pcc 2 sin a − 3U pccU g sin ( a + d)
Qe = Zg


where d is the phase angle difference between the PCC voltage and the grid voltage; a represents the angle of the
grid impedance.
The relationship between the Pe, Qe and Upcc satisfies the equation given as follows:

1 9U g 2 + 12( Pe Rg + Qe X g ) . (85)
U pcc = U g +
2 6
15
 Comparison of grid-following and grid-forming converters

(a)

(b)

(c)
Fig. 8. Controllers of the GFM converter under loss of generation: (a) VSG algorithm; (b) voltage loop; (c) inner current loop. GFM, grid-forming;
VSG, virtual synchronous generators.

Because of the existence of grid impedance and because the resistance–inductance ratio of grid impedance is
not low enough in this case study, the change of the active power will lead to the change of PCC voltage. When the
active power decreases, the voltage at the PCC will decrease accordingly, which is shown in Figure 7b.
Similarly, the reference and feedback of active power and reactive power, the output voltages and the grid
currents under the d-q axis for the GFM control are shown in Figure 8. All the control loops can work well. It is worth
mentioning that because of the droop relationship between Qe and Upcc, there is a difference between Qe and Qref
in the steady state, which is shown in Figure 8a.
The simulation results of the converter output power, frequency and RoCoF at the PCC are shown in Figure 9.
It is worth noting that for both the GFL control and GFM control, the frequency and the RoCoF are measured from
the PLL unit. However, for the GFM control, the PLL unit is just used for the measurement, and the frequency used
for the control is from the VSG algorithm loop. Compared with the GFL control, the GFM control can slow down
the changes of the converter output, reflecting the inertial response characteristics. In addition, a higher frequency

16
Gao et al.

(a) (b) (c)

Fig. 9. Simulation results under loss of generation: (a) Converter output power; (b) frequency at the PCC; (c) RoCoF at the PCC. GFL, grid-following;
GFM, grid-forming; PCC, point of common coupling; RoCoF, rate of change of frequency.

(a) (b) (c)

Fig. 10. Simulation results under a sudden load shed: (a) Converter output power; (b) frequency at the PCC; (c) RoCoF at the PCC. GFL, grid-
following; GFM, grid-forming; PCC, point of common coupling; RoCoF, rate of change of frequency.

nadir and a lower RoCoF can be provided by the GFM control, which improves the stability of the power grid during
sudden changes of generation and facilitates a strong ability to cope with the fluctuations of renewable energy
generation.

5.2. Performance under a sudden load shed

During the period spanning 2–3 s, a sudden load shed is applied to the power grid. The simulation results of the
converter output power, frequency and RoCoF at the PCC are shown in Figure 10. Compared with the GFL control,
the GFM control avoids the sudden increase in frequency, which enhances the ability of the power grid to handle the
case of load fluctuations. As shown in Figures 10b and 10c, with the GFM control, both the frequency culmination
and the RoCoF decrease significantly.

5.3. Performance under operation at different SCRs

The simulation results of frequency and RoCoF at the PCC in the case of operation at different SCRs are shown
in Figures 11 and 12, respectively. During the period spanning 2–3 s, a load step is applied to the power grid. As
shown in Figures 11 and 12, whether with GFL control or GFM control, the impact of load step on the power grid
is large when the value of SCR is small. In addition, with the GFM control, the frequency nadir is increased and
the RoCoF is decreased, which reduces the influence of sudden load increase and enhances the stability of the
power grid. When the SCR is equal to ‘1’, the GFL control cannot maintain stable operation, while the GFM control
still works well. Although the value of RoCoF is beyond the limits of the IEEE 1547-2018 standard (IEEE Std 1547-
2018, 2018), it can be solved by regulating the virtual inertia and damping coefficient. The simulation results are
consistent with the theoretical analysis in Sections 3 and 4. The GFL converter may encounter some instability

17
 Comparison of grid-following and grid-forming converters

(a) (b)
Fig. 11. Simulation results of frequency at the PCC under operation at different SCRs: (a) GFL converter (b) GFM converter. GFL, grid-following;
GFM, grid-forming; PCC, point of common coupling; SCR, short-circuit ratio.

(a) (b)
Fig. 12. Simulation results of RoCoF at the PCC under operation at different SCRs: (a) GFL converter; (b) GFM converter. GFL, grid-following;
GFM, grid-forming; PCC, point of common coupling; RoCoF, rate of change of frequency; SCR, short-circuit ratio.

issues when applied in a weak power grid, while the GFM converter is more suitable for operation with a weak
power grid.

6. Conclusion
This paper analysed the control schemes of GFL converters and GFM converters. The state-space models of
both the GFL converters and GFM converters were built. Through a case study, the dynamic responses of GFL
control and GFM control under various working conditions were compared. The grid-connected converter with the
GFM control was shown to be able to slow down the change of converter output, reflecting the inertial response
characteristics of conventional synchronous generators, and to improve the grid’s ability to cope with sudden
frequency disturbances, i.e., fluctuations of generation and load. Analysis of the simulation results showed that
the smaller the SCR becomes, the larger is the influence that frequency disturbances have on the power grid.
Furthermore, compared with the GFL converter, the GFM converter is more suitable for a weak power grid.

18
Gao et al.

References

Adapa, A. K. and John, V. (2018). Virtual resistor based IEEE Standard for Interconnection and Interoperability
active damping of LC filter in standalone voltage of Distributed Energy Resources with Associated
source inverter. In: Proceedings of APEC 2018, Electric Power Systems Interfaces. (2018). IEEE
USA, 4-8 Mar. 2018, pp. 1834–1840. Standard 1547-2018 (Revision of IEEE Standard
Chen, M., Zhou, D. and Blaabjerg, F. (2020). Modelling, 1547-2003), 2018, pp. 1-138.
Implementation, and Assessment of Virtual Lasseter, R. H., Chen, Z. and Pattabiraman, D.
Synchronous Generator in Power Systems. Journal (2020). Grid-Forming Inverters: A Critical Asset
of Modern Power Systems and Clean Energy, 8(3), for the Power Grid. IEEE Journal of Emerging
pp. 399–411. and Selected Topics in Power Electronics, 8(2),
Dahono, P. A., Bahar, Y. R., Sato, Y. and Kataoka, T. pp. 925–935.
(2001). Damping of transient oscillations on the Pattabiraman, D., Lasseter, R. H. and Jahns, T. M.
output LC filter of PWM inverters by using a virtual (2018). Comparison of grid following and grid
resistor. In: Proceedings of 4th IEEE International forming control for a high inverter penetration power
Conference on Power Electronics and Drive system. In: Proceedings of 2018 IEEE Power and
Systems, Denpasar, Indonesia, 25 Oct. 2018, Energy Society General Meeting PESGM, Portland
pp. 403–407. or USA, 5-10 Aug. 2018, pp. 1–5.
Dong, D., Wen, B., Boroyevich, D., Mattavelli, P. Peng, Q., Jiang, Q., Yang, Y., Liu, T., Wang, H. and
and Xue, Y. (2015). Analysis of Phase-Locked Blaabjerg, F. (2019). On the Stability of Power
Loop Low-Frequency Stability in Three- Electronics-Dominated Systems: Challenges and
Phase Grid-Connected Power Converters Potential Solutions. IEEE Transactions on Industry
Considering Impedance Interactions. IEEE Applications, 55(6), pp. 7657–7670.
Transactions on Industrial Electronics, 62(1), Peng, Q., Yang, Y., Liu, T. and Blaabjerg, F. (2020).
pp. 310–321. Coordination of Virtual Inertia Control and
Du, W., Tuffner, F., Schneider, K. P., Lasseter, R. H., Xie, Frequency Damping in PV Systems for Optimal
J., Chen, Z. and Bhattarai, B. P. (2020). Modeling Frequency Support. CPSS Transactions on Power
of Grid-Forming and Grid-Following Inverters for Electronics and Applications, 5(4), pp. 305–316.
Dynamic Simulation of Large-Scale Distribution Rocabert, J., Luna, A., Blaabjerg, F. and Rodríguez, P.
Systems. IEEE Transactions on Power Delivery, (2012). Control of Power Converters in AC
36(4), pp. 2035–2045. Microgrids. IEEE Transactions on Power
Fang, J., Li, H., Tang, Y. and Blaabjerg, F. (2018). Electronics, 27(11), pp. 4734–4749.
Distributed Power System Virtual Inertia Rosso, R., Wang, X., Liserre, M., Lu, X. and
Implemented by Grid-Connected Power Engelken, S. (2021). Grid-Forming Converters:
Converters. IEEE Transactions on Power Control Approaches, Grid-Synchronization, and
Electronics, 33(10), pp. 8488–8499. Future Trends – A Review. IEEE Open Journal of
Gao, X., Zhou, D., Anvari-Moghaddam, A. and Industry Applications, 2, pp. 93–109.
Blaabjerg, F. (2021). Grid-following and grid- Sangwongwanich, A., Abdelhakim, A., Yang, Y. and
forming control in power electronic based power Zhou, K. (2018). Control of single-phase and
systems: A comparative study. In: Proceedings three-phase DC/AC converters. In: F. Blaabjerg,
of IECON 2021 – 47th Annual Conference of the ed., Control of Power Electronic Converters and
IEEE Industrial Electronics Society, Toronto, 13-16 Systems. Academic Press, Amsterdam, The
Oct. 2021, pp. 1–6. Netherlands, pp. 153–173.
Gao, X., Zhou, D., Anvari-Moghaddam, A. and Blaabjerg, Wang, X., Taul, M. G., Wu, H., Liao, Y., Blaabjerg, F. and
F. (2022). Stability analysis of grid-following and Harnefors, L. (2020). Grid-Synchronization Stability
grid-forming converters based on state-space of Converter-Based Resources – An Overview.
model. In: Proceedings of 2022 International Power IEEE Open Journal of Industry Applications, 1, pp.
Electronics Conference (IPEC-Himeji 2022-ECCE 115–134.
Asia), Himeji, 15-19 May 2022, pp. 422–428. Wen, B., Boroyevich, D., Burgos, R., Mattavelli, P.
IEEE Guide for Planning DC Links Terminating at AC and Shen, Z. (2016). Analysis of D-Q Small-
Locations Having Low Short-Circuit Capacities. Signal Impedance of Grid-Tied Inverters.
(1997). IEEE Standard 1204-1997, 1997, IEEE Transactions on Power Electronics, 31(1),
pp. 1-216. pp. 675–687.

19
 Comparison of grid-following and grid-forming converters

Wu, G., Sun, H., Zhang, X., Egea-Alvarez, A., Zhao, Yang, L., Chen, Y., Luo, A., Chen, Z., Zhou, L., Zhou,
B., Xu, S., Wang, S. and Zhou, X. (2020). X., Wu, W., Tan, W. and Guerrero, J. M. (2019).
Parameter Design Oriented Analysis of the Current Effect of Phase-Locked Loop On Small-Signal
Control Stability of the Weak-Grid-Tied VSC. Perturbation Modelling and Stability Analysis for
IEEE Transactions on Power Delivery, 36(3), Three-Phase LCL-Type Inverter Connected to
pp. 1458–1470. Weak Grid. IET Renewable Power Generation,
Xie, Z., Chen, Y., Wu, W., Xu, Y., Wang, H., Guo, J. and 13(7), pp. 86–93.
Luo, A. (2019). Modeling and Control Parameters Zhong, Q. and Weiss, G. (2011). Synchronverters:
Design for Grid-Connected Inverter System Inverters That Mimic Synchronous Generators.
Considering the Effect of PLL and Grid Impedance. IEEE Transactions on Industrial Electronics, 58(4),
IEEE Access, 8, pp. 40474–40484. pp. 1259–1267.

20